multidisciplinary breast cancer case study

• - 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
• Effects of...

Effects of multidisciplinary team working on breast cancer survival: retrospective, comparative, interventional cohort study of 13 722 women

• Related content
• Peer review
• Eileen M Kesson , project manager 1 4 ,
• Gwen M Allardice , statistician 1 4 ,
• W David George , school of medicine honorary professor 2 ,
• Harry J G Burns , chief medical officer for Scotland 3 ,
• David S Morrison , director 4
• 1 NHS Greater Glasgow and Clyde, West House, Gartnavel Royal Hospital, Glasgow, UK
• 2 Western Infirmary, Glasgow
• 3 Scottish Government, Public Health Division, Edinburgh, UK
• 4 West of Scotland Cancer Surveillance Unit, Institute of Health and Wellbeing, University of Glasgow, Glasgow G12 8RZ
• Correspondence to: E M Kesson Eileen.Kesson{at}nhs.net
• Accepted 16 March 2012

Objectives To describe the effect of multidisciplinary care on survival in women treated for breast cancer.

Design Retrospective, comparative, non-randomised, interventional cohort study.

Setting NHS hospitals, health boards in the west of Scotland, UK.

Participants 14 358 patients diagnosed with symptomatic invasive breast cancer between 1990 and 2000, residing in health board areas in the west of Scotland. 13 722 (95.6%) patients were eligible (excluding 16 diagnoses of inflammatory cancers and 620 diagnoses of breast cancer at death).

Intervention In 1995, multidisciplinary team working was introduced in hospitals throughout one health board area (Greater Glasgow; intervention area), but not in other health board areas in the west of Scotland (non-intervention area).

Main outcome measures Breast cancer specific mortality and all cause mortality.

Results Before the introduction of multidisciplinary care (analysed time period January 1990 to September 1995), breast cancer mortality was 11% higher in the intervention area than in the non-intervention area (hazard ratio adjusted for year of incidence, age at diagnosis, and deprivation, 1.11; 95% confidence interval 1.00 to 1.20). After multidisciplinary care was introduced (time period October 1995 to December 2000), breast cancer mortality was 18% lower in the intervention area than in the non-intervention area (0.82, 0.74 to 0.91). All cause mortality did not differ significantly between populations in the earlier period, but was 11% lower in the intervention area than in the non-interventional area in the later period (0.89, 0.82 to 0.97). Interrupted time series analyses showed a significant improvement in breast cancer survival in the intervention area in 1996, compared with the expected survival in the same year had the pre-intervention trend continued (P=0.004). This improvement was maintained after the intervention was introduced.

Conclusion Introduction of multidisciplinary care was associated with improved survival and reduced variation in survival among hospitals. Further analysis of clinical audit data for multidisciplinary care could identify which aspects of care are most associated with survival benefits.

Introduction

Breast cancer is the commonest cancer in women worldwide. 1 Its incidence has increased over several decades, while the death rate has fallen because of improved survival. 2 3 Survival has probably improved because of a mixture of earlier detection of the disease (both through screening and earlier symptomatic presentation), 4 biological changes that have made the disease more susceptible to hormonal therapy, 5 6 and improved treatment. 7

Treatment of cancer has increasingly been provided within centralised, specialist multidisciplinary teams in Europe, the United States, and Australia. 8 9 10 11 12 These teams were established after observational evidence identified better outcomes among patients treated by specialists for various common cancers 13 , such as an 11-17% reduction in risk of death among women treated for breast cancer by specialist surgeons. 14 However, it remains unclear whether multidisciplinary care improves cancer survival and whether the costs justify clinical benefits. 15 16 17 18 Although attempts have been made to estimate the costs of multidisciplinary teams, 15 we are not aware of comparisons with costs for conventional care.

Although some evidence has shown improved outcomes in cancer associated with the introduction of multidisciplinary care, 19 20 21 a recent systematic review of 21 studies (including five on breast cancer) could not determine a causal relation between such care and cancer survival for three main reasons. 18 Firstly, the definition of multidisciplinary care was imprecise and heterogeneous. Secondly, the outcomes of multidisciplinary care could have indicated selection biases towards patients with more favourable prognostic features rather than true treatment effects. Thirdly, no adjustment was made for the confounding effects of temporal improvements in survival that might have occurred without the introduction of multidisciplinary care. For example, none of the identified studies included a contemporaneous comparison group. 18

In the early 1990s, variations in survival among hospitals managed by the Greater Glasgow health board were reported to the director of public health (HJGB). The hospital with highest survival had delivered care in a multidisciplinary manner. As a result, multidisciplinary teams were formally established in October 1995 throughout the Greater Glasgow health board area, with the aim of improving survival and minimising variations among hospitals. No substantial reorganisation of breast cancer care took place in areas managed by other health boards in the west of Scotland until 2000, when national guidance was introduced. 22

We aimed to determine whether the introduction of multidisciplinary care affected the survival of women with breast cancer. Multidisciplinary teams were introduced to Greater Glasgow health board as an NHS Scotland area-wide policy to improve the quality of care. Therefore, an experimental study design, such as the randomisation of patients to hospitals providing or not providing multidisciplinary care, would not be possible. We compared an intervention area that had introduced multidisciplinary care with neighbouring areas (without such care), before and after the organisational change. We addressed the limitations of previous studies, 18 by defining the composition of a multidisciplinary team at the outset, observing the effect of care where no other model was available, and including a contemporaneous comparison population to adjust for temporal improvements in survival.

Design and patients

We did a retrospective, comparative, non-randomised, interventional cohort study. From the Scottish Cancer Registry, we selected data for all incident, symptomatic invasive breast cancers (ICD-10, international classification of diseases (ICD), 10th revision, codes C50·1 to C50·9) diagnosed between 1 January 1990 and 31 December 2000 for residents of health board areas in the west of Scotland. We obtained data from the Scottish Cancer Registry, which is a comprehensive account of all incident cancers gathered from pathology, hospital, and death records, among other sources. The registry is linked, by probability matching, to death records from the General Register Office for Scotland and has a high level of accuracy. 23 24 Patients were grouped according to the health board of diagnosis. The Greater Glasgow health board area (population 900 000) was the intervention area, and areas managed by the remaining health boards in the west of Scotland (1 600 000) comprised the non-intervention area.

Data for tumour size was only available in the Scottish Cancer Registry from 1997 onwards. If data were available, we grouped tumour size. 14 We calculated age at diagnosis for all patients, and used four age groups: younger than 50 years, which is considered to be a proxy for premenopausal status 25 ; 50-64 years, the age band at which patients would be invited for breast screening; 65-79 years, elderly and post screening age; and 80 years and older. The oldest age group has been shown to have survival equal to that of younger women after treatment of breast cancer 26 ; the most appropriate management in this age group also remains unclear. 27

We included patients if they were diagnosed with breast cancer but had no other primary cancers, were resident in and attending an NHS hospital managed by one of the health boards in the west of Scotland, and were not diagnosed by the National Breast Screening Programme. Tumours detected by screening programmes are usually at earlier stages with better prognoses. Although screen detected patients might have been diagnosed by the programme throughout the west of Scotland, they were treated mainly in two hospitals managed by the Greater Glasgow health board.

Multidisciplinary team intervention

In 1995, multidisciplinary team care for all patients with breast cancer was introduced in the Greater Glasgow health board area (intervention area), but not in the areas managed by other health boards in the west of Scotland (non-intervention area). Five multidisciplinary teams were formed in the intervention area, each led by a specialist surgeon performing in excess of 50 operations for invasive breast cancer each year. In 1995, a multidisciplinary team was defined as a group with the following characteristics:

Comprised specialist breast cancer surgeons, pathologists, oncologists, radiologists, and specialist nurses

Worked to evidence based guidelines, written by the specialist breast surgeon whose patients had the highest survival rates before reorganisation of care delivery (WDG)

Held weekly formal meetings to discuss results and agree on adjuvant treatment for individual patients

Audited clinical activity and reported results at regular intervals

Lead clinicians from each team met regularly with the director of public health (HJGB) to discuss audit results throughout the area, with the aim to minimise deviations from guidelines and variations in practice and improve quality of care.

Doctors operating on small numbers outside the five main centres were no longer allowed to manage patients needing care for breast cancer. As a result, there was a degree of centralisation of services. However, the geographical location of the hospitals in the intervention area was such that patients did not have to travel long distances to receive care.

No formal change in delivery of care was introduced in the non-intervention area; patients continued to receive care along more traditional approaches, with surgeons making unilateral decisions regarding surgery and adjuvant treatment. In 1999, Scottish guidance was issued on the reorganisation of cancer care. 22 In 2001-02, services throughout the country began to form managed clinical networks, of which multidisciplinary care was a component. Thus, there was unlikely to be a clear difference in the use of multidisciplinary care within the west of Scotland NHS area after 2000.

Socioeconomic circumstances

Socioeconomic status was inferred for all patients on the basis of their area of residence at the time of diagnosis, using a deprivation category score, which is a validated seven category ordinal score that ranks all postcode sectors from 1 (most affluent) to 7 (most deprived), using four census variables that have been shown to best correlate with health outcomes: car ownership, male unemployment, occupational social classes IV and V (semi-skilled and unskilled occupations, respectively), and overcrowding. 28 We grouped scores further into three conventional categories: affluent (scores 1 and 2), intermediate (3 to 5), and deprived (6 and 7). 14 Data from the 1991 census showed that 51% of people in the intervention area lived in areas classified as most deprived (scores 6 and 7), compared with 18% of residents in the non-intervention area. 29

Statistical methods

We used univariate analyses to examine differences in survival trends between the intervention and non-intervention areas for two time periods, from January 1990 to September 1995 and from October 1995 to December 2000. Where appropriate, we used χ 2 and t tests to calculate any significant differences in age, deprivation, and tumour size. If needed, we transformed data to a log scale to satisfy the assumption for the t test.

We defined survival as time from incidence of breast cancer to the earlier of date of death or date of censoring. Patients diagnosed in the earlier period were censored at 30 September 2002 and those diagnosed in the later period were censored at 31 December 2007, to remove any effects of differential follow-up times between the two periods. Deaths due to breast cancer were classified as patients with a main cause of death coded by the Scottish Cancer Registry as ICD-9 codes 174.1 to 174.9 or ICD-10 codes C50.1 to C50.9.

We did an analysis of an interrupted time series to investigate the effect of the intervention while adjusting for trends in survival before the intervention. 30 Our analysis did not include patients with a date of incidence in 1995, the year the intervention was introduced. We used survival at five year follow-up (after diagnosis) as the outcome measure.

We did multivariable analyses using Cox’s proportional hazards models 31 to examine differences in survival (adjusted for age, deprivation, and year of incidence) between the intervention and non-intervention areas for each time period. The analysis also calculated outcomes in relation to breast cancer mortality and all cause mortality.

We used funnel plots to compare five year survival between hospitals. Funnel plots are control charts in which the event of interest is plotted against a measure of its precision. 32 Limits for the funnel plots were based on Wilson’s formula for binomial confidence intervals. 33 We plotted breast cancer survival at five years for each hospital against the number of patients diagnosed with breast cancer in that hospital. To compare survival between hospitals, only one investigator (DSM) knew the code indicating which hospitals comprised the intervention and non-intervention areas. Hospitals with fewer than 25 patients diagnosed in each five year time period were excluded from the funnel plots. This exclusion avoided the range of the y axis scale being too large to distinguish differences between hospitals with larger volumes. We did all analyses using Stata, version 11 software.

Of 14 358 patients reviewed, we excluded 636; 16 patients were diagnosed with inflammatory cancer (ICD for Oncology, 2nd edition, code M8530/3; seven patients in the intervention area, nine in the non-intervention area), and 620 were diagnosed only at death (death certificate only; 182, 438). Of 13 722 women in the study population, 6050 (44.1%) were in the intervention area and 7672 (55.9%) in the non-intervention area. Table 1 ⇓ describes patient characteristics.

 Age and socioeconomic characteristics of 13<thin>722 female patients with breast cancer in the west of Scotland, 1990-2000. Data are no (%)

• View inline

Age ranged from 18 to 100 years (mean 62.9 years, standard deviation 14.9); t tests of transformed data showed that ages were similar between the intervention and non-intervention areas, both before and after introduction of the intervention. We found a significant difference in deprivation (P<0.001) between the intervention and non-intervention areas, both before and after the intervention. However, deprivation had not changed significantly in either area over time. In both time periods, more patients in the intervention area than in the non-intervention area lived in the most deprived areas. Data for tumour size was available from 1997 onwards only; available data showed a mean tumour size of 23.9 mm (standard deviation 14.4); t tests of transformed data showed that tumour size was similar in both the intervention and non-intervention areas.

By the end of the follow-up period (seven years from the end of each time period), 4844 (35.3%) study patients had died of breast cancer and 8301 (60.5%) had died overall. Breast cancer survival at five years in the time period before introduction of multidisciplinary care was 71.3% (n=2201) in the intervention area and 73.6% (n=2881) in the non-intervention area. These proportions rose to 79.2% (n=2346) and 75.9% (n=2853), respectively, in the time period after the intervention was introduced. Overall survival at five years in the earlier time period was 59.9% (n=1849) in the intervention area and 61.5% (n=2408) in the non-intervention area. These proportions rose to 65.6% (n=1942) and 63.8% (n=2400), respectively, in the later time period.

Our analysis using an interrupted time series was based on breast cancer survival at five years and omitted patients with an incident cancer in 1995 (fig 1 ⇓ ). The analysis showed a significant improvement (P=0.004) in survival in the intervention area in 1996, compared with the expected survival in that year had the pre-intervention trend continued—that is, had the intervention not been introduced. We saw no corresponding improvement in survival among patients in the non-intervention area (P=0.64). Overall survival also did not improve in 1996, in either the intervention or non-intervention area. The introduction of the multidisciplinary approach initially had a significant positive impact on breast cancer survival at five years, on incident cancers in 1996, and this impact was maintained (fig 1). We saw a similar pattern in overall survival at five years, but the changes were not significant.

Fig 1 Five year survival by year of incidence with trends before and after year 1995, by intervention area

• Download figure
• Open in new tab
• Download powerpoint

Table 2 ⇓ shows the multivariable survival analyses, using a Cox’s proportional hazards model, that was adjusted for age, deprivation, and year of incidence.

 Adjusted hazard ratios of mortality in the intervention area compared with the non-intervention area

Breast cancer mortality

Before multidisciplinary care was introduced in 1995, breast cancer mortality was higher in the intervention area than in the non-intervention area (hazard ratio 1.11, 95% confidence interval 1.00 to 1.20; table 2). After the intervention was introduced, mortality was significantly lower in the intervention area than in the non-intervention area (0.82, 0.74 to 0.91). Another Cox model showed a significant interaction between time period and intervention area (P<0.001), indicating that mortality differed significantly between the intervention and non-intervention areas over time. Subgroup analyses by age group showed that the effect of the intervention was strongest in patients aged 80 years and older (P=0.001), and significant also in patients aged 65-79 years (P=0.01). We saw no significant effects in either of the younger age groups.

All cause mortality

We saw no significant difference in all cause mortality between intervention and non-intervention areas in the earlier time period before multidisciplinary care was introduced (hazard ratio 1.06, 95% confidence interval 0.98 to 1.14; table 2). However, in the later time period, mortality was 11% lower in the intervention area than in the non-intervention area (0.89, 0.82 to 0.97). As with breast cancer mortality, the effect of the intervention on all cause mortality differed significantly in both the intervention and non-intervention areas (P=0.003). Subgroup analyses by age group showed a less clear pattern for all cause mortality than that seen for breast cancer mortality. However, the effect of the intervention was most significant in patients aged 80 years or older (P=0.002); this effect was not seen in any other age group.

Funnel plot analysis

We used funnel plots to compare breast cancer survival at five years among hospitals before and after the introduction of the intervention in 1995 (fig 2 ⇓ ). A separate funnel plot was created for each time period. In the earlier time period, we plotted breast cancer survival for eight hospitals in the intervention area, and 15 hospitals in the non-intervention area. The number of patients diagnosed per hospital ranged from 30 to 881 in the intervention area and from 26 to 682 in the non-intervention area. Overall mean survival at five year follow-up was 70.3% (range 36.7-76.8% in the intervention area, 65.4-86.4% in the non-intervention area). Three of eight hospitals in the intervention area were below the 95% confidence limit and two were above. None of the non-intervention hospitals was below the 95% confidence limit but two were above.

Fig 2 Funnel plot of breast cancer survival at five years in study hospitals, by time period. Hospitals diagnosing fewer than 25 patients in each time period were excluded (of which more than 75% diagnosed fewer than three cases of breast cancer)

In the later time period, we plotted breast cancer survival for five hospitals in the intervention area, and 10 hospitals in the non-intervention area. The number of patients diagnosed per hospital ranged from 293 to 1025 in the intervention area and from 155 to 775 in the non-intervention area. Overall mean survival at five years during this period (77.0%) was higher than that during the earlier time period, with less variation among hospitals (range 78.5-80.2% in the intervention area, 69.6-80.6% in the non-intervention area). All hospitals in the intervention area lay within the 95% confidence limit. One hospital in the non-intervention area was above the 95% confidence limit and three were below the limit.

Main findings

We found that the introduction of teams providing multidisciplinary care for the treatment of breast cancer was associated with 18% lower breast cancer mortality at five years and 11% lower all cause mortality at five years, compared with similar patients treated in neighbouring areas over the same time period. The introduction of this intervention was also associated with a reduction in the number of hospitals treating breast cancer, reduced variation in survival rates among hospitals, and the reorganisation of breast cancer care away from hospitals with the poorest survival outcomes. Our use of a contemporaneous comparison group suggests that the benefits of multidisciplinary care exceeded those that would otherwise have occurred owing to ongoing improvements in treatment of breast cancer.

The effect size was of a similar magnitude to that found by Gillis and Hole between specialist and non-specialist breast cancer surgeons. 34 Before the introduction of multidisciplinary teams in the present study, patients with breast cancer in the intervention area were treated by a mixture of specialist and non-specialist surgeons, so it might be expected that only modest improvements in survival would have been achieved by introducing surgery by specialists alone. Although evidence suggests that better local and regional treatment of disease by specialist surgeons explains some of their survival advantage, 35 the survival advantage of treatment within a multidisciplinary team is unlikely to be explained by surgical specialisation alone. Our results suggest an “empirical induction period” in the intervention area, during which survival continued to increase sharply between 1996 and 1997.

Strengths and weaknesses of the study

This non-randomised intervention study compared an experimental intervention with no intervention. Although it did not have the strength of a randomised controlled trial, that study design was no longer possible because of the prolific introduction of the multidisciplinary approach of care. The principal limitation of our study was that the observed differences in survival after the introduction of multidisciplinary care in the intervention area might have been due to selection biases or confounding factors.

Although a few patients in our study might have died outside the UK and had been assumed to be alive, we have no reason to believe that this was more likely to occur in the intervention area than the non-intervention area. Selection criteria were applied in the same way to all health board areas—therefore, differential selection of patients with less advanced disease in the intervention area after September 1995 seems unlikely. Indeed, where data for tumour size were available, patients in the intervention area had similar sizes of tumours to those in the non-intervention area.

We excluded all patients with screen detected cancers because they were treated almost exclusively within the intervention area, and such cancers tend to be less advanced and less aggressive. 36 It is possible that patients treated in the intervention area presented with less advanced disease after September 1995. However, we have no reason to suspect this and available data do not support this explanation.

Women from more affluent areas have consistently been shown to have better survival from breast cancer than those from less affluent areas, 37 38 but the improvement was observed even after adjusting for deprivation. Other concurrent changes in healthcare services, not attributable to the intervention, could have occurred after 1995. However, we are not aware of differences in availability of diagnostic or treatment services between health board areas in the west of Scotland in 1996-2000. No other major interventions, either local or national, were introduced during that time. We are not aware of any formal arrangements to introduce multidisciplinary care in the non-intervention area after 1995, but we do know that trainee surgeons in the intervention area in the early 1990s took up consultant posts in other parts of Scotland. Therefore, improvements in surgical practice could have occurred in the non-intervention area as a result of multidisciplinary working. Such an effect would attenuate rather than increase the true effects of multidisciplinary care, by introducing some misclassification of intervention and non-intervention areas.

Strengths and weaknesses in relation to other studies

Comparisons have been made between contemporaneous groups of patients given multidisciplinary or conventional treatment, within the same geographical area. 39 However, the multidisciplinary approach probably selects patients with different prognostic factors from the entire patient population, introducing potential selection biases. Patients treated by a multidisciplinary team are also likely to undergo more investigations and be subject to additional clinical audits, so that information about their health and outcomes might differ systematically from those of patients not in multidisciplinary care. These selection and information biases complicate investigations into the true effects of multidisciplinary working.

Our study probably did not have these biases, because the intervention was provided for all patients within a geographical area and we obtained the same data from the Scottish Cancer Registry for patients in both intervention and non-intervention areas. We defined the composition and functions of multidisciplinary teams and used the same selection criteria for both areas, thus directly avoiding weaknesses of previous attempts to determine the effects of multidisciplinary care. 18 Owing to the lack of a precise definition of multidisciplinary care, and the variation in provision of health services in countries outside the UK our results could be difficult to generalise.

Implications for doctors and policy makers

Multidisciplinary care probably improves patient outcomes by influencing various aspects of care. These factors include adherence to guidelines and nurse education, 40 increased surgical volume and experience, 41 and improved interdisciplinary working. 42 43 Kingsmore and colleagues reported that the survival benefits of treatment by a specialist surgeon in breast cancer could be explained by better local and regional treatment, as defined by the King’s Fund consensus statement. 35 44 Although multidisciplinary care is considered standard practice in many countries, access to such care still varies. 8 10 11 45 Our results support the universal provision of cancer care by specialist, multidisciplinary teams. Furthermore, significantly improved survival was achieved among patients aged 80 years and older, which could be an important explanatory element to support the introduction of multidisciplinary care for cancer.

Unanswered questions and future research

Further work is needed to understand how the multidisciplinary approach affects specific elements of breast cancer care, to further improve the effective management of patients. We found that the benefits of multidisciplinary working were greatest in older patients. Since the intervention guidelines were not age specific, they could have given surgeons and other team members more confidence to actively treat older patients rather than managing them conservatively.

We found that multidisciplinary team care for breast cancer was associated with substantial improvements in survival compared with those we might have expected to occur. The introduction of multidisciplinary care led by specialist breast surgeons performing more than 50 operations per year, resulted in a degree of centralisation of care. Therefore, multidisciplinary teams could increase efficiency by fostering more centralised hospital care. Owing to a lack of data, we were unable to explore the financial cost of multidisciplinary working in our analysis. There is clearly scope for research in this topic; doctors and policy makers alike would benefit from cost effectiveness studies of multidisciplinary working, which could help inform implementation of future programmes. Although multidisciplinary care for cancer has become standard practice in many countries, persisting variations in access to specialist care need to be minimised.

What is already known on this topic

Care by a specialist clinician is associated with improved cancer survival

Treatment of cancer has increasingly been provided within centralised, specialist multidisciplinary teams, but it remains unclear whether such care has contributed to improvements in cancer survival

What this study adds

Introduction of multidisciplinary care was associated with substantially greater improvements in breast cancer survival over and above those that were expected to occur had the approach not been introduced

Benefits of receiving multidisciplinary care for breast cancer versus conventional care are similar to those of receiving specialist care versus non-specialist care

Introduction of multidisciplinary care was associated with the removal of poorly performing hospitals and reduced variations in outcomes among hospitals.

Cite this as: BMJ 2012;344:e2718

We thank Billy Sloan of the West of Scotland Cancer Surveillance Unit for data linkage; the Information Services Division of NHS National Services Scotland for data collection; and the late David J Hole and Tim G Cooke for their contributions to earlier evaluations of breast cancer outcomes and to the development of the West of Scotland Managed Clinical Network for Breast Cancer.

Contributors: All authors contributed to the design and analysis of the study. EMK wrote the first draft of the manuscript. All authors contributed to the redrafting of the manuscript and approved the final version. EMK is the guarantor. EMK, GMA, and DSM had full access to the data in the study and can take responsibility for the integrity of the data and the accuracy of the data analysis. The other authors had full access to all the results.

Funding: This study was supported by NHS Scotland and University of Glasgow. They had no role in the study design; in the collection, analysis, and interpretation of data; in the writing of the report; and in the decision to submit the article for publication.

Competing interests: All authors have completed the Unified Competing interest form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare: the authors have support from NHS Greater Glasgow and Clyde health board (EMK, GMA), NHS Scotland (HJGB), and the University of Glasgow (WDG, DSM) for the submitted work; no financial relationships with any organisations that might have an interest in the submitted work in the previous 3 years; no other relationships or activities that could appear to have influenced the submitted work.

Ethical approval: The West of Scotland Cancer Surveillance Unit obtained permission to obtain cancer registry data both from Caldicott Guardians of all five health boards in the west of Scotland and from the information services division of the NHS in Scotland privacy advisory committee.

Data sharing: No additional data available.

This is an open-access article distributed under the terms of the Creative Commons Attribution Non-commercial License, which permits use, distribution, and reproduction in any medium, provided the original work is properly cited, the use is non commercial and is otherwise in compliance with the license. See: http://creativecommons.org/licenses/by-nc/2.0/ and http://creativecommons.org/licenses/by-nc/2.0/legalcode .

• ↵ Althuis MD, Dozier JM, Anderson WF, Devesa SS, Brinton LA. Global trends in breast cancer incidence and mortality 1973-1997. Int J Epidemiol April 2005 ; 34 : 405 -12. OpenUrl CrossRef
• ↵ Chu KC, Tarone RE, Kessler LG, Ries LAG, Hankey BF, Miller BA, et al. Recent trends in US breast cancer incidence, survival, and mortality rates. J Natl Cancer Inst 1996 ; 88 : 1571 -9. OpenUrl Abstract / FREE Full Text
• ↵ Botha JL, Bray F, Sankila R, Parkin DM. Breast cancer incidence and mortality trends in 16 European countries. Eur J Cancer 2003 ; 39 : 1718 -29. OpenUrl CrossRef PubMed Web of Science
• ↵ Thomson CS, Brewster DH, Dewar JA, Twelves CJ. Improvements in survival for women with breast cancer in Scotland between 1987 and 1993: impact of earlier diagnosis and changes in treatment. Eur J Cancer 2004 ; 40 : 743 -53. OpenUrl CrossRef PubMed Web of Science
• ↵ Brown SBF, Mallon EA, Edwards J, Campbell FM, McGlynn LM, Elsberger B, et al. Is the biology of breast cancer changing? A study of hormone receptor status 1984-1986 and 1996-1997. Br J Cancer 2009 ; 100 : 807 -10. OpenUrl CrossRef PubMed Web of Science
• ↵ Li CI, Daling JR, Malone KE. Incidence of invasive breast cancer by hormone receptor status from 1992 to 1998. J Clin Oncol 2003 ; 21 : 28 -34. OpenUrl Abstract / FREE Full Text
• ↵ Peto R, Boreham J, Clarke M, Davies C, Beral V. UK and USA breast cancer deaths down by 25% in year 2000 at ages 20-69 years. Lancet 2000 ; 355 : 1822 . OpenUrl PubMed Web of Science
• ↵ Atun R, Ogawa T, Martin-Moreno JM. Analysis of national cancer control programmes in Europe. 2009. http://spiral.imperial.ac.uk/bitstream/10044/1/4204/1/Cancer%20Control%20vf2.pdf .
• ↵ Stitzenberg K, Meropol N. Trends in centralization of cancer surgery. Ann Surg Oncol 2010 ; 17 : 2824 -31. OpenUrl CrossRef PubMed Web of Science
• ↵ Anderson BO, Kaufman CS, Keil KD, Carlson RW. Interdisciplinary coordination for breast cancer care: a rational approach to detection, diagnosis, and treatment. Dis Manag Health Outcomes 2008 ; 16 : 7 -11. OpenUrl CrossRef Web of Science
• ↵ Department of Health UK. Cancer reform strategy. 2007. www.dh.gov.uk/en/Publicationsandstatistics/Publications/PublicationsPolicyAndGuidance/DH_081006 .
• ↵ Clinical Oncological Society of Australia, Cancer Council Australia, and the National Cancer Control Initiative 2002. Optimising cancer care in Australia. 2003. www.canceraustralia.gov.au/sites/default/files/user-upload/publications/reports/optim_cancer_care.pdf .
• ↵ Selby P, Gillis C, Haward R. Benefits from specialised cancer care. Lancet 1996 ; 348 : 313 -8. OpenUrl CrossRef PubMed Web of Science
• ↵ Kingsmore D, Ssemwogerere A, Hole D, Gillis C. Specialisation and breast cancer survival in the screening era. Br J Cancer 2003 ; 88 : 1708 -12. OpenUrl CrossRef PubMed Web of Science
• ↵ Fosker CJ, Dodwell D. Multidisciplinary team working in cancer: what is the evidence? 2010. www.bmj.com/rapid-response/2011/11/02/cost-mdt .
• ↵ Fleissig A, Jenkins V, Catt S, Fallowfield L. Multidisciplinary teams in cancer care: are they effective in the UK? Lancet Oncol 2006 ; 7 : 935 -43. OpenUrl CrossRef PubMed Web of Science
• ↵ Taylor C, Munro AJ, Glydodwellnne-Jones R, Griffith C, Trevatt P, Richards M, et al. Multidisciplinary team working in cancer: what is the evidence? BMJ 2010 ; 340 : c951 . OpenUrl FREE Full Text
• ↵ Hong NJL, Wright F, Gagliardi A, Paszat L. Examining the potential relationship between multidisciplinary cancer care and patient survival: an international literature review. J Surg Oncol 2010 ; 102 : 125 -34. OpenUrl CrossRef PubMed Web of Science
• ↵ Houssami N, Sainsbury R. Breast cancer: multidisciplinary care and clinical outcomes. Eur J Cancer 2006 ; 42 : 2480 . OpenUrl CrossRef PubMed Web of Science
• ↵ Forrest LM, McMillan DC, McArdle CS, Dunlop DJ. An evaluation of the impact of a multidisciplinary team, in a single centre, on treatment and survival in patients with inoperable non-small-cell lung cancer. Br J Cancer 2005 ; 93 : 977 -8. OpenUrl CrossRef PubMed Web of Science
• ↵ Stephens MR, Lewis WG, Brewster AE, Lord I, Blackshaw GRJC, Hodzovic I, et al. Multidisciplinary team management is associated with improved outcomes after surgery for esophageal cancer. Dis Eosphagus 2006 ; 19 : 164 -71. OpenUrl CrossRef
• ↵ NHS Scotland. Introduction of managed clinical networks within the NHS in Scotland. 1999. www.sehd.scot.nhs.uk/mels/1999_10.htm .
• ↵ Kendrick S. The development of record linkage in Scotland: the responsive application of probability matching. In: Committee on Applied and Theoretical Statistics, National Research Council; Federal Committee on Statistical Methodology, Office of Management and Budget. Record linkage techniques—1997: proceedings of an International Workshop and Exposition. National Academies Press, 1999: 319-32. www.nap.edu/catalog.php?record_id=6491 .
• ↵ Kendrick SW, Clarke JA. The Scottish medical record linkage system. Health Bull (Edinburgh) 1979 ; 51 : 72 -9. OpenUrl
• ↵ Phipps AI, Ichikawa L, Bowles EJA, Carney PA, Kerlikowske K, Miglioretti DL, et al. Defining menopausal status in epidemiologic studies: a comparison of multiple approaches and their effects on breast cancer rates. Maturitas 2010 9;67:60-6.
• ↵ Laki F, Kirova Y, Savignoni A, Campana F, Levu B, Estve M, et al. Management of operable invasive breast cancer in women over the age of 70: long-term results of a large-scale single-institution experience. Ann Surg Oncol 2010 ; 17 : 1530 -8. OpenUrl CrossRef PubMed Web of Science
• ↵ Enger SM, Thwin SS, Buist DSM, Field T, Frost F, Geiger AM, et al. Breast cancer treatment of older women in integrated health care settings. J Clin Oncol 2006 ; 24 : 4377 -83. OpenUrl Abstract / FREE Full Text
• ↵ Carstairs V, Morris R. Deprivation and health in Scotland. Aberdeen University Press, 1991.
• ↵ McLaren GB, Bain M. Deprivation and health in Scotland: insights from NHS data. NHS Scotland,1998.
• ↵ Deasy C, Bray JE, Smith K, Wolfe R, Harriss LR, Bernard SA, et al. Cardiac arrest outcomes before and after the 2005 resuscitation guidelines implementation: evidence of improvement? Resuscitation 2011 ; 82 : 984 -8. OpenUrl CrossRef PubMed Web of Science
• ↵ Cox DR. Regression models and life tables. J R Stat Soc (B) 1972 ; 34 : 187 -220. OpenUrl
• ↵ Spiegelhalter DJ. Handling over-dispersion of performance indicators. Qual Saf Health Care 2005 ; 14 : 347 -51. OpenUrl Abstract / FREE Full Text
• ↵ Newcombe RG. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med 1998 ; 17 : 857 -72. OpenUrl CrossRef PubMed Web of Science
• ↵ Gillis CR, Hole DJ. Survival outcome of care by specialist surgeons in breast cancer: a study of 3786 patients in the west of Scotland. BMJ 1996 ; 312 : 145 -8. OpenUrl Abstract / FREE Full Text
• ↵ Kingsmore D, Hole D, Gillis C. Why does specialist treatment of breast cancer improve survival? The role of surgical management. Br J Cancer 2004 ; 90 : 1920 . OpenUrl CrossRef PubMed Web of Science
• ↵ Weaver DL, Rosenberg RD, Barlow WE, Ichikawa L, Carney PA, Kerlikowske K, et al. Pathologic findings from the Breast Cancer Surveillance Consortium. Cancer 2006 ; 106 : 732 -42. OpenUrl CrossRef PubMed Web of Science
• ↵ Thomson CS, Hole DJ, Twelves CJ, Brewster DH, Black RJ. Prognostic factors in women with breast cancer: distribution by socioeconomic status and effect on differences in survival. J Epidemiol Community Health 2001 ; 55 : 308 -15. OpenUrl Abstract / FREE Full Text
• ↵ Schrijvers CTM, Mackenbach JP, Lutz J-M, Quinn MJ, Coleman MP. Deprivation and survival from breast cancer. Br J Cancer 1995 ; 72 : 738 -43. OpenUrl PubMed Web of Science
• ↵ Friedland PL, Bozic B, Dewar J, Kuan R, Meyer C, Phillips M. Impact of multidisciplinary team management in head and neck cancer patients. Br J Cancer 2011 ; 104 : 1246 -8. OpenUrl CrossRef PubMed Web of Science
• ↵ McCarthy M, Datta P, Khachatryan A, Coleman MP, Rachet B. Would compliance with cancer care standards improve survival for breast, colorectal and lung cancers? J Epidemiol Community Health 2008 ; 62 : 650 -4. OpenUrl Abstract / FREE Full Text
• ↵ Bilimoria KY, Bentrem DJ, Feinglass JM, Stewart AK, Winchester DP, Talamonti MS, et al. Directing surgical quality improvement initiatives: comparison of perioperative mortality and long-term survival for cancer surgery. J Clin Oncol 2008 ; 26 : 4626 -33. OpenUrl Abstract / FREE Full Text
• ↵ Junor EJ, Hole DJ, Gillis CR. Management of ovarian cancer: referral to a multidisciplinary team matters. Br J Cancer 1994 ; 70 : 363 -70. OpenUrl CrossRef PubMed Web of Science
• ↵ Ruhstaller T, Roe H, Thürlimann B, Nicoll JJ. The multidisciplinary meeting: an indispensable aid to communication between different specialities. Eur J Cancer 2006 ; 42 : 2459 -62. OpenUrl CrossRef PubMed Web of Science
• ↵ King’s Fund Forum Consensus Statement. Consensus development conference: treatment of primary breast cancer. BMJ 1986 ; 293 : 946 -7. OpenUrl FREE Full Text
• ↵ Saini KS, Taylor C, Ramirez AJ, Palmieri C, Gunnarsson U, Schmoll HJ, et al. Role of the multidisciplinary team in breast cancer management: results from a large international survey involving 39 countries. Ann Oncol 2012 ; 23 : 853 -9. OpenUrl Abstract / FREE Full Text

• Open access
• Published: 06 April 2024

Impact of risk factors, early rehabilitation and management of lymphedema associated with breast cancer: a retrospective study of breast Cancer survivors over 5 years

• Slobodan Tomić 1 ,
• Goran Malenković 2 ,
• Ermina Mujičić 3 ,
• Armin Šljivo   ORCID: orcid.org/0000-0003-2865-0446 3 &
• Sanja D. Tomić 2  

BMC Women's Health volume  24 , Article number:  226 ( 2024 ) Cite this article

Metrics details

Breast cancer-related lymphedema (BCRL) is a potentially disabling and often irreversible consequence of breast cancer treatment, caused by the mechanical incompetence of the lymphatic system, resulting in reduced drainage capacity and functional overload due to an excessive volume of interstitial fluid surpassing the system’s transport capacity in the arm. We wanted to determine the impact and explore the differences in independent risk factors for the occurrence of BCRL; incidence of BCRL over a five-year period at the Institute of Oncology Vojvodina in Sremska Kamenica and to answer the research question regarding the influence of the prehabilitation program on the overall incidence of BCRL during the observed five-year period.

From 2014 to 2018, a retrospective study was conducted at the Institute of Oncology of Vojvodina in Sremska Kamenica, analyzing female patients who had undergone breast cancer surgery.

The study included 150 breast cancer patients who developed secondary lymphedema following surgery with the mean age of 59.2 ± 11.3 years. Fluctuations in hospitalization rates were observed over the five-year period, with the highest number of admissions in 2014 (24.0%) and a decline in 2018 (14.0%). The most common surgical procedure performed was left quadrantectomy (24.0%), followed by right quadrantectomy (20.0%) and left amputation (15.3%). The mean number of removed lymph nodes was 15.2 ± 6.1, with no statistically significant association between the number of removed lymph nodes and the manifestation of secondary lymphedema. The severity of secondary lymphedema varied based on patient age, with a higher incidence of moderate and severe lymphedema observed in patients aged 61 years and older. Patients who underwent radical surgery were more likely to experience severe lymphedema compared to those who had conservative surgery, although this difference was not statistically significant.

In our study, the type of surgery, elapsed time since surgery, and the number of removed lymph nodes were not influencing factors for the occurrence of BCRL. However, concerning its severity, a greater number of systemic therapy modalities combined with radiotherapy were associated with a more frequent occurrence of mild and moderate BCRL. Also, the severity of BCRL varied among different age groups, with a higher incidence of moderate and severe lymphedema observed in patients aged 61 years and older. Ultimately, improving the quality of life for individuals affected by secondary lymphedema remains a crucial goal in the field of oncology.

Peer Review reports

Lymphedema associated with breast cancer (BCRL) is a potentially disabling and often irreversible complication of breast cancer treatment. It is believed to occur as a result of reduced drainage capacity or functional overload of the lymphatic system when the volume of interstitial fluid exceeds the existing transport capacity of the lymphatic system in the arm, due to the mechanical incapacity of the lymphatic system [ 1 , 2 , 3 ]. Studies have elucidated the complex interaction of inflammatory processes, disturbed lymphatic remodeling, and impaired lymphatic angiogenesis in the intricate pathogenesis of BCRL, involving multiple risk factors [ 4 , 5 , 6 ]. The likelihood of developing BCRL largely depends on risk factors that can be divided into two categories. One group consists of patient-specific or dependent factors, while the other group includes treatment-specific independent risk factors for breast cancer. Dependent factors include body mass index (BMI) at the time of diagnosis, subclinical edema, and cellulitis on the treated side [ 7 ]. Independent risk factors for BCRL related to treatment include the type of surgical intervention, the number of removed lymph nodes, radiation to regional lymph nodes, and chemotherapy [ 8 , 9 , 10 , 11 , 12 ]. Regarding the timeframe, BCRL can manifest at different time intervals during the postoperative period, with the majority of cases occurring within the first 2 years after surgery [ 3 ]. However, it is important to note that BCRL can develop even several years after the initial treatment, emphasizing the need for long-term monitoring and support for breast cancer survivors [ 4 ]. Therefore, early detection and timely interventions are integral parts of BCRL prevention. Diagnosing preclinical lymphedema can be challenging and requires preoperative assessment and monitoring [ 13 , 14 , 15 , 16 , 17 ]. The diagnosis of BCRL is based on a detailed consideration of risk factors, associated symptoms, clinical signs, and physical examination. Literature provides a wide range of non-invasive methods for assessing secondary lymphedema of the arm, including sequential measurements of arm circumference, volumetric measurements, and tissue tonometry [ 7 , 18 ]. Differences of 2 cm or more in arm circumference recorded at least at one measurement site, or when translated into volume, an increase of 10% or an increase of 200 ml at any time compared to the opposite arm, constitute clinical indicators for the diagnosis of BCRL [ 4 , 5 , 6 ]. The four stages of lymphedema based on the International Society of Lymphology framework are described as follows: Stage 0 - subclinical with no visible changes; Stage 1 - soft edema, without dermal fibrosis and with pitting that resolves upon elevation of the limb; Stage 2 - moderate with decreased elasticity due to evolving dermal fibrosis and no reduction in swelling with limb elevation; and Stage 3 - chronic and irreversible [ 19 ]. Based on the severity, breast cancer-related lymphedema can be classified into three categories: mild, moderate, and severe. The lack of standardized diagnostic criteria and consistent diagnostic methods for BCRL results in varying reported incidence rates ranging from less than 5% to over 50% [ 3 , 4 , 5 ].

Taking all of the above into consideration, with this study, we aimed to address several questions. Firstly, we wanted to determine the impact and explore differences in independent risk factors for the development of BCRL. Secondly, our goal was to establish the incidence of BCRL over a five-year period at the Institute of Oncology of Vojvodina in Sremska Kamenica. Our ultimate aim was to answer the research question regarding the influence of prehabilitation on the overall incidence of BCRL during the observed five-year period.

Patients and study design

A retrospective study was conducted at the Institute of Oncology of Vojvodina (IOV) in Sremska Kamenica on a sample of patients who underwent breast cancer surgery from 2014 to 2018. The patient sample was formed based on the following criteria: surgical treatment of breast cancer with either breast-conserving or radical surgery with axillary dissection, application of systemic therapy (chemotherapy, hormonal therapy, biological therapy, individually or in combination), undergoing radiation therapy, and participation in an early rehabilitation program. Patients who underwent sentinel node biopsy and those with oncoplastic surgical approaches were not included in the study.

Ethical approval and informed consent

The study was approved by the Ethics Committee of the Institute of Oncology of Vojvodina in Sremska Kamenica and the Ethics Committee of the Faculty of Medicine in Novi Sad (01–39/63). Every individual involved in the study was thoroughly briefed on the nature of the data employed, and they willingly and knowingly gave their informed consent for its use in the research.

Data collection and study adjustment

Relevant patient data collected from medical histories for the study included: age, time elapsed since BCRL, type of surgery, number of removed lymph nodes, values of measured sequential circumferences of limbs, severity of lymphedema, and types of therapeutic modalities (chemotherapy, biological therapy, hormonal therapy, and radiotherapy).

The criterion for diagnosing BCRL was the presence of a circumference difference of 2 cm at least at one of the 5 measurement levels compared to the contralateral arm. The assessment of upper limb circumferences was performed at five symmetric levels: over the metacarpophalangeal joints; over the wrist joint; 10 cm below the olecranon; over the olecranon; and 10 cm above the olecranon. The lymphedema circumference, expressed as the ratio between the unaffected arm’s circumference and the affected arm’s circumference, was calculated using the following formula: [(total circumference of unaffected arm - total circumference of affected arm) / total circumference of healthy arm] × 100. Based on severity, BCRL was classified into three categories: mild lymphedema (minimal) - an increase in circumference from 2 to 2.9 cm; moderate lymphedema - an increase in circumference from 3 to 4.9 cm; and severe lymphedema - an increase in circumference of 5 cm or more.

All patients included in the study, who underwent surgery at the Institute of Oncology of Vojvodina in Sremska Kamenica, were enrolled in the early rehabilitation program. This program consists of active and/or actively assisted exercises aimed at maintaining and increasing mobility and muscle strength of the operated region, as well as the entire body, towards postural and functional reeducation. The early rehabilitation program promotes a set of six to seven exercises performed in standing, lying, and sitting positions up to the pain threshold and without fatigue, 2–3 times daily, gradually increasing the number of repetitions.

As rehabilitation begins on the second postoperative day, the difficulties that patients may face during this period can serve as potential barriers to the implementation of the exercise program. These difficulties include concerns about the illness, a subjective sense of shoulder joint stiffness, pain, lack of motivation, presence of fatigue and weakness, lack of interest, fear of injury, and fear related to the presence of the drainage system. For these reasons, exercises are conducted under the supervision of a physiotherapist to create an individual plan for each patient.

The main components of the individually tailored exercise plan include the intensity, frequency, and duration of physical activity, the selection of exercises to be applied both during hospitalization and at home until the first follow-up, as well as their sequence, duration, and number of repetitions. Each session begins with deep breathing and light warm-up exercises, gradually increasing the number of repetitions for each exercise. General recommendations advise performing selected five to six exercises daily during hospitalization, initially one to two times a day under the supervision of a physiotherapist. The number of repetitions for each exercise starts with two to three times (maximum five) initially, and after the removal of the drainage and in home conditions, the number of repetitions gradually increases but does not exceed a maximum of 10 repeated movements for each exercise. After discharge, follow-up visits are scheduled for 3 weeks, then for 1 month, and further as needed. The effectiveness of the early rehabilitation program has been documented in previous studies [ 20 , 21 ].

Statistical analysis

The data collected during the research were checked for validity, encoded, and entered into a specially created database on a personal computer. After data entry, their statistical analysis and processing were carried out using the IBM SPSS (Statistical Package for Social Sciences) version 26. Descriptive and inferential statistical methods were employed in the statistical analysis and data processing. Numeric variables were presented through mean values (arithmetic mean) and measures of variability (range, standard deviation), while categorical variables were presented through frequencies and percentages.

Testing differences in the frequency distribution for nominal variables was performed using the Pearson χ2 test with Yates’ correction for continuity, and for ordinal variables, the Mantel-Haenszel χ2 test was applied. In all analyses, p  ≤ 0.05 was considered statistically significant, and the results were presented graphically and in tabular form, with the graphs created using the Excel 2016 software package.

The study included a total of 150 patients, with an average age of 59.2 ± 11.3, where the youngest patient was 23 and the oldest was 82 years old. Out of the 150 patients diagnosed with BCRL, in 2014, the highest number was recorded, with 36 patients (24.0%), slightly fewer in 2015, with 29 patients (19.3%), while in 2016 and 2017, an equal number of patients, 32 each (21.3%), was observed. In 2018, there was a decline in the number of patients with BCRL, with 21 patients (14.0%) (Fig. 1 ).

Incidence of BCRL, 2014–2018 years

Regarding surgical procedures, the largest proportion of patients underwent left quadrantectomy, with 36 (24.0%) patients, followed by right quadrantectomy with 30 (20.0%) patients. Left amputation was performed on 23 (15.3%) patients, while segmentectomy was conducted on 19 (12.7%) patients. Right mastectomy was carried out in 16 (10.7%) patients, and right amputation in 14 (9.3%) patients. Left mastectomy was performed in 11 (7.3%) patients, while tumor resection, or tumorectomy, was only performed in one patient, representing 0.7% of the total. The mean time elapsed since the surgery was 15.1 months, with the shortest period from the operation at the time of data analysis being 1 month, and the longest period being 80 months. The largest number of patients, 75 (54.91%), received a combination of chemotherapy and hormonal therapy, while only two patients (1.63%) had a combination of hormonal and biological therapy (Table 1 ).

The mean time elapsed since the surgery was 15.1 months, with the shortest period from the operation at the time of data analysis being 1 month, and the longest period being 80 months (Table 2 ).

Among the patients, 103 (68.7%) were diagnosed with mild BCRL, 32 (21.3%) had moderate BCRL, and 14 (9.3%) experienced severe BCRL. When comparing the BCRL categories based on the type of surgery, specifically conservative or radical surgical intervention, it is evident that a greater proportion of patients who underwent conservative surgery had mild BCRL compared to those who had radical surgery (63.1% vs 36.9%). Furthermore, a higher number of patients who underwent radical surgery were diagnosed with severe BCRL, as opposed to those who had conservative surgery (64.3% vs 35.7%). However, the observed differences are not statistically significant (Pearson chi-squared test; χ^2 = 5.495; df = 2; p  = 0.064) (Table 3 ). The severity of BCRL was also not statistically significant when examined in relation to the applied therapy (Pearson chi-squared test; χ^2 = 5.196; df = 6; p  = 0.591). The distribution and difference in the severity level of BCRL based on the applied therapy are presented in Table 3 . Patients who received multiple therapy modalities more frequently had mild and moderate BCRL compared to patients who received one or two therapy modalities (61.2 and 66.8% vs 38.8 and 31.3%). The manifestation of a severe form of BCRL in relation to the applied therapy was the same in both groups of patients (Table 3 ).

Regarding the number of removed lymph nodes, no statistically significant differences were observed in the manifestation of BCRL (Mantel-Haenszel chi-squared test; χ 2  = 3.150; df = 1; p  = 0.076) (Table 4 ). However, severe BCRL is more frequent in patients with a larger number of removed nodes compared to subjects with a smaller number of removed lymph nodes (78.6% vs. 21.4%). Comparing the categories of BCRL severity with the time elapsed since surgery, no statistically significant differences were observed (Mantel-Haenszel chi-squared test; χ 2  = 0.287; df = 1; p  = 0.592).

In comparing the severity categories of BCRL based on patient age, the highest number of patients with mild BCRL was found in the group aged ≤55 years ( n  = 39, 37.9%). Moderate BCRL was diagnosed in patients between the ages of 56 and 65 years, while severe BCRL was observed in half of the patients aged 66 years and older. Statistically significant differences were found when examining the severity of BCRL in relation to two age categories: patients ≤60 years old and those 61 years old and above (Mantel-Haenszel chi-squared test; χ^2 = 4.851; df = 1; p  = 0.028). Specifically, a higher incidence of moderate and severe BCRL was observed among patients aged 61 years and older (Table 5 .).

This study is one of the first in the region to investigate the occurrence of BCRL after breast cancer surgery in Serbia and the West Balkan area. The study aimed to explore the correlation between various factors such as hospitalization rates, surgical procedures, number of removed lymph nodes, elapsed time since surgery, patient age, type of surgery, and applied therapy with the severity of BCRL. Our study population, which had different surgical approaches towards breast cancer removal, showed different hospitalization rates over a five-year period from 2014 to 2018. In our study, analyzing a sample of patients operated at the Institute of Oncology of Vojvodina in Sremska Kamenica during the period from 2014 to 2018, the highest frequency of BCRL was recorded in 2014. Out of a total of 232 breast cancer patients who underwent surgery, BCRL was diagnosed in 36 (15.51%) patients. The following year, 2015, had a higher number of patients, with 291 patients, but BCRL was diagnosed in 29 (9.96%) patients. The lowest frequency of BCRL was recorded in 2018, when out of 267 patients, BCRL was diagnosed in only 21 patients (7.86%). In 2016, BCRLwas diagnosed in 32 patients (11.8%) out of a total of 252 surgeries, while in 2017, BCRL was diagnosed in 32 patients (12.69%). A higher incidence of moderate and severe BCRL was observed among patients aged 61 years and older, while a larger proportion of patients who underwent conservative surgery experienced mild BCRL compared to those who had radical surgery. No differences were found when examining the categories of BCRL severity in relation to the time that had passed since the surgery and in relation to the applied therapy modalities. Regarding hospitalization rates, the data shows fluctuations over a five-year period. The composition of the study population, including patient demographics and disease stage, could have influenced the likelihood of hospitalization. Factors such as comorbidities, disease severity, and patient preferences might have varied over the years, leading to differences in hospitalization rates. Outpatient programs for managing lymphedema, which consist of physical therapy and surgical interventions, can significantly enhance the care and outcomes of patients who have multiple coexisting health conditions. Prevention of BCRL can be treatment-specific. Reducing the extent of axillary surgery in certain circumstances, mapping the lymphatic channels of the upper extremities during surgery, and reducing the amount of nodal radiation in appropriate cases can help in prevention. According to the data, the most frequently performed procedure among patients was left quadrantectomy, accounting for 24.0% of the cases, followed by right quadrantectomy, which represented 20.0% of the patients. Although the differences were not statistically significant, a higher proportion of patients who underwent conservative surgery had mild BCRL compared to those who had radical surgery. Conversely, severe BCRL was more frequently diagnosed in patients who underwent radical surgery. These findings indicate a potential association between the type of surgery and the severity of lymphedema, although further research is necessary to establish a conclusive correlation. However, in our study, there was no compelling evidence to define the number of lymph nodes removed in correlation with BCRL. Recent studies [ 22 , 23 , 24 ] have shed light on a potential positive correlation between the number of lymph nodes removed during surgical procedures and the likelihood of developing BCRL. In other words, these studies suggest that a higher number of lymph nodes removed may increase the risk of experiencing significant lymphedema symptoms as a secondary condition. Vicini et al. [ 25 ] showed a trend of increased lymphedema when four or more lymph nodes were removed; however, this was not statistically significant. Engel et al. [ 26 ] demonstrated that taking 10 or more lymph nodes was significantly associated with lymphedema. However, there is no consensus regarding the number and lymphedema. In a recent study of 936 patients, there was also no association between the number of nodes removed and lymphedema [ 27 ].

The findings indicated that there were no statistically significant differences in the severity of lymphedema based on the elapsed time since surgery. This suggests that the severity of secondary lymphedema may not be strongly influenced by the amount of time that has passed since the initial surgery. However, it should be noted that the average duration varied among different surgical procedures, suggesting potential differences in recovery or disease progression rates.

The older age group exhibited a higher occurrence of moderate and severe BCRL compared to the younger age group. With advancing age, there may be natural changes in the lymphatic system that affect its ability to efficiently drain lymph fluid. This reduced lymphatic function can increase the risk of developing lymphedema and potentially lead to more severe symptoms [ 28 ]. Furthermore, older individuals are more likely to have multiple comorbid conditions, such as hypertension, diabetes, or cardiovascular diseases [ 29 ]. These comorbidities can affect overall health and compromise the lymphatic system’s functionality, making them more susceptible to developing severe lymphedema [ 30 ]. Aging is also associated with a decline in skin elasticity and tissue tone. This loss of elasticity can impede the ability of tissues to accommodate fluid accumulation, leading to increased swelling and severity of lymphedema symptoms [ 31 ]. Lastly, older individuals generally have a slower healing process and may take longer to recover from surgery. This delayed healing can exacerbate lymphatic dysfunction and contribute to the severity of lymphedema symptoms [ 32 ]. The results of conducted research show a wide range of BCRL occurrence frequencies. Differences in study quality, sample size estimation, sampling technique, and research methodology usually form the basis for heterogeneity in meta-analysis of data on incidence or prevalence [ 33 , 34 , 35 , 36 ]. The relatively low incidence of BCRL in our study can be explained by the specific algorithm of early rehabilitation implemented at the Institute of Oncology of Vojvodina in Sremska Kamenica. The early rehabilitation program begins on the second day after breast cancer surgery, provided there are no surgical or cardiological contraindications [ 21 ]. Patients are trained in exercises from the early rehabilitation program with the aim of preventing the development of functional complications through active movement, stretching of the postoperative scar, and activation of the “muscle pump.” Active exercises allow for proper remodeling of the postoperative scar in the axillary region and preservation of functionality in the ipsilateral arm. The kinesiotherapy set consists of 5 to 6 exercises performed daily, 2 to 3 times a day, including rhythmic flexion and extension, as well as circular movements at all levels of the ipsilateral arm, in standing and lying positions with the elevated arm, without fatigue, under the supervision of a physiotherapist during hospitalization. Education during early rehabilitation is intended for patients and family members with the aim of maintaining or restoring functional status and/or maximizing the level of functional independence while minimizing the effects of the disease and its treatment [ 22 ]. Numerous published results report the benefits of early rehabilitation interventions precisely during this period [ 8 , 37 , 38 ]. The positive effects are manifested by increased range of motion in the direction of flexion and abduction in the shoulder joint, both in the short term (3 months after surgery) and in the long term (1 year after surgery), without an increased risk of complications such as seroma formation, wound healing difficulties, postoperative pain, and BCRL [ 37 , 38 ].

This retrospective study provides valuable insights into the occurrence and severity of secondary lymphedema following breast cancer surgery among female patients in Sremska Kamenica, Serbia. The findings highlight the importance of long-term surveillance and support for breast cancer survivors, as lymphedema can manifest several years after the initial treatment. The study provides important baseline data on secondary lymphedema in this population, highlighting the need for ongoing research, advancements in treatment modalities. Early intervention, such as early rehabilitation can help reduce swelling, improve lymphatic function, and alleviate symptoms. The findings underscore the importance of multidisciplinary care involving healthcare professionals who specialize in the management of lymphedema. Ultimately, improving the quality of life for individuals affected by secondary lymphedema remains a crucial goal in the field of oncology.

Limitations

This study had several limitations. The study was conducted on a relatively small sample size of 150 female patients from a specific region (Novi Sad, Serbia). This may limit the generalizability of the findings to other populations. The study utilized a retrospective cross-sectional design, relying on data collected from medical records and patient charts. This design may introduce biases and limitations associated with retrospective data collection, including missing or incomplete information. The study was conducted in a single center, which may limit the representativeness of the findings. The results may not reflect the experiences and outcomes of breast cancer survivors in other healthcare settings or regions. The diagnosis of lymphedema was based on clinical assessment by healthcare professionals, including physical examination and patient-reported symptoms. While objective measurements were utilized when available, the reliance on clinical assessment alone may introduce subjectivity and potential variability in the diagnosis.

In our study, the type of surgery, elapsed time since surgery, and the number of removed lymph nodes were not influencing factors for the occurrence of BCRL. However, concerning its severity, a greater number of systemic therapy modalities combined with radiotherapy were associated with a more frequent occurrence of mild and moderate BCRL. Also, the severity of BCRL varied among different age groups, with a higher incidence of moderate and severe lymphedema observed in patients aged 61 years and older.

Availability of data and materials

Data and materials can be obtained by reaching out to the corresponding author upon request.

Sleigh BC, Manna B. Lymphedema. 2023 Apr 19. In: StatPearls [internet]. Treasure Island (FL): StatPearls publishing; 2023 [internet]. [cited 2023 may 22].

Rockson SG. Lymphedema after breast Cancer treatment. N Engl J Med. 2018;379:1937–44.

Article   PubMed   Google Scholar  

Rockson SG, Keeley V, Kilbreath S, Szuba A, Towers A. Cancer-associated secondary lymphoedema. Nat Rev Dis Primers. 2019;5(1):22.

He L, Qu H, Wu Q, Song Y. Lymphedema in survivors of breast cancer. Oncol Lett. 2020;19(3):2085–96.

PubMed   PubMed Central   Google Scholar  

McDuff SGR, Mina AI, Brunelle CL, Salama L, Warren LEG, Abouegylah M, et al. Timing of lymphedema after treatment for breast cancer: when are patients most at risk? Int J Radiat Oncol Biol Phys. 2019;103:62–70.

Asdourian MS, Swaroop MN, Sayegh HE, Brunelle CL, Mina AI, Zheng H, et al. Association between precautionary behaviors and breast cancer-related lymphedema in patients undergoing bilateral surgery. J Clin Oncol. 2017;35:3934–41.

Article   PubMed   PubMed Central   Google Scholar  

McEvoy MP, Ravetch E, Patel G, Fox J, Feldman S. Prevention of breast cancer-related lymphedema. Clin Breast Cancer. 2021;21(2):128–42.

DiSipio T, Rye S, Newman B, Hayes S. Incidence of unilateral arm lymphoedema after breast cancer: a systematic review and meta-analysis. Lancet Oncol. 2013;14(6):500–15.

Tsai RJ, Dennis LK, Lynch CF, Snetselaar LG, Zamba GK, Scott-Conner C. The risk of developing arm lymphedema among breast cancer survivors: a meta- analysis of treatment factors. Ann Surg Oncol. (2009) 16(7):1959–72. 10.1245/s10434–009–0452-2.

Asdourian MS, Skolny MN, Brunelle C, Seward CE, Salama L, Taghian AG. Precautions for breast cancer-related lymphoedema: risk from air travel, ipsilateral arm blood pressure measurements, skin puncture, extreme temperatures, and cellulitis. Lancet Oncol. 2016;17(9):e392–405.

Ferguson CM, Swaroop MN, Horick N, Skolny MN, Miller CL, Jammallo LS, et al. Impact of ipsilateral blood draws, injections, blood pressure measurements, and air travel on the risk of lymphedema for patients treated for breast cancer. J Clin Oncol. 2016;34(7):691–8.

McLaughlin SA, Wright MJ, Morris KT, Sampson MR, Brockway JP, Hurley KE, et al. Prevalence of lymphedema in women with breast cancer 5 years after sentinel lymph node biopsy or axillary dissection: patient perceptions and precautionary behaviors. J Clin Oncol. 2008;26(32):5220–6.

Executive C. The diagnosis and treatmen to peripheral lymphedema:2016 consensus document of the international society of lymphology. Lymphol. 2016;49(4):170–84.

Google Scholar  

Armer JM, Hulett JM, Bernas M, Ostby P, Stewart BR, Cormier JN. Best practice guidelines in assessment, risk reduction, management, and surveillance for post-breast cancer lymphedema. Curr Breast Cancer Rep. 2013;5(2):134–44.

Ostby PL, Armer JM, Dale PS, Van Loo MJ, Wilbanks CL, Stewart BR. Surveillance recommendations in reducing risk of and optimally managing breast cancer-related lymphedema. J Pers Med. 2014;4(3):424–47. https://doi.org/10.3390/jpm4030424 .

Harrington S, Gilchrist L, Sander A. Breast cancer EDGE task force outcomes: clinical measures of pain. Rehabil Oncol. 2014;32(1):13–21. https://doi.org/10.1097/01893697-201432010-00004 .

Soran A, Ozmen T, McGuire KP, Diego EJ, McAuliffe PF, Bonaventura M, et al. The importance of detection of subclinical lymphedema for the prevention of breast cancer-related clinical lymphedema after axillary lymph node dissection; a prospective observational study. Lymphat Res Biol. 2014;12(4):289–94. https://doi.org/10.1089/lrb.2014.0035 .

Goss JA, Greene AK. Sensitivity and Specificity of the Stemmer Sign for Lymphedema: A Clinical Lymphoscintigraphic Study. Plast Reconstr Surg Glob Open. 2019;7(6):e2295.

McLaughlin SA, Staley AC, Vicini F, Thiruchelvam P, Hutchison NA, Mendez J, et al. Considerations for clinicians in the diagnosis, prevention, and treatment of breast cancer-related lymphedema: recommendations from a multidisciplinary expert ASBrS panel : part 1: definitions, assessments, education, and future directions. Ann Surg Oncol. 2017;24(10):2818–26.

Tomić S, Malenković G, Lalić N, Bojović M, Tomić S. Effects of early rehabilitation treatment on the functional recovery and quality of life in patients three months after breast cancer surgery. Srp Arh Celok Lek. 2020;148(1–2):81–6.

Popović-Petrović S, Tomić S, Nedeljković M, Popović L, Matovina GEarly rehabilitation in patients with breast carcinoma. Vojnosanit Pregl. 2013;70(4):407–10.

Yen TW, Fan X, Sparapani R, Laud PW, Walker AP, Nattinger AB. A contemporary, population-based study of lymphedema risk factors in older women with breast cancer. Ann Surg Oncol. 2009;16(4):979–88. https://doi.org/10.1245/s10434-009-0347-2 .

Paskett ED, Naughton MJ, McCoy TP, Case LD, Abbott JM. The epidemiology of arm and hand swelling in premenopausal breast cancer survivors. Cancer Epidemiol Biomarkers Prev. 2007;16:775–82.

Meeske KA, Sullivan-Halley J, Smith AW, et al. Risk factors for arm lymphedema following breast cancer diagnosis in black women and white women. Breast Cancer Res Treat. 2008;113(2):383-91.

Vicini F, Shah C, Lyden M, Whitworth P. Bioelectrical impedance for detecting and monitoring patients for the development of upper limb lymphedema in the clinic. Clin Breast Cancer. 2012;12(2):133–7. https://doi.org/10.1016/j.clbc.2012.01.004 .

Engel J, Kerr J, Schlesinger-Raab A, Sauer H, Holzel D. Axilla surgery severely affects quality of life: results of a 5-year prospective study in breast cancer patients. Cancer Res Treat. 2003;79(1):47–57. https://doi.org/10.1023/A:1023330206021 .

Article   Google Scholar  

Kim HK, Ju YW, Lee JW, Kim KE, Jung J, Kim Y, et al. Association between number of retrieved sentinel lymph nodes and breast cancer-related lymphedema. J Breast Cancer. 2021;24(1):63–74. https://doi.org/10.4048/jbc.2021.24.e9 .

Shang T, Liang J, Kapron CM, Liu J. Pathophysiology of aged lymphatic vessels. Aging. 2019;11(16):6602–13. https://doi.org/10.18632/aging.102213 .

Article   CAS   PubMed   PubMed Central   Google Scholar  

Petrie JR, Guzik TJ, Touyz RM. Diabetes, hypertension, and cardiovascular disease: clinical insights and vascular mechanisms. Can J Cardiol. 2018;34(5):575–84. https://doi.org/10.1016/j.cjca.2017.12.005 .

Jiang X, Tian W, Nicolls MR, Rockson SG. The lymphatic system in obesity, insulin resistance, and cardiovascular diseases. Front Physiol. 2019:10:1402. https://doi.org/10.3389/fphys.2019.01402 .

Li K, Zhang Z, Liu NF, Feng SQ, Tong Y, Zhang JF, et al. Efficacy and safety of far infrared radiation in lymphedema treatment: clinical evaluation and laboratory analysis. Lasers Med Sci. 2017;32:485–94.

Guo S, DiPietro LA. Factors affecting wound healing. J Dent Res. 2010;89(3):219–29. https://doi.org/10.1177/0022034509359125 .

Iyigun EZ, Selamoglu D, Alco G, Pilanci KN, Ordu C, Agacayak F, et al. Bioelectrical impedance for detecting and monitoring lymphedema in patients with breast cancer. Preliminary results of the Florence nightingale breast study group. Lymphat Res Biol. 2015;13:40–5.

Bundred NJ, Stockton C, Keeley V, Riches K, Ashcroft L, Evans A, et al. Comparison of multi-frequency bioimpedance with perometry for the early detection and intervention of lymphoedema after axillary node clearance for breast cancer. Breast Cancer Res Treat. 2015;151:121–9.

Stout Gergich NL, Pfalzer LA, McGarvey C, Springer B, Gerber LH, Soballe P. Preoperative assessment enables the early diagnosis and successsful treatment of lymphedema. Cancer. 2008;112:2809–19.

Sun F, Skolny MN, Swaroop MN, Rawal B, Catalano PJ, Brunelle CL, et al. The need for preoperative baseline arm measurement to accurately quantify breast cancer-related lymphedema. Breast Cancer Res Treat. 2016;157:229–40.

Article   CAS   PubMed   Google Scholar  

Ammitzbøll G, Kjaer T, Johansen C, Lanng C, Wreford Andersen E, Kroman N, et al. Effect of progressive resistance training on health-related quality of life in the first year after breast cancer surgery-results from a randomized controlled trial. Acta Oncol. 2019;31:1–8.

Ayre K, Parker C. Lymphedema after treatment of breast cancer: a comprehensive review. J Unexplored Med Data. 2019;4:5.

Download references

Author information

Authors and affiliations.

Faculty of Medicine of University of Novi Sad, Novi Sad, Serbia

Slobodan Tomić

Department of Nursing, Faculty of Medicine, University of Novi Sad, Novi Sad, Serbia

Goran Malenković & Sanja D. Tomić

Clinical Center of University of Sarajevo, Sarajevo, Bosnia and Herzegovina

Ermina Mujičić & Armin Šljivo

You can also search for this author in PubMed   Google Scholar

Contributions

SD. T. and S. T.: They wrote the main manuscript text. They were responsible for drafting and organizing the content, conducting the literature review, and presenting the key findings and analysis in the paper. A. Š. made significant contributions to the research design and methodology. He provided expertise in a particular aspect of the study, such as data analysis, and theoretical frameworks. G.M. and E. M.: They actively participated in the review process of the manuscript. They provided critical feedback, suggestions for improvement, and ensured the overall quality and integrity of the paper.

Corresponding author

Correspondence to Armin Šljivo .

Ethics declarations

Ethical approval and consent to participate.

The study was approved by the Ethics Committee of the Institute of Oncology of Vojvodina in Sremska Kamenica and the Ethics Committee of the Faculty of Medicine in Novi Sad (01–39/63). Every subject involved in the study gave informed consent to participate in the study.

Consent for publication

Consent for publication not applicable.

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.

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/ . 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 in a credit line to the data.

Reprints and permissions

About this article

Cite this article.

Tomić, S., Malenković, G., Mujičić, E. et al. Impact of risk factors, early rehabilitation and management of lymphedema associated with breast cancer: a retrospective study of breast Cancer survivors over 5 years. BMC Women's Health 24 , 226 (2024). https://doi.org/10.1186/s12905-024-03062-7

Download citation

Received : 14 September 2023

Accepted : 29 March 2024

Published : 06 April 2024

DOI : https://doi.org/10.1186/s12905-024-03062-7

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

• Secondary lymphedema
• Breast cancer

BMC Women's Health

ISSN: 1472-6874

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

• Open access
• Published: 29 November 2023

Machine learning-based models for the prediction of breast cancer recurrence risk

• Duo Zuo 1 , 2 , 3 , 4 , 5   na1 ,
• Lexin Yang 1 , 2 , 3 , 4 , 5   na1 ,
• Yu Jin 1 , 6 ,
• Huan Qi 7 ,
• Yahui Liu 1 , 2 , 3 , 4 , 5 &
• Li Ren 1 , 2 , 3 , 4 , 5  

BMC Medical Informatics and Decision Making volume  23 , Article number:  276 ( 2023 ) Cite this article

2147 Accesses

1 Citations

Metrics details

Breast cancer is the most common malignancy diagnosed in women worldwide. The prevalence and incidence of breast cancer is increasing every year; therefore, early diagnosis along with suitable relapse detection is an important strategy for prognosis improvement. This study aimed to compare different machine algorithms to select the best model for predicting breast cancer recurrence. The prediction model was developed by using eleven different machine learning (ML) algorithms, including logistic regression (LR), random forest (RF), support vector classification (SVC), extreme gradient boosting (XGBoost), gradient boosting decision tree (GBDT), decision tree, multilayer perceptron (MLP), linear discriminant analysis (LDA), adaptive boosting (AdaBoost), Gaussian naive Bayes (GaussianNB), and light gradient boosting machine (LightGBM), to predict breast cancer recurrence. The area under the curve (AUC), accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and F1 score were used to evaluate the performance of the prognostic model. Based on performance, the optimal ML was selected, and feature importance was ranked by Shapley Additive Explanation (SHAP) values. Compared to the other 10 algorithms, the results showed that the AdaBoost algorithm had the best prediction performance for successfully predicting breast cancer recurrence and was adopted in the establishment of the prediction model. Moreover, CA125, CEA, Fbg, and tumor diameter were found to be the most important features in our dataset to predict breast cancer recurrence. More importantly, our study is the first to use the SHAP method to improve the interpretability of clinicians to predict the recurrence model of breast cancer based on the AdaBoost algorithm. The AdaBoost algorithm offers a clinical decision support model and successfully identifies the recurrence of breast cancer.

Peer Review reports

Breast cancer (BC) is one of the most common malignancies among women worldwide and a leading cause of cancer-related death in women [ 1 ]. The incidence has increased with the introduction of mammography screening, and BC cases in China account for 12.2% of all newly diagnosed breast cancers and 9.6% of all deaths from BC worldwide [ 2 ]. International studies suggest that approximately 30% of women will develop recurrence after the primary treatment for BC [ 3 ]. Patients with HR + breast cancer are at risk of recurrent disease even multiple decades after primary diagnosis [ 4 ]. Triple-negative BC have a high risk of distant relapse in the first 3 to 5 years following diagnosis [ 5 ]. Hence, the development of models to predict BC recurrence is important to aid in diagnosis and monitoring.

Breast cancer is a histologic diagnosis made based on standardized pathologic criteria. It primarily falls into invasive ductal carcinoma (60-75% of patients), invasive lobular carcinoma (5-15% of patients), and some special type carcinomas, making up the remainder of patients [ 6 ]. In BC, some pathological characteristics, such as estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2), are used to guide treatment decisions. Due to the complex causes of BC, control, early diagnosis and appropriate treatment are important strategies for improving prognosis [ 7 ]. Downregulation of endoplasmic reticulum signaling by endocrine drugs is the primary systemic treatment for ER-positive or PR-positive breast cancers. HER2 is overexpressed in approximately 20% of breast cancers and is associated with poor prognosis in the absence of systemic therapy [ 8 ]. Patients with HER2- overexpressing breast cancer benefit from HER2-targeted therapy, including anti-HER2 antibodies (such as trastuzumab and pertuzumab) and small-molecule tyrosine kinase inhibitors (such as lapatinib and neratinib) [ 9 ].

The diagnosis and monitoring of BC are the main aspects of BC therapy. The information derived from patient and primary tumor features, specifically tumor size, nodal status, tumor grade, and therapeutic modalities, has been used to build prognostic models such as PREDICT [ 10 ]. However, despite considerable efforts at the early detection of recurrent disease, evidence suggests that only a small number of recurrent cases are detected at the asymptomatic stage [ 11 , 12 ]. Multidisciplinary research or data mining is necessary to help physicians predict BC recurrence.

Recently, as artificial intelligence (AI) and its application in clinical cancer research have made rapid developments, cancer prediction performance has reached new heights [ 13 , 14 ]. Powerful AI techniques, especially machine learning (ML) and deep learning (DL), can extract clinical information from massive amounts of data to assist in proper clinical decision making [ 15 , 16 ]. These AI techniques are noninvasive techniques to diagnose the disease without harming the patient. ML is considered an objective and reproducible method for integrating multiple quantitative variables to improve diagnostic accuracy [ 17 ]. In population studies, ML can be used to effectively characterize BC risk, predict outcomes, and identify biomarkers without a priori assumptions of causation [ 18 , 19 , 20 ]. In breast cancer recurrence models, most studies have established predictive models based on imaging and pathological parameters [ 21 , 22 , 23 , 24 ]. Is it possible to use the clinical information obtained from the electronic medical records and the results of routine laboratory indicators to develop and verify the model for predicting the recurrence of BC?

This study explored and validated eleven predictive algorithms using an ML approach based on the clinicopathological and laboratory routine index data of BC patients. Our aim was to use the clinical information easily collected in clinical practice to create a clinical decision support system to identify patients at risk of recurrent cancer and promote early intervention in these patients.

Literature survey

Currently, AI techniques and statistical methods is increasingly used and developed in clinical oncology to diagnose cancers, predict patient outcomes, and inform treatment planning. In particular, rich imaging and molecular data have stimulated the application of ML and/or DL. Recently, Manoj Sharma et al. [ 25 ] proposed a comparative analysis of handcrafted features extraction approaches and DL frameworks for colon and lung cancer classification. A significant improvement in classifiers performance is observed with features extracted by deep convolutional neural networks (CNNs). The random forest (RF) classifier with DenseNet-121 extracted deep features can identify colon and lung cancer tissue with excellent results. Similarly, the authors proposed a hybrid approach for survival prediction of hepatocellular carcinoma with more accuracy and sensitivity [ 26 ]. The proposed RFGBEL model presented excellent performance in contrast to other proposed models, which achieved an accuracy of 93.92%, sensitivity of 94.73%, F-1 score of 0.93. Yala et al. [ 27 ]proposed a DL model was built to triage mammograms by setting a high-sensitivity prediction threshold so that nearly all predicted negative cases were truly negative.

Many state-of-art studies have been presented for prediction of breast cancer. Manoj Sharma et al. [ 28 ] used an ensemble model comprising three pretrained CNNs to make grading predictions for the Databiox dataset, which consists of histopathological images of invasive ductal carcinoma breast cancer diagnosed patients for this grade classification and achieved an accuracy of 94%. Dhahri et al. [ 29 ] suggested an ML-based approach in combination of Genetic Programming to distinguish between benign and malignant breast tumors using electronic health records of 569 patients collected from the Wisconsin Breast Cancer dataset. In an experiment with seven classifiers, the adaptive boosting (AdaBoost) classifier performed best, with a fair accuracy of 98.23%, making it suitable for early BC detection in controlled parametric setting. Whitney et al. [ 30 ] used both ML and DL algorithms to analyze routine H&E-stained images of early-stage ER + breast cancer patients to predict the corresponding Oncotype DX recurrence risk. Bremer et al. [ 31 ] developed a biologic signature named DCISionRT for the calculation of individual decision score (DS), which combined molecular markers and clinicopathological factors associated with recurrence or progression of ductal carcinoma in situ patients following breast-conserving surgery in a nonlinear model.

On the one hand, many studies used limited clinical information analyzed by traditional statistical methods, and on the other hand, many studies analyzed image and pathology data by ML. From the literature survey, we found a relatively limited number of studies that predicted BC recurrence solely from easily accessible clinical information and routine laboratory metrics combined with ML. This study utilizes the results of clinical information and routine laboratory indicators obtained from electronic medical records combined with a comparison of 11 proposed ML models for predicting BC recurrence and is expected to present a rational model to help clinicians and decision makers.

Materials and methods

From January 2011 to December 2018, 342 hospitalized women diagnosed with primary BC at the Tianjin Medical University Cancer Institute and Hospital (Tianjin, China) were enrolled. All patients had complete pathological and clinical laboratory test results. Data were collected retrospectively, including patient characteristics, laboratory results, tumor size, lymph node staging (based on the eighth edition of AJCC) and treatment strategies (Table  1 and Supplementary Table  1 ).

The inclusion criteria were as follows: (1) patients who met the diagnostic criteria for BC and were confirmed by pathological examination; (2) women with unilateral breast lesions for the first time; (3) patients who had not received chemotherapy, radiotherapy, or endocrine therapy; and (4) patients with complete clinical and pathological data. The exclusion criteria were as follows: (1) patients with hypertension, heart disease, diabetes, glaucoma, or other underlying diseases; (2) patients with double breast tumor, double BC, or previous breast tumor resection; (3) patients with other tumors; (4) patients with an intellectual disability or other serious mental illness; and (5) patients with liver, kidney, or other gynecological diseases. The process is shown in Fig.  1 .

Visual diagram of the detailed process for clinical design and data collection

Data preprocessing

For patient information, we converted “patients over 46 years old” to 1 and “patients ≤ 46 years old” to 0. For the diagnosis code, we converted “patients with BC recurrence” to 1 and “patients with no recurrence” to 0; We converted “patients with menopause” to 1 and “patients with no menopause” to 0; We converted “patients with primary cancer in the left breast” to 1 and “patients with primary cancer in the right breast” to 2; We converted “patients with invasive ductal carcinoma of BC” to 1, “patient with other types of invasive carcinoma” to 0, and “patients with unknown histological types of BC” to 2; We converted “patient with tumor size ≤ 2 cm” to 0, “patients with tumor size > 2 cm and ≤ 5 cm” to 1, “patients with tumor size > 5 cm” to 2, “patients with unknown tumor size” to 3; We converted “patient with lymph node staging 0” to 0, “patient with lymph node staging 1” to 1, “patient with lymph node staging 2” to 2, and “patient with lymph node staging 3” to 3; We converted “patient with unknown of histological grade” to 0, “patient with more well-differentiated histological grade” to 1, “patient with moderately differentiated histological grade” to 2, and “patient with more poorly differentiated histological grade” to 3 and “patient with undifferentiated histological grade” to 4; We converted “patient with HER2-positive of molecular subtype” to 1, “patient with Triple-Negative Breast Cancer (TNBC)” to 2, “patient with Luminal A of molecular subtype” to 3, “patient with Luminal B HER2- negative of molecular subtype” to 4, “patient with Luminal B HER2-positive of molecular subtype” to 5, and “patient with unknown of molecular subtype” to 6; We converted “patient with breast conserving therapy” to 0, “patient with mastectomy” to 1, “patient with chemotherapy” to 2, and “patient with unknown treatment strategies” to 3. Missing value imputations on missing at random data of laboratory indicators were used random forest by Python package (Sklearn, 1.0.2) and multiple imputation by R package (mice, 4.1.2).

Machine learning models

The prediction model was developed by using the following algorithms: logistic regression (LR) [ 32 ], random forest (RF) [ 33 ], support vector classification (SVC) [ 34 ], extreme gradient boosting (XGBoost) [ 35 ], gradient boosting decision tree (GBDT) [ 36 ], decision tree [ 37 ], multilayer perceptron (MLP) [ 38 ], linear discriminant analysis (LDA) [ 39 ], AdaBoost [ 40 ], Gaussian naive Bayes (GaussianNB) [ 41 ], and light gradient boosting machine (LightGBM) [ 42 ]. All ML analyses were performed by Python 3.8.8. The study samples were randomly divided into a training set (n = 239) and a testing set (n = 103) at a ratio of 7:3 [ 43 , 44 ]. In the process of training, we used a 3-fold inner cross-validation approach to estimate the models [ 45 , 46 ]. In the test set, the AUC, accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and F1 score were estimated. The best prediction model was selected by evaluating the largest AUC [ 47 , 48 ]. We applied the Shapley Additive Explanation (SHAP) to explain the best-performing predictive model. Feature ranking was obtained by computing SHAP values. The features were ordered by the mean absolute value of the SHAP values for each feature [ 44 ]. Combining ML with SHAP could provide an explicit explanation of the efficacy prediction [ 44 , 47 , 49 ]. The process is shown in Fig.  2 .

Flowchart of the machine learning development model for predicting recurrence of breast cancer. Abbreviations: BC, breast cancer; LR, logistic regression [ 32 ]; RF, random forest [ 33 ]; SVC, support vector classification [ 34 ]; XGBoost, extreme gradient boosting [ 35 ]; GBDT, gradient boosting decision tree [ 36 ]; decision tree [ 37 ]; MLP, multilayer perceptron [ 38 ]; LDA, linear discriminant analysis [ 39 ]; AdaBoost, adaptive boosting [ 40 ]; GaussianNB, Gaussian naive Bayes [ 41 ]; LightGBM, light gradient boosting machine [ 42 ]; SHAP, Shapley Additive Explanation; DCA, decision curve analysis

Statistical analyses

Categorical data were analyzed using the chi-square test [ 50 ]. If the two sets of continuous variables were normally distributed, a two-tailed Student’s t -test was used for comparison. If the two sets of continuous variables were nonnormally distributed, a Mann‒Whitney test was used for comparison. The cumulative risk curve was drawn by Kaplan‒Meier methods. The cumulative risk incidence between the two groups was compared based on Kaplan‒Meier analysis and the log-rank test. SHAP and decision curve analysis (DCA) were performed using Python 3.8.8. Statistical analysis was conducted by SPSS statistics 25.0. All statistical tests were two-tailed, and p  < 0.05 was considered significant.

Clinical features

In all, 342 BC patients (average age, 46.65 years; range, 19–77 years) from January 2011 to December 2018 were identified, and 256 (74.9%) had recurrence, 86 (25.1%) patients had no recurrence. Table  1 summarizes the clinical and tumor histological characteristics of patients. The most frequent molecular subtypes were luminal B HER2-neg (43.6%) > HER2-pos (19.3%) > TN (18.1%) > luminal B HER2-pos (8.5%) > luminal A (6.7%) [missing data for molecular subtypes were grouped as unknown (3.8%)]. Compared with patients without recurrence, patients with recurrence had multiple lymph node involvement and invasive ductal-type disease, and the histological grade and tumor size in these patients were significantly higher (Table  1 ). In clinical laboratory characteristics, D-dimer, CEA, CA125, CA15-3, WBC, NEUT, NLR, Fbg and α2-AP levels played critical roles in the differential diagnosis of patients with BC recurrence and no recurrence (Supplementary Table  1 ). All of these clinical features are easily obtainable from the electronic medical records of BC patients, and there are a total of 25 clinical features.

Machine learning-based prediction of BC recurrence

We hypothesized that the comprehensive integration of clinical features might provide important clues to predict BC recurrence outcomes. Therefore, we obtained 25 clinical features from electronic medical records, all of which were used for the development of predictive models for BC recurrence. We tested the performance of eleven ML models, including AdaBoost, LightGBM, XGBoost, decision tree, GBDT, LDA, GaussianNB, SVC, LR, RF and MLP, using the discovery cohort. We selected and then tested eleven types of ML models as clinical decision-support systems for predicting BC recurrence. During the development of these models, the clinical features of 70% of the patients were randomly selected for training. In addition, we performed a 3-fold internal cross-validation used to assess the effectiveness of the predictive ability of a model built based on the training set, and externally validated the accuracy of the predictive ability of the model by going through a test set based on an independent sample size (Table 2 and Supplementary Table  2 ). Furthermore, to evaluate the performance of a ML model, a confusion matrix was used (Supplementary Table  3 ). The prediction performances of these eleven ML models were compared, and the most accurate prediction model was chosen. The model obtained by AdaBoost had the best discrimination (AUC = 0.987) (Fig.  3 ). The sensitivity, specificity, PPV, NPV, F1 score and accuracy of the model for predicting BC recurrence were 94.7%, 97.6%, 90.0%, 98.8%, 92.3% and 97.1%, respectively (Table  2 ).

Comparison of the area under the receiver operating characteristic curves for eleven machine learning algorithms. Abbreviations: LR, logistic regression; RF, random forest; SVC, support vector classification; XGBoost, extreme gradient boosting; GBDT, gradient boosting decision tree; MLP, multilayer perceptron; LDA, linear discriminant analysis; AdaBoost, adaptive boosting; GaussianNB, Gaussian naive Bayes; LightGBM, light gradient boosting machine

The AdaBoost algorithm was adopted in the establishment of the prediction model. To better understand how features in the prediction model of BC recurrence based on the AdaBoost algorithm contribute to the prediction results, we calculated the SHAP value of each feature. The top 20 features were selected by the importance ranking of the average absolute SHAP value, which was based on the AdaBoost algorithm model (Fig.  4 a). According to the importance ranking of the average absolute SHAP value, the top 4 features [carcinoma antigen 125 (CA125), carcinoembryonic antigen (CEA), fibrinogen (Fbg) and tumor diameter] were assessed as the most important variables. Figure  4 b is a violin plot of each feature showing the correlation between the value of each feature and the SHAP value. The larger the absolute value of a feature’s SHAP, the greater the hint that this feature has a greater impact on the AdaBoost-based prediction model. Red dots represent the higher values for this feature, while blue dots represent lower values for this feature. Higher CA125, Fbg, carcinoma antigen 15 − 3 (CA15-3), D-dimer and coagulation factor VIII (FVIII) concentrations, red blood cell (RBC) count, N stage, larger tumor diameter and lower CEA, α2-antiplasmin (α2-AP) and tissue polypeptide specific antigen (TPSA) concentrations were associated with a higher predicted probability of 5-year BC recurrence. Furthermore, different molecular subtypes also had a certain impact on the outcome of BC recurrence.

SHAP values and feature interaction scores in AdaBoost-based prediction. ( a ) The top 20 most important features for the prediction of BC recurrence (ranked from most to least important). ( b ) The distribution of the impacts of each feature on the model output. The colors represent the feature values: red for larger values and blue for smaller values. Abbreviations: CA125, carcinoma antigen 125; CEA: carcinoembryonic antigen; Fbg: fibrinogen; CA15-3, carcinoma antigen 15 − 3; FVIII, coagulation factor VIII; TPSA, tissue polypeptide-specific antigen; α2-AP, α2-antiplasmin; RBC, red blood cell; NEUT, neutrophils; PLR, platelet-to-lymphocyte ratio; WBC, white blood cell; PLT, platelet, SHAP, Shapley Additive Explanation

Clinical use

Several single clinical features were found to be significant predictive markers of BC recurrence. There were significant correlations between CA125 expression (cutoff 4.71 U/ml) and BC prediction. The cutoff value was selected based on the probability threshold of Youden’s index. The study population was divided into high-risk groups and low-risk groups based on the cutoff value. Based on Kaplan‒Meier analysis and the log-rank test, there was a significant difference in progression-free survival between the two groups ( p < 0.0001, Fig.  5 a). CA125 expression levels affected the risk of recurrence, with higher expression levels associated with a higher five-year risk of recurrence and a shorter progression-free survival in patients. Similarly, higher expression levels of CA15-3, Fbg, D-dimer and FVIII were correlated with a worse prognosis in patients (Fig.  5 b–e).

In clinical practice, machine model prediction is not simply predictive of patients who will likely have BC recurrence or be free of recurrence. The clinical application value of the model was evaluated by DCA. We expressed the net benefit as a function of the decision threshold in the decision curve, and the threshold probability reflected the cost‒benefit ratio. The DCA of the 11 ML algorithms is shown in Fig.  5 f, which shows that when the threshold probability of a patient was greater than 1%, using the AdaBoost algorithm model to guide clinical intervention provided more benefit than either the intervention for all (black line) or none (dotted line). Compared with the other algorithms, the net benefit in this range had obvious superiority. When 1% was taken as the prediction probability, the net benefit of the AdaBoost algorithm was significantly higher than that of the other algorithms.

Kaplan‒Meier plots and decision curve analysis (DCA). ( a – e ) Kaplan‒Meier plot of progression-free survival (PFS) based on CA125, CA15-3, Fbg, D-Dimer and FVIII expression (* p  < 0.05, **** p  < 0.0001). ( f ) DCA of different ML algorithms. The y-axis measures the net benefit. The dotted line (Treat None) represents the net benefit of the prediction of nonrecurrence for all BC patients; the black line (Treat All) represents the net benefits of the outcomes of recurrence for all BC patients. The model with the highest clinical value was determined by quantifying net benefits under different thresholds

In this study, an AdaBoost-based model was trained and tested as a decision-making tool, which is expected to predict the recurrence risk of BC. In addition, the most important variable features were selected by SHAP from many clinical characteristics related with BC. With this type of analysis, clinicians can use the model established by the proposed algorithm to identify BC patients with high recurrence risk, and it is expected to improve the risk stratification of patients in clinical practice.

BC, the most common malignancy diagnosed in women worldwide, is a highly heterogeneous disease presenting with a broad range of clinical and molecular characteristics. In recent years, there has been a steady decline in BC mortality, and early detection of BC recurrence allows for more effective salvage treatment. Women with early BC are at an ongoing risk of relapse, even after successful surgery and treatment of the primary tumor [ 51 ]. Following initial treatment, BC can recur locally, regionally (nodes), or at distant metastatic sites. For women with HR-negative disease, the risk of recurrence is mainly confined to the first 5 years after diagnosis [ 52 , 53 ]. Patients with HR-positive tumors are at risk of late recurrence even after triamcinolone therapy [ 54 ]. Most recurrences of BC occur distally, with the most common sites of metastasis being the bones, liver and lungs [ 55 ]. Although BC tumor markers such as CA15-3 and CEA can be used to detect early tumor recurrence, the serum test is not suitable alone for BC follow-up [ 56 , 57 ]. To establish a predictive model for BC recurrence with comprehensive integration of relevant clinical factors, we collected 25 clinically relevant features that are clinically common and readily available from patients with BC from electronic medical records, including CA125 expression, coagulation function (Fbg, FVIII and D-dimer), tumor diameter, molecular subtype, and previous strategy of therapy, among others. These factors are evaluated in routine clinical practice and do not require additional cost or effort.

CA125 is expressed by normal bronchial, endometrial, ovarian and corneal epithelial cells, and it was first identified in mice immunized with ovarian cancer cells [ 58 ]. CA125 is best known as a biomarker for monitoring epithelial ovarian cancer [ 59 ]. In addition, CA125 is a repeating peptide epitope of the mucin MUC16, which promotes breast cancer cell proliferation and metastasis [ 60 , 61 ]. An increase in the concentration of CA125 is an indicator of disease recurrence [ 62 ]. In a study by Jager et al., CA 125 levels in 26 patients with a single site of metastasis out of 250 metastatic BC patients were analyzed, suggesting that an elevated CA 125 level in metastatic BC patients is related to pleural disease [ 63 ]. Another prospective study also demonstrated the value of metastatic sites. Among nine patients with pleural-based disease, eight (89%) had an elevated CA 125 level, and progressive disease correlated with elevated CA 125 levels in all cases [ 64 ]. This suggests that lesions close to the pleura can induce an inflammatory reaction and result in elevated CA 125 levels. In a retrospective review of 51 patients with treated BC, progressive disease in 21 patients correlated with an elevation in CA125 in 57% of patients and one false-negative reduction [ 65 ]. Continuous biomarker monitoring has the potential to predict the diagnosis of recurrence at the minimum asymptomatic stage [ 66 ]. Our data demonstrate that an increase in CA 125 may also result in the earlier detection of recurrent or progressive disease, which is likely to alter survival and affect quality of life. The patients with values above the cutoff of CA125 presented a significantly shorter median PFS than those with values below the cutoff. The optimal use of this marker in breast carcinomatosis is unclear, but its possible use in combination with other tumor markers, such as CA15-3 or CEA, is expected to be of complementary value for clinical decision making and to improve our understanding of the function of CA125 in human pathology.

CEA is a cytoplasmic glycoprotein that is highly expressed in most tumor tissues and is commonly used as a marker to assess cancer risk and prognosis [ 67 , 68 ]. However, this study reveals conclusions that are inconsistent with these several studies, and low CEA expression levels are involved in predicting the recurrence risk of BC according to SHAP values. Similarly, in 105 patients with metastatic BC, 39 patients (37%) with low CEA levels had significantly shorter median survival times after recurrence (18 versus 28 months) than patients with high CEA levels [ 69 ]. Low CEA levels may indicate complex and heterogeneous disease; thus, there might be a subtype of BC with rapid proliferation and low CEA secretion [ 69 ]. Preoperative serum levels of CEA were associated with molecular subtypes of BC, and CEA expression levels were significantly lower in patients with triple-negative metastatic BC than in those with other subtypes [ 70 ].

Malignant tumor growth and dissemination are associated with the development of a subclinical hypercoagulable state [ 71 ]. The patient’s coagulation abnormalities worsened with cancer progression and metastasis. In this respect, circulating thrombotic biomarkers may represent a novel noninvasive factor for better prediction of disease recurrence risk [ 72 ]. In our study, fibrinogen, FVIII and D-dimer had potential value in predicting BC recurrence. D-dimer is the primary degradation product of cross-linked fibrin, representing an index of both coagulation and fibrinolysis activation. The pathogenesis of cancer coagulation activation is complex and variable. Laboratory results indicate that fibrinolysis and fibrinolysis processes are similar in the progression of malignant tumors and are increasingly present in patients with metastases [ 73 ]. In BC, high fibrinogen levels were associated with poorer overall survival [ 74 , 75 ]. Some studies have shown that D-dimers are useful indicators for monitoring metastasis in cancer patients, and increased D-dimer levels are associated with the rate of progression and poor prognoses [ 76 , 77 ], which is consistent with our data.

A growing body of evidence suggests that the risk of recurrence depends heavily on the biology of BC [ 78 , 79 , 80 ]. The classification of subtypes shows the heterogeneity of BC, which has been shown to be of prognostic value in BC. Several studies have revealed associations between molecular subtypes of BC and local recurrence rates. The subtypes are ER-positive luminal A (luminal A), ER-positive luminal B (luminal B), HER2 enriched, basal-like, and normal breast-like. In a study of 2985 patients classified into different subtypes, HER2-enriched and basal subtypes showed a significantly higher risk of regional relapse after breast-conserving therapy [ 81 ]. Luminal B tumors have poorer outcomes than luminal A tumors due to the expression of some proliferating genes, such as Ki-67, CCNB1 and MYBL2 [ 82 ]. In addition, tumor size and lymph node status were significant predictors of disease-free survival and overall survival. In a cohort of 15,819 women with invasive BC, the rate of lymph node metastasis increased with increasing tumor volume in BC patients with tumors smaller than 100 cm 3 , increasing BC mortality [ 83 ].

The strong heterogeneity of BC represents a serious issue for treatment monitoring [ 84 ], and predicting the individual risk of recurrence of primary BC will enable physicians to choose the best treatment strategy. In this respect, AI holds great promise to enable the evaluation of tumor aggressiveness, individual risk of recurrence, and response to specific treatments in BC [ 85 ]. AI is applied to assist cancer diagnosis and prognosis, given its unprecedented accuracy level, which is even higher than that of general statistical experts [ 14 ]. Previous studies have mainly applied AI to two main approaches to BC diagnosis, relying on image analysis and pathological data [ 86 ]. While AI in digital breast pathology and breast imaging shows great promise in reducing false positive rates in breast cancer screening, images might suffer from technical bias [ 86 , 87 ]. In this study, we used clinical characteristics, pathological molecular typing, and laboratory indicators, which provide a detailed fingerprint of tumors to predict recurrent BC by ML-based AI.

ML, as a narrow form of AI, has been proven to be a powerful tool in the prediction of disease outcomes [ 88 , 89 , 90 , 91 , 92 ]. In our study, prediction models based on 11 ML algorithms were tested using 25 easily obtainable clinical features from electronic medical records. Compared with the prediction performance of every single clinical feature, ML-based AdaBoost using the combination of clinical features showed more significant performance. Several recent studies have used ML methods to predict cancer recurrence and survival outcomes. For example, a study showed three prediction models combined with digitized images of fine needle aspiration of breast masses that can be used to predict BC reoccurrence time as accurately as 1 year [ 93 ]. In addition, Tahmassebi A reported using ML with multiparametric magnetic resonance imaging to predict pathological complete response and survival in patients treated with neoadjuvant chemotherapy [ 94 ]. A breast cancer recurrence and metastasis risk assessment framework was developed from histopathological images using image features and ML technologies [ 23 ]. In contrast to these studies, we tested more models based on different algorithms for predicting BC recurrence within a five-year follow-up period through easily accessible clinical information and routine laboratory indicators. We found that AdaBoost can be used to predict recurrence/nonrecurrence with an accuracy of 97.1%, a high sensitivity of 94.7% and a high specificity of 97.6%.

To our knowledge, we used AdaBoost in combination with SHAP for the first time to predict the recurrence of BC. Second, by searching for keywords [(((((((conventional laboratory indicators) OR (routinely measured blood biomarkers)) OR (routinely measured blood indicators)) OR (routinely peripheral blood indicators)) OR (conventional peripheral blood indicators)) AND ((breast cancer) OR (breast carcinoma))) AND (Recurrence)) AND ((((((((((((Machine Learning) OR (logistic regression)) OR (random forest)) OR (support vector machine)) OR (XGBoost)) OR (gradient boosting decision tree)) OR (decision tree)) OR (multilayer perceptron)) OR (linear discriminant analysis)) OR (AdaBoost)) OR (Gaussian naive Bayes)) OR (LightGBM))] on the PubMed website, we believe that our study is the first to use the features of traditional laboratory indicators and clinical information easily available from electronic medical records in AdaBoost’s model to predict the recurrence of BC. AdaBoost is one of the best boosting algorithms. AdaBoost can boost a weak learning algorithm with an accuracy slightly better than random guessing into an arbitrarily accurate strong learning algorithm, bringing about a new method and new insights into the design of the learning algorithm [ 95 ]. Even if many base classifier instances are used, AdaBoost rarely overfits the solution and minimizes the exponential loss function by fitting the stepwise additive model [ 96 ]. Due to the minimization of the classification error, which can be best approximated as exponential loss, AdaBoost performs very well on a wide range of classification problems [ 97 ] AdaBoost could be a helpful tool for physicians to predict BC recurrence. Additionally, we use SHAP to interpret AdaBoost predictions based on SHAP values and feature interaction scores. We found that correlated variables reflecting tumor biomarkers (CA125, CEA, CA15-3), clinicopathological features (tumor diameter, N stage, molecular subtype), and coagulation abnormalities (Fbg, FVIII, D-dimer) have important weights in predicting the recurrence of BC. This may result in more sustainable health for patients, thereby reducing the psychological, social and economic burden on society.

Our study has several limitations. First, the study population was relatively small. Although we evaluated 342 patients, 103 of whom were randomly included in the test set as an independent sample, a larger cohort is needed for future external validation of the accuracy of the prediction model. Second, although we initially evaluated the value of 25 available clinical features for predicting recurrence, we need more clinical information, such as gene mutations, to optimize these prediction models and provide a valuable basis for individualized treatment. Thus, future studies should be conducted to validate the feasibility of the proposed algorithm.

This study described the application of clinical information and laboratory parameters-based ML in patients with BC recurrence, generating a AdaBoost algorithm model that reliably predicts the probability of BC recurrence. In our study, ML combined with the explainability method of SHAP makes the black box model of ML explainable, which is more suitable for the clinical scenario of predicting breast cancer recurrence. In addition, the addition of DCA highlights the clinical value of AdaBoost. We suggest the use of this approach as an auditable decision aid that contributes to patient healthcare and research.

Data Availability

The data generated in this study are available upon request from the corresponding author.

Siegel RL, Miller KD, Fuchs HE, Jemal A, Cancer Statistics. 2021. CA Cancer J Clin. 2021;71(1):7–33.

Fan L, Strasser-Weippl K, Li JJ, St Louis J, Finkelstein DM, Yu KD, et al. Breast cancer in China. Lancet Oncol. 2014;15(7):e279–89.

Article   PubMed   Google Scholar  

Wapnir IL, Anderson SJ, Mamounas EP, Geyer CE Jr., Jeong JH, Tan-Chiu E, et al. Prognosis after ipsilateral breast Tumor recurrence and locoregional recurrences in five National Surgical adjuvant breast and Bowel Project node-positive adjuvant Breast cancer trials. J Clin Oncol. 2006;24(13):2028–37.

Pan H, Gray R, Braybrooke J, Davies C, Taylor C, McGale P, et al. 20-Year risks of Breast-Cancer recurrence after stopping endocrine therapy at 5 years. N Engl J Med. 2017;377(19):1836–46.

Article   PubMed   PubMed Central   Google Scholar  

Foulkes WD, Smith IE, Reis-Filho JS. Triple-negative Breast cancer. N Engl J Med. 2010;363(20):1938–48.

Article   CAS   PubMed   Google Scholar  

Rakha EA, Tse GM, Quinn CM. An update on the pathological classification of Breast cancer. Histopathology. 2023;82(1):5–16.

Rautenberg T, Siebert U, Arnold D, Bennouna J, Kubicka S, Walzer S, et al. Economic outcomes of sequences which include monoclonal antibodies against vascular endothelial growth factor and/or epidermal growth factor receptor for the treatment of unresectable metastatic Colorectal cancer. J Med Econ. 2014;17(2):99–110.

Piccart-Gebhart MJ, Procter M, Leyland-Jones B, Goldhirsch A, Untch M, Smith I, et al. Trastuzumab after adjuvant chemotherapy in HER2-positive Breast cancer. N Engl J Med. 2005;353(16):1659–72.

Waks AG, Winer EP. Breast Cancer Treatment: a review. JAMA. 2019;321(3):288–300.

Wishart GC, Azzato EM, Greenberg DC, Rashbass J, Kearins O, Lawrence G, et al. PREDICT: a new UK prognostic model that predicts survival following Surgery for invasive Breast cancer. Breast Cancer Res. 2010;12(1):R1.

Pourzand A, Tajaddini A, Pirouzpanah S, Asghari-Jafarabadi M, Samadi N, Ostadrahimi AR, et al. Associations between Dietary Allium vegetables and risk of Breast Cancer: a hospital-based Matched Case-Control Study. J Breast Cancer. 2016;19(3):292–300.

Zapater-Moros A, Gámez-Pozo A, Prado-Vázquez G, Trilla-Fuertes L, Arevalillo JM, Díaz-Almirón M, et al. Probabilistic graphical models relate immune status with response to neoadjuvant chemotherapy in Breast cancer. Oncotarget. 2018;9(45):27586–94.

Jiang F, Jiang Y, Zhi H, Dong Y, Li H, Ma S, et al. Artificial intelligence in healthcare: past, present and future. Stroke Vasc Neurol. 2017;2(4):230–43.

Huang S, Yang J, Fong S, Zhao Q. Artificial intelligence in cancer diagnosis and prognosis: opportunities and challenges. Cancer Lett. 2020;471:61–71.

Murdoch TB, Detsky AS. The inevitable application of big data to health care. JAMA. 2013;309(13):1351–2.

Kolker E, Özdemir V, Kolker E. How Healthcare can refocus on its super-customers (Patients, n = 1) and customers (doctors and nurses) by leveraging lessons from Amazon, Uber, and Watson. Omics. 2016;20(6):329–33.

Daimiel Naranjo I, Gibbs P, Reiner JS, Lo Gullo R, Sooknanan C, Thakur SB et al. Radiomics and Machine learning with multiparametric breast MRI for Improved Diagnostic accuracy in Breast Cancer diagnosis. Diagnostics (Basel). 2021;11(6).

Chen Z, Wang M, De Wilde RL, Feng R, Su M, Torres-de la Roche LA, et al. A machine learning model to predict the Triple negative Breast Cancer Immune Subtype. Front Immunol. 2021;12:749459.

Article   CAS   PubMed   PubMed Central   Google Scholar  

Ma M, Liu R, Wen C, Xu W, Xu Z, Wang S, et al. Predicting the molecular subtype of Breast cancer and identifying interpretable imaging features using machine learning algorithms. Eur Radiol. 2022;32(3):1652–62.

Rasool A, Bunterngchit C, Tiejian L, Islam MR, Qu Q, Jiang Q. Improved Machine Learning-based predictive models for Breast Cancer diagnosis. Int J Environ Res Public Health. 2022;19(6).

Yu Y, Tan Y, Xie C, Hu Q, Ouyang J, Chen Y, et al. Development and validation of a Preoperative Magnetic Resonance Imaging Radiomics-Based Signature to Predict Axillary Lymph Node Metastasis and Disease-Free Survival in patients with early-stage Breast Cancer. JAMA Netw Open. 2020;3(12):e2028086.

Lee J, Yoo SK, Kim K, Lee BM, Park VY, Kim JS, et al. Machine learning–based radiomics models for prediction of locoregional recurrence in patients with Breast cancer. Oncol Lett. 2023;26(4):422.

Liu X, Yuan P, Li R, Zhang D, An J, Ju J, et al. Predicting Breast cancer recurrence and Metastasis risk by integrating color and texture features of histopathological images and machine learning technologies. Comput Biol Med. 2022;146:105569.

Romeo V, Cuocolo R, Sanduzzi L, Carpentiero V, Caruso M, Lama B et al. MRI Radiomics and Machine Learning for the prediction of Oncotype Dx Recurrence score in invasive Breast Cancer. Cancers (Basel). 2023;15(6).

Kumar N, Sharma M, Singh VP, Madan C, Mehandia S. An empirical study of handcrafted and dense feature extraction techniques for lung and colon Cancer classification from histopathological images. Biomed Signal Process Control. 2022;75:103596.

Article   Google Scholar  

Sharma M, Kumar N. Improved hepatocellular carcinoma fatality prognosis using ensemble learning approach. J Ambient Intell Humaniz Comput. 2022;13(12):5763–77.

Yala A, Schuster T, Miles R, Barzilay R, Lehman C. A deep learning model to Triage Screening mammograms: a Simulation Study. Radiology. 2019;293(1):38–46.

Kumaraswamy E, Kumar S, Sharma M. An invasive ductal carcinomas Breast Cancer Grade classification using an ensemble of convolutional neural networks. Diagnostics [Internet]. 2023; 13(11).

Dhahri H, Al Maghayreh E, Mahmood A, Elkilani W, Faisal Nagi M. Automated Breast Cancer diagnosis based on machine learning algorithms. J Healthc Eng. 2019;2019:4253641.

Whitney J, Corredor G, Janowczyk A, Ganesan S, Doyle S, Tomaszewski J, et al. Quantitative nuclear histomorphometry predicts oncotype DX risk categories for early stage ER + Breast cancer. BMC Cancer. 2018;18(1):610.

Bremer T, Whitworth PW, Patel R, Savala J, Barry T, Lyle S, et al. A Biological signature for breast ductal carcinoma in situ to predict Radiotherapy Benefit and assess recurrence risk. Clin Cancer Res. 2018;24(23):5895–901.

Vetter TR, Schober P. Regression: the Apple does not fall far from the Tree. Anesth Analg. 2018;127(1):277–83.

Chen Z, He N, Huang Y, Qin WT, Liu X, Li L. Integration of a deep learning classifier with a Random Forest Approach for Predicting Malonylation sites. Genomics Proteom Bioinf. 2018;16(6):451–9.

Huang S, Cai N, Pacheco PP, Narrandes S, Wang Y, Xu W. Applications of support Vector Machine (SVM) Learning in Cancer Genomics. Cancer Genomics Proteomics. 2018;15(1):41–51.

CAS   PubMed   Google Scholar  

Ma B, Meng F, Yan G, Yan H, Chai B, Song F. Diagnostic classification of cancers using extreme gradient boosting algorithm and multi-omics data. Comput Biol Med. 2020;121:103761.

Seto H, Oyama A, Kitora S, Toki H, Yamamoto R, Kotoku J, et al. Gradient boosting decision tree becomes more reliable than logistic regression in predicting probability for Diabetes with big data. Sci Rep. 2022;12(1):15889.

Wang L, Zhu L, Jiang J, Wang L, Ni W. Decision tree analysis for evaluating Disease activity in patients with rheumatoid arthritis. J Int Med Res. 2021;49(10):3000605211053232.

Yang S, Fu C, Lian X, Dong X, Zhang Z. Understanding human-virus protein-protein interactions using a human protein complex-based analysis Framework. mSystems. 2019;4(2).

Xu L, Raitoharju J, Iosifidis A, Gabbouj M. Saliency-based Multilabel Linear Discriminant Analysis. IEEE Trans Cybern. 2022;52(10):10200–13.

Sorayaie Azar A, Babaei Rikan S, Naemi A, Bagherzadeh Mohasefi J, Pirnejad H, Bagherzadeh Mohasefi M, et al. Application of machine learning techniques for predicting survival in Ovarian cancer. BMC Med Inform Decis Mak. 2022;22(1):345.

Ontivero-Ortega M, Lage-Castellanos A, Valente G, Goebel R, Valdes-Sosa M. Fast Gaussian Naïve Bayes for searchlight classification analysis. NeuroImage. 2017;163:471–9.

Rufo DD, Debelee TG, Ibenthal A, Negera WG. Diagnosis of Diabetes Mellitus using gradient boosting machine (LightGBM). Diagnostics (Basel). 2021;11(9).

Guan X, Du Y, Ma R, Teng N, Ou S, Zhao H, et al. Construction of the XGBoost model for early Lung cancer prediction based on metabolic indices. BMC Med Inform Decis Mak. 2023;23(1):107.

Ma J, Bo Z, Zhao Z, Yang J, Yang Y, Li H et al. Machine learning to predict the response to Lenvatinib combined with Transarterial Chemoembolization for Unresectable Hepatocellular Carcinoma. Cancers (Basel). 2023;15(3).

Yang PT, Wu WS, Wu CC, Shih YN, Hsieh CH, Hsu JL. Breast cancer recurrence prediction with ensemble methods and cost-sensitive learning. Open Med (Wars). 2021;16(1):754–68.

Bangalore Yogananda CG, Shah BR, Vejdani-Jahromi M, Nalawade SS, Murugesan GK, Yu FF, et al. A fully automated deep Learning Network for Brain Tumor Segmentation. Tomography. 2020;6(2):186–93.

Mao Y, Huang Y, Xu L, Liang J, Lin W, Huang H, et al. Surgical methods and social factors are Associated with Long-Term Survival in follicular thyroid carcinoma: construction and validation of a Prognostic Model based on machine learning algorithms. Front Oncol. 2022;12:816427.

Lan X, Wang X, Qi J, Chen H, Zeng X, Shi J, et al. Application of machine learning with multiparametric dual-energy computed tomography of the breast to differentiate between benign and malignant lesions. Quant Imaging Med Surg. 2022;12(1):810–22.

Miranda E, Adiarto S, Bhatti FM, Zakiyyah AY, Aryuni M, Bernando C. Understanding Arteriosclerotic Heart Disease patients using Electronic Health records: a machine learning and Shapley Additive exPlanations Approach. Healthc Inf Res. 2023;29(3):228–38.

Wang G, Zhang Y, Li S, Zhang J, Jiang D, Li X, et al. A machine learning-based Prediction Model for Cardiovascular Risk in Women with Preeclampsia. Front Cardiovasc Med. 2021;8:736491.

Sestak I, Cuzick J. Markers for the identification of late Breast cancer recurrence. Breast Cancer Res. 2015;17(1):10.

Cheang MC, Voduc D, Bajdik C, Leung S, McKinney S, Chia SK, et al. Basal-like Breast cancer defined by five biomarkers has superior prognostic value than triple-negative phenotype. Clin Cancer Res. 2008;14(5):1368–76.

Esserman LJ, Moore DH, Tsing PJ, Chu PW, Yau C, Ozanne E, et al. Biologic markers determine both the risk and the timing of recurrence in Breast cancer. Breast Cancer Res Treat. 2011;129(2):607–16.

Saphner T, Tormey DC, Gray R. Annual hazard rates of recurrence for Breast cancer after primary therapy. J Clin Oncol. 1996;14(10):2738–46.

Fisher B, Jeong JH, Anderson S, Bryant J, Fisher ER, Wolmark N. Twenty-five-year follow-up of a randomized trial comparing radical mastectomy, total mastectomy, and total mastectomy followed by irradiation. N Engl J Med. 2002;347(8):567–75.

Kokko R, Holli K, Hakama M. Ca 15 – 3 in the follow-up of localised Breast cancer: a prospective study. Eur J Cancer. 2002;38(9):1189–93.

Molina R, Zanón G, Filella X, Moreno F, Jo J, Daniels M, et al. Use of serial carcinoembryonic antigen and CA 15.3 assays in detecting relapses in Breast cancer patients. Breast Cancer Res Treat. 1995;36(1):41–8.

Bast RC Jr., Feeney M, Lazarus H, Nadler LM, Colvin RB, Knapp RC. Reactivity of a monoclonal antibody with human ovarian carcinoma. J Clin Invest. 1981;68(5):1331–7.

Duffy MJ, Bonfrer JM, Kulpa J, Rustin GJ, Soletormos G, Torre GC, et al. CA125 in Ovarian cancer: European Group on Tumor markers guidelines for clinical use. Int J Gynecol Cancer. 2005;15(5):679–91.

Chaudhary S, Appadurai MI, Maurya SK, Nallasamy P, Marimuthu S, Shah A, et al. MUC16 promotes triple-negative Breast cancer lung Metastasis by modulating RNA-binding protein ELAVL1/HUR. Breast Cancer Res. 2023;25(1):25.

Reinartz S, Failer S, Schuell T, Wagner U. CA125 (MUC16) gene silencing suppresses growth properties of ovarian and Breast cancer cells. Eur J Cancer. 2012;48(10):1558–69.

van der Burg ME, Lammes FB, Verweij J. The role of CA 125 and conventional examinations in diagnosing Progressive carcinoma of the ovary. Surg Gynecol Obstet. 1993;176(4):310–4.

PubMed   Google Scholar  

Jäger W, Kissing A, Cilaci S, Melsheimer R, Lang N. Is an increase in CA 125 in Breast cancer patients an indicator of pleural metastases? Br J Cancer. 1994;70(3):493–5.

Krämer S, Jäger W, Lang N. CA 125 is an indicator for pleural metastases in Breast cancer. Anticancer Res. 1997;17(4b):2967–70.

Jensen JL, Maclean GD, Suresh MR, Almeida A, Jette D, Lloyd S, et al. Possible utility of serum determinations of CA 125 and CA 27.29 in Breast cancer management. Int J Biol Markers. 1991;6(1):1–6.

De Rose F, Meduri B, De Santis MC, Ferro A, Marino L, Colciago RR, et al. Rethinking Breast cancer follow-up based on individual risk and recurrence management. Cancer Treat Rev. 2022;109:102434.

Chen X, Wang X, He H, Liu Z, Hu JF, Li W. Combination of circulating Tumor cells with serum carcinoembryonic antigen enhances clinical prediction of non-small cell Lung cancer. PLoS ONE. 2015;10(5):e0126276.

Yang W, Luo Y, Hu S, Li Y, Liu Q. Value of combined detection of serum carcino-embryonic antigen, carbohydrate antigen 19 – 9 and cyclooxygenase-2 in the diagnosis of Colorectal cancer. Oncol Lett. 2018;16(2):1551–6.

PubMed   PubMed Central   Google Scholar  

Krieger G, Wander HE, Kneba M, Prangen M, Bandlow G, Nagel GA. Metastatic Breast cancer with constantly low CEA blood levels. A subgroup with unfavorable prognosis? J Cancer Res Clin Oncol. 1984;108(3):341–4.

Wang W, Xu X, Tian B, Wang Y, Du L, Sun T, et al. The diagnostic value of serum Tumor markers CEA, CA19-9, CA125, CA15-3, and TPS in metastatic Breast cancer. Clin Chim Acta. 2017;470:51–5.

Caine GJ, Stonelake PS, Lip GY, Kehoe ST. The hypercoagulable state of malignancy: pathogenesis and current debate. Neoplasia. 2002;4(6):465–73.

Giaccherini C, Marchetti M, Masci G, Verzeroli C, Russo L, Celio L, et al. Thrombotic biomarkers for risk prediction of malignant Disease recurrence in patients with early stage Breast cancer. Haematologica. 2020;105(6):1704–11.

Falanga A, Marchetti M, Vignoli A, Balducci D. Clotting mechanisms and cancer: implications in thrombus formation and Tumor progression. Clin Adv Hematol Oncol. 2003;1(11):673–8.

Wen J, Yang Y, Ye F, Huang X, Li S, Wang Q, et al. The preoperative plasma fibrinogen level is an Independent prognostic factor for overall survival of Breast cancer patients who underwent surgical treatment. Breast. 2015;24(6):745–50.

Krenn-Pilko S, Langsenlehner U, Stojakovic T, Pichler M, Gerger A, Kapp KS, et al. An elevated preoperative plasma fibrinogen level is associated with poor disease-specific and overall survival in Breast cancer patients. Breast. 2015;24(5):667–72.

Dirix LY, Salgado R, Weytjens R, Colpaert C, Benoy I, Huget P, et al. Plasma fibrin D-dimer levels correlate with tumour volume, progression rate and survival in patients with metastatic Breast cancer. Br J Cancer. 2002;86(3):389–95.

Tekeşin K, Bayrak S, Esatoğlu V, Özdemir E, Özel L, Melih Kara V. D-Dimer and Carcinoembryonic Antigen levels: useful indicators for Predicting the Tumor Stage and Postoperative Survival. Gastroenterol Res Pract. 2016;2016:4295029.

Foulkes WD, Reis-Filho JS, Narod SA. Tumor size and survival in Breast cancer—a reappraisal. Nat Reviews Clin Oncol. 2010;7(6):348–53.

Article   CAS   Google Scholar  

Prat A, Pineda E, Adamo B, Galván P, Fernández A, Gaba L, et al. Clinical implications of the intrinsic molecular subtypes of Breast cancer. Breast. 2015;24(Suppl 2):26–35.

Holm J, Eriksson L, Ploner A, Eriksson M, Rantalainen M, Li J, et al. Assessment of Breast Cancer risk factors reveals subtype heterogeneity. Cancer Res. 2017;77(13):3708–17.

Voduc KD, Cheang MC, Tyldesley S, Gelmon K, Nielsen TO, Kennecke H. Breast cancer subtypes and the risk of local and regional relapse. J Clin Oncol. 2010;28(10):1684–91.

Perou CM, Jeffrey SS, van de Rijn M, Rees CA, Eisen MB, Ross DT, et al. Distinctive gene expression patterns in human mammary epithelial cells and breast cancers. Proc Natl Acad Sci U S A. 1999;96(16):9212–7.

Sopik V, Narod SA. The relationship between tumour size, nodal status and distant metastases: on the origins of Breast cancer. Breast Cancer Res Treat. 2018;170(3):647–56.

Dagogo-Jack I, Shaw AT. Tumour heterogeneity and resistance to cancer therapies. Nat Reviews Clin Oncol. 2018;15(2):81–94.

Nassif AB, Talib MA, Nasir Q, Afadar Y, Elgendy O. Breast cancer detection using artificial intelligence techniques: a systematic literature review. Artif Intell Med. 2022;127:102276.

La Porta CA, Zapperi S. Artificial intelligence in Breast cancer diagnostics. Cell Rep Med. 2022;3(12):100851.

Chan RC, To CKC, Cheng KCT, Yoshikazu T, Yan LLA, Tse GM. Artificial intelligence in Breast cancer histopathology. Histopathology. 2023;82(1):198–210.

Heo J, Yoon JG, Park H, Kim YD, Nam HS, Heo JH. Machine learning-based model for prediction of outcomes in Acute Stroke. Stroke. 2019;50(5):1263–5.

Stark GF, Hart GR, Nartowt BJ, Deng J. Predicting Breast cancer risk using personal health data and machine learning models. PLoS ONE. 2019;14(12):e0226765.

Dong J, Feng T, Thapa-Chhetry B, Cho BG, Shum T, Inwald DP, et al. Machine learning model for early prediction of acute kidney injury (AKI) in pediatric critical care. Crit Care. 2021;25(1):288.

Kate RJ, Pearce N, Mazumdar D, Nilakantan V. A continual prediction model for inpatient acute kidney injury. Comput Biol Med. 2020;116:103580.

Tack C. Artificial intelligence and machine learning | applications in musculoskeletal physiotherapy. Musculoskelet Sci Pract. 2019;39:164–9.

Gupta SR. Prediction time of Breast cancer Tumor recurrence using machine learning. Cancer Treat Res Commun. 2022;32:100602.

Tahmassebi A, Wengert GJ, Helbich TH, Bago-Horvath Z, Alaei S, Bartsch R, et al. Impact of machine learning with Multiparametric magnetic resonance imaging of the breast for early prediction of response to Neoadjuvant Chemotherapy and survival outcomes in Breast Cancer patients. Invest Radiol. 2019;54(2):110–7.

Cao Y, Miao Q-G, Liu J-C, Gao L. Advance and prospects of AdaBoost Algorithm. Acta Automatica Sinica. 2013;39(6):745–58.

Stiglic G, Pajnkihar M. Evaluation of major online Diabetes risk calculators and computerized predictive models. PLoS ONE. 2015;10(11):e0142827.

Baig MM, Awais MM, El-Alfy E-SM. AdaBoost-based artificial neural network learning. Neurocomputing. 2017;248:120–6.

Download references

Acknowledgements

Thanks to Tianjin Medical University Cancer Institute and Hospital for supporting this study.

This work was sponsored by grants from Tianjin Health Research Project (TJWJ2022ZD003), Tianjin Health Research Project (TJWJ2023MS005), and Tianjin Key Medical Discipline (Specialty) Construction Project (TJYXZDXK-012 A) and Tianjin Key Medical Discipline (Specialty) Construction Project (TJYXZDXK-009 A) in China.

Author information

Duo Zuo and Lexin Yang contributed equally to this work.

Authors and Affiliations

Department of Clinical Laboratory, Tianjin Medical University Cancer Institute & Hospital, Tianjin, 300060, China

Duo Zuo, Lexin Yang, Yu Jin, Yahui Liu & Li Ren

National Clinical Research Center for Cancer, Tianjin, 300060, China

Duo Zuo, Lexin Yang, Yahui Liu & Li Ren

Tianjin’s Clinical Research Center for Cancer, Tianjin, 300060, China

Key Laboratory of Cancer Prevention and Therapy, Tianjin, 300060, China

Key Laboratory of Breast Cancer Prevention and Therapy, Tianjin Medical University, Ministry of Education, Tianjin, 300060, China

Tongji University Cancer Center, Shanghai Tenth People’s Hospital, School of Medicine, Tongji University, Shanghai, 200072, China

China Mobile Group Tianjin Company Limited, Tianjin, 300308, China

You can also search for this author in PubMed   Google Scholar

Contributions

D.Z., H.Q., LX.Y., Y.J., and YH.L. performed bioinformatics and statistical analysis. D.Z., LX.Y., and Y.J. interpreted the results and wrote the manuscript. D.Z. and LX.Y. prepared Figs. 1, 2, 3, 4 and 5. D.Z. and L.R. revised the manuscript. L.R. supervised the study. All authors reviewed the manuscript.

Corresponding author

Correspondence to Li Ren .

Ethics declarations

Ethics approval and consent to participate.

The study complies with the Declaration of Helsinki and has been approved by the Ethical Committee of Tianjin Medical University Cancer Institute and Hospital (No. bc2021230 and No. bc2023158), and consent was obtained for the raw data collection in the patients’ electronic records. The need for informed consent was waived by the Ethics Committee/Institutional Review Board of Tianjin Medical University Cancer Institute and Hospital.

Consent for publication

Not applicable.

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.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Supplementary material 1

: Supplementary Table 1 . Clinical laboratory characteristics of breast cancer patients; Supplementary Table 2 . The 3-fold cross-validation results of 11 machine learning models; Supplementary Table 3 . Confusion matrix of 11 machine learning models

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/ . 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 in a credit line to the data.

Reprints and permissions

About this article

Cite this article.

Zuo, D., Yang, L., Jin, Y. et al. Machine learning-based models for the prediction of breast cancer recurrence risk. BMC Med Inform Decis Mak 23 , 276 (2023). https://doi.org/10.1186/s12911-023-02377-z

Download citation

Received : 21 July 2023

Accepted : 17 November 2023

Published : 29 November 2023

DOI : https://doi.org/10.1186/s12911-023-02377-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

• Breast cancer
• Machine learning
• Artificial intelligence
• Disease recurrence
• Prediction model

BMC Medical Informatics and Decision Making

ISSN: 1472-6947

• General enquiries: [email protected]

Breast cancer: multidisciplinary care and clinical outcomes

Affiliation.

• 1 Screening and Test Evaluation Program (STEP), School of Public Health, A27, University of Sydney, Sydney, NSW 2006, Australia. [email protected]
• PMID: 16904313
• DOI: 10.1016/j.ejca.2006.05.023

A multidisciplinary approach to the management of breast cancer is the standard of care in developed health systems. We performed a systematic review to assess the extent and quality of evidence on whether multidisciplinary care (MDC), or related aspects of care contribute to clinical outcomes in breast cancer, and in particular whether these influence survival. Only two primary studies have looked at MDC and neither of these studies considered long-term outcomes. The studies of MDC (case series) provide weak evidence that MDC may alter treatment patterns. Several population-based cohort studies showed that related aspects of team work, specialist (surgeon) and hospital workload and specialisation, are associated with improved survival. This group of studies used better quality design with more clearly defined outcome measures, and most of the studies have allowed for possible confounding variables. Evidence of a survival benefit was most consistent for specialist (surgeon) effect. However, the reasons behind the improved survival reported in these studies are unclear, and it is unlikely that this is entirely attributable to treatment patterns. We conclude that although intrinsically multidisciplinary care should be associated with better survival, there remains a paucity of evidence to support this. Studies of the long-term clinical effects of MDC in breast cancer should be a priority for future evaluation.

Publication types

• Research Support, Non-U.S. Gov't
• Systematic Review
• Breast Neoplasms / mortality
• Breast Neoplasms / therapy*
• Patient Care Team / organization & administration
• Patient Care Team / standards*
• Survival Analysis
• Treatment Outcome

The association of FSCN1 (rs852479, rs1640233) and HOTAIR (rs920778) polymorphisms with the risk of breast cancer in Egyptian women

• Original Article
• Open access
• Published: 08 April 2024
• Volume 51 , article number  495 , ( 2024 )

Cite this article

You have full access to this open access article

• Eman Reda Galal   ORCID: orcid.org/0000-0001-9167-6812 1 ,
• Dina A. Abdelhakam 2 ,
• Lamiaa Khalaf Ahmed 1 ,
• Yasmine Elhusseny 3 ,
• Sherif El Prince Sayed 4 &
• Noha H. Eltaweel 5  

Breast cancer (BC) is one of the most prevalent cancers that contribute to mortality among women worldwide. Despite contradictory findings, considerable evidence suggests that single nucleotide polymorphisms (SNPs) in the FSCN1 and HOTAIR genes may have a causative impact on the development of BC. This case–control study was conducted to evaluate the association of genotype frequency in FSCN1 rs852479, rs1640233, and HOTAIR rs920778 with susceptibility and prognosis of BC, as well as the impact of clinical stages and hormonal features.

Methods and results

FSCN1 (rs852479, rs1640233) and HOTAIR (rs920778) were genotyped using TaqMan real-time PCR assay in 200 BC patients and 200 cancer-free controls, all representing Egyptian women. Genotypic analyses in association with clinicopathological factors and disease risk were assessed. As a result, a significant association with BC risk was observed for CC genotype frequency of  FSCN1 rs852479 A > C (OR = 0.395, 95% CI 0.204–0.76,  p - value  = 0.005). However, no significant correlation was detected between the FSCN1 rs1640233 C > T and HOTAIR rs920778 C > T polymorphic variants and susceptibility to BC. Interestingly, CC genotype of FSCN1 rs1640233 was more likely to progress tumor size and lymph node invasion in BC cases ( p - value  = 0.04 and 0.02, respectively). Moreover, it was revealed that there was a non-significant correlation between the haplotype distributions of FSCN1 rs852479 and rs1640233 and the probability of BC.

Conclusions

Based on the sample size and genetic characteristics of the subjects involved in the present study, our findings indicated that  FSCN1 rs852479 may contribute to BC susceptibility in a sample of the Egyptian population.

Avoid common mistakes on your manuscript.

Introduction

Breast cancer (BC) is the most prevalent malignancy in women worldwide. According to its death rate, it is considered the second most frequent cause of cancer mortality among women [ 1 ]. Globally, there were 2.3 million women diagnosed with BC and 685,000 deaths from this disease in 2020 [ 2 ]. The incidence and mortality rates of BC vary according to the region [ 3 ]. An estimated 313,510 new instances of invasive BC in women and 611,720 cancer-related deaths will occur in the United States in 2024 [ 4 ]. In Egypt, more than 22,000 new cases of cancer are diagnosed each year, making it the main cause of cancer-related mortality among Egyptian women [ 5 ]. The ratio of mortality/incidence rate of BC cases in Egypt was approximately double the ratio (41%) when compared with developed countries (23%) [ 6 ]. Many variables are associated with the risk of BC, including age, environmental, gynecological, and genetic factors [ 7 ].

Polymorphisms of a DNA sequence caused by a single nucleotide variation in humans are known as single-nucleotide polymorphisms (SNPs), which are the most prevalent types of genetic variations in the human genome. SNPs in genes can potentially alter the protein structure or affect the expression level of the gene product [ 8 , 9 ], which in turn changes disease susceptibility, affecting tumorigenesis and cancer progression as well as drug resistance [ 10 , 11 ]. Certain genetic polymorphisms can predict an individual’s susceptibility to BC and also influence disease management and progression [ 12 ].

Fascin-1 ( FSCN1 ) is a 55-kDa actin-bundling protein coded by a gene located on chromosome 7p22.1 with about 13.84 kb in length and includes five exons. Human FSCN1 is thought to be involved in the assembly of actin filament bundles found in lamellipodia, filopodia, microspikes, and stress fibers [ 13 , 14 ]. FSCN1 is abundantly expressed in many types of normal cells, including neurons, endothelial cells, glial cells, mesenchymal, and antigen-presenting dendritic cells, and is low or absent in normal epithelial cells [ 15 ]. Based on the occurrence of FSCN1 in different organs, it is predictably participating in more biological functions in the human body [ 16 ]. In contrast to normal tissues, increased FSCN1 expression has been associated with several types of malignancies, including lung, colon, breast, ovary, and oral squamous cell carcinoma [ 17 , 18 , 19 , 20 , 21 ]. As an oncogene, FSCN1 can influence mitochondrial remodeling in cancerous cells, in addition to promoting invasion, tumor migration, metastatic colonization, cancer cell self-renewal, and drug resistance. In BC, FSCN1 is crucial for predicting aggressive tumor behavior, especially in advanced stages [ 22 ]. Recent evidence suggests that aberrant STAT3 signaling accelerates the growth of breast tumors by downregulating the expression of downstream target genes that regulate angiogenesis, such as hypoxia-inducible factor-1 (HIF-1) and nuclear factor-kappaB (NF-κB), and by binding to the promoter of the FSCN1 gene, triggering its expression [ 23 ]. Upregulation of FSCN1 enhances the severity and prognosis of human BC and can serve as a diagnostic marker to differentiate triple-negative subtypes of BC from other types of the disease [ 14 ]. Interestingly, different SNPs have been reported in FSCN1 to modulate the risk of BC development [ 24 ].

HOX transcript antisense RNA ( HOTAIR ) is a transcript that originates from the antisense strand of the HOXC gene cluster with an approximate length of 2.2 kb. The human HOTAIR gene is found between  HOXC11  and  HOXC12  genes on the long arm of chromosome 12q13.13 [ 25 ]. It is an example of an oncogenic long noncoding RNA (lncRNA), which has emerged as a master regulator of cancer [ 12 ]. The HOTAIR gene controls several cellular and biochemical processes to promote the proliferation, invasion, survival, drug resistance, and prognosis of various tumors. Some reports indicate that polymorphisms of the HOTAIR gene are associated with a variety of cancers, including breast [ 25 ], pancreatic [ 26 ], gastric [ 27 ], thyroid [ 28 ], and colorectal cancers [ 29 ]. HOTAIR gene expression in BC cells is modulated by numerous epigenetic and transcriptional mechanisms [ 25 ]. Several SNPs, located in the intronic region of the HOTAIR gene, have been reported to regulate its expression level [ 30 , 31 , 32 ]. These SNPs are expected to be related to the occurrence, progression, recurrence, and metastasis of BC and serve as a novel therapeutic target for the disease [ 33 , 34 ].

Recently, the relationship of FSCN1 and HOTAIR polymorphisms with breast tumor development has been investigated [ 34 , 35 , 36 ]. However, some conclusions are still controversial and require further analysis to fully understand the relationship between these genes’ polymorphisms and BC risk. Therefore, this study was conducted to elucidate the association between FSCN1 rs852479, rs1640233, and HOTAIR (rs920778) with the risk or prognosis of BC concerning several clinicopathological variables in the Egyptian population.

Subjects and methods

Study subjects.

This study enrolled 200 Egyptian women with BC (cases) and 200 healthy women without BC (controls), matched by age and comparable socioeconomic factors. All participants were recruited from Beni-Suef University Hospital in the period between 2021 and 2023.

The study protocol was approved by the Ethics Committee of the Faculty of Pharmacy (Girls), Al-Azhar University (REC number: 436), and all study procedures were conducted in accordance with the Declaration of Helsinki. All study participants provided fully informed written consent at the time of study entry.

All samples underwent genotyping for three SNPs: FSCN1 rs852479, rs1640233, and HOTAIR rs920778 to evaluate the association between gene polymorphisms and BC risk. Clinical examinations were detected to assess the impact of polymorphisms on BC patients based on menopausal status, tumor size, lymph node invasion, and histological grade. Additionally, BC prognostic biomarkers, including estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2), were investigated.

DNA extraction

A peripheral blood sample (3 ml) was withdrawn from all study participants under complete aseptic conditions. Genomic DNA was extracted from blood samples using the salting‐out method [ 37 ]. The concentration and quality of the DNA were checked by measuring the absorbance at 260 and 280 nm using a UV spectrophotometer, NanoDrop 2000 (Thermo-Fisher Scientific, Wilmington, USA). Pure preparations of DNA have OD260/OD280 values of 1.7–2.0. The extracted DNA concentration ranged from 50 to 100 ng DNA/μl. The extracted DNA samples were maintained at a temperature of − 20 °C until the genotyping procedure.

Polymorphisms genotyping

Genotyping of FSCN1 rs852479, rs1640233, and HOTAIR rs920778 was done by TaqMan real‐time PCR method using the pre-designed assays for allelic discrimination, containing specific TaqMan probes with fluorescent dyes for each allele. The total PCR volume was 20 µl, containing 5 μl DNA, 10 μl TaqMan Universal PCR Master Mix, 0.05 μl (40×) Assay Mix, and 4.5 μl RNase‐free water. The PCR reaction conditions were the same for the three SNPs, with a pre-denaturation cycle at 95 °C for 10 min, followed by 45 cycles of 95 °C denaturation for 10 s, 60 °C annealing for 30 s, and final extension at 72 °C for 30 s. For genotyping quality control, deionized water was used to replace template DNA as a negative control. The PCR results (changing fluorescence level) were analyzed using the provided software.

Sample size and statistical analysis

The sample size was calculated using G*Power software version 3.1.9.7 for power analysis and sample size [ 38 ]. A total sample size of 400 was required, 200 in each group, with a power of 80% and a significance level of 5%. SPSS 22.0 software package (SPSS Inc., Chicago, IL, USA) was utilized for statistical analysis. Categorical variables are expressed as frequencies and percentages, while continuous variables are given as mean ± standard deviation (SD). Differences in clinical characteristics were compared between patients and healthy control groups using independent-sample t-tests (continuous variables) and chi-square tests (categorical variables). Hardy–Weinberg equilibrium (HWE) analysis was performed for each SNP assay. The chi-square (χ 2 ) test was used to test differences between the two groups for each SNP genotype and allele. Allele frequencies were calculated with the gene counting method. The most common genotypes were selected as the reference. Odds ratios (OR) were calculated with a 95% confidence interval (CI) to estimate the degree of the association between genotypes and the risk of BC. SNPStats ( https://www.snpstats.net/ ) was used to perform haplotype analysis test for linkage disequilibrium (LD) [ 39 ]. The significance level was set at a p-value  < 0.05.

General demographic and clinicopathological characteristics of the studied groups

The demographic and clinicopathological features of BC patients and controls in this study are shown in Table  1 . The mean age of the controls at the time of enrollment was not significantly different from that of the BC cases (48.39 vs. 49.5 years, respectively, p-value  = 0.394). Regarding BC cases, 47% were premenopausal, and 53% were postmenopausal. In respect to ER, 126 cases (63%) tested positive, and 125 cases (62.5%) tested positive for PR, while 45 (22.5%) of the cases were positive for HER2. Referring to the Nottingham prognostic index (NPI), the percentage of BC patients with T 1, 2, 3, and 4 was 9, 37.5, 28.5, and 25%, respectively, and N 0, 1, 2, and 3 emerged in 28, 43.5, 19, and 9.5% of patients, respectively. Concerning the histology grades of BC, 9 patients were classified as grade I (4.5%), 141 as grade II (70.5%), and 50 as grade III (25%).

Distribution frequencies of genotypes and alleles in BC patients and controls

The distribution patterns of FSCN1 rs852479 and rs1640233, and HOTAIR rs920778 genotypes for all subjects are shown in Table  2 and Fig.  1 . In the healthy controls and cases groups, all genotypic frequencies were in HWE ( p-value  > 0.05). Genotype analysis of  FSCN1 polymorphism in both controls and cases revealed that most of those with rs852479 SNP were homozygous for the AA genotype, while with rs1640233 SNP, most of them were homozygous for CC genotype. Likewise, for HOTAIR rs920778 SNP, most of the controls and cases were homozygous for CC genotype (Table  2 ).

SNPs genotypes and allele frequencies of FSCN1 and HOTAIR genes in different studied groups A Frequency of FSCN1 gene rs852479, B FSCN1 gene rs1640233, and C HOTAIR gene rs920778 genotypes and alleles among studied groups

According to the logistic regression analysis for each genetic polymorphism in BC patients and controls as given in Table  2 , we observed that women with CC genotype frequency of  FSCN1 rs852479 A > C have a significantly high incidence of developing BC when compared with AC genotype (CC vs. AC, OR = 0.395; 95% CI 0.204–0.76,  p-value  = 0.005; OR = 1.053, 95% CI 0.67–1.64,  p-value  = 0.82, respectively). In addition, those with the C allele of the FSCN1 rs852479 polymorphism were more likely than those with the A allele to develop BC (C allele OR = 0.63; 95% CI 0.46–0.88, p-value  = 0.01) (Table  2 ; Fig.  1 A). Regarding FSCN1 rs1640233 C > T and HOTAIR rs920778 C > T polymorphism, differences in all genotypes were not significant for BC patients compared with healthy controls (Table  2 ; Fig.  1 B, C ).

The association between genotypes and clinicopathological features of BC patients

In this study, the possible relationship between some clinicopathological parameters of patients with BC and the distribution of SNP genotypes was explored (Table  3 ). Regarding the clinical characteristics, only the  FSCN1 rs1640233 polymorphism of the CC genotype was significantly associated with developing tumor size and lymph node involvement among BC cases ( p-value  = 0.04 and 0.02, respectively). Otherwise, no significant differences were found in the frequencies of FSCN1 rs852479 and HOTAIR rs920778 genotypes in the patients’ group based on all evaluated features ( p-value  > 0.05).

The association of FSCN1 haplotype frequencies with BC in the studied groups

Association analysis between the risk of BC and haplotypes of  FSCN1 rs852479 and rs1640233 among BC cases and controls is summarized in Table  4 . LD was estimated for the two SNPs (r 2  = 0.65, D′ = 0.88), which is expected under linkage disequilibrium. The haplotypes’ distribution showed that the AC haplotype was the most frequent in both cases and controls (69.9 and 77.2%, respectively), while AT haplotype showed the lowest frequency among both groups. Overall, none of the considered haplotypes were significantly associated with the development of BC ( p-value  > 0.05).

BC is a complicated and heterogeneous disease with a multifaceted etiology caused by a combination of genetic and lifestyle-related factors. Various studies suggest that SNP genotyping may contribute to risk assessment and guide BC management. In the present case–control study, we evaluated the frequency distributions of the FSCN1 (rs852479, rs1640233) and HOTAIR (rs920778) SNPs and their associations with BC susceptibility in Egyptian women.

Regarding FSCN1 , our findings revealed that women with CC genotype frequency of rs852479 C > A are significantly associated with a high risk of developing BC when compared with AC genotype (CC vs. AC, OR = 0.395, 95% CI 0.204–0.76,  p-value  = 0.005; OR = 1.053, 95% CI 0.67–1.64,  p-value  = 0.82; respectively). Furthermore, the FSCN1 rs852479 C allele polymorphism is attributed to increased BC risk when compared with the frequency of A allele (C allele OR = 0.63; 95% CI 0.46–0.88, p-value  = 0.01). In contrast, the rs1640233 SNP polymorphism of patients and controls did not differ significantly across all genotypes ( p-value  > 0.05).

Wang et al. [ 24 ] investigated the relationship between six SNPs of the FSCN1 gene in a cohort of Han Chinese women. There were no significant variations detected in the genotypes’ frequency of the rs8772, rs3801004, rs2966447, rs852479, and rs1640233 polymorphisms between BC patients and the healthy control group [ 24 ]. Nevertheless, another study revealed that Egyptian females with the FSCN1 rs3801004 C > G polymorphisms had a significantly higher risk of BC [ 36 ]. Liu et al. [ 40 ] suggest that there might be an association between FSCN1 and the development of BC. They further confirmed the possible functional relevance of FSCN1 expression in the development of Triple-Negative Breast Cancer (TNBC) because it was substantially higher in TNBC than in the non-TNBC subtype. Consequently, these results assist in elucidating the functional significance of FSCN1 in the pathogenesis of TNBC and may provide perspectives on the mechanisms behind cancerous progression [ 40 ].

Concerning HOTAIR rs920778 C > T polymorphism, we reported no significant difference between the cases and control group ( p-value  > 0.05) for all genotypes and alleles. Our results are consistent with a recent study on the Egyptian population, which discovered that the rs920778 C > T polymorphism was not significantly related to BC progression [ 35 ]. According to prior research, the allelic frequencies of the  HOTAIR  gene (rs12826786, rs1899663, and rs4759314) were not statistically different between BC patients and cancer-free controls and were not likely to develop BC [ 41 ].

Contrary to our findings, it has been observed that there was a significant relationship between the rs920778 polymorphism and a high incidence of BC in women from Turkey [ 42 ], Iran [ 43 ], India [ 44 ], and China [ 30 , 45 ]. Based on a meta-analysis of 4 studies with 4936 cases and 5214 healthy controls investigating the association of four HOTAIR SNPs with BC vulnerability, it was found that rs920778 polymorphism significantly lowered the risk of BC under heterozygous, homozygous, and recessive models among the West Asians, and increased BC risk under dominant and allele models within the East Asian population [ 34 ]. Furthermore, some reports have indicated that HOTAIR SNP rs920778 exhibits variable results in the same population but in distinct cancer types such as gastric [ 46 ] and breast [ 42 ], which suggests that there are variations in the polymorphism throughout different malignancies.

These disparities in the results could be caused by genetic diversity among ethnic populations resulting from different gene–gene and gene–environment interactions, or they could be the result of additional constraints associated with the number of cases and sampling techniques. As elucidation, the HapMap data ( https://www.ncbi.nlm.nih.gov/snp/rs920778 ) indicates that there are notable variations in the allele frequency of the HOTAIR rs920778 polymorphism between various ethnic communities. Additionally, the assessment of HOTAIR expression in tumor samples could help in better recognition of the role of these polymorphisms in cancer progression, which ought to be investigated further [ 32 ].

Numerous investigations have been conducted on the relationship between gene polymorphisms involved in different cellular processes and the risk and clinicopathological aspects of BC. When we analyzed the clinical aspects of rs852479 and rs1640233 FSCN1 and rs920778 HOTAIR  genotypic frequencies among BC patients, we found that CC genotype of FSCN1 rs1640233 was significantly associated with developed tumor size and lymph node invasion ( p-value  = 0.04 and 0.02; respectively). Besides, no statistically significant differences were identified in the frequencies of FSCN1 rs852479 and HOTAIR rs920778 genotypes concerning all evaluated parameters ( p-value  > 0.05). Within the same context, other investigations suggested that there was no significant correlation between the clinicopathological aspects of BC patients and the HOTAIR rs920778 polymorphism [ 30 ]. Interestingly, Hassanzarei et al. [ 43 ] discovered that the frequencies of different HOTAIR genotypes in the Iranian population weren’t associated with any clinicopathological features except for rs920778, which was significantly related to ER status. Conversely, Bayram et al. [ 42 ] found that the CC genotype of HOTAIR rs920778 polymorphism was associated with advanced TNM classification, larger tumor size, poor histological grade, and the presence of distant metastasis in BC patients but was not related to other clinic-laboratory or hormonal parameters.

In a comparison of clinic-pathological aspects with FSCN1 genotypes, Wang et al. [ 24 ] discovered that BC patients with the  FSCN1  rs852479 and rs1640233 were not statistically correlated to any clinical status of the tumor. Using immunohistochemistry, Min et al. [ 47 ] investigated FSCN1 expression in a microarray of 194 samples from patients with invasive breast cancer. Findings suggested a strong correlation between the expression of FSCN1 and some clinicopathological characteristics, such as high histological grade, tumor necrosis, and status of ER- and PR-negativity. They further found that FSCN1 expression was significantly associated with BC survival, especially in patients with advanced-stage BC [ 47 ]. Moreover, in Chinese and African-American women, FSCN1 expression is suggested to be associated with TNBC and also linked to more severe clinical aspects and negative hormone receptors [ 48 , 49 ]. Haplotype analyses may provide evidence about the genetic involvement in disease incidence [ 50 ]. We examined the impacts of different haplotype combinations of two  FSCN1  SNPs rs852479 and rs1640233 upon the risk of BC, and no significant relation between haplotypes and BC susceptibility was detected. Overall, as related to other functional polymorphisms in other genes, the effect of genetic polymorphisms of FSCN1 and HOTAIR on predisposition to BC would be affected by additional factors in these genes or perhaps other genes, and the assessment should be customized on a population-specific criterion.

The findings of this study suggest that FSCN1  rs852479 C > A polymorphism is implicated in BC risk and development among Egyptian women. Furthermore, CC variant of FSCN1 rs1640233 C > T has been found to be significantly associated with some BC prognostic factors, potentially worsening the prognosis for those carrying the polymorphism. Otherwise, no significant relationship between the HOTAIR rs920778 C > T polymorphism and BC risk in our patients was detected. To our knowledge, this is the first study regarding  FSCN1  rs852479 and rs1640233 polymorphisms and their association with BC susceptibility in Egyptian women. Further studies are needed to be conducted in larger patient cohorts to explore specific clinical and pathological characteristics as well as in patients from different populations.

Data availability

The data sets generated and/or analyzed over the course of the study are not publicly available but are available from the corresponding author upon reasonable request.

Abbreviations

• Breast cancer

Single nucleotide polymorphisms

HOX Transcript antisense RNA

Triple-negative breast cancer

Ahmed H, Hussein M (2022) Impacts of genetic polymorphisms on breast cancer. Int J Oncol Res 5:037. https://doi.org/10.23937/2643-4563/1710037

Article   Google Scholar  

Sung H, Ferlay J, Siegel RL et al (2021) Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 71(3):209–249. https://doi.org/10.3322/CAAC.21660

Article   PubMed   Google Scholar  

Lima SM, Kehm RD, Terry MB (2021) Global breast cancer incidence and mortality trends by region, age-groups, and fertility patterns. EClinicalMedicine 38:100985. https://doi.org/10.1016/j.eclinm.2021.100985

Article   PubMed   PubMed Central   Google Scholar  

Siegel RL, Giaquinto AN, Jemal A (2024) Cancer statistics, 2024. CA Cancer J Clin 74(1):12–49. https://doi.org/10.3322/caac.21820

Abdelaziz A, Shawki M, Shaaban A et al (2020) Breast cancer awareness among Egyptian women and the impact of caring for patients with breast cancer on family caregivers’ knowledge and behaviour. Res Oncol 17(1):1–8. https://doi.org/10.21608/RESONCOL.2020.42340.1114

Azim HA, Elghazawy H, Ghazy RM et al (2023) Clinicopathologic features of breast cancer in Egypt—contemporary profile and future needs: a systematic review and meta-analysis. JCO Glob Oncol 9:e2200387. https://doi.org/10.1200/go.22.00387

Osman Y, Elsharkawy T, Hashim TM et al (2022) Study of single nucleotide polymorphisms associated with breast cancer patients among Arab ancestries. Int J Breast Cancer 2022:2442109. https://doi.org/10.1155/2022/2442109

Article   CAS   PubMed   PubMed Central   Google Scholar  

Robert F, Pelletier J (2018) Exploring the impact of single-nucleotide polymorphisms on translation. Front Genet 9:507. https://doi.org/10.3389/FGENE.2018.00507

Ali HM, Ellakwa DES, Elaraby NM, Zaher AM, Amr KS (2023) Study the association of microRNA polymorphisms (miR-146a, miR-4513) with the risk of coronary heart diseases in Egyptian population. J Biochem Mol Toxicol 37(3):e23284. https://doi.org/10.1002/JBT.23284

Article   CAS   PubMed   Google Scholar  

Berry NK, Scott RJ, Rowlings P, Enjeti AK (2019) Clinical use of SNP-microarrays for the detection of genome-wide changes in haematological malignancies. Crit Rev Oncol Hematol 142:58–67. https://doi.org/10.1016/J.CRITREVONC.2019.07.016

Deng N, Zhou H, Fan H, Yuan Y (2017) Single nucleotide polymorphisms and cancer susceptibility. Oncotarget 8(66):110635. https://doi.org/10.18632/ONCOTARGET.22372

Van Veen EM, Brentnall AR, Byers H et al (2018) Use of single-nucleotide polymorphisms and mammographic density plus classic risk factors for breast cancer risk prediction. JAMA Oncol 4(4):476–482. https://doi.org/10.1001/JAMAONCOL.2017.4881

Li CH, Chan MH, Liang SM, Chang YC, Hsiao M (2022) Fascin-1: Updated biological functions and therapeutic implications in cancer biology. BBA Adv 2:100052. https://doi.org/10.1016/J.BBADVA.2022.100052

Liu H, Zhang Y, Li L et al (2021) Fascin actin-bundling protein 1 in human cancer: promising biomarker or therapeutic target? Mol Ther Oncolytics 20:240–264. https://doi.org/10.1016/J.OMTO.2020.12.014

Poli G, Ruggiero C, Cantini G et al (2019) Fascin-1 is a novel prognostic biomarker associated with tumor invasiveness in adrenocortical carcinoma. J Clin Endocrinol Metab 104(5):1712–1724. https://doi.org/10.1210/jc.2018-01717

Abbasi A, Noroozinia F, Anvar S, Abbasi MA, Hosseinzadeh S, Seyed Mokhtari A (2020) Fascin overexpression is associated with higher grades of breast cancer. Pol J Pathol 70(4):264–268. https://doi.org/10.5114/PJP.2019.93128

Zhao W, Gao J, Wu J et al (2015) Expression of Fascin-1 on human lung cancer and paracarcinoma tissue and its relation to clinicopathological characteristics in patients with lung cancer. Onco Targets Ther 8:2571. https://doi.org/10.2147/OTT.S81915

Piskor BM, Pryczynicz A, Lubowicka E et al (2018) Immunohistochemical expression of Fascin-1 in colorectal cancer in relation to clinical and pathological parameters. Folia Histochem Cytobiol 1(2):106–112. https://doi.org/10.5603/FHC.A2018.0011

Gupta I, Vranic S, Al-Thawadi H et al (2021) Fascin in gynecological cancers: an update of the literature. Cancers 13(22):5760. https://doi.org/10.3390/CANCERS13225760

Park SH, Song JY, Kim YK et al (2014) Fascin1 expression in high-grade serous ovarian carcinoma is a prognostic marker and knockdown of fascin1 suppresses the proliferation of ovarian cancer cells. Int J Oncol 44(3):637–646. https://doi.org/10.3892/IJO.2013.2232

Natesan SC, Ramakrishnan BP, Krishnapillai R, Thomas P (2019) Immunohistochemical expression of fascin in oral epithelial dysplasia and oral squamous cell carcinoma. World J Dent 10(5):340–345. https://doi.org/10.5005/jp-journals-10015-1658

Sarantelli E, Mourkakis A, Zacharia LC, Stylianou A, Gkretsi V (2023) Fascin-1 in cancer cell metastasis: old target-new insights. Int J Mol Sci 24(14):11253. https://doi.org/10.3390/IJMS241411253

Chang D, Liu XX, Liu R, Sun JW (2023) The role and regulatory mechanism of FSCN1 in breast tumorigenesis and progression. Yi Chuan 45(2):115–127. https://doi.org/10.16288/J.YCZZ.22-346

Wang CQ, Tang CH, Wang Y et al (2017) FSCN1 gene polymorphisms: biomarkers for the development and progression of breast cancer. Sci Rep 7(1):15887. https://doi.org/10.1038/S41598-017-16196-6

Raju GSR, Pavitra E, Bandaru SS et al (2023) HOTAIR: a potential metastatic, drug-resistant and prognostic regulator of breast cancer. Mol Cancer 22(1):1–15. https://doi.org/10.1186/S12943-023-01765-3

Jiang D, Xu L, Ni J, Zhang J, Cai M, Shen L (2019) Functional polymorphisms in LncRNA HOTAIR contribute to susceptibility of pancreatic cancer. Cancer Cell Int 19(1):1–8. https://doi.org/10.1186/S12935-019-0761-X

Pan W, Liu L, Wei J, Ge Y, Zhang J, Chen H et al (2015) A functional lncrna hotair genetic variant contributes to gastric cancer susceptibility. Mol Carcinog 55(1):90–96. https://doi.org/10.1002/mc.22261

Madadi Rad R, Pouladi N, Bosharani NN, Alizadeh A (2022) Association between HOTAIR rs1899663 G>T gene polymorphism and thyroid cancer susceptibility. J Babol Univ Med Sci 24(1):95–102

Google Scholar  

Kim JO, Jun HH, Kim EJ et al (2020) Genetic variants of HOTAIR associated with colorectal cancer susceptibility and mortality. Front Oncol 10:72. https://doi.org/10.3389/FONC.2020.00072

Lv Z, Kou C, Chen N et al (2021) Single nucleotide polymorphisms in HOTAIR are related to breast cancer risk and prognosis in the Northeastern Chinese population. Front Oncol 11:706428. https://doi.org/10.3389/FONC.2021.706428

Qi Q, Wang J, Huang B et al (2016) Association of HOTAIR polymorphisms rs4759314 and rs920778 with cancer susceptibility on the basis of ethnicity and cancer type. Oncotarget 7(25):38775–38784. https://doi.org/10.18632/ONCOTARGET.9608

Tian T, Li C, Xiao J et al (2016) Quantitative assessment of the polymorphisms in the HOTAIR lncRNA and cancer risk: a meta-analysis of 8 case-control studies. PLoS ONE 11(3):e0152296. https://doi.org/10.1371/JOURNAL.PONE.0152296

Hajjari M, Rahnama S (2019) Association between SNPs of long non-coding RNA HOTAIR and risk of different cancers. Front Genet 10:113. https://doi.org/10.3389/fgene.2019.00113

Wang B, Yuan F, Zhang F, Miao Z, Jiang D (2023) A systematic review and meta-analysis of the association between HOTAIR polymorphisms and susceptibility to breast cancer. Arch Med Sci 19(1):128–137. https://doi.org/10.5114/AOMS.2019.87537

Anber N, Tarabay MM, Elmougy R, Abdel-Dayem MA, Elbendary EY (2023) Association of HOTAIR gene rs920778 (C > T) and rs4759314 (A > G) polymorphism with breast cancer in Egyptian women. Mol Biol Rep 50(11):9153. https://doi.org/10.1007/S11033-023-08725-6

Ibrahim NE, Hamed RMR, Refaat A, Mosaad YO, Mekawy DM (2023) Genetic polymorphism in FSCN1 rs3801004 C/G and CD44 rs353639 A/C, as prognostic factor in Egyptian breast cancer patients. Asian Pac J Cancer Prev 24(10):3517–3523. https://doi.org/10.31557/APJCP.2023.24.10.3517

Miller SA, Dykes DD, Polesky HF (1988) A simple salting out procedure for extracting DNA from human nucleated cells. Nucleic Acids Res 16(3):1215–1215. https://doi.org/10.1093/NAR/16.3.1215

Kang H (2021) Sample size determination and power analysis using the G*Power software. J Educ Eval Health Prof 18:17. https://doi.org/10.3352/jeehp.2021.18.17

Solé X, Guinó E, Valls J, Iniesta R, Moreno V (2006) SNPStats: a web tool for the analysis of association studies. Bioinformatics 22(15):1928–1929. https://doi.org/10.1093/BIOINFORMATICS/BTL268

Liu Z, Ning G, Xu R, Cao Y, Meng A, Wang Q (2016) Fscn1 is required for the trafficking of TGF-β family type I receptors during endoderm formation. Nat Commun 7:12603. https://doi.org/10.1038/NCOMMS12603

Khorshidi HR, Taheri M, Noroozi R, Soudyab M, Sayad A, Ghafouri-Fard S (2017) Investigation of the association of HOTAIR single nucleotide polymorphisms and risk of breast cancer in an Iranian population. Int J Cancer Manag 10(5):7498. https://doi.org/10.5812/IJCM.7498

Bayram S, Sümbül AT, Batmacı CY, Genç A (2015) Effect of HOTAIR rs920778 polymorphism on breast cancer susceptibility and clinicopathologic features in a Turkish population. Tumour Biol 36(5):3863–3870. https://doi.org/10.1007/S13277-014-3028-0

Hassanzarei S, Hashemi M, Sattarifard H, Hashemi SM, Bahari G, Ghavami S (2017) Genetic polymorphisms of HOTAIR gene are associated with the risk of breast cancer in a sample of southeast Iranian population. Tumor Biol 39(10):1–8. https://doi.org/10.1177/1010428317727539

Article   CAS   Google Scholar  

Rajagopal T, Seshachalam A, Akshaya RL et al (2020) Association of HOTAIR (rs920778 and rs1899663) and NME1 (rs16949649 and rs2302254) gene polymorphisms with breast cancer risk in India. Gene 762:145033. https://doi.org/10.1016/J.GENE.2020.145033

Yan R, Cao J, Song C et al (2015) Polymorphisms in lncRNA HOTAIR and susceptibility to breast cancer in a Chinese population. Cancer Epidemiol 39(6):978–985. https://doi.org/10.1016/J.CANEP.2015.10.025

Bayram S, Ülger Y, Sümbül AT, Kaya BY, Rencüzoğulları A, Genç A, Sevgiler Y, Bozkurt O, Rencüzoğulları E (2015) A functional HOTAIR rs920778 polymorphism does not contributes to gastric cancer in a Turkish population: a case-control study. Fam Cancer 14(4):561–567. https://doi.org/10.1007/s10689-015-9813-0

Min KW, Chae SW, Kim DH et al (2015) Fascin expression predicts an aggressive clinical course in patients with advanced breast cancer. Oncol Lett 10(1):121. https://doi.org/10.3892/OL.2015.3191

Wang CQ, Li Y, Huang BF et al (2017) EGFR conjunct FSCN1 as a novel therapeutic strategy in triple-negative breast cancer. Sci Rep 7(1):15654. https://doi.org/10.1038/S41598-017-15939-9

Esnakula AK, Ricks-Santi L, Kwagyan J et al (2014) Strong association of fascin expression with triple negative breast cancer and basal-like phenotype in African-American women. J Clin Pathol 67(2):153–160. https://doi.org/10.1136/JCLINPATH-2013-201698

D’Amelio AM, Monroy C, El-Zein R, Etzel CJ (2012) Using haplotype analysis to elucidate significant associations between genes and Hodgkin lymphoma. Leuk Res 36(11):1359. https://doi.org/10.1016/J.LEUKRES.2012.07.014

Download references

Open access funding provided by The Science, Technology & Innovation Funding Authority (STDF) in cooperation with The Egyptian Knowledge Bank (EKB).

Author information

Authors and affiliations.

Biochemistry and Molecular Biology Department, Faculty of Pharmacy (Girls), Al-Azhar University, Cairo, Egypt

Eman Reda Galal & Lamiaa Khalaf Ahmed

Department of Clinical Pathology, Faculty of Medicine, Ain Shams University, Cairo, Egypt

Dina A. Abdelhakam

Medical Biochemistry and Molecular Biology Department, School of Medicine, Newgiza University, Giza, Egypt

Yasmine Elhusseny

Department of General Surgery, Faculty of Medicine, Beni-Suef University, Beni Suef, Egypt

Sherif El Prince Sayed

Medical Molecular Genetics Department, Human Genetics and Genome Project Institute, National Research Centre, Cairo, Egypt

Noha H. Eltaweel

You can also search for this author in PubMed   Google Scholar

Contributions

NHE, ERG, LKA, and YE: performed the molecular experiments, collected the DNA samples, and contributed to validation and interpretation of the data; ERG and NHE: made statistics and analysis of data; SEPS and DAA: participated in the management of the study subjects, examined the patients, collected biological samples, and performed clinical-pathological analysis; ERG: wrote the first draft of the manuscript, reviewed the manuscript before submission, and corresponded with the journal. All authors have read, revised, and approved the final form of the manuscript.

Corresponding author

Correspondence to Eman Reda Galal .

Ethics declarations

Competing interests.

The authors declare no competing interests.

Conflicts of interest

The authors have no relevant financial or non-financial interests to disclose.

Ethical approval

This study has been approved by the Ethics Committee of the Faculty of Pharmacy (Girls), Al-Azhar University, Cairo, Egypt (REC number: 436).

Informed consent

Written informed consent was obtained from all participants following a full explanation of the study.

Additional information

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 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

Galal, E.R., Abdelhakam, D.A., Ahmed, L.K. et al. The association of FSCN1 (rs852479, rs1640233) and HOTAIR (rs920778) polymorphisms with the risk of breast cancer in Egyptian women. Mol Biol Rep 51 , 495 (2024). https://doi.org/10.1007/s11033-024-09459-9

Download citation

Received : 22 February 2024

Accepted : 19 March 2024

Published : 08 April 2024

DOI : https://doi.org/10.1007/s11033-024-09459-9

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

• Gene polymorphism
• Find a journal
• Publish with us
• Track your research

CASE REPORT article

Sunitinib for adenocarcinoma of the rete testis: a case report.

• Department of Urology, The First Affiliated Hospital of Guangxi Medical University, Nanning, Guangxi, China

Background: Adenocarcinoma of the rete testis (AORT) is an extremely rare and aggressive tumor with a poor prognosis. Its etiology and pathological characteristics have not been extensively studied, making accurate diagnosis and appropriate management challenging. AORT, an invasive testicular tumor with a mortality rate of 46%, treatment typically involves radical orchiectomy, retroperitoneal pelvic lymph node dissection (RPLND), adjuvant chemotherapy, and/or ongoing monitoring, but the response to conventional radiation and chemotherapy is limited. At present, no effective targeted therapy for AORT has been found.

Case description: In this case report, we present the clinical scenario of a 50-year-old male patient initially diagnosed with a right testicular hydrocele, who subsequently underwent eversion of the parietal tunica vaginalis. Postoperative pathological analysis revealed metastatic clear cell renal cell carcinoma (ccRCC). PET/CT demonstrated findings suggestive of left renal upper pole carcinoma with involvement of the right scrotum, para-aortic region, bilateral iliac vessels, bilateral inguinal region, and multiple metastases. Sunitinib, a tyrosine kinase inhibitor, is commonly employed in the treatment of ccRCC. The patient underwent treatment with sunitinib for a duration of 20 months, resulting in the inactivation of multiple metastases. Following this, a radical orchiectomy was performed, and the postoperative pathology confirmed the presence of AORT. This article provides a comprehensive account of the patient's medical history, diagnostic process, treatment modalities, and subsequent follow-up observations.

Conclusions: This case report highlights the successful use of targeted therapy with sunitinib in a patient with AORT. The patient showed a positive response to targeted therapy. This study not only provides a novel foundation for the treatment of AORT, but also offers valuable insights for future treatment strategies in managing this particular form of testicular cancer.

● The successful use of targeted therapy with sunitinib in a patient with AORT and multiple lymph node metastases. What is known and what is new?

● AORT, an invasive testicular tumor with a mortality rate of 46%, treatment typically involves radical orchiectomy, RPLND, adjuvant chemotherapy, and/or ongoing monitoring, but the response to conventional radiation and chemotherapy is limited.

● This AORT patient showed a positive response to sunitinib, leading to the inactivation of all lymph node metastatic lesions and subsequent curative surgery. What is the implication, and what should change now?

● The successful response to targeted therapy in this case highlights the potential of personalized treatment approaches for rare and aggressive tumors such as AORT. However, further studies are needed to evaluate the efficacy and optimal duration of targeted therapy in AORT.

Introduction

Adenocarcinoma of the rete testis (AORT) is an extremely rare malignant tumor that arises from the rete testis, a network of tubules located in the mediastinum testis. It accounts for 1%-2% of all testicular tumors and has a poor prognosis due to its aggressive nature and limited treatment options ( 1 ). The etiology and pathological characteristics of AORT are not well understood, and its rarity often leads to misdiagnosis and delayed treatment. In this case report, we aim to provide a comprehensive overview of the diagnostic and therapeutic journey of a patient with AORT, highlighting the challenges faced and the successful outcome achieved through targeted therapy and curative surgery.

Case presentation

A 50-year-old male patient presented to a local hospital on January, 2022, with a chief complaint of "enlarged right testicle for more than 10 years." The patient denies having a family medical history, psychosocial history, or genetic history. Physical examination revealed significant enlargement of the right testicle, which was soft in consistency without palpable nodules or tenderness. Ultrasound examination indicated testicular hydrocele. Initial diagnosis was right testicular hydrocele, and the patient underwent testicular tunica vaginalis inversion surgery. The surgical procedure for the patient went smoothly. However, pathology after testicular tunica vaginalis inversion surgery suggested a high possibility of metastatic clear cell renal cell carcinoma(ccRCC). The patient was subsequently referred to our hospital for further evaluation and management.

Laboratory investigations, including tumor marker analysis, did not show elevated levels of alpha-fetoprotein (AFP), lactate dehydrogenase (LDH), carbohydrate antigen 199 (CA199), carbohydrate antigen 125 (CA125), carcinoembryonic antigen (CEA), β-human chorionic gonadotropin(β-HCG), or prostate-specific antigen (PSA). Scrotal ultrasound and computed tomography (CT) scans revealed a mass in the right scrotum, raising suspicion of malignancy ( Figure 1 ).

Figure 1 Scrotal ultrasound before targeted therapy (A) , scrotal ultrasound after targeted therapy (B) , scrotal CT before targeted therapy (C) , scrotal CT after targeted therapy (D) .

Further imaging with positron emission tomography combined with computed tomography (PET/CT) demonstrated findings suggestive of left renal upper pole carcinoma with involvement of the right scrotum, para-aortic region, bilateral iliac vessels, bilateral inguinal region, and multiple metastases ( Figure 2 ).

Figure 2 PET/CT imaging: Testicular before targeted therapy (right), after targeted therapy (left) (A) , Inguinal lymph nodes before targeted therapy (right), after targeted therapy (left) (B) .

This patient and his family refused percutaneous renal biopsy or resection of the left kidney tumor. Based on the pathological results and imaging findings, the final diagnosis was determined as metastases ccRCC. After thorough communication with the patient and family, a decision was made to initiate targeted therapy with sunitinib, a tyrosine kinase inhibitor known to inhibit tumor growth and angiogenesis. The treatment regimen consisted of 4 weeks of medication followed by a 2-week drug-free period, and continued for 18 months.

During the course of targeted therapy, the patient underwent regular follow-up imaging studies to monitor the response to treatment. Gradual reduction in the size of the right testicular mass was observed on the imaging studies, indicating a positive response to sunitinib. A subsequent PET/CT scan revealed the following results after targeted therapy: A) The tumor activity in the right scrotal lesion was still present but showed improvement compared to previous scans; B) The previously identified metastatic lymph nodes showed no tumor activity; C) No new lesions were observed ( Figure 2 ). These findings suggested a favorable response to targeted therapy, paving the way for curative surgery.

The patient discontinued sunitinib in August 2023 and underwent curative surgery for right testicular cancer in November 2023. The patient did not undergo kidney surgery because PET/CT indicated that the left renal lesion had disappeared. The postoperative pathological evaluation indicated: 1) Both tumors were of the same origin; 2) Considering the clinical and imaging findings, excluding metastasis from ccRCC, the diagnosis was determined as AORT ( Figure 3 ). The timeline with relevant data as Table 1 .

Figure 3 The visible component in the image resembles adenocarcinoma of the rete testis (AORT).

Table 1 Immunohistochemical staining results.

AORT is an extremely rare disease, with fewer than 80 cases reported in the literature since its initial description in 1945 ( 2 ). The etiology and pathological characteristics of AORT remain poorly understood, making accurate diagnosis and appropriate management challenging. The majority of reported cases occur in Caucasian patients, and the tumor primarily affects middle-aged to elderly individuals ( 1 ).

The exact cause of AORT is still unclear, although some studies have proposed associations between tumor occurrence and various conditions and chemical exposures. These include hydrocele, chronic epididymitis or orchitis, testicular trauma, and exposure to certain chemicals such as asbestos and heavy metals ( 3 – 6 ). However, further research is needed to establish a definitive link between these factors and the development of AORT.

The diagnosis of AORT is often challenging due to its rarity and nonspecific clinical presentation. Patients may present with symptoms such as testicular enlargement, pain, or a palpable mass ( 2 ). However, these symptoms can also be seen in other testicular conditions, leading to misdiagnosis or delayed diagnosis.

Cases reported by Kitano et al. showed that serum CA 199 levels increase with the progression of AORT metastasis ( 7 ). There have also been reports of primary AORT associated with elevated serum CEA ( 8 ). However, as for AORT, there are currently no definitive tumor biomarkers. In this case, the patient’s AFP, β-HCG, CEA, CA 199, and other markers were within the normal range, with the only elevation being in serum ferritin. To date, there is no research demonstrating that serum ferritin can serve as a biomarker for AORT, and its role requires further investigation.

Imaging studies, including scrotal ultrasound and CT scans, play a crucial role in the initial evaluation of patients with suspected AORT. PET/CT scans are useful in identifying metastatic lesions and assessing treatment response.

The diagnosis of AORT is typically confirmed by postoperative pathology ( 9 ). Testicular gate localization, the transition from benign to malignant epithelium, and supportive immunostaining contribute to its accurate diagnosis ( 1 ). Research has shown that AE 1/AE 3 cytokeratin is the most consistent positive stain, followed by EMA,and vimentin ( 10 , 11 ). In this case, the patient’s immunohistochemical results indicated EMA (+),and vimentin (+) ( Table 2 ),consistent with the pathological diagnosis of AORT.

Table 2 The timeline with relevant data.

The management of AORT is complex and requires a multidisciplinary approach. Due to the rarity of the disease, there is no standard treatment protocol, and management decisions are often based on individual cases and expert opinions. Surgical resection, including radical orchiectomy, is the mainstay of treatment for localized disease. However, in cases of advanced or metastatic disease, systemic therapy options such as chemotherapy, targeted therapy, or immunotherapy may be considered.

AORT, a rare and invasive testicular tumor with a mortality rate of 46%, requires a multidisciplinary approach ( 1 ). Treatment typically involves radical orchiectomy, retroperitoneal pelvic lymph node dissection (RPLND), adjuvant chemotherapy, and/or ongoing monitoring, but the response to conventional radiation and chemotherapy is limited ( 2 ). In addition, AORT tends to metastasize to retroperitoneal lymph nodes, some cases undergoing RPLND have shown better outcomes, but this difference is not statistically significant ( 12 ). As far, it remains unclear whether RPLND changes patient outcomes. Due to the rarity of the disease, there is no standard treatment protocol. Paclitaxel, ifosfamide, and cisplatin are potential new drugs for treating AORT ( 13 ). Studies suggest that platinum-based chemotherapy may be effective against AORT, the patient with AORT who received bleomycin, etoposide, and platinum treatment had no further enlargement of metastatic lymph nodes after 7 months of discontinuing chemotherapy, and the patient remained alive at 20 months after discontinuation of chemotherapy ( 14 ).

In the case presented, the patient underwent targeted therapy with sunitinib, a tyrosine kinase inhibitor that has shown efficacy in the treatment of various solid tumors. Sunitinib inhibits multiple receptor tyrosine kinases, including vascular endothelial growth factor receptors (VEGFRs), platelet-derived growth factor receptors (PDGFRs), and stem cell factor receptor (KIT), thereby inhibiting tumor growth and angiogenesis ( 15 ). The patient showed a positive response to sunitinib, with significant reduction in the size of the testicular mass and complete inactivation of metastatic lymph nodes.

Analysis of existing data shows a median survival time of 33 months for patients with AORT. The 3-year and 5-year survival rates are 45% and 20%, respectively, with a 3-year survival rate of 18% for metastatic patients ( 12 ). In this case, the patient’s scrotal mass decreased after treatment with sunitinib, and the tumor activity in the metastatic lesions disappeared. Previous studies have shown that tumor confinement to the testicle is the most important prognostic factor for survival ( 12 ). The conditions for performing radical surgery for testicular tumors were created for the patient, and the patient did not experience any significant discomfort or dissatisfaction during the treatment process.

The successful response to targeted therapy in this case highlights the potential of personalized treatment approaches for rare and aggressive tumors such as AORT. Due to the inherent limitations of a case report, including potential biases and the lack of a comparison with a control group, the findings of this study have limited generalizability. Therefore, further studies are needed to evaluate the efficacy and optimal duration of targeted therapy in AORT. Additionally, the long-term outcomes and potential side effects of targeted therapy in AORT patients require careful monitoring and investigation.

This case report highlights the successful use of targeted therapy with sunitinib in a patient with AORT and multiple lymph node metastases. The patient showed a positive response to targeted therapy, leading to the inactivation of all lymph node metastatic lesions and subsequent curative surgery. This study not only provides a novel foundation for the treatment of AORT, but also offers valuable insights for future diagnoses and treatment strategies in managing this particular form of testicular cancer.

Data availability statement

The original contributions presented in the study are included in the article/supplementary material. Further inquiries can be directed to the corresponding author.

Ethics statement

The studies involving humans were approved by First Affiliated Hospital of Guangxi Medical University ethics board. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study. Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.

Author contributions

KL: Writing – original draft, Writing – review & editing. DC: Writing – original draft, Writing – review & editing. MH: Writing – original draft, Writing – review & editing. JY: Writing – original draft, Writing – review & editing. HM: Writing – original draft, Writing – review & editing.

The author(s) declare that no financial support was received for the research, authorship, and/or publication of this article.

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.

1. Al-Obaidy KI, Idrees MT, Grignon DJ, Ulbright TM. Adenocarcinoma of the rete testis. Am J Surg Pathol . (2019) 43:670–81. doi: 10.1097/PAS.0000000000001219

PubMed Abstract | CrossRef Full Text | Google Scholar

2. Mitchell CF, Pearce S. Adenocarcinoma of the rete testis: A case report. Cureus . (2022) 14(2):e22210. doi: 10.7759/cureus.22210

3. Stein JP, Freeman JA, Esrig D, Chandrasoma PT, Skinnerner DG. Papillary adenocarcinoma of therets testis: A case report and review of the literature. Urol . (1994) 44:588–94. doi: 10.1016/S0090-4295(94)80067-7

4. Spataro V, Caldiera S, Rusca T, Sessa C, Cavalli F. Adenocarcinoma of the rete testis. J Urol . (2000) 164:1307–8. doi: 10.1016/S0022-5347(05)67169-8

5. Rubegni P, Poggiali S, De Santi M, Stefania M, Bilenchi R, Miracco C, et al. Cutaneous metastases from adenocarcinoma of the rete testis. J Cutaneous Pathol . (2006) 33:181–4. doi: 10.1111/j.0303-6987.2006.00395.x

CrossRef Full Text | Google Scholar

6. Amin MB. Selected other problematic testicular and paratesticular lesions: rete testis neoplasms and pseudotumors, mesothelial lesions and secondary tumors. Modern Pathol . (2005) 18:S131–S45. doi: 10.1016/S0893-3952(22)04463-5

7. Kitano H, Sentani K, Goto K, Sekino Y, Yamanaka R, Nagasaka K, et al. Primary adenocarcinoma of the rete testis with elevated serum CA19-9 antigen levels. Int Cancer Conf J . (2020) 9:240–3. doi: 10.1007/s13691-020-00437-z

8. Takeda S, Yamashita S, Kawamorita N, Mitsuzuka K, Fujishima F, Watanabe M, et al. Primary adenocarcinoma of the rete testis with elevation of serum carcinoembryonic antigen: a case report. Int Cancer Conf J . (2017) 6:84–7. doi: 10.1007/s13691-017-0280-3

9. Wu C-A, Chen Y-H, Man K-M, Shen J-L, Chen W-C. Papillary adenocarcinoma of rete testis mimics inflammatory lump: A case report. Case Rep Urol . (2011) 2011:1–3. doi: 10.1155/2011/857812

10. Huang PW, Chang KM. Adenocarcinoma of the rete testis with prominent papillary structure and clear neoplastic cells: Morphologic and immunohistochemical findings and differential diagnosis. Indian J Pathol Microbiol . (2015) 58(2):232–4. doi: 10.4103/0377-4929.155328

11. Glazier DB, Vates TS, Cummings KB, Antoun S. Adenocarcinoma of the rete testis. World J Urol . (1996) 14(6):397–400. doi: 10.1007/BF00183123

12. Sanchez-Chapado M, Angulo JC, Haas GP. Adenocarcinoma of the rete testis. Urol . (1995) 46:468–75. doi: 10.1016/S0090-4295(99)80257-X

13. Chovanec M, Mego M, Syčová-Milá Z, Obertová J, Rajec J, Palacka P, et al. Adenocarcinoma of the rete testis – a rare case of testicular Malignancy. Klinicka Onkologie . (2014) 27:136–7. doi: 10.14735/amko2014136

14. Owa S, Sasaki T, Uchida K, Watanabe S, Kato M, Sugino Y, et al. A case of adenocarcinoma of the rete testis with durable response to cisplatin-based chemotherapy. IJU Case Rep . (2021) 4:267–71. doi: 10.1002/iju5.12298

15. Jin J, Xie Y, Zhang J-S, Wang J-Q, Dai S-J, He W-f, et al. Sunitinib resistance in renal cell carcinoma: From molecular mechanisms to predictive biomarkers. Drug Resist Updat . (2023) 67:100929. doi: 10.1016/j.drup.2023.100929

Keywords: case report, sunitinib, AORT, diagnostic, targeted therapy

Citation: Li K, Chen D, He M, Yu J and Mi H (2024) Sunitinib for adenocarcinoma of the rete testis: a case report. Front. Oncol. 14:1358133. doi: 10.3389/fonc.2024.1358133

Received: 19 December 2023; Accepted: 25 March 2024; Published: 08 April 2024.

Reviewed by:

Copyright © 2024 Li, Chen, He, Yu and Mi. 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: Hua Mi, [email protected]

† These authors have contributed equally to this work

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.

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
• Ann Med Surg (Lond)
• v.85(6); 2023 Jun
• PMC10289736

The effect of multidisciplinary team on survival rates of women with breast cancer: a systematic review and meta-analysis

Eko a. pangarsa.

a Hematology Medical Oncology Division, Department of Internal Medicine, Faculty of Medicine, Diponegoro University/Dr. Kariadi General Hospital Semarang, Indonesia

Daniel Rizky

Kevin tandarto, budi setiawan, damai santosa, jessica n. hadiyanto.

b Department of Internal Medicine, Dr. Kariadi General Hospital/Faculty of Medicine, Diponegoro University, Indonesia

Salma Kyana

Catharina suharti, background:.

Breast cancer is quite frequent all around the world. This disease was responsible for an estimated 2.1 million malignancies in 2022, making it the seventh-highest cause of cancer deaths globally. A multidisciplinary team (MDT) care policy was developed in the United Kingdom (UK) in 1995 to enhance the quality of care for cancer patients. The purpose of this systematic review and meta-analysis study is to assess the effects of MDT on breast cancer survival rates.

This study was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020. Systematic search was conducted in several international databases including Google Scholar, PubMed, EBSCOhost, and Proquest from 2012 to 2022. The authors used RevMan 5.4 to do the meta-analysis of the pooled hazard ratio. Newcastle–Ottawa Scale to measure the risk of bias. Newcastle–Ottawa Scale evaluated participant selection, comparability, and reporting of results using eight subscale items. Egger’s test funnel plot was used to assess the potential publication bias for this study.

A total of 1187 studies were identified from research database. The authors found a total of six studies from six different countries (China, the UK, Taiwan, Australia, Africa, and France) included for this study. Based on the meta-analysis of the pooled hazard ratio of the included studies, the authors found that the overall effect size of the study was 0.80 (CI 95%: 0.73–0.88).

Conclusions:

Breast cancer patients who participated in well-organized MDT discussions had a greater survival rate than those who did not.

Introduction

• To the best of our knowledge, no systematic review and meta-analysis study has yet been conducted to assess the effect of multidisciplinary team (MDT) on breast cancer survival rate.
• Breast cancer is quite frequent all around the world. It was responsible for an estimated 2.1 million malignancies in 2018, making it the seventh-highest cause of cancer deaths globally.
• MDT treatment can prevent 98.8% of all drug mistakes and enhance overall care quality.
• We included the pooled odds ratio to see the survival of breast cancer patients in the MDT and non-MDT groups.
• Funnel plot with slightly asymmetry.

Breast cancer is quite frequent all around the world. It was responsible for an estimated 2.1 million malignancies in 2022, making it the seventh-highest cause of cancer deaths globally 1 , 2 . This disease affects one in every nine women in industrialized nations and one in every 20 in less developed countries with 2.3 million new cases diagnosed, surpassed lung cancer as the most prevalent cancer in the world in 2020 3 . About 45.4% of the 2.3 million breast cancer cases diagnosed in 2020 were in Asia 4 .

Survival rates of breast cancer differ globally, with higher survival rates in developed compared to less developed countries 5 , 6 . For instance, the 5-year survival rate in developed countries such as the United States of America and the United Kingdom (UK) were 85–90% between the years 2017–2019 7 . In developing countries this rate ranged between 40–60%. The higher mortality of breast cancer in developing countries could be related to low awareness of screening needs, a lack of early detection programs, and a lack of diagnosis and treatment facilities 8 – 11 .

A multidisciplinary team (MDT) care policy was developed in the UK in 1995 to enhance the quality of care for cancer patients. MDT comprises a variety of professions, including medical, nursing, and allied workers, as well as diagnostic experts, who work together to identify the best treatment plan for each patient 12 – 15 . Previous research indicates that MDT care can aid in clinical decision-making. MDT treatment can prevent 98.8% of all drug mistakes and enhance overall care quality 12 , 16 . After the introduction of multidisciplinary care in the UK, breast cancer mortality in the intervention region was 18% lower than in the nonintervention area 17 .

Even though a few studies on the effects of MDT on the study about survival rate in breast cancer patients have been published, there is still a knowledge gap in this subject. There are relatively few data on the role of MDT in breast cancer survival 18 . To the best of our knowledge, no systematic review and meta-analysis study has yet been conducted to assess the effect of MDT on breast cancer survival rate. The purpose of this systematic review and meta-analysis study is to assess the effects of MDT on breast cancer survival rates.

This study was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 and AMSTAR 2 (Assessing the Methodological Quality of Systematic Review 2) 19 – 21 . The AMSTAR 2 scores for this study high-quality. Figure ​ Figure1 1 shows the selection procedure for the studies.

PRISMA 2020 Flow Diagram Showing the Different Phases Involved in Searching for Relevant Publication to Assess the Effect of MDT on Survival Rate of Women With Breast Cancer.

Search strategy

Systematic search was conducted in several international databases including Google Scholar, PubMed, EBSCOhost, and Proquest from 2012 to 2022. The search strategy was a combination of keywords based on medical subject headings (MeSH) and free texts including: ‘Multidisciplinary team’ or ‘interprofessional team’ or ‘multiprofessional team’ or ‘interdisciplinary team’) AND (‘Breast Neoplasms’[MeSH] or ‘breast cancer’ or ‘breast carcinoma’ or ‘breast tumor’ or ‘breast malignant’ or ‘breast neoplasm’) AND (‘Survival’ [MeSH] or ‘Survival analysis’ or ‘Survival Rate’ or ‘Kaplan-Meier’ or ‘Proportional Hazard Model’ or ‘Hazard Ratio’ or ‘Cox Model’ or ‘Cox Regression’) AND (‘Cohort’ or ‘Prospective’ or ‘Retrospective’ or ‘Follow-up’ or ‘Longitude’.

Eligibility criteria

The search strategy exported every result into Mendeley reference manager. The titles and abstracts of the research were used to examine the papers after duplicate studies were eliminated. We included observational retrospective or prospective cohort studies if they were: published in English and Indonesian language; estimated the survival rate in women with breast cancer; and investigated the role of a MDT in breast cancer treatment. We contacted the authors whenever additional information was required.

Studies that did not report survival rates or did not provide enough data, as well as those that reported survival rates after relapse, were excluded. Review articles, letters to the editor, and case report studies were also excluded from this study.

Study selection

The systematic search results were input into the reference management software, and after excluding duplicate articles, two reviewers (K.T. and J.N.H) independently reviewed the retrieved articles by title, abstract, and full text of records based on eligibility criteria. Disagreements between the two reviewers were settled by consensus and, if necessary, by a third party (S.K.).

Quality assessment

We used the Newcastle–Ottawa Scale to measure the risk of bias. Newcastle–Ottawa Scale evaluated participant selection, comparability, and reporting of results using eight subscale items. Cohort studies use up to nine points out of the total number of subscale questions. K.T. and J.N.H. made this critical judgment 22 . Discussions were made to settle disagreements. Without an agreement, the third reviewer (S.K.) viewpoint was included in the final decision in this study.

Data extraction

Three reviewers (K.T., J.N.H., and S.K.) extracted data independently using predefined sheets that included the following information: first author name, year of publication, study period, country, income of the country, sample size, definition of MDT, staging of breast cancer, study design, median follow-up time, mean age with SD, and 1-, 2-, 3-, 5-, 10-, survival rates with proportions variable.

Statistical analysis

We used RevMan 5.4 to calculate the overall effect size of MDT on breast cancer survival rates. We chose a 95% CI, which indicated that survival rates of the MDT group would be statistically significant if the probability ( P ) was less than 0.05 (5%) and the CI did not cross the middle line (or 0 value). To create the data extraction sheet, we used Microsoft Excel 2013.

To measure between-study heterogeneity, we utilized the I 2 Index to calculate the fraction of total variance due to between-study variation, as well as the χ 2 test at the 10% significant level ( P <0.1). Given the study heterogeneity, we utilized the Der-Simonies and Laird random-effect models to compute the pooled hazard ratio of survival rates of MDT in women with breast cancers. We used Egger’s test funnel plot to assess the potential publication bias for this study.

Ethical clearance

The authors of this article have not undertaken any human or animal studies. All studies carried out adhered to the ethical guidelines outlined in each case.

Study characteristics

A total of 1187 studies were identified from the research database (Table ​ (Table1). 1 ). After removing the duplicate records, 967 records were screened, and 50 were assessed for eligibility. We found a total of six studies from six countries (China, the UK, Taiwan, Australia, Africa, and France) included for this study 12 , 16 , 17 , 23 – 25 . All the studies had a low risk of bias. The six included studies involved 87 057 women diagnosed with breast cancer from various demographics.

Summaries of the included studies

The definition of MDT

Brandao et al. 25 defined the MDT meetings were performed on a weekly basis and lasted 1 h, with the participation of at least one member from the surgery, oncology, pathology, and radiology departments and with trainees from these different specialties. Kesson et al. also defined an MDT team comprised of specialist breast cancer surgeons, pathologists, oncologists, radiologists, and specialist nurses, worked to evidence-based guidelines. Meeting of the MDT usually held weekly formal meetings to discuss results and agree on adjuvant treatment for individual patients, audited clinical activity and reported results at regular intervals, and lead clinicians from each team met regularly with the director of public health to discuss audit results throughout the area. The aim of the meeting is to minimize deviations from guidelines and variations in practice and improve the quality of care 12 . Other studies also defined MDT as the same meaning.

The effect of MDT in survival rate of breast cancer patients

Prior to the introduction of multidisciplinary treatment in 1995, breast cancer mortality was greater in the intervention region than in the nonintervention group (hazard ratio 1.11, 95% confidence range 1.00–1.20) 23 . After the intervention was implemented, mortality in the intervention region was much lower than in the nonintervention group (0.82, 0.74–0.91). The MDT treatment group had a considerably reduced recurrence rate than the non-MDT care group (HR, 0.84; 95% CI: 0.70–0.99, P <0.05). The MDT care group had a substantially decreased relative risk of death than the non-MDT care group (HR, 0.89; 95% CI: 0.82–0.96) 24 .

Noncompliance with MDT guidelines was associated with worse disease outcomes, regardless of whether the noncompliance was for adjuvant chemotherapy, radiation, endocrine treatment, or targeted therapy. Thus, efforts to promote compliance with MDT guidelines for breast cancer patients may aid in improving their outcomes 16 .

According to one study, implementing an MDT after educating a few critical health providers resulted in a substantial reduction in mortality among patients with early breast cancer. After controlling for other prognostic variables, this advantage remained substantial. Well-organized MDT patients outlive the non-MDT group (log-rank test, P =0.013), but disorganized MDT patients had the reverse effect (log-rank test, P =0.001) 17 , 23 . Significantly more breast cancer patients who presented to an MDT got surgery alone, in combination with a systemic therapy, or in combination with all three treatment options ( P <0.01). Tsai et al. also found that after adjusting for demographic characteristics, charlson comorbidity index score, monthly salary, urbanization level, cancer stage, hospital ownership, treatment modality, and physician’s service, the Cox proportional hazards model was used to analyze the relative risk of death between the MDT care and non-MDT groups, as well as between the recurrence and non-recurrence groups 17 .

Disease-free survival in breast cancer patients with MDT

Brandao et al. found that the 3-year disease-free survival rate in the pre-MDT group was 41.7% (95% CI: 30.2–52.8) and 56.8% (95% CI: 45.3–66.8) in the post-MDT group. The proportion of patients with loco-regional relapses (with or without concurrent distant relapse) or death as the first disease-free survival event was higher in the pre-MDT group compared to the post-MDT group (29 vs. 18%; and 23 vs. 10%, respectively), but these differences were not statistically significant ( P =0.07) 25 .

Stage, tumor subtype, and type of first-line chemotherapy treatment were substantially linked with overall survival in both the general population and patients with early breast cancer. Age above 70 years old was related with a higher probability of noncompliance than age under 50 years old (OR 1.68, 95% CI 1.21–2.17, P <0.001) 17 , 23 , 24 .

Meta-analysis of the included studies

The forest plot of included studies for meta-analysis of the effect of MDT on breast cancer survival analysis are presented in Figure ​ Figure2. 2 . Based on the meta-analysis of the pooled hazard ratio of the included studies, we found that the overall effect size of the study was 0.80 (CI 95%: 0.73–0.88). The heterogeneity was 53%, which means that these studies had moderate heterogeneity between the included studies. The P -value of the study was significant ( P <0.00001).

Forest plot of the included studies.

The Egger’s test funnel plot of the included study is presented in Figure ​ Figure3. 3 . Based on the result, we concluded that there was a potential risk of bias for this meta-analysis study (asymmetry of the scatter plot in the triangle).

Funnel plot of the included studies.

Risk of bias

Table ​ Table2 2 shows the quality assessment of the included retrospective cohort studies. Most retrospective cohort studies had sufficient cohort selection, comparability, and outcome evaluation.

Quality assessment of the included studies

Compared to similar patients treated in neighboring areas over the same period, the introduction of teams offering multidisciplinary care for the treatment of breast cancer was related to 18% reduced breast cancer mortality at 5 years and 11% lower all-cause mortality at 5 years 26 , 27 . The implementation of this strategy was also linked to a decrease in the number of hospitals providing treatment for patients with breast cancer, a narrowing of the survival rate gap across hospitals, and a reallocation of breast cancer care away from the facilities with the worst survival rates 28 . Patients with breast cancer who received MDT treatment had a considerably lower death risk. Although multidisciplinary treatment should, on its face, be linked to more remarkable survival, there is still a shortage of data to support these results 29 – 31 . The underlying processes of MDT treatment, including team focus, improved structure and procedure, and increased team performance, were examined in a few studies, leading to superior survival rates and patient satisfaction outcomes 26 , 28 .

There are other meta-analysis studies that investigated the role of MDT in other type of cancer. In a study by Peng et al. , it was observed that in patients with colorectal cancer, the MDT group had a higher rate of overall survival than the non-MDT group (HR=0.81, 95% CI 0.69–0.94, P =0.005) 32 . The MDT group was linked to a better overall survival in a subgroup analysis of stage IV colorectal cancer (HR=0.73, 95% CI 0.59–0.90, P =0.004). However, there was no discernible difference in postoperative mortality between the MDT and non-MDT groups (OR=0.84, 95% CI 0.44–1.61, P =0.60). Study by Shang et al. , also found that with moderate heterogeneity ( I 2 =68%, P =0.01), exposure groups of patients with head and neck cancer treated with MDT showed a higher survival rate (Hazard ratio=0.84, 95% CI (0.76–0.92), P =0.0004) 33 . Our meta-analysis study also discovered that the MDT group had a higher overall survival rate for patients with breast cancer.

The most crucial components of MDT care are team members, evidence-based recommendations, routine formal meetings, and individualized treatment plans. The MDT program sometimes struggled with issues such as an overwhelming caseload, low MDT meeting attendance, poor collaboration, a lack of leadership, job ambiguity, and a disregard for holistic requirements 34 . According to previous research, the majority of MDT are made up of pathologists (84%), radiologists (73%), radiation oncologists (90%), medical oncologists (95%), surgical oncologists (95%), and specialized nurses (49%). Most nations (82%) have MDT sessions once every week 35 .

The impact of MDT care on patients with breast cancer varied among these trials. The many MDT settings, such as teamwork, performance, and leadership, might be the primary culprit 23 . Prior research emphasized the necessity for high-quality markers to gage the effectiveness of MDT. MDT care models and structures are effective based on research. Therefore, we suggest that MDT be incorporated into future strategies to lower the risk of breast cancer recurrence 36 .

Multiple experts participate in MDT, allowing for the easy discussion of opposing viewpoints. A meeting can manage difficult choices and clear the path for progress. MDT give people a place to exchange information and responsibilities, enhance service delivery, foster better communication, and increase understanding of one another. Continually employing a diverse approach to problem-solving lowers treatment decision isolation 37 . Research in the UK that looked at 370 MDT for chronic diseases concluded that these sessions had various methods and arrangements. They discovered that patients who lived in underprivileged regions had a lower likelihood of having the MDT’s decisions implemented 37 – 39 .

A systematic review identified factors associated with poor-quality decision-making in MDT, including time constraints, an excessive caseload, low MDT meeting attendance, poor teamwork, and a lack of leadership. Teamwork that is dysfunctional can be caused by poor communication within the team and role ambiguity (such as a poor definition or understanding of roles within the team) 40 . The effectiveness of decision-making in MDT meetings has been a specific area of attention for several studies. In a variety of tumor types, several studies have looked at how final treatment plans compared to MDT recommendations. Up to 15% of instances saw a discrepancy in the actual treatments. The main causes were a failure to take into account all of the patient’s information, including comorbidities and treatment preferences. This could be brought on by a number of problems, such as failing to gather this data from patients prior to MDT discussions, not having enough time to prepare and/or present properly in meetings, or failing to include nurses’ input in MDT discussions 41 .

Breast cancer patients treated with MDT may not always have favorable outcomes. To optimize results, MDT should be well-organized and involve multiple disciplines, including surgeons, medical oncologists, imaging doctors, and pathology doctors. Detailed information should be delivered before the meeting, and the number of patients discussed should be limited. The discussion time per patient should be at least 20–30 min. Lu et al. demonstrated that breast cancer patients who were treated with disorganized MDT may even have worse survival outcomes compared to those who do not receive MDT at all. In the UK, a number of methods have been created and evaluated to rate the effectiveness of teamwork in cancer teams. They include a team self-assessment tool that enables anonymous team member self-assessment of teamwork over the entire pathway as well as independent observational tools to evaluate teamwork in MDT sessions. This latter tool is a part of the MDT-FIT (Feedback for Improving Teamworking) evidence-based team-improvement initiative, which was created on behalf of the UK National Cancer Action Team 41 . It is an assessment-and-feedback process that gives teams the space to reflect on how they are operating as a team and select actions for improvement. It is based on input and testing with over 100 MDTs. The review of 10 breast cancer teams within a big cancer network is currently in its final stages. Teams normally recommend six to eight areas for improvement, and the majority are put into practice within 6–9 months 40 , 41 .

Breast cancer is a highly complex disease that is influenced by many variables such as staging, tumor characteristics, tumor size, vascular invasion, nodal involvement, and socioeconomics, all of which affect patient outcomes. According to previous research, individuals with breast cancer who had low levels of education and socioeconomic status in their neighborhoods fared 1.4–2.7 times worse overall than those with high levels of education, socioeconomic status, and neighborhood. According to another study, low socioeconomic level people had a considerably greater mortality risk than high socioeconomic level people (HR, 1.08; 95% CI: 1.05–1.11) 42 . According to earlier research, the 5-year survival rate fell as breast cancer progressed (from 97.5 to 18.4%) 43 – 45 . Similarly, a prior study found that from stage I to stage IV, the 5-year survival rate was 98–23.4%. From the stage I through stage IV of breast cancer, there was a 61.61–5.11% 5-year survival rate 46 . In this systematic review and meta-analysis study, we were unable to assess the effects of these variables as not all of the included studies provided subgroup analysis of these variables on patient outcomes. Lu et al. and Kesson et al. did not mention the analysis of those variables with MDT approaches 12 , 20 . Yang et al. found that after adjusting for factors such as tumor size, histological grade, axillary lymph node status, lymphovascular invasion, estrogen receptor, progesteron receptor , and Ki-67 labeling index through multivariable analysis, subjects who did not included in the MDT approach had a higher risk of significantly recurrence (HR 1.50, 95% CI 1.03–2.18, P <0.033) 21 . Tsai et al. found that after adjusting socioeconomic variables, cancer stage, and treatment modality, breast cancer patients who received MDT care had a significantly lower risk of recurrence and mortality 17 . Brandao et al. found that despite of the superiority of the MDT approach in the overall population, this method also showed survival benefits in the early stage of breast cancer, stage III, histological grade II–III, estrogen receptor positive, and HER2 negative subtype population. Although there was no survival benefits with the introduction of the MDT approach in patients with metastatic breast cancers 22 . According to Rogers et al. , after taking into account factors such as tumor stage, comorbidities, age, and treatment received, there was no significant difference in survival between breast cancer patients who were presented to an MDT before treatment and those who were not (HR 1.84, 95% CI 0.91–3.74, P <0.09) 16 .

Although MDT discussion has been suggested as the best approach to care for cancer patients and has gained widespread support, there is insufficient proof of their implementation in Low and Middle-Income Countries (LMIC). Almost all cancer patients and their relatives in LMIC suffer some monetary hardship during the phases of cancer diagnosis and treatment. Financial toxicity is the acronym for the negative impact that an excessive financial burden brought on by a cancer diagnosis has on the health of patients, their families, and society as a whole. In LMIC, like in Indonesia, the affordability of patients is an essential factor in receiving medical care 16 , 47 . MDT may be a valuable mechanism and a central theme in LMICs where discussions of financial situation, disease severity, and treatment costs can spark heated debate and controversy. Each participant will represent their area of expertise while discussing each patient’s case based on clinical merit in this tumor board 47 .

However, from a public health standpoint, this might provide problems with access and resources. There are intricate relationships between MDT presentation patterns and outcomes across tumor streams 12 , 35 , 42 . With all regards to improving cancer care, implementing MDT achieves better outcomes. This could be achieved by implementing the national policy of MDT for cancer care. Proactively discover healthcare innovations and use virtual MDT practices to improve patient outcomes. To improve skills and raise knowledge of MDT programs, consider making participation in at least one MDT meeting a benchmark for oncology trainee 47 , 48 . Therefore, to achieve optimal outcome in cancer care, the European Partnership for Action Against Cancer (EPAAC) identified MDT as a key element in cancer care 48 .

The limitation of this study is that it only included six studies, which may not be representative of the general population of breast cancer patients. Further research is needed, especially in RCT models, to see the effect of MDT on the prognosis of patients with breast cancer. We cannot assess all the variables that influence the success of MDT in breast cancer patients in this systematic review and meta-analysis study.

Conclusions

We concluded that breast cancer patients who participated in well-organized MDT discussions had a greater survival rate than those who did not. This leads to the conclusion of MDT implementation should be encouraged in all countries moreover in LMIC.

Ethical approval

This research did not involve human subjects; therefore, it was exempt from ethical clearance.

Informed consent

Not applicable.

Sources of funding

This study received no funding.

Author contributions

This work was carried out in collaboration among all authors. E.A.P., D.R., and C.S. contributed to the conception of the review and interpreted the literatures based on the level of evidence and revised the manuscript. J.N.H., K.T., and S.K. participate in reviewing preparation of the manuscript. B.S. and D. S. participate in preparation and critical review of the manuscripts. In addition, all authors read and approved the manuscript.

Declaration of competing interest

Registration of research studies.

• Name of the registry: Research Registry.
• Unique Identifying number or registration ID: researchregistry8933.
• Hyperlink to your specific registration (must be publicly accessible and will be checked): https://www.researchregistry.com/browse-the-registry#home/registrationdetails/6450bfdfe7a6cb00289ea30f/ .

Eko Adhi Pangarsa, Division of Hematology Medical Oncology, Department of Internal Medicine, Dr. Kariadi General Hospital/Faculty of Medicine, Diponegoro University. E-mail: .moc.liamg@09asragnapoke

Provenance and peer review

Not commissioned, externally peer reviewed.

Acknowledgments

The authors want to thank you Mika L. Tobing, M.D, Suyono, MD, Ika Kartiyani, M.D, Vina Yunarvika, M.D, and Desta Nur Ewika, M.D for helping this research.

Sponsorships or competing interests that may be relevant to content are disclosed at the end of this article

Published online 25 May 2023