drug drug interactions research articles

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• Front Pharmacol

Adverse drug reactions and correlations with drug–drug interactions: A retrospective study of reports from 2011 to 2020

Huaqiao jiang.

1 Department of Pharmacy, Jinshan Hospital, Fudan University, Shanghai, China

2 Department of Nursing, Jinshan Hospital, Fudan University, Shanghai, China

Weifang Ren

Zhonghong fang, xiaofang tan.

Myriam Jaam , Qatar University, Qatar

Monica Danial , Universiti Sains Malaysia (USM), Malaysia

Associated Data

The raw data supporting the conclusion of this article will be made available by the authors, without undue reservation.

Introduction: Adverse drug reactions (ADRs) represent a public health problem worldwide that deserves attention due to the impact on mortality, morbidity, and healthcare costs. Drug–drug interactions (DDIs) are an important contributor to ADRs. Most of the studies focused only on potential DDIs (pDDIs), while the detailed data are limited regarding the ADRs associated with actual DDIs.

Methods: This retrospective study evaluated ADRs reported between 2011 and 2020 in a tertiary hospital. The causality and severity of ADRs were evaluated through the Naranjo Algorithm and Hartwig’s scale, respectively. Preventability classification was based on the modified Schoumock and Thornton scale. For ADRs with at least two suspected drugs, pDDIs were identified according to the Lexi-Interact. We further checked whether the ADR description in the reports corresponded to the clinical consequences of the pDDIs.

Results: A total of 1,803 ADRs were reported, of which 36.77% ADRs were classified as mild, 43.26% as moderate, and 19.97% as severe. The assessment of causality showed that the distributions of definite, probable, and possible categories were 0.33%, 58.68%, and 40.99%, respectively. A total of 53.97% of ADRs were identified as preventable ADRs, while 46.03% were recognized as unpreventable. The severity of ADRs was significantly correlated with age, the number of suspected drugs and preventability. Antimicrobial agents were the most common implicated pharmacological group, and the most frequently affected system was the gastrointestinal system. Considering individual drugs, aspirin was the most frequently reported drug. Among 573 ADRs with at least two suspected drugs, 105 ADRs were caused by actual DDIs, of which only 59 and 6 ADRs were caused by actual DDIs in category D and X, respectively. The most frequent drugs involved in actual DDIs of category D were aspirin and heparin, with the majority of ADRs being gastrointestinal bleeding.

Conclusion: This study analyzed the pattern of ADRs in detail and obtained clinical evidence about ADRs associated with actual DDIs. These findings may be useful to compare patterns between different centers and to design preventive strategies for ADRs. Continuous education and training should be provided for physicians regarding the knowledge and recognition of ADRs associated with DDIs.

Introduction

According to the World Health Organization (WHO), an adverse drug reaction (ADR) is an unintended and noxious response that is detected in patients after the use of drugs for the prophylaxis, diagnosis or treatment of a disease at doses normally used ( Edwards and Aronson, 2000 ). ADRs, as a major threat in the healthcare system, contribute significantly to mortality, morbidity, extended hospital stays, and increased healthcare costs ( Khan, 2013 ; Angamo et al., 2016 ). A meta-analysis showed that the percentage of ADR-induced admissions in patients over 60 years old was accurately estimated to be 8.7% ( Oscanoa et al., 2017 ). To minimize the consequences of ADRs, it is necessary to study ADRs in terms of their early identification and prevention and to motivate healthcare professionals to report ADRs ( Arulappen et al., 2018 ).

According to a WHO report, 60% of ADRs are preventable ( Lau et al., 2003 ). Drug–drug interactions (DDIs) are an important cause of preventable ADRs. The increasing number of patients with multimorbidity and the growing complexity of therapeutic agents have led to widespread polypharmacy, which could result in the rising numbers of potential DDIs (pDDIs), especially in elderly individuals ( Obreli-Neto et al., 2012a ; Scondotto et al., 2018 ). Although there are several databases available that could be used to evaluate pDDIs, the clinical relevance and actual clinical importance of majority pDDIs remain insufficiently characterized and underestimated ( Roblek et al., 2015 ). Actual DDIs are identified on the basis of clinical evidence, such as laboratory test results or symptoms, consequently, the frequency of actual DDIs is much lower than that of pDDIs ( Magro et al., 2012 ; Zheng et al., 2018 ). Over the past years, a substantial number of articles have been published about ADRs due to DDIs ( Leone et al., 2010 ; Obreli-Neto et al., 2012a ; Obreli Neto et al., 2012b ; Kovacevic et al., 2019 ; Letinier et al., 2020 ; Magro et al., 2020 ). A 6-year retrospective study in Bengbu in China showed that among the ADRs reported between nervous system drugs in hospitalized patients, 12.14% of the ADRs were associated with potential and actual DDIs, and actual DDIs were present in 6.21% of all ADRs ( Shi et al., 2014 ). However, the incidence of ADRs resulting from DDIs could not be accurately estimated primarily because of differences in study designs and populations ( Mirosevic Skvrce et al., 2011 ).

In this context, the present study aimed to describe the distribution of ADRs, assess causality, preventability and severity of ADRs, and determine factors involved in the severity of ADRs in a tertiary hospital between 2011 and 2020. Additionally, we described and analyzed the most frequent drugs suspected to cause ADRs and the organ system classes affected by ADRs. Furthermore, we evaluated the pDDIs among the ADRs with more than one suspected drug, estimated the incidence of ADRs due to actual DDIs and characterized ADRs caused by actual DDIs.

Materials and methods

Data collection.

In this retrospective single-center study, all the ADRs was collected from the National ADR Monitoring system in Jinshan Hospital of Fudan University, between 01 January 2011 and 31 December 2020. Jinshan Hospital is a tertiary general hospital with a 700-bed capacity in the Jinshan district of Shanghai. In 2020, there were 28,533 hospital admissions, and 1.28 million outpatient and emergency department visits. ADR reports were filled out according to a specific ADR report format and submitted in paper based or electronic way by healthcare professionals, including physicians, pharmacists, and nurses.

Once received, the reported ADRs were reviewed and evaluated by ADR surveillance unit of the pharmacy department. Only the reported ADRs followed the WHO definition ( Edwards and Aronson, 2000 ) and without any uncertainty or mistakes were accepted after exclusion of duplicates and uploaded to ADR Monitoring system. A series of exclusion criteria were applied to ensure a robust data set for analysis. Exclusion criteria included the following: 1) ADRs with doubtful causality with Naranjo’s algorithm ( Naranjo et al., 1981 ). 2) ADR forms with insufficient information 3) ADRs symptoms similar to the original disease.

The demographic and other information relevant to ADRs were documented, including gender, age, diagnosis, admission department, suspected drugs, concomitant medications, drug details, organ system involved in the ADR, the management and outcome of the ADRs, and the type of reporter. One report could describe one or more ADRs. The incriminated drugs were classified by pharmacological group according to the WHO Anatomical Therapeutic Chemical Classification (ATC). The involved system organ classes were determined according to WHO Adverse Reaction Terminologies (WHO-ART). Two investigators cross checked the data for accuracy. Flowchart depicting the study process was shown in Figure 1 .

Flowchart depicting the study process.

Causality, preventability, and severity assessment

Each ADR was further evaluated for various parameters, such as causality, severity and preventability, using previously validated and recognized approaches. The assessment of causality was performed using the Naranjo Algorithm, which consists of 10 individually scored criteria. ADRs were categorized as possible ADRs (1–4), probable ADRs (5–8) or definite ADRs (≥9) based on the total score ( Naranjo et al., 1981 ). Severity classification was based on Hartwig’s scale, which showed the criteria and matched levels used for ADR severity assessment. ADRs were considered as severe if they resulted in one of the following outcomes: the requirement for intensive medical care, permanent harm to the patient, or the death of the patient ( Hartwig et al., 1992 ). The preventability of ADRs was assessed by the modified Schoumock and Thornton scale and classified into definitely preventable, probably preventable and not preventable reactions ( Schumock and Thornton, 1992 ). In our study, both definitely and probably preventable ADRs were considered as one category of preventable reactions.

Evaluation of potential drug–drug interactions

For ADRs caused by two or more suspected drugs, pDDIs were identified by the software Lexi-Interact in UpToDate. The evaluation results of pDDIs were classified into five levels of risk as no known interaction (A), no action needed (B), monitor therapy (C), consider therapy modification (D), and avoid combination (X). We further verified whether the clinical consequences of pDDIs corresponded to the description of the ADR in the report, and if consistent, the pDDI was considered the actual DDI. Two clinical pharmacists independently assessed the probability, severity and preventability of ADRs as well as the consistency between ADRs and pDDIs. Any discrepancies were resolved by discussion.

Statistical analysis

Descriptive statistics were applied to describe the population as well as the clinical characteristics of ADRs and pDDIs. The categorical data were presented as numbers and proportions. Sankey diagrams of severity in preventable and unpreventable ADRs were plotted with the R package alluvial. The Mann–Whitney U test was used to evaluate the correlation between gender and the severity of ADRs. Spearman’s rank tests were performed to determine the association of age, the number of suspected drugs and the category of preventability with the severity of ADRs. The Kruskal–Wallis H test was performed to evaluate the correlation between the route of administration and the severity of ADRs. Statistical analysis was performed using IBM SPSS Statistics version 25. A p -value < 0.05 was considered statistically significant.

Department and reporter distribution of adverse drug reactions

From January 2011 to December 2020, a total of 1,803 ADRs were reported by healthcare professionals in our hospital, although the number of ADRs reported was relatively small between 2011 and 2013. During this 10-year period, pharmacists contributed 55.69% of all ADR reports, followed by physicians (43.98%). The frequency of ADRs reported by nurses was low, accounting for only 0.33%. The annual number of reports was no more than 221 during 2011–2018, however, this number subsequently increased significantly over the next 2 years, reaching 388 in 2020 ( Figure 2 ). A small proportion of ADRs were reported by pharmacists between 2011 and 2013, however, since 2014, more than half of ADR reports have been submitted by pharmacists. Detailed data by the year and distribution of reporters were shown in Figure 2 . In our study, the highest percentage of ADRs was collected from the gastroenterology department (26.8%), followed by the departments of emergency and critical care medicine (11.4%), cardiology department (7.9%), and neurology department (7.8%) ( Figure 3 ). The proportions of ADRs collected from clinical departments were presented in Figure 3 .

The total number of adverse drug reaction (ADR) reports and the distribution of reporters from different occupations by year during 2011–2020.

The percentage of adverse drug reactions (ADRs) from different clinical departments.

Causality, preventability, and severity assessment of adverse drug reactions

ADRs were further analyzed for causality, preventability and severity, as shown in Table 1 . The assessment of causality according to the Naranjo Algorithm showed that the numbers of definite, probable and possible ADRs were 6 (0.33%), 1,058 (58.68%), and 739 (40.99%), respectively. According to Hartwig’s Severity Assessment Scale, 663 (36.77%) ADRs were classified as mild, 780 (43.26%) as moderate, and 360 (19.97%) as severe. The evaluation of the preventability of ADRs using the modified Schumock and Thornton criteria revealed that 973 (53.97%) ADRs were identified as preventable ADRs, including 93 as definitely preventable and 880 as probably preventable, while 830 (46.03%) ADRs were recognized as unpreventable. Symptomatic or specific treatment was given for 1,045 (57.96%) ADRs. According to the records of ADR reports, the majority of ADRs (81.09%) had improved, 238 (13.20%) patients had recovered from their ADRs, and 103 (5.71%) ADRs continued or their status was unclear. Suspected drugs were withdrawn in 1,700 (94.29%) ADR reports, but an altered dose or no change in therapy was observed in 103 (5.71%) reports. The visual design follows the principle of the Sankey diagram, which links the ADR characteristics by lines and signifies the quantities via line width, stratified by preventability ( Figure 4 ).

Assessment and pattern of adverse drug reactions.

Sankey diagram of severity in preventable and unpreventable ADRs. (A,B) The causality assessment, fate of the suspected drug, treatment given and outcome of ADRs matched with ADR severity in preventable ADRs. (C,D) The causality assessment, fate of the suspected drug, treatment given and outcome of ADRs matched with ADR severity in unpreventable ADRs. The causality assessment of ADRs (A1 and A2), A1: Definite/Probable, A2: Possible. ADR severity (B1, B2, and B3), B1: Mild, B2: Moderate, B3: Severe. Fate of the suspected drug (C1 and C2), C1: Drug withdrawn, C2: Dose altered/No change. Treatment given (D1 and D2), D1: Treatment given, D2: No treatment. Outcome of ADRs (E1, E2, and E3), E1: Recovered, E2: Improved, E3: Continuing/Unclear.

Characteristics of adverse drug reactions according to severity assessment

The characteristics of ADRs according to severity assessment were shown in Table 2 . A total of 1,803 ADRs were identified among 1,779 patients. Multiple ADRs in the same patient may be identified with different severity scale, so Table 2 depicted the distribution of mild, moderate, and severe reactions between different gender and age based on ADRs rather than patients. Concerning patient gender and ADRs, 46.87% males and 53.13% females experienced ADRs over the past 10 years. The proportion of mild ADRs was higher in females (60.33%) than in males (39.67%), however, the ratio of males to females was approximately 1:1 among those experiencing moderate and severe ADRs. The Mann–Whitney U test revealed significant differences in the mild, moderate, and severe ADR distributions between the males and females.

Comparison of mild, moderate, and severe reactions.

The R value represents Spearman’s correlation coefficient.

As shown in Table 2 , the percentage of ADRs was highest among elderly individuals over 65 years of age (42.87%), followed by the 41–64-year (38.38%) and 18–40-year (14.03%) age groups. The minimum number of ADRs was observed in the age group under 18 years (4.71%). More than half of severe ADRs occurred in elderly individuals over 65 years of age. The majority of ADRs (67.72%) were identified with only one suspected drug, followed by 24.35% with two suspected drugs, and only 143 (7.93%) ADRs were found with ≥3 suspected drugs. According to Hartwig’s scale, 46.03% of ADRs were classified as unpreventable ADRs, 48.81% as probably preventable ADRs and 5.16% as definitely preventable ADRs. The percentage of unpreventable ADRs significantly decreased with ADR severity (mild 63.65% vs. moderate 43.72% vs. severe 18.61%). The statistical results revealed significant positive correlations of ADR severity with age (Spearman’s R = 0.167, p < 0.001), the number of suspected drugs (Spearman’s R = 0.136, p < 0.001) and ADR preventability (Spearman’s R = 0.299, p < 0.001).

The route of administration was classified according to the first suspected drug in the ADR reports. More than half of ADRs were associated with oral medicines regardless of their severity rating. Unexpectedly, the proportion of ADRs associated with intravenous drugs gradually decreased with increasing severity. The Kruskal–Wallis H test was further carried out and showed a significant association between the route of administration and the severity of ADRs ( Table 2 ).

Drugs involved in adverse drug reactions and effects on organ systems

The pharmacological groups implicated in the ADRs were summarized in Table 3 . Systemic antimicrobial agents were the most commonly implicated drugs (22.75%), with 14.53% and 39.32% of their associated ADRs being classified as severe and preventable ADRs, respectively. Cardiovascular agents were the second most frequently reported class of drugs responsible for ADRs (12.41%), followed by medications for the alimentary tract and metabolism (12.06%). Drugs acting on the blood and blood-forming organs represented 11.75% of the reports (32.45% severe and 83.11% preventable ADRs). Drugs acting on the musculoskeletal system were implicated in 10.77% of the ADRs (39.71% severe and 71.48% preventable ADRs). Traditional Chinese medicines were implicated in 10.07% of the reports (18.53% severe ADRs and 54.44% preventable ADRs).

Pharmacology groups according to the WHO-ATC code and their pattern in ADRs.

WHO-ART, WHO Adverse Reaction Terminologies; ADR, adverse drug reaction.

The frequency of commonly prescribed drugs among total and severe ADRs was shown in Table 4 . When individual drugs were considered, aspirin was responsible for a maximum number of both total and severe ADRs, far more than any other drugs. Among the total ADRs, levofloxacin (82) was the second most frequent causative drug, followed by compound pseudoephedrine hydrochloride (65) and clopidogrel (48). In addition to aspirin, the drugs most frequently involved in severe ADRs were clopidogrel (17), levofloxacin (16), compound pseudoephedrine hydrochloride (14), and diclofenac sodium (13).

Top 10 incriminated drugs in total and severe ADRs based on frequency.

ADR, adverse drug reaction.

Upon a review of the outcomes of ADRs, the most frequently affected system was the gastrointestinal system (30.83%), with the clinical symptoms of nausea, vomiting, abdominal pain, diarrhea, abdominal distention, and so on. In addition, the commonly reported reactions were skin and appendage disorders (22.44%) and liver and biliary system disorders (14.19%). A more detailed description was presented in Table 5 .

Organs or systems involved in ADRs according to WHO classification.

WHO, World Health Organization; ADR, adverse drug reaction.

Adverse drug reactions caused by drug–drug interactions

pDDIs were evaluated in 573 of 1,803 ADR reports (31.78%) involving more than one suspected drug. 156 ADRs were identified with pDDIs of category C, D, and X, of which 100 ADRs were identified with only one pDDI and 56 ADRs with multiple pDDIs. Table 6 showed that 208 pDDIs of category C were identified in 112 ADRs, 74 pDDIs of category D in 58 ADRs, and 11 pDDIs of category X in 10 ADRs. Furthermore, we checked whether the reported ADRs were consistent with the potential clinical consequences of pDDIs. The results showed 105 ADRs were caused by actual DDIs, accounting for 18.32% of the ADR reports with more than one suspected drug. Among them, 59 and 6 ADRs were caused by actual DDIs in the category D and X, respectively.

Distribution of the potential drug–drug interactions with category C, D, and X in ADRs.

ADR, adverse drug reaction; pDDI, potential drug–drug interaction.

Tables 7 , ​ ,8 8 summarized the ADRs caused by actual DDIs belonging to category X and D, respectively. Potassium chloride and promethazine were the drug–drug combination most involved in ADRs caused by actual DDIs in category X, with severe and adverse clinical consequences to the gastrointestinal system. The most frequent drugs involved in actual DDIs of category D were aspirin ( n = 34) and heparin ( n = 26), and the great majority of ADRs caused by DDIs were associated with gastrointestinal bleeding. Aspirin/heparin ( n = 10) and heparin/clopidogrel ( n = 10), followed by aspirin/warfarin ( n = 6) and aspirin/ibuprofen ( n = 5), were the drug–drug combinations most involved in ADRs caused by DDIs of category D.

ADRs caused by actual drug–drug interactions belonging to category X.

ADRs caused by actual drug–drug interactions belonging to category D.

ADR, adverse drug reaction; CNS, central nervous system.

In this study, physicians and pharmacists were the groups that reported the great majority of ADRs, and the frequency of ADRs reported by nursing staff was low, which may be due to their extensive workload in everyday practice, inattention and unawareness toward ADR reporting or worry about legal implications ( Singh et al., 2017 ). The reporter distribution of ADRs varies widely in different studies because of differences in healthcare structures as well as the awareness and motivation of healthcare professionals. The number of ADRs was relatively small, especially for ADRs reported by pharmacists between 2011 and 2013, indicating underreporting in pharmacovigilance. The key to improving ADR reporting rates is adequate pharmacovigilance education and training for healthcare professionals ( Barzaga Arencibia et al., 2012 ).

In the present study, we analyzed the pattern of ADRs based on the causality, severity, and preventability in our hospital, all of which vary among different hospitals due to differences in the population characteristics and hospital specialties. Naranjo’s causality assessment showed that only 0.33% of reports were definite because of limited use of dechallenge and rechallenge processes for ethical reasons as well as the retrospective study design without the ability to assess the ADR completely. The suspected drugs were withdrawn among 94.29% of ADRs, and for the remaining 5.71% of ADRs, the suspected drug doses were altered or rechallenge processes were initiated. In this study, 19.97% of ADRs were classified as severe. Severe ADRs, as major concerns for public health, are a contributing factor of hospitalizations and morbidity ( Rottenkolber et al., 2011 ; Marques et al., 2014 ). The analysis indicated a preventability rate of 53.97% among ADRs, comparable with the results of studies conducted in Romania and Jordan showing that 41% and 44.7% of ADRs were preventable, respectively ( Farcas et al., 2014 ; Al Damen and Basheti, 2019 ). However, the data from a study showed lower preventability for ADRs (12%) compared with our finding ( Dequito et al., 2011 ). As described in previous studies, insufficient monitoring, inappropriate dosing, and DDIs were the most frequent factors involved in ADR preventability ( Farcas et al., 2014 ; Al Damen and Basheti, 2019 ). Incriminated drugs were withdrawn in 94.29% of the reports, which is in line with a previous study in a psychiatric department of a tertiary care teaching hospital in India ( Patel et al., 2015 ). The high proportion of withdrawal may be due to the reporting nature of ADRs that troublesome ADRs are more likely to be detected.

There may be significant difference between male and female regarding the ADR prevalence due to factors such as body mass index, fat composition, hormonal effects, drug susceptibility, or genetic differences in the levels of enzymes ( Haile et al., 2013 ; Rukmangathen et al., 2020 ). However, we demonstrated that females had only slightly higher incidence of ADRs than males in the present study. The frequency of ADRs increased with age, with the highest prevalence of ADRs in elderly individuals over 65 years (42.87%), followed by individuals 41–64 years of age (38.38%), which is in concordance with the findings of a previous study ( Shepherd et al., 2012 ). Older patients are particularly vulnerable to ADRs owing to the multiple-drug regimens used for chronic diseases and physiological changes in this population, such as reduced gastrointestinal motility and gastric blood flow, impaired repair mechanisms, and lower mucosal protection ( Marusic et al., 2014 ). A systematic review of ADRs in elderly individuals revealed that comorbid complexity was positively associated with ADR occurrence ( Alhawassi et al., 2014 ). In the present study, there were statistically significant differences in the incidence of severe ADRs in the different gender and age groups, and polypharmacy increased the proportion of severe ADRs.

Anti-infectives for systemic use were the most common pharmacological group, accounting for 22.75% of total ADRs in our study, which is in line with previous studies ( Haile et al., 2013 ; Marques et al., 2014 ). The excessive use of antibiotics may be responsible for the increased risk of ADRs. Cardiovascular system agents (12.41%) were the second most frequently incriminated pharmacological class of ADRs in our study, among them, 65.20% were preventable ADRs. A systematic review showed that cardiovascular medicines were commonly associated with preventable drug-related admissions ( Howard et al., 2007 ). In another study, cardiovascular agents were identified as the second most frequently responsible drugs linked to preventable ADRs ( Farcas et al., 2014 ).

The system most frequently affected by ADRs in this study was the gastrointestinal system, accounting for 30.83%, probably due to more than half of the suspected drugs being administered orally. This was followed by skin and appendage disorders (22.44%). This observation is consistent with the findings of a prospective observational study of hospitalized pediatric patients, which reported gastrointestinal system disorders (51.56%) and skin and appendage disorders (18.75%) as the most frequent manifestations of ADRs ( Kurian et al., 2016 ).

As DDIs are usually predictable and manageable, ADRs caused by DDIs may be prevented by monitoring the patient closely or replacing the responsible drugs with other medications. To reduce the risk of DDIs and improve patient safety, it is essential that healthcare professionals regularly review the medication regimens, recognize potentially interacting drug pairs, and withdraw unnecessary drugs ( Magro et al., 2020 ). A prospective study showed that the number of patients with pDDIs and actual DDIs decreased by 18% and 43%, respectively, with an intervention based on a computerized clinical decision support system containing information on drug combinations ( Bertsche et al., 2010 ). However, reporters less frequently recognize actual DDIs due to the limited availability of DDI databases or alerting drug-interaction systems ( Mirosevic Skvrce et al., 2011 ). Therefore, it is important to increase the knowledge of pharmacovigilance through the additional education of healthcare providers.

In a previous study, we investigated the prevalence of pDDIs and their association with characteristics in outpatient prescriptions ( Ren et al., 2020 ). However, to assess the clinical impact of DDIs on public health, only ADRs associated with DDIs should be considered. In our study, 105 ADR reports were induced by actual DDIs, accounting for 18.32% of the ADR reports with more than one suspected drug. This percentage was close to the proportion reported by Magro et al. (2020) . According to the online version of DRUGDEX ® system, they verified DDI among serious ADRs containing at least two suspected or concomitant drugs in the National Pharmacovigilance database from Veneto Region, and identified 17.4% ADR reports associated with a DDI. However, the results of another study performed in an Italian spontaneous reporting database showed that regarding patients treated with at least two drugs, 6.5% of ADR reports was associated with a DDI using the DRUGDEX ® system ( Leone et al., 2010 ). Similarly, a prospective cohort study conducted in the primary public health system of the Ourinhos microregion in Brazil revealed that the incidence of DDI-related ADRs was 6% in elderly outpatients using DDI-checker programs (DrugDigest ® , Drugs ® , Micromedex ® , and Medscape ® ) ( Obreli-Neto et al., 2012b ).

In the present study, aspirin and heparin were the drugs most frequently associated with actual DDIs of category D, with symptom of gastrointestinal bleeding. Similarly, a prospective observational study conducted in the cardiology unit of an Indian hospital showed that heparin and aspirin were the most common drugs responsible for DDIs, and bleeding was the most frequent clinical consequence ( Mateti et al., 2011 ). Furthermore, aspirin, which is widely used for the prevention of vascular events, was reported to increase the baseline risk of gastrointestinal bleeding by approximately 60% among older persons aged over 70 years in a randomized controlled trial ( Mahady et al., 2021 ).

Although the study had important findings regarding the pattern of ADRs and the role of actual DDIs in ADRs over the past decade along with a large sample size, several limitations should be taken into consideration. First, as a retrospective study, data were collected from the clinical records of ADRs always with incomplete information, such as information on concomitant drugs, comorbidities, lifestyle, diet, and so on. Prospective studies will be carried out to clarify and reduce this limitation in the future. Second, this study was conducted at a single institution, limiting the generalizability of its findings due to the differences in population characteristics and prescribing patterns. Last, the single source of the DDI screening database used in this study may hinder the identification of DDIs because consistent criteria for DDI identification and assessment are currently lacking.

This study of ADR data collected over 10 years revealed that almost all ADRs were reported by pharmacists and physicians in our hospital, and the severity of ADRs was significantly correlated with age, the number of suspected drugs and preventability. Systemic antimicrobial agents were the most frequently incriminated pharmacological group, and aspirin was responsible for the largest proportion of total and severe ADRs. The gastrointestinal system was the system most frequently affected by ADRs. As observed in this study, aspirin and heparin were the most common drugs in actual DDIs of category D, resulting in gastrointestinal bleeding.

Active pharmacovigilance programs are important to accurately identify and assess ADRs in the clinical setting, further minimize drug-induced harm and improve the quality of patient care. Our findings obtained clinical evidence about ADRs associated with actual DDIs in our hospital. It will be necessary to make clinicians aware of the possibility of DDI-related ADRs and achieve a clear understanding of drug pairs resulting in DDI-related ADRs, in order to guide the prescribing practices and minimize the harms from actual DDIs. Moreover, rigorous prescription and frequent monitoring of drug therapy are essential for reducing the risk of ADRs.

Acknowledgments

The authors acknowledge the ADR reporting system of Jinshan Hospital, Fudan University.

Data availability statement

Ethics statement.

This study was approved by the Ethics Committee of Jinshan Hospital, Fudan University at Shanghai, China (approval No. JIEC 2022-S29).

Author contributions

HJ, ZF, XL, and NZ contributed to the conception and design of the study. HJ, YL, WR, YL, and XT contributed to the recording and statistical analysis of the data. HJ and XL wrote the first draft of the manuscript. XL and NZ made critical revisions to the manuscript. All authors approved the final version of the manuscript.

This work was supported by the Fourth Training Program for the Outstanding Young Talents, Jinshan Health Commission (#JSYQ201904), Key Construction Project on Clinical Pharmacy of Shanghai (#2019-1229).

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

• Al Damen L., Basheti I. (2019). Preventability analysis of adverse drug reactions in a Jordanian hospital: A prospective observational study . Int. J. Clin. Pharm. 41 ( 6 ), 1599–1610. 10.1007/s11096-019-00925-0 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Alhawassi T. M., Krass I., Bajorek B. V., Pont L. G. (2014). A systematic review of the prevalence and risk factors for adverse drug reactions in the elderly in the acute care setting . Clin. Interv. Aging 9 , 2079–2086. 10.2147/CIA.S71178 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Angamo M. T., Chalmers L., Curtain C. M., Bereznicki L. R. (2016). Adverse-drug-reaction-related hospitalisations in developed and developing countries: A review of prevalence and contributing factors . Drug Saf. 39 ( 9 ), 847–857. 10.1007/s40264-016-0444-7 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Arulappen A. L., Danial M., Sulaiman S. A. S. (2018). Evaluation of reported adverse drug reactions in antibiotic usage: A retrospective study from a tertiary care hospital, Malaysia . Front. Pharmacol. 9 , 809. 10.3389/fphar.2018.00809 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Barzaga Arencibia Z., Lopez Leyva A., Mejias Pena Y., Gonzalez Reyes A. R., Fernandez Manzano E., Choonara I., et al. (2012). Pharmacovigilance in children in camaguey province, Cuba . Eur. J. Clin. Pharmacol. 68 ( 7 ), 1079–1084. 10.1007/s00228-012-1222-9 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Bertsche T., Pfaff J., Schiller P., Kaltschmidt J., Pruszydlo M. G., Stremmel W., et al. (2010). Prevention of adverse drug reactions in intensive care patients by personal intervention based on an electronic clinical decision support system . Intensive Care Med. 36 ( 4 ), 665–672. 10.1007/s00134-010-1778-8 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Dequito A. B., Mol P. G., van Doormaal J. E., Zaal R. J., van den Bemt P. M., Haaijer-Ruskamp F. M., et al. (2011). Preventable and non-preventable adverse drug events in hospitalized patients: A prospective chart review in The Netherlands . Drug Saf. 34 ( 11 ), 1089–1100. 10.2165/11592030-000000000-00000 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Edwards I. R., Aronson J. K. (2000). Adverse drug reactions: Definitions, diagnosis, and management . Lancet 356 ( 9237 ), 1255–1259. 10.1016/s0140-6736(00)02799-9 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Farcas A., Bucsa C., Sinpetrean A., Leucuta D., Mogosan C., Dumitrascu D., et al. (2014). Preventability analysis of adverse drug reactions detected in two internal medicine departments in Romania . Intern. Emerg. Med. 9 ( 2 ), 187–193. 10.1007/s11739-012-0843-4 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Haile D. B., Ayen W. Y., Tiwari P. (2013). Prevalence and assessment of factors contributing to adverse drug reactions in wards of a tertiary care hospital, India . Ethiop. J. Health Sci. 23 ( 1 ), 39–48. [ PMC free article ] [ PubMed ] [ Google Scholar ]
• Hartwig S. C., Siegel J., Schneider P. J. (1992). Preventability and severity assessment in reporting adverse drug reactions . Am. J. Health. Syst. Pharm. 49 ( 9 ), 2229–2232. 10.1093/ajhp/49.9.2229 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Howard R. L., Avery A. J., Slavenburg S., Royal S., Pipe G., Lucassen P., et al. (2007). Which drugs cause preventable admissions to hospital? A systematic review . Br. J. Clin. Pharmacol. 63 ( 2 ), 136–147. 10.1111/j.1365-2125.2006.02698.x [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Khan L. M. (2013). Comparative epidemiology of hospital-acquired adverse drug reactions in adults and children and their impact on cost and hospital stay – A systematic review . Eur. J. Clin. Pharmacol. 69 ( 12 ), 1985–1996. 10.1007/s00228-013-1563-z [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Kovacevic M., Vezmar Kovacevic S., Radovanovic S., Stevanovic P., Miljkovic B. (2019). Adverse drug reactions caused by drug-drug interactions in cardiovascular disease patients: Introduction of a simple prediction tool using electronic screening database items . Curr. Med. Res. Opin. 35 ( 11 ), 1873–1883. 10.1080/03007995.2019.1647021 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Kurian J., Mathew J., Sowjanya K., Chaitanya K. R., Ramesh M., Sebastian J., et al. (2016). Adverse drug reactions in hospitalized pediatric patients: A prospective observational study . Indian J. Pediatr. 83 ( 5 ), 414–419. 10.1007/s12098-015-2002-1 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Lau P. M., Stewart K., Dooley M. J. (2003). Comment: Hospital admissions resulting from preventable adverse drug reactions . Ann. Pharmacother. 37 ( 2 ), 303–304. 10.1177/106002800303700229 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Leone R., Magro L., Moretti U., Cutroneo P., Moschini M., Motola D., et al. (2010). Identifying adverse drug reactions associated with drug-drug interactions: Data mining of a spontaneous reporting database in Italy . Drug Saf. 33 ( 8 ), 667–675. 10.2165/11534400-000000000-00000 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Letinier L., Ferreira A., Marceron A., Babin M., Micallef J., Miremont-Salame G., et al. (2020). Spontaneous reports of serious adverse drug reactions resulting from drug-drug interactions: An analysis from the French pharmacovigilance database . Front. Pharmacol. 11 , 624562. 10.3389/fphar.2020.624562 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Magro L., Arzenton E., Leone R., Stano M. G., Vezzaro M., Rudolph A., et al. (2020). Identifying and characterizing serious adverse drug reactions associated with drug-drug interactions in a spontaneous reporting database . Front. Pharmacol. 11 , 622862. 10.3389/fphar.2020.622862 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Magro L., Moretti U., Leone R. (2012). Epidemiology and characteristics of adverse drug reactions caused by drug-drug interactions . Expert Opin. Drug Saf. 11 ( 1 ), 83–94. 10.1517/14740338.2012.631910 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Mahady S. E., Margolis K. L., Chan A., Polekhina G., Woods R. L., Wolfe R., et al. (2021). Major GI bleeding in older persons using aspirin: Incidence and risk factors in the ASPREE randomised controlled trial . Gut 70 ( 4 ), 717–724. 10.1136/gutjnl-2020-321585 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Marques J., Ribeiro-Vaz I., Pereira A. C., Polonia J. (2014). A survey of spontaneous reporting of adverse drug reactions in 10 years of activity in a pharmacovigilance centre in Portugal . Int. J. Pharm. Pract. 22 ( 4 ), 275–282. 10.1111/ijpp.12078 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Marusic S., Sicaja M., Obreli Neto P. R., Franic M., Marinovic I., Bacic-Vrca V. (2014). Adverse drug reactions in elderly patients following discharge from an internal medicine clinic . Int. J. Clin. Pharmacol. Ther. 52 ( 10 ), 906–913. 10.5414/CP202041 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Mateti U., Rajakannan T., Nekkanti H., Rajesh V., Mallaysamy S., Ramachandran P. (2011). Drug-drug interactions in hospitalized cardiac patients . J. Young Pharm. 3 ( 4 ), 329–333. 10.4103/0975-1483.90246 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Mirosevic Skvrce N., Macolic Sarinic V., Mucalo I., Krnic D., Bozina N., Tomic S. (2011). Adverse drug reactions caused by drug-drug interactions reported to Croatian agency for medicinal products and medical devices: A retrospective observational study . Croat. Med. J. 52 ( 5 ), 604–614. 10.3325/cmj.2011.52.604 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Naranjo C. A., Busto U., Sellers E. M., Sandor P., Ruiz I., Roberts E. A., et al. (1981). A method for estimating the probability of adverse drug reactions . Clin. Pharmacol. Ther. 30 ( 2 ), 239–245. 10.1038/clpt.1981.154 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Obreli-Neto P. R., Nobili A., de Lyra D. P., Jr., Pilger D., Guidoni C. M., de Oliveira Baldoni A., et al. (2012b). Incidence and predictors of adverse drug reactions caused by drug-drug interactions in elderly outpatients: A prospective cohort study . J. Pharm. Pharm. Sci. 15 ( 2 ), 332–343. 10.18433/j3cc86 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Obreli-Neto P. R., Nobili A., de Oliveira Baldoni A., Guidoni C. M., de Lyra Junior D. P., Pilger D., et al. (2012a). Adverse drug reactions caused by drug-drug interactions in elderly outpatients: A prospective cohort study . Eur. J. Clin. Pharmacol. 68 ( 12 ), 1667–1676. 10.1007/s00228-012-1309-3 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Oscanoa T. J., Lizaraso F., Carvajal A. (2017). Hospital admissions due to adverse drug reactions in the elderly. A meta-analysis . Eur. J. Clin. Pharmacol. 73 ( 6 ), 759–770. 10.1007/s00228-017-2225-3 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Patel T. K., Bhabhor P. H., Desai N., Shah S., Patel P. B., Vatsala E., et al. (2015). Adverse drug reactions in a psychiatric department of tertiary care teaching hospital in India: Analysis of spontaneously reported cases . Asian J. Psychiatr. 17 , 42–49. 10.1016/j.ajp.2015.07.003 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Ren W., Liu Y., Zhang J., Fang Z., Fang H., Gong Y., et al. (2020). Prevalence of potential drug-drug interactions in outpatients of a general hospital in China: A retrospective investigation . Int. J. Clin. Pharm. 42 ( 4 ), 1190–1196. 10.1007/s11096-020-01068-3 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Roblek T., Vaupotic T., Mrhar A., Lainscak M. (2015). Drug-drug interaction software in clinical practice: A systematic review . Eur. J. Clin. Pharmacol. 71 ( 2 ), 131–142. 10.1007/s00228-014-1786-7 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Rottenkolber D., Schmiedl S., Rottenkolber M., Farker K., Salje K., Mueller S., et al. (2011). Adverse drug reactions in Germany: Direct costs of internal medicine hospitalizations . Pharmacoepidemiol. Drug Saf. 20 ( 6 ), 626–634. 10.1002/pds.2118 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Rukmangathen R., Brahmanapalli V. D., Thammisetty D. P., Pemmasani D., Gali S. D., Atmakuru R. B. (2020). Study of adverse drug reactions to antiretroviral therapy in a tertiary care hospital, Tirupati . Perspect. Clin. Res. 11 ( 4 ), 158–163. 10.4103/picr.PICR_133_18 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Schumock G. T., Thornton J. P. (1992). Focusing on the preventability of adverse drug reactions . Hosp. Pharm. 27 ( 6 ), 538. [ PubMed ] [ Google Scholar ]
• Scondotto G., Pojero F., Pollina Addario S., Ferrante M., Pastorello M., Visconti M., et al. (2018). The impact of polypharmacy and drug interactions among the elderly population in Western Sicily, Italy . Aging Clin. Exp. Res. 30 ( 1 ), 81–87. 10.1007/s40520-017-0755-2 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Shepherd G., Mohorn P., Yacoub K., May D. W. (2012). Adverse drug reaction deaths reported in United States vital statistics, 1999-2006 . Ann. Pharmacother. 46 ( 2 ), 169–175. 10.1345/aph.1P592 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Shi Q. P., He X. D., Yu M. L., Zhu J. X., Liu Y., Ding F., et al. (2014). A 6-year retrospective study of adverse drug reactions due to drug-drug interactions between nervous system drugs . Int. J. Clin. Pharmacol. Ther. 52 ( 5 ), 392–401. 10.5414/CP202050 [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Singh P., Agrawal M., Hishikar R., Joshi U., Maheshwari B., Halwai A. (2017). Adverse drug reactions at adverse drug reaction monitoring center in Raipur: Analysis of spontaneous reports during 1 year . Indian J. Pharmacol. 49 ( 6 ), 432–437. 10.4103/ijp.IJP_781_16 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
• Zheng W. Y., Richardson L. C., Li L., Day R. O., Westbrook J. I., Baysari M. T. (2018). Drug-drug interactions and their harmful effects in hospitalised patients: A systematic review and meta-analysis . Eur. J. Clin. Pharmacol. 74 ( 1 ), 15–27. 10.1007/s00228-017-2357-5 [ PubMed ] [ CrossRef ] [ Google Scholar ]

• Introduction
• Conclusions
• Article Information

ABC-110 indicates Study of Opaganib in Coronavirus Disease 2019 Pneumonia (COVID-19); ACE-ID-201, A Phase 2, Open Label, Randomized Study of the Efficacy and Safety of Acalabrutinib with Best Supportive Care Vs Best Supportive Care in Subjects Hospitalized with COVID-19; AMMURAVID, Factorial Randomized Trial of Remdesivir and Baricitinib Plus Dexamethasone for COVID-19; ColCOVID, Colchicine Counteracting Inflammation in COVID-19 Pneumonia; COMBAT-19, Mavrilimumab in Severe COVID-19 Pneumonia and Hyper-inflammation; COV-BARRIER, Study of Baricitinib (LY3009104) in Children With COVID-19; COVER, COVID Ivermectin - Randomized, Double-blind, Multi Centre Phase II, Proof of Concept, Dose Finding Clinical Trial on Ivermectin for the Early Treatment of COVID-19; DEF-IVID19, Defibrotide in COVID-19 Pneumonia - Use of Defibrotide to Reduce Progression of Acute Respiratory Failure Rate in Patients With COVID-19 Pneumonia; FibroCov, Open-label, Randomized, Parallel-arm Study Investigating the Efficacy and Safety of Intravenous Administration of Pamrevlumab Vs Standard of Care in Patients With COVID-19; HS216C17, Clinical Study to Evaluate the Performance and Safety of Favipiravir in COVID-19; IVIG/H/COVID-19, High Dose Intravenous Polyvalent Immunoglobulin (IVIG) in Patients With Early Inflammatory COVID-19; RANCONA, A Randomized Clinical Trial of Nafamostat: A Potent Transmembrane Protease Serine 2 (TMPRSS2) Inhibitor for the Treatment of Covid-19; REPAVID-19, Reparixin in COVID-19 Pneumonia - Efficacy and Safety; SOBI.IMMUNO-101, Efficacy and Safety of Emapalumab and Anakinra in Reducing Hyperinflammation and Respiratory Distress in Patients With COVID-19 Infection; SOLIDARITY, Efficacy of Different Anti-viral Drugs in COVID 19 Infected Patients; STAUNCH, Steroids and Unfractionated Heparin in Critically Ill Patients With Pneumonia From COVID-19 Infection; TOGETHER, Trial to Evaluate the Effect of Peginterferon Lambda for the Treatment of COVID-19; XPORT-CoV-1001, Evaluation of Activity and Safety of Oral Selinexor in Participants With Severe COVID-19 Infection.

eFigure. Flowchart of the Systematic Review

eTable 1. Potential DDIs for Each Drug Administered Against COVID-19 With the Degree of Severity of the Associated Adverse Outcomes and/or ADRs Found by Using the Drug Interaction Checkers

eTable 2. Reviewed Studies With Reported DDIs and Their Predictability (YES or NO) by Using the DDI Checkers

See More About

Sign up for emails based on your interests, select your interests.

Customize your JAMA Network experience by selecting one or more topics from the list below.

• Academic Medicine
• Acid Base, Electrolytes, Fluids
• Allergy and Clinical Immunology
• American Indian or Alaska Natives
• Anesthesiology
• Anticoagulation
• Art and Images in Psychiatry
• Artificial Intelligence
• Assisted Reproduction
• Bleeding and Transfusion
• Caring for the Critically Ill Patient
• Challenges in Clinical Electrocardiography
• Climate and Health
• Climate Change
• Clinical Challenge
• Clinical Decision Support
• Clinical Implications of Basic Neuroscience
• Clinical Pharmacy and Pharmacology
• Complementary and Alternative Medicine
• Consensus Statements
• Coronavirus (COVID-19)
• Critical Care Medicine
• Cultural Competency
• Dental Medicine
• Dermatology
• Diabetes and Endocrinology
• Diagnostic Test Interpretation
• Drug Development
• Electronic Health Records
• Emergency Medicine
• End of Life, Hospice, Palliative Care
• Environmental Health
• Equity, Diversity, and Inclusion
• Facial Plastic Surgery
• Gastroenterology and Hepatology
• Genetics and Genomics
• Genomics and Precision Health
• Global Health
• Guide to Statistics and Methods
• Hair Disorders
• Health Care Delivery Models
• Health Care Economics, Insurance, Payment
• Health Care Quality
• Health Care Reform
• Health Care Safety
• Health Care Workforce
• Health Disparities
• Health Inequities
• Health Policy
• Health Systems Science
• History of Medicine
• Hypertension
• Images in Neurology
• Implementation Science
• Infectious Diseases
• Innovations in Health Care Delivery
• JAMA Infographic
• Law and Medicine
• Leading Change
• Less is More
• LGBTQIA Medicine
• Lifestyle Behaviors
• Medical Coding
• Medical Devices and Equipment
• Medical Education
• Medical Education and Training
• Medical Journals and Publishing
• Mobile Health and Telemedicine
• Narrative Medicine
• Neuroscience and Psychiatry
• Notable Notes
• Nutrition, Obesity, Exercise
• Obstetrics and Gynecology
• Occupational Health
• Ophthalmology
• Orthopedics
• Otolaryngology
• Pain Medicine
• Palliative Care
• Pathology and Laboratory Medicine
• Patient Care
• Patient Information
• Performance Improvement
• Performance Measures
• Perioperative Care and Consultation
• Pharmacoeconomics
• Pharmacoepidemiology
• Pharmacogenetics
• Pharmacy and Clinical Pharmacology
• Physical Medicine and Rehabilitation
• Physical Therapy
• Physician Leadership
• Population Health
• Primary Care
• Professional Well-being
• Professionalism
• Psychiatry and Behavioral Health
• Public Health
• Pulmonary Medicine
• Regulatory Agencies
• Reproductive Health
• Research, Methods, Statistics
• Resuscitation
• Rheumatology
• Risk Management
• Scientific Discovery and the Future of Medicine
• Shared Decision Making and Communication
• Sleep Medicine
• Sports Medicine
• Stem Cell Transplantation
• Substance Use and Addiction Medicine
• Surgical Innovation
• Surgical Pearls
• Teachable Moment
• Technology and Finance
• The Art of JAMA
• The Arts and Medicine
• The Rational Clinical Examination
• Tobacco and e-Cigarettes
• Translational Medicine
• Trauma and Injury
• Treatment Adherence
• Ultrasonography
• Users' Guide to the Medical Literature
• Vaccination
• Venous Thromboembolism
• Veterans Health
• Women's Health
• Workflow and Process
• Wound Care, Infection, Healing

Get the latest research based on your areas of interest.

Others also liked.

• Download PDF
• X Facebook More LinkedIn
• CME & MOC

Conti V , Sellitto C , Torsiello M, et al. Identification of Drug Interaction Adverse Events in Patients With COVID-19 : A Systematic Review . JAMA Netw Open. 2022;5(4):e227970. doi:10.1001/jamanetworkopen.2022.7970

Manage citations:

© 2024

• Permissions

Identification of Drug Interaction Adverse Events in Patients With COVID-19 : A Systematic Review

• 1 Department of Medicine, Surgery, and Dentistry, Scuola Medica Salernitana, University of Salerno, Baronissi, Italy
• 2 Clinical Pharmacology Unit, San Giovanni di Dio e Ruggi d’Aragona University Hospital, Salerno, Italy
• 3 Doctoral School, Department of Medicine, Surgery and Dentistry “Scuola Medica Salernitana,” University of Salerno, Baronissi, Italy
• 4 Postgraduate Department of Medicine, Surgery and Dentistry “Scuola Medica Salernitana,” University of Salerno, Baronissi, Italy
• 5 Department of Medical and Surgical Sciences, Alma Mater Studiorum, University of Bologna, Bologna, Italy
• 6 Oncology Unit, San Giovanni di Dio e Ruggi d’Aragona University Hospital, Salerno, Italy
• 7 Department of Medicine and Health Sciences, University of Molise, Campobasso, Italy
• 8 Infectious Diseases Unit, San Giovanni di Dio e Ruggi d’Aragona University Hospital, Salerno, Italy

Question   Is it possible to assess adverse events associated with drug-drug interactions (DDIs) by drug interaction checkers in patients with COVID-19?

Findings   The DDIs identified in this systematic review involved 46 different drugs, with 575 DDIs for 58 drug pairs (305 associated with at least 1 adverse drug reaction) reported. Drug interaction checkers could have identified such events, including severe and life-threatening ones.

Meaning   Notwithstanding the emergency context of the COVID-19 pandemic, DDI-related adverse events should never be overlooked to customize the most effective and safest therapy.

Importance   During the COVID-19 pandemic, urgent clinical management of patients has mainly included drugs currently administered for other diseases, referred to as repositioned drugs. As a result, some of these drugs have proved to be not only ineffective but also harmful because of adverse events associated with drug-drug interactions (DDIs).

Objective   To identify DDIs that led to adverse clinical outcomes and/or adverse drug reactions in patients with COVID-19 by systematically reviewing the literature and assessing the value of drug interaction checkers in identifying such events.

Evidence Review   After identification of the drugs used during the COVID-19 pandemic, the drug interaction checkers Drugs.com, COVID-19 Drug Interactions, LexiComp, Medscape, and WebMD were consulted to analyze theoretical DDI-associated adverse events in patients with COVID-19 from March 1, 2020, through February 28, 2022. A systematic literature review was performed by searching the databases PubMed, Scopus, and Cochrane for articles published from March 1, 2020, through February 28, 2022, to retrieve articles describing actual adverse events associated with DDIs. The drug interaction checkers were consulted again to evaluate their potential to assess such events.

Findings   The DDIs identified in the reviewed articles involved 46 different drugs. In total, 575 DDIs for 58 drug pairs (305 associated with at least 1 adverse drug reaction) were reported. The drugs most involved in DDIs were lopinavir and ritonavir. Of the 6917 identified studies, 20 met the inclusion criteria. These studies, which enrolled 1297 patients overall, reported 115 DDI-related adverse events: 15 (26%) were identifiable by all tools analyzed, 29 (50%) were identifiable by at least 1 of them, and 14 (24%) remained nonidentifiable.

Conclusions and Relevance   The main finding of this systematic review is that the use of drug interaction checkers could have identified several DDI-associated adverse drug reactions, including severe and life-threatening events. Both the interactions between the drugs used to treat COVID-19 and between the COVID-19 drugs and those already used by the patients should be evaluated.

The COVID-19 pandemic has overwhelmed a completely unprepared world. Physicians have been faced with the challenge of caring for infected patients in the absence of consolidated scientific evidence and guidelines. 1 As a consequence, they have used drugs already approved for other diseases, referred to as repositioned drugs. 1 , 2 Especially at the beginning of the pandemic, the potential efficacy of these repositioned drugs against SARS-CoV-2 was often based on in vitro or in vivo evidence. 3 Some of these drugs have been used without considering their potential to cause adverse outcomes associated with drug-drug interactions (DDIs). 4 , 5

Drug-drug interactions, determined by pharmacokinetic and pharmacodynamic mechanisms, occur with high frequency in polytreated patients, such as patients with COVID-19. 5 The increase in adverse outcomes associated with DDIs and/or adverse drug reactions (ADRs) leads to increased hospital admissions and health care costs. Therefore, it is essential to avoid potential DDIs when establishing therapy. Drug interaction checkers are tools used to identify potential DDIs, supporting safe prescribing. This study aimed to identify DDIs that led to adverse clinical outcomes and/or ADRs in patients with COVID-19 by systematically reviewing the literature and assessing the value of drug interaction checkers in identifying such events.

The study design for this systematic review involved 4 steps. Step 1 involved the identification of all drugs used during the pandemic by consulting the European Medicines Agency and the Italian Medicines Agency websites, ClinicalTrials.gov database, and literature data. Step 2 involved searching for potential DDIs that involved each drug identified in step 1 using the following drug interaction checkers: Drugs.com, COVID-19 Drug Interactions, LexiComp, Medscape, and WebMD. Step 3 involved a literature systematic review to identify articles that reported adverse clinical outcomes and/or ADRs related to DDIs among COVID-19 treatments and with coadministered drugs. Step 4 involved evaluating whether the DDIs identified in step 3 could have been identified by using the tools listed in step 2.

To conduct a comprehensive systematic literature search, we used both controlled vocabulary and free-text terms. The following Medical Subject Heading terms were applied by using the Boolean operator AND: DDIs, COVID-19, patients with COVID-19, comedications, and ADRs. The PubMed, Scopus, and Cochrane databases were searched from the pandemic inception (March 1, 2020) up to February 28, 2022. Our research was limited to articles that involved patients with COVID-19 without sex and age restriction. Articles of any language that identified potential associations between DDIs and relevant clinical outcomes in patients with COVID-19 were included. A systematic review was performed, which identified 6917 studies, following the recommendations of the Meta-analysis of Observational Studies in Epidemiology ( MOOSE ) reporting guideline 6 and the PRISMA statement of reporting systematic review and meta-analysis. 7 This study did not need approval from an ethics committee or written informed consent from patients because it is a systematic review without meta-analysis.

Inclusion criteria were as follows: articles involving patients with a diagnosis of COVID-19, case reports and case series, letters to the editor and communications, observational studies, and interventional clinical trials. Exclusion criteria were as follows: articles that did not report a direct association between DDIs and clinically relevant outcomes in patients with COVID-19, reviews and meta-analyses, conference papers and book chapters, and studies in silico or based on in vitro experiments.

The drug interaction checkers used in this study were Drugs.com, COVID-19 Drug Interactions, LexiComp, Medscape, and WebMD. Drugs.com 8 generates a list of DDIs that are marked by a colored dot. Major DDIs (highly clinically significant; avoid combinations) are in red, moderate DDIs (moderately clinically significant; usually avoid combinations; use it only under special circumstances) are in orange, and minor DDIs (minimally clinically significant; minimize risk; assess risk and consider an alternative drug; take steps to circumvent the interaction risk and/or institute a monitoring plan) are in yellow. In COVID-19 Drug Interactions (made by Liverpool University), 9 the drugs are divided according to the risk of clinically significant interaction as follows: do not coadminister (with a red circle), potential interaction (with an orange square), potential weak interaction (with a yellow triangle), and no interaction expected (with a green rhombus). The LexiComp interactions tool 10 identifies DDIs, assigning the following risk rating: A, no known interaction; B, no action needed; C, monitor therapy; D, consider therapy modification; and X, avoid combination. LexiComp reports the drug class and the mechanism responsible for the interaction. The Medscape tool 11 classifies the DDIs as follows: contraindicated (in red), serious–use alternative (in orange), monitor closely (in green), and minor (in blue). The degree of severity is indicated by different shades of red: contraindicated (in dark red), serious–use alternative (in red), monitor closely (in pink), and minor (in light pink). The WebMD tool 12 classifies the DDI risk as follows: don’t use together (in red), serious (in orange), monitor closely (in yellow), and minor (in green).

The drugs used during the COVID-19 pandemic were identified by consulting the website of the European Medicines Agency, 13 Italian Medicines Agency, 14 ClinicalTrials.gov, 15 and literature data ( Figure ). The Figure shows all the 46 drugs listed in chronological order according to their period of use. Most of them were used under the concept of repurposing; some have been included in clinical trials or administered as off-label or compassionate use.

eTable 1 in the Supplement reports the number of potential DDIs for each drug administered against COVID-19 with the degree of severity of the associated adverse outcomes and/or ADRs found by using the drug interaction checkers. The tools differed from each other regarding the number of potential DDIs identified and the classification of the severity grade of the DDI-associated clinical outcomes. Drugs.com identified the largest numbmer of DDIs, followed by Medscape, WebMD, LexiComp, and COVID-19 Drug Interactions. Drugs.com identified the lirgest number of highly clinically significant DDI-associated adverse outcomes, followed by LexiComp, COVID-19 Drug Interactions, WebMD, and Medscape (eTable 1 in the Supplement ).

The drugs most involved in DDIs were lopinavir and ritonavir, followed by nirmatrelvir and ritonavir, darunavir and cobicistat, chloroquine, acetazolamide, and hydroxychloroquine. The drug interaction checkers agree with each other (even if with different classification methods) in considering lopinavir and ritonavir as the drug involved in the most serious DDI-associated ADRs (eTable 1 in the Supplement ).

A systematic review was performed to identify adverse clinical outcomes and/or ADRs related to DDIs among treatments of COVID-19 and between COVID-19 treatments and drugs coadministered in patients with COVID-19. The PRISMA algorithm 7 shows the research workflow (eFigure in the Supplement ). Twenty articles 16 - 35 that involved 46 interacting drugs that led to DDI-associated adverse outcomes were evaluated ( Table 1 ). The most frequent DDIs were hydroxychloroquine and lopinavir-ritonavir. Most DDIs resulted in QT-interval prolongation. Such a dangerous alteration was found 20 times, and in 12 of these 20 cases, it occurred together with other adverse outcomes, even leading to the death of 8 patients ( Table 1 ).

Eleven DDI-associated ADRs were diarrhea and vomiting as well as liver disorders. Six neurologic or psychiatric DDIs were reported. Three of 6 were serious neurovascular hemorrhages. One of them involved corticosteroids, hydroxychloroquine, and unfractionated heparin, another one implicated the aforementioned drugs coadministered with apixaban, and the last one included fractionated heparin. Surprisingly, none of the reviewed studies reported DDIs that involved ritonavir and anticoagulants. However, all drug interaction checkers agreed that the most severe DDIs occurred with ritonavir and direct factor Xa inhibitors.

Drugs coadministered with hydroxychloroquine were lopinavir-ritonavir (24 cases), acetazolamide (20 cases), and darunavir-cobicistat (15 cases). Of the 53 DDIs in which hydroxychloroquine was involved, 31 were associated with QT-interval prolongation. Four of these 31 DDIs led to patient death.

Only 3 of 56 DDIs involved coadministration of chloroquine with other drugs, including ceftriaxone, acetazolamide, and oseltamivir 18 or memantine and letrozole. 31 The major complication linked to chloroquine and hydroxychloroquine, in monotherapy or in combination and in short or low-dose regimens, was again QT-interval prolongation, which also caused fatal arrhythmias. 18

The DDIs identified in the reviewed articles involved 46 different drugs ( Table 1 ). Many of them were administered for patients’ comorbidities. Table 2 lists all the drugs reported in the articles and identified as triggers of DDIs by at least 1 of the drug interaction checkers used. Drugs.com was the most complete tool. Conversely, COVID-19 Drug Interactions, WebMD, Medscape, and LexiComp did not include some medications, such as memantine, letrozole, and magnesium sulfate ( Table 2 ).

In total, 575 DDIs for 58 drug pairs (305 associated with at least 1 ADR) were reported. Such DDIs were identified as follows: 70 by Medscape, 68 by COVID-19 Drug Interactions, 64 by Drugs.com, 55 by WebMD, and 48 by LexiComp. In 271 of 580 cases, no interactions were found. LexiComp reported the fewest DDIs, classified into B (no action needed) (10 [20%]), C (monitor therapy) (22 [45%]), D (consider modifying therapy) (10 [23%]), and X (avoid combinations) (6 [12%]).

The number of the identified severe-moderate DDI-associated adverse events was comparable among Drugs.com, Medscape, and WebMD. An equivalent classification was found using the latter 2 tools. Most DDIs were classified as major (30 [48%]) and moderate (32 [49%]) by Drugs.com, as serious (32 [46%]) and monitor closely (31 [44%]) by Medscape, and as serious (23 [43%]) and monitor closely (26 [46%]) by WebMD ( Table 2 ). In addition, DDI-associated adverse events were classified as minor by Drugs.com in 2 cases (3%), by Medscape in 5 cases (7%), and by WebMD in 4 cases (7%).

COVID-19 Drug Interactions identified DDI-associated adverse outcomes as follows: 15 (22%) as do not coadminister, 32 (46%) as potential interaction, and 3 (4%) as potential weak interaction ( Table 2 ). According to Medscape and WebMD, the most severe DDIs were caused by the association of amiodarone with lopinavir and ritonavir and lopinavir and ritonavir with simvastatin, classified as contraindicated (2 [3%]) by Medscape and as don’t use together (2 [4%]) by WebMD.

Globally, the reviewed studies described 15 patients taking lopinavir and ritonavir plus simvastatin 25 and only 1 taking lopinavir and ritonavir plus amiodarone. 24 The studies 24 , 25 reported liver toxicity (related to lopinavir and ritonavir plus simvastatin) and orthostatic syncope (related to lopinavir and ritonavir plus amiodarone). For all the tools, besides these serious DDI-associated adverse outcomes already described, the remaining 301 can be divided into 117 (39%) classified as severe, 132 (43%) as moderate, and 52 (17%) as minor.

eTable 2 in the Supplement details the last step of the study. Of the 6917 studies identified, 20 studies, which enrolled 1297 patients, reported 115 DDI-related adverse events: 15 (26%) were identifiable by all tools analyzed, 29 (50%) were identifiable by at least 1 of them, and 14 (24%) remained nonidentifiable. Most of these involved psychotic disorders or cutaneous reactions.

Therapeutic strategy to treat COVID-19 has rapidly changed during the pandemic, above all based on experimental and real-world data and following the concept of repurposing. Some drugs have fallen out of use, whereas others represent a cornerstone of treatment. 2 , 36 - 39 Both real-world data and results of clinical trials have highlighted the need to review all steps of the care process from the beginning of the pandemic to today. 37 In particular, what seems clear is the large variability in the therapeutic response of patients with COVID-19 and therefore the urgent need to use a personalized approach. 38 - 40 One important issue is that patients with comorbidities (thus polytreated), who represent most patients with COVID-19, are likely to experience ADRs, including those related to DDIs. Therefore, regardless of the drugs used for SARS-CoV-2 clearance and to treat COVID-19, it is crucial to take into account the risk of DDIs. 41

The current study was planned to analyze DDI-associated clinical outcomes that occurred in clinical practice during the pandemic and to investigate whether and how drug interaction checkers might be useful to assess them. Our main finding is that the use of these tools could have identified several DDI-associated ADRs, including severe and life-threatening events. However, the interactions between the drugs used to treat COVID-19 and between the COVID-19 drugs and those already used by the patients should be evaluated.

At the beginning of the pandemic, chloroquine and hydroxychloroquine were largely used because of their ability (assessed in vitro) to modify cellular pH, thus interfering with SARS-CoV-2 replication and its fusion with the host cells. 2 Then, as shown in the current study, hydroxychloroquine was recognized to interfere with the antiviral agents lopinavir-ritonavir, darunavir-cobicistat, and acetazolamide, causing QT-interval prolongation, ventricular arrhythmias, and torsade de pointes. 19 , 31 In the study by Borba et al, 18 several patients treated with chloroquine died after drug administration. Most patients (89.6%) with increased QT-interval prolongation were taking oseltamivir as well as acetazolamide and ceftriaxone. Crescioli et al 19 reported 5 deaths among 23 patients. These patients had developed QT-interval prolongation after the coadministration of hydroxychloroquine with at least 1 of the following drugs: darunavir-cobicistat, acetazolamide, amiodarone, lopinavir-ritonavir, haloperidol, citalopram, and trazodone. Martínez-López-de-Castro et al 25 reported that 3 of 44 deceased patients also had alteration of the QT interval associated with DDIs.

Lopinavir-ritonavir and darunavir-cobicistat were involved in most of the DDI-associated ADRs. Of importance, all the drug interaction checkers used in our study could have identified such events. This finding is not surprising, because these antivirals are inhibitors of cytochrome CYP3A4, which is the most involved isoenzyme of drug metabolism.

The interaction among hydroxychloroquine, darunavir-cobicistat, and tocilizumab can also lead to psychiatric disorders, such as behavioral disturbances, psychosis, agitation, delirium, and aggression. However, psychiatric ADRs were difficult to identify by the DDI tools. Martínez-López-de-Castro et al 25 evaluated 2 patients taking hydroxychloroquine and lopinavir-ritonavir who experienced psychiatric disorders, whereas Anmella et al 16 described 1 patient treated with acetazolamide, hydroxychloroquine, lopinavir-ritonavir, paroxetine, risperidone, and topiramate who had disturbing behavior.

None of the interaction tools identified the cutaneous ADRs that emerged from the systematic review. Martínez-López-de-Castro et al 25 identified 8 patients with COVID-19 who reported cutaneous reactions following administration of acetazolamide plus hydroxychloroquine and hydroxychloroquine plus lopinavir-ritonavir. Skroza et al 30 described erythematous rash, urticaria, and varicella-like blisters in 18 patients and 1 patient with a history of COVID-19 and late-onset urticarial vasculitis after healing.

Therapy must be chosen wisely, especially when dealing with drugs known to favor DDIs, such as anticoagulants. 42 In this regard, Ghani et al 22 described 3 patients treated with hydroxychloroquine and unfractionated or fractionated heparins or apixaban who had subarachnoid, severe cerebral edema, and intraparenchymal hemorrhages. A recent review 43 also highlighted the risk of QT-interval prolongation and cardiomyopathy attributable to the possibility of interaction between apixaban and hydroxychloroquine because of a mechanism of inhibition of CYP2C8 and P-glycoprotein.

Several potential DDIs that involved anticancer drugs used for the treatment of COVID-19 were also found (eTable 1 in the Supplement ). This finding is important considering that anticancer agents have a narrow therapeutic index and the ADRs are responsible for approximately 12% of hospitalizations in oncology units, almost 3 times more than in other medical areas. 44 , 45 Anticancer drugs belonging to the targeted therapy are mainly associated with QT liability and interact with concomitant medications, increasing the likelihood of life-threatening ventricular arrhythmia. 43 , 46 Nevertheless, our systematic review retrieved only 2 studies that reported potential DDIs that involved anticancer drugs. Szekely et al 31 indicated a potential DDI that involved letrozole coadministered with chloroquine and memantine, leading to torsade de pointes. However, none of the 5 drug interaction checkers detected such a DDI. Treon et al 34 documented a tachyarrhythmia potentially associated with acetazolamide, hydroxychloroquine, and ibrutinib administration. However, 4 of 5 drug interaction checkers recognized acetazolamide and hydroxychloroquine but not ibrutinib as responsible drugs for this DDI. No other DDI-associated adverse outcomes that involved ibrutinib were found despite this drug being a P-glycoprotein inhibitor and CYP3A4 substrate. 47

The experience of the pandemic offers the opportunity to improve therapy for patients with other diseases, such as rheumatological diseases, who have variable responses to the disease-modifying antirheumatic drugs. Identifying pretherapeutic and on-treatment factors associated with drug effectiveness is essential in this field. 48 The same goes for all drugs, including antivirals, anticoagulants, hypoglycemic agents, and antibiotics, whose use is not avoidable, especially in hospitalized patients. Recently, 2 oral antivirals were approved. One of them is molnupiravir, originally developed against influenza viruses. 49 , 50 The other one is an association of 2 protease inhibitors, nirmatrelvir and ritonavir. 51

Drug interaction checkers identified potential DDIs that involved nirmatrelvir-ritonavir and several drugs, such as colchicine, statins, antithrombotic, immunosuppressant, and antineoplastic agents, and DDIs that involved fluvoxamine combined with antidepressants, antiplatelet agents, benzodiazepines, and fentanyl. Conversely, only LexiComp identified a DDI between molnupiravir and cladribine. The reviewed studies 16 - 35 did not report DDI-associated clinical outcomes, conceivably because of the recent use of these COVID-19 drugs. However, potential DDIs should never be underestimated. In particular, even if nirmatrelvir-ritonavir has been specifically developed for the treatment of COVID-19, the presence of ritonavir should be emphasized.

This study has some limitations. Only 5 (although widely used and consolidated) available drug interaction checkers were accessed, with the risk of overlooking some DDI-associated ADRs that occurred in clinical practice. However, the concomitant use of tools with different classification methods can complicate the assessment of the DDI-associated outcomes. Similarly, we may have neglected studies included in gray literature (eg, congress proceedings) and emerging sources (eg, preprint websites). Moreover, except for the study by Crescioli et al, 19 which used the Naranjo algorithm, the other reviewed studies did not implement a causality assessment to ascertain the relationship between DDIs and the ADRs described. However, the aim of drug interaction checkers is to highlight the risk of DDI-associated ADRs to help physicians and patients to follow the most appropriate therapy and set up monitoring actions.

The findings of this systematic review of drug interactions among patients with COVID-19 reported in databases and the literature suggest that extreme caution should be used in choosing COVID-19 therapy, especially in polytreated patients. Although a critical emergency, such as the COVID-19 pandemic, might justify an urgent clinical approach, possible DDIs should never be ignored when choosing the most effective and safest therapy. In this context, support could and can still derive from drug interaction checkers, which help to perform a therapeutic reconciliation by stopping use of or withholding drugs and by intensifying clinical monitoring. Attention must be paid to concomitantly examine different sources of information to manage old and new drugs. The COVID-19 pandemic offers learning and opportunity to draw on new ideas and stimuli to optimize the care of all patients with complex conditions.

Accepted for Publication: March 2, 2022.

Published: April 19, 2022. doi:10.1001/jamanetworkopen.2022.7970

Open Access: This is an open access article distributed under the terms of the CC-BY License . © 2022 Conti V et al. JAMA Network Open .

Corresponding Author: Valeria Conti, PhD, Department of Medicine, Surgery, and Dentistry, Scuola Medica Salernitana, University of Salerno, Via S Allende, 84081 Baronissi, SA, Italy ( [email protected] ).

Author Contributions: Drs Conti and Filippelli had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Drs Conti and Sellitto contributed equally to this work.

Concept and design: Conti, Sellitto, Pagliano, Filippelli.

Acquisition, analysis, or interpretation of data: Conti, Sellitto, Torsiello, Manzo, De Bellis, Stefanelli, Bertini, Costantino, Maci, Raschi, Sabbatino, Corbi.

Drafting of the manuscript: Conti, Sellitto, Torsiello, Bertini, Sabbatino.

Critical revision of the manuscript for important intellectual content: Conti, Sellitto, Torsiello, Manzo, De Bellis, Stefanelli, Costantino, Maci, Raschi, Sabbatino, Corbi, Pagliano, Filippelli.

Supervision: Conti, Sellitto, Manzo, Stefanelli, Sabbatino, Corbi, Pagliano, Filippelli.

Conflict of Interest Disclosures: Dr Raschi reported receiving personal fees from Novartis outside the submitted work. No other disclosures were reported.

Additional Contributions: Native English-speaker Jan Festa, Morgan School, Salerno, Italy, revised the manuscript. She was not compensated for her work.

• Register for email alerts with links to free full-text articles
• Access PDFs of free articles
• Manage your interests
• Save searches and receive search alerts

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• My Account Login
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Open access
• Published: 02 September 2021

A machine learning framework for predicting drug–drug interactions

• Suyu Mei 1 &
• Kun Zhang 2  

Scientific Reports volume  11 , Article number:  17619 ( 2021 ) Cite this article

8245 Accesses

19 Citations

11 Altmetric

Metrics details

• Biochemistry
• Biological techniques
• Biotechnology
• Computational biology and bioinformatics
• Drug discovery
• Systems biology

Understanding drug–drug interactions is an essential step to reduce the risk of adverse drug events before clinical drug co-prescription. Existing methods, commonly integrating heterogeneous data to increase model performance, often suffer from a high model complexity, As such, how to elucidate the molecular mechanisms underlying drug–drug interactions while preserving rational biological interpretability is a challenging task in computational modeling for drug discovery. In this study, we attempt to investigate drug–drug interactions via the associations between genes that two drugs target. For this purpose, we propose a simple f drug target profile representation to depict drugs and drug pairs, from which an l 2 -regularized logistic regression model is built to predict drug–drug interactions. Furthermore, we define several statistical metrics in the context of human protein–protein interaction networks and signaling pathways to measure the interaction intensity, interaction efficacy and action range between two drugs. Large-scale empirical studies including both cross validation and independent test show that the proposed drug target profiles-based machine learning framework outperforms existing data integration-based methods. The proposed statistical metrics show that two drugs easily interact in the cases that they target common genes; or their target genes connect via short paths in protein–protein interaction networks; or their target genes are located at signaling pathways that have cross-talks. The unravelled mechanisms could provide biological insights into potential adverse drug reactions of co-prescribed drugs.

Similar content being viewed by others

Generative AI for designing and validating easily synthesizable and structurally novel antibiotics

Kyle Swanson, Gary Liu, … Jonathan M. Stokes

A small-molecule TNIK inhibitor targets fibrosis in preclinical and clinical models

Feng Ren, Alex Aliper, … Alex Zhavoronkov

RNAi-based drug design: considerations and future directions

Qi Tang & Anastasia Khvorova

Introduction

Drug–drug interactions (DDIs) have been recognized as a major cause of adverse drug reactions (ADRs) that leads to rising healthcare costs 1 . Antagonistic drug–drug interactions may occur when a patient takes more than one drug concurrently and potentially result in adverse side effects and toxicities 2 . In many cases, drug–drug interactions are hardly detected during the clinical trial phase, and arbitrary co-prescription of drugs without prior knowledge potentially poses serious threats to patient health and life 3 . Cytochrome-P450 (CYP450) isoforms (e.g., CYP1A2, CYP2C8, CYP2C9, CYP2C19, CYP2D6 and CYP3A4/5) take the responsibility to metabolize the majority of available drugs and frequently cause antagonistic drug–drug interactions 4 . For instance, CYP1A2 metabolizes both drug Theophylline and Duloxetine. If the stronger substrate Duloxetine competes with the weaker substrate Theophylline to bind to the active site of CYP1A2, breakdown of Theophylline will be reduced, leading to increased plasma levels of theophylline and potential side-effects like headache, nausea and vomiting 5 . To reduce the risk of potential adverse drug reactions, it is crucial to examine in advance whether co-prescribed drugs interact. Drug–drug interactions could be identified via in vitro or in vivo experiments as well as in silico computational methods. However, the former two approaches are very costly and in some cases are impossible to be carried out because the serious side effects DDIs elicited in experiments could do irreversible damages to human health 6 . With the advancement of pharmacogenomics, recent years have witnessed much effort to develop data-driven in silico computational methods to predict drug–drug interactions and their efficacy, although the “black-box” machine learning and artificial intelligence models sometimes frustrates the experimental pharmacologists in terms of multidisciplinary gap and practical successes 7

As regards drug–drug interactions, existing computational methods could be roughly classified into three categories, namely similarity-based methods 8 , 9 , 10 , 11 , networks-based methods 12 , 13 , 14 , 15 , 16 and machine learning methods 17 , 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 . Similarity-based methods directly infer drug–drug interactions on the basis of similarity scores between drug profiles. Vilar et al. 8 have reviewed several drug profiles, such as pharmaceutical profiles, gene expression profiles and phenome profiles, which have been used to infer drug repurposing, drug adverse effects and drug–drug interactions. Among these profiles, drug structural profiles could be well interpreted based on the assumption that structurally similar drugs tend to target the same or functionally-associated genes to produce similar drug efficacies 9 . The other major concern of similarity-based methods is to develop effective metrics to measure similarity between drug profiles. Ferdousi et al. 10 choose the optimum measure from a dozen of similarity metrics between drug target profiles (e.g., inner product, Jaccard similarity, Russell-Rao similarity and Tanimoto coefficient) to infer DDIs. In spite of simple and intuitive interpretation, similarity-based methods are easily affected by noise, for instance, the thresholding of similarity scores is seriously affected by false DDIs.

The second category of methods, i.e., networks-based methods, could be further classified into drug similarity networks-based methods 12 , 13 , 14 and protein–protein interaction (PPI) networks-based methods 15 , 16 . Drug similarity networks-based methods s predict novel links/DDIs via networks inference on the drug–drug similarity networks constructed via a variety of drug similarity metrics, e.g., matrix factorization 12 , 13 , block coordinate descent optimization 14 . Similar to the similarity-based methods 8 , 9 , 10 , 11 , these methods also resort to the similarities between drug structural profiles to infer DDIs. Comparatively, networks-based methods are more robust against noise than direct similarity-based methods. However, drug–drug interactions do not mean direct reactions between two structurally-similar drug molecules but synergistic enhancement or antagonistic attenuation of each other’s efficacy. When two drugs take actions on the same genes, associated metabolites or cross-talk signaling pathways, the biological events that two co-prescribed drugs influence or alter each other’s therapeutic effects may very well happen 10 . In this sense, the knowledge about what two drugs target is more useful and interpretable than drug structural similarity to infer drug–drug interactions, especially for the potential interactions between two drugs that are not structurally similar.

The PPI networks-based methods 15 , 16 assume that two drugs would produce unexpected perturbations to each other’s therapeutic efficacy if they simultaneously act on the same or associated genes, so that these methods have the merit of capturing the underlying mechanism of drug–drug interactions. Park et al. 15 assume two drugs interact if they cause close perturbation within the same pathway or distant perturbation within two cross-talk pathways, wherein the distant perturbation is captured via random walk algorithm on PPI networks. Huang et al. 16 also consider drug actions in the context of PPI networks. In their method, the target genes together with their neighbouring genes in PPI networks are defined as the target-centred system for a drug, and then a metric called S-score is proposed to measure the similarity between two drugs’ target-centered systems to infer drug–drug interactions. To date, PPI networks are far from complete and contain a certain level of noise so as to be restricted in the application to inferring drug–drug interactions.

The third category of methods, i.e., machine learning methods, has been widely used to infer drug–drug interactions 17 , 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 . Most of these methods focus on improving the performance of drug–drug interactions prediction via data integration. In these methods, data integration attempts to capture multiple aspects of information of a single data source or combining multiple heterogeneous data sources. Dhami et al. 17 attempt to combine multiple similarity metrics (e.g., molecular feature similarity, string similarity, molecular fingerprint similarity, molecular access system) from the sole data of drug SMILES representation. The other methods 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 all combine multiple data sources. Data integration often combines diverse feature information such as drug adverse drug reactions (ADR) 18 , 19 , 20 , 23 , 24 , target similarity 18 , 19 , 20 , 22 , 23 , 24 , PPI networks 23 , 24 , signaling pathways 19 and so on. Among these features, the information of drug chemical structures in the form of SMILES descriptors is most frequently used 17 , 18 , 19 , 20 , 21 , 22 , 23 , 24 . The machine learning frameworks used to integrate heterogeneous data include ensemble learning 18 , 19 , kernel methods 17 , 20 and deep learning 21 , 22 . Empirical studies show that data integration surely enrich the description of drugs from multiple aspects and accordingly improves the performance of drug–drug interaction prediction. However, data integration suffers from two major drawbacks. One drawback is that data integration increases data complexity. In most cases, we do not know which information is the most important and indispensable for predicting drug–drug interactions. Some information may contribute less to the prediction task. More importantly, data integration renders data constraint more demanding. Once any aspect of feature information is not available, e.g., drug molecular structure, the trained model may fail to work. Actually, single-task learning without data integration also can achieve satisfactory predictive performance, e.g., deep learning on available DDI networks only 25 . The other drawback of data integration is that the molecular mechanisms underlying drug–drug interactions is often ignored or drowned by the information flood. As results, the model is trained like a black-box and the predictions are hard to interpret in biological sense. Recent studies have revealed some molecular mechanisms drug–drug interactions, e.g., targeted gene profile and signaling pathway profile 26 . This information needs to be considered to increase model interpretability.

In this study, we attempt to simplify the computational modeling for drug–drug interaction prediction on the basis of potential drug perturbations on associated genes and signaling pathways. We assume that two drugs potentially interact when a drug alters the other drug’s therapeutic effects through targeted genes or signaling pathways. For this sake, only the known target genes of drugs taken from DrugBank 27 are used to train a predictive model without the information of drug structures or adverse drug reactions that are hard to represent and potentially are not available. The drug target profile is actually a binary vector indicating the presence or absence of a gene and the target profiles of two drugs are simply combined into a feature vector to depict a drug pair. To counteract the potential impact of noise, we choose l 2 -regularized logistic regression as the base learner. The proposed framework is evaluated via cross validation and independent test, wherein the external test data are taken from the comprehensive database 28 . We further propose several statistical metrics based on protein–protein interaction networks and signaling pathways to measure the intensity that drugs act on each other.

Data and methods

The known drug–drug interactions and drug–gene interactions are extracted from DrugBank 27 . As we use drug target profile to represent drugs and drug pairs, only the drugs that have been discovered to target at least one human gene are studied in this work. As results, we totally extract 6066 drugs and 2940 targeted human genes from DrugBank 27 . There are 915,413 drug–drug interactions and 23,169 drug–gene interactions associated with these drugs. As drug–drug interaction prediction is essentially a problem of binary supervised learning, we use the 915,413 drug pairs as the positive training data and randomly sample another 915,413 drug pairs from the 6066 drugs as the negative training data. The two classes of data are ensured to have no overlap.

The comprehensive database 28 provides a large repository for drug–drug interactions from experiments and text mining, some of which come from scattered databases such as DrugBank 27 , KEGG 29 , OSCAR 30 ( https://oscar-emr.com/ ), VA NDF-RT 31 and so on. After removing the drug–drug interactions that already exist in DrugBank 27 , we totally obtain 13 external datasets as positive independent test data, for instance, the largest 8188 drug–drug interactions from KEGG 29 . To estimate the risk of model bias, we randomly sample 8188 drug pairs as negative independent test data. These drug pairs are not overlapped with the training data and the positive independent test data.

To quantitatively estimate the intensity that two drugs perturbate each other’s efficacy, we build up comprehensive physical protein–protein interaction (PPI) networks from existing databases (HPRD 32 , BioGRID 33 , IntAct 34 , HitPredict 35 . We totally obtain 171,249 physical PPIs. From NetPath 36 , we obtain 27 immune signaling pathways with IL1–IL11 merged into one pathway for simplicity. From Reactome 37 , we obtain 1846 human signaling pathways.

Drug target profile-based feature construction

Drugs act on their target genes to produce desirable therapeutic efficacies. In most cases, drug perturbations could disperse to other genes through PPI networks or signaling pathways, so as to accidentally yield synergy or antagonism to the drugs targeting the indirectly affected genes. In this study, we depict drugs and drug pairs using drug target profile only. For each drug \({d}_{i}\) in the DDI-associated drug set \(D\) , its targeted human gene set is denoted as \({G}_{{d}_{i}}\) . The entire target gene set is defined as follows.

For each drug \({d}_{i}\) , drug target profile is formally defined as follows.

Then the drug target profile of a drug pair ( \({d}_{i},{d}_{j}\) ) is defined by combining the target profile of \({d}_{i}\) and \({d}_{j}\) as follows.

The genes \(g\notin G\) are discarded. The simple feature representation of drug target profile intuitively reveals the co-occurrence patterns of genes that a drug or drug pair targets. As an intuitive example, assuming the entire gene set \(G=\{TF,ALB,XDH,ORM1,ORM2\}\) , drug Patisiran (DB14582) targets the genes {ALB, ORM1, ORM2} and drug Bismuth Subsalicylate (DB01294) targets the genes {ALB, TF}, then Patisiran is represented with the vector [0, 1, 0, 1, 1] and Bismuth Subsalicylate is represented with the vector [1, 1, 0, 0, 0]. The drug pair (Patisiran, Bismuth Subsalicylate) is represented with the combined vector [1, 2, 0, 1, 1], which is used as the input of the base learner. All the data including the training set and the test set have the same feature descriptors. It is noted that all the target genes are chosen to represent drugs and drug pairs without giving priority or importance to the features, because the known target genes are very sparse and many target genes are unknown. If feature selection with importance weights is conducted, many drugs and drug pairs would be represented with null vector.

L 2 -regularized logistic regression as base learner

L 2 -regularized logistic regression 38 , well-known for its fast fitting large training data and penalizing potential noise and overtraining, is adopted as the base learner in this study. Given the training data x and labels y with each instance \({x}_{i}\) corresponding a class label \({y}_{i}\) , i.e., \(({x}_{i},{y}_{i}),i=\mathrm{1,2},...,l;{x}_{i}\in {R}^{n};{y}_{i}\in \{-1,+1\}\) , the decision function of logistic regression is defined as \(f(x)=\frac{1}{1+\mathit{exp}(-y{\omega }^{T}x)}\) . L 2 -regularized logistic regression derives the weight vector \(\omega\) via solving the optimization problem

where \(C\) denotes penalty parameter or regularizer. The second term penalizes potential noise/outlier or overtraining. The optimization problem ( 4 ) is solved via its dual form

where \({\alpha }_{i}\) denotes Lagrangian operator and \({Q}_{ij}={y}_{i}{y}_{j}{x}_{i}^{T}{x}_{j}\) . To simplify the parameter tuning, the regularizer C as defined in Formula ( 4 ) is chosen within the set \(\{{2}^{i}|-16\le i\le 16,i\in I\}\) , where I denotes the integer set.

Metrics for model performance and intensity of drug–drug interactions

Metrics for binary classification.

Frequently-used performance metrics for supervised classification include Receiver Operating Characteristic curve AUC (ROC-AUC), sensitivity (SE), precision (PR), Matthews correlation coefficient (MCC), accuracy and F1 score. Except that ROC-AUC is calculated based on the outputs of decision function \(f(x)\) , all the other metrics are calculated via confusion matrix M. The element \({M}_{i,j}\) records the counts that class i are classified to class j . From M , we first define several intermediate variables as Formula ( 6 ). Then we further define the performance metrics PR l , SE l and MCC l for each class label as Formula ( 7 ). The overall accuracy and MCC are defined by Formula ( 8 ).

where L denotes the number of labels and equals to 2 in this study. F1 score is defined as follows.

Metrics for intensity of drug–drug interactions

Two drugs perturbate each other’s efficacy through their targeted genes and the association between the targeted genes determines the interaction intensity of two drugs. If two drugs target common genes or different genes connected via short paths in PPI networks, we deem it as close interaction; if two drugs target different genes via long paths in PPI networks or across signaling pathways, we deem it as distant interaction; otherwise, the two drugs may not interact. If two drugs target common genes, the interaction could be regarded as most intensive and the intensity can be measured by Jaccard index. Given a drug pair ( \({d}_{i},{d}_{j}\) ), the Jaccard index between the two drugs is defined as follows

where \({G}_{{d}_{i}}\) and \({G}_{{d}_{j}}\) denote the target gene set of \({d}_{i}\) and \({d}_{j}\) , respectively. The larger the Jaccard index is, the more intensively the drugs interact. We use the threshold \(\xi\) to measure the level of interaction intensity. We further estimate the percentage of drug pairs whose interaction intensity exceeds \(\xi\) as follows

where \(U\) denotes the set of drug–drug interactions. If \(\xi ={ min}_{\forall ({d}_{i},{d}_{j})\in U}\frac{1}{|{ G}_{{d}_{i}}\cup { G}_{{d}_{j}}|}\) , then \({Sim}_{U}\) gives the percentage of drug pairs that target at least one common gene.

Two drugs may also interact through their target genes communicating via protein–protein interactions, although they do not target common genes. In these cases, we need to consider all the paths between two target genes in PPI networks. Given a gene pair ( \({g}_{i},{g}_{j}\) ), we use breadth-first graph search algorithm to search for all the paths between \(\mathrm{them}\) in human PPI networks, denotes as \({P}_{({g}_{i},{g}_{j})}\) . The length of the shortest path and longest path s denoted as \({S}_{({g}_{i},{g}_{j})}\) and \({L}_{({g}_{i},{g}_{j})}\) , respectively. We use the distance between target genes in terms of path length in PPI networks to define the distance between drugs. The average number of paths \({Avg}_{({d}_{i},{d}_{j})}\) , the shortest distance \({S}_{({d}_{i},{d}_{j})}\) and the longest distance \({L}_{({d}_{i},{d}_{j})}\) between drug \({d}_{i}\) and \({d}_{j}\) are defined as follows.

\({Avg}_{({d}_{i},{d}_{j})}\) indicates the number of paths through which two drugs interact. \({S}_{({d}_{i},{d}_{j})}\) indicates the most economical and effective way that two drugs interact. \({L}_{({d}_{i},{d}_{j})}\) indicates how far two drugs could alter each other’s efficacy, i.e., action range between two drugs. These three metrics are proposed to measure the interaction intensities between two drugs. Especially, \({S}_{({d}_{i},{d}_{j})}=0\) indicates that drug \({d}_{i}\) and \({d}_{j}\) target common genes, and \({Avg}_{({d}_{i},{d}_{j})}=0\) indicates that there are no paths between drug \({d}_{i}\) and \({d}_{j}\) and the two drugs do not interact.

Assuming K signaling pathways in total, if there exists a target gene \({g}_{j}\) of drug \({d}_{i}\) located in a signaling pathway \({Sig}_{k}\) , denoted as \({{g}_{j}\in Sig}_{k}\) , the pathway set associated with \({g}_{j}\) is defined as \({Sig}_{{g}_{j}}=\{{{{Sig}_{k}|g}_{j}\in Sig}_{k},k=\mathrm{1,2},\dots ,K\}\) . The signaling pathways targeted by \({d}_{i}\) is defined as \({\bigcup }_{{g}_{j}\in { G}_{{d}_{i}}}{Sig}_{{g}_{j}}\) , and then the common target signaling pathways between \({d}_{i}\) and \({d}_{j}\) are defined as \({Sig}_{({d}_{i},{d}_{j})}={\bigcup }_{{g}_{j}\in { G}_{{d}_{i}}}{Sig}_{{g}_{j}}\bigwedge {\bigcup }_{{g}_{j}\in { G}_{{d}_{j}}}{Sig}_{{g}_{j}}\) . The common target cellular processes between \({d}_{i}\) and \({d}_{j}\) are constructed in the same way, except that the signaling pathways are replaced with the GO terms of biological processes in GOA database 39 .

Performance of cross validation and independent test

The results of fivefold cross validation show that the proposed framework fairly encouraging performance (see Fig.  1 A for ROC-AUC scores and Table 1 for other metrics). The metrics of SP, SE and MCC on the two classes show that the proposed framework is less biased, e.g., 0.9556 on the positive class, 0.9402 on the negative class in terms of sensitivity and 0.9007 overall MMC. These results show that drug target profile alone is sufficient to separate interacting drug pairs from non-interacting drug pairs with a high accuracy (Accuracy = 94.79%). Drug takes effect via its targeted genes and the direct or indirect association or signaling between targeted genes underlies the mechanism of drug–drug interaction. From this aspect, drug target profile intuitively and effectively elucidates the molecular mechanism behind drug–drug interactions. Drug target profile could represent not only the genes targeted by structurally similar drugs but also the genes targeted by structurally dissimilar drugs, so that it is less biased than drug structural profile. The results also show that neither data integration nor drug structural information is indispensable for drug–drug interaction prediction. To more objectively gain knowledge about whether or not the model behaves stably, we evaluate the model performance with varying k -fold cross validation ( k  = 3, 5, 7, 10, 15, 20, 25) (see the Supplementary Fig. S1 ). The results show that the proposed framework achieves nearly constant performance in terms of Accuracy, MCC and ROC-AUC score with varying k -fold cross validation.

Performance of cross validation and independent test. ( A ) ROC curve and AUC score for fivefold cross validation. ( B ) Statistics of independent test data size. ( C ) Recall rates on the independent test data.

Cross validation still is prone to overfitting, though that the validation set is disjoint with the training set for each fold. We further conduct independent test on 13 external DDI datasets and one negative independent test data to estimate how well the proposed framework generalizes to unseen examples. The size of the independent test data varies from 3 to 8188 (see Fig.  1 B). The performance of independent test is in Fig.  1 C. The proposed framework achieves recall rates on the independent test data all above 0.8 except the dataset “DDI Corpus 2013”. On the experimental DDIs from KEGG 26 , OSCAR 27 and VA NDF-RT 28 , the proposed framework achieves recall rate 0.9497, 0.8992 and 0.9730, respectively (see Table 1 ). On the negative independent test data, the proposed framework also achieves 0.9373 recall rate, which indicates a low risk of predictive bias. The independent test performance also shows that the proposed framework trained using drug target profile generalizes well to unseen drug–drug interactions with less bias.

Comparisons with existing methods

Existing methods infer drug–drug interactions majorly via drug structural similarities in combination with data integration in many cases. Structurally similar drugs tend to target common or associated genes so that they interact to alter each other’s therapeutic efficacy. These methods surely capture a fraction of drug–drug interactions. However, structurally dissimilar drugs may also interact through their targeted genes, which cannot be captured by the existing methods based on drug structural similarities. In our proposed framework, direct or indirect associations between the target genes of two drugs are assumed to be the major driving force that induces drug–drug interactions, so as to capture both structurally-similar and structurally-dissimilar drug–drug interactions. From biological insights, the proposed framework is easier to interpret. From computational point of view, the proposed framework uses drug target profiles only and greatly reduces data complexity as compared to existing data integration methods.

From performance point of view, the proposed framework also outperforms existing methods. The performance comparisons are provided in Table 2 . All the existing methods achieve fairly high ROC-AUC scores except Cheng et al. 15 (ROC-AUC = 0.67). Unfortunately, these methods show a high risk of bias. For instance, the model proposed by Vilar et al. 9 , trained via drug structural profiles, is highly biased towards the negative class with sensitivity 0.68 and 0.96 on the positive and the negative class, respectively. The data integration method proposed by Zhang et al. 19 achieves encouraging performance of cross validation (ROC-AUC score = 0.957, PR = 0.785, SE = 0.670) but only recognizes 7 out of 20 predicted DDIs (equivalent to 35% recall rate of independent test), although it exploits a large amount of feature information such as drug substructures, drug targets, drug enzymes, drug transporters, drug pathways, drug indications and drug side-effects. Similarly, Gottlieb et al. 23 achieve fairly good performance of cross validation but achieve only 53% recall rate of independent test.

Deep learning, the most promising revolutionary technique to date in machine learning and artificial intelligence, has been used to predict the effects and types of drug–drug interactions 21 , 22 . The most related deep learning framework proposed by Karim et al. 25 automatically learns feature representations from the structures of available drug–drug interaction networks to predict novel DDIs. This method also achieves satisfactory performance (ROC-AUC score = 0.97, MCC = 0.79, F1 score = 0.91), but the learned features are hard to interpret and to provide biological insights into the molecular mechanisms underlying drug–drug interactions.

Analyses of molecular mechanisms behind drug–drug interactions

Jaccard index between two drugs.

The more common genes two drugs target, the more intensively the two drugs potentially interact. As presented in Formula ( 10 ), the interaction intensity is measured with Jaccard index. The percentage of drug pairs whose interaction intensity exceeds ξ is illustrated in Fig.  2 . The threshold of interaction intensity assumes \(\xi ={ min}_{\forall ({d}_{i},{d}_{j})\in U}\frac{1}{|{ G}_{{d}_{i}}\cup { G}_{{d}_{j}}|}\) and \(\xi =0.5\) in Fig.  2 A,B, respectively. The statistics are derived from the training data. We can see that interacting drugs tend to target much more common genes than non-interacting drugs.

Statistics of common target genes between interacting and non-interacting drugs.

Average number of paths between two drugs

The average number of paths between the garget genes of two drugs as defined in Formula ( 12 ) also measures the interaction intensity between drugs. To reduce the time of paths search, we only randomly choose 9692 interacting drug pairs and 9692 non-interacting drug pairs as examples for the analyses of molecular mechanism behind drug–drug interactions. The average number of paths of top twenty drug pairs are illustrated in Fig.  3 A. We can see that interacting drug pairs have their target genes more heavily connected than non-interacting drug pairs, which also means the more paths two drugs are connected through, the more probably the two drugs interact to alter each other’s effects. As shown in Fig.  3 B, non-interacting drugs are more likely to be unreachable to each other than interacting drugs.

The statistics of average number of paths, shortest path lengths and longest path lengths between two drugs.

Shortest path length between two drugs

For the randomly sampled 9692 interacting drug pairs and 9692 non-interacting drug pairs, the length of the shortest paths between two drugs’ target genes ranges from 0 to 5 (see Fig.  3 C). We can see that interacting drug pairs significantly outnumber non-interacting drug pairs when the shortest path length is equal to 0, that’s, that two drugs target common genes. With the increase of the shortest path length, non-interacting drug pairs gradually outnumber interacting drug pairs. These results show that drug–drug interactions tend to occur between drugs that target common genes or whose target genes come across via shorter shortest paths. The shorter the shortest path is, the more efficiently the drugs interact.

Longest path length between two drugs

For the randomly sampled drug pairs, the length of the longest paths between two drugs’ target genes ranges from 0 to 8 (see Fig.  3 D). Non-interacting drug pairs outnumber interacting drug pairs when the longest path ranges from 3 to 5, but conversely interacting drug pairs significantly outnumber non-interacting drug pairs when the longest path length equals to 6. These results to some extent show that interacting drugs could exert far-reaching perturbations on each other with a longer range of action than non-interacting drugs. The metrics \({Avg}_{({d}_{i},{d}_{j})},\) \({S}_{({d}_{i},{d}_{j})}\) and \({L}_{({d}_{i},{d}_{j})}\) defined in Formula ( 12 ) could measure the tendency of drug–drug interaction in terms of interaction intensity, interaction efficiency and action range. When the shortest path length equals to 0 and the longest path length equals to 6, the randomly sampled interacting and on-interacting drug pairs show a significant statistical difference.

Common target pathways between two drugs

We map the target genes onto the signaling pathways from NetPath 36 and Reactome 37 to investigate that interacting drugs tend to target common signaling pathways. Computational results show that interacting drug pairs tend to target more common signaling pathways than non- interacting drug pairs (see Fig.  4 A for NetPath pathways and Fig.  4 B for Reactome pathways). If the target genes of two drugs are located in the same signaling pathway, the two drugs are more likely to perturbate each other’s efficacies.

Statistics of common signaling pathways that two drugs target and common cellular processes that two drugs are involved in.

Common cellular processes between two drugs

As shown in Fig.  4 C, interacting drugs are more likely to get involved in common cellular processes than non-interacting drugs. This phenomenon is not hard to understand. Two drugs whose target genes are involved in common cellular processes more likely alter each other’s therapeutic effects.

Predictions and clinical implications

We randomly sample 99,986 drug pairs as the prediction set, which are not overlapped with the training data and the independent test data. Thereinto, 43,719 drug pairs are predicted to interact by the proposed framework (see Supplementary File S1 ). These predictions to some extent contain a certain level of false interactions. For each prediction, a confidence level in the form of probability could be chosen to filter out the weak interactions (e.g., 0.7 probability as a threshold). These predictions are further analysed from the aspect of cellular processes (see Supplementary File S2 ) and signaling pathways (see Supplementary File S3 ) to help us understand the molecular mechanisms underlying drug–drug interactions. We choose the drug Nabiximols and Glucosamine as a case study.

Nabiximols (C 42 H 60 O 4 ), extracted from Cannabis sativa L., is often used to treat neuropathic pain and intractable cancer pain, with the pharmacological effects of analgesic, muscle relaxant, anxiolytic, neuroprotective and anti-psychotic activity ( https://www.drugbank.ca/drugs/DB14011 ). Glucosamine (C 6 H 13 NO 5 ), as a precursor for glycosaminoglycans that are a major component of joint cartilage, is often used to rebuild cartilage and treat osteoarthritis ( https://www.drugbank.ca/drugs/DB01296 ). According to DrugBank 27 , Nabiximols targets 57 human genes and Glucosamine targets six human genes. Based on these target genes, we could analyse the cellular processes and signaling pathways through which Nabiximols and Glucosamine take effect.

Common cellular processes between Nabiximols and Glucosamine

Two drugs mediate common cellular processes via common target genes or associated target genes involved in the same cellular processes. Computational results show that Nabiximols and Glucosamine get involved 68 common cellular processes. For clarity, only 21 cellular processes and the associated target genes are illustrated in Fig.  5 . The rest cellular processes are provided in Supplementary File S2 . As shown in Fig.  5 , Nabiximols and Glucosamine mediate the common cellular processes of exogenous drug catabolic process (GO:0042738) and drug metabolic process (GO:0017144) via the common gene CYP2C19 . Association via different target genes is one major way that two drugs mediate common cellular processes. For instance, Nabiximols and Glucosamine mediate the common cellular processes of negative regulation of smooth muscle cell proliferation (GO:0048662) via Nabiximols-targeted gene PPARG and Glucosamine-targeted gene IFNG . For another example, Nabiximols and Glucosamine mediate the common cellular processes of regulation of reactive oxygen species (ROS) metabolic process (GO:2000377) via Nabiximols-targeted gene CYP1B1 and Glucosamine-targeted gene TNF . Among the predicted drug–drug interactions, many drug pairs do not target common genes but they are found to mediate common cellular processes via different target genes (see Supplementary File S2 ). For instance, drug Nabiximols (DB14011) and Gallium nitrate (DB05260) are not found to target common genes in DrugBank 27 , but they are predicted to target the common cellular processes of neutrophil chemotaxis (GO:0030593), positive regulation of NF-kappaB transcription factor activity (GO:0051092), etc.

Common cellular processes of target genes between DB14011|Nabiximols and DB01296|Glucosamine predicted to interact. Red triangle nodes denote drugs; green circle nodes denote drug target genes; light red circle nodes denote common target genes; and yellow diamond nodes denote biological processes of gene ontology. This drawing is created by Cytoscape version 2.8.2 ( https://cytoscape.org/ ).

Common signaling pathways between Nabiximols and Glucosamine

The common Reactome signaling pathways that Nabiximols and Glucosamine mediate are illustrated in Fig.  6 . Among the target genes, the common target gene CYP2C19 is located in four Reactome signaling pathways, i.e., Synthesis of epoxy (EET) and dihydroxyeicosatrienoic acids (DHET) (R-HSA-2142670), Xenobiotics (R-HSA-211981), CYP2E1 reactions (R-HSA-211999) and Synthesis of (16-20)-hydroxyeicosatetraenoic acids (HETE) (R-HSA-2142816). Apart from common garget genes, association via different target genes also leads to two drugs mediating common signaling pathways. For instance, Nabiximols and Glucosamine mediate the common signaling pathway of Neutrophil degranulation (R-HSA-6798695) via Nabiximols-targeted gene ALOX5 and Glucosamine-targeted gene MMP9 . Two drugs that do not target common genes also potentially mediate the same signaling pathways (see Supplementary File S3 ). For instance, drug Nabiximols (DB14011) and SF1126 (DB05210) have not been reported to target common genes in DrugBank 27 , but they are predicted to mediate several common signaling pathways, e.g., Regulation of PTEN gene transcription (R-HSA-8943724), Interleukin-4 and Interleukin-13 signaling (R-HSA-6785807), G alpha (q) signaling events (R-HSA-416476).

Common target Reactome signaling pathways between DB14011|Nabiximols and DB01296|Glucosamine predicted to interact. Red triangle nodes denote drugs; green circle nodes denote drug target genes; light red circle nodes denote common target genes; and blue hexagon nodes denote Reactome signaling pathways. This drawing is created by Cytoscape version 2.8.2 ( https://cytoscape.org/ ).

Only after co-prescribed drugs have clinically done damages to patient health and life, could drug–drug interactions be detected and reported in most cases. For this reason, we need resort to computational methods to predict whether two drugs interact and produce undesirable side effects before clinical co-prescription. Existing computational methods focus on integrating multiple heterogeneous data sources to increase model performance, among which drug structural profile is the most frequently used feature information. These methods heavily depend on drug structures and assume that structurally similar drugs often target common or associated genes so as to alter each other’s therapeutic efficacies. This assumption surely captures a fraction of drug–drug interactions but shows bias, because it ignores a large fraction of interactions between structurally dissimilar drugs. The other major drawback of these methods lies in the high data complexity. In these methods, we do not know which information contributes most to the model performance and it is hard to interpret the molecular mechanisms behind drug–drug interactions. Furthermore, data integration would fail when the required data are not available, e.g., drug structures, drug side-effects, clinical records. Lastly, proper representation of drug molecule structures and extracting features from drug SMILES remain challenging in the progress of computational modelling for drug development. In this study, we use drug target profile to depict drugs and drug pairs to achieve two goals. One goal is to simplify the modeling processes via reducing data complexity and relieving dependency on drug molecular structures. The other goal is to computationally model the molecular mechanisms underlying drug–drug interactions so that the model is biologically interpretable. Drugs act on their target genes to produce desirable therapeutic efficacies. We assume that the perturbations of two drugs come across through common target genes, paths in PPI networks or signaling pathways, synergistic enhancement or antagonistic counteract of therapeutic effects of individual drugs would take place. As compared to the existing methods, this proposed framework bases the assumption of drug–drug interactions on drug–targeted genes instead of drug structural similarities. We use the known drug–drug interactions from DrugBank 27 as the positive training data and randomly sample the same size of drug pairs as the negative training data to train an l 2 -regualrized logistic regression model. K -fold cross validation is a common practice used to estimate model performance, but the performance varies with the choice of k . The best practice is to choose k at intervals (e.g., k  = 3, 5, 10, 15, …) or even conduct leave-one-out cross validation, so that we could more objectively know whether or not the model behaves stably. However, this practice is computationally prohibitive to large training data (915,413 positive examples and 915,413 negative examples) and thirteen external test datasets with tedious model parameters tuning. Actually, it is hard to obtain a training set representative of and infinitely approximate to the population distribution via varying k -folds. Nevertheless, we still evaluate the model performance with varying k -fold cross validation ( k  = 3, 5, 7, 10, 15, 20, 25). The results show that the performance in terms of Accuracy, MCC and ROC-AUC score is fairly stable with k varying widely. Apart from horizontally randomizing examples (X-randomization), some statistical machine learning models such as Random Forest also conduct vertical feature randomization (Y-randomization) to obtain different views or to evaluate feature importance. Because the known target genes are very sparse and thus random sampling of feature subsets potentially results in null vector representation of drug pairs, we choose all the features in this study.

Empirical studies show that the proposed framework achieves fairly encouraging performance of fivefold cross validation and independent test on thirteen external datasets, which significantly outperforms the existing methods. Furthermore, the encouraging performance on the randomly sampled negative independent test data shows that the proposed framework is less biased. Nevertheless, the proposed framework yields a little large fraction of false interactions, which is largely due to the quality of randomly sampled negative training data. This problem could be to some extent solved by choosing a higher threshold of probability to filter out the weak predictions. In addition, drug target profile simplifies computational modeling, but meanwhile restricts the application of the proposed framework in that the target genes have not been reported for many less-studied drugs. This problem could be solved with the accumulation of the knowledge about drug target genes. The proposed framework could to some extent to be generalized to the other problems concerned with drug discovery, e.g., drug combinatorial synergy and antagonism, drug side-effects, drug–food interaction, etc., in which drug target profile could still be useful. Whether or not drug target profile representation is sufficient to solve these problems need to be further investigated.

We further propose several statistical metrics based on protein–protein interaction networks and signaling pathways to measure the intensity that drugs act on each other. These metrics show that two drugs tend to interact more efficiently if their perturbations could come across via shorter shortest paths in PPI networks, and the perturbations would be more far-reaching if longer shortest paths between the two drugs. Lastly, we use the common cellular processes and signaling pathways that two drugs target to understand the mechanisms underlying drug–drug interactions. The unravelled mechanisms are useful to provide biological insights into potential pharmacological risks of known drug–drug interactions.

Conclusions

Drug target profile representation of drugs and drug pairs simplifies the modeling processes for drug–drug interactions by reducing both data complexity and dependency on drug molecular structures. Meanwhile, Drug target profile representation renders the proposed framework biologically interpretable in terms of molecular mechanisms underlying drug–drug interactions.

Code availability

The source code and tools for this proposed framework are publicly available at https://github.com/suyumei/DrugDrugIntact.git .

Wienkers, L. C. & Heath, T. G. Predicting in vivo drug interactions from in vitro drug discovery data. Nat. Rev. Drug Discovery 4 , 825–833 (2005).

Article   CAS   Google Scholar  

Edwards, I. R. & Aronson, J. K. Adverse drug reactions: Definitions, diagnosis, and management. Lancet 356 , 1255–1259 (2000).

Leape, L. L. et al. Systems analysis of adverse drug events. ADE Prevention Study Group. JAMA 274 , 35–43 (1995).

Steyn, S. J. & Varma, M. V. S. Cytochrome-P450-mediated drug–drug interactions of substrate drugs: Assessing clinical risk based on molecular properties and an extended clearance classification system. Mol. Pharm. 17 (8), 3024–3032 (2020).

Deodhar, M. et al. Mechanisms of CYP450 inhibition: Understanding drug–drug interactions due to mechanism-based inhibition in clinical practice. Pharmaceutics 12 (9), 846 (2020).

Duke, J. D. et al. Literature based drug interaction prediction with clinical assessment using electronic medical records: Novel myopathy associated drug interactions. PLoS Comput. Biol. 8 , e1002614 (2012).

Medina-Franco, J. L. et al. Rationality over fashion and hype in drug design [version 1; peer review: 2 approved]. F1000Research 10 (Chem Inf Sci), 397 (2021).

Article   Google Scholar  

Vilar, S. & Hripcsak, G. The role of drug profiles as similarity metrics: Applications to repurposing, adverse effects detection and drug–drug interactions. Brief Bioinform. 18 , 670–681 (2017).

CAS   PubMed   Google Scholar  

Vilar, S. et al. Drug–drug interaction through molecular structure similarity analysis. J. Am. Med. Inform. Assoc. 19 , 1066–1074 (2012).

Ferdousi, R., Safdari, R. & Omidi, Y. Computational prediction of drug–drug interactions based on drugs functional similarities. J. Biomed. Inform. 70 , 54–64 (2017).

Vilar, S. et al. Similarity-based modeling in large-scale prediction of drug–drug interactions. Nat. Protoc. 9 , 2147–2163 (2014).

Zhang, W., Chen, Y., Li, D. & Yue, X. Manifold regularized matrix factorization for drug–drug interaction prediction. J. Biomed. Inform. 88 , 90–97 (2018).

Shtar, G., Rokach, L. & Shapira, B. Detecting drug–drug interactions using artificial neural networks and classic graph similarity measures. PLoS ONE 14 , e0219796 (2019).

Zhang, P., Wang, F., Hu, J. & Sorrentino, R. Label propagation prediction of drug–drug interactions based on clinical side effects. Sci. Rep. 5 , 12339 (2015).

Article   ADS   Google Scholar  

Park, K., Kim, D., Ha, S. & Lee, D. Predicting pharmacodynamic drug–drug interactions through signaling propagation interference on protein–protein interaction networks. PLoS ONE 10 , e0140816 (2015).

Huang, J. et al. Systematic prediction of pharmacodynamic drug–drug interactions through protein–protein-interaction network. PLoS Comput Biol 9 , e1002998 (2013).

Article   ADS   CAS   Google Scholar  

Dhami, D. S., Kunapuli, G., Das, M., Page, D. & Natarajan, S. Drug–drug interaction discovery: Kernel learning from heterogeneous similarities. Smart Health (Amst.) 9–10 , 88–100 (2018).

Cheng, F. & Zhao, Z. Machine learning-based prediction of drug–drug interactions by integrating drug phenotypic, therapeutic, chemical, and genomic properties. J. Am. Med. Inform. Assoc. 21 , e278–e286 (2014).

Zhang, W. et al. Predicting potential drug–drug interactions by integrating chemical, biological, phenotypic and network data. BMC Bioinform. 18 , 18 (2017).

Song, D. et al. Similarity-based machine learning support vector machine predictor of drug–drug interactions with improved accuracies. J. Clin. Pharm. Ther. 44 , 268–275 (2019).

Ryu, J. Y., Kim, H. U. & Lee, S. Y. Deep learning improves prediction of drug–drug and drug-food interactions. Proc. Natl. Acad. Sci. USA 115 , E4304–E4311 (2018).

Lee, G., Park, C. & Ahn, J. Novel deep learning model for more accurate prediction of drug–drug interaction effects. BMC Bioinform. 20 , 415 (2019).

Gottlieb, A., Stein, G. Y., Oron, Y., Ruppin, E. & Sharan, R. INDI: a computational framework for inferring drug interactions and their associated recommendations. Mol. Syst. Biol. 8 , 592 (2012).

Qian, S., Liang, S. & Yu, H. Leveraging genetic interactions for adverse drug–drug interaction prediction. PLoS Comput. Biol. 15 , e1007068 (2019).

Karim, M.R., Cochez, M., Jares, J.B., Uddin, M., Beyan, O., Decker, S. Drug–drug interaction prediction based on knowledge graph embeddings and convolutional-LSTM network. (2019). arXiv:1908.01288.

Jia, J. et al. Mechanisms of drug combinations: Interaction and network perspectives. Nat. Rev. Drug Discov. 8 , 111–128 (2009).

Wishart, D. S. et al. DrugBank 5.0: A major update to the DrugBank database for 2018. Nucleic Acids Res 46 , D1074–D1082 (2018).

Ayvaz, S. et al. Toward a complete dataset of drug–drug interaction information from publicly available sources. J. Biomed. Inform. 55 , 206–217 (2015).

Kanehisa, M. et al. Data, information, knowledge and principle: Back to metabolism in KEGG. Nucleic Acids Res 42 (Database issue), D199–D205 (2014).

Crowther, N. R., Holbrook, A. M., Kenwright, R. & Kenwright, M. Drug interactions among commonly used medications. Chart simplifies data from critical literature review. Can. Fam. Phys. 43 , 1972–1976 (1997) ( 1979–1981 ).

CAS   Google Scholar  

Olvey, E. L., Clauschee, S. & Malone, D. C. Comparison of critical drug–drug interaction listings: The department of Veterans Affairs medical system and standard reference compendia. Clin. Pharmacol. Ther. 87 , 48–51 (2010).

Keshava Prasad, T. S. et al. Human protein reference database—2009 update. Nucleic Acids Res. 37 (Database issue), D767–D772 (2009).

Chatr-Aryamontri, A. et al. The BioGRID interaction database: 2015 update. Nucleic Acids Res 43 (Database issue), D470–D478 (2015).

Orchard, S., Ammari, M., Aranda, B., Breuza, L. & Briganti, L. The MIntAct project–IntAct as a common curation platform for 11 molecular interaction databases. Nucleic Acids Res. (Database issue) 42 , D358–D363 (2014).

López, Y., Nakai, K., Patil, A. HitPredict version 4: Comprehensive reliability scoring of physical protein–protein interactions from more than 100 species. Database (Oxford) . 2015:bav117 (2015).

Kandasamy, K. et al. NetPath: A public resource of curated signal transduction pathways. Genome Biol. 11 , R3 (2010).

Fabregat, A. et al. The Reactome pathway knowledgebase. Nucleic Acids Res. 46 (Database issue), D649–D655 (2018).

Fan, R., Chang, K., Hsieh, C., Wang, X. & Lin, C. LIBLINEAR: A library for large linear classification. Mach. Learn Res. 9 , 1871–1874 (2008).

MATH   Google Scholar  

Barrell, D. et al. The GOA database in 2009—An integrated Gene Ontology Annotation resource. Nucleic Acids Res. 37 (Database issue), D396–D403 (2009).

Download references

Acknowledgements

This work is partly supported by the funding from the NIH Grants 2U54MD007595. The contents are solely the responsibility of the authors and do not necessarily represent the official views of the NIH.

Author information

Authors and affiliations.

Software College, Shenyang Normal University, Shenyang, 110034, China

Bioinformatics Core of Xavier RCMI Center for Cancer Research, Department of Computer Science, Xavier University of Louisiana, New Orleans, LA, 70125, USA

You can also search for this author in PubMed   Google Scholar

Contributions

M.S. conducted the study and wrote the paper. Z.K. revised the manuscript. All the authors reviewed the final manuscript.

Corresponding authors

Correspondence to Suyu Mei or Kun Zhang .

Ethics declarations

Competing interests.

The authors declare no competing interests.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Supplementary figure s1., supplementary information 1., supplementary information 2., supplementary information 3., rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Mei, S., Zhang, K. A machine learning framework for predicting drug–drug interactions. Sci Rep 11 , 17619 (2021). https://doi.org/10.1038/s41598-021-97193-8

Download citation

Received : 07 May 2021

Accepted : 18 August 2021

Published : 02 September 2021

DOI : https://doi.org/10.1038/s41598-021-97193-8

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

This article is cited by

Recent trends and perspectives of artificial intelligence-based machine learning from discovery to manufacturing in biopharmaceutical industry.

• Ravi Maharjan
• Jae Chul Lee
• Seong Hoon Jeong

Journal of Pharmaceutical Investigation (2023)

Machine learning-based quantitative prediction of drug exposure in drug-drug interactions using drug label information

• Ha Young Jang
• Jihyeon Song

npj Digital Medicine (2022)

By submitting a comment you agree to abide by our Terms and Community Guidelines . If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

Sign up for the Nature Briefing: Translational Research newsletter — top stories in biotechnology, drug discovery and pharma.

• Search Menu
• Chemical Biology and Nucleic Acid Chemistry
• Computational Biology
• Critical Reviews and Perspectives
• Data Resources and Analyses
• Gene Regulation, Chromatin and Epigenetics
• Genome Integrity, Repair and Replication
• Methods Online
• Molecular Biology
• Nucleic Acid Enzymes
• RNA and RNA-protein complexes
• Structural Biology
• Synthetic Biology and Bioengineering
• Advance Articles
• Breakthrough Articles
• Special Collections
• Scope and Criteria for Consideration
• Author Guidelines
• Data Deposition Policy
• Database Issue Guidelines
• Web Server Issue Guidelines
• Submission Site
• About Nucleic Acids Research
• Editors & Editorial Board
• Information of Referees
• Self-Archiving Policy
• Dispatch Dates
• Advertising and Corporate Services
• Journals Career Network
• Journals on Oxford Academic
• Books on Oxford Academic

Article Contents

Introduction, materials and methods, database content and usage, discussion and conclusion, supplementary data, acknowledgements.

• < Previous

DDInter: an online drug–drug interaction database towards improving clinical decision-making and patient safety

The authors wish it to be known that, in their opinion, the first three authors should be regarded as joint First Authors.

• Article contents
• Figures & tables
• Supplementary Data

Guoli Xiong, Zhijiang Yang, Jiacai Yi, Ningning Wang, Lei Wang, Huimin Zhu, Chengkun Wu, Aiping Lu, Xiang Chen, Shao Liu, Tingjun Hou, Dongsheng Cao, DDInter: an online drug–drug interaction database towards improving clinical decision-making and patient safety, Nucleic Acids Research , Volume 50, Issue D1, 7 January 2022, Pages D1200–D1207, https://doi.org/10.1093/nar/gkab880

• Permissions Icon Permissions

Drug-drug interaction (DDI) can trigger many adverse effects in patients and has emerged as a threat to medicine and public health. Despite the continuous information accumulation of clinically significant DDIs, there are few open-access knowledge systems dedicated to the curation of DDI associations. To facilitate the clinicians to screen for dangerous drug combinations and improve health systems, we present DDInter, a curated DDI database with comprehensive data, practical medication guidance, intuitive function interface, and powerful visualization to the scientific community. Currently, DDInter contains about 0.24M DDI associations connecting 1833 approved drugs (1972 entities). Each drug is annotated with basic chemical and pharmacological information and its interaction network. For DDI associations, abundant and professional annotations are provided, including severity, mechanism description, strategies for managing potential side effects, alternative medications, etc. The drug entities and interaction entities are efficiently cross-linked. In addition to basic query and browsing, the prescription checking function is developed to facilitate clinicians to decide whether drugs combinations can be used safely. It can also be used for informatics-based DDI investigation and evaluation of other prediction frameworks. We hope that DDInter will prove useful in improving clinical decision-making and patient safety. DDInter is freely available, without registration, at http://ddinter.scbdd.com/ .

DDInter provides detailed annotations of each DDI association and enables users to conduct data query, data retrieval and prescription checking.

Drug–drug interaction (DDI) is one of the most important concerns in clinical rational administration and post-marketing pharmacovigilance ( 1 , 2 ). When taking two or more drugs at the same time or in succession, the activity of one drug may be alerted significantly due to the presence of other drugs, which is described as DDI. DDIs can be roughly classified into two main types: pharmacokinetic (PK) and pharmacodynamic (PD). PK interactions occur when one drug alters the absorption, distribution, metabolism and/or excretion (ADME) of another drug ( 3 , 4 ), while PD interactions occur when one drug alters the pharmacological effects of another drug without affecting its pharmacokinetics ( 5 ).

With the continuous increase of new drugs, prescriptions containing multiple drugs have been common treatment options ( 6 ), especially for patients accompanied with many chronic diseases such as cancer, diabetes, and cardiovascular disorders. This tends to increase the risk of clinically relevant DDIs and create new challenges to therapeutic management. The occurrence of DDIs often triggers unexpected pharmacological effects ( 7 ). In a few cases, DDIs are beneficial and can be exploited as therapeutic strategies for improving efficiency, avoiding toxicity, or minimizing drug resistance, such as the combination of β-lactams and clavulanic acid ( 8 ). However, most DDIs are unpleasant and detrimental, which can make patients exposed to the risks of side effects and toxicity and even deteriorate their physical conditions ( 9 ). It has been reported that DDIs could be associated with up to 30% of all the reported adverse drug events, leading to increased hospitalizations and emergency department visits ( 10 ). A statistical analysis has revealed that ∼15% of older adults taking multiple medications were at risk for potential major DDIs ( 11 ). To minimize potential injury caused by unfavorable DDIs, physicians need to prescribe appropriate drugs to avoid risky drug combinations ( 12 ).

Databases containing rich DDI information are valuable tools for prescribing, which could warn physicians and pharmacists about potential risks timely. Currently, there are some commercial systems dedicated to providing exhaustive and professional DDI information as clinical decision support, such as Drug-Reax software from Micromedex ® Healthcare Series and Lexi-Interact ® software of Lexicomp. However, the subscription fees may prevent their broader access to medical practitioners, especially those in resource-limited areas. Although some free online tools, such as DrugBank ( 13 ), SuperDRUG2 ( 14 ), SuperCYPsPred ( 15 ) and Transformer ( 16 ), have slightly touched upon DDIs, the lack of sufficient mechanisms and management annotations make it difficult to give practical guidance to rational medication. Therefore, there is still an urgent need to develop comprehensive, professional, and open-access DDI databases to improve clinical decision-making and patient safety.

Here, we present DDInter, a comprehensive and practical DDI database, currently containing ∼0.24 M DDI pairs connecting 1833 approved drugs that have been reviewed and curated by a clinical pharmacist team. DDInter provides abundant information for each drug–drug pair, including interaction mechanism, severity level, strategy for managing potential risks, alternatives for drug replacement, literature citations, etc. Concretely, users could easily browse drug entries and interactions, retrieve the basic information and interaction networks of drugs, and carry out prescription checking with the help of the interaction checker module. To help users better understand and explore the searching results, several data visualization tools are embedded to display complex relationships dynamically. The web interface of DDInter is user-friendly and the entities are efficiently cross-linked. DDInter aims to provide a professional DDI knowledge base for the broad community. Physicians and pharmacists could gain practical guidance for dosage adjustment, drug replacement, as well as risk judgment and management. Meanwhile, data scientists could employ the database as the resource for the detection of potential DDIs and evaluation of other prediction tools.

Data collection and curation

DDI information is largely dispersed in scientific literature and labels of pharmaceutical products. The PubMed database was systematically searched to identify the studies that reported the effects of one drug upon another (as of March 2021). Literature pertaining to drug-food or drug–gene interactions and other irrelevant references were ruled out. Then, all the medication guides that had been approved by the US FDA for avoiding serious adverse events were downloaded from Drugs@FDA, with duplicates excluded. Finally, a total of 9460 scientific literature and the labels of pharmaceutical products were employed as the document library for interaction annotation. Most of these documents focused on specific drug classes that were inclined to interfere the activity of co-administered drugs, such as Ca-channel blockers and NSAIDs. About forty thousand DDIs were downloaded from the remarkable article published by a team from Stanford University in Sci Transl Med ( 17 ). These DDIs were collected from authoritative resources and had been confirmed significant after statistical correction of uncharacterized bias. Other DDIs were extracted from documents through semi-automated text-mining ( 18 ). For each drug pair, we recorded the names, the descriptions of mechanisms, and the strategies for managing potential risks and strictly followed the document contents without any further category-based deduction that had been used in other resources. When chemical or pharmacological classes were adopted to refer to a series of drugs that could exert influence on the paired drugs, we extracted the class-drug pairs and then translated them into multiple DDIs based on the drug lists reported in the documents. Certainly, it is common for some DDIs to be confronted with information missing although we have endeavored to collect more data. We used the drug names to query for standard names (generic names) and synonymies (chemical names, brand names, etc.) in DrugBank and Drugs@FDA. The standardization of drug names and merging of interactions were conducted based on the lexicon of drug synonymies. It should be noted that different administration routes of a drug were regarded as different records because their interaction networks may be distinct, especially when the route of administration of a drug might influence its pharmacological effects. In addition, we extracted the interaction descriptions of all the 236 834 DDIs and made further refinements, leading to 5560 distinct and high-quality mechanism descriptions. This suggested that many DDIs shared the same interactions. For each drug entry, a simple pharmacological summary, anatomical therapeutic chemical (ATC) codes, and some external links were collected to enable a better understanding of the specific drug. The small-molecule drugs in DDInter were also annotated with the basic chemical information, including molecular formula, molecular weight, IUPAC name, InChI, InChI key and canonical SMILES.

It is only the first step to identify the nature and the mode of action of DDIs. When many warnings are given, alert fatigue may undermine the usefulness of the interaction checking ( 19 ). Therefore, the registered pharmacist team from Xiangya hospital intensively reviewed these interactions and classified them into different risk levels (major, moderate, minor, and unknown) as suggested by DRUGDEX ( 20 ) and other similar resources. To achieve the maximum rationality, each of the interaction entries was reviewed by at least two pharmacists and the third person would be engaged in the annotation when conflicting opinions occurred. Generally, major interactions were highly clinically significant and the drug combinations should be strictly avoided; moderate interactions may result in exacerbation of the disease of the patient and/or change in therapy; minor interactions were minimally clinically significant and usually they do not require changes in therapy; unknown interactions meant that the interaction description was unavailable or incomplete. In addition to severity level, each of the interaction entries was also manually annotated with interaction mechanisms, including absorption, distribution, metabolism, excretion, synergy and antagonism. More details of the severity and mechanism annotations have been summarized in the website and Supplementary data.

Alternative medications

To allow for convenience and to facilitate clinical prescription, alternative medications were provided based on the ATC code ( 21 ), a widely accepted drug classification scheme in both academic and clinical practice. It categorized drugs into a hierarchy with five different levels based on their therapeutic, pharmacological, and chemical attributes. In DDInter, the third level of the ATC code was used to derive alternative medications, i.e. pharmacological subgroup. Taking atropine and cyproheptadine as an example, the alternatives of atropine share the same anatomical, therapeutic, and pharmacological subgroups with atropine, but have no interactions with cyproheptadine. For drugs with multiple ATC codes, such as aspirin and metformin, it is hard to distinguish which is responsible for the occurrence of DDIs, and therefore alternatives of all the ATC codes were displayed.

Potential metabolism interactions

Cytochrome P450 (CYP) enzymes are responsible for the metabolism of approximately two-thirds of the clinical drugs and therefore play an important role in many metabolic DDIs ( 22 , 23 ). It is necessary to record the CYPs-mediated metabolic profile as additional information to make better elucidation and understanding of potential metabolic interactions. For each small-molecule drug, the two activity modes (substrate and inhibitor) of five major CYP isoenzymes (1A2, 3A4, 2C9, 2C19 and 2D6) were predicted using ADMETlab 2.0 ( 24 ), an integrated online platform for ADMET property evaluation. According to the original paper, the molecules with prediction probability values higher than 0.7 are more likely to be substrates or inhibitors. Metabolic interactions will take place when both of the interacted drugs are inhibitors or substrates of one particular enzyme, including substrate–substrate interaction, inhibitor–inhibitor interaction and inhibitor–substrate interaction.

Online database implementation

DDinter was built based on the Python web framework of Django 3.0 and Bootstrap 4.3.1. The web interface was developed using HTML5, CSS and JavaScript. All the data was stored and managed using MySQL. For molecule visualization, RDKit ( 25 ) was applied to generate 2D images, and 3Dmol.js ( 26 ) was used to display the 3D structures of drugs. All the online data visualizations, including the relation graph and sunburst chart of drug entries, and the bar chart of potential metabolism interactions, were supported by ECharts 4.0 ( 27 ), an open-sourced JavaScript library for the rapid construction of interactive visualization. The website has been tested thoroughly to ensure the functionality across multiple operating systems and web browsers.

Data summary and analysis

DDInter is a comprehensive database that annotates a total of 236 834 clinically relevant DDI associations connecting 1833 US FDA-approved drugs (1972 drug entities). About 82% of the DDI entities were annotated with mechanism categories and severity risks, and the distributions have been described in Figure 1 . Synergy subclass accounts for the largest proportion reaching 46.53%, followed by the interactions mediated via metabolism actions (17.81%) and antagonism actions (8.01%). The most frequently appeared synergy DDI is QTc-interval prolongation ( 28 ), which is a rhythm disturbance and can be caused by an extensive list of medications, such as antiarrhythmic agents, azole antifungal, antipsychotic and antidepressant agents. The concurrent use of these medications should be avoided unless the benefits overweight the risk. And frequent cardiac monitoring is recommended for high-risk populations. Regarding severity levels, 16.67% of the DDIs fall under the category major, 61.28% under the category moderate, and 4.14% under the category minor. It is easy to observe that moderate DDIs take the largest share, which typically need simple interventions to minimize adverse effects such as dosage adjustment and symptoms monitoring. Although moderate DDIs are not life-threatening, clinicians are bound to handle these co-administrations seriously and make patients informed of possible side effects. Here, the proportion of minor DDIs may be grossly underestimated because minor DDIs that do not have serious influence on the quality of life could be scarcely reported ( 29 ).

The statistics of the DDI associations in DDInter based on mechanisms ( A ) and risk levels ( B ). Associations with unavailable or incomplete description of interaction details were labeled as ‘Unknown’.

To gain insight into the DDI associations in the disease area, the degree (number of DDIs involved) of each drug in the DDInter network was calculated and linked to the corresponding ATC codes. The number of drugs and average degrees per therapeutic area have been summarized in Table 1 . Note that drugs with multiple ATC codes were counted once in all the therapeutic areas. On average, the degree of drugs in our database is ∼240. The high degrees of drugs targeting the nervous system or cardiovascular system are clear, exceeding 400 and 350, respectively. A closer analysis demonstrated that 64% and 56% of the drugs targeting these two systems appear in >400 DDI associations. The ten most connected drugs in DDInter are shown in Table 2 , of which six drugs fall under the nervous system. Interestingly, the drugs related to the cardiovascular system are not in the list although their average degree is second only to the nervous system, which may be explained by the even distribution of degrees within the system. The second connected drug is dexamethasone, which has 16 ATC codes covering multiple therapeutic areas like endocrine, rheumatic, allergic, respiratory, etc. The character of multiple pharmacological activities suggests that dexamethasone can influence various therapeutic targets or signaling pathways, which is an important precipitating cause of DDIs. Besides, phenytoin and warfarin are special, whose high degrees are largely contributed by their narrow therapeutic index ( 30 , 31 ). The non-linear pharmacokinetic character of these drugs makes them sensitive to the effects exerted by other drugs. Collectively, the drugs targeting the cardiovascular and nervous systems have high DDI risks and more caution should be given to patients treated by drugs with multiple pharmacological activities or narrow therapeutic index.

DDI involvement of drugs categorized by the therapeutic area

*The statistic was based on the approved drugs recorded in the DDInter database only. The ATC code of each drug was extracted from DrugBank or DrugCentral, and the drugs that could not be mapped to the ATC codes were labelled ‘Unclassified’. ATC, WHO Anatomical Therapeutic Chemical Classification System.

Information of the most connected ten drugs in the DDInter database

Web design and interface

DDInter offers an effective and user-friendly web interface to make full advantage of the wealthy data, accessible at http://ddinter.scbdd.com/ . The search bar in the home page allows users to query the interaction networks of drugs by entering drug names. The ‘Browse’ module provides an overview of the data content, while the ‘Interaction checker’ is designed for prescribing checking. The ‘Download’ page provides links to download DDIs of different therapeutic areas. Some explanation of the annotations, statistic information, detailed tutorials, and user terms and conditions are readily accessible in the ‘About’ page.

Data browsing

Two types of entries, drugs and interactions, could be browsed. All the drugs entities are assigned with unique identifiers and shown in the form of molecular structures. The interactive filter located on the left side of the page allows users to explore a subset of the original data, such as small-molecule drugs, biotech drugs, and drugs targeting different therapeutic areas. Clicking on a specific drug ID will open the drug information page that displays affluent interaction contents as well as chemical and pharmacological descriptions. For interaction browsing, severity level and mechanism filters are both provided to guide users to find the entries of interest. These interaction entities are shown in the tabular format with basic information. When clicking on the ‘view’ button of specific interaction, a list of drug–drug pairs corresponding to this entry would be accessed, where users could obtain concrete interaction content from the provided hyperlinks. By default, drug entities and interaction entities are presented in order of identifier and risk level, respectively.

Data retrieval

The quick search of the DDI information of drugs is available in the middle of the homepage. The text-based search enables users to enter the generic name or synonyms of a drug. Clicking on the search icon will jump to a new page where all the DDI associations of this drug are summarized. If the typed terms could match with multiple drug entities in the database, a list of suitable suggestions will be provided, with hyperlinks to the corresponding full pages. The queried drug information page consists of two major parts: basic chemical and pharmacological annotations, and drug interaction network. As shown in Figure 2A , the ‘Basic Information’ field describes drug structures and other characteristics, including DDInter identifier, drug type, molecular formula and weight, pharmacological summary, ATC codes, useful external links, and some chemical representations. To facilitate the understanding of the DDI associations, a relation graph and a sunburst chart are provided to demonstrate the distributions of all the involved DDIs intuitively (Figure 2B ). Both mechanism-based and risk-based distribution schemes are provided and the conversion is achieved through the switch on the top right corner of the graph. In the DDI relation graph, the secondary nodes represent action mechanisms or risk levels, and interacted drugs are linked to these nodes closely. Users can hover over a connection line to learn the specific interaction information and click on the legend entities to filter out those nodes of less interest. The sunburst chart depicts different hierarchies of DDI associations, where each section can be further expanded by simple clicking. These graphs can be downloaded from the website in PNG format. Below the distribution graphs, all the DDI entries involved are presented in a table (Figure 2C ), containing severity level, identifiers and names of the associated drugs, mechanism of action, and hyperlinks to pages of drug–drug pairs. Specific DDI information can be accessed from these hyperlinks. In addition, a filtering tool is provided to help users to refine the results.

The web interface of DDInter. ( A ) Basic chemical and pharmacological information of the queried drug (caffeine, DDInter263). ( B ) Visualization of DDI network. The relation graph displays the distributions of all the involved DDIs and the sunburst chart shows different hierarchies of DDI associations. ( C ) A list of drugs interacting with the query drug. ( D ) The page of specific drug–drug pair shows extensive information of interactions. ( E ) The interaction checker module helps physicians screen for risks in prescriptions.

Interaction information of drug–drug pairs is mainly displayed in the tabular format in the DDI page (Figure 2D ). The ‘Interaction’ field shows expanded descriptions mechanisms, including pharmacologic conflicts, synergetic toxicity, metabolic enzymes competition, etc. The ‘Management’ field displays the strategies for managing potential side effects, including avoiding combinations, monitoring potential toxicity, adjusting the dosage of drugs, changing time of administration, etc. These annotations are valuable for clinical decision-making, from which physicians could judge whether the DDI risks are tolerated and receive guidance on dose adjustment and therapeutic monitor. If available, the source from which the information is extracted is also provided to allow users to trace back to the original documents. Clinically, replacing a drug by another drug with similar efficacy but lower interaction risks is frequently used in DDI management. Therefore, DDInter provides the alternatives of each drug in DDI associations based on the third level (pharmacological subgroup) of the ATC codes. The metabolic profiles targeting the most important five CYP isoenzymes are presented in a bar chart at the bottom of the page, which can be used to investigate the potential metabolic interactions that have been not detected or summarized. Prediction values higher than 0.7 mean that corresponding drugs are inclined to be substrates or inhibitors of specific enzymes and lead to the occurrence of DDI events. Users can browse the concrete predicted values, switch the chart types and scaling, and download the full chart by using the operation broad in the upper-right corner.

Interaction checker

Prescribing multiple drugs to treat the disorders of patients is common in clinical practice. To facilitate the screening of potential DDIs in prescriptions and improve patient safety, the ‘interaction checker’ module was designed and developed, accessible from the top navigation bar. This module is based on the complete DDI database and allows for the checking of no more than five drugs at once. The detected DDI associations are shown in separate report cards, with information on risk levels, interaction mechanisms, and management (Figure 2E ). A relation chart is provided to visualize the DDI network of the chosen drugs. Simple counting statistics of DDI risks are carried out and the risk status of each DDI association is color-coded. If users are interested in a specific DDI entry, they can click on the ‘see more details’ button to jump to the pages of drug–drug pairs where complete information is summarized.

Downloads and updates

All the DDI associations can be downloaded from the website without login or registration. To facilitate the physicians to focus the DDIs of specific therapeutic areas, the dataset was split into multiple sub-datasets according to the ATC codes. Meanwhile, we hope these datasets could help bioinformaticians to detect potential DDIs and to evaluate other DDI prediction methods. In the future, we will continue to maintain and update the database. The data in DDInter will be updated every 6 months based on the evolving scientific literature and we will integrate more reliable data from other resources to present the most comprehensive landscape of knowledge of clinical-related DDIs.

The understanding and management of DDI events is a real challenge in clinical pharmacy. Some clinical groups and researchers have proposed some strategies to address this confusing area. For instance, the American Geriatrics Society (AGS) has published the Beers Criteria ( 32 ), containing lists of potentially inappropriate medications that should be avoided in elderly people. Pharmacists have also summarized some characters of drugs with high DDI risks, including narrow therapeutic indexes, inhibitors or inducers of CYP450 enzymes, treatment for chronic diseases, etc. ( 33 ). Some data scientists proposed various methods for drug–drug-interaction predictions ( 34 , 35 ). However, these empirical rules cover only a small part of DDI associations and more clinical-related DDIs reported in scientific documents have not been summarized and provided to physicians. And the reliability of predicted DDIs may be influenced by the accuracy of methods. Hence, it is urgent to develop a DDI knowledge system to screen for potential DDI risks proactively. In this work, we proposed DDInter to help physicians and pharmacists detect inappropriate medications and manage clinical outcomes. The manually curated platform contains about 0.24M DDI associations connecting 1833 approved drugs by bringing together the DDI information scattering in literature and product labels. It supplies detailed and professional information of each DDI association, including severity level, mechanism description, management of concurrence, alternative medications, etc. Additionally, it integrates various functions including data browsing, retrieval, and interaction checker to support clinical decision-making. Some data visualization tools are embedded to help users to understand and explore the searching results. The comparison of DDInter and other excellent public medicine resources have been summarized in Supplementary Table S1 for interested readers. To our knowledge, DDInter is the only open-access database that provides professional DDI information to support clinical medication.

Nevertheless, several limitations in the DDInter framework should be elaborated. First, it lacks a ‘gold standard’ to define the clinical significance of DDIs. Since the practical experience of pharmacists is subjective and biased, the definition of severity levels may not be perfectly suitable for all the circumstances worldwide. It would be necessary for users to learn about the annotation standards of our database before using it in clinical practices. The second is data incompleteness. Although we have endeavored to collect information from multiple resources, it is impossible to claim that all the DDIs identified are included. Moreover, certain DDI associations may be concealed by complex disease status and have never been reported yet. Thirdly, this work merely focuses on interactions between two drugs, while three or more drug interactions are not included. In the future, we plan to integrate more data from other resources into DDInter. In summary, the management of DDI events is a complex and important issue and multidisciplinary collaborations are required to address the challenges of DDIs. We hope that our database will prove useful in improving clinical decision-making and patient safety.

Supplementary Data are available at NAR Online.

We would like to thank the pharmacist team from Xiangya hospital for their professional technical support in the data curation and valuable feedback about clinical demands. The studies meet with the approval of the university's review board.

National Science Foundation of China [2217030451, 21575128, 81773632]; Changsha Municipal Natural Science Foundation [kq2014144]; Changsha Science and Technology Bureau project [kq2001034]; Zhejiang Provincial Natural Science Foundation of China [LZ19H300001]; Key R&D Program of Zhejiang Province [2020C03010]; Leading Talent of ‘Ten Thousand Plan’-National High-Level Talents Special Support Plan; HKBU Strategic Development Fund project [SDF19-0402-P02]; Key Research Project of Ningxia Hui Autonomous Region in 2021 (Major Project) [2021BEG01001]. Funding for open access charge: HKBU Strategic Development Fund project [SDF19-0402-P02].

Conflict of interest statement . None declared.

Control, C.f.D. National Hospital Ambulatory Medical Care Survey 2009, Centers for Disease Control . National Center for Health Statistics . 2010 ;

Google Scholar

Google Preview

Wienkers L.C. , Heath T.G. Predicting in vivo drug interactions from in vitro drug discovery data . Nat. Rev. Drug Discov. 2005 ; 4 : 825 – 833 .

Nebert D.W. , Russell D.W. Clinical importance of the cytochromes P450 . Lancet . 2002 ; 360 : 1155 – 1162 .

Nigam S.K. What do drug transporters really do? . Nat. Rev. Drug Discov. 2015 ; 14 : 29 – 44 .

Niu J. , Straubinger R.M. , Mager D.E. Pharmacodynamic drug-drug interactions . Clin. Pharmacol. Ther. 2019 ; 105 : 1395 – 1406 .

Onder G. , Marengoni A. Polypharmacy . JAMA . 2017 ; 318 : 1728 .

Ryu J.Y. , Kim H.U. , Lee S.Y. Deep learning improves prediction of drug–drug and drug-food interactions . Proc. Natl. Acad. Sci. U.S.A. 2018 ; 115 : E4304 – E4311 .

Guieu B. , Jourdan J.P. , Dreneau A. , Willand N. , Rochais C. , Dallemagne P. Desirable drug–drug interactions or when a matter of concern becomes a renewed therapeutic strategy . Drug Discov. Today . 2021 ; 26 : 315 – 328 .

Vilar S. , Friedman C. , Hripcsak G. Detection of drug–drug interactions through data mining studies using clinical sources, scientific literature and social media . Brief. Bioinform. 2018 ; 19 : 863 – 877 .

Pirmohamed M. , Orme M. Drug interactions of clinical importance . Davies's Textbook of Adverse Drug Reactions . 1998 ; 888 – 912 .

Qato D.M. , Wilder J. , Schumm L.P. , Gillet V. , Alexander G.C. Changes in prescription and over-the-counter medication and dietary supplement use among older adults in the United States, 2005 vs 2011 . JAMA Intern. Med. 2016 ; 176 : 473 – 482 .

Qiu Y. , Zhang Y. , Deng Y. , Liu S. , Zhang W. A comprehensive review of computational methods for drug-drug interaction detection . IEEE/ACM Trans. Comput. Biol. Bioinf. 2021 ; https://doi.org/10.1109/TCBB.2021.3081268 .

Wishart D.S. , Feunang Y.D. , Guo A.C. , Lo E.J. , Marcu A. , Grant J.R. , Sajed T. , Johnson D. , Li C. , Sayeeda Z. et al. . DrugBank 5.0: a major update to the DrugBank database for 2018 . Nucleic Acids Res. 2018 ; 46 : D1074 – D1082 .

Siramshetty V.B. , Eckert O.A. , Gohlke B.O. , Goede A. , Chen Q. , Devarakonda P. , Preissner S. , Preissner R. SuperDRUG2: a one stop resource for approved/marketed drugs . Nucleic Acids Res. 2018 ; 46 : D1137 – D1143 .

Banerjee P. , Dunkel M. , Kemmler E. , Preissner R. SuperCYPsPred-a web server for the prediction of cytochrome activity . Nucleic Acids Res. 2020 ; 48 : W580 – W585 .

Hoffmann M.F. , Preissner S.C. , Nickel J. , Dunkel M. , Preissner R. , Preissner S. The transformer database: biotransformation of xenobiotics . Nucleic Acids Res. 2014 ; 42 : D1113 – D1117 .

Tatonetti N.P. , Ye P.P. , Daneshjou R. , Altman R.B. Data-driven prediction of drug effects and interactions . Sci. Transl. Med. 2012 ; 4 : 125ra131 .

Wang W. , Yang X. , Yang C. , Guo X. , Zhang X. , Wu C. Dependency-based long short term memory network for drug–drug interaction extraction . BMC Bioinformatics . 2017 ; 18 : 578 .

Greenberg M. , Ridgely M.S. Clinical decision support and malpractice risk . JAMA . 2011 ; 306 : 90 – 91 .

Klasco R. DRUGDEX® system (electronic version). Thomson MICROMEDEX . Greenwood Village, Colorado, USA . 2009 ; 1 : 06 – 09 .

MacDonald K. , Potvin K. Interprovincial variation in access to publicly funded pharmaceuticals: a review based on the WHO anatomical therapeutic chemical classification system . Can. Pharm. J. 2004 ; 137 : 29 – 34 .

Sikka R. , Magauran B. , Ulrich A. , Shannon M. Bench to bedside: pharmacogenomics, adverse drug interactions, and the cytochrome P450 system . Acad. Emerg. Med. 2005 ; 12 : 1227 – 1235 .

Lynch T. , Price A. The effect of cytochrome P450 metabolism on drug response, interactions, and adverse effects . Am. Fam. Physician . 2007 ; 76 : 391 – 396 .

Xiong G. , Wu Z. , Yi J. , Fu L. , Yang Z. , Hsieh C. , Yin M. , Zeng X. , Wu C. , Lu A. et al. . ADMETlab 2.0: an integrated online platform for accurate and comprehensive predictions of ADMET properties . Nucleic Acids Res. 2021 ; 49 : W5 – W14 .

Landrum G. RDKit: open-source cheminformatics . 2016 ;

Rego N. , Koes D 3Dmol.js: molecular visualization with WebGL . Bioinformatics . 2015 ; 31 : 1322 – 1324 .

Li D. , Mei H. , Shen Y. , Su S. , Zhang W. , Wang J. , Zu M. , Chen W. ECharts: a declarative framework for rapid construction of web-based visualization . Visual Informatics . 2018 ; 2 : 136 – 146 .

Sheikh-Taha M. , Asmar M. Polypharmacy and severe potential drug–drug interactions among older adults with cardiovascular disease in the United States . BMC Geriatr . 2021 ; 21 : 233 .

Triplitt C. Drug interactions of medications commonly used in diabetes . Diabetes Spectrum . 2006 ; 19 : 202 – 211 .

Leite P.M. , Martins M.A.P. , Castilho R.O. Review on mechanisms and interactions in concomitant use of herbs and warfarin therapy . Biomed. Pharmacother. 2016 ; 83 : 14 – 21 .

Walia K.S. , Khan E.A. , Ko D.H. , Raza S.S. , Khan Y.N. Side effects of antiepileptics–a review . Pain Pract . 2004 ; 4 : 194 – 203 .

By the American Geriatrics Society Beers Criteria Update Expert, P. American Geriatrics Society 2015 updated beers criteria for potentially inappropriate medication use in older adults . J. Am. Geriatr. Soc. 2015 ; 63 : 2227 – 2246 .

Mallet L. , Spinewine A. , Huang A. The challenge of managing drug interactions in elderly people . Lancet . 2007 ; 370 : 185 – 191 .

Zheng Y. , Peng H. , Zhang X. , Zhao Z. , Gao X. , Li J. DDI-PULearn: a positive-unlabeled learning method for large-scale prediction of drug–drug interactions . BMC Bioinformatics . 2019 ; 20 : 661 .

Nyamabo A.K. , Yu H. , Shi J.Y. SSI-DDI: substructure-substructure interactions for drug–drug interaction prediction . Brief. Bioinform. 2021 ; bbab133 .

Author notes

Supplementary data, email alerts, citing articles via.

• Editorial Board

Affiliations

• Online ISSN 1362-4962
• Print ISSN 0305-1048
• Copyright © 2024 Oxford University Press
• About Oxford Academic
• Publish journals with us
• University press partners
• What we publish
• New features  
• Open access
• Institutional account management
• Rights and permissions
• Get help with access
• Accessibility
• Advertising
• Media enquiries
• Oxford University Press
• Oxford Languages
• University of Oxford

Oxford University Press is a department of the University of Oxford. It furthers the University's objective of excellence in research, scholarship, and education by publishing worldwide

• Copyright © 2024 Oxford University Press
• Cookie settings
• Cookie policy
• Privacy policy
• Legal notice

This Feature Is Available To Subscribers Only

Sign In or Create an Account

This PDF is available to Subscribers Only

For full access to this pdf, sign in to an existing account, or purchase an annual subscription.

ORIGINAL RESEARCH article

Adverse drug reactions and correlations with drug–drug interactions: a retrospective study of reports from 2011 to 2020.

• 1 Department of Pharmacy, Jinshan Hospital, Fudan University, Shanghai, China
• 2 Department of Nursing, Jinshan Hospital, Fudan University, Shanghai, China

Introduction: Adverse drug reactions (ADRs) represent a public health problem worldwide that deserves attention due to the impact on mortality, morbidity, and healthcare costs. Drug–drug interactions (DDIs) are an important contributor to ADRs. Most of the studies focused only on potential DDIs (pDDIs), while the detailed data are limited regarding the ADRs associated with actual DDIs.

Methods: This retrospective study evaluated ADRs reported between 2011 and 2020 in a tertiary hospital. The causality and severity of ADRs were evaluated through the Naranjo Algorithm and Hartwig’s scale, respectively. Preventability classification was based on the modified Schoumock and Thornton scale. For ADRs with at least two suspected drugs, pDDIs were identified according to the Lexi-Interact. We further checked whether the ADR description in the reports corresponded to the clinical consequences of the pDDIs.

Results: A total of 1,803 ADRs were reported, of which 36.77% ADRs were classified as mild, 43.26% as moderate, and 19.97% as severe. The assessment of causality showed that the distributions of definite, probable, and possible categories were 0.33%, 58.68%, and 40.99%, respectively. A total of 53.97% of ADRs were identified as preventable ADRs, while 46.03% were recognized as unpreventable. The severity of ADRs was significantly correlated with age, the number of suspected drugs and preventability. Antimicrobial agents were the most common implicated pharmacological group, and the most frequently affected system was the gastrointestinal system. Considering individual drugs, aspirin was the most frequently reported drug. Among 573 ADRs with at least two suspected drugs, 105 ADRs were caused by actual DDIs, of which only 59 and 6 ADRs were caused by actual DDIs in category D and X, respectively. The most frequent drugs involved in actual DDIs of category D were aspirin and heparin, with the majority of ADRs being gastrointestinal bleeding.

Conclusion: This study analyzed the pattern of ADRs in detail and obtained clinical evidence about ADRs associated with actual DDIs. These findings may be useful to compare patterns between different centers and to design preventive strategies for ADRs. Continuous education and training should be provided for physicians regarding the knowledge and recognition of ADRs associated with DDIs.

Introduction

According to the World Health Organization (WHO), an adverse drug reaction (ADR) is an unintended and noxious response that is detected in patients after the use of drugs for the prophylaxis, diagnosis or treatment of a disease at doses normally used ( Edwards and Aronson, 2000 ). ADRs, as a major threat in the healthcare system, contribute significantly to mortality, morbidity, extended hospital stays, and increased healthcare costs ( Khan, 2013 ; Angamo et al., 2016 ). A meta-analysis showed that the percentage of ADR-induced admissions in patients over 60 years old was accurately estimated to be 8.7% ( Oscanoa et al., 2017 ). To minimize the consequences of ADRs, it is necessary to study ADRs in terms of their early identification and prevention and to motivate healthcare professionals to report ADRs ( Arulappen et al., 2018 ).

According to a WHO report, 60% of ADRs are preventable ( Lau et al., 2003 ). Drug–drug interactions (DDIs) are an important cause of preventable ADRs. The increasing number of patients with multimorbidity and the growing complexity of therapeutic agents have led to widespread polypharmacy, which could result in the rising numbers of potential DDIs (pDDIs), especially in elderly individuals ( Obreli-Neto et al., 2012a ; Scondotto et al., 2018 ). Although there are several databases available that could be used to evaluate pDDIs, the clinical relevance and actual clinical importance of majority pDDIs remain insufficiently characterized and underestimated ( Roblek et al., 2015 ). Actual DDIs are identified on the basis of clinical evidence, such as laboratory test results or symptoms, consequently, the frequency of actual DDIs is much lower than that of pDDIs ( Magro et al., 2012 ; Zheng et al., 2018 ). Over the past years, a substantial number of articles have been published about ADRs due to DDIs ( Leone et al., 2010 ; Obreli-Neto et al., 2012a ; Obreli Neto et al., 2012b ; Kovacevic et al., 2019 ; Letinier et al., 2020 ; Magro et al., 2020 ). A 6-year retrospective study in Bengbu in China showed that among the ADRs reported between nervous system drugs in hospitalized patients, 12.14% of the ADRs were associated with potential and actual DDIs, and actual DDIs were present in 6.21% of all ADRs ( Shi et al., 2014 ). However, the incidence of ADRs resulting from DDIs could not be accurately estimated primarily because of differences in study designs and populations ( Mirosevic Skvrce et al., 2011 ).

In this context, the present study aimed to describe the distribution of ADRs, assess causality, preventability and severity of ADRs, and determine factors involved in the severity of ADRs in a tertiary hospital between 2011 and 2020. Additionally, we described and analyzed the most frequent drugs suspected to cause ADRs and the organ system classes affected by ADRs. Furthermore, we evaluated the pDDIs among the ADRs with more than one suspected drug, estimated the incidence of ADRs due to actual DDIs and characterized ADRs caused by actual DDIs.

Materials and methods

Data collection.

In this retrospective single-center study, all the ADRs was collected from the National ADR Monitoring system in Jinshan Hospital of Fudan University, between 01 January 2011 and 31 December 2020. Jinshan Hospital is a tertiary general hospital with a 700-bed capacity in the Jinshan district of Shanghai. In 2020, there were 28,533 hospital admissions, and 1.28 million outpatient and emergency department visits. ADR reports were filled out according to a specific ADR report format and submitted in paper based or electronic way by healthcare professionals, including physicians, pharmacists, and nurses.

Once received, the reported ADRs were reviewed and evaluated by ADR surveillance unit of the pharmacy department. Only the reported ADRs followed the WHO definition ( Edwards and Aronson, 2000 ) and without any uncertainty or mistakes were accepted after exclusion of duplicates and uploaded to ADR Monitoring system. A series of exclusion criteria were applied to ensure a robust data set for analysis. Exclusion criteria included the following: 1) ADRs with doubtful causality with Naranjo’s algorithm ( Naranjo et al., 1981 ). 2) ADR forms with insufficient information 3) ADRs symptoms similar to the original disease.

The demographic and other information relevant to ADRs were documented, including gender, age, diagnosis, admission department, suspected drugs, concomitant medications, drug details, organ system involved in the ADR, the management and outcome of the ADRs, and the type of reporter. One report could describe one or more ADRs. The incriminated drugs were classified by pharmacological group according to the WHO Anatomical Therapeutic Chemical Classification (ATC). The involved system organ classes were determined according to WHO Adverse Reaction Terminologies (WHO-ART). Two investigators cross checked the data for accuracy. Flowchart depicting the study process was shown in Figure 1 .

FIGURE 1 . Flowchart depicting the study process.

Causality, preventability, and severity assessment

Each ADR was further evaluated for various parameters, such as causality, severity and preventability, using previously validated and recognized approaches. The assessment of causality was performed using the Naranjo Algorithm, which consists of 10 individually scored criteria. ADRs were categorized as possible ADRs (1–4), probable ADRs (5–8) or definite ADRs (≥9) based on the total score ( Naranjo et al., 1981 ). Severity classification was based on Hartwig’s scale, which showed the criteria and matched levels used for ADR severity assessment. ADRs were considered as severe if they resulted in one of the following outcomes: the requirement for intensive medical care, permanent harm to the patient, or the death of the patient ( Hartwig et al., 1992 ). The preventability of ADRs was assessed by the modified Schoumock and Thornton scale and classified into definitely preventable, probably preventable and not preventable reactions ( Schumock and Thornton, 1992 ). In our study, both definitely and probably preventable ADRs were considered as one category of preventable reactions.

Evaluation of potential drug–drug interactions

For ADRs caused by two or more suspected drugs, pDDIs were identified by the software Lexi-Interact in UpToDate. The evaluation results of pDDIs were classified into five levels of risk as no known interaction (A), no action needed (B), monitor therapy (C), consider therapy modification (D), and avoid combination (X). We further verified whether the clinical consequences of pDDIs corresponded to the description of the ADR in the report, and if consistent, the pDDI was considered the actual DDI. Two clinical pharmacists independently assessed the probability, severity and preventability of ADRs as well as the consistency between ADRs and pDDIs. Any discrepancies were resolved by discussion.

Statistical analysis

Descriptive statistics were applied to describe the population as well as the clinical characteristics of ADRs and pDDIs. The categorical data were presented as numbers and proportions. Sankey diagrams of severity in preventable and unpreventable ADRs were plotted with the R package alluvial. The Mann–Whitney U test was used to evaluate the correlation between gender and the severity of ADRs. Spearman’s rank tests were performed to determine the association of age, the number of suspected drugs and the category of preventability with the severity of ADRs. The Kruskal–Wallis H test was performed to evaluate the correlation between the route of administration and the severity of ADRs. Statistical analysis was performed using IBM SPSS Statistics version 25. A p -value < 0.05 was considered statistically significant.

Department and reporter distribution of adverse drug reactions

From January 2011 to December 2020, a total of 1,803 ADRs were reported by healthcare professionals in our hospital, although the number of ADRs reported was relatively small between 2011 and 2013. During this 10-year period, pharmacists contributed 55.69% of all ADR reports, followed by physicians (43.98%). The frequency of ADRs reported by nurses was low, accounting for only 0.33%. The annual number of reports was no more than 221 during 2011–2018, however, this number subsequently increased significantly over the next 2 years, reaching 388 in 2020 ( Figure 2 ). A small proportion of ADRs were reported by pharmacists between 2011 and 2013, however, since 2014, more than half of ADR reports have been submitted by pharmacists. Detailed data by the year and distribution of reporters were shown in Figure 2 . In our study, the highest percentage of ADRs was collected from the gastroenterology department (26.8%), followed by the departments of emergency and critical care medicine (11.4%), cardiology department (7.9%), and neurology department (7.8%) ( Figure 3 ). The proportions of ADRs collected from clinical departments were presented in Figure 3 .

FIGURE 2 . The total number of adverse drug reaction (ADR) reports and the distribution of reporters from different occupations by year during 2011–2020.

FIGURE 3 . The percentage of adverse drug reactions (ADRs) from different clinical departments.

Causality, preventability, and severity assessment of adverse drug reactions

ADRs were further analyzed for causality, preventability and severity, as shown in Table 1 . The assessment of causality according to the Naranjo Algorithm showed that the numbers of definite, probable and possible ADRs were 6 (0.33%), 1,058 (58.68%), and 739 (40.99%), respectively. According to Hartwig’s Severity Assessment Scale, 663 (36.77%) ADRs were classified as mild, 780 (43.26%) as moderate, and 360 (19.97%) as severe. The evaluation of the preventability of ADRs using the modified Schumock and Thornton criteria revealed that 973 (53.97%) ADRs were identified as preventable ADRs, including 93 as definitely preventable and 880 as probably preventable, while 830 (46.03%) ADRs were recognized as unpreventable. Symptomatic or specific treatment was given for 1,045 (57.96%) ADRs. According to the records of ADR reports, the majority of ADRs (81.09%) had improved, 238 (13.20%) patients had recovered from their ADRs, and 103 (5.71%) ADRs continued or their status was unclear. Suspected drugs were withdrawn in 1,700 (94.29%) ADR reports, but an altered dose or no change in therapy was observed in 103 (5.71%) reports. The visual design follows the principle of the Sankey diagram, which links the ADR characteristics by lines and signifies the quantities via line width, stratified by preventability ( Figure 4 ).

TABLE 1 . Assessment and pattern of adverse drug reactions.

FIGURE 4 . Sankey diagram of severity in preventable and unpreventable ADRs. (A,B) The causality assessment, fate of the suspected drug, treatment given and outcome of ADRs matched with ADR severity in preventable ADRs. (C,D) The causality assessment, fate of the suspected drug, treatment given and outcome of ADRs matched with ADR severity in unpreventable ADRs. The causality assessment of ADRs (A1 and A2), A1: Definite/Probable, A2: Possible. ADR severity (B1, B2, and B3), B1: Mild, B2: Moderate, B3: Severe. Fate of the suspected drug (C1 and C2), C1: Drug withdrawn, C2: Dose altered/No change. Treatment given (D1 and D2), D1: Treatment given, D2: No treatment. Outcome of ADRs (E1, E2, and E3), E1: Recovered, E2: Improved, E3: Continuing/Unclear.

Characteristics of adverse drug reactions according to severity assessment

The characteristics of ADRs according to severity assessment were shown in Table 2 . A total of 1,803 ADRs were identified among 1,779 patients. Multiple ADRs in the same patient may be identified with different severity scale, so Table 2 depicted the distribution of mild, moderate, and severe reactions between different gender and age based on ADRs rather than patients. Concerning patient gender and ADRs, 46.87% males and 53.13% females experienced ADRs over the past 10 years. The proportion of mild ADRs was higher in females (60.33%) than in males (39.67%), however, the ratio of males to females was approximately 1:1 among those experiencing moderate and severe ADRs. The Mann–Whitney U test revealed significant differences in the mild, moderate, and severe ADR distributions between the males and females.

TABLE 2 . Comparison of mild, moderate, and severe reactions.

As shown in Table 2 , the percentage of ADRs was highest among elderly individuals over 65 years of age (42.87%), followed by the 41–64-year (38.38%) and 18–40-year (14.03%) age groups. The minimum number of ADRs was observed in the age group under 18 years (4.71%). More than half of severe ADRs occurred in elderly individuals over 65 years of age. The majority of ADRs (67.72%) were identified with only one suspected drug, followed by 24.35% with two suspected drugs, and only 143 (7.93%) ADRs were found with ≥3 suspected drugs. According to Hartwig’s scale, 46.03% of ADRs were classified as unpreventable ADRs, 48.81% as probably preventable ADRs and 5.16% as definitely preventable ADRs. The percentage of unpreventable ADRs significantly decreased with ADR severity (mild 63.65% vs. moderate 43.72% vs. severe 18.61%). The statistical results revealed significant positive correlations of ADR severity with age (Spearman’s R = 0.167, p < 0.001), the number of suspected drugs (Spearman’s R = 0.136, p < 0.001) and ADR preventability (Spearman’s R = 0.299, p < 0.001).

The route of administration was classified according to the first suspected drug in the ADR reports. More than half of ADRs were associated with oral medicines regardless of their severity rating. Unexpectedly, the proportion of ADRs associated with intravenous drugs gradually decreased with increasing severity. The Kruskal–Wallis H test was further carried out and showed a significant association between the route of administration and the severity of ADRs ( Table 2 ).

Drugs involved in adverse drug reactions and effects on organ systems

The pharmacological groups implicated in the ADRs were summarized in Table 3 . Systemic antimicrobial agents were the most commonly implicated drugs (22.75%), with 14.53% and 39.32% of their associated ADRs being classified as severe and preventable ADRs, respectively. Cardiovascular agents were the second most frequently reported class of drugs responsible for ADRs (12.41%), followed by medications for the alimentary tract and metabolism (12.06%). Drugs acting on the blood and blood-forming organs represented 11.75% of the reports (32.45% severe and 83.11% preventable ADRs). Drugs acting on the musculoskeletal system were implicated in 10.77% of the ADRs (39.71% severe and 71.48% preventable ADRs). Traditional Chinese medicines were implicated in 10.07% of the reports (18.53% severe ADRs and 54.44% preventable ADRs).

TABLE 3 . Pharmacology groups according to the WHO-ATC code and their pattern in ADRs.

The frequency of commonly prescribed drugs among total and severe ADRs was shown in Table 4 . When individual drugs were considered, aspirin was responsible for a maximum number of both total and severe ADRs, far more than any other drugs. Among the total ADRs, levofloxacin (82) was the second most frequent causative drug, followed by compound pseudoephedrine hydrochloride (65) and clopidogrel (48). In addition to aspirin, the drugs most frequently involved in severe ADRs were clopidogrel (17), levofloxacin (16), compound pseudoephedrine hydrochloride (14), and diclofenac sodium (13).

TABLE 4 . Top 10 incriminated drugs in total and severe ADRs based on frequency.

Upon a review of the outcomes of ADRs, the most frequently affected system was the gastrointestinal system (30.83%), with the clinical symptoms of nausea, vomiting, abdominal pain, diarrhea, abdominal distention, and so on. In addition, the commonly reported reactions were skin and appendage disorders (22.44%) and liver and biliary system disorders (14.19%). A more detailed description was presented in Table 5 .

TABLE 5 . Organs or systems involved in ADRs according to WHO classification.

Adverse drug reactions caused by drug–drug interactions

pDDIs were evaluated in 573 of 1,803 ADR reports (31.78%) involving more than one suspected drug. 156 ADRs were identified with pDDIs of category C, D, and X, of which 100 ADRs were identified with only one pDDI and 56 ADRs with multiple pDDIs. Table 6 showed that 208 pDDIs of category C were identified in 112 ADRs, 74 pDDIs of category D in 58 ADRs, and 11 pDDIs of category X in 10 ADRs. Furthermore, we checked whether the reported ADRs were consistent with the potential clinical consequences of pDDIs. The results showed 105 ADRs were caused by actual DDIs, accounting for 18.32% of the ADR reports with more than one suspected drug. Among them, 59 and 6 ADRs were caused by actual DDIs in the category D and X, respectively.

TABLE 6 . Distribution of the potential drug–drug interactions with category C, D, and X in ADRs.

Tables 7 , 8 summarized the ADRs caused by actual DDIs belonging to category X and D, respectively. Potassium chloride and promethazine were the drug–drug combination most involved in ADRs caused by actual DDIs in category X, with severe and adverse clinical consequences to the gastrointestinal system. The most frequent drugs involved in actual DDIs of category D were aspirin ( n = 34) and heparin ( n = 26), and the great majority of ADRs caused by DDIs were associated with gastrointestinal bleeding. Aspirin/heparin ( n = 10) and heparin/clopidogrel ( n = 10), followed by aspirin/warfarin ( n = 6) and aspirin/ibuprofen ( n = 5), were the drug–drug combinations most involved in ADRs caused by DDIs of category D.

TABLE 7 . ADRs caused by actual drug–drug interactions belonging to category X.

TABLE 8 . ADRs caused by actual drug–drug interactions belonging to category D.

In this study, physicians and pharmacists were the groups that reported the great majority of ADRs, and the frequency of ADRs reported by nursing staff was low, which may be due to their extensive workload in everyday practice, inattention and unawareness toward ADR reporting or worry about legal implications ( Singh et al., 2017 ). The reporter distribution of ADRs varies widely in different studies because of differences in healthcare structures as well as the awareness and motivation of healthcare professionals. The number of ADRs was relatively small, especially for ADRs reported by pharmacists between 2011 and 2013, indicating underreporting in pharmacovigilance. The key to improving ADR reporting rates is adequate pharmacovigilance education and training for healthcare professionals ( Barzaga Arencibia et al., 2012 ).

In the present study, we analyzed the pattern of ADRs based on the causality, severity, and preventability in our hospital, all of which vary among different hospitals due to differences in the population characteristics and hospital specialties. Naranjo’s causality assessment showed that only 0.33% of reports were definite because of limited use of dechallenge and rechallenge processes for ethical reasons as well as the retrospective study design without the ability to assess the ADR completely. The suspected drugs were withdrawn among 94.29% of ADRs, and for the remaining 5.71% of ADRs, the suspected drug doses were altered or rechallenge processes were initiated. In this study, 19.97% of ADRs were classified as severe. Severe ADRs, as major concerns for public health, are a contributing factor of hospitalizations and morbidity ( Rottenkolber et al., 2011 ; Marques et al., 2014 ). The analysis indicated a preventability rate of 53.97% among ADRs, comparable with the results of studies conducted in Romania and Jordan showing that 41% and 44.7% of ADRs were preventable, respectively ( Farcas et al., 2014 ; Al Damen and Basheti, 2019 ). However, the data from a study showed lower preventability for ADRs (12%) compared with our finding ( Dequito et al., 2011 ). As described in previous studies, insufficient monitoring, inappropriate dosing, and DDIs were the most frequent factors involved in ADR preventability ( Farcas et al., 2014 ; Al Damen and Basheti, 2019 ). Incriminated drugs were withdrawn in 94.29% of the reports, which is in line with a previous study in a psychiatric department of a tertiary care teaching hospital in India ( Patel et al., 2015 ). The high proportion of withdrawal may be due to the reporting nature of ADRs that troublesome ADRs are more likely to be detected.

There may be significant difference between male and female regarding the ADR prevalence due to factors such as body mass index, fat composition, hormonal effects, drug susceptibility, or genetic differences in the levels of enzymes ( Haile et al., 2013 ; Rukmangathen et al., 2020 ). However, we demonstrated that females had only slightly higher incidence of ADRs than males in the present study. The frequency of ADRs increased with age, with the highest prevalence of ADRs in elderly individuals over 65 years (42.87%), followed by individuals 41–64 years of age (38.38%), which is in concordance with the findings of a previous study ( Shepherd et al., 2012 ). Older patients are particularly vulnerable to ADRs owing to the multiple-drug regimens used for chronic diseases and physiological changes in this population, such as reduced gastrointestinal motility and gastric blood flow, impaired repair mechanisms, and lower mucosal protection ( Marusic et al., 2014 ). A systematic review of ADRs in elderly individuals revealed that comorbid complexity was positively associated with ADR occurrence ( Alhawassi et al., 2014 ). In the present study, there were statistically significant differences in the incidence of severe ADRs in the different gender and age groups, and polypharmacy increased the proportion of severe ADRs.

Anti-infectives for systemic use were the most common pharmacological group, accounting for 22.75% of total ADRs in our study, which is in line with previous studies ( Haile et al., 2013 ; Marques et al., 2014 ). The excessive use of antibiotics may be responsible for the increased risk of ADRs. Cardiovascular system agents (12.41%) were the second most frequently incriminated pharmacological class of ADRs in our study, among them, 65.20% were preventable ADRs. A systematic review showed that cardiovascular medicines were commonly associated with preventable drug-related admissions ( Howard et al., 2007 ). In another study, cardiovascular agents were identified as the second most frequently responsible drugs linked to preventable ADRs ( Farcas et al., 2014 ).

The system most frequently affected by ADRs in this study was the gastrointestinal system, accounting for 30.83%, probably due to more than half of the suspected drugs being administered orally. This was followed by skin and appendage disorders (22.44%). This observation is consistent with the findings of a prospective observational study of hospitalized pediatric patients, which reported gastrointestinal system disorders (51.56%) and skin and appendage disorders (18.75%) as the most frequent manifestations of ADRs ( Kurian et al., 2016 ).

As DDIs are usually predictable and manageable, ADRs caused by DDIs may be prevented by monitoring the patient closely or replacing the responsible drugs with other medications. To reduce the risk of DDIs and improve patient safety, it is essential that healthcare professionals regularly review the medication regimens, recognize potentially interacting drug pairs, and withdraw unnecessary drugs ( Magro et al., 2020 ). A prospective study showed that the number of patients with pDDIs and actual DDIs decreased by 18% and 43%, respectively, with an intervention based on a computerized clinical decision support system containing information on drug combinations ( Bertsche et al., 2010 ). However, reporters less frequently recognize actual DDIs due to the limited availability of DDI databases or alerting drug-interaction systems ( Mirosevic Skvrce et al., 2011 ). Therefore, it is important to increase the knowledge of pharmacovigilance through the additional education of healthcare providers.

In a previous study, we investigated the prevalence of pDDIs and their association with characteristics in outpatient prescriptions ( Ren et al., 2020 ). However, to assess the clinical impact of DDIs on public health, only ADRs associated with DDIs should be considered. In our study, 105 ADR reports were induced by actual DDIs, accounting for 18.32% of the ADR reports with more than one suspected drug. This percentage was close to the proportion reported by Magro et al. (2020) . According to the online version of DRUGDEX ® system, they verified DDI among serious ADRs containing at least two suspected or concomitant drugs in the National Pharmacovigilance database from Veneto Region, and identified 17.4% ADR reports associated with a DDI. However, the results of another study performed in an Italian spontaneous reporting database showed that regarding patients treated with at least two drugs, 6.5% of ADR reports was associated with a DDI using the DRUGDEX ® system ( Leone et al., 2010 ). Similarly, a prospective cohort study conducted in the primary public health system of the Ourinhos microregion in Brazil revealed that the incidence of DDI-related ADRs was 6% in elderly outpatients using DDI-checker programs (DrugDigest ® , Drugs ® , Micromedex ® , and Medscape ® ) ( Obreli-Neto et al., 2012b ).

In the present study, aspirin and heparin were the drugs most frequently associated with actual DDIs of category D, with symptom of gastrointestinal bleeding. Similarly, a prospective observational study conducted in the cardiology unit of an Indian hospital showed that heparin and aspirin were the most common drugs responsible for DDIs, and bleeding was the most frequent clinical consequence ( Mateti et al., 2011 ). Furthermore, aspirin, which is widely used for the prevention of vascular events, was reported to increase the baseline risk of gastrointestinal bleeding by approximately 60% among older persons aged over 70 years in a randomized controlled trial ( Mahady et al., 2021 ).

Although the study had important findings regarding the pattern of ADRs and the role of actual DDIs in ADRs over the past decade along with a large sample size, several limitations should be taken into consideration. First, as a retrospective study, data were collected from the clinical records of ADRs always with incomplete information, such as information on concomitant drugs, comorbidities, lifestyle, diet, and so on. Prospective studies will be carried out to clarify and reduce this limitation in the future. Second, this study was conducted at a single institution, limiting the generalizability of its findings due to the differences in population characteristics and prescribing patterns. Last, the single source of the DDI screening database used in this study may hinder the identification of DDIs because consistent criteria for DDI identification and assessment are currently lacking.

This study of ADR data collected over 10 years revealed that almost all ADRs were reported by pharmacists and physicians in our hospital, and the severity of ADRs was significantly correlated with age, the number of suspected drugs and preventability. Systemic antimicrobial agents were the most frequently incriminated pharmacological group, and aspirin was responsible for the largest proportion of total and severe ADRs. The gastrointestinal system was the system most frequently affected by ADRs. As observed in this study, aspirin and heparin were the most common drugs in actual DDIs of category D, resulting in gastrointestinal bleeding.

Active pharmacovigilance programs are important to accurately identify and assess ADRs in the clinical setting, further minimize drug-induced harm and improve the quality of patient care. Our findings obtained clinical evidence about ADRs associated with actual DDIs in our hospital. It will be necessary to make clinicians aware of the possibility of DDI-related ADRs and achieve a clear understanding of drug pairs resulting in DDI-related ADRs, in order to guide the prescribing practices and minimize the harms from actual DDIs. Moreover, rigorous prescription and frequent monitoring of drug therapy are essential for reducing the risk of ADRs.

Data availability statement

The raw data supporting the conclusion of this article will be made available by the authors, without undue reservation.

Ethics statement

This study was approved by the Ethics Committee of Jinshan Hospital, Fudan University at Shanghai, China (approval No. JIEC 2022-S29).

Author contributions

HJ, ZF, XL, and NZ contributed to the conception and design of the study. HJ, YL, WR, YL, and XT contributed to the recording and statistical analysis of the data. HJ and XL wrote the first draft of the manuscript. XL and NZ made critical revisions to the manuscript. All authors approved the final version of the manuscript.

This work was supported by the Fourth Training Program for the Outstanding Young Talents, Jinshan Health Commission (#JSYQ201904), Key Construction Project on Clinical Pharmacy of Shanghai (#2019-1229).

Acknowledgments

The authors acknowledge the ADR reporting system of Jinshan Hospital, Fudan University.

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Al Damen, L., and Basheti, I. (2019). Preventability analysis of adverse drug reactions in a Jordanian hospital: A prospective observational study. Int. J. Clin. Pharm. 41 (6), 1599–1610. doi:10.1007/s11096-019-00925-0

PubMed Abstract | CrossRef Full Text | Google Scholar

Alhawassi, T. M., Krass, I., Bajorek, B. V., and Pont, L. G. (2014). A systematic review of the prevalence and risk factors for adverse drug reactions in the elderly in the acute care setting. Clin. Interv. Aging 9, 2079–2086. doi:10.2147/CIA.S71178

Angamo, M. T., Chalmers, L., Curtain, C. M., and Bereznicki, L. R. (2016). Adverse-drug-reaction-related hospitalisations in developed and developing countries: A review of prevalence and contributing factors. Drug Saf. 39 (9), 847–857. doi:10.1007/s40264-016-0444-7

Arulappen, A. L., Danial, M., and Sulaiman, S. A. S. (2018). Evaluation of reported adverse drug reactions in antibiotic usage: A retrospective study from a tertiary care hospital, Malaysia. Front. Pharmacol. 9, 809. doi:10.3389/fphar.2018.00809

Barzaga Arencibia, Z., Lopez Leyva, A., Mejias Pena, Y., Gonzalez Reyes, A. R., Fernandez Manzano, E., Choonara, I., et al. (2012). Pharmacovigilance in children in camaguey province, Cuba. Eur. J. Clin. Pharmacol. 68 (7), 1079–1084. doi:10.1007/s00228-012-1222-9

Bertsche, T., Pfaff, J., Schiller, P., Kaltschmidt, J., Pruszydlo, M. G., Stremmel, W., et al. (2010). Prevention of adverse drug reactions in intensive care patients by personal intervention based on an electronic clinical decision support system. Intensive Care Med. 36 (4), 665–672. doi:10.1007/s00134-010-1778-8

Dequito, A. B., Mol, P. G., van Doormaal, J. E., Zaal, R. J., van den Bemt, P. M., Haaijer-Ruskamp, F. M., et al. (2011). Preventable and non-preventable adverse drug events in hospitalized patients: A prospective chart review in The Netherlands. Drug Saf. 34 (11), 1089–1100. doi:10.2165/11592030-000000000-00000

Edwards, I. R., and Aronson, J. K. (2000). Adverse drug reactions: Definitions, diagnosis, and management. Lancet 356 (9237), 1255–1259. doi:10.1016/s0140-6736(00)02799-9

Farcas, A., Bucsa, C., Sinpetrean, A., Leucuta, D., Mogosan, C., Dumitrascu, D., et al. (2014). Preventability analysis of adverse drug reactions detected in two internal medicine departments in Romania. Intern. Emerg. Med. 9 (2), 187–193. doi:10.1007/s11739-012-0843-4

Haile, D. B., Ayen, W. Y., and Tiwari, P. (2013). Prevalence and assessment of factors contributing to adverse drug reactions in wards of a tertiary care hospital, India. Ethiop. J. Health Sci. 23 (1), 39–48.

PubMed Abstract | Google Scholar

Hartwig, S. C., Siegel, J., and Schneider, P. J. (1992). Preventability and severity assessment in reporting adverse drug reactions. Am. J. Health. Syst. Pharm. 49 (9), 2229–2232. doi:10.1093/ajhp/49.9.2229

Howard, R. L., Avery, A. J., Slavenburg, S., Royal, S., Pipe, G., Lucassen, P., et al. (2007). Which drugs cause preventable admissions to hospital? A systematic review. Br. J. Clin. Pharmacol. 63 (2), 136–147. doi:10.1111/j.1365-2125.2006.02698.x

Khan, L. M. (2013). Comparative epidemiology of hospital-acquired adverse drug reactions in adults and children and their impact on cost and hospital stay – A systematic review. Eur. J. Clin. Pharmacol. 69 (12), 1985–1996. doi:10.1007/s00228-013-1563-z

Kovacevic, M., Vezmar Kovacevic, S., Radovanovic, S., Stevanovic, P., and Miljkovic, B. (2019). Adverse drug reactions caused by drug-drug interactions in cardiovascular disease patients: Introduction of a simple prediction tool using electronic screening database items. Curr. Med. Res. Opin. 35 (11), 1873–1883. doi:10.1080/03007995.2019.1647021

Kurian, J., Mathew, J., Sowjanya, K., Chaitanya, K. R., Ramesh, M., Sebastian, J., et al. (2016). Adverse drug reactions in hospitalized pediatric patients: A prospective observational study. Indian J. Pediatr. 83 (5), 414–419. doi:10.1007/s12098-015-2002-1

Lau, P. M., Stewart, K., and Dooley, M. J. (2003). Comment: Hospital admissions resulting from preventable adverse drug reactions. Ann. Pharmacother. 37 (2), 303–304. doi:10.1177/106002800303700229

Leone, R., Magro, L., Moretti, U., Cutroneo, P., Moschini, M., Motola, D., et al. (2010). Identifying adverse drug reactions associated with drug-drug interactions: Data mining of a spontaneous reporting database in Italy. Drug Saf. 33 (8), 667–675. doi:10.2165/11534400-000000000-00000

Letinier, L., Ferreira, A., Marceron, A., Babin, M., Micallef, J., Miremont-Salame, G., et al. (2020). Spontaneous reports of serious adverse drug reactions resulting from drug-drug interactions: An analysis from the French pharmacovigilance database. Front. Pharmacol. 11, 624562. doi:10.3389/fphar.2020.624562

Magro, L., Arzenton, E., Leone, R., Stano, M. G., Vezzaro, M., Rudolph, A., et al. (2020). Identifying and characterizing serious adverse drug reactions associated with drug-drug interactions in a spontaneous reporting database. Front. Pharmacol. 11, 622862. doi:10.3389/fphar.2020.622862

Magro, L., Moretti, U., and Leone, R. (2012). Epidemiology and characteristics of adverse drug reactions caused by drug-drug interactions. Expert Opin. Drug Saf. 11 (1), 83–94. doi:10.1517/14740338.2012.631910

Mahady, S. E., Margolis, K. L., Chan, A., Polekhina, G., Woods, R. L., Wolfe, R., et al. (2021). Major GI bleeding in older persons using aspirin: Incidence and risk factors in the ASPREE randomised controlled trial. Gut 70 (4), 717–724. doi:10.1136/gutjnl-2020-321585

Marques, J., Ribeiro-Vaz, I., Pereira, A. C., and Polonia, J. (2014). A survey of spontaneous reporting of adverse drug reactions in 10 years of activity in a pharmacovigilance centre in Portugal. Int. J. Pharm. Pract. 22 (4), 275–282. doi:10.1111/ijpp.12078

Marusic, S., Sicaja, M., Obreli Neto, P. R., Franic, M., Marinovic, I., and Bacic-Vrca, V. (2014). Adverse drug reactions in elderly patients following discharge from an internal medicine clinic. Int. J. Clin. Pharmacol. Ther. 52 (10), 906–913. doi:10.5414/CP202041

Mateti, U., Rajakannan, T., Nekkanti, H., Rajesh, V., Mallaysamy, S., and Ramachandran, P. (2011). Drug-drug interactions in hospitalized cardiac patients. J. Young Pharm. 3 (4), 329–333. doi:10.4103/0975-1483.90246

Mirosevic Skvrce, N., Macolic Sarinic, V., Mucalo, I., Krnic, D., Bozina, N., and Tomic, S. (2011). Adverse drug reactions caused by drug-drug interactions reported to Croatian agency for medicinal products and medical devices: A retrospective observational study. Croat. Med. J. 52 (5), 604–614. doi:10.3325/cmj.2011.52.604

Naranjo, C. A., Busto, U., Sellers, E. M., Sandor, P., Ruiz, I., Roberts, E. A., et al. (1981). A method for estimating the probability of adverse drug reactions. Clin. Pharmacol. Ther. 30 (2), 239–245. doi:10.1038/clpt.1981.154

Obreli-Neto, P. R., Nobili, A., de Lyra, D. P., Pilger, D., Guidoni, C. M., de Oliveira Baldoni, A., et al. (2012b). Incidence and predictors of adverse drug reactions caused by drug-drug interactions in elderly outpatients: A prospective cohort study. J. Pharm. Pharm. Sci. 15 (2), 332–343. doi:10.18433/j3cc86

CrossRef Full Text | Google Scholar

Obreli-Neto, P. R., Nobili, A., de Oliveira Baldoni, A., Guidoni, C. M., de Lyra Junior, D. P., Pilger, D., et al. (2012a). Adverse drug reactions caused by drug-drug interactions in elderly outpatients: A prospective cohort study. Eur. J. Clin. Pharmacol. 68 (12), 1667–1676. doi:10.1007/s00228-012-1309-3

Oscanoa, T. J., Lizaraso, F., and Carvajal, A. (2017). Hospital admissions due to adverse drug reactions in the elderly. A meta-analysis. Eur. J. Clin. Pharmacol. 73 (6), 759–770. doi:10.1007/s00228-017-2225-3

Patel, T. K., Bhabhor, P. H., Desai, N., Shah, S., Patel, P. B., Vatsala, E., et al. (2015). Adverse drug reactions in a psychiatric department of tertiary care teaching hospital in India: Analysis of spontaneously reported cases. Asian J. Psychiatr. 17, 42–49. doi:10.1016/j.ajp.2015.07.003

Ren, W., Liu, Y., Zhang, J., Fang, Z., Fang, H., Gong, Y., et al. (2020). Prevalence of potential drug-drug interactions in outpatients of a general hospital in China: A retrospective investigation. Int. J. Clin. Pharm. 42 (4), 1190–1196. doi:10.1007/s11096-020-01068-3

Roblek, T., Vaupotic, T., Mrhar, A., and Lainscak, M. (2015). Drug-drug interaction software in clinical practice: A systematic review. Eur. J. Clin. Pharmacol. 71 (2), 131–142. doi:10.1007/s00228-014-1786-7

Rottenkolber, D., Schmiedl, S., Rottenkolber, M., Farker, K., Salje, K., Mueller, S., et al. (2011). Adverse drug reactions in Germany: Direct costs of internal medicine hospitalizations. Pharmacoepidemiol. Drug Saf. 20 (6), 626–634. doi:10.1002/pds.2118

Rukmangathen, R., Brahmanapalli, V. D., Thammisetty, D. P., Pemmasani, D., Gali, S. D., and Atmakuru, R. B. (2020). Study of adverse drug reactions to antiretroviral therapy in a tertiary care hospital, Tirupati. Perspect. Clin. Res. 11 (4), 158–163. doi:10.4103/picr.PICR_133_18

Schumock, G. T., and Thornton, J. P. (1992). Focusing on the preventability of adverse drug reactions. Hosp. Pharm. 27 (6), 538.

Scondotto, G., Pojero, F., Pollina Addario, S., Ferrante, M., Pastorello, M., Visconti, M., et al. (2018). The impact of polypharmacy and drug interactions among the elderly population in Western Sicily, Italy. Aging Clin. Exp. Res. 30 (1), 81–87. doi:10.1007/s40520-017-0755-2

Shepherd, G., Mohorn, P., Yacoub, K., and May, D. W. (2012). Adverse drug reaction deaths reported in United States vital statistics, 1999-2006. Ann. Pharmacother. 46 (2), 169–175. doi:10.1345/aph.1P592

Shi, Q. P., He, X. D., Yu, M. L., Zhu, J. X., Liu, Y., Ding, F., et al. (2014). A 6-year retrospective study of adverse drug reactions due to drug-drug interactions between nervous system drugs. Int. J. Clin. Pharmacol. Ther. 52 (5), 392–401. doi:10.5414/CP202050

Singh, P., Agrawal, M., Hishikar, R., Joshi, U., Maheshwari, B., and Halwai, A. (2017). Adverse drug reactions at adverse drug reaction monitoring center in Raipur: Analysis of spontaneous reports during 1 year. Indian J. Pharmacol. 49 (6), 432–437. doi:10.4103/ijp.IJP_781_16

Zheng, W. Y., Richardson, L. C., Li, L., Day, R. O., Westbrook, J. I., and Baysari, M. T. (2018). Drug-drug interactions and their harmful effects in hospitalised patients: A systematic review and meta-analysis. Eur. J. Clin. Pharmacol. 74 (1), 15–27. doi:10.1007/s00228-017-2357-5

Keywords: adverse drug reactions, drug–drug interactions, causality, severity, preventability

Citation: Jiang H, Lin Y, Ren W, Fang Z, Liu Y, Tan X, Lv X and Zhang N (2022) Adverse drug reactions and correlations with drug–drug interactions: A retrospective study of reports from 2011 to 2020. Front. Pharmacol. 13:923939. doi: 10.3389/fphar.2022.923939

Received: 19 April 2022; Accepted: 19 July 2022; Published: 22 August 2022.

Reviewed by:

Copyright © 2022 Jiang, Lin, Ren, Fang, Liu, Tan, Lv and Zhang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

*Correspondence: Xiaoqun Lv, [email protected] ; Ning Zhang, [email protected]

† These authors have contributed equally to this work and share last authorship

‡ ORCID: Xiaoqun Lv, orcid.org/0000-0002-6130-9904 ; Ning Zhang, orcid.org/0000-0002-8376-1037

Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.

• Research article
• Open access
• Published: 03 September 2018

Potential drug-to-drug interactions: a cross-sectional study among older patients discharged from hospital to home care

• Rose Mari Olsen   ORCID: orcid.org/0000-0002-7009-8803 1 &
• Hege Sletvold 1  

Safety in Health volume  4 , Article number:  8 ( 2018 ) Cite this article

4136 Accesses

8 Citations

1 Altmetric

Metrics details

There are major challenges in ensuring medication safety and preventing adverse drug events in older people. Older people are especially vulnerable to drug-to-drug interactions (DDIs). However, when older patients are transferred from hospital to home care, information related to DDIs is limited. The objectives of this study were to (1) identify and describe potential DDIs in older patients discharged from hospital to home care and (2) identify patient and hospital transfer characteristics associated with the potential DDIs.

This is a cross-sectional study of patients discharged from medical hospital wards to home care in central Norway. Nursing transfer documents, including medication lists, were reviewed from records of 99 older inpatients on the day of discharge. The patients’ drug regimens were screened using the Norwegian drug interaction database, www.interaksjoner.no . Descriptive statistics and univariable and multivariable linear regression analyses were used to analyze the data.

The mean age of the sample was 82.9 years; 58.6% were female. In total, 274 DDIs were identified. Major DDIs were identified in two patients, moderate DDIs in 80 patients, and minor DDIs in 40 patients. At least one potential DDI was found for 84 of the patients, with an average of 2.77 DDIs per patient. The most frequent DDIs were related to the concomitant use of alendronate and calcium. Warfarin treatments were frequently linked to DDIs. Potential DDIs were associated with the number of prescribed drugs, age, and living situation.

Conclusions

This study shows that potentially clinically relevant DDIs are common for older patients transferred from hospital to home care and pose a risk for these patients’ health. Monitoring for potential DDIs is highly important to ensure patient safety, and home care nurses might play a significant role through awareness and early recognition.

Drugs are essential in the care and treatment of older patients, and used correctly, drugs contribute to better health and increased quality of life for many. However, there are major challenges in ensuring medication safety and preventing adverse drug events (ADEs) in older people [ 1 , 2 , 3 ]. A specific type of ADE, a drug-to-drug interaction (DDI), occurs when one drug changes the effect of another drug. Older people are especially vulnerable to DDIs and DDI-related adverse events because of age-related pharmacodynamic and pharmacokinetic changes, an increased risk for multimorbidity and, consequently, polypharmacy [ 4 ]. DDIs contribute to adverse drug reactions (ADR) and the burden of iatrogenic illnesses in older people [ 5 ]. They also increase hospital visits and admissions [ 6 ], hospital readmission [ 7 ], and mortality [ 8 ] and represent a significant burden in terms of healthcare costs [ 9 ].

For older people, discharge from hospital to home is an especially vulnerable situation. They often have complex care needs caused by multimorbidity and multiple functional limitations [ 10 , 11 ]. In addition, the overlap of acute and chronic diseases when transferred from hospital may increase older patients’ susceptibility to ADRs [ 5 ], including ADRs caused by DDIs. During hospitalization, extensive changes may be made in the patients’ medication regimen [ 12 ], with prescription of new drugs causing new potential DDIs. After discharge, the patients most often face a drastic decrease in the number of supporting personnel to assist them in successfully complying with an often new and more complex drug regimen [ 13 ]. The patient might be without helping relatives and with sparse follow-up care from home health nursing; thus, the self-management skills of the patient are essential. The general practitioner (GP) must decide whether or not to maintain the change in the patient’s medication regimen; however, incomplete communication about medication management between the hospital and GP at the time of discharge is common [ 12 , 14 ]. In addition, discrepancies may occur between the hospital discharge medication list and the list of medications that the patient actually uses at home [ 15 , 16 ].

Although numerous studies have evaluated potentially inappropriate medication use among older people, information related to DDIs in older patients transferred from hospital to home care is limited. The objectives of this study were to (1) identify and describe potential DDIs in older patients discharged from hospital to home care and (2) identify patient and hospital transfer characteristics associated with the potential DDIs.

Study setting and sample

A cross-sectional study using consecutive sampling was carried out at a general medicine ward and a geriatric unit at a local hospital in central Norway from August 2010 to June 2011. The sample consisted of nursing transfer documents and medication lists from records of older inpatients admitted from their own homes. The criteria for inclusion were that the patients were 70+ years of age and consent competent and that they received home health nursing both before and after the hospitalization. Contact persons at the hospital wards evaluated all patients for the inclusion criteria and enrolled the participants. A sample size of 100 was determined appropriate [ 17 , 18 ]. Recruitment was challenging because many of the patients did not fulfill the criteria for receiving home health nursing after hospitalization: due to bad condition, they were discharged to a higher care level in primary care (i.e., nursing home). Overall, 111 patients met the inclusion criteria. Of these, nine refused to participate and three were excluded due to lack of information regarding medical prescriptions, leaving 99 patients enrolled in the study.

Data collection

Hospital transfer documents and background data were retrieved (by the contact persons) from the patient records on the day of discharge. Information on prescription drugs (drug names, dosage, and frequency) was collected from the physicians’ discharge notes. These were attached to the nursing discharge notes because they contained medication information, intended for medication administration by home health nurses. Background data were used to identify patient characteristics (age, gender, living situation, housing situation, and distance from hospital) and hospital transfer characteristics (type of hospitalization, readmission status, medical department facility, and length of hospital stay).

Drug use and potential drug-to-drug interactions

All drugs were classified according to the Anatomical Therapeutic Chemical Classification (ATC) code. Variables of drug use included; prescribed drugs (in total), scheduled drugs, and drugs “as required” (pro re nata).

The potential DDIs were identified and evaluated by the interdisciplinary research team, i.e., a registered nurse (first author) and a pharmacist (second author). We used the database www.interaksjoner.no to check for DDIs. The database is maintained by the Norwegian Medicines Agency [ 19 ]. It contains information on both pharmacodynamic and pharmacokinetic drug interactions classified into three categories according to the “traffic light system” in terms of clinical relevance, i.e., assumed severity (major, moderate, minor). Red alerts in the database concern drug combinations to avoid (major), yellow alerts concern drugs that may be combined, but precautions need to be taken, e.g., dose changes or monitoring of clinical and/or laboratory parameters (moderate), and green alerts concern drugs where there is only a theoretical chance of a DDI, and drugs may be combined (minor). Some DDIs that appear in the database are clinically relevant only under specific circumstances, e.g., an interaction between warfarin and paracetamol, with a resulting increase in the international normalized ratio (INR) is only relevant if the doses of paracetamol are high, and treatment continues for several days. This was taken into account in our evaluation of the potential DDIs. However, it was not possible to assess whether all such circumstances were fulfilled, due to uncertainty in this information (e.g., paracetamol is often taken “as required”). The number of DDIs was classified on the basis of the pharmaceutical preparations instead of the pharmacological substances. Thus, the preparation Calcigran Forte® for example, which contains calcium and vitamin D, has only one DDI in combination with hydrochlorothiazide, although both calcium and vitamin D increase the risk of hypercalcemia in combination with hydrochlorothiazide. The database contains information on DDIs for ATC group levels (e.g., A10 drugs used in diabetes) that may not be clinically relevant for all specific drugs in this ATC group, and these DDIs were not included in the results. For example, metformin (A10BA02) and the ACE-inhibitor enalapril apparently interact according to the database, resulting in an increased hypoglycemic effect; however, metformin does not increase the risk of hypoglycemia, and the DDI is not clinically relevant in this case.

Statistical analysis

SPSS version 23.0 for Windows (SPSS Inc., Chicago, IL, USA) was used for statistical analysis. Descriptive statistics were expressed as medians and ranges or means and standard deviations (SD) for continuous variables (where appropriate) and as frequencies and percentages for categorical variables. Univariable analyses were performed to estimate the effect of covariates (patient and hospital transfer characteristics—used as categorical variables and number of prescribed drugs—used as a continuous variable) on the occurrence of potential DDIs. Variables showing a trend in association in univariable analysis ( p  < 0.2) were included in the multivariable linear regression analysis (using the Enter method) to control confounding effects. Several models were tested, and variables were removed until the remaining individual variables had a p value < 0.1. Two-sided p values < 0.05 were considered significant. Modeling results are reported as unstandardized and standardized regression coefficients, p values, and 95% confidence intervals.

Characteristics of the study population

The mean age of the study population ( N  = 99) was 82.9 years (SD = 6.1; range = 70 to 95 years). As shown in Table  1 , 58.6% were females. Most of the patients lived alone (69.7%), in their own homes (71.7%), and within a half hour driving time distance from the hospital (53.5%). With regard to the medical department, 32.3% of the patients were admitted to the geriatric unit and 67.7% to the general medicine ward. Most admissions were urgent (83.8%) and 25.3% were readmissions (within 30 days after discharge). The length of hospital stay varied between 1 and 24 days (mean = 7.12; SD = 4.5). The most common causes of hospitalization were pneumonia (31.3%) and chronic obstructive pulmonary disease (12.1%).

The total number of drugs prescribed were 1108, including 978 scheduled drugs and 130 drugs to be taken “as required.” On average, the patients used 11.3 drugs (median = 11, SD = 3.5), with a range of 2 to 25 drugs. All except one patient used five drugs or more. The number of scheduled drugs per patient was on average 9.88 (SD = 3.3; range = 2–21), and for “as required” drugs, the average per patient was 1.30 (SD = 1.4; range = 0–6).

In total, the patients used 156 different drugs (i.e., active substances). Of these, the most frequently prescribed drugs were prednisolone and metoprolol, both used by 51 of the patients (51.5%).

Potential drug-to-drug interactions

Among the 99 patients in the sample, we found at least one potential DDI for 84 (84.8%) of the patients: major DDIs were identified in two patients, moderate DDIs in 80 patients, and minor DDIs in 40 patients. In total, 274 DDIs were identified, with an average of 2.77 DDIs per patient (median = 2.0; SD = 2.36). The maximum number of potential DDIs identified was 10 in a regimen comprising 17 drugs. According to severity classification, 2 (0.7%) of the DDIs were major, 205 (74.8%) were moderate, and 67 (24.5%) were minor.

Antibiotic agents were involved in 38 DDIs, including the two drug combinations causing potential DDIs classified as major (identified in two different medication regimes): ciprofloxacin (ATC J01M A02) and melatonin (ATC N05C H01), and erythromycin (ATC J01F A01) and simvastatin (ATC C10A A01). In addition, amoxicillin, mycostatin, pivmecillinam, metronidazole, and phenoxymethylpenicillin were identified in potential DDIs.

Table  2 shows the most common potentially clinically relevant (moderate severity) DDIs identified, and a description of the mechanism, and possible precautions, for handling of the DDIs. These DDIs involved regularly used drugs, e.g., alendronate, calcium, warfarin, and paracetamol. Six drug combinations accounted for 37% of the moderate DDIs. Warfarin was frequently involved in potential DDIs, accounting for 24% ( N  = 66) of all the DDIs. In total, 80 dissimilar drug interaction pairs were identified.

Factors associated with potential DDIs

Results from both univariable linear regression and the multivariable linear regression analyses assessing the relationship between DDIs and independent variables are given in Table  3 . Univariable analysis showed a statistically significant correlation between the number of potential DDIs and living situation ( p value < 0.001), number of prescribed drugs ( p value < 0.001), age ( p value = 0.008), and housing situation ( p value = 0.033). There was no significant relationship with gender, distance from hospital, type of hospitalization, length of hospital stay, readmission status, or hospital facility.

Multivariable linear regression analyses showed that age, number of prescribed drugs, and living situation were significantly associated with DDIs, after controlling for the housing situation. These variables accounted for approximately 47% of the variability in DDIs ( p value < 0.001) (Table  3 ). Age has a negative influence on DDIs, i.e., patients of a younger age are more likely to experience DDIs. Patients living with someone are also more likely to have DDIs than patients living alone. For every one-unit increase in the number of drugs, the number of DDIs increases by on average 0.33.

The objectives of this study were to identify and describe potential DDIs in older patients discharged from hospital to home care and to identify patient and hospital transfer characteristics associated with potential DDIs. Potential DDIs, particularly moderate ones, were common in this population (84.8%). This is in agreement with Marusic et al. [ 20 ] who found potential DDIs in 85.6% (190 out of 222) of older patients discharged from an internal medicine clinic in Croatia, but somewhat in contrast to Pasina et al. [ 8 ] who reported potential DDIs in 60.5% (1642 out of 2712) of an equivalent patient group in Italy. A major limitation in our study is the lack of information on whether or not the potential DDIs actually led to ADRs or a change of therapeutic effect. Marusic et al. [ 20 ] followed patients for 30 consecutive days after discharge and detected actual DDIs in 21 (9.5%) of the patients, where two lacked therapeutic effect and 19 experienced ADRs. Thus, our results almost certainly overestimate the prevalence of actual DDIs. Furthermore, the patients may not take all drugs as prescribed, and prescribers may be well aware of the DDIs, e.g., have adjusted the dosage or followed monitoring parameters.

The most common DDIs identified in this study (Table  2 ) are in accordance with other studies. Warfarin is well known for its DDI potential [ 20 ], as are digitalis glycosides [ 9 ]. In addition, a high number of DDIs concerning alendronate and calcium carbonate/vitamin D 3 were detected in this sample, and these drugs are frequently combined for the treatment of osteoporosis. In order to avoid reduced absorption of alendronate, drugs need to be taken at different time points during the day [ 21 ]. Therefore, the prevention of this DDI is left to the patient or the home health nurse administering the drugs. This may be challenging, not least because the patients at discharge often are in a vulnerable state, and may have physical and cognitive impairment. Corbett et al. [ 15 ] found that 40% of older patients had one or more medication discrepancies at the patient level, e.g., intentional and non-intentional non-adherence and not filling prescriptions, when transferred from hospital to home. Consequently, the patient’s self-management skills that are taught by the home health nurse are essential. In the study of Sino et al. [ 22 ], 80.3% of the home health nurses felt responsible for their older patients’ proper medication use, but the mean score for knowledge of drug interactions was 77% of the maximum score.

Fortunately, the DDIs most frequently described in our study may be preventable and/or manageable if the prescribers have the complete overview of the patient’s drug use. However, drugs taken “as required” may cause special challenges in terms of DDIs, generating transient DDIs that may be difficult to detect [ 4 ]. In this study, the concomitant use of warfarin and paracetamol may be worth noting (Table  2 ). Paracetamol in high doses may increase the INR [ 19 ]. A meta-analysis of seven randomized controlled trials ( N  = 225 patients) found that paracetamol was associated with a mean 0.62 INR increase (95% CI 0.46–0.78) compared to a placebo, when the daily dosage of paracetamol was in the range of 1–4 g [ 23 ]. Thus, the patients using warfarin that increased their doses of paracetamol without informing the physician may be at increased risk of bleeding. Corbett et al. [ 15 ] reported that older patients transferred from hospital to home reported substituting prescribed pain medication with pain medication they had at home. To substitute paracetamol for non-steroidal anti-inflammatory drugs (NSAIDs), resulting in a contraindicated DDI [ 19 ], may have deleterious effects on the risk of bleeding. Moreover, the warfarin interaction potential requires awareness, and this study identified warfarin as responsible for a quarter of the potential DDIs, including combination with the corticosteroid prednisolone. A study of 32 patients on long-term warfarin therapy that initiated short-term oral corticosteroid therapy found a mean difference between pre- and post-INR values of 1.24 (95% CI 0.86–1.62), and 16 patients (50%) required a modification of warfarin therapy [ 24 ]. Thus, identification of warfarin drug combinations to avoid, or ones that require precautions, is essential to ensure the necessary monitoring and adjustment of dosages.

As the patients in the present study were transferred from hospital, the use of antibiotics in the study sample was common and can be related to the frequent causes of hospitalization, i.e., pneumonia and chronic obstructive pulmonary disease. Therefore, we chose to focus on DDIs involving prescribed antibiotic drugs. Antibiotic agents were involved in 38 DDIs, of which two were contraindicated. These DDIs may pose a particular challenge to patient safety, as the GP may neither be involved in patient care nor get timely information from the hospital doctor after discharge. Viktil et al. [ 12 ] found that only 24 out of 105 discharge notes of older patients discharged from general medicine departments at Norwegian hospitals were received within a week by the GPs. In the same study, they found that extensive changes were made in drug regimens, both during hospitalization and in the initial months after discharge (3.4 versus 4.4 drug changes per patient, respectively). Bakken et al. [ 25 ] studied patients transferred from hospital to an intermediate-care nursing home unit or hospital ward, and they found an increase in the number of drugs from admission to discharge, mainly due to treatment of infections and pain.

The finding of an association between DDIs and the number of prescribed drugs is in accordance with several other studies [e.g., 9 , 26 ]. Age as a predictor of DDIs has also been reported in previous research [e.g., 9 ]. However, in contrast to most other studies, we found that the number of DDIs decreases with age: Patients aged 70–84 years had significantly more DDIs than those aged 85–95, even when controlling for other variables. Fialova et al. [ 27 ], in a study of potentially inappropriate medication (PIM) use in older patients ( N  = 2707) receiving home care in eight European countries, found that PIM use was negatively associated with age for patients of 85 years and older. The authors suggest that this finding could be explained by greater physician awareness of PIM in the oldest patients, or by a higher mortality rate in this age group. Although Fialova et al. did not include DDIs in their criteria for PIM, their explanation may also be transferrable to our findings.

Our result showing an association between DDIs and living situation, i.e., those living with someone are more likely to have DDIs, was somewhat surprising. A possible explanation may be that patients living with someone receive help from their relatives and, consequently, receive less follow-up care from the healthcare services. Unfortunately, we do not have information about frequency of assistance provided by the home health nursing services in order to investigate this hypothesis.

Strengths and weaknesses

The small sample size, and the use of consecutive sampling, limits the validity and generalizability of the results. Although the patients in the sample were representative of those in the hospital in terms of age and gender, a selection bias may have occurred since we only enrolled patients with informed consent (and excluded those with cognitive impairments, e.g., dementia).

This cross-sectional study used correlational analysis in certain aspects, and therefore, only associations and not causal relationships can be established. There is no way to establish temporal sequences—that is, for example, which came first the “DDIs” or the “Number of prescribed drugs used.” According to R-squared, the model explained 47% of the variance in DDI. This fact suggests that other patient and transfer characteristics were likely to be of importance.

The identification and assessment of potential DDIs was performed by a registered nurse (first author) and a pharmacist (second author), and this interdisciplinary approach strengthens the analysis. We chose to use the Norwegian drug interaction database [ 19 ], as this is a well-known database in Norway, based on data from the Norwegian Medicines Agency’s decision support system. Both prescribers and pharmacists use this database. DDI databases differ in the level of documentation and classification of DDIs [ 28 ], and our results may have been different if we had used other databases.

This study was concerned with potential DDIs based on medication lists, and no attempt was made to determine whether the patients actually took the medications or whether the interaction resulted in clinically relevant DDIs. Studies have shown that errors in the medication lists at the time of discharge from hospitals are common [ 16 ]. Future research needs to assess the prevalence of clinically relevant DDIs in older patients discharged from hospital to home care.

Implications

Older patients discharged from hospital to home care are frequently exposed to potential DDIs, and this requires special awareness among healthcare professionals. In their roles as caregivers and administrators of medications, with regular contact with the patients, home care nurses are particularly well positioned to be the most astute observers of DDIs and can recognize and monitor relevant clinical symptoms. Thereby, nurse-led structured medication monitoring can be effective in preventing DDIs, as previously described as an intervention to reduce adverse drug reactions [ 29 ]. However, further studies are needed to explore the potential benefits of such interventions. A qualitative study from Sweden show that registered nurses working in home care settings for older patients can contribute in pharmacovigilance regarding these patients’ drug treatment [ 30 ]. Ensuring safe and effective drug treatment for the individual patient should involve interprofessional strategies. An interdisciplinary team including GPs, nurses, and pharmacists can utilize the integrated medicines management (IMM) model [ 31 ], which includes medication reconciliation, medication review, and patient education, with the aim of preventing and reducing DDIs as well as other ADEs. Thereby, a multi-disciplined, professional team can contribute to medication safety for home-dwelling older patients after hospital discharge.

The present study shows that potential DDIs are frequent among older patients discharged from hospital to home and that DDIs might be associated with the number of prescribed drugs, patient age, and living situation. Monitoring for potential DDIs is highly important to ensure patient safety. In addition to the GP and pharmacist, home care nurses, who visit their patients in their homes on a regular basis, can assist in early recognition of potential DDIs in home-dwelling, older patients after hospital discharge and thereby contribute to the safe use of medicines.

Abbreviations

Adverse drug events

Adverse drug reactions

Anatomical Therapeutic Chemical Classification

• Drug-to-drug interactions

General practitioner

International normalized ratio

Potentially inappropriate medication

Standard deviation

Scripcaru G, Mateus C, Nunes C. Adverse drug events-analysis of a decade. A Portuguese case-study, from 2004 to 2013 using hospital database. PLoS One. 2017;12:e0178626. https://doi.org/10.1371/journal.pone.0178626 .

Article   PubMed   PubMed Central   CAS   Google Scholar  

Field TS, Gurwitz JH, Harrold LR, et al. Risk factors for adverse drug events among older adults in the ambulatory setting. J Am Geriatr Soc. 2004;52:1349–54.

Article   PubMed   Google Scholar  

Gallagher PF, O’Connor MN, O’Mahony D. Prevention of potentially inappropriate prescribing for elderly patients: a randomized controlled trial using STOPP/START criteria. Clin Pharmacol Ther. 2011;89:845–54.

Article   PubMed   CAS   Google Scholar  

Mallet L, Spinewine A, Huang A. Prescribing in elderly people 2: the challenge of managing drug interactions in elderly people. Lancet. 2007;370:185–91.

Marengoni A, Pasina L, Concoreggi C, et al. Understanding adverse drug reactions in older adults through drug–drug interactions. Eur J Intern Med. 2014;25:843–6.

Dechanont S, Maphanta S, Butthum B, et al. Hospital admissions/visits associated with drug–drug interactions: a systematic review and meta-analysis. Pharmacoepidemiol Drug Saf. 2014;23:489–97.

Hauviller L, Eyvrard F, Garnault V, et al. Hospital re-admission associated with adverse drug reactions in patients over the age of 65 years. Eur J Clin Pharmacol. 2016;72:631–9.

Pasina L, Djade CD, Nobili A, et al. Drug-drug interactions in a cohort of hospitalized elderly patients. Pharmacoepidemiol Drug Saf. 2013;22:1054–60.

PubMed   Google Scholar  

Moura CS, Acurcio FA, Belo NO. Drug-drug interactions associated with length of stay and cost of hospitalization. J Pharm Pharm Sci. 2009;12:266–72.

Marengoni A, Angleman S, Melis R, et al. Aging with multimorbidity: a systematic review of the literature. Ageing Res Rev. 2011;10:430–9.

Nardi R, Scanelli G, Corrao S, et al. Co-morbidity does not reflect complexity in internal medicine patients. Eur J Intern Med. 2007;18:359–68.

Viktil KK, Blix HS, Eek AK, et al. How are drug regimen changes during hospitalisation handled after discharge: a cohort study. BMJ Open. 2012;2:e001461. https://doi.org/10.1136/bmjopen-2012-001461 .

Article   PubMed   PubMed Central   Google Scholar  

Hesselink G, Flink M, Olsson M, et al. Are patients discharged with care? A qualitative study of perceptions and experiences of patients, family members and care providers. BMJ Qual Saf. 2012;21(Suppl 1):i39–49.

Uitvlugt E, Siegert C, Janssen M, et al. Completeness of medication-related information in discharge letters and post-discharge general practitioner overviews. Int J Clin Pharm. 2015;37:1206–12.

Corbett CF, Setter SM, Daratha KB, et al. Nurse identified hospital to home medication discrepancies: implications for improving transitional care. Geriatr Nurs. 2010;31:188–96.

Michaelsen MH, Mccague P, Bradley CP, et al. Medication reconciliation at discharge from hospital: a systematic review of the quantitative literature. Pharmacy. 2015;3:53–71.

Green SB. How many subjects does it take to do a regression analysis. Multivar Behav Res. 1991;26:499–510.

Article   CAS   Google Scholar  

Soper, D.S. A-priori sample size calculator for multiple regression. 2018. https://www.danielsoper.com/statcalc/calculator.aspx?id=1 . Accessed 29 Jan 2018.

Google Scholar  

interaksjoner.no [Interaction database for norwegian clinicians]. 2018. http://interaksjoner.azurewebsites.net/ . Accessed 29 Jan 2018.

Marusic S, Bacic-Vrca V, Obreli Neto P, et al. Actual drug–drug interactions in elderly patients discharged from internal medicine clinic: a prospective observational study. Eur J Clin Pharmacol. 2013;69:1717–24.

Rosen HN, Rosen CJ, Schmader KE, et al. The use of bisphosphonates in postmenopausal women with osteoporosis. UpToDate2017. Available from: www.uptodate.com . Accessed 29 Jan 2018.

Sino CGM, Munnik A, Schuurmans MJ. Knowledge and perspectives of Dutch home healthcare nurses regarding medication frequently used by older people. Int J Older People Nursing. 2013;8:131–8.

Article   Google Scholar  

Caldeira D, Costa J, Barra M, et al. How safe is acetaminophen use in patients treated with vitamin K antagonists? A systematic review and meta-analysis. Thromb Res. 2015;135:58–61.

Hazlewood KA, Fugate SE, Harrison DL. Effect of oral corticosteroids on chronic warfarin therapy. Ann Pharmacother. 2006;40:2101–6.

Bakken MS, Ranhoff AH, Engeland A, et al. Inappropriate prescribing for older people admitted to an intermediate-care nursing home unit and hospital wards. Scand J Prim Health Care. 2012;30:169–75.

Johnell K, Klarin I. The relationship between number of drugs and potential drug-drug interactions in the elderly. Drug Saf. 2007;30:911–8.

Fialová D, Topinková E, Gambassi G, et al. Potentially inappropriate medication use among elderly home care patients in Europe. JAMA. 2005;293:1348–58.

Kongsholm G, Nielsen A, Damkier P. Drug interaction databases in medical literature: transparency of ownership, funding, classification algorithms, level of documentation, and staff qualifications. A systematic review. Eur J Clin Pharmacol. 2015;71:1397–402.

Gabe ME, Davies GA, Murphy F, et al. Adverse drug reactions: treatment burdens and nurse-led medication monitoring. J Nurs Manag. 2011;19:377–92.

Johansson-Pajala RM, Jorsäter Blomgren K, Bastholm-Rahmner P, et al. Nurses in municipal care of the elderly act as pharmacovigilant intermediaries: a qualitative study of medication management. Scand J Prim Health Care. 2016;34:37–45.

Scullin C, Scott MG, Hogg A, et al. An innovative approach to integrated medicines management. J Eval Clin Pract. 2007;13:781–8.

Download references

Acknowledgements

The authors wish to thank the patients who participated in the study, and the contact persons at the hospital wards who performed the inclusion and data collection. The authors also acknowledge Linda Amundstuen Reppe for the valuable support and helpful discussions.

Available of data and materials

The datasets analyzed during the current study are available from the corresponding author on reasonable request.

Author information

Authors and affiliations.

Faculty of Nursing and Health Sciences, Nord University, PB 474, N-7801, Namsos, Norway

Rose Mari Olsen & Hege Sletvold

You can also search for this author in PubMed   Google Scholar

Contributions

RMO designed the study and collected the data. RMO and HS identified and evaluated the potential DDIs. RMO conducted the statistical analysis, created the tables, and led the drafting of the manuscript. RMO and HS were both involved in critically revising the manuscript for important intellectual content and both read and approved the final manuscript.

Corresponding author

Correspondence to Rose Mari Olsen .

Ethics declarations

Ethics approval and consent to participate.

The research project was approved by The Committee for Medical and Health Research Ethics of Norway (no. 2009/815) and carried out in accordance with the Declaration of Helsinki. Informed consent was obtained from all patients.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License ( http://creativecommons.org/licenses/by/4.0/ ), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/ ) applies to the data made available in this article, unless otherwise stated.

Reprints and permissions

About this article

Cite this article.

Olsen, R.M., Sletvold, H. Potential drug-to-drug interactions: a cross-sectional study among older patients discharged from hospital to home care. Saf Health 4 , 8 (2018). https://doi.org/10.1186/s40886-018-0075-z

Download citation

Received : 19 June 2018

Accepted : 19 August 2018

Published : 03 September 2018

DOI : https://doi.org/10.1186/s40886-018-0075-z

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Older patients
• Patient safety
• Hospital discharge
• Primary care

Safety in Health

ISSN: 2056-5917

• Submission enquiries: [email protected]
• General enquiries: [email protected]

• - Google Chrome

Intended for healthcare professionals

• Access provided by Google Indexer
• My email alerts
• BMA member login
• Username * Password * Forgot your log in details? Need to activate BMA Member Log In Log in via OpenAthens Log in via your institution

Search form

• Advanced search
• Search responses
• Search blogs
• Distinct...

Distinct electrocardiographic findings of a drug-drug interaction

• Related content
• Peer review
• Yubin Zhang , resident physician 1 ,
• Tong Liu , professor of medicine 2 ,
• Gan-Xin Yan , professor of medicine 3 4 5
• 1 Department of Electrocardiogram , the First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, China
• 2 Tianjin Key Laboratory of Ionic-Molecular Function of Cardiovascular disease, Department of Cardiology, Tianjin Institute of Cardiology, Second Hospital of Tianjin Medical University, Tianjin, China
• 3 Lankenau Medical Center and Lankenau Institute for Medical Research, Wynnewood, PA, USA
• 4 Fuwai Huazhong Hospital, Chinese Academy of Medical Sciences, Zhengzhou, Henan, China
• 5 Sidney Kimmel Medical College at Thomas Jefferson University, Philadelphia, PA, USA
• Correspondence to: G-X Yan yanganxinmd{at}comcast.net

A man in his 80s presented to the emergency department with a one day history of light headedness, dyspnoea, poor appetite, and fatigue. He had a history of hypertension, chronic atrial fibrillation, and atrial flutter, and his drug history included warfarin, metoprolol, digoxin, and furosemide. He reported a fever after experiencing symptoms of an upper respiratory tract infection and had taken a four day course of azithromycin empirically before presenting.

In the emergency department, the patient’s blood pressure was 100/68 mm Hg and his heart rate was 49 beats/min. The results of laboratory tests were: brain natriuretic peptide 1560 ng/L (range 0-100 ng/L), serum creatinine 1.5 mg/dL (range 0.7 -1.3 mg/dL for man), serum potassium 3.8 mmol/L (range 3.5-5.5 mmol/L), and serum digoxin 4.23 nmol/L (range 1.02-2.56 nmol/L). A 12 lead electrocardiogram (ECG) was recorded at admission ( fig 1 ). An ECG six months previously had shown atrial fibrillation, prominent U waves, and a controlled ventricular rate. The ECG was repeated two days after the patient stopped digoxin and azithromycin ( fig 2 ).

12 lead electrocardiogram in the emergency department

• Download figure
• Open in new tab
• Download powerpoint

12 lead electrocardiogram two days after discontinuation of digoxin and azithromycin

What do the ECGs show?

What is the most likely diagnosis?

How would you manage this condition?

1. What do the ECGs show?

Figure 1 shows atrial fibrillation, junctional escape beats followed by several consecutive relatively narrow QRS complexes with a beat-to-beat change in their axis, and an alternating left and right bundle branch block pattern. The junctional escape beats during atrial fibrillation indicate a high degree atrioventricular block. Non-sustained relatively narrow complex tachycardia favours the diagnosis of bidirectional ventricular tachycardia. Another differential diagnosis is atrial fibrillation with aberrant conduction. In lead aVR, however, the initial q-wave >40 ms in odd beats (black arrowheads) and notching on the initial downstroke of predominantly negative QRS in even beats (hollow arrowheads) support bidirectional ventricular tachycardia ( fig 3 ). 1 Figure 4 shows atypical atrial flutter with varying atrioventricular conduction, downsloping ST segment depression in V5 to V6 (“reverse tick” sign), and U waves in precordial leads ( fig 4 ).

12 lead electrocardiogram two days after discontinuation of digoxin and azithromycin showing atypical atrial flutter with varying atrioventricular conduction, downsloping ST segment depression in V5 to V6, and U waves in precordial leads

2. What is the most likely diagnosis?

Digoxin toxicity—the patient’s clinical symptoms, age, ECG findings of bidirectional ventricular tachycardia and high degree atrioventricular block together with the raised serum digoxin level suggest digoxin toxicity likely due to concomitant use of azithromycin. 2 3 4

Patients with digoxin toxicity commonly present with gastrointestinal symptoms of nausea or vomiting. Risk factors for digoxin toxicity include drug interactions such as concomitant use of macrolide antibiotics and certain class Ic, III, and IV antiarrhythmic drugs such as flecainide, amiodarone, and verapamil, 4 as well as electrolyte abnormalities such as hypokalaemia or hypomagnesaemia. 2 3 5 Macrolide antibiotics increase serum digoxin level by reducing intestinal and renal transport of digoxin through inhibition of the P-glycoprotein. 4 5

Typical ECG features of digoxin use include downsloping ST segment depression (“reverse tick” sign), commonly seen in the limb and left precordial leads, and U waves that tend to manifest in the mid-precordial leads.

3. How would you manage this condition?

Patients who are haemodynamically unstable should be stabilised according to life support protocols. The mainstay of management is to discontinue any drugs that induce digoxin toxicity and monitor serum electrolyte levels, providing supportive care as needed. 6

Digoxin toxicity can be rapidly reversed with digoxin specific antibody (Fab) fragment. Indications for the use of digoxin specific antibody include hyperkalaemia, haemodynamically compromised bradycardia due to high degree atrioventricular block or ventricular arrhythmias, and digoxin levels >19.21 nmol/L.

Learning points

Consider digoxin toxicity in patients who are taking digoxin and present with nausea, vomiting, decreased appetite, and risk factors such as drug interactions.

ECG findings such as bidirectional ventricular tachycardia and high degree atrioventricular block in patients taking digoxin highly indicate digoxin toxicity.

Digoxin specific antibody should be used in patients with haemodynamic instability or hyperkalaemia.

Patient outcome

After digoxin toxicity was diagnosed, both azithromycin and digoxin were discontinued. Digoxin specific antibody was not used in this patient because of stable junctional narrow QRS rhythm, normal serum potassium levels, and haemodynamic stability. In addition, his non-sustained bidirectional ventricular tachycardia was brief, ranging from three to six beats, and became less frequent after initial supportive care. With disappearance of high degree atrioventricular block and bidirectional ventricular tachycardia two days later, the patient’s symptoms improved. His creatinine and brain natriuretic peptide levels also returned to within normal range. At discharge, the patient continued to have atrial fibrillation and flutter with varying atrioventricular conduction at a well controlled ventricular rate without digoxin.

Competing interests: The BMJ has judged that there are no disqualifying financial ties to commercial companies. The authors declare the following other interests: none.

Further details of The BMJ policy on financial interests are here: https://www.bmj.com/about-bmj/resources-authors/forms-policies-and-checklists/declaration-competing-interests

Patient consent obtained.

Provenance and peer review: Not commissioned; externally peer reviewed.

• Vereckei A ,
• Szénási G ,
• Altemose GT ,
• Chapman M ,
• Hargreaves M ,
• Schneider H ,
• Raja Rao MP ,
• Panduranga P ,
• Sulaiman K ,
• Al-Jufaili M
• from the American Heart Association Electrocardiography and Arrhythmias Committee of the Council of Clinical Cardiology
• Mamdani MM ,
• Juurlink DN
• Kusumoto FM ,
• Schoenfeld MH ,
• Barrett C ,

A review of machine learning-based methods for predicting drug–target interactions

• Published: 12 April 2024
• Volume 12 , article number  30 , ( 2024 )

Cite this article

• Wen Shi 1 , 3   na1 ,
• Hong Yang   ORCID: orcid.org/0000-0002-4328-335X 1   na1 ,
• Linhai Xie 2 ,
• Xiao-Xia Yin 1 &
• Yanchun Zhang 3 , 4  

The prediction of drug–target interactions (DTI) is a crucial preliminary stage in drug discovery and development, given the substantial risk of failure and the prolonged validation period associated with in vitro and in vivo experiments. In the contemporary landscape, various machine learning-based methods have emerged as indispensable tools for DTI prediction. This paper begins by placing emphasis on the data representation employed by these methods, delineating five representations for drugs and four for proteins. The methods are then categorized into traditional machine learning-based approaches and deep learning-based ones, with a discussion of representative approaches in each category and the introduction of a novel taxonomy for deep neural network models in DTI prediction. Additionally, we present a synthesis of commonly used datasets and evaluation metrics to facilitate practical implementation. In conclusion, we address current challenges and outline potential future directions in this research field.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price includes VAT (Russian Federation)

Instant access to the full article PDF.

Rent this article via DeepDyve

Institutional subscriptions

Ambudkar SV, Gottesman MM, editors. ABC Transporters: Biomedical, Cellular, and Molecular Aspects, Methods in Enzymology, vol. 292. San Diego (CA): Academic Press; 1998.

Al-Absi HR, Refaee MA, Rehman AU, Islam MT, Belhaouari SB, Alam T. Risk factors and comorbidities associated to cardiovascular disease in qatar: a machine learning based case-control study. IEEE Access. 2021;9:29929–41.

Article   Google Scholar  

Vermaas JV, Sedova A, Baker MB, Boehm S, Rogers DM, Larkin J, Glaser J, Smith MD, Hernandez O, Smith JC. Supercomputing pipelines search for therapeutics against covid-19. Comput Sci Eng. 2020;23(1):7–16.

Peska L, Buza K, Koller J. Drug-target interaction prediction: a Bayesian ranking approach. Comput Methods Programs Biomed. 2017;152:15–21.

Yamanishi Y, Araki M, Gutteridge A, Honda W, Kanehisa M. Prediction of drug-target interaction networks from the integration of chemical and genomic spaces. Bioinformatics. 2008;24(13):i232–40.

Speck-Planche A, Kleandrova VV, Luan F, Cordeiro MND. A ligand-based approach for the in silico discovery of multi-target inhibitors for proteins associated with HIV infection. Mol Biosyst. 2012;8(8):2188–96.

Yang Y, Zhu Z, Wang X, Zhang X, Mu K, Shi Y, Peng C, Xu Z, Zhu W. Ligand-based approach for predicting drug targets and for virtual screening against covid-19. Briefings Bioinform. 2021;22(2):1053–64.

Aziz F, Cardoso VR, Bravo-Merodio L, Russ D, Pendleton SC, Williams JA, Acharjee A, Gkoutos GV. Multimorbidity prediction using link prediction. Sci Rep. 2021;11(1):16392.

Valentin JP, Guillon JM, Jenkinson S, Kadambi V, Ravikumar P, Roberts S, Rosenbrier-Ribeiro L, Schmidt F, Armstrong D. In vitro secondary pharmacological profiling: an iq-drusafe industry survey on current practices. J Pharmacol Toxicol Methods. 2018;93:7–14.

da Silva Rocha SF, Olanda CG, Fokoue HH, Sant’Anna CM. Virtual screening techniques in drug discovery: review and recent applications. Curr Top Med Chem. 2019;19(19):1751–67.

Mousavian Z, Masoudi-Nejad A. Drug-target interaction prediction via chemogenomic space: learning-based methods. Expert Opin Drug Metab Toxicol. 2014;10(9):1273–87.

Li J, Zheng S, Chen B, Butte AJ, Swamidass SJ, Lu Z. A survey of current trends in computational drug repositioning. Brief Bioinform. 2016;17(1):2–12.

Chen R, Liu X, Jin S, Lin J, Liu J. Machine learning for drug-target interaction prediction. Molecules. 2018;23(9):2208.

Zhang W, Lin W, Zhang D, Wang S, Shi J, Niu Y. Recent advances in the machine learning-based drug-target interaction prediction. Curr Drug Metabol. 2019;20(3):194–202.

Bagherian M, Sabeti E, Wang K, Sartor MA, Nikolovska-Coleska Z, Najarian K. Machine learning approaches and databases for prediction of drug-target interaction: a survey paper. Brief Bioinform. 2021;22(1):247–69.

Stokes JM, Yang K, Swanson K, Jin W, Cubillos-Ruiz A, Donghia NM, MacNair CR, French S, Carfrae LA, Bloom-Ackermann Z, et al. A deep learning approach to antibiotic discovery. Cell. 2020;180(4):688–702.

Thafar MA, Alshahrani M, Albaradei S, Gojobori T, Essack M, Gao X. Affinity2vec: drug-target binding affinity prediction through representation learning, graph mining, and machine learning. Sci Rep. 2022;12(1):4751.

Honda S, Shi S, Ueda H.R. Smiles transformer: Pre-trained molecular fingerprint for low data drug discovery. arXiv preprint arXiv:1911.04738 2019.

Wang YB, You ZH, Yang S, Yi HC, Chen ZH, Zheng K. A deep learning-based method for drug-target interaction prediction based on long short-term memory neural network. BMC Med Inform Decis Mak. 2020;20(2):1–9.

Google Scholar  

Zhao ZY, Huang WZ, Zhan XK, Pan J, Huang YA, Zhang SW, Yu CQ et al. An ensemble learning-based method for inferring drug-target interactions combining protein sequences and drug fingerprints. BioMed Res Int. 2021;2021.

Zhao Z, Bourne PE. Harnessing systematic protein–ligand interaction fingerprints for drug discovery. Drug Discov Today. 2022;27(10):103319.

Shao J, Gong Q, Yin Z, Yin Z, Pan W, Pandiyan S, Wang L. S2dv: converting smiles to a drug vector for predicting the activity of anti-hbv small molecules. Brief Bioinform. 2022;23(2):bbab593.

Kurata H, Tsukiyama S, Manavalan B. iacvp: markedly enhanced identification of anti-coronavirus peptides using a dataset-specific word2vec model. Brief Bioinform. 2022;23(4):bbac265.

Lim J, Ryu S, Kim JW, Kim WY. Molecular generative model based on conditional variational autoencoder for de novo molecular design. J Chem. 2018;10(1):1–9.

Li P, Wang J, Qiao Y, Chen H, Yu Y, Yao X, Gao P, Xie G, Song S. Learn molecular representations from large-scale unlabeled molecules for drug discovery. arXiv preprint arXiv:2012.11175 . 2020.

Zhang J, Liu B. A review on the recent developments of sequence-based protein feature extraction methods. Curr Bioinform. 2019;14(3):190–9.

Cui F, Zhang Z, Zou Q. Sequence representation approaches for sequence-based protein prediction tasks that use deep learning. Brief Funct Genom. 2021;20(1):61–73.

Zhang YF, Wang X, Kaushik AC, Chu Y, Shan X, Zhao MZ, Xu Q, Wei DQ. Spvec: a word2vec-inspired feature representation method for drug-target interaction prediction. Front Chem. 2020;7:895.

Wang H, Wang J, Dong C, Lian Y, Liu D, Yan Z. A novel approach for drug-target interactions prediction based on multimodal deep autoencoder. Front Pharmacol. 2020;10:1592.

Jiang M, Li Z, Zhang S, Wang S, Wang X, Yuan Q, Wei Z. Drug-target affinity prediction using graph neural network and contact maps. RSC Adv. 2020;10(35):20701–12.

Wang P, Zheng S, Jiang Y, Li C, Liu J, Wen C, Patronov A, Qian D, Chen H, Yang Y. Structure-aware multimodal deep learning for drug-protein interaction prediction. J Chem Inform Model. 2022;62(5):1308–17.

Zhao Y, Zheng K, Guan B, Guo M, Song L, Gao J, Qu H, Wang Y, Shi D, Zhang Y. Dldti: a learning-based framework for drug-target interaction identification using neural networks and network representation. J Transl Med. 2020;18:1–15.

Lee H, Kim W. Comparison of target features for predicting drug-target interactions by deep neural network based on large-scale drug-induced transcriptome data. Pharmaceutics. 2019;11(8):377.

Li X, Xiong H, Li X, Wu X, Zhang X, Liu J, Bian J, Dou D. Interpretable deep learning: interpretation, interpretability, trustworthiness, and beyond. Knowl Inform Syst. 2022;64(12):3197–234.

Liberzon A, Birger C, Thorvaldsdóttir H, Ghandi M, Mesirov JP, Tamayo P. The molecular signatures database hallmark gene set collection. Cell Syst. 2015;1(6):417–25.

Thafar MA, Olayan RS, Ashoor H, Albaradei S, Bajic VB, Gao X, Gojobori T, Essack M. Dtigems+: drug-target interaction prediction using graph embedding, graph mining, and similarity-based techniques. J Cheminform. 2020;12(1):1–17.

Mohamed SK, Nováček V, Nounu A. Discovering protein drug targets using knowledge graph embeddings. Bioinformatics. 2020;36(2):603–10.

Thafar MA, Olayan RS, Albaradei S, Bajic VB, Gojobori T, Essack M, Gao X. Dti2vec: drug-target interaction prediction using network embedding and ensemble learning. J Cheminform. 2021;13(1):1–18.

Perlman L, Gottlieb A, Atias N, Ruppin E, Sharan R. Combining drug and gene similarity measures for drug-target elucidation. J Comput Biol. 2011;18(2):133–45.

Shi JY, Yiu SM In: 2015 IEEE International Conference on Bioinformatics and Biomedicine (BIBM) (IEEE, 2015): 1636–1641

Shi JY, Li JX, Lu HM, Zhang Y. In: Intelligence Science and Big Data Engineering. Big Data and Machine Learning Techniques: 5th International Conference, IScIDE 2015, Suzhou, June 14–16, 2015, Revised Selected Papers, Part II 5 (Springer, 2015): 477–486

Shi JY, Yiu SM, Li Y, Leung HC, Chin FY. Predicting drug-target interaction for new drugs using enhanced similarity measures and super-target clustering. Methods. 2015;83:98–104.

Buza K, Peška L. Drug-target interaction prediction with bipartite local models and hubness-aware regression. Neurocomputing. 2017;260:284–93.

Pahikkala T, Airola A, Pietilä S, Shakyawar S, Szwajda A, Tang J, Aittokallio T. Toward more realistic drug-target interaction predictions. Brief Bioinform. 2015;16(2):325–37.

He T, Heidemeyer M, Ban F, Cherkasov A, Ester M. Simboost: a read-across approach for predicting drug-target binding affinities using gradient boosting machines. J Cheminform. 2017;9(1):1–14.

Yu H, Chen J, Xu X, Li Y, Zhao H, Fang Y, Li X, Zhou W, Wang W, Wang Y. A systematic prediction of multiple drug-target interactions from chemical, genomic, and pharmacological data. PLoS ONE. 2012;7(5): e37608.

Fu G, Ding Y, Seal A, Chen B, Sun Y, Bolton E. Predicting drug target interactions using meta-path-based semantic network analysis. BMC Bioinform. 2016;17(1):1–10.

Wang MY, Li P, Qiao Pl, et al. The virtual screening of the drug protein with a few crystal structures based on the adaboost-svm. Comput Math Methods Med. 2016:2016

Olayan RS, Ashoor H, Bajic VB. Ddr: efficient computational method to predict drug-target interactions using graph mining and machine learning approaches. Bioinformatics. 2018;34(7):1164–73.

Ezzat A, Wu M, Li XL, Kwoh CK. Computational prediction of drug-target interactions using chemogenomic approaches: an empirical survey. Brief Bioinform. 2019;20(4):1337–57.

Xia Z, Wu LY, Zhou X, Wong ST. In: BMC systems biology, vol. 4 (BioMed Central, 2010): 1–16

Van Laarhoven T, Nabuurs SB, Marchiori E. Gaussian interaction profile kernels for predicting drug-target interaction. Bioinformatics. 2011;27(21):3036–43.

Wang Y, Zeng J. Predicting drug-target interactions using restricted Boltzmann machines. Bioinformatics. 2013;29(13):i126–34.

Aghakhani S, Qabaja A, Alhajj R. Integration of k-means clustering algorithm with network analysis for drug-target interactions network prediction. Int J Data Min Bioinform. 2018;20(3):185–212.

Yang F, Xue F, Zhang Y, Karypis G. Kernelized multitask learning method for personalized signaling adverse drug reactions. IEEE Trans Knowl Data Eng. 2021;35(2):1681–94.

Zheng Y, Tang P, Qiu W, Wang H, Guo J, Huang Z. in International conference on database systems for advanced applications (Springer, 2023):336–352

Yin XX, Jian Y, Zhang Y, Zhang Y, Wu J, Lu H, Su MY. Automatic breast tissue segmentation in mris with morphology snake and deep denoiser training via extended stein’s unbiased risk estimator. Health Inform Sci Syst. 2021;9:1–21.

Ma T, Xiao C, Zhou J, Wang F. Drug similarity integration through attentive multi-view graph auto-encoders. arXiv preprint arXiv:1804.10850 . 2018.

Vázquez J, López M, Gibert E, Herrero E, Luque FJ. Merging ligand-based and structure-based methods in drug discovery: an overview of combined virtual screening approaches. Molecules. 2020;25(20):4723.

Ravì D, Wong C, Deligianni F, Berthelot M, Andreu-Perez J, Lo B, Yang GZ. Deep learning for health informatics. IEEE J Biomed Health Inform. 2016;21(1):4–21.

Abbasi K, Razzaghi P, Poso A, Ghanbari-Ara S, Masoudi-Nejad A. Deep learning in drug target interaction prediction: current and future perspectives. Curr Med Chem. 2021;28(11):2100–13.

Thafar M, Raies AB, Albaradei S, Essack M, Bajic VB. Comparison study of computational prediction tools for drug-target binding affinities. Front Chem. 2019;7:782.

Nikraftar Z, Keyvanpour MR. A comparative analytical review on machine learning methods in Drugtarget interactions prediction. Curr Comput-Aided Drug Des. 2023;19(5):325–55.

Liu X, Yan M, Deng L, Li G, Ye X, Fan D, Pan S, Xie Y. Survey on graph neural network acceleration: An algorithmic perspective. arXiv preprint arXiv:2202.04822 . 2022

Huang L, Lin J, Liu R, Zheng Z, Meng L, Chen X, Li X, Wong KC. Coadti: multi-modal co-attention based framework for drug–target interaction annotation. Brief Bioinform. 2022;23(6):bbac446.

Hua Y, Song X, Feng Z, Wu X. Mfr-dta: a multi-functional and robust model for predicting drug–target binding affinity and region. Bioinformatics. 2023;39(2):btad056.

Gim M, Choe J, Baek S, Park J, Lee C, Ju M, Lee S, Kang J, Kang J. Arkdta: attention regularization guided by non-covalent interactions for explainable drug–target binding affinity prediction. Bioinformatics. 2023;39(1):i448–57.

Yuan Y, Chang S, Zhang Z, Li Z, Li S, Xie P, Yau WP, Lin H, Cai W, Zhang Y, et al. A novel strategy for prediction of human plasma protein binding using machine learning techniques. Chemom Intell Lab Syst. 2020;199: 103962.

Askr H, Elgeldawi E, Aboul Ella H, Elshaier YA, Gomaa MM, Hassanien AE. Deep learning in drug discovery: an integrative review and future challenges. Artif Intell Rev. 2023;56(7):5975–6037.

Öztürk H, Özgür A, Ozkirimli E. Deepdta: deep drug-target binding affinity prediction. Bioinformatics. 2018;34(17):i821–9.

Öztürk H, Ozkirimli E, Özgür A. Widedta: prediction of drug-target binding affinity. arXiv preprint arXiv:1902.04166 . 2019.

Lee I, Keum J, Nam H. Deepconv-dti: prediction of drug-target interactions via deep learning with convolution on protein sequences. PLoS Comput Biol. 2019;15(6): e1007129.

Zheng X, He S, Song X, Zhang Z, Bo X. In: Artificial neural networks and machine learning-ICANN 2018: 27th international conference on artificial neural networks, Rhodes, October 4–7, 2018, Proceedings, Part I 27 (Springer, 2018): 104–114

Rifaioglu AS, Cetin Atalay R, Cansen Kahraman D, Doğan T, Martin M, Atalay V. Mdeepred: novel multi-channel protein featurization for deep learning-based binding affinity prediction in drug discovery. Bioinformatics. 2021;37(5):693–704.

Nguyen T, Le H, Quinn TP, Nguyen T, Le TD, Venkatesh S. Graphdta: predicting drug-target binding affinity with graph neural networks. Bioinformatics. 2021;37(8):1140–7.

Wu Y, Gao M, Zeng M, Zhang J, Li M. Bridgedpi: a novel graph neural network for predicting drug-protein interactions. Bioinformatics. 2022;38(9):2571–8.

Zhang H, Hu J, Zhang X. In: International conference on intelligent computing (Springer, 2022): 533–546

Yang Z, Zhong W, Zhao L, Chen CYC. Mgraphdta: deep multiscale graph neural network for explainable drug-target binding affinity prediction. Chem Sci. 2022;13(3):816–33.

Mukherjee S, Ghosh M, Basuchowdhuri P. In: Proceedings of the 2022 SIAM international conference on data mining (SDM) (SIAM, 2022): 729–737

Zhao Q, Xiao F, Yang M, Li Y, Wang J. In: 2019 IEEE international conference on bioinformatics and biomedicine (BIBM) (IEEE, 2019): 64–69

Chen L, Tan X, Wang D, Zhong F, Liu X, Yang T, Luo X, Chen K, Jiang H, Zheng M. Transformercpi: improving compound-protein interaction prediction by sequence-based deep learning with self-attention mechanism and label reversal experiments. Bioinformatics. 2020;36(16):4406–14.

Ghimire A, Tayara H, Xuan Z, Chong KT. Csatdta: prediction of drug-target binding affinity using convolution model with self-attention. Int J Mol Sci. 2022;23(15):8453.

Monteiro NR, Oliveira JL, Arrais JP. Dtitr: end-to-end drug-target binding affinity prediction with transformers. Comput Biol Med. 2022;147: 105772.

Cowen L, Ideker T, Raphael BJ, Sharan R. Network propagation: a universal amplifier of genetic associations. Nat Rev Genet. 2017;18(9):551–62.

Luo Y, Zhao X, Zhou J, Yang J, Zhang Y, Kuang W, Peng J, Chen L, Zeng J. A network integration approach for drug-target interaction prediction and computational drug repositioning from heterogeneous information. Nat Commun. 2017;8(1):573.

Eslami Manoochehri H, Nourani M. Drug-target interaction prediction using semi-bipartite graph model and deep learning. BMC bioinform. 2020;21:1–16.

Wang Z, Zhou M, Arnold C. Toward heterogeneous information fusion: bipartite graph convolutional networks for in silico drug repurposing. Bioinformatics. 2020;36(1):i525–33.

Li X, Ma D, Ren Y, Luo J, Li Y. Large-scale prediction of drug-protein interactions based on network information. Curr Comput-Aided Drug Des. 2022;18(1):64–72.

Peng J, Li J, Shang X. A learning-based method for drug-target interaction prediction based on feature representation learning and deep neural network. BMC Bioinform. 2020;21(13):1–13.

Peng J, Wang Y, Guan J, Li J, Han R, Wei Z, Shang X. An end-to-end heterogeneous graph representation learning-based framework for drug–target interaction prediction. Brief Bioinform. 2021;22(5):bbaa430.

Zhou D, Xu Z, Li W, Xie X, Peng S. Multidti: drug–target interaction prediction based on multi-modal representation learning to bridge the gap between new chemical entities and known heterogeneous network. Bioinformatics. 2021;37(23):4485–92.

Wang S, Du Z, Ding M, Rodriguez-Paton A, Song T. Kg-dti: a knowledge graph based deep learning method for drug-target interaction predictions and Alzheimer’s disease drug repositions. Appl Intell. 2022;52(1):846–57.

Wang W, Liang S, Yu M, Liu D, Zhang H, Wang X, Zhou Y. Gchn-dti: predicting drug-target interactions by graph convolution on heterogeneous networks. Methods. 2022;206:101–7.

Tunyasuvunakool K, Adler J, Wu Z, Green T, Zielinski M, Žídek A, Bridgland A, Cowie A, Meyer C, Laydon A, et al. Highly accurate protein structure prediction for the human proteome. Nature. 2021;596(7873):590–6.

Baek M, DiMaio F, Anishchenko I, Dauparas J, Ovchinnikov S, Lee GR, Wang J, Cong Q, Kinch LN, Schaeffer RD, et al. Accurate prediction of protein structures and interactions using a three-track neural network. Science. 2021;373(6557):871–6.

Liebeschuetz JW, Cole JC, Korb O. Pose prediction and virtual screening performance of gold scoring functions in a standardized test. J Comput-Aided Mol Des. 2012;26:737–48.

Clark JJ, Orban ZJ, Carlson HA. Predicting binding sites from unbound versus bound protein structures. Sci Rep. 2020;10(1):15856.

Akdel M, Pires DE, Pardo EP, Jänes J, Zalevsky AO, Mészáros B, Bryant P, Good LL, Laskowski RA, Pozzati G, et al. A structural biology community assessment of alphafold2 applications. Nat Struct Mol Biol. 2022;29(11):1056–67.

Laskowski RA, Watson JD, Thornton JM. Protein function prediction using local 3d templates. J Mol Biol. 2005;351(3):614–26.

Zhang Z, Chen L, Zhong F, Wang D, Jiang J, Zhang S, Jiang H, Zheng M, Li X. Graph neural network approaches for drug-target interactions. Curr Opin Struct Biol. 2022;73: 102327.

Jiménez J, Skalic M, Martinez-Rosell G, De Fabritiis G. K deep: protein-ligand absolute binding affinity prediction via 3d-convolutional neural networks. J Chem Inform Model. 2018;58(2):287–96.

Tsubaki M, Tomii K, Sese J. Compound-protein interaction prediction with end-to-end learning of neural networks for graphs and sequences. Bioinformatics. 2019;35(2):309–18.

Lim J, Ryu S, Park K, Choe YJ, Ham J, Kim WY. Predicting drug-target interaction using a novel graph neural network with 3d structure-embedded graph representation. J Chem Inform Model. 2019;59(9):3981–8.

Li S, Zhou J, Xu T, Huang L, Wang F, Xiong H, Huang W, Dou D, Xiong H. In: Proceedings of the 27th ACM SIGKDD conference on knowledge discovery & data mining 2021: 975–985

Aykent S, Xia T. In: Proceedings of the 28th ACM SIGKDD conference on knowledge discovery and data mining 2022: 4–14

Stärk H, Ganea O, Pattanaik L, Barzilay R, Jaakkola T. In: International conference on machine learning (PMLR, 2022), pp. 20503–20521

Zhou G, Gao Z, Ding Q, Zheng H, Xu H, Wei Z, Zhang L, Ke G. Uni-mol: a universal 3d molecular representation learning framework 2023.

da Silva Rocha SF, Olanda CG, Fokoue HH, Sant’Anna CM. Virtual screening techniques in drug discovery: review and recent applications. Curr Top Med Chem. 2019;19(19):1751–67.

Li Y, Qiao G, Gao X, Wang G. Supervised graph co-contrastive learning for drug-target interaction prediction. Bioinformatics. 2022;38(10):2847–54.

Wu J, Lv X, Jiang S. In: Advances in intelligent automation and soft computing (Springer, 2022), pp. 376–383

Gilson MK, Liu T, Baitaluk M, Nicola G, Hwang L, Chong J. Bindingdb in 2015: a public database for medicinal chemistry, computational chemistry and systems pharmacology. Nucleic Acids Res. 2016;44(D1):D1045–53.

Schomburg I, Chang A, Ebeling C, Gremse M, Heldt C, Huhn G. Brenda, the enzyme database: updates and major new developments. Nucleic acids Res. 2004;32(1):D431–3.

Gaulton A, Hersey A, Nowotka M, Bento AP, Chambers J, Mendez D, Mutowo P, Atkinson F, Bellis LJ, Cibrián-Uhalte E, et al. The Chembl database in 2017. Nucleic Acids Res. 2017;45(D1):D945–54.

Wishart DS, Feunang YD, Guo AC, Lo EJ, Sajed T, Johnson D, Li C, Sayeeda Z, Sayeeda Z, et al. Drugbank 5.0: a major update to the drugbank database for 2018. Nucleic acids Res. 2018;46(D1):D1074–82.

Günther S, Kuhn M, Dunkel M, Campillos M, Senger C, Petsalaki E, Ahmed J, Urdiales EG, Gewiess A. Supertarget and matador: resources for exploring drug-target relationships. Nucleic Acids Res. 2007;36(1):D919–22.

Wang R, Fang X, Lu Y, Yang CY, Wang S. The pdbbind database: methodologies and updates. J Med Chem. 2005;48(12):4111–9.

Kim S, Thiessen PA, Bolton EE, Chen J, Fu G, Gindulyte A, Han L, He J, He S, Shoemaker BA, et al. Pubchem substance and compound databases. Nucleic Acids Res. 2016;44(D1):D1202–13.

Kuhn M, Szklarczyk D, Pletscher-Frankild S, Blicher TH, Von Mering C, Jensen LJ, Bork P. Stitch 4: integration of protein-chemical interactions with user data. Nucleic Acids Res. 2014;42(D1):D401–7.

Günther S, Kuhn M, Dunkel M, Campillos M, Senger C, Petsalaki E, Ahmed J, Urdiales EG, Gewiess A, Jensen LJ, et al. Supertarget and matador: resources for exploring drug-target relationships. Nucleic Acids Res. 2007;36(1):D919–22.

Magariños MP, Carmona SJ, Crowther GJ, Ralph SA, Roos DS, Shanmugam D, Van Voorhis WC, Agüero F. Tdr targets: a chemogenomics resource for neglected diseases. Nucleic Acids Res. 2012;40(D1):D1118–27.

Chen X, Ji ZL, Chen YZ. Ttd: therapeutic target database. Nucleic Acids Res. 2002;30(1):412–5.

Tang J, Szwajda A, Shakyawar S, Xu T, Hintsanen P, Wennerberg K, Aittokallio T. Making sense of large-scale kinase inhibitor bioactivity data sets: a comparative and integrative analysis. J Chem Inform Model. 2014;54(3):735–43.

Ezzat A, Wu M, Li XL, Kwoh CK. Drug-target interaction prediction via class imbalance-aware ensemble learning. BMC Bioinform. 2016;17(19):267–76.

Samara KA, Al Aghbari Z, Abusafia A. Glimpse: a glioblastoma prognostication model using ensemble learning’a surveillance, epidemiology, and end results study. Health Inform Sci Syst. 2021;9:1–13.

Hu L, Fu C, Ren Z, Cai Y, Yang J, Xu S, Xu W, Tang D. Sselm-neg: spherical search-based extreme learning machine for drug-target interaction prediction. BMC Bioinform. 2023;24(1):38.

Download references

Acknowledgements

This work was supported by the National Natural Science Foundation of China (No. 62376065), Natural Science Foundation of Guangdong (No. 2022A1515010102) and Joint Research Fund of Guangzhou and University (No. 2024A03J0323).

Author information

Wen Shi and Hong Yang have contributed equally to this work.

Authors and Affiliations

Cyberspace Institute of Advanced Technology, Guangzhou University, Guangzhou, 510006, China

Wen Shi, Hong Yang & Xiao-Xia Yin

State Key Laboratory of Proteomics, National Center for Protein Sciences (Beijing), Beijing, 102206, China

School of Computer Science and Technology, Zhejiang Normal University, Jinhua, 321004, China

Wen Shi & Yanchun Zhang

Department of New Networks, Peng Cheng Laboratory, Shenzhen, 518000, China

Yanchun Zhang

You can also search for this author in PubMed   Google Scholar

Corresponding authors

Correspondence to Linhai Xie or Yanchun Zhang .

Ethics declarations

Conflict of interest.

The authors have no Conflict of interest to declare that are relevant to the content of this article.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Shi, W., Yang, H., Xie, L. et al. A review of machine learning-based methods for predicting drug–target interactions. Health Inf Sci Syst 12 , 30 (2024). https://doi.org/10.1007/s13755-024-00287-6

Download citation

Received : 11 September 2023

Accepted : 04 March 2024

Published : 12 April 2024

DOI : https://doi.org/10.1007/s13755-024-00287-6

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Drug–target interactions
• Machine learning
• Data representation
• Deep neural network models
• Find a journal
• Publish with us
• Track your research

• Share full article

Advertisement

Supported by

A Conversation With …

Teen Drug Use Habits Are Changing, For the Good. With Caveats.

Dr. Nora Volkow, who leads the National Institutes of Drug Abuse, would like the public to know things are getting better. Mostly.

By Matt Richtel

Historically speaking, it’s not a bad time to be the liver of a teenager. Or the lungs.

Regular use of alcohol, tobacco and drugs among high school students has been on a long downward trend.

In 2023, 46 percent of seniors said that they’d had a drink in the year before being interviewed; that is a precipitous drop from 88 percent in 1979, when the behavior peaked, according to the annual Monitoring the Future survey, a closely watched national poll of youth substance use. A similar downward trend was observed among eighth and 10th graders, and for those three age groups when it came to cigarette smoking. In 2023, just 15 percent of seniors said that they had smoked a cigarette in their life, down from a peak of 76 percent in 1977 .

Illicit drug use among teens has remained low and fairly steady for the past three decades, with some notable declines during the Covid-19 pandemic.

In 2023, 29 percent of high school seniors reported using marijuana in the previous year — down from 37 percent in 2017, and from a peak of 51 percent in 1979.

There are some sobering caveats to the good news. One is that teen overdose deaths have sharply risen, with fentanyl-involved deaths among adolescents doubling from 2019 to 2020 and remaining at that level in the subsequent years.

Dr. Nora Volkow has devoted her career to studying use of drugs and alcohol. She has been the director of the National Institute on Drug Abuse since 2003. She sat down with The New York Times to discuss changing patterns and the reasons behind shifting drug-use trends.

What’s the big picture on teens and drug use?

People don’t really realize that among young people, particularly teenagers, the rate of drug use is at the lowest risk that we have seen in decades. And that’s worth saying, too, for legal alcohol and tobacco.

What do you credit for the change?

One major factor is education and prevention campaigns. Certainly, the prevention campaign for cigarette smoking has been one of the most effective we’ve ever seen.

Some of the policies that were implemented also significantly helped, not just making the legal age for alcohol and tobacco 21 years, but enforcing those laws. Then you stop the progression from drugs that are more accessible, like tobacco and alcohol, to the illicit ones. And teenagers don’t get exposed to advertisements of legal drugs like they did in the past. All of these policies and interventions have had a downstream impact on the use of illicit drugs.

Does social media use among teens play a role?

Absolutely. Social media has shifted the opportunity of being in the physical space with other teenagers. That reduces the likelihood that they will take drugs. And this became dramatically evident when they closed schools because of Covid-19. You saw a big jump downward in the prevalence of use of many substances during the pandemic. That might be because teenagers could not be with one another.

The issue that’s interesting is that despite the fact schools are back, the prevalence of substance use has not gone up to the prepandemic period. It has remained stable or continued to go down. It was a big jump downward, a shift, and some drug use trends continue to slowly go down.

Is there any thought that the stimulation that comes from using a digital device may satisfy some of the same neurochemical experiences of drugs, or provide some of the escapism?

Yes, that’s possible. There has been a shift in the types of reinforcers available to teenagers. It’s not just social media, it’s video gaming, for example. Video gaming can be very reinforcing, and you can produce patterns of compulsive use. So, you are shifting one reinforcer, one way of escaping, with another one. That may be another factor.

Is it too simplistic to see the decline in drug use as a good news story?

If you look at it in an objective way, yes, it’s very good news. Why? Because we know that the earlier you are using these drugs, the greater the risk of becoming addicted to them. It lowers the risk these drugs will interfere with your mental health, your general health, your ability to complete an education and your future job opportunities. That is absolutely good news.

But we don’t want to become complacent.

The supply of drugs is more dangerous, leading to an increase in overdose deaths. We’re not exaggerating. I mean, taking one of these drugs can kill you.

What about vaping? It has been falling, but use is still considerably higher than for cigarettes: In 2021, about a quarter of high school seniors said that they had vaped nicotine in the preceding year . Why would teens resist cigarettes and flock to vaping?

Most of the toxicity associated with tobacco has been ascribed to the burning of the leaf. The burning of that tobacco was responsible for cancer and for most of the other adverse effects, even though nicotine is the addictive element.

What we’ve come to understand is that nicotine vaping has harms of its own, but this has not been as well understood as was the case with tobacco. The other aspect that made vaping so appealing to teenagers was that it was associated with all sorts of flavors — candy flavors. It was not until the F.D.A. made those flavors illegal that vaping became less accessible.

My argument would be there’s no reason we should be exposing teenagers to nicotine. Because nicotine is very, very addictive.

Anything else you want to add?

We also have all of this interest in cannabis and psychedelic drugs. And there’s a lot of interest in the idea that psychedelic drugs may have therapeutic benefits. To prevent these new trends in drug use among teens requires different strategies than those we’ve used for alcohol or nicotine.

For example, we can say that if you take drugs like alcohol or nicotine, that can lead to addiction. That’s supported by extensive research. But warning about addiction for drugs like cannabis and psychedelics may not be as effective.

While cannabis can also be addictive, it’s perhaps less so than nicotine or alcohol, and more research is needed in this area, especially on newer, higher-potency products. Psychedelics don’t usually lead to addiction, but they can produce adverse mental experiences that can put you at risk of psychosis.

Matt Richtel is a health and science reporter for The Times, based in Boulder, Colo. More about Matt Richtel

• OU Homepage
• The University of Oklahoma

OU Research Contributes to National Conversation on Neuropsychiatric Side Effects in Children Taking Asthma Drug

Study sheds light on effectiveness of black box warning for Singulair

TULSA, OKLA. – A University of Oklahoma study about a “black box warning” for the asthma drug Singulair continues to influence a national conversation about the medication and its reported neuropsychiatric side effects in children and adolescents. The U.S. Food and Drug Administration assigns black box warnings, sometimes called boxed warnings, as the highest safety-related consumer warning the organization assigns to medications, intended to bring attention to the risks of taking the medication.

In 1998, the FDA approved Singulair, known as montelukast in its generic form, for treating asthma and hay fever. It became a frequently prescribed drug for children struggling with asthma and allergies because it is available in a cherry-flavored chewable pill and may decrease the need for steroids or daily inhaler use. However, in the 26 years since its approval, concerned parents have sounded the alarm about dramatic and sometimes deadly mental health changes they have seen in their children, prompting the FDA to issue a black box warning in 2020.

In their study, researchers at OU-TU School of Community Medicine on the University of Oklahoma’s Tulsa campus sought to understand whether Singulair’s reported negative side effects – including depression, aggression and suicidal thoughts – decreased after the black box warning. Researchers analyzed adverse events reported to the FDA two years before and two years after the warning was issued. In children ages one to 10, reports about most harmful side effects decreased after the warning was issued. However, for youth ages 11 to 17, the outcome was mixed. Reports about side effects actually increased for five of the eight mental health symptoms. Overall, prescriptions for Singulair have decreased only slightly since the black box warning was issued.

In February, the New York attorney general cited the OU findings among other research in a letter to the FDA urging the agency to sound a new, louder alarm about the negative side effects of Singulair in children. Families, too, continue to push for more restrictions on the drug.

A limitation of the study is that the black box warning was issued at the beginning of the COVID-19 pandemic, which had its own effect on the mental health of youth. But for OU researchers, their study

underscores the importance of conversations between doctors and families, said the study’s lead author, Samer Abdelkader, D.O., a pediatrics resident physician in the OU-TU School of Community Medicine.

“As a clinician interested in public health, I hope we can maximize the intent of these warnings and mitigate potential negative impacts on our patients,” Abdelkader said. “I think we can enhance our patient care with better conversations about the benefits and risks of this medication and come to a more informed decision on whether this is the right treatment for each patient.”

Asthma is the most common chronic disease in children, and hay fever (also known as allergic rhinitis) affects the lives of one in five young people, according to the Centers for Disease Control and Prevention. Singulair treats the inflammation and airway swelling that can be dangerous in both conditions. The FDA approved the drug for treating asthma in children as young as 12 months old and for allergies in babies as young as six months old. It is prescribed to millions of children and adolescents each year.

Parents were the first to raise awareness about mental changes they saw in their children who were taking Singulair. Aggression, anxiety, depression, hyperactivity, sleep problems and suicidal thinking were among the symptoms, and several high-profile suicides brought further attention to the drug’s reported problems. Without parents’ advocacy, issues with the drug likely would not have come to light as soon, said Amy Hendrix-Dicken, a senior research assistant and co-author of the paper.

“Parents are the greatest partners in providing health care to children,” she said. “Parents know their kids; our health care providers see them for 15 minutes, and there’s only so much you can glean from a 15-minute visit. We encourage families and patients to speak up, and if they feel like their provider isn’t listening, to find someone who will.”

New information like the black box warning can take a long time to make its way into doctors’ offices large and small, and federal agencies may communicate potential drug problems in differing ways. But health care providers should take the initiative to learn everything they can about the drugs they prescribe, the study’s authors said.

“I no longer say that a side effect cannot be from a medication,” said study co-author and pharmacist Michelle Condren, a professor and vice chair of research in the OU Department of Pediatrics. “I may say, ‘I haven’t seen this side effect before’ or ‘Let me look in the medical literature,’ but I’ll never say that a side effect isn’t possible. This has changed the way I communicate with patients and families. We never want to discount someone who is concerned that a medication could be causing a side effect, but to partner with them to figure things out.”

About the Project

The research team’s study was published in The Journal of Pediatric Pharmacology and Therapeutics and can be accessed at https://doi.org/10.5863/1551-6776-28.8.704.

About the University of Oklahoma

Founded in 1890, the University of Oklahoma is a public research university located in Norman, Oklahoma. As the state’s flagship university, OU serves the educational, cultural, economic and health care needs of the state, region and nation. OU was named the state’s highest-ranking university in  U.S. News & World Report’s  most recent Best Colleges list .  For more information about the university, visit  ou.edu .

More Research News

Researchers discover cell ‘crosstalk’ that triggers cancer cachexia.

New research from the University of Oklahoma reveals a previously unknown chain of events sparking the development of cancer cachexia, a debilitating muscle-wasting condition that almost always occurs in people diagnosed with pancreatic cancer. The research, led by Min Li, Ph.D., a professor in the OU College of Medicine, is published in the journal Cancer Cell.

OU Researcher Receives $3.1M Grant for Clean Hydrogen Technologies

Hanping Ding, Ph.D., an assistant professor in the School of Aerospace and Mechanical Engineering at the University of Oklahoma, has been awarded a $3.1 million grant from the Hydrogen and Fuel Cell Technologies Office in the Department of Energy through the Bipartisan Infrastructure Law to further research in clean hydrogen production. The funding is part of a $750 million effort in President Biden’s Investing in American agenda.

Genetic Testing May Provide Improved Medication Therapy Safety and Efficacy for Pediatric Cystic Fibrosis Patients

Second-year University of Oklahoma pediatric resident Dr. Caroline Thompson has a professional and personal connection to cystic fibrosis patient care for children. A childhood friend had the disease and made a lasting impression on Thompson. Now, Thompson has received the Cystic Fibrosis Foundation’s Medical Resident Research Award for a pilot study evaluating pharmacogenomic-directed therapy for pediatric patients at the Oklahoma Cystic Fibrosis Center Tulsa.

More OU News

• Accessibility
• Sustainability
• OU Job Search
• Legal Notices
• Resources and Offices
• OU Report It!

Many Cancer Drugs Still Unproven 5 Years After Accelerated Approval

By Robin Foster HealthDay Reporter

MONDAY, April 8, 2024 (HealthDay News) -- New research questions the effectiveness of the U.S. Food and Drug Administration's accelerated drug approval program after finding that many cancer drugs remain unproven five years later.

The study, published Sunday in the Journal of the American Medical Association and presented simultaneously at the American Association of Cancer Research's annual meeting in San Diego, found that 46 cancer drugs were granted accelerated approval between 2013 and 2017.

Of those, 41% showed no benefit after five years of follow-up. And of the 63% that were converted to regular approval, less than half (43%) demonstrated any clinical benefit in confirmatory trials.

“Five years after the initial accelerated approval, you should have a definitive answer,” Dr. Ezekiel Emanuel , a cancer specialist and bioethicist at the University of Pennsylvania who was not involved in the study, told the Associated Press . “Thousands of people are getting those drugs. That seems a mistake if we don’t know whether they work or not.”

U.S. Cities With the Most Homelessness

First created in 1992 to get new HIV drugs to desperate patients as quickly as possible, more than 80% of the program's accelerated approvals now go to cancer drugs, researchers found.

The program allows the FDA to grant early approval to drugs that show promising results for treating debilitating or fatal diseases. In exchange, drug companies are expected to do rigorous testing and produce better evidence before gaining full approval.

When it comes to withdrawing disappointing drugs from the market, the FDA or the drug company makes that call. Sometimes, the FDA has decided that less definitive evidence is good enough for a full approval, the AP reported.

Meanwhile, it’s not clear how much cancer patients understand about drugs with accelerated approval, study co-author Dr. Edward Scheffer Cliff , of Harvard Medical School, told the AP .

“We raise the question: Is that uncertainty being conveyed to patients?” Cliff said.

Still, drugs that got accelerated approval may be the only option for patients with rare or advanced cancers, Dr. Jennifer Litton , of MD Anderson Cancer Center in Houston, told the AP .

It’s important for doctors to carefully explain the evidence, Litton said.

“It might be shrinking of tumor. It might be how long the tumor stays stable,” Litton told the AP . “You can provide the data you have, but you shouldn’t overpromise.”

The accelerated approval program recently got some updates from Congress that gave the FDA more authority and streamlined the process for withdrawing drugs when companies don’t meet their commitments, the AP reported.

Those changes allow the agency “to withdraw approval for a drug approved under accelerated approval, when appropriate, more quickly,” FDA spokesperson Cherie Duvall-Jones told the AP .

More information

The FDA has more on the its accelerated approval program .

SOURCES: Journal of the American Medical Association , April 7, 2024; Associated Press

Copyright © 2024 HealthDay . All rights reserved.

Join the Conversation

Tags: cancer , FDA

America 2024

Health News Bulletin

Stay informed on the latest news on health and COVID-19 from the editors at U.S. News & World Report.

Sign in to manage your newsletters »

Sign up to receive the latest updates from U.S News & World Report and our trusted partners and sponsors. By clicking submit, you are agreeing to our Terms and Conditions & Privacy Policy .

You May Also Like

The 10 worst presidents.

U.S. News Staff Feb. 23, 2024

Cartoons on President Donald Trump

Feb. 1, 2017, at 1:24 p.m.

Photos: Obama Behind the Scenes

April 8, 2022

Photos: Who Supports Joe Biden?

March 11, 2020

U.S.: Threat From Iran ‘Very Credible’

Cecelia Smith-Schoenwalder April 12, 2024

Inflation Up, Consumer Sentiment Steady

Tim Smart April 12, 2024

House GOP Hands Johnson a Win

Aneeta Mathur-Ashton April 12, 2024

A Watershed Moment for America

Lauren Camera April 12, 2024

The Politically Charged Issue of EVs

Johnson Turns to Trump Amid GOP Dissent

Aneeta Mathur-Ashton April 11, 2024

An Ozempic baby boom? Some GLP-1 users report unexpected pregnancies.

Across social media, women who have used Ozempic or similar medications for diabetes or weight loss are reporting an unexpected side effect — surprise pregnancies.

The Facebook group “I got pregnant on Ozempic,” has more than 500 members. Numerous posts on Reddit and TikTok discuss unplanned pregnancies while on Ozempic and similar drugs which can spur significant weight loss by curbing appetite and slowing the digestive process. The drugs are known as “Glucagon-like peptide 1” or GLP-1 drugs.

The reports of an Ozempic baby boom are anecdotal, and it’s not known how widespread the phenomenon is. Experts say significant weight loss can affect fertility. Others speculate that the GLP-1 drugs could interfere with the absorption of oral contraceptives, causing birth control failures.

“I got pregnant on a GLP-1,” posted Deb Oliviara, 32, on her @Dkalsolive TikTok account, which has 36,000 followers. She had noted in another video that she’d previously suffered two miscarriages and a stillbirth.

Oliviara, who lives in Michigan, said in a direct message that she had been using Ozempic for three months before getting pregnant. “I was three weeks along when I found out,” Oliviara said. “I am now 3 months pregnant, and baby is doing amazing.”

“My little Mounjaro baby is almost 6 months old after trying for over 10 years with PCOS!” another woman commented on the post, referring to polycystic ovary syndrome , a hormonal health condition that is a leading cause of infertility.

Paige Burnham, 29, who lives in Louisville, had lost about 80 pounds while using Ozempic, also known as semaglutide, for Type 2 diabetes when she began feeling nauseous on a trip to Disney World. She assumed the symptom was due to the drug. “My most typical Ozempic side effect was nausea,” she said.

But she learned the symptom was actually morning sickness due to pregnancy — a surprise since she and her partner had tried for four years to conceive. She stopped taking Ozempic and gave birth to a healthy baby boy, Creed, in March 2023.

A lack of research on pregnancy and GLP-1 drugs

Little is known about the effects of Ozempic and similar drugs on women who want to get pregnant or who become pregnant while taking the drugs because they were specifically excluded from early clinical trials of the drug.

A spokesman for Novo Nordisk, which makes Ozempic and Wegovy, said the company is collecting data to evaluate the safety of becoming pregnant while using Wegovy, the version of semaglutide approved for weight loss.

“Pregnancy or intention to become pregnant were exclusion criteria in our trials with semaglutide in both obesity and type 2 diabetes,” the company said in a statement.

Get Well+Being tips straight to your inbox

Eli Lilly, maker of the GLP-1 drugs Mounjaro and Zepbound, did not respond to requests for comment.

The biggest concern among women who become pregnant using a GLP-1 is whether the drug poses a risk to the fetus. While women like Burnham and Oliviara have posted reassuring stories of delivering healthy babies, doctors say it’s important to use backup birth control and stop the drug immediately if you become pregnant.

A Novo Nordisk spokesman said in a statement that there isn’t enough available data to know if the drug poses a risk for birth defects, miscarriage or other adverse events related to pregnancy. Based on animal reproduction studies for Wegovy, the company said there “may be potential risks to the fetus from exposure to semaglutide during pregnancy.”

The company recommends stopping Wegovy at least two months before a planned pregnancy.

According to Ozempic’s prescribing information , pregnant rats administered Ozempic showed fetal structural abnormalities, fetal growth problems and embryonic mortality. In rabbits and cynomolgus monkeys, there were early pregnancy losses or structural abnormalities as well as marked maternal body weight loss.

Controlling diabetes is important for a healthy pregnancy, and experts say patients taking Ozempic for diabetes should discuss the risks and benefits with their doctor.

Why drugs like Ozempic might affect pregnancy risk

While it’s unclear whether women taking a GLP-1 have a higher risk of unplanned pregnancies, doctors say there are a few explanations why some women are getting pregnant while using the drugs.

Weight loss can have an effect on ovulation and fertility, said Lora Shahine , a reproductive endocrinologist with a fertility practice in Seattle and Bellevue, Wash.

“I think that with weight loss and balancing of hormones and improved insulin resistance, the hormonal access clicks back in, and all of a sudden they start ovulating again — they might not have been ovulating for years,” said Shahine, who is also an associate clinical professor at the University of Washington.

Stephanie Fein , an internist in Los Angeles who specializes in helping women lose weight for their fertility, said that losing just 5 to 10 percent of body weight can help someone conceive. “No one knows exactly the reason,” she said. “Fat is hormonally active. We know it has effects on estrogen, and it will impact ovulation and possibly egg development.”

The drugs also may interfere with oral contraceptives in some patients, doctors say. The GLP-1 drugs help people lose weight by slowing gastric emptying, curbing hunger and leaving people feel full sooner. It may be that the GLP-1 drugs also affect the absorption of oral contraceptives, said William Dietz, physician and chair of the STOP Obesity Alliance at the Milken Institute School of Public Health at George Washington University. “This may mean that birth control medications are metabolized or ineffective,” he said.

Dietz said most experts recommend discontinuing GLP-1 medications when pregnancy is detected. “I don’t think we know the impact of these drugs on fetal development,” he added.

Shahine recommends that women using oral contraceptives who are taking a GLP-1 drug use a second form of birth control. The drugs also aren’t recommended for mothers who are breastfeeding. Animal studies have shown semaglutide is present in the milk of lactating rats treated with the drug.

After Burnham stopped breastfeeding, she resumed taking Ozempic. Because of her past struggles with infertility, she doesn’t want to take birth control, although she said she is concerned about getting pregnant too soon. “I’m not ready yet,” she said.

Amy Klein is the author of “The Trying Game: Get Through Fertility Treatment and Get Pregnant without Losing Your Mind.”

Read more from Well+Being

Well+Being shares news and advice for living well every day. Sign up for our newsletter to get tips directly in your inbox.

Are you taking your meds wrong ? Many patients make these common mistakes.

Centenarians give their advice about everything.

The wall sit is a simple exercise that can lower your blood pressure.

Tart cherries — more specifically, tart cherry juice — may help with inflammation and pain.

Do you self-sabotage ? Here’s how to stop.