correlational study research paper

Have a language expert improve your writing

Run a free plagiarism check in 10 minutes, automatically generate references for free.

• Knowledge Base
• Methodology
• Correlational Research | Guide, Design & Examples

Correlational Research | Guide, Design & Examples

Published on 5 May 2022 by Pritha Bhandari . Revised on 5 December 2022.

A correlational research design investigates relationships between variables without the researcher controlling or manipulating any of them.

A correlation reflects the strength and/or direction of the relationship between two (or more) variables. The direction of a correlation can be either positive or negative.

Table of contents

Correlational vs experimental research, when to use correlational research, how to collect correlational data, how to analyse correlational data, correlation and causation, frequently asked questions about correlational research.

Correlational and experimental research both use quantitative methods to investigate relationships between variables. But there are important differences in how data is collected and the types of conclusions you can draw.

Prevent plagiarism, run a free check.

Correlational research is ideal for gathering data quickly from natural settings. That helps you generalise your findings to real-life situations in an externally valid way.

There are a few situations where correlational research is an appropriate choice.

To investigate non-causal relationships

You want to find out if there is an association between two variables, but you don’t expect to find a causal relationship between them.

Correlational research can provide insights into complex real-world relationships, helping researchers develop theories and make predictions.

To explore causal relationships between variables

You think there is a causal relationship between two variables, but it is impractical, unethical, or too costly to conduct experimental research that manipulates one of the variables.

Correlational research can provide initial indications or additional support for theories about causal relationships.

To test new measurement tools

You have developed a new instrument for measuring your variable, and you need to test its reliability or validity .

Correlational research can be used to assess whether a tool consistently or accurately captures the concept it aims to measure.

There are many different methods you can use in correlational research. In the social and behavioural sciences, the most common data collection methods for this type of research include surveys, observations, and secondary data.

It’s important to carefully choose and plan your methods to ensure the reliability and validity of your results. You should carefully select a representative sample so that your data reflects the population you’re interested in without bias .

In survey research , you can use questionnaires to measure your variables of interest. You can conduct surveys online, by post, by phone, or in person.

Surveys are a quick, flexible way to collect standardised data from many participants, but it’s important to ensure that your questions are worded in an unbiased way and capture relevant insights.

Naturalistic observation

Naturalistic observation is a type of field research where you gather data about a behaviour or phenomenon in its natural environment.

This method often involves recording, counting, describing, and categorising actions and events. Naturalistic observation can include both qualitative and quantitative elements, but to assess correlation, you collect data that can be analysed quantitatively (e.g., frequencies, durations, scales, and amounts).

Naturalistic observation lets you easily generalise your results to real-world contexts, and you can study experiences that aren’t replicable in lab settings. But data analysis can be time-consuming and unpredictable, and researcher bias may skew the interpretations.

Secondary data

Instead of collecting original data, you can also use data that has already been collected for a different purpose, such as official records, polls, or previous studies.

Using secondary data is inexpensive and fast, because data collection is complete. However, the data may be unreliable, incomplete, or not entirely relevant, and you have no control over the reliability or validity of the data collection procedures.

After collecting data, you can statistically analyse the relationship between variables using correlation or regression analyses, or both. You can also visualise the relationships between variables with a scatterplot.

Different types of correlation coefficients and regression analyses are appropriate for your data based on their levels of measurement and distributions .

Correlation analysis

Using a correlation analysis, you can summarise the relationship between variables into a correlation coefficient : a single number that describes the strength and direction of the relationship between variables. With this number, you’ll quantify the degree of the relationship between variables.

The Pearson product-moment correlation coefficient, also known as Pearson’s r , is commonly used for assessing a linear relationship between two quantitative variables.

Correlation coefficients are usually found for two variables at a time, but you can use a multiple correlation coefficient for three or more variables.

Regression analysis

With a regression analysis , you can predict how much a change in one variable will be associated with a change in the other variable. The result is a regression equation that describes the line on a graph of your variables.

You can use this equation to predict the value of one variable based on the given value(s) of the other variable(s). It’s best to perform a regression analysis after testing for a correlation between your variables.

It’s important to remember that correlation does not imply causation . Just because you find a correlation between two things doesn’t mean you can conclude one of them causes the other, for a few reasons.

Directionality problem

If two variables are correlated, it could be because one of them is a cause and the other is an effect. But the correlational research design doesn’t allow you to infer which is which. To err on the side of caution, researchers don’t conclude causality from correlational studies.

Third variable problem

A confounding variable is a third variable that influences other variables to make them seem causally related even though they are not. Instead, there are separate causal links between the confounder and each variable.

In correlational research, there’s limited or no researcher control over extraneous variables . Even if you statistically control for some potential confounders, there may still be other hidden variables that disguise the relationship between your study variables.

Although a correlational study can’t demonstrate causation on its own, it can help you develop a causal hypothesis that’s tested in controlled experiments.

A correlation reflects the strength and/or direction of the association between two or more variables.

• A positive correlation means that both variables change in the same direction.
• A negative correlation means that the variables change in opposite directions.
• A zero correlation means there’s no relationship between the variables.

A correlational research design investigates relationships between two variables (or more) without the researcher controlling or manipulating any of them. It’s a non-experimental type of quantitative research .

Controlled experiments establish causality, whereas correlational studies only show associations between variables.

• In an experimental design , you manipulate an independent variable and measure its effect on a dependent variable. Other variables are controlled so they can’t impact the results.
• In a correlational design , you measure variables without manipulating any of them. You can test whether your variables change together, but you can’t be sure that one variable caused a change in another.

In general, correlational research is high in external validity while experimental research is high in internal validity .

A correlation is usually tested for two variables at a time, but you can test correlations between three or more variables.

A correlation coefficient is a single number that describes the strength and direction of the relationship between your variables.

Different types of correlation coefficients might be appropriate for your data based on their levels of measurement and distributions . The Pearson product-moment correlation coefficient (Pearson’s r ) is commonly used to assess a linear relationship between two quantitative variables.

Cite this Scribbr article

If you want to cite this source, you can copy and paste the citation or click the ‘Cite this Scribbr article’ button to automatically add the citation to our free Reference Generator.

Bhandari, P. (2022, December 05). Correlational Research | Guide, Design & Examples. Scribbr. Retrieved 14 May 2024, from https://www.scribbr.co.uk/research-methods/correlational-research-design/

Is this article helpful?

Pritha Bhandari

Other students also liked, a quick guide to experimental design | 5 steps & examples, quasi-experimental design | definition, types & examples, qualitative vs quantitative research | examples & methods.

Importance and use of correlational research

Affiliation.

• 1 School of Nursing and Midwifery, Trinity College Dublin, Dublin, Republic of Ireland.
• PMID: 27424963
• DOI: 10.7748/nr.2016.e1382

Background: The importance of correlational research has been reported in the literature yet few research texts discuss design in any detail.

Aim: To discuss important issues and considerations in correlational research, and suggest ways to avert potential problems during the preparation and application of the design.

Discussion: This article targets the gap identified in the literature regarding correlational research design. Specifically, it discusses the importance and purpose of correlational research, its application, analysis and interpretation with contextualisations to nursing and health research.

Conclusion: Findings from correlational research can be used to determine prevalence and relationships among variables, and to forecast events from current data and knowledge. In spite of its many uses, prudence is required when using the methodology and analysing data. To assist researchers in reducing mistakes, important issues are singled out for discussion and several options put forward for analysing data.

Implications for practice: Correlational research is widely used and this paper should be particularly useful for novice nurse researchers. Furthermore, findings generated from correlational research can be used, for example, to inform decision-making, and to improve or initiate health-related activities or change.

Keywords: correlation; correlational research; data analysis; measurement tools; nurses; nursing research; quantitative; variables.

• Nursing Research*

Want to create or adapt books like this? Learn more about how Pressbooks supports open publishing practices.

7.2 Correlational Research

Learning objectives.

• Define correlational research and give several examples.
• Explain why a researcher might choose to conduct correlational research rather than experimental research or another type of nonexperimental research.

What Is Correlational Research?

Correlational research is a type of nonexperimental research in which the researcher measures two variables and assesses the statistical relationship (i.e., the correlation) between them with little or no effort to control extraneous variables. There are essentially two reasons that researchers interested in statistical relationships between variables would choose to conduct a correlational study rather than an experiment. The first is that they do not believe that the statistical relationship is a causal one. For example, a researcher might evaluate the validity of a brief extraversion test by administering it to a large group of participants along with a longer extraversion test that has already been shown to be valid. This researcher might then check to see whether participants’ scores on the brief test are strongly correlated with their scores on the longer one. Neither test score is thought to cause the other, so there is no independent variable to manipulate. In fact, the terms independent variable and dependent variable do not apply to this kind of research.

The other reason that researchers would choose to use a correlational study rather than an experiment is that the statistical relationship of interest is thought to be causal, but the researcher cannot manipulate the independent variable because it is impossible, impractical, or unethical. For example, Allen Kanner and his colleagues thought that the number of “daily hassles” (e.g., rude salespeople, heavy traffic) that people experience affects the number of physical and psychological symptoms they have (Kanner, Coyne, Schaefer, & Lazarus, 1981). But because they could not manipulate the number of daily hassles their participants experienced, they had to settle for measuring the number of daily hassles—along with the number of symptoms—using self-report questionnaires. Although the strong positive relationship they found between these two variables is consistent with their idea that hassles cause symptoms, it is also consistent with the idea that symptoms cause hassles or that some third variable (e.g., neuroticism) causes both.

A common misconception among beginning researchers is that correlational research must involve two quantitative variables, such as scores on two extraversion tests or the number of hassles and number of symptoms people have experienced. However, the defining feature of correlational research is that the two variables are measured—neither one is manipulated—and this is true regardless of whether the variables are quantitative or categorical. Imagine, for example, that a researcher administers the Rosenberg Self-Esteem Scale to 50 American college students and 50 Japanese college students. Although this “feels” like a between-subjects experiment, it is a correlational study because the researcher did not manipulate the students’ nationalities. The same is true of the study by Cacioppo and Petty comparing college faculty and factory workers in terms of their need for cognition. It is a correlational study because the researchers did not manipulate the participants’ occupations.

Figure 7.2 “Results of a Hypothetical Study on Whether People Who Make Daily To-Do Lists Experience Less Stress Than People Who Do Not Make Such Lists” shows data from a hypothetical study on the relationship between whether people make a daily list of things to do (a “to-do list”) and stress. Notice that it is unclear whether this is an experiment or a correlational study because it is unclear whether the independent variable was manipulated. If the researcher randomly assigned some participants to make daily to-do lists and others not to, then it is an experiment. If the researcher simply asked participants whether they made daily to-do lists, then it is a correlational study. The distinction is important because if the study was an experiment, then it could be concluded that making the daily to-do lists reduced participants’ stress. But if it was a correlational study, it could only be concluded that these variables are statistically related. Perhaps being stressed has a negative effect on people’s ability to plan ahead (the directionality problem). Or perhaps people who are more conscientious are more likely to make to-do lists and less likely to be stressed (the third-variable problem). The crucial point is that what defines a study as experimental or correlational is not the variables being studied, nor whether the variables are quantitative or categorical, nor the type of graph or statistics used to analyze the data. It is how the study is conducted.

Figure 7.2 Results of a Hypothetical Study on Whether People Who Make Daily To-Do Lists Experience Less Stress Than People Who Do Not Make Such Lists

Data Collection in Correlational Research

Again, the defining feature of correlational research is that neither variable is manipulated. It does not matter how or where the variables are measured. A researcher could have participants come to a laboratory to complete a computerized backward digit span task and a computerized risky decision-making task and then assess the relationship between participants’ scores on the two tasks. Or a researcher could go to a shopping mall to ask people about their attitudes toward the environment and their shopping habits and then assess the relationship between these two variables. Both of these studies would be correlational because no independent variable is manipulated. However, because some approaches to data collection are strongly associated with correlational research, it makes sense to discuss them here. The two we will focus on are naturalistic observation and archival data. A third, survey research, is discussed in its own chapter.

Naturalistic Observation

Naturalistic observation is an approach to data collection that involves observing people’s behavior in the environment in which it typically occurs. Thus naturalistic observation is a type of field research (as opposed to a type of laboratory research). It could involve observing shoppers in a grocery store, children on a school playground, or psychiatric inpatients in their wards. Researchers engaged in naturalistic observation usually make their observations as unobtrusively as possible so that participants are often not aware that they are being studied. Ethically, this is considered to be acceptable if the participants remain anonymous and the behavior occurs in a public setting where people would not normally have an expectation of privacy. Grocery shoppers putting items into their shopping carts, for example, are engaged in public behavior that is easily observable by store employees and other shoppers. For this reason, most researchers would consider it ethically acceptable to observe them for a study. On the other hand, one of the arguments against the ethicality of the naturalistic observation of “bathroom behavior” discussed earlier in the book is that people have a reasonable expectation of privacy even in a public restroom and that this expectation was violated.

Researchers Robert Levine and Ara Norenzayan used naturalistic observation to study differences in the “pace of life” across countries (Levine & Norenzayan, 1999). One of their measures involved observing pedestrians in a large city to see how long it took them to walk 60 feet. They found that people in some countries walked reliably faster than people in other countries. For example, people in the United States and Japan covered 60 feet in about 12 seconds on average, while people in Brazil and Romania took close to 17 seconds.

Because naturalistic observation takes place in the complex and even chaotic “real world,” there are two closely related issues that researchers must deal with before collecting data. The first is sampling. When, where, and under what conditions will the observations be made, and who exactly will be observed? Levine and Norenzayan described their sampling process as follows:

Male and female walking speed over a distance of 60 feet was measured in at least two locations in main downtown areas in each city. Measurements were taken during main business hours on clear summer days. All locations were flat, unobstructed, had broad sidewalks, and were sufficiently uncrowded to allow pedestrians to move at potentially maximum speeds. To control for the effects of socializing, only pedestrians walking alone were used. Children, individuals with obvious physical handicaps, and window-shoppers were not timed. Thirty-five men and 35 women were timed in most cities. (p. 186)

Precise specification of the sampling process in this way makes data collection manageable for the observers, and it also provides some control over important extraneous variables. For example, by making their observations on clear summer days in all countries, Levine and Norenzayan controlled for effects of the weather on people’s walking speeds.

The second issue is measurement. What specific behaviors will be observed? In Levine and Norenzayan’s study, measurement was relatively straightforward. They simply measured out a 60-foot distance along a city sidewalk and then used a stopwatch to time participants as they walked over that distance. Often, however, the behaviors of interest are not so obvious or objective. For example, researchers Robert Kraut and Robert Johnston wanted to study bowlers’ reactions to their shots, both when they were facing the pins and then when they turned toward their companions (Kraut & Johnston, 1979). But what “reactions” should they observe? Based on previous research and their own pilot testing, Kraut and Johnston created a list of reactions that included “closed smile,” “open smile,” “laugh,” “neutral face,” “look down,” “look away,” and “face cover” (covering one’s face with one’s hands). The observers committed this list to memory and then practiced by coding the reactions of bowlers who had been videotaped. During the actual study, the observers spoke into an audio recorder, describing the reactions they observed. Among the most interesting results of this study was that bowlers rarely smiled while they still faced the pins. They were much more likely to smile after they turned toward their companions, suggesting that smiling is not purely an expression of happiness but also a form of social communication.

Naturalistic observation has revealed that bowlers tend to smile when they turn away from the pins and toward their companions, suggesting that smiling is not purely an expression of happiness but also a form of social communication.

sieneke toering – bowling big lebowski style – CC BY-NC-ND 2.0.

When the observations require a judgment on the part of the observers—as in Kraut and Johnston’s study—this process is often described as coding . Coding generally requires clearly defining a set of target behaviors. The observers then categorize participants individually in terms of which behavior they have engaged in and the number of times they engaged in each behavior. The observers might even record the duration of each behavior. The target behaviors must be defined in such a way that different observers code them in the same way. This is the issue of interrater reliability. Researchers are expected to demonstrate the interrater reliability of their coding procedure by having multiple raters code the same behaviors independently and then showing that the different observers are in close agreement. Kraut and Johnston, for example, video recorded a subset of their participants’ reactions and had two observers independently code them. The two observers showed that they agreed on the reactions that were exhibited 97% of the time, indicating good interrater reliability.

Archival Data

Another approach to correlational research is the use of archival data , which are data that have already been collected for some other purpose. An example is a study by Brett Pelham and his colleagues on “implicit egotism”—the tendency for people to prefer people, places, and things that are similar to themselves (Pelham, Carvallo, & Jones, 2005). In one study, they examined Social Security records to show that women with the names Virginia, Georgia, Louise, and Florence were especially likely to have moved to the states of Virginia, Georgia, Louisiana, and Florida, respectively.

As with naturalistic observation, measurement can be more or less straightforward when working with archival data. For example, counting the number of people named Virginia who live in various states based on Social Security records is relatively straightforward. But consider a study by Christopher Peterson and his colleagues on the relationship between optimism and health using data that had been collected many years before for a study on adult development (Peterson, Seligman, & Vaillant, 1988). In the 1940s, healthy male college students had completed an open-ended questionnaire about difficult wartime experiences. In the late 1980s, Peterson and his colleagues reviewed the men’s questionnaire responses to obtain a measure of explanatory style—their habitual ways of explaining bad events that happen to them. More pessimistic people tend to blame themselves and expect long-term negative consequences that affect many aspects of their lives, while more optimistic people tend to blame outside forces and expect limited negative consequences. To obtain a measure of explanatory style for each participant, the researchers used a procedure in which all negative events mentioned in the questionnaire responses, and any causal explanations for them, were identified and written on index cards. These were given to a separate group of raters who rated each explanation in terms of three separate dimensions of optimism-pessimism. These ratings were then averaged to produce an explanatory style score for each participant. The researchers then assessed the statistical relationship between the men’s explanatory style as college students and archival measures of their health at approximately 60 years of age. The primary result was that the more optimistic the men were as college students, the healthier they were as older men. Pearson’s r was +.25.

This is an example of content analysis —a family of systematic approaches to measurement using complex archival data. Just as naturalistic observation requires specifying the behaviors of interest and then noting them as they occur, content analysis requires specifying keywords, phrases, or ideas and then finding all occurrences of them in the data. These occurrences can then be counted, timed (e.g., the amount of time devoted to entertainment topics on the nightly news show), or analyzed in a variety of other ways.

Key Takeaways

• Correlational research involves measuring two variables and assessing the relationship between them, with no manipulation of an independent variable.
• Correlational research is not defined by where or how the data are collected. However, some approaches to data collection are strongly associated with correlational research. These include naturalistic observation (in which researchers observe people’s behavior in the context in which it normally occurs) and the use of archival data that were already collected for some other purpose.

Discussion: For each of the following, decide whether it is most likely that the study described is experimental or correlational and explain why.

• An educational researcher compares the academic performance of students from the “rich” side of town with that of students from the “poor” side of town.
• A cognitive psychologist compares the ability of people to recall words that they were instructed to “read” with their ability to recall words that they were instructed to “imagine.”
• A manager studies the correlation between new employees’ college grade point averages and their first-year performance reports.
• An automotive engineer installs different stick shifts in a new car prototype, each time asking several people to rate how comfortable the stick shift feels.
• A food scientist studies the relationship between the temperature inside people’s refrigerators and the amount of bacteria on their food.
• A social psychologist tells some research participants that they need to hurry over to the next building to complete a study. She tells others that they can take their time. Then she observes whether they stop to help a research assistant who is pretending to be hurt.

Kanner, A. D., Coyne, J. C., Schaefer, C., & Lazarus, R. S. (1981). Comparison of two modes of stress measurement: Daily hassles and uplifts versus major life events. Journal of Behavioral Medicine, 4 , 1–39.

Kraut, R. E., & Johnston, R. E. (1979). Social and emotional messages of smiling: An ethological approach. Journal of Personality and Social Psychology, 37 , 1539–1553.

Levine, R. V., & Norenzayan, A. (1999). The pace of life in 31 countries. Journal of Cross-Cultural Psychology, 30 , 178–205.

Pelham, B. W., Carvallo, M., & Jones, J. T. (2005). Implicit egotism. Current Directions in Psychological Science, 14 , 106–110.

Peterson, C., Seligman, M. E. P., & Vaillant, G. E. (1988). Pessimistic explanatory style is a risk factor for physical illness: A thirty-five year longitudinal study. Journal of Personality and Social Psychology, 55 , 23–27.

Research Methods in Psychology Copyright © 2016 by University of Minnesota is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License , except where otherwise noted.

Designing and Carrying Out Correlational Studies Using Real-World Data

• First Online: 10 February 2022

Cite this chapter

• Charles P. Friedman 4 ,
• Jeremy C. Wyatt 5 &
• Joan S. Ash 6  

Part of the book series: Health Informatics ((HI))

1459 Accesses

This chapter describes the benefits and challenges of carrying out studies of informatics resources using routine or real-world data. It describes how to carry out such studies in a professional manner using a trusted or virtual environment. It then examines the challenges of inferring causality from correlation, and offers strategies to overcome each of these.

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

Access this chapter

• Available as EPUB and PDF
• Read on any device
• Instant download
• Own it forever
• Compact, lightweight edition
• Dispatched in 3 to 5 business days
• Free shipping worldwide - see info
• Durable hardcover edition

Tax calculation will be finalised at checkout

Purchases are for personal use only

Institutional subscriptions

Assessment bias is also known as ascertainment bias.

Agiro A, Chen X, Eshete B, Sutphen R, Bourquardez Clark E, Burroughs CM, et al. Data linkages between patient-powered research networks and health plans: a foundation for collaborative research. J Am Med Inform Assoc 2019;26:594–602. Available from https://doi.org/10.1093/jamia/ocz012 . Accessed 2 July 2021.

Agniel D, Kohane IS, Weber GM. Biases in electronic health record data due to processes within the healthcare system: retrospective observational study. BMJ. 2018;361:k1479.

Article   Google Scholar  

Andersen B. Methodological errors in medical research. Oxford, UK: Blackwell Scientific Publications; 1990.

Google Scholar  

Baker L. Hilarious graphs (and pirates) prove that correlation is not causation. Data Science Central. Issaquah, WA: Data Science Central, TechTarget; 2018. Available from https://www.datasciencecentral.com/profiles/blogs/hilarious-graphs-and-pirates-prove-that-correlation-is-not . Accessed 29 June 2021.

Benchimol EI, Smeeth L, Guttmann A, Harron K, Moher D, Petersen I, et al. RECORD Working Committee. The REporting of studies conducted using observational routinely-collected health data (RECORD) statement. PLoS Med 2015;12:e1001885. Available from https://doi.org/10.1371/journal.pmed.1001885 . Accessed 29 June 2021.

Britto MT, Fuller SC, Kaplan HC, Kotagal U, Lannon C, Margolis PA, et al. Using a network organisational architecture to support the development of learning healthcare systems. BMJ Qual Safety. 2018;27:937–46.

Byar DP. Why data bases should not replace randomized clinical trials. Biometrics. 1980;36:337–42.

Article   CAS   Google Scholar  

Collins R, Bowman L, Landray M, Peto R. The magic of randomization versus the myth of real-world evidence. N Engl J Med. 2020;382:674–8.

Cooper GF, Bahar I, Becich MJ, Benos PV, Berg J, Espino JU, et al. Center for Causal Discovery Team. The center for causal discovery of biomedical knowledge from big data. J Am Med Inform Assoc 2015;22:1132–6. Available from https://doi.org/10.1093/jamia/ocv059 . Accessed 29 June 2021.

Davey Smith G. Capitalising on Mendelian randomization to assess the effects of treatments. James Lind Library Bull (JLL Bull) 2006;1:1. Available from https://www.jameslindlibrary.org/articles/capitalising-on-mendelian-randomization-to-assess-the-effects-of-treatments/ . Accessed 29 June 2021.

Deeny SR, Steventon A. Making sense of the shadows: priorities for creating a learning healthcare system based on routinely collected data. BMJ Qual Safety. 2015;24:505–15.

Equator Network, UK EQUATOR Centre. Results of Search for Reporting Guidelines: Observational Studies. Oxford, UK: Equator Network, UK EQUATOR Centre, Centre for Statistics in Medicine (CSM), NDORMS, University of Oxford; 2021. Available from https://www.equator-network.org/?post_type=eq_guidelines&eq_guidelines_study_design=observational-studies&eq_guidelines_clinical_specialty=0&eq_guidelines_report_section=0&s=+&eq_guidelines_study_design_sub_cat=0 . Accessed 29 June 2021.

Eriksson T, Maclure M, Kragstrup J. Consultation with the general practitioner triggered by advice from social network members. Scand J Prim Health Care. 2004;22:54–9.

Ford I, Murray H, Packard CJ, Shepherd J, Macfarlane PW, Cobbe SM. Long-term follow-up of the West of Scotland Coronary Prevention Study. N Eng J Med. 2000;357:1477–86.

Friedman CP, Wong AK, Blumenthal D. Achieving a nationwide learning health system. Sci Transl Med 2010;2:57cm29. Available from https://doi.org/10.1126/scitranslmed.3001456 . Accessed 29 June 2021.

Gray E, Marti J, Wyatt JC, Brewster DH, Hall PS, SATURNE Advisory Group. Chemotherapy effectiveness in trial-underrepresented groups with early breast cancer: a retrospective cohort study. PLoS Med. 2019;16:e1003006.

Grimes DA, Schulz KF. Bias and causal associations in observational research. Lancet. 2002;359:248–52.

Gymrek M, McGuire AL, Golan D, Halperin E, Erlich Y. Identifying personal genomes by surname inference. Science. 2013;339:321–4.

Hemkens LG, Contopoulos-Ioannidis DG, Ioannidis JP. Routinely collected data and comparative effectiveness evidence: promises and limitations. CMAJ 2016;188:158–164. Available from https://doi.org/10.1503/cmaj.150653 . Accessed 29 Jun 2021.

Hill AB. The environment and disease: association or causation? Proc R Soc Med. 1965;58:295–300.

CAS   PubMed   PubMed Central   Google Scholar  

Hilton Boon M, Craig P, Thomson H, Campbell M, Moore L. Regression discontinuity designs in health: a systematic review. Epidemiology. 2021;32:87–93.

Hundepool A, Domingo-Ferrer J, Franconi L, Giessing S, Lenz R, Longhurst J, et al. A CENtre of EXcellence for Statistical Disclosure Control Handbook on Statistical Disclosure Control Version 1.2 . Brussels, Belgium: CENtre of EXcellence for Statistical Disclosure Control (CENEX SDC), Portal on Collaboration in Research and Methodology for Official Statistics (CROS), ESSnet, European Commission; 2010.

Ioannidis JPA. Contradicted and initially stronger effects in highly cited clinical research. JAMA. 2005;294:218–28.

Julious SA, Mullee MA. Confounding and Simpson’s paradox. BMJ. 1994;309:1480–1.

Kahneman D. Thinking, fast and slow. New York: Macmillan; 2011.

Klein MS, Ross FV, Adams DL, Gilbert CM. Effect of online literature searching on length of stay and patient care costs. Acad Med. 1994;69(6):489–95.

König K, Pechmann A, Thiele S, Walter MC, Schorling D, Tassoni A, et al. De-duplicating patient records from three independent data sources reveals the incidence of rare neuromuscular disorders in Germany. Orphanet J Rare Dis. 2019;14:152.

Lawrence ND. Data readiness levels: turning data from palid to vivid. In: Lawrence ND, editors. inverseprobability.com : Neil Lawrence’s Homepage. Cambridge, UK: Neil D. Lawrence; 2017. Available from https://inverseprobability.com/2017/01/12/data-readiness-levels . Accessed 29 June 2021.

Maclure M. The case-crossover design: a method for studying transient effects on the risk of acute events. Am J Epidemiol. 1991;133:144–53.

Matthews GJ, Harel O. Data confidentiality: a review of methods for statistical disclosure limitation and methods for assessing privacy. Stat Surv. 2011;5:1–29.

O’Keefe CM, Rubin DB. Individual privacy versus public good: protecting confidentiality in health research. Stat Med. 2015;34:3081–103.

Ory HW. Association between oral contraceptives and myocardial infarction. A review. JAMA. 1977;237:2619–22.

Riaz H, Khan MS, Siddiqi TJ, Usman MS, Shah N, Goyal A, et al. Association between obesity and cardiovascular outcomes: a systematic review and meta-analysis of Mendelian randomization studies. JAMA Netw Open. 2018;1:e183788.

Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70:41–55.

Seltzer CC, Bosse R, Garvey AJ. Mail survey response by smoking status. Am J Epidemiol. 1974;100:453–7.

Sherman RE, Anderson SA, Dal Pan GJ, Gray GW, Gross T, Hunter NL, et al. Real-world evidence - what is it and what can it tell us? N Engl J Med. 2016;375:2293–7.

Simpson EH. The interpretation of interaction in contingency tables. J R Stat Soc Ser B (Methodol). 1951;13:238–41.

Suissa S, Ernst P. Avoiding immortal time bias in observational studies. Eur Respir J. 2020;55:2000138.

Tsopra R, Wyatt JC, Beirne P, Rodger K, Callister M, Ghosh D, et al. Level of accuracy of diagnoses recorded in discharge summaries: a cohort study in three respiratory wards. J Eval Clin Pract. 2019;25:36–43.

Tufte E. The visual display of quantitative information. New York: Graphics Press; 1992.

United Kingdom National Health Service (NHS) Digital. Hospital Episode Statistics for Admitted Patient Care and Outpatient Data: Official Statistics Database. Leeds, UK: United Kingdom National Health Service (NHS) Digital, United Kingdom National Health Service (NHS); 2021. Available from https://digital.nhs.uk/data-and-information/publications/statistical/hospital-episode-statistics-for-admitted-patient-care-outpatient-and-accident-and-emergency-data . Accessed 29 June 2021.

Waruru A, Natukunda A, Nyagah LM, Kellogg TA, Zielinski-Gutierrez E, Waruiru W, et al. Where no universal health care identifier exists: comparison and determination of the utility of score-based persons matching algorithms using demographic data. JMIR Pub Health Surveill. 2018;4:e10436.

Yao XI, Wang X, Speicher PJ, Hwang ES, Cheng P, Harpole DH, et al. Reporting and guidelines in propensity score analysis: a systematic review of cancer and cancer surgical studies. J Natl Cancer Inst 2017;109:djw323. Available from https://doi.org/10.1093/jnci/djw323 . Accessed 29 June 2021.

Zhou Y, Matsouaka RA, Thomas L. Propensity score weighting under limited overlap and model misspecification. Stat Methods Med Res. 2020;29:3721–56.

Download references

Author information

Authors and affiliations.

Department of Learning Health Sciences, University of Michigan Medical School, Ann Arbor, MI, USA

Charles P. Friedman

Department of Primary Care, Population Sciences and Medical Education, School of Medicine, University of Southampton, Southampton, UK

Jeremy C. Wyatt

Department of Medical Informatics and Clinical Epidemiology, School of Medicine, Oregon Health & Science University, Portland, OR, USA

Joan S. Ash

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Charles P. Friedman .

Answers to Self-Tests

Self-test 13.1.

Spurious correlation

The key observation of concern here is that the 3500 app users form only 1.4% of the overall population (3500/247,000) eligible to use the app. So, a study team needs to ask themselves what factors caused those relatively few individuals to use the app, and might these factors also have led to the observed weight loss ? Some possible factors could include:

Pressure from family and friends (and support) to lose weight

A forthcoming significant future event, e.g. a wedding or graduation

A health scare, e.g. a heart attack, suffered by the patients themselves or a close family member

Developing a new relationship that caused them to “turn over a new leaf” in their lives

Any of these factors are plausible explanations for why weight loss app users might start and continue to use the app, but are also reasons why that person might lose weight even without the app. So, use of the app might be a marker of the small proportion of people in the population who really intend to lose weight, and are being supported by others close to them to lose weight—rather than the cause. To investigate this possibility, a random subset of app users and non-users could be surveyed, asking them about any of the above factors. A qualitative investigation (using interviews or focus groups) of factors associated with success at weight loss (both promoters and inhibitors) would be useful to carry out before finalizing the survey questions to ensure that factors identified as important by responders are included.

Selection bias

Generally, consistent app users are younger than non-users and more likely to own and use a smartphone, which puts them in a income group than non-app users or people who do not own a smart phone. So, what we could be seeing here is bigger weight loss in the subset of people who own and use a smart phone compared to those who do not. This selection bias could be investigated by comparing data on the educational level and household income of app users and non-users, and adding these variable into a regression model to predict weight loss based on a number of factors that are known to be associated with success at losing weight. The person’s use or not of the smart phone app could then be added to this model to investigate if this new model explains more of the observed variation in weight loss, and if the coefficient for this variable is significant.

Information bias

If the person’s weight at the start and the end of the study period is recorded by the app in app users but is recorded informally (in a notebook, diary or vague remembrance of weight a year ago) in the non-users, this could lead to an information bias. Assume that the weight loss based on data recorded in the app is correct. However, it is possible that the weight loss for non-users is underestimated because the non-app users have forgotten or misplaced their weight recording from a year ago, and innocently substitute a figure that is closer to their current weight than it was in reality. If this is the case, the lack of an objective record in non-app users of their weight a year ago would tend to reduce their apparent weights loss, and thus make it appear less than the measured weight loss in app users. To eliminate this bias, we would need an objective record of the baseline weight for the entire population.

Confounding

It is possible that the app was provided to a subset of the population specifically selected by the health care delivery network as part of a risk management exercise to reduce their chances of developing a weight-related illness (such as diabetes, stroke, heart attacks). In addition, other actions may also have formed part of this program, for example regular contacts with a dietician, feedback on their weight and exercise regime compared to that of others in the program, or incentives such as a lower renewal fee if they were successful at reaching their weight loss goal at year end. All of these would be confounders in any study that claimed to measure the impact of the app on weight loss. A searching question directed back to the health care delivery network about potential reasons for app use would help the study team uncover whether the app was targeted to a specific population, and if the users also received other weight loss support.

Self-Test 13.2

The histogram shows how the percent online time increased dramatically in 2020, then reduced slightly in 2021 and 2022, but still remained higher than for the earlier years. Two of the added value scores (physiology and pathology) remained roughly the same, but the anatomy added value score appeared to drop in 2020 and the following years.

The X-Y plot of mean added value score against percent online time shows a negative correlation, with a slight drop in the mean score with increasing online time. This is confirmed by the regression line, showing all data points close to the line, and R squared of 0.97. From this analysis, one would be inclined to conclude that online learning does not promote student learning.

The histogram clearly indicates that the effects differ across the subjects, and the averaging across the subjects conceals this effect. It appears that the increased time spent online had little effect on the added value scores for two subjects, but was associated with a large drop in the anatomy program added value scores. This might be because anatomy is better taught face to face, or it might be another factor, such as a change in program admission criteria, in teaching staff or the materials used. For example, an increase in the baseline anatomy skills of the students in 2020 and onwards could explain the decrease in added value. Equally, the explanation for reduced anatomy added value scores may be an effect of the COVID pandemic which started in 2020, but this would be expected to affect all three subjects.

To investigate if the “online effect” is larger for anatomy than for the two other subjects, the slopes of the regression lines for anatomy alone versus the mean of all three subjects could be compared. The final graph below shows that the reduction in added value with online teaching is greater for anatomy with a regression coefficient of 0.29, almost double the coefficient of 0.17 for the mean across all three subjects.

Further data that could be sought to understand the reason for the observed changes includes:

The numbers of students in the program each year, as some of these changes could simply be due to small numbers in certain years.

The actual entering scores of students for each subject and year, as interpretation of the added value score varies depending on the starting point.

Data about teaching staff numbers and skills and student participation rates and satisfaction year by year and across the three subjects, to help understand why online learning may have led to lower added value for anatomy, if this is true.

Further information about the types (and quality) of online learning experience offered across the three subjects. It is possible, for example, that the online offerings for anatomy differed from those in physiology and pathology.

Rights and permissions

Reprints and permissions

Copyright information

© 2022 Springer Nature Switzerland AG

About this chapter

Friedman, C.P., Wyatt, J.C., Ash, J.S. (2022). Designing and Carrying Out Correlational Studies Using Real-World Data. In: Evaluation Methods in Biomedical and Health Informatics. Health Informatics. Springer, Cham. https://doi.org/10.1007/978-3-030-86453-8_13

Download citation

DOI : https://doi.org/10.1007/978-3-030-86453-8_13

Published : 10 February 2022

Publisher Name : Springer, Cham

Print ISBN : 978-3-030-86452-1

Online ISBN : 978-3-030-86453-8

eBook Packages : Medicine Medicine (R0)

Share this chapter

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

• Publish with us

Policies and ethics

• Find a journal
• Track your research

• Primary research
• Open access
• Published: 28 August 2013

Correlational study: illness representations and coping styles in caregivers for individuals with schizophrenia

• Shyhrete Rexhaj 1 , 2 ,
• Nataly Viens Python 2 ,
• Diane Morin 3 ,
• Charles Bonsack 1 &
• Jérôme Favrod 1 , 2  

Annals of General Psychiatry volume  12 , Article number:  27 ( 2013 ) Cite this article

30k Accesses

25 Citations

2 Altmetric

Metrics details

Caring for individuals with schizophrenia can create distress for caregivers which can, in turn, have a harmful impact on patient progress. There could be a better understanding of the connections between caregivers’ representations of schizophrenia and coping styles. This study aims at exploring those connections.

This correlational descriptive study was conducted with 92 caregivers of individuals suffering from schizophrenia. The participants completed three questionnaires translated and validated in French: (a) a socio-demographic questionnaire, (b) the Illness Perception Questionnaire for Schizophrenia and (c) the Family Coping Questionnaire.

Our results show that illness representations are slightly correlated with coping styles. More specifically, emotional representations are correlated to an emotion-focused coping style centred on coercion, avoidance and resignation.

Our results are coherent with the Commonsense Model of Self-Regulation of Health and Illness and should enable to develop new interventions for caregivers.

Problem statement

Between 30% and 91% of individuals with schizophrenia live in a family setting [ 1 – 3 ]. The decreased length of stay in hospital and restrictions on involuntary treatments mean that family-based caregivers provide an important support during periods of psychological instability. This support implies that caregivers rely on a variety of strategies to confront the consequences resulting from the psychological instability of the schizophrenia patient.

Burden, distress, illness representations and coping strategies

The literature indicates that there are interactions between the different concepts of family burden, distress, illness representations, the expressed emotion (EE) and coping strategies. It has been demonstrated that caregivers’ representations of negative consequences of the illness for the patient are positively correlated with their objective burden, while representations of the negative consequences of schizophrenia for the caregiver are positively correlated to a subjective burden [ 4 , 5 ]. The caregivers’ subjective burden is also associated with negative emotional responses to the illness [ 5 ]. Also, caregivers having an elevated and hostile type of EE usually consider the patient responsible for the causes of illness [ 6 , 7 ]. Furthermore, in comparison with caregivers having a low EE, those having a high, critical-type EE tend to underestimate their possibilities of controlling problems themselves and perceive the illness as unlikely to be controlled by treatment as well as attribute more negative consequences of the illness for the patient and themselves [ 4 ].

Distress experienced by caregivers is also correlated with illness representations such as the following: (1) illness outbreaks are chronic, (2) the feeling that treatment does not help control the illness, (3) the perception that the patient can have greater personal control over the illness, (4) the perception that the illness brings about negative consequences for the patient and (5) that the illness brings about negative consequences for the caregiver and (6) representations that the illness elicits painful emotions such as anxiety and fear [ 4 , 5 , 8 ].

Coping strategies by caregivers of patients with schizophrenia have been explored by several authors who mainly based their observations using models or frameworks derived from Lazarus and Folkman’s theory of stress coping [ 9 – 12 ]. A review of the literature showed that there are strategies that are more or less efficient for confronting stress. More specifically for caregivers, strategies like coercion, avoidance and resignation are associated with suffering and patient relapse [ 13 , 14 ].

Birchwood and Cochrane [ 14 ] explored coping strategies used by caregivers of schizophrenia patients. Their results detail eight essential coping categories: coercion, avoidance, ignorance/acceptance, constructive, resignation, reassurance, disorganization and collusion. They also showed that coercion is the first predictor in patient relapses. As concerns specificities of different coping styles, Magliano et al. [ 15 ] conducted a study which explored coping strategies in relation to physical and somatic symptoms of caregivers as well as the association between these two variables. The Family Coping Questionnaire (FCQ) was used [ 10 ]. Their results indicated that emotion-focused coping (coercion, avoidance and resignation) is positively correlated with participant anxiety and depression [ 15 ]. In another study, Knudson and Coyle came up with a procedure which incited certain caregivers to modify their coping strategies [ 11 ]. In their qualitative study, the caregivers of individuals with schizophrenia were invited to describe their coping strategy. At the onset of the illness, the caregivers tended to use a problem-focused coping strategy. However, if the symptoms became persistent, they progressively opted for emotion-focused strategies, which enabled them to attain a position of acceptance and, ultimately, of well-being. Nevertheless, the conceptualization of emotion-focused coping done according to factor analysis in the study of Magliano et al. only included strategies such as resignation, avoidance and coercion [ 10 ]. Acceptance of the illness has seldom been studied and could be a potential functional human response in some instances.

Magliano et al. reported that the high level of burden is associated with a reduction in social interests, a reduction of social support network as well as a resignation and an avoidance of contact with the family member suffering from schizophrenia. It also appears that in the absence of any specific intervention with these caregivers, this result can remain unchanged for as long as a year [ 15 ]. Specific to the subjective burden, it appears that emotional and cognitive reactions are associated with the use of coping strategies that are specifically emotional, such as avoidance and isolation [ 16 , 17 ]. These strategies have been identified as being linked to an increase in caregiver distress [ 18 ]. Increased distress can result in a deterioration of the family atmosphere through elevated EE levels [ 12 ]. These elements lead to a further increase in burden [ 19 ].

Theoretical framework

In terms of theory, in the 1960s, Leventhal and his colleagues developed the Commonsense Model of Self-Regulation of Health and Illness (SRM) [ 20 ]. The results from their initial research indicated that modifying the personal representations of health and the development of an action plan were the two determining factors for the creation of health-promoting actions. Their conclusions led to the development of a series of other studies meant to define the different characteristics of health and illness representations. According to the model, representations are cognitive and emotional constructions of a health problem. The SRM includes eight concepts or dimensions: (1) internal and external stimuli, (2) treatment system, (3) representation of the illness and the treatment, (4) coping procedures, (5) evaluation of the cognitive treatment of the information, (6) emotional representation, (7) coping responses and (8) evaluation of the emotional treatment of the information.

Coping involves procedures which enable an individual to collect information and control the problem, as well as different responses such as distraction or relaxation. Coping can be described as having two primary functions. The first is centred on problem solving and the second on the immediate regulation of the emotion elicited by the problem. According to the SRM, the choice of a coping strategy used by caregivers is associated with the kind of representation of the illness that the caregiver has developed. In this way, the caregiver’s explicative model seems to have an influence on their choice of coping strategy. This choice can then, in turn, lighten or increase caregiver burden [ 17 ]. Effectively, negative representations of the illness can lead to the use of unsuitable coping strategies [ 21 , 22 ].

Goal of the study

The goal of this study was to explore the associations between illness representations and three forms of coping styles—(1) problem-focused coping, (2) emotion-focused coping and (3) social support-focused coping—for caregivers of individuals with schizophrenia.

Design and recruitment

This correlational descriptive study was conducted with 92 caregivers of individuals with schizophrenia. Participants were members of French-speaking social support organizations, were recruited using a convenience sampling strategy and met the following criteria: (1) being 18 years or older, (2) living in Switzerland or France, (3) being able to speak French, (4) acting as carer for a family member or close friend who suffers from schizophrenia and (5) having had at least a 1-h contact with this person in the last month. As the first step of the recruiting strategy, social support organizations were met at the time of meeting with members in order to provide a complete presentation of the study. Then, the members who were present received an envelope containing an informed consent sheet, an information letter and the necessary questionnaires, as well as a pre-addressed and pre-stamped envelope for the return of all completed documents.

A second recruiting strategy was conducted using an electronic survey. The presidents of the social support organizations sent to all members of their respective groups the survey link leading to a dedicated website that included the same documentation as the paper version.

Instruments

Participants filled out three self-administered forms: (a) a socio-demographic questionnaire, (b) the Illness Perception Questionnaire for Schizophrenia : Relatives’ version and (c) the Family Coping Questionnaire. Authorization to translate these questionnaires into French using independent backward translation was granted by the authors.

The socio-demographic questionnaire

According to the literature, the most significant socio-demographic variables which influence illness representations and coping strategies are the following: (1) caregiver gender [ 23 ], (2) caregiver age, (3) caregiver education level [ 13 ], (4) the length of contact with the patient [ 19 , 24 ], (5) professional and social support [ 25 , 26 ] such as involvement in a Profamille program [ 27 ], (6) the nature of their connection with the patient [ 28 ], (7) the length of the illness [ 29 ], (8) whether the caregiver lives with the patient or not [ 30 ], (9) age of the patient, (10) patient gender and (11) the patient’s professional support. With the further goal of comparing our results with other studies already reported in the literature, we collected data on the number of people living in a household and the civil status of the caregivers.

The Illness Perception Questionnaire for Schizophrenia: Relatives’ version

Illness representations were measured using a self-administered questionnaire entitled ‘Illness Perception Questionnaire for Schizophrenia: Relatives’ version (IPQS: Relatives)’ [ 5 ]. It involves 13 sub-scales (150 items): identity, timeline (acute/chronic), timeline (cyclic), negative consequences for the individual, negative consequences for the caregiver, personal control (feeling of powerlessness—patient), personal control (feeling of powerlessness—caregiver), personal blame (patient responsibility), personal blame (caregiver responsibility), therapeutic control, mental health problem coherence, emotional representations and causes. For the current study, the identity and the causes of the illness were not considered. The responses provided by the caregivers regarding their perceptions of the illness were measured using a 5-level Likert scale: 0 = do not agree at all, 1 = do not agree, 2 = more or less agree = 3 = agree and 4 = completely agree. Results show a dimensional reliability situated between α = 0.63 and α = 0.83 [ 5 ].

The Family Coping Questionnaire

Coping styles were measured using a self-administered questionnaire entitled ‘Family Coping Questionnaire (FCQ)’. The version of the FCQ, used in this study, included 27 items measuring seven dimensions (information gathering, positive communication, social involvement, coercion, avoidance, resignation, the patient’s social involvement). A factor analysis enabled the authors to identify three main factors. Looking at these three factors and at the conceptual definition of coping, it is clear that the seven dimensions can be regrouped into three coping modes: (1) problem-focused coping (the patient’s social involvement, positive communication and information gathering are positively correlated with each other and negatively with avoidance, (2) emotion-focused coping (coercion, avoidance and resignation are positively correlated) and (3) social support-focused coping (social involvement is associated with avoidance) [ 10 ]. Our study catalogued these strategies. Caregiver responses to different situations were measured using a 5-level Likert scale: 1 = never, 2 = rarely, 3 = sometimes, 4 = very often and 5 = not applicable. Its validity was demonstrated during the BIOMED 1 study, conducted in five European countries [ 15 ]. Results showed a dimensional reliability between α = 0.68 and α = 0.83 [ 13 ].

Data analysis

Data were treated using the computing program ‘IBM SPSS Statistics® version 20’. Descriptive statistics were used to describe socio-demographic characteristics. To test associations between illness representations and coping styles, bivariate correlational analyses were used. The p value threshold used was set to <0.05.

Ethical considerations

All participants were required to sign an informed consent form for the paper version of the survey or confirm their consent in order to access the electronic version of the survey. The research protocol received full authorization by the Canton of Vaud’s Ethics Committee for human-based research.

Results and discussion

While Table  1 presents the socio-demographic characteristics of participants, Table  2 presents the main characteristics of the individuals with schizophrenia cared for. As showed, the final sample involved 92 middle-aged individuals (mean age = 56 years, standard deviation (SD) = 12.6) including 68% of women. Most participants were married or lived in a household as a married couple (70.3%). The majority of respondents were either mothers or fathers (66.3%) of the individual with schizophrenia. All participants had completed some level of education; vocational school, apprenticeship and university-level education were the most often cited. Most participants reported that they lived with one or several other individuals. Most of them (62%) participated to the Profamille program, a psycho-educational program for family members and friends of individuals with schizophrenia [ 31 ]. Among the 92 participants, 37 (40.2%) reported that they lived with the schizophrenia patient.

As concerns the individual with schizophrenia, results show that the duration of the illness is long (mean = 15.1 years, SD = 9.7). The mean age of the individual with schizophrenia taken care for by the participants of this study was 34 years and 77.2% were male. Individuals with schizophrenia call upon community services in a variety of ways. In 88% of all cases, a psychiatrist was involved in patient follow-up. In 34.8% of all cases, a general medical practitioner was involved in patient follow-up. In our sample, 33.7% of the schizophrenia patients received nursing care. Also, a significant number (33.7%) of the individuals with schizophrenia in our sample were involved in sheltered workshops. Other community services were also used at more than 10% (social worker, 20.7%; psychologist, 13%; day centre, 17.4%).

Results presented in Table  3 show that caregivers have a great conception of the illness as being chronic and as having recurring symptoms per cycle. They have a strong perception that the illness brings about negative consequences for the individual with schizophrenia. While at the same time, they perceive fewer negative consequences for themselves caused by the patient’s illness. They have a feeling of control over the illness and that the patient can also have control. They do not think that the patient is responsible for his or her illness. They do not see themselves as responsible for the appearance of the illness. Treatment is perceived as helpful for better controlling the illness. For these caregivers, the illness has meaning and coherence. They more or less agree that the illness brings about negative emotions like sorrow or anxiety, measured here under the category of emotional representations. Participants in this study mostly used problem-focused and social support-focused coping styles. Emotion-focused coping was used less compared to the two other styles (median = 15, min to max = 7.00–28.00).

Correlations between illness representations and coping styles are presented in Table  4 . It can be seen that all the statistically significant correlations are somewhat low ranging from r = 0.23 to r = 0.41. The statistically significant correlations are the following:

Representations that the illness brings about negative consequences for the caregiver have a moderate positive correlation with problem-focused coping ( r = 0.31, p = 0.006) and emotion-focused coping ( r = 0.35, p = 0.002). There was also a slight negative correlation between these representations and social support-focused coping ( r = −0.29, p = 0.012).

Representations that the illness brings about negative consequences for the patient have a slight positive correlation with emotion-focused coping ( r = 0.23, p = 0.037).

The presence of a feeling of control by the caregiver is moderately positively correlated with problem-focused coping ( r = 0.31, p = 0.006).

Representations that the caregiver might be responsible for the onset of the illness have a moderate positive correlation with emotion-focused coping ( r = 0.34, p = 0.001).

Representations that the patient is responsible for the onset of the illness have a slight positive correlation with emotion-focused coping ( r = 0.25, p = 0.020).

Representations that treatment helps to control the illness have a slight positive correlation with problem-focused coping ( r = 0.23, p = 0.040).

A lack of meaning attributed to the illness has a moderate positive correlation with emotion-focused coping ( r = 0.31, p = 0.005).

Representations that the illness brings about negative emotions have a moderate positive correlation with emotion-focused coping ( r = 0.41, p = 0.000) and a moderate negative correlation with social support-focused coping ( r = −0.39, p = 0.000).

It is timely to note that this study is the first known to use French-validated tools to measure correlations between representations of schizophrenia and coping styles within a caregiver sample. For most variables, our results align with several other studies in this field [ 8 , 9 , 13 , 15 , 32 ]. Nevertheless, the caregivers in this study were less likely to have the perception that the patient was at fault or that they themselves were responsible for the onset of the illness than compared to the results published by Lobban and collaborators [ 5 ]. Our results also show that emotion-focused coping is moderately correlated with (1) negative consequences for the caregivers, (2) the feeling of being at fault, (3) the feeling that the mental health problem is not coherent and (4) the overall score of the emotional representation scale. Problem-focused coping is itself moderately correlated with (1) the perception of negative consequences for caregivers and (2) the feeling of control. Social support-focused coping has a moderate negative correlation with the overall scores of the emotional representation scale. The other correlations are non-significant or less than 0.30 with a slight scaling effect [ 33 ].

The SRM predicts that illness representations influence coping procedures and that emotional representations influence coping responses. These different factors also influence each other. More specifically, when a health threat occurs, it is handled by the processing system , the second concept of SRM. This system consists of two types of threat management: (1) the representations of illness and treatment and (2) the emotional representations. Both types can influence each other. Each type of threat management includes three steps to process information: representation (illness and treatment), coping procedures and evaluation. Regarding the first type of threat management, the first stage of data processing is the development of a representation of the disease and the treatment. This concept determines the goals and coping procedures to achieve them [ 34 ]. Coping procedures are the second stage of information processing. They are a wide range of cognitive and behavioural measures undertaken in response to the cognitive representation, such as problem-focused coping [ 22 , 34 ]. The third step of information processing is the evaluation of the cognitive treatment of information. Regarding the second type of threat management, the first step of information processing is the formation of an emotional representation [ 22 ]. Based on this emotional representation, goals are set and coping responses are determined to achieve them. Coping responses are the second step of information processing. They are a wide range of strategies focused on managing emotions that are applied in response to the emotional representation. The last step of information processing for this second type is the evaluation of the emotional treatment of the information. The results of the evaluation step produce a feedback at the preceding steps .

Our results are quite coherent with this predictive model, especially for emotion-focused coping and the score of emotional representations, consequences for the caregiver and the feeling of being at fault. Consequences for the patient and the feeling of control are associated with problem-focused coping. The fact that the consequences for the caregiver are associated as much with problem-focused coping as with emotion-focused coping underscores the presence of interactions between the emotional and cognitive variables.

A study similar to ours [ 8 ] did not establish a positive correlation between the representations of the illness that bring about negative consequences for the caregiver and problem-focused coping. It is important to note that these authors did not conceptualize coping as we did, and therefore, comparing our results is a limited exercise. However, the abovementioned correlation leads one to think that the Profamille psycho-educational program followed by a significant portion of our participants had an impact on this coping style without actually modifying the representations of the illness as conceived by the participants.

Our sample pool was particular in that it was composed of individuals belonging to family-type assistance programs and that most of the participants had followed some kind of psycho-educational program. On average, our participants had been caring for their patient for 15 years. This could explain the weak level of reliance on emotion-focused coping. Nevertheless, the correlations observed suggest that the SRM model is still valid over time.

In terms of intervention, the psycho-educational programs available for the caregivers tend to focus on knowledge acquisition of the illness and treatment and positive communication skills training to reduce stress within the family. Our data suggest that it would be of interest to take better into account the emotional representations of the illness. In everyday practice, first-line health care professionals have to care for caregivers’ painful emotions such as anger, guilt, sadness and fear. However, the required skills to help caregivers to better manage their painful emotions are difficult to define precisely. The development and evaluation of the best professional strategies to help caregivers to better manage their painful emotions appear to be an interesting line of research for the future. Refining the best professionals’ skills to help caregivers to deal with negative emotions may influence more directly emotional representations and coping skills in caregivers.

Given the scarcity of studies to compare to, further studies could be useful to better identify coping styles that improve emotional regulation without negative consequences for the patient, like acceptance or changing value systems. In this way, a holistic understanding of care, as suggested by several authors [ 35 , 36 ], could be further promoted. Thus, interventions could be focused on the development of alternatives to less effective or even painful strategies like resignation, coercion and avoidance. Furthermore, the tools used to measure emotion-focused coping tend to highlight more negative emotional coping methods like coercion, avoidance or resignation. It is important to develop a measuring instrument that also takes into account the positive strategies focused on emotion.

Limitations of the study

This was the first study using the IPQS: Relatives and the FCQ in French. Participant recruitment took place within the context of social support organizations according to a convenience sampling method. Therefore, our results could be cautiously generalized to other programs having a similar context to ours, in terms of culture and health care systems.

In the empirical literature, as well as in the Commonsense Model of Self-Regulation of Health and Illness, problem-focused and social support-focused coping styles are recognized as efficient for decreasing caregiver loads and increasing a patient’s chances of better coping with schizophrenia. The results of this study also show that illness representations influence the choice of coping styles. The coping style approach is useful to develop further targeted and feasible interventions. Further studies in this field should focus on the evaluation of interventions based on illness representations in order to prevent the use of unsuitable coping styles. Also, it will be necessary to identify this phenomenon in a caregiver population of individuals who are not involved into any mutual support group in order to adapt appropriate care and support procedures.

Abbreviations

Expressed emotion

Family coping questionnaire

Relatives, illness perception questionnaire for schizophrenia: relatives’ version

Standard deviation

Commonsense model of self-regulation of health and illness.

van Wijngaarden B, Schene A, Koeter M, Becker T, Knapp M, Knudsen HC, Tansella M, Thornicroft G, Vazquez-Barquero JL, Lasalvia A, Leese M, EPSILON Study Group: People with schizophrenia in five countries: conceptual similarities and intercultural differences in family caregiving. Schizophrenia Bulletin. 2003, 29 (3): 573-586. 10.1093/oxfordjournals.schbul.a007029.

Article   PubMed   Google Scholar  

Heider D, Angermeyer MC, Winkler I, Schomerus G, Bebbington PE, Brugha T, Azorin JM, Toumi M: A prospective study of quality of life in schizophrenia in three European countries. Schizophrenia Research. 2007, 93 (1–3): 194-202.

Lauber C, Eichenberger A, Luginbuhl P, Keller C, Rossler W: Determinants of burden in caregivers of patients with exacerbating schizophrenia. European Psychiatry: The Journal of the Association of European Psychiatrists. 2003, 18 (6): 285-289. 10.1016/j.eurpsy.2003.06.004.

Article   Google Scholar  

Barrowclough C, Lobban F, Hatton C, Quinn J: An investigation of models of illness in carers of schizophrenia patients using the illness perception questionnaire. The British Journal of Clinical Psychology/The British Psychological Society. 2001, 40 (Pt 4): 371-385.

Article   CAS   PubMed   Google Scholar  

Lobban F, Barrowclough C, Jones S: Assessing cognitive representations of mental health problems. II. The illness perception questionnaire for schizophrenia: relatives’ version. The British Journal of Clinical Psychology/The British Psychological Society. 2005, 44 (Pt 2): 163-179.

Barrowclough C, Tarrier N, Johnston M: Distress, expressed emotion, and attributions in relatives of schizophrenia patients. Schizophrenia Bulletin. 1996, 22 (4): 691-702. 10.1093/schbul/22.4.691.

Lobban F, Barrowclough C, Jones S: Does expressed emotion need to be understood within a more systemic framework? An examination of discrepancies in appraisals between patients diagnosed with schizophrenia and their relatives. Social Psychiatry and Psychiatric Epidemiology. 2006, 41 (1): 50-55. 10.1007/s00127-005-0993-z.

Fortune DG, Smith JV, Garvey K: Perceptions of psychosis, coping, appraisals, and psychological distress in the relatives of patients with schizophrenia: an exploration using self-regulation theory. The British Journal of Clinical Psychology/The British Psychological Society. 2005, 44 (Pt 3): 319-331.

Moller-Leimkuhler AM: Burden of relatives and predictors of burden. Baseline results from the Munich 5-year-follow-up study on relatives of first hospitalized patients with schizophrenia or depression. European Archives of Psychiatry and Clinical Neuroscience. 2005, 255 (4): 223-231. 10.1007/s00406-004-0550-x.

Magliano L, Guarneri M, Marasco C, Tosini P, Morosini PL, Maj M: A new questionnaire assessing coping strategies in relatives of patients with schizophrenia: development and factor analysis. Acta Psychiatrica Scandinavica. 1996, 94 (4): 224-228. 10.1111/j.1600-0447.1996.tb09853.x.

Knudson B, Coyle A: Parents’ experiences of caring for sons and daughters with schizophrenia: a qualitative analysis of coping. European Journal of Psychotherapy & Counselling. 2002, 5 (5): 169-183.

Barrowclough C, Parle M: Appraisal, psychological adjustment and expressed emotion in relatives of patients suffering from schizophrenia. The British Journal of Psychiatry: The Journal of Mental Science. 1997, 171: 26-30. 10.1192/bjp.171.1.26.

Article   CAS   Google Scholar  

Magliano L, Fadden G, Economou M, Xavier M, Held T, Guarneri M, Marasco C, Tosini P, Maj M: Social and clinical factors influencing the choice of coping strategies in relatives of patients with schizophrenia: results of the BIOMED I study. Social Psychiatry and Psychiatric Epidemiology. 1998, 33 (9): 413-419. 10.1007/s001270050074.

Birchwood M, Cochrane R: Families coping with schizophrenia: coping styles, their origins and correlates. Psychological Medicine. 1990, 20 (4): 857-865. 10.1017/S0033291700036552.

Magliano L, Fadden G, Economou M, Held T, Xavier M, Guarneri M, Malangone C, Marasco C, Maj M: Family burden and coping strategies in schizophrenia: 1-year follow-up data from the BIOMED I study. Social Psychiatry and Psychiatric Epidemiology. 2000, 35 (3): 109-115. 10.1007/s001270050192.

Gonzalez-Torres MA, Oraa R, Aristegui M, Fernandez-Rivas A, Guimon J: Stigma and discrimination towards people with schizophrenia and their family members. A qualitative study with focus groups. Social Psychiatry and Psychiatric Epidemiology. 2007, 42 (1): 14-23. 10.1007/s00127-006-0126-3.

Magliano L, Fadden G, Madianos M, de Almeida JM, Held T, Guarneri M, Marasco C, Tosini P, Maj M: Burden on the families of patients with schizophrenia: results of the BIOMED I study. Social Psychiatry and Psychiatric Epidemiology. 1998, 33 (9): 405-412. 10.1007/s001270050073.

Magliano L, Veltro F, Guarneri M, Marasco C: Clinical and socio-demographic correlates of coping strategies in relatives of schizophrenic patients. European Psychiatry: The Journal of the Association of European Psychiatrists. 1995, 10 (3): 155-158. 10.1016/0767-399X(96)80104-0.

Madianos M, Economou M, Dafni O, Koukia E, Palli A, Rogakou E: Family disruption, economic hardship and psychological distress in schizophrenia: can they be measured?. European Psychiatry: The Journal of the Association of European Psychiatrists. 2004, 19 (7): 408-414. 10.1016/j.eurpsy.2004.06.028.

Leventhal H, Leventhal EA, Schafer P: Vigilant coping and health behavior: a life span problem. Aging, Health, and Behavior. Edited by: IMORA. 1991, Baltimore: Johns Hopkins, 109-140.

Google Scholar  

Magliano L, Guarneri M, Fiorillo A, Marasco C, Malangone C, Maj M: A multicenter Italian study of patients’ relatives’ beliefs about schizophrenia. Psychiatric Services. 2001, 52 (11): 1528-1530. 10.1176/appi.ps.52.11.1528.

Leventhal H, Leventhal EA, Cameron LD: Representations, procedures, and affect in illness self-regulation: a perceptual-cognitive approach. Handbook of Health Psychology. Edited by: Baum AS, Revenson TA, Singer JE. 2001, New York: Erbaum, 19-48.

Mitsonis C, Voussoura E, Dimopoulos N, Psarra V, Kararizou E, Latzouraki E, Zervas I, Katsanou MN: Factors associated with caregiver psychological distress in chronic schizophrenia. Social Psychiatry and Psychiatric Epidemiology. 2010, 47 (2): 331-337.

Koukia E, Madianos MG: Is psychosocial rehabilitation of schizophrenic patients preventing family burden? A comparative study. Journal of Psychiatric and Mental Health Nursing. 2005, 12 (4): 415-422. 10.1111/j.1365-2850.2005.00852.x.

Magliano L, Fiorillo A, Malangone C, Marasco C, Guarneri M, Maj M, National Mental Health Project Working Group: The effect of social network on burden and pessimism in relatives of patients with schizophrenia. The American Journal of Orthopsychiatry. 2003, 73 (3): 302-309.

Magliano L, Fiorillo A, Malangone C, De Rosa C, Maj M, Family Intervention Working Group: Implementing psychoeducational interventions in Italy for patients with schizophrenia and their families. Psychiatric Services. 2006, 57 (2): 266-269. 10.1176/appi.ps.57.2.266.

Hodé Y, Krychowski R, Beck N, Vonthron R, Rouyèr N, Huentz M, Risacher P, Fattah S, Pasche M, Pasche P: Effet d’un programme psychoéducatif sur l’humeur des familles des malades soufrant de schizophrénie. Journal de thérapie comportementale et cognitive. 2008, 18: 104-107. 10.1016/j.jtcc.2008.06.003.

Stalberg G, Ekerwald H, Hultman CM: At issue: siblings of patients with schizophrenia: sibling bond, coping patterns, and fear of possible schizophrenia heredity. Schizophrenia Bulletin. 2004, 30 (2): 445-458. 10.1093/oxfordjournals.schbul.a007091.

Moller-Leimkuhler AM: Multivariate prediction of relatives’ stress outcome one year after first hospitalization of schizophrenic and depressed patients. European Archives of Psychiatry and Clinical Neuroscience. 2006, 256 (2): 122-130. 10.1007/s00406-005-0619-1.

Schene AH, van Wijngaarden B: A survey of an organization for families of patients with serious mental illness in The Netherlands. Psychiatric Services. 1995, 46 (8): 807-813.

Cromier H, Guimond G, Joncas J, Leblanc G, Morin R, Vaillancourt S: Profamille, programme d’intervention de groupe auprès des familles de personnes atteintes de schizophrénie. 1991, Québec: Unité de psychiatrie sociale et préventive du CHUL et Centre de recherche de l’Université Laval Robert Giffard

Kartalova-O’Doherty Y, Doherty DT: Coping strategies and styles of family carers of persons with enduring mental illness: a mixed methods analysis. Scandinavian Journal of Caring Sciences. 2008, 22 (1): 19-28. 10.1111/j.1471-6712.2007.00583.x.

Article   PubMed Central   PubMed   Google Scholar  

Field AP: Discovering Statistics Using SPSS: (and Sex and Drugs and Rock 'n' Roll). 2009, Thousand Oaks: Sage, 3

Leventhal H, Leventhal EA, Contrada RJ: Self regulation, health, and behavior: a perceptual-cognitive approach. Psychology and Health. 1998, 13: 717-733. 10.1080/08870449808407425.

Dallaire C, Toupin I: Les théories à spectre modéré, les théories propres aux situations et les interventions infirmières. Le savoir infirmier: au coeur de la discipline et de la profession infirmière. Edited by: Dallaire IC. 2008, Boucherville: Gaëtan Morin, 239-264.

Pépin J, Kérouac S, Ducharme F: La pensée infirmière. 2010, Montréal: Chenelière Éducation, 3

Download references

Acknowledgements

This work has been supported by a donation from Dr. Alexander Engelhorn.

Author information

Authors and affiliations.

Community Psychiatry Service, Department of Psychiatry, University Hospital Centre of Vaud, Site de Cery, Prilly, 1008, Switzerland

Shyhrete Rexhaj, Charles Bonsack & Jérôme Favrod

School of Nursing Sciences, La Source, University of Applied Sciences of Western Switzerland, Av. Vinet 30, Lausanne, 1004, Switzerland

Shyhrete Rexhaj, Nataly Viens Python & Jérôme Favrod

Institut Universitaire de Formation et de Recherche en Soins, University of Lausanne, Biopôle 2, Rue de la Corniche 10, Lausanne, 1010, Switzerland

Diane Morin

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Shyhrete Rexhaj .

Additional information

Competing interests.

The authors declare that they have no competing interests.

Authors’ contributions

SR, DM and JF contributed to the conception and design of the study. SR contributed to the acquisition of data. SR and JF performed the statistical analysis and drafted the first manuscript. NVP and CB critically reviewed and revised the manuscript. All authors read and approved the final manuscript.

Authors’ original submitted files for images

Below are the links to the authors’ original submitted files for images.

Authors’ original file for figure 1

Authors’ original file for figure 2, authors’ original file for figure 3, authors’ original file for figure 4, rights and permissions.

This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License ( http://creativecommons.org/licenses/by/2.0 ), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Reprints and permissions

About this article

Cite this article.

Rexhaj, S., Python, N.V., Morin, D. et al. Correlational study: illness representations and coping styles in caregivers for individuals with schizophrenia. Ann Gen Psychiatry 12 , 27 (2013). https://doi.org/10.1186/1744-859X-12-27

Download citation

Received : 12 June 2013

Accepted : 20 August 2013

Published : 28 August 2013

DOI : https://doi.org/10.1186/1744-859X-12-27

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

• Representations of schizophrenia
• Nursing care

Annals of General Psychiatry

ISSN: 1744-859X

• Submission enquiries: [email protected]

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
• R Soc Open Sci
• v.9(12); 2022 Dec

Published correlational effect sizes in social and developmental psychology

Josefína weinerová.

1 Department of Psychology, University of Cambridge, Cambridge, England

Dénes Szűcs

John p. a. ioannidis.

2 Meta-Research Innovation Center at Stanford (METRICS), Stanford University, Stanford, CA, USA

Associated Data

• Weinerová J, Szűcs D, Ioannidis JPA. 2022. Published correlational effect sizes in social and developmental psychology. Dryad Digital Repository . ( 10.5061/dryad.bg79cnpdw) [ CrossRef ]
• Weinerová J, Szűcs D, Ioannidis JPA. 2022. Published correlational effect sizes in social and developmental psychology. Figshare . ( 10.6084/m9.figshare.c.6350055) [ CrossRef ]

All data and code are available from the Dryad Digital Repository: https://doi.org/10.5061/dryad.bg79cnpdw [ 15 ]. The protocol and analysis plans were preregistered via OSF and can be accessed at https://osf.io/u96yn/ .

Supplementary material is available online [ 38 ].

The distribution of effect sizes may offer insights about the research done and reported in a scientific field. We have evaluated 12 412 manually collected correlation effect sizes (Sample 1) and 31 157 computer-extracted correlation effect sizes (Sample 2) published in journals focused on social or developmental psychology. Sample 1 consisted of 243 studies from six journals published in 2010 and 2019. Sample 2 consisted of 5012 papers published in 10 journals between 2010 and 2019. The 25th, 50th and 75th effect size percentiles were 0.08, 0.17 and 0.33, and 0.17, 0.31 and 0.52 in Samples 1 and 2, respectively. Sample 2 percentiles were probably larger because Sample 2 only included effect sizes from the text but not from tables. In text authors may have emphasized larger correlations. Large sample sizes were associated with smaller reported correlations. In Sample 1 about 70% of studies specified a directional hypothesis. In 2010 no papers had power calculations, while in 2019 14% of papers had power calculations. These data offer empirical insights into the distribution of reported correlations and may inform the interpretation of effect sizes. They also demonstrate the importance of computation of statistical power and highlight potential reporting bias.

1.  Introduction

Calculating effect sizes allows researchers to characterize the magnitude and practical importance of their findings (i.e. substantive significance), to compare findings across studies and determine statistical power and required sample sizes. However, interpreting effect sizes is not necessarily straightforward. Cohen [ 1 ] proposed conventional benchmarks, according to which an effect size can be considered small, medium or large. While this approach is widely used, it lacks context: what can be considered small or large effect size may differ between different research fields. Hence, using universal effect size benchmarks across different fields and research areas may lead to the overestimation or underestimation of effect size magnitudes in specific fields [ 2 , 3 ]. Due to this problem, researchers proposed that a particular effect size should be compared with the specific distribution of published effect sizes [ 2 – 4 ]. Nevertheless, field-specific published effect size distributions may also misrepresent true effect sizes because published effect sizes may be highly exaggerated [ 5 , 6 ]. To compare the above approaches, here, we explored the distribution of published statistically significant and non-significant effect sizes and compared them with Cohen's [ 1 ] effect size benchmarks in the fields of social and developmental psychology. While some previous studies on the topic have included non-significant effect sizes, to the best of our knowledge none have compared statistically significant and non-significant effect sizes in their analyses. However, this is important as statistically non-significant effect sizes may be less exposed to effect size exaggeration. When compiling effect size distributions, we have also considered the interplay of degrees of freedom (in the case of correlation degrees of freedom = sample size minus 2) and published effect sizes as studies with larger degrees of freedom tend to publish smaller statistically significant effect sizes [ 6 , 7 ].

Table 1 shows results from studies in the psychological science and related areas evaluating effect size distributions in specific fields [ 2 – 4 , 8 , 9 , 11 , 12 ]. Apart from Hemphill [ 8 ] and Hartgerink et al . [ 10 ], all of these studies used the 25th, 50th and 75th percentiles as proxies for small, medium and large effect sizes. While Cohen did not suggest using these percentiles as benchmarks, their use is motivated by the definition of effect sizes put forward by Cohen [ 13 ]. He defined a medium effect ‘ likely to be visible to naked eye of a careful observer (It has since been noted in effect size surveys that it approximates the average size of observed effects in various fields) ’ [ 14 , p. 156]. Small effect size was defined as noticeably smaller than medium effect size but non-trivial. Large effect size was defined as different from the medium effect size by the same degree but in the opposite direction [ 13 ].

Table 1.

Summary of studies analysing effect size distribution across different fields and subfields in the psychological sciences and related areas. Note. If other types of effect sizes were collected, they were transformed to the type used for the analysis. NA = data not available. Type of effect size is specified in the 3rd column.

The percentile values from the above studies can be used as an empirical comparison against Cohen's benchmarks. For example, a recent analysis of 708 correlation coefficients collected from meta-analyses focusing on individual differences studies estimated the 25th, 50th and 75th percentiles of the effect size distribution at r = 0.11, r = 0.19 and r = 0.29, respectively [ 4 ]. These effect sizes are much lower than effects generally considered small, medium and large, respectively. Conversely, Quintana [ 3 ] looked at effect size estimates from meta-analyses of heart rate variability case control studies and found that Cohen's benchmarks slightly underestimate the effect size magnitudes in that area of research. Certainly, it should be acknowledged that the observed effect sizes may represent a combination of true effects and bias. Bias typically (but not necessarily always) will tend to make them bigger.

1.1. Limitations of studies on effect size distributions

To date, most studies focusing on the estimation of effect size distributions in different fields have used data from meta-analyses. This approach has the advantage of accessing effect sizes from a specific field or related to a specific question, but it is subject to limitations. First, meta-analyses are at risk of relying on exaggerated published effect sizes. Second, the decision to exclude or include certain papers represents a secondary source of researcher degrees of freedom in addition to the one present in the primary literature. Third, only a few relevant papers to date have incorporated statistically non-significant effect sizes in its sample [ 3 , 10 ]. However, including these effect sizes in analyses is important because non-significant results are probably less exposed to effect size inflation bias than statistically significant results. Fourth, while a few relevant studies commented on the relationship of sample sizes and statistical significance [ 2 , 3 , 12 ], to the best of our knowledge no studies focused on the interpretation of effect sizes have examined how effect size distribution will be affected by sample size distribution.

1.2. The current study

Here, we address four limitations of the literature. First, we collected a large amount of data from the primary literature including correlations, sample sizes and p -values. Second, considering sample sizes in analyses allowed us to determine how effect size distributions vary as a function of sample size. Third, we collected both statistically significant and non-significant effect sizes that allowed us to gain a more balanced impression about expected effect sizes. Fourth, to gain an impression of potentially changing research practices we also determined the temporal change in effect size and sample size distributions during the last decade.

We collected two samples of correlation coefficients from the social and developmental psychology literature. In Sample 1 we have manually extracted 12 412 records of statistical information including correlation coefficients ( r ), sample size ( N ) and p -values from 178 papers including 243 studies published in 2010 and 2019 in six major journals in social and developmental psychology. By records, we mean each individual occurrence of correlation coefficients fulfilling the inclusion criteria within the study. In Sample 2 we have used a computer algorithm to extract 31 157 statistical records: r values, degrees of freedom (d.f.) and p -values from all papers published in 10 social and developmental psychology journals (six of which have been used for Sample 1 data collection as well) between the years 2010 and 2019. The data enabled us to assess the distribution of published sample sizes and correlation effect sizes, compare them between two subfields of psychology and to understand how correlation distributions have changed during the past decade.

2.  Methods

The study has been preregistered on the Open Science Framework at https://osf.io/u96yn/ .

2.1. Sample 1

Sample 1 consisted of 12 412 correlation coefficients and related statistical information which were extracted from 243 studies published in 2019 and 2010. For the year 2019, we collected 6895 correlations from 127 studies. For the year 2010, we collected 5517 correlations from 116 studies. All records collected were included in the analysis.

In our preregistration, we had set out the aim to collect records from half a year's worth of issues for each journal in each year. This turned out to be unmanageable due to the number of potential records.

We manually extracted statistical information from social and developmental psychology papers published as pdf files in 2010 and 2019. Information on journals, papers and studies is shown in table 2 . We sampled papers from three journals focused more on social psychology and three journals focused more on developmental psychology. The specialization of the journals was determined using the subject categories on SCImago Journal and Country Rank website [ 14 ]. We specifically chose those journals that focus primarily on empirical articles and have issues available for both the year 2019 and 2010. Additionally, the three journals for each subfield were selected so that their impact factors span different levels to keep the sample as representative of the field as possible.

Table 2.

Data collected in Sample 1. Note. The table shows the names, 5-year impact factors and issues of journals used for Sample 1 data collection. In the papers/studies column, we see the number of papers (before slash) and the number of studies (after slash) collected from a given journal in either 2010 or 2019. Data were collected from all eligible papers contained in the mentioned issues. 5-year IF = 5-year impact factor; papers = number of collected separate scientific reports published as a pdf file; studies = number of experiments with separately defined sample, methods and results reported within the papers; records = number of statistical records collected from each journal (1 record = 1 correlation with associated data).

Issues from each journal were selected at random without previously reading any papers. Once we realized that collecting half a year worth of issues was unrealistic, we tried to balance the choice of issues over the whole year, but this was not always possible due to some data being already collected at this point. Data for year 2019 were collected prior to data for year 2010. Data from all eligible studies within these selected issues were extracted. If there were multiple studies reported within a paper, the data for each study were recorded separately.

Studies were included in the data if they reported correlation coefficients in their Results section or in the electronic supplementary material. The method for data extraction was developed on a sample of 58 studies from 33 papers (2923 records) published in the 2019 issues from all six journals in table 2 . These papers were part of the final sample as well.

The pdf files were downloaded from the online platforms of each journal. The extracted data included: all correlation coefficients reported in Results sections or electronic supplementary material, in tables and text; reported significance levels or p -values corresponding to these correlation coefficients; the type of correlation coefficient used; whether the correlation was computed between two different constructs or between the same construct measured at two time points; whether correlation coefficients were reported in the Results sections or in electronic supplementary material and whether correlation coefficients were reported in correlation tables or in the text.

Additionally, we also extracted the following information for each study: journal name; first author of study; topic of the paper; overall sample size; use of power calculation to estimate sample size; whether alternative hypotheses predicted a directional effect, an effect in either direction, or a threshold value at which the effect would be considered large enough to provide evidence for the hypothesis; any comments on the null hypothesis or the null hypothesis set other than specifying the null hypothesis as r = 0 correlation. For the 2019 papers, we also recorded whether each study was preregistered (preregistration was not yet typically used in psychology in 2010).

When examining reported alternative hypotheses, we found that it was not possible to distinguish between hypotheses set a priori and post hoc . This caveat also applies to preregistered studies as it is possible that only certain correlations were included in the preregistration. When study preregistration was available, the hypotheses reported in the published studies and in the preregistrations were compared.

The median number of correlations extracted per study was 18 for the year 2010 and 34 for 2019. This increase in the median of the number of correlations published per study was mostly due to the Social Psychological and Personality Science journal (SPPS; increase by 200%), the Journal of Personality and Social Psychology (JPSP; increase by 150%) and the Journal of Child Psychology and Psychiatry (JCPP; increase by 94%). For the Developmental Psychology journal (DevPsy) the median number of correlations per study increased by 42% and for the Child Development journal (ChildDev) by 32%. For the European Journal of Personality (EJP) the number of correlations published in 2019 was 30% lower than the one for 2010.

2.2. Sample 2

In total 31 157 records were extracted for Sample 2. Of these records, 579 came from papers also included in Sample 1 (see details below).

Sample 2 was collected by an automated text mining algorithm adapted from Szűcs & Ioannidis [ 6 ]. We collected data from journal pdf files published in the same six journals as used in Sample 1 and the following four additional journals: Journal of Applied Developmental Psychology (5-year impact factor: 2.905), Journal of Experimental Child Psychology (5-year impact factor: 3.366), Journal of Experimental Social Psychology (5-year impact factor: 3.666) and Journal of Research in Personality (5-year impact factor: 3.365). All issues published between 2010 and 2019 were scanned for data. Table 3 summarizes the number of papers and the number of correlation records per paper by year, subfield and journal.

Table 3.

The total number of records/papers extracted for different subfields and journals for the years 2010–2019. Note. ChildDev = Child Development, DevPsy = Developmental Psychology, JAppDevPsy = Journal of Applied Developmental Psychology, JCPP = Journal of Child Psychology and Psychiatry, JExpChildPsy = Journal of Experimental Child Psychology, EJP = European Journal of Personality, JExpSocPsy = Journal of Experimental Social Psychology, JPSP = Journal of Personality and Social Psychology, JResInPer = Journal of Research in Personality, SPPS = Social Psychological and Personality Science. The absence of r values in specific issues of journals is due to character encoding problems which have prevented our algorithm from extracting the data.

2.2.1. Computerized extraction for Sample 2

The computer algorithm searched for specific word and symbol combinations for reporting r values, degrees of freedom and p -values. The algorithm searched the text of papers but not the tables for data records. In psychology, r values are often reported as ‘ r (d.f.) = x.xx, p = y.yy’. The algorithm thus first identified whether pdf files included the symbol combination ‘ r =’ or ‘ r (d.f.)=’ neglecting spaces between the characters ‘ r ’, ‘(‘, ‘)’, ‘=’ and the string ‘d.f.’. If such combination was identified, the algorithm then extracted 56 characters starting with the ‘ r =’.

Numerical values right after ‘ r =’ and in the range of −1 to 1 were detected as r values. Values included in parentheses after ‘ r ’ (e.g. r (d.f.) =) were detected as degrees of freedom. Values reported after correlation values and preceded by ‘ p =’, ‘ p <’ or ‘ p >’ were detected as p -values (irrespective of intervening spaces). The algorithm collected r values even in the case when degrees of freedom or p -values were not reported. The algorithm script is available at doi:10.5061/dryad.bg79cnpdw [ 15 ].

2.2.2. Validation of the extraction procedure

Throughout the algorithm development phase, the efficiency of the text mining algorithm was validated by about a dozen separate checks on text mining outcomes. When errors were found, the text mining algorithm was perfected further to avoid the detected errors.

Error checks revealed that the algorithm has misidentified some negative correlations as positive. This happened due to very difficult-to-predict idiosyncratic changes in the character coding of pdf files. In order to avoid any sign errors we have only used absolute values in the analysis of Sample 2.

After the above initial checks to further validate the extraction algorithm the second phase of validation included drawing a sample of 20 random papers from the final data sample and manually verifying the accuracy of data extraction. The algorithm performed well. The randomly selected papers for validation included 392 data records in the text of papers. The algorithm has successfully identified 93% of them. The algorithm has missed 7% of r values and did not commit any false positive errors. The most common causes of missing records were line breaks within a correlation report and the use of subscript characters after the r value (e.g. ‘ r extraversion =’ ). The list of studies used for validation can be found in the electronic supplementary material.

Additionally, during the data analysis we have found that the algorithm has not detected any data in some journals for some years. Upon checking these journals, we found this was due to character encoding issues within the pdf versions of the article. These issues do not allow for the possibility to even search for the symbol combination ‘ r =’ using the search bar. However, we are interested in whole subfields rather than specific journals and this problem has not created an imbalance in the amount of data for each subfield.

2.3. Overlap between Sample 1 and Sample 2

In Sample 1 there were 752 correlation values from 100 papers which were detected in text and could therefore theoretically be detected by the computerized extraction for Sample 2. Out of those, 28 papers with 173 values were not detected by the computerized extraction (e.g. because of special symbols used or because of verbal description of correlations, such as ‘correlation was 0.38’ instead of ‘ r = 0.38’). This means that in total there was 72 papers and 579 correlation values which were included in both samples. The computerized extraction detected additional 90 correlation values in those shared papers. Considering that during validation the computer algorithm did not falsely detect any non-existent correlations, Sample 2 probably included the 90 additional records because Sample 1 records were collected only from the Results sections of the papers whereas the computerized extraction method collected all r values from all sections of papers. The distribution of correlation values from the overlapping studies is shown in figure 1 .

The density distribution of r values from overlapping papers in Sample 1 ( N = 579) and Sample 2 ( N = 669).

Figure 1 shows overall high level of correspondence between the two samples. The 25th, 50th and 75th percentiles of r values in overlapping studies were 0.18, 0.29 and 0.47 for Sample 1 data and 0.19, 0.30 and 0.50 for Sample 2 data. This suggests that the additional correlations detected by the computerized extraction ( N = 90) tended towards larger effect sizes (also consistent with the peak in density distribution between 0.75 and 1.00 for Sample 2 data in figure 1 ).

2.4. General data analysis

All analysis steps were performed in R programming software [ 16 ]. The code used for computation, analysis and visualizations can be found at https://osf.io/x45mj/ . All analyses were done on absolute (unsigned) r values.

3.  Results

3.1. sample 1, 3.1.1. distributions of r values and sample sizes.

The number of r values reported in one study ranged between 1 and 441, with a median of 26 r values. Ninety-four per cent of the collected r values were reported in correlation tables. Studies with larger sample size reported more correlations in tables: studies with sample size less than 100 reported 1080 correlation values and 85% of these were presented in tables. By contrast, studies with sample size greater than or equal to 100 reported 10 606 records and 95.6% of these were presented in tables. This could have important implications for studies using computerized methods to extract statistical information. Our data suggest that computerized methods extracting only data from text will not detect the majority of correlational effect sizes presented within the papers.

To better understand the interaction between sample size and r value magnitude we have mapped the cumulative 25th, 50th and 75th percentiles of correlation values by the maximum degrees of freedom associated with correlation values. Figure 2 a shows that the magnitude of the percentiles decreased with including larger and larger degrees of freedom. As the percentiles for the larger degrees of freedom also include all r values with smaller degrees of freedom this suggests that with increasing degrees of freedom, overall sample r values are becoming smaller. Note that the percentile values were not weighted by sample size. That is, even without giving larger weight to larger studies as typically done in meta-analyses there is a substantial decrease in r values due to larger studies reporting much smaller effect sizes than smaller studies. Figure 2 b confirms this trend also for data from Sample 2.

( a ) The cumulative 25th, 50th and 75th percentiles of absolute r values by maximum degrees of freedom associated with correlation values for ( a ) Sample 1 ( N = 12 406) and ( b ) Sample 2 ( N = 3292). r values without corresponding degrees of freedom were excluded from this analysis. Degrees of freedom are plotted on log 10 scale. The magnitude of 25th, 50th and 75th percentile decreases with increasing d.f. in both samples. The black triangles denote the median correlation for every 1000 values of Sample 1 and every 300 values in Sample 2 (we have used different amounts for this calculation because Sample 2 data have less available values for this analysis. The median of correlation values for groups within the data follows the same broad trend as the cumulative percentiles and decreases with the increase in degrees of freedom.

We were interested in seeing how the distribution of reportedly significant and non-significant r values compared with the significance boundary. To this end, figure 3 a , b shows the bivariate distribution of sample sizes and r values, for records reported as statistically significant and non-significant, respectively. The 25th, 50th and 75th percentile for all (statistically significant and non-significant) 12 412 correlations was 0.08, 0.17 and 0.33. The median sample size was 230, the 25th and 75th percentiles were 140 and 564 respectively. The density of statistically significant r values was highest near the significance boundary. There is also high concentration of non-significant r values with r < 0.1 between 100 < N < 300.

Bivariate distribution of sample sizes and r values for records reported as ( a ) statistically significant and ( b ) non-significant. The count of observations is colour coded. Sample sizes are plotted on a log 10 -spaced scale. The red line denotes the significance threshold p ≤ α , where α = 0.05 for two-sided test. The blue line denotes the significance threshold p ≤ α , where α = 0.05 for one-sided test.

Figure 3 a,b also indicates that some reportedly significant and non-significant r values cross the significance boundary. There are only few reported correlations which are too small to be statistically significant ( figure 3 a ), and this may represent misreporting of significance status or statistical information. There are more reported correlations which are too large to be non-significant ( figure 3 b ); in these cases, misreporting could also be at play, but it is also possible that some adjustment for multiplicity has been performed by the authors.

A total of 7163 (58%) correlations were reported as statistically significant, and 4825 (39%) correlations were reported as non-significant. The remaining 3% (424 correlations) were reported without specified significance. A total of 144 correlations were reported as r = 0.

Figure 4 a shows the probability density of sample sizes for statistically significant and non-significant r values. Figure 4 b , c shows the probability density and cumulative density of correlation values, respectively. As expected, with increasing sample size, the proportion of significant r values increases. This is because with larger sample size, smaller effect size will be detected as significant. For significant correlations, the 25th, 50th and 75th r value percentiles were 0.18, 0.29 and 0.44. For non-significant correlations, these percentiles were 0.03, 0.06 and 0.11, respectively.

( a ) The univariate density distribution of sample sizes separately depending on whether the r value associated with given sample size was reported as significant ( N = 7163) or non-significant ( N = 4825). While the two distributions are largely overlapping with smaller sample size values, the prevalence of significant values increases at larger sample sizes. ( b ) The univariate density distribution of correlation values depending on their significance. ( c ) The cumulative probability distribution of correlation values depending on their significance.

In Sample 1, 752 records were collected from text (6%). Out of those 119 records had no specified significance status. We were interested in whether authors do report significant correlations more in text rather than tables. Our results would indicate that this is correct, as about 83% of correlations in the text were statistically significant, whereas only 59% of correlations in tables were statistically significant. However, the overall percentage of in-text correlations is very low in Sample 1, which makes this result tentative.

The 25th, 50th and 75th percentiles of r values were 0.16, 0.28 and 0.46 for those reported in text and 0.07, 0.17 and 0.32 for those reported in tables. Table 4 shows the 25th, 50th and 75th percentiles for those correlations reported as significant and non-significant in text and tables.

Table 4.

The proportions and 25th, 50th and 75th percentiles for correlations reported in text and in tables depending on the reported significance.

We also wanted to compare the distribution of r values and sample sizes across different years and subfields of psychology. Table 5 shows r value quartiles for different years and for different research areas. We calculated the exact p -values for those r values which were reported without a p -value or α ( N = 424) and found that 384 r values were significant at α = 0.05 assuming a two-sided test. 396 r values were significant at α = 0.05 assuming a one-sided test. Note that assuming a one-sided test increases power, therefore smaller effect sizes will be detected as statistically significant.

Table 5.

The 25th, 50th, and 75th percentiles of r values for subsets of data for the years 2010 and 2019 and for journals falling under social and developmental psychology. Note. r -value quartiles are shown separately for correlations reported as significant and non-significant. In developmental psychology 4558 (57%) records were reported as statistically significant, 3110 (39%) records were reported as non-significant, and 342 (4%) records were reported with no p -value. In social psychology 2605 (59%) records were reported as significant, 1715 (39%) as non-significant and 82 (2%) with no p -value.

Figure 5 shows the expected and measured values of non-significant correlations. The disparity between the observed and expected values of non-significant effect sizes could be caused by the fact that non-significant results are a mixture of some results that arise from studies that target nil-null effects and some other studies that target non-null effects but end up being non-significant, or by the tendency to preferentially report ‘just non-significant’ values within the studies' results.

The expected and measured mean values of non-significant correlation effect sizes compared with the significance thresholds for two-sided (dotted line) and one-sided (solid line) test. Expected mean of non-significant correlation effect sizes if the null hypothesis was true (leftmost dashed line) was calculated as an integral of the critical threshold and 0, given particular degrees of freedom.

3.1.2. Differences between years

In total there were 5517 r values for the year 2010 and 6895 for the year 2019. Table 5 summarizes the 25th, 50th and 75th percentiles of r values and sample sizes. There is a notable increase in the median sample size from 2010 to 2019. Conversely, r values remain remarkably similar. This is curious, as with increasing sample size, the effect size which will be detected as significant decreases.

3.1.3. Differences between subfields

Table 5 also summarizes the 25th, 50th and 75th percentiles for r values and sample sizes for the subfields of social and developmental psychology. Sample sizes increased for both developmental (median: 147 to 267) and social psychology (median: 145 to 300) from 2010 to 2019. Conversely, r values were similar between the two fields, and they also did not change over time.

3.1.4. Specification of study hypotheses

Studies were coded as having a directional hypothesis if at least one hypothesis specified the sign of an effect. Out of the 243 studies, 176 (72%) contained at least one hypothesis where the sign of the effect was specified; 67 of the studies (28%) had hypotheses which did not specify the expected sign of the effects. In 2010 there were 79 studies (68%) specifying the sign of an expected effect and 37 (32%) without specifying the sign. In 2019, there were 97 studies (76%) with directional hypothesis and 30 (24%) without a directional hypothesis. When using nil-null hypothesis and assuming high enough sample size, directional hypotheses have 50% chance of being found significant and non-directional hypothesis literally 100% chance. We have, however, not collected data on how the given hypotheses were actually tested.

Out of the 127 studies published in 2019, 18 (14%) included a power analysis. The expected effect size was reported in 13 of those. The rest did not specify the effect size, or the power calculation was done a posteriori on the detected effect. No power calculations were reported in studies published in 2010.

3.1.5. Preregistrations

Seven studies in Sample 1 (4%) contained a link to a preregistration document. These studies included 329 correlations and came from five papers (three studies were part of one paper). The studies were published in the European Journal of Personality (five studies) and Social Psychological and Personality Science (two studies). Studies were preregistered on the Open Science Framework 1 (six studies) or the AsPredicted platform 2 (one study).

In two cases there was an extra hypothesis in the preregistration not stated in the published paper. In one case no hypotheses were mentioned in the preregistration. In one case the preregistered hypotheses were more precise (directional as opposed to explorative) than those stated in the study. In the four remaining studies, hypotheses were the same in both the preregistration and in the published paper. The 25th, 50th and 75th percentiles for correlations collected from preregistered studies were r = 0.07, 0.18 and 0.35 and N = 157, 213 and 264.

3.2. Sample 2

3.2.1. distributions of r values and degrees of freedom.

A total of 31 157 correlations were extracted for the years 2010–2019. The 25th, 50th and 75th percentiles for all r values were 0.17, 0.31 and 0.52. Figure 6 shows the distribution of r values in each year. Figure 7 a shows the median r values with bootstrapped 95% confidence intervals. The median r value decreased from 0.35 in 2010 to 0.26 in 2019. Note that this is in contrast with the results in Sample 1 where the correlation values remained similar between the years. Figure 7 b , c shows the 25th, 50th and 75th percentiles for r values and degrees of freedom across the years for different subfields. Degrees of freedom were reported with 3292 r values (11%). The 25th, 50th and 75th percentiles for the degrees of freedom were 37, 72 and 144. Degrees of freedom increased from a median of 53 in 2010 to a median of 114.5 in 2019 which was also true in Sample 1. As an exception from this trend, there was a decrease in degree of freedom from the year 2018 (median = 184) to 2019 (median = 114.5).

Probability distribution of all r values in Sample 2 per year.

( a ) Medians of all Sample 2 data across years with 95% bootstrap confidence intervals. The bootstrap method used is ‘standard normal interval’ and computed in R as part of the boot package [ 17 ]. N of simulations = 10 000. ( b ) The 25th, 50th and 75th percentiles of r values for developmental ( N = 13 514) and social ( N = 17 643) psychology. ( c ) The 25th, 50th and 75th percentiles of all collected degrees of freedom for developmental ( N = 1813) and social ( N = 1479) psychology. Degrees of freedom for social psychology have increased across the years while they have stayed about the same for developmental psychology. Degrees of freedom are plotted on a log 10 -spaced scale.

Table 6 shows the number of records with degrees of freedom for each year and the correlation between degrees of freedom and r value magnitude. In early years there was a negative correlation between the degrees of freedom and r value magnitude, but this correlation gradually disappeared by 2019.

Table 6.

The number of records including degrees of freedom per year and the Spearman's rho correlations of absolute r values and degrees of freedom for each year. Note. Spearman's rho correlation was used after the Shapiro–Wilk normality test has shown that the data is not normally distributed.

3.2.2. Differences between subfields

The comparison of subfields of psychology in Sample 2 yielded the following results. The 25th, 50th and 75th percentiles for all r values collected from journals focused on developmental psychology were 0.20, 0.35 and 0.56 ( N = 13 514). In the case of journals focused on social psychology the respective r values were 0.15, 0.28 and 0.49 ( N = 17 643). The overall percentiles of degrees of freedom were 20, 53 and 97 ( N = 1813) for developmental psychology and 51, 103 and 237 ( N = 1479) for social psychology. The sample sizes of degree of freedom values are smaller than the sample sizes of correlation values because degrees of freedom could be extracted only for a subset of the r values in Sample 2.

The 25th, 50th and 75th percentiles of r values and degrees of freedom for the two subfields across the years are shown in table 7 . The magnitudes of degrees of freedom have increased across the years while r value magnitudes have decreased for both fields.

Table 7.

The 25th, 50th and 75th percentiles for each year divided into r values and degrees of freedom published in journals specializing in developmental psychology and journals specializing in social psychology. Note. r = r values; d.f. = degrees of freedom. For both subfields the median r value for 2019 is slightly lower than the median r value for 2010. For both subfields the median d.f. for 2019 is higher than the median d.f. for 2010. There is no clear decreasing pattern across years for either subfield. The number of records collected for each group is shown.

4.  Discussion

In this study, we have determined the distribution of correlation coefficients in five developmental and five social psychology journals. We used two samples. Sample 1 included correlations from both text and tables. Sample 2 included correlations only from the text. Sample 2 correlations were larger than those of Sample 1, probably because the correlations chosen to be presented in the text were a biased subsample of all correlations calculated. Hence, larger correlations may have been reported in the text than in tables. In Sample 2 the magnitude of correlations decreased over the decade we examined, whereas they remained stable in Sample 1.

4.1. Implications for power calculations

In Sample 1 the 25th, 50th and 75th percentiles of all correlation effect sizes were r = 0.08, 0.17 and 0.33, much smaller than Cohen's estimates for small, medium and large effect sizes ( r = 0.1, 0.3 and 0.5, respectively). Considering only correlations clearly reported as statistically significant the 25th, 50th and 75th percentiles were 0.18, 0.29 and 0.44, reasonably close to Cohen's [ 1 ] estimates. However, considering only correlations reported as statistically non-significant, the respective percentiles were very small (0.03, 0.06 and 0.11). This is to be expected as larger magnitudes of effect sizes would by definition be detected as statistically significant.

If we consider the effect sizes from both statistically significant and non-significant reports to be the best overall effect size estimates, the implications for power calculations are clearly profound: the relationship between power and effect size is not linear, even a seemingly small decrease in effect size may translate to a considerable change in sample size. For example, relying on Cohen's benchmarks would suggest that a sample size of 84 is required to detect a ‘medium-sized’ effect of r = 0.3. By contrast, our data suggest that a 219% larger sample size of 268 would be necessary ( r = 0.17, power = 0.8, α = 0.05, two-sided test).

It is a question whether field-wide estimates of effect size distribution such as those presented here are the right basis for power calculation, or whether one should focus on the distribution of effect sizes from previous studies looking at similar problems. The latter approach would probably be more precise as long as studies looking at the specific question of interest are well-powered and unbiased. One has to question in each case whether the available results are less biased in the studies targeting similar questions or in the larger discipline. Discipline-wide effect size distributions can also be interesting in order to determine the overall distribution of effects one can expect within a wider field. Here we have focused on psychological sciences, but similar considerations may apply also to other scientific fields, e.g. empirical distributions of effect sizes have been studied also in medical disciplines [ 18 – 20 ].

4.2. Differences between samples

There was a pronounced difference in the percentiles of the two samples we collected. In Sample 2 the 25th, 50th and 75th percentiles were 0.1–0.2 larger than in Sample 1: 0.17, 0.31 and 0.52. This difference can be attributed to the fact that the automatic extraction algorithm used to collect Sample 2 was able to extract only correlations reported in the text. By contrast, in Sample 1 only 6% of all records were collected from text. Indeed, considering only Sample 1, we found that correlation effect sizes were about 0.1–0.2 larger in the text than in tables. This disparity may arise because correlation tables often include correlations between all observed variables whereas correlations mentioned in the text are more likely to be interpreted by the narrative of papers, so they are more likely to be related to variables in the focus of the study, or how the study had been written up. Text bodies may also contain larger correlations, because they draw more attention and are considered worth mentioning in the text, regardless of whether they reflect primary analyses or secondary, exploratory ones. Alternatively, the disparity could also be caused by the mean degrees of freedom being smaller in Sample 2 than in Sample 1. Given that in Sample 1 we can see higher proportion of significant correlations in text, the former explanation seems to play at least a partial role. In any case, this points to an important problem for researchers deciding between manual and computerized collection of correlational effect sizes to study their distribution, as computerized extraction methods usually sample only information from text.

4.3. Temporal developments in effect size and sample size distributions

The recent awareness of a high number of false positive findings in published literature [ 21 , 22 ] has led to calls to increase sample sizes in various research areas [ 6 , 23 – 25 ]. In our manually collected Sample 1 data, we found a median sample size increase in both developmental (from 147 to 267) and social psychology (145 to 300). Effect sizes remained stable in both subfields (0.17–0.18 and 0.18–0.16 in developmental and social psychology respectively). By contrast, in our Sample 2 data we found that median degrees of freedom increased in both fields but more modestly so, especially in developmental psychology. Between 2010–2012 and 2017–2019 effect sizes declined in both fields: 0.36 to 0.30 in developmental, and 0.29 to 0.23 in social psychology (Sample 2).

The correlation effect sizes in Sample 2 were negatively correlated with associated sample size in the earliest publication years we studied. However, this negative correlation gradually disappeared by 2019. This may be explained by less selective reporting of correlations in text in recent years.

4.4. Correlation distributions differ between subfields of psychology

Multiple authors pointed out that using universal effect size benchmarks may lead to the underestimation or overestimation of the effect sizes in research subfields [ 2 – 4 ]. Here, we found that the 25th, 50th and 75th r value percentiles were very similar in developmental and social psychology studies in Sample 1 (0.07, 0.17, 0.33, and 0.08, 0.18, 0.33, respectively). In Sample 2 there was more pronounced difference in r value percentiles between the two fields (0.20, 0.35 and 0.56, and 0.15, 0.28 and 0.49 for developmental and social psychology respectively). If Sample 2 picked up correlations more likely to be the foci of the studies (reported in the text), this would suggest that while the distribution of reported correlations in the two fields is very similar, the focus of the two fields is on correlations of different magnitudes. Alternatively, researchers in developmental psychology may prefer to highlight larger correlations.

The between-field comparison of r values and degrees of freedom was hindered by the fact that degrees of freedom could be extracted only for a subset of values in Sample 2, as most often they were not reported with each r value.

4.5. Sample size and effect sizes

When relying on null-hypothesis significance testing, having larger sample sizes allows one to detect smaller effect sizes as statistically significant. In fields where low sample sizes are typical this will lead to exaggeration of effect sizes in the published literature because studies with low sample sizes can only produce statistically significant results if effects are relatively large [ 26 ]. However, due to sampling variability large effects sizes will be detected from time to time even if the true effects tested are small or null [ 6 ]. This argument is supported by findings from large-scale replication efforts. The Open Science Collaboration [ 27 ] study conducted replication of 100 psychology studies and found that the mean effect size of the replications ( r = 0.197) was roughly half the magnitude of the mean effect size published in the original studies ( r = 0.403). Furthermore, a recent report focused on emotion research showed that effect sizes reported in highly cited observational and experimental studies are on average about twice larger compared with the largest sample studies on the same topic [ 7 ].

Hence, published field-specific effect size distributions probably depend on study sample sizes. Our data showed clear evidence of this expectation: We found that as records with larger and larger sample sizes were considered, smaller and smaller median r values were found.

Schäfer & Schwarz [ 2 ] have suggested an alternative explanation for the above potential regularity arguing that the negative relationship between sample size and effect size may also arise because in research fields with large effects researchers have learned that small sample size is sufficient while in fields with small effects researchers would aim for larger sample sizes. However, this suggestion cannot explain the pattern of our data: here, we observed a strong decline in the magnitude of cumulative percentiles of r values with increasing sample sizes even within the same research field and even while the decrease in r value magnitudes across years was relatively small. Hence, it is likely that published correlation magnitudes are driven by study sample sizes simply because larger studies can publish more statistically significant small effects. This suggests that researchers cannot simply determine ‘true’ expected effect sizes by looking at some published papers. Rather, the decline of effect sizes with increasing study sample size must be considered when trying to determine expected effect sizes.

4.6. Power calculations

Only 14% of the Sample 1 studies published in 2019 included a statistical power calculation (there were no statistical power calculations in studies published in 2010). This is more than twice the number reported by Szűcs and Ioannidis [ 24 ] for studies published in neuroimaging journals in 2017 (6.9% out of 130 studies) and 2018 (6.4% out of 140 studies). This may indicate either increased focus on statistical power calculations with time or may indicate that power calculations are more frequent in social and developmental psychology than in neuroimaging papers.

4.7. Preregistrations

Only seven studies in Sample 1 (4%) contained a link to a preregistration document (these studies included a total of 329 correlations). Schäfer & Schwarz [ 2 ] found that the median r value for preregistered studies was 55% lower than the median value for non-preregistered studies. However, in our sample, the 50th, and 75th r value percentiles for preregistered studies were slightly higher than those for all Sample 1, and the 25th percentile was slightly smaller. However, as our sample only included very few preregistered studies, our findings may not adequately represent effect and sample size differences between preregistered versus non preregistered studies.

It is noteworthy that very few studies in the sample were preregistered and that they have come from only two journals. This suggests that preregistration in 2019 was not yet widespread practice in many impactful journals in social and developmental psychology. Those studies that were preregistered have mainly used the Open Science Framework 3

4.8. Hypothesis precision

Multiple authors pointed out that most hypotheses in psychological science tend to be directional at best [ 28 – 31 ]. Our data suggest that the situation has remained unchanged since the 1960s. No studies specified the looked-for effect size. Twenty-eight per cent of Sample 1 studies only specified hypotheses predicting an association but not the direction of the association; 72% of Sample 1 studies contained at least one hypothesis which specified a predicted direction of an effect; 8% more studies had directional hypotheses in 2019 than in 2010.

Specifying only the sign (direction) of an effect has been shown to lead to many false positives [ 32 ]. In studies with very high statistical power there is 50% probability of rejecting the null hypotheses. When not even the direction of an effect is specified, such studies can reject the null hypotheses with near certainty [ 28 , 30 ]. If the study hypotheses were specified after the results were known (HARKing; [ 33 ]), all studies of any sample size would be almost certain to detect a significant effect. It is also noteworthy that if studies do not specify an effect size sought then principled statistical power calculation becomes impossible.

4.9. Ambient correlational noise

Most psychological variables tend to correlate with each other simply because they are affected by the interactions of many background factors [ 30 ]. Hence, any randomly selected variables are likely to be at least remotely connected through a background network of interacting variables [ 31 , 34 ], leading to shared variance. Consequently, a randomly chosen pair of variables is likely to have non-zero absolute correlation [ 30 , 35 , 36 ]. This phenomenon is termed the ‘ambient correlational noise’ [ 30 ] or the ‘crud factor’ [ 37 ]. In our data the middle 95% of the overall correlation distribution ranged between 0 and 0.64 for Sample 1. The middle 95% of the distribution of non-significant correlations was between 0 and 0.21. That is, most absolute correlation values clearly departed from zero. It is also of note that the means of observed statistically non-significant correlations were larger than the means expected if the nil-null hypothesis was true. This could be because the nil-null hypothesis is correct only in a subset of the studies and therefore nearly all studies would measure larger than zero effect sizes. Nevertheless, this could also reflect the ambient correlational noise within the data.

These points are very important to consider when ‘real’ (not ‘statistical’) significance of results is evaluated: published studies often interpret statistically significant effect sizes in the 0.1 ≤ r ≤ 0.2 range. If the sample size is large enough, departure from zero will also be detected as statistically significant (and confidence intervals will exclude zero). However, since many correlations are likely to be small, correlation effect sizes of this magnitude may just reflect ambient correlational noise.

Given the problem of unspecified effect size magnitudes in hypotheses, there is a chance that many statistically significant results may arise due to the ambient correlational noise rather than due to the hypothesized associations. While sample size increase is usually viewed as an increase in the quality of the study, it is important to keep in mind that it should be accompanied by corresponding changes to the design of the study and optimally by a principled argument about the expected effect size and most recently, by preregistration. Otherwise, increased sample size may just lead to an increased number of false positive results [ 30 ].

5.  Limitations

The automatic extraction algorithm used for Sample 2 could extract only correlations from the text but not from tables. The manually extracted data allowed us to determine that correlations reported in text were much less common than correlations reported in tables. However, we could not determine the ratio of correlations reported in tables to those reported in the text in Sample 2. Further, for analyses involving degrees of freedom and p -values in Sample 2, only a subset of the data could be used because these values were not reported with most r values.

While the validation procedure for Sample 2 has offered satisfactory results, it focused on checking papers from which there was at least one correlation successfully extracted. Therefore, there is a possibility that some papers containing correlations may not have been included in the sample if they only presented correlations in tables. However, when we compared the density distribution of correlation values from studies which were part of both samples, we found excellent correspondence.

Since we have collected all r values within the Results section (Sample 1) or within the text of the paper (Sample 2), some of these r values probably targeted the same or similar questions and are calculated on the same or similar sample, meaning that many or all of the r values collected within one paper are likely to be inherently correlated. Additionally, the researcher biases within the methods and analysis are likely to be the same for r values collected from the same papers. Given that we have shown that larger studies report smaller r values, this could skew the overall distribution of r values if either smaller studies or larger studies consistently reported larger number of r values. However, there was only weak correlation between the overall study sample size and number of r values reported ( r = 0.13).

Finally, for the purpose of this study the two subfields presented have been defined by overall specialization of the journals considered in sampling.

6.  Conclusion

As expected, we found that effect size distributions strongly depended on sample sizes: the larger the maximum sample size the smaller the corresponding effect size distribution quartiles. This suggests that effect sizes cannot be considered to be fixed and independent from sample sizes. Rather, larger studies will measure smaller effect size distributions than smaller studies. This observation has major implications for power calculations: many small and probably underpowered studies will report larger effect size quartiles than larger, well-powered studies. If power calculations are then based on effect sizes from small studies, future studies will also be small and underpowered and will also report relatively large effect sizes. Our observation also suggests that without considering how sample size affects effect size distributions, it cannot be determined whether large effects arise due to effect size inflation in small studies or whether they can really be expected in a field. Similarly, non-significant effect sizes should also be considered in effect size distributions because they may represent effects that are too small to be detected by underpowered studies.

1 https://osf.io/

2 https://aspredicted.org/

3 https://osf.io/

Data accessibility

Authors' contributions.

J.W.: conceptualization, formal analysis, methodology, writing—original draft; D.S.: conceptualization, methodology, software, supervision, writing—review and editing; J.P.A.I.: conceptualization, supervision, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

The Meta-Research Innovation Center at Stanford (METRICS) is supported by a grant from the Laura and John Arnold Foundation. The work of J.P.A. Ioannidis is supported by an unrestricted grant from Sue and Bob O'Donnell.

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

• View all journals
• My Account Login
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Open access
• Published: 13 May 2024

Long-term weight loss effects of semaglutide in obesity without diabetes in the SELECT trial

• Donna H. Ryan 1 ,
• Ildiko Lingvay   ORCID: orcid.org/0000-0001-7006-7401 2 ,
• John Deanfield 3 ,
• Steven E. Kahn 4 ,
• Eric Barros   ORCID: orcid.org/0000-0001-6613-4181 5 ,
• Bartolome Burguera 6 ,
• Helen M. Colhoun   ORCID: orcid.org/0000-0002-8345-3288 7 ,
• Cintia Cercato   ORCID: orcid.org/0000-0002-6181-4951 8 ,
• Dror Dicker 9 ,
• Deborah B. Horn 10 ,
• G. Kees Hovingh 5 ,
• Ole Kleist Jeppesen 5 ,
• Alexander Kokkinos 11 ,
• A. Michael Lincoff   ORCID: orcid.org/0000-0001-8175-2121 12 ,
• Sebastian M. Meyhöfer 13 ,
• Tugce Kalayci Oral 5 ,
• Jorge Plutzky   ORCID: orcid.org/0000-0002-7194-9876 14 ,
• André P. van Beek   ORCID: orcid.org/0000-0002-0335-8177 15 ,
• John P. H. Wilding   ORCID: orcid.org/0000-0003-2839-8404 16 &
• Robert F. Kushner 17  

Nature Medicine ( 2024 ) Cite this article

13k Accesses

2260 Altmetric

Metrics details

• Health care
• Medical research

In the SELECT cardiovascular outcomes trial, semaglutide showed a 20% reduction in major adverse cardiovascular events in 17,604 adults with preexisting cardiovascular disease, overweight or obesity, without diabetes. Here in this prespecified analysis, we examined effects of semaglutide on weight and anthropometric outcomes, safety and tolerability by baseline body mass index (BMI). In patients treated with semaglutide, weight loss continued over 65 weeks and was sustained for up to 4 years. At 208 weeks, semaglutide was associated with mean reduction in weight (−10.2%), waist circumference (−7.7 cm) and waist-to-height ratio (−6.9%) versus placebo (−1.5%, −1.3 cm and −1.0%, respectively; P  < 0.0001 for all comparisons versus placebo). Clinically meaningful weight loss occurred in both sexes and all races, body sizes and regions. Semaglutide was associated with fewer serious adverse events. For each BMI category (<30, 30 to <35, 35 to <40 and ≥40 kg m − 2 ) there were lower rates (events per 100 years of observation) of serious adverse events with semaglutide (43.23, 43.54, 51.07 and 47.06 for semaglutide and 50.48, 49.66, 52.73 and 60.85 for placebo). Semaglutide was associated with increased rates of trial product discontinuation. Discontinuations increased as BMI class decreased. In SELECT, at 208 weeks, semaglutide produced clinically significant weight loss and improvements in anthropometric measurements versus placebo. Weight loss was sustained over 4 years. ClinicalTrials.gov identifier: NCT03574597 .

Similar content being viewed by others

Effects of a personalized nutrition program on cardiometabolic health: a randomized controlled trial

Two-year effects of semaglutide in adults with overweight or obesity: the STEP 5 trial

What is the pipeline for future medications for obesity?

The worldwide obesity prevalence, defined by body mass index (BMI) ≥30 kg m − 2 , has nearly tripled since 1975 (ref. 1 ). BMI is a good surveillance measure for population changes over time, given its strong correlation with body fat amount on a population level, but it may not accurately indicate the amount or location of body fat at the individual level 2 . In fact, the World Health Organization defines clinical obesity as ‘abnormal or excessive fat accumulation that may impair health’ 1 . Excess abnormal body fat, especially visceral adiposity and ectopic fat, is a driver of cardiovascular (CV) disease (CVD) 3 , 4 , 5 , and contributes to the global chronic disease burden of diabetes, chronic kidney disease, cancer and other chronic conditions 6 , 7 .

Remediating the adverse health effects of excess abnormal body fat through weight loss is a priority in addressing the global chronic disease burden. Improvements in CV risk factors, glycemia and quality-of-life measures including personal well-being and physical functioning generally begin with modest weight loss of 5%, whereas greater weight loss is associated with more improvement in these measures 8 , 9 , 10 . Producing and sustaining durable and clinically significant weight loss with lifestyle intervention alone has been challenging 11 . However, weight-management medications that modify appetite can make attaining and sustaining clinically meaningful weight loss of ≥10% more likely 12 . Recently, weight-management medications, particularly those comprising glucagon-like peptide-1 receptor agonists, that help people achieve greater and more sustainable weight loss have been developed 13 . Once-weekly subcutaneous semaglutide 2.4 mg, a glucagon-like peptide-1 receptor agonist, is approved for chronic weight management 14 , 15 , 16 and at doses of up to 2.0 mg is approved for type 2 diabetes treatment 17 , 18 , 19 . In patients with type 2 diabetes and high CV risk, semaglutide at doses of 0.5 mg and 1.0 mg has been shown to significantly lower the risk of CV events 20 . The SELECT trial (Semaglutide Effects on Heart Disease and Stroke in Patients with Overweight or Obesity) studied patients with established CVD and overweight or obesity but without diabetes. In SELECT, semaglutide was associated with a 20% reduction in major adverse CV events (hazard ratio 0.80, 95% confidence interval (CI) 0.72 to 0.90; P  < 0.001) 21 . Data derived from the SELECT trial offer the opportunity to evaluate the weight loss efficacy, in a geographically and racially diverse population, of semaglutide compared with placebo over 208 weeks when both are given in addition to standard-of-care recommendations for secondary CVD prevention (but without a focus on targeting weight loss). Furthermore, the data allow examination of changes in anthropometric measures such as BMI, waist circumference (WC) and waist-to-height ratio (WHtR) as surrogates for body fat amount and location 22 , 23 . The diverse population can also be evaluated for changes in sex- and race-specific ‘cutoff points’ for BMI and WC, which have been identified as anthropometric measures that predict cardiometabolic risk 8 , 22 , 23 .

This prespecified analysis of the SELECT trial investigated weight loss and changes in anthropometric indices in patients with established CVD and overweight or obesity without diabetes, who met inclusion and exclusion criteria, within a range of baseline categories for glycemia, renal function and body anthropometric measures.

Study population

The SELECT study enrolled 17,604 patients (72.3% male) from 41 countries between October 2018 and March 2021, with a mean (s.d.) age of 61.6 (8.9) years and BMI of 33.3 (5.0) kg m − 2 (ref. 21 ). The baseline characteristics of the population have been reported 24 . Supplementary Table 1 outlines SELECT patients according to baseline BMI categories. Of note, in the lower BMI categories (<30 kg m − 2 (overweight) and 30 to <35 kg m − 2 (class I obesity)), the proportion of Asian individuals was higher (14.5% and 7.4%, respectively) compared with the proportion of Asian individuals in the higher BMI categories (BMI 35 to <40 kg m − 2 (class II obesity; 3.8%) and ≥40 kg m − 2 (class III obesity; 2.2%), respectively). As the BMI categories increased, the proportion of women was higher: in the class III BMI category, 45.5% were female, compared with 20.8%, 25.7% and 33.0% in the overweight, class I and class II categories, respectively. Lower BMI categories were associated with a higher proportion of patients with normoglycemia and glycated hemoglobin <5.7%. Although the proportions of patients with high cholesterol and history of smoking were similar across BMI categories, the proportion of patients with high-sensitivity C-reactive protein ≥2.0 mg dl −1 increased as the BMI category increased. A high-sensitivity C-reactive protein >2.0 mg dl −1 was present in 36.4% of patients in the overweight BMI category, with a progressive increase to 43.3%, 57.3% and 72.0% for patients in the class I, II and III obesity categories, respectively.

Weight and anthropometric outcomes

Percentage weight loss.

The average percentage weight-loss trajectories with semaglutide and placebo over 4 years of observation are shown in Fig. 1a (ref. 21 ). For those in the semaglutide group, the weight-loss trajectory continued to week 65 and then was sustained for the study period through week 208 (−10.2% for the semaglutide group, −1.5% for the placebo group; treatment difference −8.7%; 95% CI −9.42 to −7.88; P  < 0.0001). To estimate the treatment effect while on medication, we performed a first on-treatment analysis (observation period until the first time being off treatment for >35 days). At week 208, mean weight loss in the semaglutide group analyzed as first on-treatment was −11.7% compared with −1.5% for the placebo group (Fig. 1b ; treatment difference −10.2%; 95% CI −11.0 to −9.42; P  < 0.0001).

a , b , Observed data from the in-trial period ( a ) and first on-treatment ( b ). The symbols are the observed means, and error bars are ±s.e.m. Numbers shown below each panel represent the number of patients contributing to the means. Analysis of covariance with treatment and baseline values was used to estimate the treatment difference. Exact P values are 1.323762 × 10 −94 and 9.80035 × 10 −100 for a and b , respectively. P values are two-sided and are not adjusted for multiplicity. ETD, estimated treatment difference; sema, semaglutide.

Categorical weight loss and individual body weight change

Among in-trial (intention-to-treat principle) patients at week 104, weight loss of ≥5%, ≥10%, ≥15%, ≥20% and ≥25% was achieved by 67.8%, 44.2%, 22.9%, 11.0% and 4.9%, respectively, of those treated with semaglutide compared with 21.3%, 6.9%, 1.7%, 0.6% and 0.1% of those receiving placebo (Fig. 2a ). Individual weight changes at 104 weeks for the in-trial populations for semaglutide and placebo are depicted in Fig. 2b and Fig. 2c , respectively. These waterfall plots show the variation in weight-loss response that occurs with semaglutide and placebo and show that weight loss is more prominent with semaglutide than placebo.

a , Categorical weight loss from baseline at week 104 for semaglutide and placebo. Data from the in-trial period. Bars depict the proportion (%) of patients receiving semaglutide or placebo who achieved ≥5%, ≥10%, ≥15%, ≥20% and ≥25% weight loss. b , c , Percentage change in body weight for individual patients from baseline to week 104 for semaglutide ( b ) and placebo ( c ). Each patient’s percentage change in body weight is plotted as a single bar.

Change in WC

WC change from baseline to 104 weeks has been reported previously in the primary outcome paper 21 . The trajectory of WC change mirrored that of the change in body weight. At week 208, average reduction in WC was −7.7 cm with semaglutide versus −1.3 cm with placebo, with a treatment difference of −6.4 cm (95% CI −7.18 to −5.61; P  < 0.0001) 21 .

WC cutoff points

We analyzed achievement of sex- and race-specific cutoff points for WC by BMI <35 kg m − 2 or ≥35 kg m − 2 , because for BMI >35 kg m − 2 , WC is more difficult technically and, thus, less accurate as a risk predictor 4 , 25 , 26 . Within the SELECT population with BMI <35 kg m − 2 at baseline, 15.0% and 14.3% of the semaglutide and placebo groups, respectively, were below the sex- and race-specific WC cutoff points. At week 104, 41.2% fell below the sex- and race-specific cutoff points for the semaglutide group, compared with only 18.0% for the placebo group (Fig. 3 ).

WC cutoff points; Asian women <80 cm, non-Asian women <88 cm, Asian men <88 cm, non-Asian men <102 cm.

Waist-to-height ratio

At baseline, mean WHtR was 0.66 for the study population. The lowest tertile of the SELECT population at baseline had a mean WHtR <0.62, which is higher than the cutoff point of 0.5 used to indicate increased cardiometabolic risk 27 , suggesting that the trial population had high WCs. At week 208, in the group randomized to semaglutide, there was a relative reduction of 6.9% in WHtR compared with 1.0% in placebo (treatment difference −5.87% points; 95% CI −6.56 to −5.17; P  < 0.0001).

BMI category change

At week 104, 52.4% of patients treated with semaglutide achieved improvement in BMI category compared with 15.7% of those receiving placebo. Proportions of patients in the BMI categories at baseline and week 104 are shown in Fig. 4 , which depicts in-trial patients receiving semaglutide and placebo. The BMI category change reflects the superior weight loss with semaglutide, which resulted in fewer patients being in the higher BMI categories after 104 weeks. In the semaglutide group, 12.0% of patients achieved a BMI <25 kg m − 2 , which is considered the healthy BMI category, compared with 1.2% for placebo; per study inclusion criteria, no patients were in this category at baseline. The proportion of patients with obesity (BMI ≥30 kg m − 2 ) fell from 71.0% to 43.3% in the semaglutide group versus 71.9% to 67.9% in the placebo group.

In the semaglutide group, 12.0% of patients achieved normal weight status at week 104 (from 0% at baseline), compared with 1.2% (from 0% at baseline) for placebo. BMI classes: healthy (BMI <25 kg m − 2 ), overweight (25 to <30 kg m − 2 ), class I obesity (30 to <35 kg m − 2 ), class II obesity (35 to <40 kg m − 2 ) and class III obesity (BMI ≥40 kg m − 2 ).

Weight and anthropometric outcomes by subgroups

The forest plot illustrated in Fig. 5 displays mean body weight percentage change from baseline to week 104 for semaglutide relative to placebo in prespecified subgroups. Similar relationships are depicted for WC changes in prespecified subgroups shown in Extended Data Fig. 1 . The effect of semaglutide (versus placebo) on mean percentage body weight loss as well as reduction in WC was found to be heterogeneous across several population subgroups. Women had a greater difference in mean weight loss with semaglutide versus placebo (−11.1% (95% CI −11.56 to −10.66) versus −7.5% in men (95% CI −7.78 to −7.23); P  < 0.0001). There was a linear relationship between age category and degree of mean weight loss, with younger age being associated with progressively greater mean weight loss, but the actual mean difference by age group is small. Similarly, BMI category had small, although statistically significant, associations. Those with WHtR less than the median experienced slightly lower mean body weight change than those above the median, with estimated treatment differences −8.04% (95% CI −8.37 to −7.70) and −8.99% (95% CI −9.33 to −8.65), respectively ( P  < 0.0001). Patients from Asia and of Asian race experienced slightly lower mean weight loss (estimated treatment difference with semaglutide for Asian race −7.27% (95% CI −8.09 to −6.46; P  = 0.0147) and for Asia −7.30 (95% CI −7.97 to −6.62; P  = 0.0016)). There was no difference in weight loss with semaglutide associated with ethnicity (estimated treatment difference for Hispanic −8.53% (95% CI −9.28 to −7.76) or non-Hispanic −8.52% (95% CI −8.77 to 8.26); P  = 0.9769), glycemic status (estimated treatment difference for prediabetes −8.53% (95% CI −8.83 to −8.24) or normoglycemia −8.48% (95% CI −8.88 to −8.07; P  = 0.8188) or renal function (estimated treatment difference for estimated glomerular filtration rate (eGFR) <60 or ≥60 ml min −1  1.73 m − 2 being −8.50% (95% CI −9.23 to −7.76) and −8.52% (95% CI −8.77 to −8.26), respectively ( P  = 0.9519)).

Data from the in-trial period. N  = 17,604. P values represent test of no interaction effect. P values are two-sided and are not adjusted for multiplicity. The dots show estimated treatment differences, and the error bars show 95% CIs. Details of the statistical models are available in Methods . ETD, estimated treatment difference; HbA1c, glycated hemoglobin; MI, myocardial infarction; PAD, peripheral artery disease; sema, semaglutide.

Safety and tolerability according to baseline BMI category

We reported in the primary outcome of the SELECT trial that adverse events (AEs) leading to permanent discontinuation of the trial product occurred in 1,461 patients (16.6%) in the semaglutide group and 718 patients (8.2%) in the placebo group ( P  < 0.001) 21 . For this analysis, we evaluated the cumulative incidence of AEs leading to trial product discontinuation by treatment assignment and by BMI category (Fig. 6 ). For this analysis, with death modeled as a competing risk, we tracked the proportion of in-trial patients for whom drug was withdrawn or interrupted for the first time (Fig. 6 , left) or cumulative discontinuations (Fig. 6 , right). Both panels of Fig. 6 depict a graded increase in the proportion discontinuing semaglutide, but not placebo. For lower BMI classes, discontinuation rates are higher in the semaglutide group but not the placebo group.

Data are in-trial from the full analysis set. sema, semaglutide.

We reported in the primary SELECT analysis that serious adverse events (SAEs) were reported by 2,941 patients (33.4%) in the semaglutide arm and by 3,204 patients (36.4%) in the placebo arm ( P  < 0.001) 21 . For this study, we analyzed SAE rates by person-years of treatment exposure for BMI classes (<30 kg m − 2 , 30 to <35 kg m − 2 , 35 to <40 kg m − 2 , and ≥40 kg m − 2 ) and provide these data in Supplementary Table 2 . We also provide an analysis of the most common categories of SAEs. Semaglutide was associated with lower SAEs, primarily driven by CV event and infections. Within each obesity class (<30 kg m − 2 , 30 to <35 kg m − 2 , 35 to <40 kg m − 2 , and ≥40 kg m − 2 ), there were fewer SAEs in the group receiving semaglutide compared with placebo. Rates (events per 100 years of observation) of SAEs were 43.23, 43.54, 51.07 and 47.06 for semaglutide and 50.48, 49.66, 52.73 and 60.85 for placebo, with no evidence of heterogeneity. There was no detectable difference in hepatobiliary or gastrointestinal SAEs comparing semaglutide with placebo in any of the four BMI classes we evaluated.

The analyses of weight effects of the SELECT study presented here reveal that patients assigned to once-weekly subcutaneous semaglutide 2.4 mg lost significantly more weight than those receiving placebo. The weight-loss trajectory with semaglutide occurred over 65 weeks and was sustained up to 4 years. Likewise, there were similar improvements in the semaglutide group for anthropometrics (WC and WHtR). The weight loss was associated with a greater proportion of patients receiving semaglutide achieving improvement in BMI category, healthy BMI (<25 kg m − 2 ) and falling below the WC cutoff point above which increased cardiometabolic risk for the sex and race is greater 22 , 23 . Furthermore, both sexes, all races, all body sizes and those from all geographic regions were able to achieve clinically meaningful weight loss. There was no evidence of increased SAEs based on BMI categories, although lower BMI category was associated with increased rates of trial product discontinuation, probably reflecting exposure to a higher level of drug in lower BMI categories. These data, representing the longest clinical trial of the effects of semaglutide versus placebo on weight, establish the safety and durability of semaglutide effects on weight loss and maintenance in a geographically and racially diverse population of adult men and women with overweight and obesity but not diabetes. The implications of weight loss of this degree in such a diverse population suggests that it may be possible to impact the public health burden of the multiple morbidities associated with obesity. Although our trial focused on CV events, many chronic diseases would benefit from effective weight management 28 .

There were variations in the weight-loss response. Individual changes in body weight with semaglutide and placebo were striking; still, 67.8% achieved 5% or more weight loss and 44.2% achieved 10% weight loss with semaglutide at 2 years, compared with 21.3% and 6.9%, respectively, for those receiving placebo. Our first on-treatment analysis demonstrated that those on-drug lost more weight than those in-trial, confirming the effect of drug exposure. With semaglutide, lower BMI was associated with less percentage weight loss, and women lost more weight on average than men (−11.1% versus −7.5% treatment difference from placebo); however, in all cases, clinically meaningful mean weight loss was achieved. Although Asian patients lost less weight on average than patients of other races (−7.3% more than placebo), Asian patients were more likely to be in the lowest BMI category (<30 kg m − 2 ), which is known to be associated with less weight loss, as discussed below. Clinically meaningful weight loss was evident in the semaglutide group within a broad range of baseline categories for glycemia and body anthropometrics. Interestingly, at 2 years, a significant proportion of the semaglutide-treated group fell below the sex- and race-specific WC cutoff points, especially in those with BMI <35 kg m − 2 , and a notable proportion (12.0%) fell below the BMI cutoff point of 25 kg m − 2 , which is deemed a healthy BMI in those without unintentional weight loss. As more robust weight loss is possible with newer medications, achieving and maintaining these cutoff point targets may become important benchmarks for tracking responses.

The overall safety profile did not reveal any new signals from prior studies, and there were no BMI category-related associations with AE reporting. The analysis did reveal that tolerability may differ among specific BMI classes, since more discontinuations occurred with semaglutide among lower BMI classes. Potential contributors may include a possibility of higher drug exposure in lower BMI classes, although other explanations, including differences in motivation and cultural mores regarding body size, cannot be excluded.

Is the weight loss in SELECT less than expected based on prior studies with the drug? In STEP 1, a large phase 3 study of once-weekly subcutaneous semaglutide 2.4 mg in individuals without diabetes but with BMI >30 kg m − 2 or 27 kg m − 2 with at least one obesity-related comorbidity, the mean weight loss was −14.9% at week 68, compared with −2.4% with placebo 14 . Several reasons may explain the observation that the mean treatment difference was −12.5% in STEP 1 and −8.7% in SELECT. First, SELECT was designed as a CV outcomes trial and not a weight-loss trial, and weight loss was only a supportive secondary endpoint in the trial design. Patients in STEP 1 were desirous of weight loss as a reason for study participation and received structured lifestyle intervention (which included a −500 kcal per day diet with 150 min per week of physical activity). In the SELECT trial, patients did not enroll for the specific purpose of weight loss and received standard of care covering management of CV risk factors, including medical treatment and healthy lifestyle counseling, but without a specific focus on weight loss. Second, the respective study populations were quite different, with STEP 1 including a younger, healthier population with more women (73.1% of the semaglutide arm in STEP 1 versus 27.7% in SELECT) and higher mean BMI (37.8 kg m − 2 versus 33.3 kg m − 2 , respectively) 14 , 21 . Third, major differences existed between the respective trial protocols. Patients in the semaglutide treatment arm of STEP 1 were more likely to be exposed to the medication at the full dose of 2.4 mg than those in SELECT. In SELECT, investigators were allowed to slow, decrease or pause treatment. By 104 weeks, approximately 77% of SELECT patients on dose were receiving the target semaglutide 2.4 mg weekly dose, which is lower than the corresponding proportion of patients in STEP 1 (89.6% were receiving the target dose at week 68) 14 , 21 . Indeed, in our first on-treatment analysis at week 208, weight loss was greater (−11.7% for semaglutide) compared with the in-trial analysis (−10.2% for semaglutide). Taken together, all these issues make less weight loss an expected finding in SELECT, compared with STEP 1.

The SELECT study has some limitations. First, SELECT was not a primary prevention trial, and the data should not be extrapolated to all individuals with overweight and obesity to prevent major adverse CV events. Although the data set is rich in numbers and diversity, it does not have the numbers of individuals in racial subgroups that may have revealed potential differential effects. SELECT also did not include individuals who have excess abnormal body fat but a BMI <27 kg m − 2 . Not all individuals with increased CV risk have BMI ≥27 kg m − 2 . Thus, the study did not include Asian patients who qualify for treatment with obesity medications at lower BMI and WC cutoff points according to guidelines in their countries 29 . We observed that Asian patients were less likely to be in the higher BMI categories of SELECT and that the population of those with BMI <30 kg m − 2 had a higher percentage of Asian race. Asian individuals would probably benefit from weight loss and medication approaches undertaken at lower BMI levels in the secondary prevention of CVD. Future studies should evaluate CV risk reduction in Asian individuals with high CV risk and BMI <27 kg m − 2 . Another limitation is the lack of information on body composition, beyond the anthropometric measures we used. It would be meaningful to have quantitation of fat mass, lean mass and muscle mass, especially given the wide range of body size in the SELECT population.

An interesting observation from this SELECT weight loss data is that when BMI is ≤30 kg m − 2 , weight loss on a percentage basis is less than that observed across higher classes of BMI severity. Furthermore, as BMI exceeds 30 kg m − 2 , weight loss amounts are more similar for class I, II and III obesity. This was also observed in Look AHEAD, a lifestyle intervention study for weight loss 30 . The proportion (percentage) of weight loss seems to be less, on average, in the BMI <30 kg m − 2 category relative to higher BMI categories, despite their receiving of the same treatment and even potentially higher exposure to the drug for weight loss 30 . Weight loss cannot continue indefinitely. There is a plateau of weight that occurs after weight loss with all treatments for weight management. This plateau has been termed the ‘set point’ or ‘settling point’, a body weight that is in harmony with the genetic and environmental determinants of body weight and adiposity 31 . Perhaps persons with BMI <30 kg m − 2 are closer to their settling point and have less weight to lose to reach it. Furthermore, the cardiometabolic benefits of weight loss are driven by reduction in the abnormal ectopic and visceral depots of fat, not by reduction of subcutaneous fat stores in the hips and thighs. The phenotype of cardiometabolic disease but lower BMI (<30 kg m − 2 ) may be one where reduction of excess abnormal and dysfunctional body fat does not require as much body mass reduction to achieve health improvement. We suspect this may be the case and suggest further studies to explore this aspect of weight-loss physiology.

In conclusion, this analysis of the SELECT study supports the broad use of once-weekly subcutaneous semaglutide 2.4 mg as an aid to CV event reduction in individuals with overweight or obesity without diabetes but with preexisting CVD. Semaglutide 2.4 mg safely and effectively produced clinically significant weight loss in all subgroups based on age, sex, race, glycemia, renal function and anthropometric categories. Furthermore, the weight loss was sustained over 4 years during the trial.

Trial design and participants

The current work complies with all relevant ethical regulations and reports a prespecified analysis of the randomized, double-blind, placebo-controlled SELECT trial ( NCT03574597 ), details of which have been reported in papers describing study design and rationale 32 , baseline characteristics 24 and the primary outcome 21 . SELECT evaluated once-weekly subcutaneous semaglutide 2.4 mg versus placebo to reduce the risk of major adverse cardiac events (a composite endpoint comprising CV death, nonfatal myocardial infarction or nonfatal stroke) in individuals with established CVD and overweight or obesity, without diabetes. The protocol for SELECT was approved by national and institutional regulatory and ethical authorities in each participating country. All patients provided written informed consent before beginning any trial-specific activity. Eligible patients were aged ≥45 years, with a BMI of ≥27 kg m − 2 and established CVD defined as at least one of the following: prior myocardial infarction, prior ischemic or hemorrhagic stroke, or symptomatic peripheral artery disease. Additional inclusion and exclusion criteria can be found elsewhere 32 .

Human participants research

The trial protocol was designed by the trial sponsor, Novo Nordisk, and the academic Steering Committee. A global expert panel of physician leaders in participating countries advised on regional operational issues. National and institutional regulatory and ethical authorities approved the protocol, and all patients provided written informed consent.

Study intervention and patient management

Patients were randomly assigned in a double-blind manner and 1:1 ratio to receive once-weekly subcutaneous semaglutide 2.4 mg or placebo. The starting dose was 0.24 mg once weekly, with dose increases every 4 weeks (to doses of 0.5, 1.0, 1.7 and 2.4 mg per week) until the target dose of 2.4 mg was reached after 16 weeks. Patients who were unable to tolerate dose escalation due to AEs could be managed by extension of dose-escalation intervals, treatment pauses or maintenance at doses below the 2.4 mg per week target dose. Investigators were allowed to reduce the dose of study product if tolerability issues arose. Investigators were provided with guidelines for, and encouraged to follow, evidence-based recommendations for medical treatment and lifestyle counseling to optimize management of underlying CVD as part of the standard of care. The lifestyle counseling was not targeted at weight loss. Additional intervention descriptions are available 32 .

Sex, race, body weight, height and WC measurements

Sex and race were self-reported. Body weight was measured without shoes and only wearing light clothing; it was measured on a digital scale and recorded in kilograms or pounds (one decimal with a precision of 0.1 kg or lb), with preference for using the same scale throughout the trial. The scale was calibrated yearly as a minimum unless the manufacturer certified that calibration of the weight scales was valid for the lifetime of the scale. Height was measured without shoes in centimeters or inches (one decimal with a precision of 0.1 cm or inches). At screening, BMI was calculated by the electronic case report form. WC was defined as the abdominal circumference located midway between the lower rib margin and the iliac crest. Measures were obtained in a standing position with a nonstretchable measuring tape and to the nearest centimeter or inch. The patient was asked to breathe normally. The tape touched the skin but did not compress soft tissue, and twists in the tape were avoided.

The following endpoints relevant to this paper were assessed at randomization (week 0) to years 2, 3 and 4: change in body weight (%); proportion achieving weight loss ≥5%, ≥10%, ≥15% and ≥20%; change in WC (cm); and percentage change in WHtR (cm cm −1 ). Improvement in BMI category (defined as being in a lower BMI class) was assessed at week 104 compared with baseline according to BMI classes: healthy (BMI <25 kg m − 2 ), overweight (25 to <30 kg m − 2 ), class I obesity (30 to <35 kg m − 2 ), class II obesity (35 to <40 kg m − 2 ) and class III obesity (≥40 kg m − 2 ). The proportions of individuals with BMI <35 or ≥35 kg m − 2 who achieved sex- and race-specific cutoff points for WC (indicating increased metabolic risk) were evaluated at week 104. The WC cutoff points were as follows: Asian women <80 cm, non-Asian women <88 cm, Asian men <88 cm and non-Asian men <102 cm.

Overall, 97.1% of the semaglutide group and 96.8% of the placebo group completed the trial. During the study, 30.6% of those assigned to semaglutide did not complete drug treatment, compared with 27.0% for placebo.

Statistical analysis

The statistical analyses for the in-trial period were based on the intention-to-treat principle and included all randomized patients irrespective of adherence to semaglutide or placebo or changes to background medications. Continuous endpoints were analyzed using an analysis of covariance model with treatment as a fixed factor and baseline value of the endpoint as a covariate. Missing data at the landmark visit, for example, week 104, were imputed using a multiple imputation model and done separately for each treatment arm and included baseline value as a covariate and fit to patients having an observed data point (irrespective of adherence to randomized treatment) at week 104. The fit model is used to impute values for all patients with missing data at week 104 to create 500 complete data sets. Rubin’s rules were used to combine the results. Estimated means are provided with s.e.m., and estimated treatment differences are provided with 95% CI. Binary endpoints were analyzed using logistic regression with treatment and baseline value as a covariate, where missing data were imputed by first using multiple imputation as described above and then categorizing the imputed data according to the endpoint, for example, body weight percentage change at week 104 of <0%. Subgroup analyses for continuous and binary endpoints also included the subgroup and interaction between treatment and subgroup as fixed factors. Because some patients in both arms continued to be followed but were off treatment, we also analyzed weight loss by first on-treatment group (observation period until first time being off treatment for >35 days) to assess a more realistic picture of weight loss in those adhering to treatment. CIs were not adjusted for multiplicity and should therefore not be used to infer definitive treatment effects. All statistical analyses were performed with SAS software, version 9.4 TS1M5 (SAS Institute).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

Data will be shared with bona fide researchers who submit a research proposal approved by the independent review board. Individual patient data will be shared in data sets in a deidentified and anonymized format. Information about data access request proposals can be found at https://www.novonordisk-trials.com/ .

Obesity and overweight. World Health Organization https://www.who.int/news-room/fact-sheets/detail/obesity-and-overweight (2021).

Cornier, M. A. et al. Assessing adiposity: a scientific statement from the American Heart Association. Circulation 124 , 1996–2019 (2011).

Article   PubMed   Google Scholar  

Afshin, A. et al. Health effects of overweight and obesity in 195 countries over 25 years. N. Engl. J. Med. 377 , 13–27 (2017).

Jensen, M. D. et al. 2013 AHA/ACC/TOS guideline for the management of overweight and obesity in adults: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines and The Obesity Society. J. Am. Coll. Cardiol. 63 , 2985–3023 (2014).

Poirier, P. et al. Obesity and cardiovascular disease: pathophysiology, evaluation, and effect of weight loss: an update of the 1997 American Heart Association Scientific Statement on Obesity and Heart Disease from the Obesity Committee of the Council on Nutrition, Physical Activity, and Metabolism. Circulation 113 , 898–918 (2006).

Dai, H. et al. The global burden of disease attributable to high body mass index in 195 countries and territories, 1990–2017: an analysis of the Global Burden of Disease Study. PLoS Med. 17 , e1003198 (2020).

Article   PubMed   PubMed Central   Google Scholar  

Ndumele, C. E. et al. Cardiovascular–kidney–metabolic health: a presidential advisory from the American Heart Association. Circulation 148 , 1606–1635 (2023).

Garvey, W. T. et al. American Association of Clinical Endocrinologists and American College of Endocrinology comprehensive clinical practice guidelines for medical care of patients with obesity. Endocr. Pr. 22 , 1–203 (2016).

Article   Google Scholar  

Ryan, D. H. & Yockey, S. R. Weight loss and improvement in comorbidity: differences at 5%, 10%, 15%, and over. Curr. Obes. Rep. 6 , 187–194 (2017).

Wing, R. R. et al. Benefits of modest weight loss in improving cardiovascular risk factors in overweight and obese individuals with type 2 diabetes. Diabetes Care 34 , 1481–1486 (2011).

Article   CAS   PubMed   PubMed Central   Google Scholar  

Wadden, T. A., Tronieri, J. S. & Butryn, M. L. Lifestyle modification approaches for the treatment of obesity in adults. Am. Psychol. 75 , 235–251 (2020).

Tchang, B. G. et al. Pharmacologic treatment of overweight and obesity in adults. in (eds. Feingold, K. R. et al.) Endotext https://www.ncbi.nlm.nih.gov/books/NBK279038/ (MDText.com, 2000).

Müller, T. D., Blüher, M., Tschöp, M. H. & DiMarchi, R. D. Anti-obesity drug discovery: advances and challenges. Nat. Rev. Drug Discov. 21 , 201–223 (2022).

Wilding, J. P. H. et al. Once-weekly semaglutide in adults with overweight or obesity. N. Engl. J. Med. 384 , 989–1002 (2021).

Article   CAS   PubMed   Google Scholar  

Wegovy (semaglutide) summary of product characteristics. European Medicines Agency https://www.ema.europa.eu/en/documents/product-information/wegovy-epar-product-information_en.pdf (2023).

WEGOVY (semaglutide) prescribing information. Food and Drug Administration https://www.accessdata.fda.gov/drugsatfda_docs/label/2023/215256s007lbl.pdf (2023).

Sorli, C. et al. Efficacy and safety of once-weekly semaglutide monotherapy versus placebo in patients with type 2 diabetes (SUSTAIN 1): a double-blind, randomised, placebo-controlled, parallel-group, multinational, multicentre phase 3a trial. Lancet Diabetes Endocrinol. 5 , 251–260 (2017).

Ozempic (semaglutide) summary of product characteristics. European Medicines Agency https://www.ema.europa.eu/en/documents/product-information/ozempic-epar-product-information_en.pdf (2023).

OZEMPIC (semaglutide) prescribing information. Food and Drug Administration https://www.accessdata.fda.gov/drugsatfda_docs/label/2017/209637lbl.pdf (2017).

Marso, S. P. et al. Semaglutide and cardiovascular outcomes in patients with type 2 diabetes. N. Engl. J. Med. 375 , 1834–1844 (2016).

Lincoff, A. M. et al. Semaglutide and cardiovascular outcomes in obesity without diabetes. N. Engl. J. Med. 389 , 2221–2232 (2023).

Ross, R. et al. Waist circumference as a vital sign in clinical practice: a consensus statement from the IAS and ICCR Working Group on Visceral Obesity. Nat. Rev. Endocrinol. 16 , 177–189 (2020).

Snijder, M. B., van Dam, R. M., Visser, M. & Seidell, J. C. What aspects of body fat are particularly hazardous and how do we measure them? Int. J. Epidemiol. 35 , 83–92 (2006).

Lingvay, I. et al. Semaglutide for cardiovascular event reduction in people with overweight or obesity: SELECT study baseline characteristics. Obesity 31 , 111–122 (2023).

Basset, J. The Asia-Pacific perspective: redefining obesity and its treatment. International Diabetes Institute, World Health Organization Regional Office for the Western Pacific, International Association for the Study of Obesity & International Obesity Task Force https://www.vepachedu.org/TSJ/BMI-Guidelines.pdf (2000).

Hu, F. in Obesity Epidemiology (ed. Hu, F.) 53–83 (Oxford University Press, 2008).

Browning, L. M., Hsieh, S. D. & Ashwell, M. A systematic review of waist-to-height ratio as a screening tool for the prediction of cardiovascular disease and diabetes: 0·5 could be a suitable global boundary value. Nutr. Res. Rev. 23 , 247–269 (2010).

Sattar, N. et al. Treating chronic diseases without tackling excess adiposity promotes multimorbidity. Lancet Diabetes Endocrinol. 11 , 58–62 (2023).

Obesity classification. World Obesity https://www.worldobesity.org/about/about-obesity/obesity-classification (2022).

Unick, J. L. et al. Effectiveness of lifestyle interventions for individuals with severe obesity and type 2 diabetes: results from the Look AHEAD trial. Diabetes Care 34 , 2152–2157 (2011).

Speakman, J. R. et al. Set points, settling points and some alternative models: theoretical options to understand how genes and environments combine to regulate body adiposity. Dis. Model. Mech. 4 , 733–745 (2011).

Ryan, D. H. et al. Semaglutide effects on cardiovascular outcomes in people with overweight or obesity (SELECT) rationale and design. Am. Heart J. 229 , 61–69 (2020).

Download references

Acknowledgements

Editorial support was provided by Richard Ogilvy-Stewart of Apollo, OPEN Health Communications, and funded by Novo Nordisk A/S, in accordance with Good Publication Practice guidelines ( www.ismpp.org/gpp-2022 ).

Author information

Authors and affiliations.

Pennington Biomedical Research Center, Baton Rouge, LA, USA

Donna H. Ryan

Department of Internal Medicine/Endocrinology and Peter O’ Donnell Jr. School of Public Health, University of Texas Southwestern Medical Center, Dallas, TX, USA

Ildiko Lingvay

Institute of Cardiovascular Science, University College London, London, UK

John Deanfield

VA Puget Sound Health Care System and University of Washington, Seattle, WA, USA

Steven E. Kahn

Novo Nordisk A/S, Søborg, Denmark

Eric Barros, G. Kees Hovingh, Ole Kleist Jeppesen & Tugce Kalayci Oral

Endocrinology and Metabolism Institute, Cleveland Clinic, Cleveland, OH, USA

Bartolome Burguera

Institute of Genetics and Cancer, University of Edinburgh, Edinburgh, UK

Helen M. Colhoun

Obesity Unit, Department of Endocrinology, Hospital das Clínicas, University of São Paulo, São Paulo, Brazil

Cintia Cercato

Internal Medicine Department D, Hasharon Hospital-Rabin Medical Center, Faculty of Medicine, Tel Aviv University, Tel Aviv, Israel

Dror Dicker

Center for Obesity Medicine and Metabolic Performance, Department of Surgery, University of Texas McGovern Medical School, Houston, TX, USA

Deborah B. Horn

First Department of Propaedeutic Internal Medicine, School of Medicine, National and Kapodistrian University of Athens, Athens, Greece

Alexander Kokkinos

Department of Cardiovascular Medicine, Cleveland Clinic, and Cleveland Clinic Lerner College of Medicine of Case Western Reserve University, Cleveland, OH, USA

A. Michael Lincoff

Institute of Endocrinology & Diabetes, University of Lübeck, Lübeck, Germany

Sebastian M. Meyhöfer

Cardiovascular Division, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA, USA

Jorge Plutzky

University of Groningen, University Medical Center Groningen, Department of Endocrinology, Groningen, the Netherlands

André P. van Beek

Department of Cardiovascular and Metabolic Medicine, University of Liverpool, Liverpool, UK

John P. H. Wilding

Northwestern University Feinberg School of Medicine, Chicago, IL, USA

Robert F. Kushner

You can also search for this author in PubMed   Google Scholar

Contributions

D.H.R., I.L. and S.E.K. contributed to the study design. D.B.H., I.L., D.D., A.K., S.M.M., A.P.v.B., C.C. and J.P.H.W. were study investigators. D.B.H., I.L., D.D., A.K., S.M.M., A.P.v.B., C.C. and J.P.H.W. enrolled patients. D.H.R. was responsible for data analysis and manuscript preparation. All authors contributed to data interpretation, review, revisions and final approval of the manuscript.

Corresponding author

Correspondence to Donna H. Ryan .

Ethics declarations

Competing interests.

D.H.R. declares having received consulting honoraria from Altimmune, Amgen, Biohaven, Boehringer Ingelheim, Calibrate, Carmot Therapeutics, CinRx, Eli Lilly, Epitomee, Gila Therapeutics, IFA Celtics, Novo Nordisk, Pfizer, Rhythm, Scientific Intake, Wondr Health and Zealand Pharma; she declares she received stock options from Calibrate, Epitomee, Scientific Intake and Xeno Bioscience. I.L. declares having received research funding (paid to institution) from Novo Nordisk, Sanofi, Mylan and Boehringer Ingelheim. I.L. received advisory/consulting fees and/or other support from Altimmune, AstraZeneca, Bayer, Biomea, Boehringer Ingelheim, Carmot Therapeutics, Cytoki Pharma, Eli Lilly, Intercept, Janssen/Johnson & Johnson, Mannkind, Mediflix, Merck, Metsera, Novo Nordisk, Pharmaventures, Pfizer, Regeneron, Sanofi, Shionogi, Structure Therapeutics, Target RWE, Terns Pharmaceuticals, The Comm Group, Valeritas, WebMD and Zealand Pharma. J.D. declares having received consulting honoraria from Amgen, Boehringer Ingelheim, Merck, Pfizer, Aegerion, Novartis, Sanofi, Takeda, Novo Nordisk and Bayer, and research grants from British Heart Foundation, MRC (UK), NIHR, PHE, MSD, Pfizer, Aegerion, Colgate and Roche. S.E.K. declares having received consulting honoraria from ANI Pharmaceuticals, Boehringer Ingelheim, Eli Lilly, Merck, Novo Nordisk and Oramed, and stock options from AltPep. B.B. declares having received honoraria related to participation on this trial and has no financial conflicts related to this publication. H.M.C. declares being a stockholder and serving on an advisory panel for Bayer; receiving research grants from Chief Scientist Office, Diabetes UK, European Commission, IQVIA, Juvenile Diabetes Research Foundation and Medical Research Council; serving on an advisory board and speaker’s bureau for Novo Nordisk; and holding stock in Roche Pharmaceuticals. C.C. declares having received consulting honoraria from Novo Nordisk, Eli Lilly, Merck, Brace Pharma and Eurofarma. D.D. declares having received consulting honoraria from Novo Nordisk, Eli Lilly, Boehringer Ingelheim and AstraZeneca, and received research grants through his affiliation from Novo Nordisk, Eli Lilly, Boehringer Ingelheim and Rhythm. D.B.H. declares having received research grants through her academic affiliation from Novo Nordisk and Eli Lilly, and advisory/consulting honoraria from Novo Nordisk, Eli Lilly and Gelesis. A.K. declares having received research grants through his affiliation from Novo Nordisk and Pharmaserve Lilly, and consulting honoraria from Pharmaserve Lilly, Sanofi-Aventis, Novo Nordisk, MSD, AstraZeneca, ELPEN Pharma, Boehringer Ingelheim, Galenica Pharma, Epsilon Health and WinMedica. A.M.L. declares having received honoraria from Novo Nordisk, Eli Lilly, Akebia Therapeutics, Ardelyx, Becton Dickinson, Endologix, FibroGen, GSK, Medtronic, Neovasc, Provention Bio, ReCor, BrainStorm Cell Therapeutics, Alnylam and Intarcia for consulting activities, and research funding to his institution from AbbVie, Esperion, AstraZeneca, CSL Behring, Novartis and Eli Lilly. S.M.M. declares having received consulting honoraria from Amgen, AstraZeneca, Bayer, Boehringer Ingelheim, Daichii-Sankyo, esanum, Gilead, Ipsen, Eli Lilly, Novartis, Novo Nordisk, Sandoz and Sanofi; he declares he received research grants from AstraZeneca, Eli Lilly and Novo Nordisk. J.P. declares having received consulting honoraria from Altimmune, Amgen, Esperion, Merck, MJH Life Sciences, Novartis and Novo Nordisk; he has received a grant, paid to his institution, from Boehringer Ingelheim and holds the position of Director, Preventive Cardiology, at Brigham and Women’s Hospital. A.P.v.B. is contracted via the University of Groningen (no personal payment) to undertake consultancy for Novo Nordisk, Eli Lilly and Boehringer Ingelheim. J.P.H.W. is contracted via the University of Liverpool (no personal payment) to undertake consultancy for Altimmune, AstraZeneca, Boehringer Ingelheim, Cytoki, Eli Lilly, Napp, Novo Nordisk, Menarini, Pfizer, Rhythm Pharmaceuticals, Sanofi, Saniona, Tern Pharmaceuticals, Shionogi and Ysopia. J.P.H.W. also declares personal honoraria/lecture fees from AstraZeneca, Boehringer Ingelheim, Medscape, Napp, Menarini, Novo Nordisk and Rhythm. R.F.K. declares having received consulting honoraria from Novo Nordisk, Weight Watchers, Eli Lilly, Boehringer Ingelheim, Pfizer, Structure and Altimmune. E.B., G.K.H., O.K.J. and T.K.O. are employees of Novo Nordisk A/S.

Peer review

Peer review information.

Nature Medicine thanks Christiana Kartsonaki, Peter Rossing, Naveed Sattar and Vikas Sridhar for their contribution to the peer review of this work. Primary Handling Editor: Sonia Muliyil, in collaboration with the Nature Medicine team.

Additional information

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

Extended data

Extended data fig. 1 effect of semaglutide treatment or placebo on waist circumference from baseline to week 104 by subgroups..

Data from the in-trial period. N  = 17,604. P values represent test of no interaction effect. P values are two-sided and not adjusted for multiplicity. The dots show estimated treatment differences and the error bars show 95% confidence intervals. Details of the statistical models are available in Methods . BMI, body mass index; CI, confidence interval; CV, cardiovascular; CVD, cardiovascular disease; eGFR, estimated glomerular filtration rate; ETD, estimated treatment difference; HbA1c, glycated hemoglobin; MI, myocardial infarction; PAD, peripheral artery disease; sema, semaglutide.

Supplementary information

Reporting summary, supplementary tables 1 and 2.

Supplementary Table 1. Baseline characteristics by BMI class. Data are represented as number and percentage of patients. Renal function categories were based on the eGFR as per Chronic Kidney Disease Epidemiology Collaboration. Albuminuria categories were based on UACR. Smoking was defined as smoking at least one cigarette or equivalent daily. The category ‘Other’ for CV inclusion criteria includes patients where it is unknown if the patient fulfilled only one or several criteria and patients who were randomized in error and did not fulfill any criteria. Supplementary Table 2. SAEs according to baseline BMI category. P value: two-sided P value from Fisher’s exact test for test of no difference.

Rights and permissions

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

Reprints and permissions

About this article

Cite this article.

Ryan, D.H., Lingvay, I., Deanfield, J. et al. Long-term weight loss effects of semaglutide in obesity without diabetes in the SELECT trial. Nat Med (2024). https://doi.org/10.1038/s41591-024-02996-7

Download citation

Received : 01 March 2024

Accepted : 12 April 2024

Published : 13 May 2024

DOI : https://doi.org/10.1038/s41591-024-02996-7

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

Quick links

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

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Help | Advanced Search

Computer Science > Machine Learning

Title: exploring correlations of self-supervised tasks for graphs.

Abstract: Graph self-supervised learning has sparked a research surge in training informative representations without accessing any labeled data. However, our understanding of graph self-supervised learning remains limited, and the inherent relationships between various self-supervised tasks are still unexplored. Our paper aims to provide a fresh understanding of graph self-supervised learning based on task correlations. Specifically, we evaluate the performance of the representations trained by one specific task on other tasks and define correlation values to quantify task correlations. Through this process, we unveil the task correlations between various self-supervised tasks and can measure their expressive capabilities, which are closely related to downstream performance. By analyzing the correlation values between tasks across various datasets, we reveal the complexity of task correlations and the limitations of existing multi-task learning methods. To obtain more capable representations, we propose Graph Task Correlation Modeling (GraphTCM) to illustrate the task correlations and utilize it to enhance graph self-supervised training. The experimental results indicate that our method significantly outperforms existing methods across various downstream tasks.

Submission history

Access paper:.

• HTML (experimental)
• Other Formats

References & Citations

• Google Scholar
• Semantic Scholar

BibTeX formatted citation

Bibliographic and Citation Tools

Code, data and media associated with this article, recommenders and search tools.

• Institution

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs .

• Open access
• Published: 14 May 2024

Rain, rain, go away, come again another day: do climate variations enhance the spread of COVID-19?

• Masha Menhat 1 ,
• Effi Helmy Ariffin   ORCID: orcid.org/0000-0002-8534-0113 2 ,
• Wan Shiao Dong 3 ,
• Junainah Zakaria 2 ,
• Aminah Ismailluddin 3 ,
• Hayrol Azril Mohamed Shafril 4 ,
• Mahazan Muhammad 5 ,
• Ahmad Rosli Othman 6 ,
• Thavamaran Kanesan 7 ,
• Suzana Pil Ramli 8 ,
• Mohd Fadzil Akhir 2 &
• Amila Sandaruwan Ratnayake 9  

Globalization and Health volume  20 , Article number:  43 ( 2024 ) Cite this article

114 Accesses

3 Altmetric

Metrics details

The spread of infectious diseases was further promoted due to busy cities, increased travel, and climate change, which led to outbreaks, epidemics, and even pandemics. The world experienced the severity of the 125 nm virus called the coronavirus disease 2019 (COVID-19), a pandemic declared by the World Health Organization (WHO) in 2019. Many investigations revealed a strong correlation between humidity and temperature relative to the kinetics of the virus’s spread into the hosts. This study aimed to solve the riddle of the correlation between environmental factors and COVID-19 by applying RepOrting standards for Systematic Evidence Syntheses (ROSES) with the designed research question. Five temperature and humidity-related themes were deduced via the review processes, namely 1) The link between solar activity and pandemic outbreaks, 2) Regional area, 3) Climate and weather, 4) Relationship between temperature and humidity, and 5) the Governmental disinfection actions and guidelines. A significant relationship between solar activities and pandemic outbreaks was reported throughout the review of past studies. The grand solar minima (1450-1830) and solar minima (1975-2020) coincided with the global pandemic. Meanwhile, the cooler, lower humidity, and low wind movement environment reported higher severity of cases. Moreover, COVID-19 confirmed cases and death cases were higher in countries located within the Northern Hemisphere. The Blackbox of COVID-19 was revealed through the work conducted in this paper that the virus thrives in cooler and low-humidity environments, with emphasis on potential treatments and government measures relative to temperature and humidity.

• The coronavirus disease 2019 (COIVD-19) is spreading faster in low temperatures and humid area.

• Weather and climate serve as environmental drivers in propagating COVID-19.

• Solar radiation influences the spreading of COVID-19.

• The correlation between weather and population as the factor in spreading of COVID-19.

Graphical abstract

Introduction

The revolution and rotation of the Earth and the Sun supply heat and create differential heating on earth. The movements and the 23.5° inclination of the Earth [ 1 ] separate the oblate-ellipsoid-shaped earth into northern and southern hemispheres. Consequently, the division results in various climatic zones at different latitudes and dissimilar local temperatures (see Fig.  1 ) and affects the seasons and length of a day and night in a particular region [ 2 ]. Global differential heating and climate variability occur due to varying solar radiation received by each region [ 3 ]. According to Trenberth and Fasullo [ 4 ] and Hauschild et al. [ 5 ] the new perspective on the issue of climate change can be affected relative to the changes in solar radiation patterns. Since the study by Trenberth and Fasullo [ 4 ] focused on climate model changes from 1950 to 2100, it was found that the role of changing clouds and trapped sunlight can lead to an opening of the aperture for solar radiation.

The annual average temperature data for 2021 in the northern and southern hemispheres ( Source: meteoblue.com ). Note: The black circles mark countries with high Coronavirus disease 2019 (COVID-19) infections

Furthermore, the heat from sunlight is essential to humans; several organisms could not survive without it. Conversely, the spread of any disease-carrying virus tends to increase with less sunlight exposure [ 6 ]. Historically, disease outbreaks that led to epidemic and pandemic eruptions were correlated to atmospheric changes. Pandemic diseases, such as the flu (1918), Asian flu (1956–1958), Hong Kong flu (1968), and recently, the coronavirus disease 2019 (COVID-19) (2019), recorded over a million death toll each during the winter season or minimum temperature conditions [ 7 ]. The total number of COVID-19 cases is illustrated in Fig.  2 .

A graphical representation of the total number of COVID-19 cases across various periods between 2020 and 2021. ( Source : www.worldometers.info ). Note: The black circles indicate countries with high numbers COVID-19-infections

In several previous outbreaks, investigations revealed a significant association between temperature and humidity with a particular focus on the transmission dynamics of the infection from the virus into the hosts [ 8 , 9 , 10 ]. Moreover, disease outbreaks tended to heighten in cold temperatures and low humidity [ 11 ]. Optimal temperature and sufficient relative humidity during evaporation are necessary for cloud formation, resulting in the precipitated liquid falling to the ground as rain, snow, or hail due to the activity of solar radiation balancing [ 4 ].

Consequently, the radiation balancing processes in the atmosphere are directly linked to the living beings on the earth, including plants and animals, and as well as viruses and bacterias. According to Carvalho et al. [ 12 ]‘s study, the survival rate of the Coronaviridae Family can decrease during summer seasons. Nevertheless, numerous diseases were also developed from specific viruses, such as influenza, malaria, and rubella, and in November 2019, a severe health threat originated from a 125 nm size of coronavirus, had resulted in numerous deaths worldwide.

Transmission and symptoms of COVID-19

The COVID-19, or severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), is an infectious disease caused by a newly discovered pathogenic virus from the coronavirus family, the novel coronavirus (2019-nCoV) [ 13 ]. The first case was recorded in Wuhan, China, in December 2019 [ 14 ]. The pathogenic virus is transmitted among humans when they breathe in air contaminated with droplets and tiny airborne particles containing the virus [ 14 , 15 , 16 , 17 , 18 ].

According to the World Health Organization (WHO), the most common symptoms of COVID-19 infection include fever, dry cough, and tiredness. Nevertheless, older people and individuals with underlying health problems (lung and heart problems, high blood pressure, diabetes, or cancer) are at higher risk of becoming seriously ill and developing difficulty breathing [ 19 ]. The COVID-19 was initially only predominant in China but rapidly spread to other countries globally. The remarkably swift acceleration of the number of infections and mortality forced WHO to declare COVID-19 a global public health emergency on the 30th of January 2020, which was later declared as a pandemic on the 11th of March 2020 [ 20 ].

Since no vaccine was available then, WHO introduced the COVID-19 preventative measures to reduce the chances of virus transmission. The guideline for individual preventative included practising hand and respiratory hygiene by regularly cleaning hands with soap and water or alcohol-based sanitisers, wear a facemask and always maintaining at least a one-meter physical distance [ 21 ]. Nevertheless, the worldwide transmission of COVID-19 has resulted in fear and forced numerous countries to impose restrictions rules, such as lockdown, travel bans, closed country borders, restrictions on shipping activities, and movement limitations, to diminish the spread of COVID-19 [ 22 ].

According to WHO, by the 2nd of December 2020, 63,379,338 confirmed cases and 1,476,676 mortalities were recorded globally. On the 3rd of December 2021, 263,655,612 confirmed cases and deaths were recorded, reflecting increased COVID-19 infections compared to the previous year. The American and European regions documented the highest COVID-19 patients with 97,341,769 and 88,248,591 cases, respectively (see Fig. 2 ), followed by Southeast Asia with 44,607,287, Eastern Mediterranean accounted 16,822,791, Western Pacific recorded 6,322,034, and Africa reported the lowest number of cases at 6,322,034 [ 19 ].

Recently, an increasing number of studies are investigating the association between environmental factors (temperature and humidity) and the viability, transmission, and survival of the coronavirus [ 23 , 24 , 25 , 26 ]. The results primarily demonstrated that temperature was more significantly associated with the transmission of COVID-19 [ 27 , 28 , 29 ] and its survival period on the surfaces of objects [ 30 ]. Consequently, the disease was predominant in countries with low temperature and humidity [ 31 ], which was also proven by Diao et al. [ 32 ]‘s study demonstrating higher rates of COVID-19 transmission in China, England, Germany, and Japan.

A comprehensive systematic literature review (SLR) is still lacking despite numerous research on environmental factors linked to coronavirus. Accordingly, this article aimed to fill the gap in understanding and identifying the correlation between environmental factors and COVID-19 by analysing existing reports. Systematically reviewing existing literature is essential to contribute to the body of knowledge and provide beneficial information for public health policymakers.

Methodology

The present study reviewed the protocols, formulation of research questions, selection of studies, appraisal of quality, and data abstraction and analysis.

The protocol review

The present SLR was performed according to the reporting standards for systematic evidence syntheses (ROSES) and followed or adapted the guidelines as closely as possible. Thus, in this study, a systematic literature review was guided by the ROSES review protocol (Fig.  3 ). Compared to preferred reporting items for systematic review and meta-analysis (PRISMA), ROSES is a review protocol specifically designed for a systematic review in the conservation or environment management fields [ 33 ]. Compared to PRISMA, ROSES offers several advantages, as it is tailored to environmental systematic review, which reduces emphasis on quantitative synthesis (e.g. meta-analysis etc.) that is only reliable when used with appropriate data [ 34 ].

The flow diagram guide by ROSES protocol and Thematical Analysis

The current SLR started by determining the appropriate research questions, followed by the selection criteria, including the review, specifically on the keywords employed and the selection of journals database. Subsequently, the appraisal quality process and data abstraction and analysis were conducted.

Formulation of research questions

The entire process of this SLR was guided by the specific research questions, while sources to be reviewed and data abstraction and analysis were in line with the determined research question [ 35 , 36 ]. In the present article, a total of five research questions were formed, namely:

What the link between solar activity and COVID-19 pandemic outbreaks?

Which regions were more prone to COVID-19?

What were the temporal and spatial variabilities of high temperature and humidity during the spread of COVID-19?

What is the relationship between temperature and humidity in propagating COVID-19?

How did the government’s disinfection actions and guidelines can be reducing the spread of COVID-19?

Systematic searching strategies

Selection of studies.

In this stage of the study, the appropriate keywords to be employed in the searching process were determined. After referring to existing literature, six main keywords were chosen for the searching process, namely COVID-19, coronavirus, temperature, humidity, solar radiation and population density. The current study also utilised the boolean operators (OR, AND, AND NOT) and phrase searching.

Scopus was employed as the main database during the searching process, in line with the suggestion by Gusenbauer and Haddaway [ 37 ], who noted the strength of the database in terms of quality control and search and filtering functions. Furthermore, Google Scholar was selected as the supporting database. Although Halevi et al. [ 38 ] expressed concerns about its quality, Haddaway et al. [ 39 ] reported that due to its quantity, Google Scholar was suitable as a supporting database in SLR studies.

In the first stage of the search, 2550 articles were retrieved, which were then screened. The suitable criteria were also determined to control the quality of the articles reviewed [ 40 ]. The criteria are: any documents published between 2000 to 2022, documents that consist previously determined keywords, published in English, and any environment-related studies that focused on COVID-19. Based on these criteria, 2372 articles were excluded and 178 articles were proceeded to the next step namely eligibility. In the eligibility process, the title and the abstract of the articles were examined to ensure its relevancy to the SLR and in this process a total of 120 articles were excluded and only 58 articles were processed in the next stage.

Appraisal of the quality

The study ensured the rigor of the chosen articles based on best evidence synthesis. In the process, predefined inclusion criteria for the review were appraised by the systematic review team based on previously established guidelines and the studies were then judged as being scientifically admissible or not [ 40 ]. Hence, by controlling the quality based on the best evidence synthesis, the present SLR controls its quality by including articles that are in line with the inclusion criteria. It means that any article published within the timeline (in the year 2000 and above), composed of predetermined keywords, in English medium, and environment-related investigations focusing on COVID-19 are included in the review. Based on this process, all 58 articles fulfilled all the inclusion criteria and are considered of good quality and included in the review.

Data abstraction and analysis

The data abstraction process in this study was performed based on five research questions (please refer to 2.2, formulation of research questions). The data that was able to answer the questions were abstracted and placed in a table to ease the data analysis process. The primary data analysis technique employed in the current study was qualitative and relied on thematic analysis.

The thematic technique is a descriptive method that combines data flexibly with other information evaluation methods [ 41 ], aiming to identify the patterns in studies. Any similarities and relationships within the abstracted data emerge as patterns. Subsequently, suitable themes and sub-themes would be developed based on obtained patterns [ 42 ]. Following the thematic process, five themes were selected in this study.

Background of the selected articles

The current study selected 58 articles for the SLR. Five themes were developed based on the thematic analysis from the predetermined research questions: the link between solar activity and pandemic outbreaks, regional area, climate and weather, the relationship between temperature and humidity, and government disinfection action guidelines. Among the articles retrieved between 2000 and 2022; two were published in 2010, one in 2011, four in 2013, three in 2014, two in 2015, six in 2016 and 2017, respectively, one in 2018, six in 2019, twelve in 2020, eight in 2021, and seven in 2022.

Temperature- and humidity-related themes

The link between solar activity and pandemic outbreaks.

Numerous scientists have investigated the relationship between solar activities and pandemic outbreaks over the years ([ 43 ]; A [ 27 , 44 , 45 ].). Nuclear fusions from solar activities have resulted in minimum and maximum solar sunspots. Maximum solar activities are characterised by a high number of sunspots and elevated solar flare frequency and coronal mass injections. Minimum solar sunspot occurrences are identified by low interplanetary magnetic field values entering the earth [ 1 ].

A diminished magnetic field was suggested to be conducive for viruses and bacteria to mutate, hence the onset of pandemics. Nonetheless, Hoyle and Wickramasinghe [ 46 ] reported that the link between solar activity and pandemic outbreaks is only speculative. The literature noted that the data recorded between 1930 and 1970 demonstrated that virus transmissions and pandemic occurrences were coincidental. Moreover, no pandemic cases were reported in 1979, when minimum solar activity was recorded [ 47 ].

Chandra Wickramasinghe et al. [ 48 ] suggested a significant relationship between pandemic outbreaks and solar activities as several grand solar minima, including Sporer (1450–1550 AD), Mounder (1650–1700 AD), and Dalton (1800–1830) minimums, were recorded coinciding with global pandemics of diseases, such as smallpox, the English sweat, plague, and cholera pandemics. Furthermore, since the Dalton minimum, which recorded minimum sunspots, studies from 2002 to 2015 have documented the reappearance of previous pandemics. For example, influenza subtype H1N1 1918/1919 episodically returned in 2009, especially in India, China, and other Asian countries. Zika virus, which first appeared in 1950, flared and became endemic in 2015, transmitted sporadically, specifically in African countries. Similarly, SARS-CoV was first recorded in China in 2002 and emerged as an outbreak, MERS-CoV, in middle east countries a decade later, in 2012.

In 2020, the World Data Centre Sunspot Index and Long-term Solar Observations ( http://sidc.be ) confirmed that a new solar activity was initiated in December 2019, during which a novel coronavirus pandemic also occurred, and present a same as the previous hypothesis. Nevertheless, a higher number of pandemic outbreaks were documented during low minimum solar activities, including Ebola (1976), H5N1 (Nipah) (1967–1968), H1N1 (2009), and COVID-19 (2019–current). Furthermore, Wickramasinghe and Qu [ 49 ] reported that since 1918 or 1919, more devastating and recurrent pandemics tend to occur, particularly after a century. Consequently, within 100 years, a sudden surge of influenza was recorded, and novel influenza was hypothesised to emerge.

Figure  4 demonstrates that low minimum solar activity significantly reduced before 2020, hence substantiating the claim that pandemic events are closely related to solar activities. Moreover, numerous studies (i.e. [ 43 ], Chandra [ 46 , 47 , 48 ]) reported that during solar minimums, new viruses could penetrate the surfaces of the earth and high solar radiation would result in lower infection rates, supporting the hypothesis mentioned above.

The number of sunspots in the last 13 years. Note : The yellow curve indicates the daily sunspot number and the 2010–2021 delineated curve illustrates the minimum solar activity recorded (source: http://sidc.be/silso )

Regional area

In early December 2019, Wuhan, China, was reported as the centre of the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) outbreak [ 50 ]. Chinese health authorities immediately investigated and controlled the spread of the disease. Nevertheless, by late January 2020, the WHO announced that COVID-19 was a global public health emergency. The upgrade was due to the rapid rise in confirmed cases, which were no longer limited to Wuhan [ 28 ]. The disease had spread to 24 other countries, which were mainly in the northern hemisphere, particularly the European and Western Pacific regions, such as France, United Kingdom, Spain, South Korea, Japan, Malaysia, and Indonesia [ 51 , 52 ]. The migration or movement of humans was the leading agent in the spread of COVID-19, resulting in an almost worldwide COVID-19 pandemic [ 53 ].

The first hotspots of the epidemic outspread introduced by the Asian and Western Pacific regions possessed similar winter climates with an average temperature and humidity rate of 5–11 °C and 47–79%. Consequently, several publications reviewed in the current study associated the COVID-19 outbreak with regional climates (i.e. [ 1 , 29 , 54 , 55 ]) instead of its close connection to China. This review also discussed the effects of a range of specific climatological variables on the transmission and epidemiology of COVID-19 in regional climatic conditions.

America and Europe documented the highest COVID-19 cases, outnumbering the number reported in Asia [ 19 ] and on the 2nd of December 2020, the United States of America (USA) reported the highest number of confirmed COVID-19 infections, with over 13,234,551 cases and 264,808 mortalities (Da S [ 56 ].). The cases in the USA began emerging in March 2020 and peaked in late November 2020, during the wintertime in the northern hemisphere (December to March) [ 53 ]. Figure  5 demonstrates the evolution of the COVID-19 pandemic in several country which represent comparison two phase of summer and one phase of winter. Most of these countries tend to increase of COVID cases close to winter season. Then, it can be worsening on phase two of summer due to do not under control of human movement although the normal trend it is presenting during winter phase.

The evolution of the COVID-19 pandemic from the 15th of February 2020 to the 2nd of December 2020 ( Source: https://www.worldometers.info/coronavirus )

The coronavirus spread aggressively across the European region, which recorded the second highest COVID-19 confirmed cases after America. At the end of 2020, WHO reported 19,071,275 Covid-19 cases in the area, where France documented 2,183,275 cases, the European country with the highest number of confirmed cases, followed by the United Kingdom (1,629,661 cases) and Spain (1,652,801 cases) [ 19 ]. Europe is also located in the northern hemisphere and possesses a temperate climate.

The spatial and temporal transmission patterns of coronavirus infection in the European region were similar to America and the Eastern Mediterranean, where the winter season increased COVID-19 cases. Typically, winter in Europe occurs at the beginning of October and ends in March. Hardy et al. [ 57 ] also stated that temperature commonly drops below freezing (approximately − 1 °C) when snow accumulates between December to mid-March, resulting in an extreme environment. Figure 5 indicates that COVID-19 cases peaked in October when the temperature became colder [ 21 ]. Similarly, the cases were the highest in the middle of the year in Australia and South Asian countries, such as India, that experience winter and monsoon, respectively, during the period.

In African regions, the outbreak of COVID-19 escalated rapidly from June to October before falling from October to March, as summer in South Africa generally occurs from November to March, while winter from June to August. Nevertheless, heavy rainfall generally transpires during summer, hence the warm and humid conditions in South Africa and Namibia during summer, while the opposite happens during winter (cold and dry). Consequently, the outbreak in the region recorded an increasing trend during winter and subsided during the summer, supporting the report by Gunthe et al. [ 58 ]. Novel coronavirus disease presents unique and grave challenges in Africa, as it has for the rest of the world. However, the infrastructure and resources have limitations for Africa countries facing COVID-19 pandemic and the threat of other diseases [ 59 ].

Conclusively, seasonal and regional climate patterns were associated with COVID-19 outbreaks globally. According to Kraemer et al. [ 60 ], they used real-time mobility data in Wuhan and early measurement presented a positive correlation between human mobility and spread of COVID-19 cases. However, after the implementation of control measures, this correlation dropped and growth rates became negative in most locations, although shifts in the demographics of reported cases were still indicative of local chains of transmission outside of Wuhan.

Climate and weather

The term “weather” represents the changes in the environment that occur daily and in a short period, while “climate” is defined as atmospheric changes happening over a long time (over 3 months) in specific regions. Consequently, different locations would experience varying climates. Numerous reports suggested climate and weather variabilities as the main drivers that sped or slowed the transmission of SARS-CoV-2 worldwide [ 44 , 61 , 62 , 63 ].

From a meteorological perspective, a favourable environment has led to the continued existence of the COVID-19 virus in the atmosphere [ 64 ]. Studies demonstrated that various meteorological conditions, such as the rate of relative humidity (i.e. [ 28 ]), precipitation (i.e. [ 65 ]), temperature (i.e. [ 66 ]), and wind speed factors (i.e. [ 54 ]), were the crucial components that contributed to the dynamic response of the pandemic, influencing either the mitigation or exacerbation of novel coronavirus transmission. In other words, the environment was considered the medium for spreading the disease when other health considerations were put aside. Consequently, new opinions, knowledge, and findings are published and shared to increase awareness, thus encouraging preventive measures within the public.

The coronavirus could survive in temperatures under 30 °C with a relative humidity of less than 80% [ 67 ], suggesting that high temperatures and lower relative humidity contributed to the elicitation of COVID-19 cases [ 18 , 51 , 58 , 68 ]. Lagtayi et al. [ 7 ] highlighted temperature as a critical factor, evidently from the increased transmission rate of MERS-Cov in African states with a warm and dry climate. Similarly, the highest COVID-19 cases were recorded in dry temperate regions, especially in western Europe (France and Spain), China, and the USA, while the countries nearer to the equator were less affected. Nevertheless, the temperature factor relative to viral infections depends on the protein available in the viruses. According to Chen and Shakhnovich [ 69 ], there is a good correlation between decreasing temperature and the growth of proteins in virus. Consequently, preventive measures that take advantage of conducive environments for specific viruses are challenging.

Precipitation also correlates with influenza [ 43 ]. A report demonstrated that regions with at least 150 mm of monthly precipitation threshold level experienced fewer cases than regions with lower precipitation rates. According to Martins et al. [ 70 ], influenza and COVID-19 can be affected by climate, where virus can be spread through the respiratory especially during rainfall season. The daily spread of Covid-19 cases in tropical countries, which receive high precipitation levels, are far less than in temperate countries [ 27 ]. Likewise, high cases of COVID-19 were reported during the monsoon season (mid-year) in India during which high rainfall is recorded [ 71 ]. Moreover, the majority of the population in these regions has lower vitamin D levels, which may contribute to weakened immune responses during certain seasons [ 27 ].

Rainfall increases the relative atmospheric humidity, which is unfavourable to the coronaviruses as its transmission requires dry and cold weather. Moreover, several reports hypothesised that rain could wash away viruses on object surfaces, which is still questioned. Most people prefer staying home on rainy days, allowing less transmission or close contact. Conversely, [ 72 ] exhibited that precipitation did not significantly impact COVID-19 infectiousness in Oslo, Norway due the location in northern hemisphere which are during winter season presenting so cold.

Coşkun et al. [ 54 ] and Wu et al. [ 29 ] claimed that wind could strongly correlate with the rate of COVID-19 transmission. Atmospheric instability (turbulent occurrences) leads to increased wind speed and reduces the dispersion of particulate matter (PM 2.5 and PM 10 ) in the environment and among humans. An investigation performed in 55 cities in Italy during the COVID-19 outbreak proved that the areas with low wind movement (stable atmospheric conditions) possessed a higher correlation coefficient and exceeded the threshold value of the safe level of PM 2.5 and PM 10 . Resultantly, more individuals were recorded infected with the disease in the regions. As mentioned in Martins et al. [ 70 ] the COVID-19 can be affected by climate and the virus can be spread through respiratory which is the virus moving in the wind movement.

The relationship between temperature and humidity

Climatic parameters, such as temperature and humidity, were investigated as the crucial factors in the epidemiology of the respiratory virus survival and transmission of COVID-19 ([ 61 ]; S [ 73 , 74 ].). The rising number of confirmed cases indicated the strong transmission ability of COVID-19 and was related to meteorological parameters. Furthermore, several studies found that the disease transmission was associated with the temperature and humidity of the environment [ 55 , 64 , 68 , 75 ], while other investigations have examined and reviewed environmental factors that could influence the epidemiological aspects of Covid-19.

Generally, increased COVID-19 cases and deaths corresponded with temperature, humidity, and viral transmission and mortality. Various studies reported that colder and dryer environments favoured COVID-19 epidemiologically [ 45 , 76 , 77 ]. As example tropical region, the observations indicated that the summer (middle of year) and rainy seasons (end of the year) could effectively diminish the transmission and mortality from COVID-19. High precipitation statistically increases relative air humidity, which is unfavourable for the survival of coronavirus, which prefers dry and cold conditions [ 32 , 34 , 78 , 79 ]. Consequently, warmer conditions could reduce COVID-19 transmission. A 1 °C increase in the temperature recorded a decrease in confirmed cases by 8% increase [ 45 ].

Several reports established that the minimum, maximum, and average temperature and humidity correlated with COVID-19 occurrence and mortality [ 55 , 80 , 81 ]. The lowest and highest temperatures of 24 and 27.3 °C and a humidity between 76 and 91% were conducive to spreading the virulence agents. The propagation of the disease peaked at the average temperature of 26 °C and humidity of 55% before gradually decreasing with elevated temperature and humidity [ 78 ].

Researchers are still divided on the effects of temperature and humidity on coronavirus transmission. Xu et al. [ 26 ] confirmed that COVID-19 cases gradually increased with higher temperature and lower humidity, indicating that the virus was actively transmitted in warm and dry conditions. Nevertheless, several reports stated that the spread of COVID-19 was negatively correlated with temperature and humidity [ 10 , 29 , 63 ]. The conflicting findings require further investigation. Moreover, other factors, such as population density, elderly population, cultural aspects, and health interventions, might potentially influence the epidemiology of the disease and necessitate research.

Governmental disinfection actions and guidelines

The COVID-19 is a severe health threat that is still spreading worldwide. The epidemiology of the SAR-CoV-2 virus might be affected by several factors, including meteorological conditions (temperature and humidity), population density, and healthcare quality, that permit it to spread rapidly [ 16 , 17 ]. Nevertheless, in 2020, no effective pharmaceutical interventions or vaccines were available for the diagnosis, treatment, and epidemic prevention against COVID-19 [ 73 , 82 ]. Consequently, after 2020 the governments globally have designed and executed non-pharmacological public health measures, such as lockdown, travel bans, social distancing, quarantine, public place closure, and public health actions, to curb the spread of COVID-19 infections and several studies have reported on the effects of these plans [ 13 , 83 ].

The COVID-19 is mainly spread via respiratory droplets from an infected person’s mouth or nose to another in close contact [ 84 ]. Accordingly, WHO and most governments worldwide have recommended wearing facemasks in public areas to curb the transmission of COVID-19. The facemasks would prevent individuals from breathing COVID-19-contaminated air [ 85 ]. Furthermore, the masks could hinder the transmission of the virus from an infected person as the exhaled air is trapped in droplets collected on the masks, suspending it in the atmosphere for longer. The WHO also recommended adopting a proper hand hygiene routine to prevent transmission and employing protective equipment, such as gloves and body covers, especially for health workers [ 86 ].

Besides wearing protective equipment, social distancing was also employed to control the Covid-19 outbreak [ 74 , 87 ]. Social distancing hinders the human-to-human transmission of the coronavirus in the form of droplets from the mouth and nose, as evidenced by the report from Sun and Zhai [ 88 ]. Conversely, Nair & Selvaraj [ 89 ] demonstrated that social distancing was less effective in communities and cultures where gatherings are the norm. Nonetheless, the issue could be addressed by educating the public and implementing social distancing policies, such as working from home and any form of plague treatment.

Infected persons, individuals who had contact with confirmed or suspected COVID-19 patients, and persons living in areas with high transmission rates were recommended to undergo quarantine by WHO. The quarantine could be implemented voluntarily or legally enforced by authorities and applicable to individuals, groups, or communities (community containment) [ 90 ]. A person under mandatory quarantine must stay in a place for a recommended 14-day period, based on the estimated incubation period of the SARS-CoV-2 [ 19 , 91 ]. According to Stasi et al. [ 92 ], 14-days period for mandatory quarantine it is presenting a clinical improvement after they found 5-day group and 10-day group can be decrease number of patient whose getting effect of COVID-19 from 64 to 54% respectively. This also proven by Ahmadi et al. [ 43 ] and Foad et al. [ 93 ], quarantining could reduce the transmission of COVID-19.

Lockdown and travel bans, especially in China, the centre of the coronavirus outbreak, reduced the infection rate and the correlation of domestic air traffic with COVID-19 cases [ 17 ]. The observations were supported by Sun & Zhai [ 88 ] and Sun et al. [ 94 ], who noted that travel restrictions diminished the number of COVID-19 reports by 75.70% compared to baseline scenarios without restrictions. Furthermore, example in Malaysia, lockdowns improved the air quality of polluted areas especially in primarily at main cities [ 95 ]. As additional, Martins et al. [ 70 ] measure the Human Development Index (HDI) with the specific of socio-economic variables as income, education and health. In their study, the income and education levels are the main relevant factors that affect the socio-economic.

A mandatory lockdown is an area under movement control as a preventive measure to stop the coronavirus from spreading to other areas. Numerous governments worldwide enforced the policy to restrict public movements outside their homes during the pandemic. Resultantly, human-to-human transmission of the virus was effectively reduced. The lockdown and movement control order were also suggested for individuals aged 80 and above or with low or compromised immunities, as these groups possess a higher risk of contracting the disease [ 44 ].

Governments still enforced movement orders even after the introduction of vaccines by Pfizer, Moderna, and Sinovac, as the vaccines only protect high-risk individuals from the worst effects of COVID-19. Consequently, in most countries, after receiving the first vaccine dose, individuals were allowed to resume life as normal but were still required to follow the standard operating procedures (SOP) outlined by the government.

The government attempted to balance preventing COVID-19 spread and recovering economic activities, for example, local businesses, maritime traders, shipping activities, oil and gas production and economic trades [ 22 , 96 ]. Nonetheless, the COVID-19 cases demonstrated an increasing trend during the summer due to the higher number of people travelling and on vacation, primarily to alleviate stress from lockdowns. Several new variants were discovered, including the Delta and Omicron strains, which spread in countries such as the USA and the United Kingdom. The high number of COVID-19 cases prompted the WHO to suggest booster doses to ensure full protection.

As mentioned in this manuscript, the COVID-19 still uncertain for any kind factors that can be affected on spreading of this virus. However, regarding many sources of COVID-19 study, the further assessment on this factor need to be continue to be sure, that we ready to facing probably in 10 years projection of solar minimum phase can be held in same situation for another pandemic.

The sun has an eleven-year cycle known as the solar cycle, related to its magnetic field, which controls the activities on its surface through sunspots. When the magnetic fields are active, numerous sunspots are formed on its surface, hence the sun produces more radiation energy emitted to the earth. The condition is termed solar maximum (see Fig.  6 , denoted by the yellow boxes). Alternatively, as the magnetic field of the sun weakens, the number of sunspots decreases, resulting in less radiation energy being emitted to the earth. The phenomenon is known as the solar minimum (see Fig. 6 , represented by the blue boxes).

The emergence and recurrence of pandemics every 5 years in relation to solar activities ( Source: www.swpc.noaa.gov/ ). Note: The yellow boxes indicate the solar maximum, while the blue boxes represent the solar minimum

The magnetic field of the sun protects the earth from cosmic or galactic cosmic rays emitted by supernova explosions, stars, and gamma-ray bursts [ 97 ]. Nevertheless, galactic cosmic rays could still reach the earth during the solar minimum, the least solar radiation energy period. In the 20th and early 21st centuries, several outbreaks of viral diseases that affected the respiratory system (pneumonia or influenza), namely the Spanish (1918–1919), Asian (1957–1958) and Hong Kong (1968) flu, were documented. Interestingly, the diseases that claimed numerous lives worldwide occurred at the peak of the solar maximum.

Figure  6 illustrates the correlation between the number of sunspots and disease outbreaks from 1975 to 2021, including COVID-19, that began to escalate in December 2019. Under the solar minimum conditions, the spread of Ebola (1976), H5N1 (1997–1998), H1N1 (2009), and COVID-19 (2019-2020) were documented, while the solar maximum phenomenon recorded SARS (2002) and H7N9 (2012–2013) or MERS outbreaks. Nonetheless, solar activity through the production of solar sunspots began to decline since the 22nd solar cycle. Accordingly, further studies are necessary to investigate the influence such solar variations could impart or not on pandemic development.

Despite the findings mentioned above, the sun and cosmic radiations could influence the distribution or outspread of disease-spreading viruses. The rays could kill the viruses via DNA destruction or influence their genetic mutations, which encourage growth and viral evolution. Nevertheless, the connection between radiation and the evolutionary process requires further study by specialists in the field it is become true or not.

The spread of viral diseases transpires naturally in our surroundings and occurs unnoticed by humans. According to records, the spread of pandemic diseases, including the Black Death (fourteenth century) and the Spanish flu (1919), was significantly influenced by the decline and peak of solar activities. Furthermore, in the past 20 years, various diseases related to the influenza virus have been recorded. According to the pattern observed, if all diseases were related to the solar cycle (solar maximum and minimum), the viral diseases would reoccur every 5 to 6 years since they first appeared between 1995 and 2020. Accordingly, the next pandemic might occur around 2024 or 2025 and need to have a proper study for prove these statements. Nonetheless, the activities on the surface of the sun have been weakening since the 23rd solar cycle and it can be proven later after the proper study can be make it.

The beginning of the COVID-19 spread, only several countries with the same winter climate with an average temperature of 5–11 °C and an average humidity rate of 47–79% located at latitudes 30–50 N reported cases. The areas included Wuhan distribution centres in China, the United Kingdom, France, Spain, South Korea, Japan, and the USA (see Fig.  5 ). Other than biological aspects, the higher number of confirmed cases recorded in colder environments was due to the human body secreting less lymphoproliferative hormone, leading to decreased immunogenicity effects and increased risk of infection [ 24 ]. Consequently, the virus could attack and rapidly infect humans during the period [ 1 , 54 ].

The lymphoproliferative response is a protective immune response that plays a vital role in protecting and eradicating infections and diseases. On the other hand, staying in warm conditions or being exposed to more sunlight would lower the risks of infection. According to Asyary and Veruswati [ 98 ], sunlight triggers vitamin D, which increases immunity and increases the recovery rates of infected individuals.

Researchers believe that viruses could survive in the environment for up to 3 to 4 years or even longer. The survival rate of the microorganisms is relatively high, which is related to their biological structures, adaptability on any surfaces, and transmission medium to spread diseases. Viruses possess simple protein structures, namely the spike, membrane, and envelope protein; therefore, when they enter living organisms (such as through the respiratory system), the viruses are easily transmitted.

Once they have entered a host, the viruses duplicate exponentially and swarm the lungs. Subsequently, after the targeted organs, such as the lungs, are invaded, the viruses attack the immune system and create confusion in protective cells to destroy healthy cells. The situation is still considered safe in younger and healthy individuals as their immune systems could differentiate and counter-attack the viruses, curing them. Nonetheless, in elders and individuals with several chronic diseases, most of their protective cells are dead, hence their immune system is forced to work hard to overcome the infection. Pneumonia and death tend to occur when the situation is overwhelming [ 85 ]. Consequently, the viruses are harmful to humans as they could multiply in a short period, enter the blood, and overrun the body.

The coronavirus could attach to surfaces without a host, including door knobs and steel and plastic materials. The microorganisms could survive alone, but virologists have yet to determine how long. If someone touches any surface with the virus, the individual would then be infected. The situation would worsen if the infected person contacted numerous people and became a super spreader. A super spreader does not exhibit any symptoms and continuously transmits the virus without realising it. An infected individual transmits the coronavirus via droplets from coughs or sneezes. Nevertheless, scientists have yet to determine if coronavirus is spread via airborne or droplets, hence requiring thorough evaluation [ 99 ].

The COVID-19 virus mutates over time, and it can be changing any times. Mutations alter the behaviour and genetic structure of the virus, resulting in a new strain. Numerous research have been conducted to procure vaccines and anti-viral medications, but mutations have led to evolutionary disadvantages. The novel strains are more infectious than the original ones. As of November 2020, approximately six new coronavirus strains have been detected, each displaying different transmission behaviours [ 100 ].

Recent studies demonstrated that the mutated viruses exhibit little variability, allowing scientists to produce viable vaccines [ 71 ]. Furthermore, different types of vaccines are manufactured by different countries, which could be advantageous. Currently, most countries also recommend booster doses to attain extra protection after receiving the mandatory two vaccine doses. In same time, the social and physical interactions between humans also necessitate to be aware.

The COVID-19 virus is primarily transmitted through droplets produced by an infected person. Accordingly, physical distancing, a one-metre minimum distance between individuals [ 19 ], and following the SOP might prevent or avoid spreading the disease. Moreover, self-quarantine, school closures, working from home, cancelling large events, limiting gatherings, and avoiding spending long periods in crowded places are essential strategies in enforcing physical distancing at a community level. The policies are essential precautions that could reduce the further spreading of coronavirus and break the chain of transmission.

Government support also need to control the spread of COVID-19 with the strict SOP. The SOP enforcement in public places would enhance adherence to the new practice among the public and the community, aiding in curbing disease transmission. Practising limited meetings and social gatherings, avoiding crowded places, workplace distancing, preventing non-necessary travels of high-risk family members, especially those with chronic disease, and adhering to the recommended SOP could reduce coronavirus outbreaks. Nonetheless, individual awareness is also necessary to achieve COVID-19 spread prevention.

Many researchers are focused on identifying the primary drivers of pandemic outbreaks. Seasonal, temperature, and humidity differences significantly impacted COVID-19 growth rate variations. It is crucial to highlight the potential link between the recurrence of pandemics every 5 years and solar activities, which can influence temperature and humidity variations. Notable variations in COVID-19 mortality rates were observed between northern and southern hemisphere countries, with the former having higher rates. One hypothesis suggests that populations in the northern hemisphere may receive insufficient sunlight to maintain optimal vitamin D levels during winter, possibly leading to higher mortality rates.

The first COVID-19 case was detected in Wuhan, China, which is in the northern hemisphere. The number of cases rapidly propagated in December during the winter season. At the time, the temperature in Wuhan was recorded at 13–18 °C. Accordingly, one theory proposes that the survival and transmission of the coronavirus were due to meteorological conditions, namely temperatures between 13 and 18 °C and 50–80% humidity.

Daily rainfall directly impacts humidity levels. The coronavirus exhibited superior survival rates in cold and dry conditions. Furthermore, transmissible gastroenteritis (TGEV) suspensions and possibly other coronaviruses remain viable longer in their airborne states, which are more reliably collected in low relative humidity than in high humidity. Consequently, summer rains would effectively reduce COVID-19 transmission in southern hemisphere regions.

In southern hemisphere regions, the summer seasons are accompanied by a high average temperature at the end and beginning of the year. Countries with temperatures exceeding 24 °C reported fewer infections. As temperatures rise from winter to summer, virus transmission is expected to decline. Nonetheless, the activities and transmission of the virus were expected to decrease during winter to summer transitions, when the countries would be warmer. The peak intensity of infections strongly depends on the level of seasonal transmissions.

Social distancing plays a critical role in preventing the overload of healthcare systems. Many respiratory pathogens, including those causing mild common cold-like syndromes, show seasonal fluctuations, often peaking in winter. This trend can be attributed to increased indoor crowding, school reopening, and climatic changes during autumn.

The spread of COVID-19 to neighbouring regions can be attributed to population interactions. Migration patterns, such as the movement from northern to southern regions during the warmer months, have significant epidemiological impacts. This trend mirrors the behavior of influenza pandemics where minor outbreaks in spring or summer are often followed by major waves in autumn or winter.

Availability of data and materials

Not applicable.

Abbreviations

Novel coronavirus

Coronavirus disease 2019

Deoxyribonucleic acid

Swine influenza

Influenza A virus subtype H5N1

Asian Lineage Avian Influenza A(H7N9) Virus

Middle East respiratory syndrome

Middle East respiratory syndrome Coronavirus

Particulate matter

Preferred Reporting Items for Systematic Reviews and Meta-Analyses

RepOrting standards for Systematic Evidence Syntheses

Severe Acute Respiratory Syndrome

Severe Acute Respiratory Syndrome Coronavirus

Syndrome coronavirus 2

Systematic literature review

Standard operating procedure

Transmissible gastroenteritis Virus

United States of America

World Health Organization

Apanovich I. Climate and man. Opposition or natural stage of the earth’s evolution? Norwegian journal of development of the international. Science. 2019;26(25):12–27.

Google Scholar  

Borah P, Singh MK, Mahapatra S. Estimation of degree-days for different climatic zones of north-East India. Sustain Cities Soc. 2015;14(1):70–81.

Article   Google Scholar  

Chen D, Chen HW. Using the Köppen classification to quantify climate variation and change: an example for 1901-2010. Environmental Development. 2013;6(1):69–79.

Trenberth KE, Fasullo JT. Global warming due to increasing absorbed solar radiation. Geophys Res Lett. 2009;36

Hauschild MZ, Huijbregts MAJ, Guinée L, Lane J, Fantke P, Zelm v R, et al. Life Cycle Impact Assessment – The Complete World of Life Cycle Assessment; 2015. p. 345.

Book   Google Scholar  

Nakada LYK, Urban RC. COVID-19 pandemic: environmental and social factors influencing the spread of SARS-CoV-2 in São Paulo. Brazil Environmental Science and Pollution Research. 2021;28(30):40322–8.

Article   CAS   PubMed   Google Scholar  

Lagtayi, R., Lairgi, L., Daya, A., & Khouya, A. (2021). The impact of the average temperature, humidity, wind speed, altitude and population density on daily COVID-19 infections’ evolution. January, 9094.

Majumder MS, Liu D, Poirier C, Mandl KD, Lipsitch M, The MS. The role of absolute humidity on transmission rates of the COVID-19 outbreak; 2020.

Wiersinga WJ, Rhodes A, Cheng AC, Peacock SJ, Prescott HC. Pathophysiology, transmission, diagnosis, and treatment of coronavirus disease 2019 (COVID-19): A review. JAMA - Journal of the American Medical Association. 2020;324(8):782–93.

Zhang X, Maggioni V, Houser P, Xue Y, Mei Y. The impact of weather condition and social activity on COVID-19 transmission in the United States. J Environ Manag. 2022;302:114085.

Article   CAS   Google Scholar  

Mäkinen TM, Juvonen R, Jokelainen J, Harju TH, Peitso A, Bloigu A, et al. Cold temperature and low humidity are associated with increased occurrence of respiratory tract infections. Respir Med. 2009;103(3):456–62.

Article   PubMed   Google Scholar  

Carvalho FRS, Henriques DV, Correia O, Schmalwieser AW. Potential of solar UV radiation for inactivation of Coronaviridae family. Photochem Photobiol. 2021;97:213–20.

Ali I, Alharbi OM. COVID-19: disease, management, treatment, and social impact. Sci Total Environ. 2020;728:138861.

Article   CAS   PubMed   PubMed Central   Google Scholar  

Jin Y, Yang H, Ji W, Wu W, Chen S, Zhang W, et al. Virology, epidemiology, pathogenesis, and control of covid-19. Viruses. 2020;12(4):1–17.

Sohrabi C, Alsafi Z, O’Neill N, Khan M, Kerwan A, Al-Jabir A, et al. World health organization declares global emergency: A review of the 2019 novel coronavirus (COVID-19). Int J Surg. 2020;76:71–6.

Article   PubMed   PubMed Central   Google Scholar  

Wang J, Tang K, Feng K, Lv W. High temperature and high humidity reduce the transmission of COVID-19; 2020a.

Wang L, Duan Y, Zhang W, Liang J, Xu J, Zhang Y, et al. Epidemiologic and clinical characteristics of 26 cases of covid-19 arising from patient-to-patient transmission in Liaocheng, China. Clinical Epidemiology. 2020b;12:387–91.

Xie J, Zhu Y. Science of the Total environment association between ambient temperature and COVID-19 infection in 122 cities from China. Sci Total Environ. 2020;724:138201.

World Health Organization. (2020a). Director-General’s opening remarks at the media briefing on COVID-19-10April 2020. https://www.who.int/dg/speeches/detail/who-director-general-s-opening-remarks-at-the-media-briefing-on-covid-19—10- april-2020.

Bostock B. South Korea is testing 200,000 members of a doomsday church linked to more than 60% of its coronavirus cases; 2020.

World Health Organization. (2020b). Health topics/coronavirus. https://www.who.int/health-topics/coronavirus#tab=tab_1.

Menhat M, Mohd Zaideen IM, Yusuf Y, Salleh NHM, Zamri MA, Jeevan J. The impact of Covid-19 pandemic: A review on maritime sectors in Malaysia. Ocean Coast Manag. 2021;209:105638.

Byun WS, Heo SW, Jo G, Kim JW, Kim S, Lee S, et al. Is coronavirus disease (COVID-19) seasonal? A critical analysis of empirical and epidemiological studies at global and local scales. Environ Res. 2021;196:110972.

Dhakal P, Pokhrel P, B. Seasonal variation and COVID-19 pandemic in Nepal. Nepal Medical Journal. 2020;3(2):77–80.

Mehmet Ş. Science of the Total environment impact of weather on COVID-19 pandemic in Turkey. 728; 2020.

Xu H, Yan C, Fu Q, Xiao K, Yu Y, Han D, et al. Science of the Total environment possible environmental effects on the spread of COVID-19 in China. Sci Total Environ. 2020;731:139211.

Rosario DKA, Mutz YS, Bernardes PC, Conte-Junior CA. Relationship between COVID-19 and weather: case study in a tropical country. Int J Hyg Environ Health. 2020;229:113587.

Wang J, Tang K, Feng K, Lin X, Lv W, Chen K, et al. Impact of temperature and relative humidity on the transmission of COVID-19: A modelling study in China and the United States. BMJ Open. 2021;11(2):1–16.

Wu Y, Jing W, Liu J, Ma Q, Yuan J, Wang Y, et al. Effects of temperature and humidity on the daily new cases and new deaths of COVID-19 in 166 countries. Sci Total Environ. 2020;729:1–7.

Casanova LM, Jeon S, Rutala WA, Weber DJ, Sobsey MD. Effects of air temperature and relative humidity on coronavirus survival on surfaces. Appl Environ Microbiol. 2010;76(9):2712–7.

Islam N, Bukhari Q, Jameel Y, Shabnam S, Erzurumluoglu AM, Siddique MA, et al. COVID-19 and climatic factors: A global analysis. Environ Res. 2021;193:110355.

Diao Y, Kodera S, Anzai D, Gomez-Tames J, Rashed EA, Hirata A. Influence of population density, temperature, and absolute humidity on spread and decay durations of COVID-19: A comparative study of scenarios in China, England, Germany, and Japan. One Health. 2021;12:100203.

Haddaway NR, Macura B, Whaley P, Pullin AS. ROSES reporting standards for systematic evidence syntheses: pro forma, flow-diagram and descriptive summary of the plan and conduct of environmental systematic reviews and systematic maps. Environ Evid. 2018;7(1):4–11.

Sharif N, Dey SK. Impact of population density and weather on COVID-19 pandemic and SARS-CoV-2 mutation frequency in Bangladesh. Epidemiol Infect. 2021:1–10.

Kraus S, Breier Dasí-Rodríguez S. El arte de elaborar una revisión bibliográfica sistemática en la investigación sobre el espíritu empresarial. Int Entrep Manag J. 2020;16:1023–42.

Xiao Y, Watson M. Guidance on conducting a systematic literature review. J Plan Educ Res. 2019;39(1):93–112.

Gusenbauer M, Haddaway NR. Which academic search systems are suitable for systematic reviews or meta-analyses? Evaluating retrieval qualities of Google scholar, PubMed, and 26 other resources. Res Synth Methods. 2020;11(2):181–217.

Halevi G, Moed H, Bar-Ilan J. Suitability of Google scholar as a source of scientific information and as a source of data for scientific evaluation—review of the literature. Journal of Informetrics. 2017;11(3):823–34.

Haddaway NR, Collins AM, Coughlin D, Kirk S. The role of google scholar in evidence reviews and its applicability to grey literature searching. PLoS One. 2015;10(9):1–17.

Littlewood C, Chance-Larsen K, McLean S. Quality appraisal as a part of the systematic review. International Journal of Physiotherapy and Rehabilitation. 2010;1(1):53–8.

Vaismoradi M, Turunen H, Bondas T. Content analysis and thematic analysis: implications for conducting a qualitative descriptive study. Nurs Health Sci. 2013;15(3):398–405.

Braun V, Clarke V. Using thematic analysis in psychology. Qual Res Psychol. 2006;3:77–101.

Ahmadi M, Sharifi A, Dorosti S, Jafarzadeh Ghoushchi S, Ghanbari N. Investigation of effective climatology parameters on COVID-19 outbreak in Iran. Sci Total Environ. 2020;729

Gupta A, Banerjee S, Das S. Significance of geographical factors to the COVID-19 outbreak in India. Modeling Earth Systems and Environment. 2020;6(4):2645–53.

Pequeno P, Mendel B, Rosa C, Bosholn M, Souza JL, Baccaro F, et al. Air transportation, population density and temperature predict the spread of COVID-19 in Brazil. PeerJ. 2020;2020(6):1–15.

Hoyle F, Wickramasinghe NC. Sunspots and influenza [6]. Nature. 1990;343(6256):304.

Wickramasinghe NC, Rocca MC, Tokoro G, Temple R. Journal of infectious diseases. Scienctific Research and Community. 2020;1(4):1–10.

Wickramasinghe NC, Steele EJ, Wainwright M, Tokoro G, Fernando M, Qu J. Sunspot cycle minima and pandemics : A case for vigilance at the present time. Journal of Astrobiology & Outreach. 2017;5:2332–519.

Wickramasinghe NC, Qu J. Are we approaching a new influenza pandemic. Virol Curr Res. 2018;2(107):2.

Guan W, Ni Z, Hu Y, Liang W, Ou C, He J, et al. Disease 2019 in China; 2020.

Bashir MF, Ma B, Bilal K, Bashir MA, Tan D, Bashir M. Correlation between climate indicators and COVID-19 pandemic in New York, USA. Sci Total Environ. 2020;728:138835.

Cucinotta D, Vanelli M. WHO declares COVID-19 a pandemic. Acta Biomed. 2020;91:157–60.

PubMed   PubMed Central   Google Scholar  

Lin C, Lau AKH, Fung JCH, Guo C, Chan JWM, Yeung DW, et al. A mechanism-based parameterisation scheme to investigate the association between transmission rate of COVID-19 and meteorological factors on plains in China. Sci Total Environ. 2020;737:140348.

Coşkun H, Yıldırım N, Gündüz S. The spread of COVID-19 virus through population density and wind in Turkey cities. Sci Total Environ. 2021;751

Yang HY, Lee JKW. The impact of temperature on the risk of covid-19: A multinational study. Int J Environ Res Public Health. 2021;18(8)

Candido DD, Watts A, Abade L, Kraemer MUG, Pybus OG, Croda J, et al. Routes for COVID-19 importation in Brazil. Journal of Travel Medicine. 2020;27(3):1–3.

Hardy JP, Groffman PM, Fitzhugh RD, Henry KS, Welman AT, Demers JD, et al. Snow depth manipulation and its influence on soil frost and water dynamics in a northern hardwood forest. Biogeochemistry. 2001;56(2):151–74.

Gunthe SS, Swain B, Patra SS, Amte A. On the global trends and spread of the COVID-19 outbreak: preliminary assessment of the potential relation between location-specific temperature and UV index. Journal of Public Health (Germany). 2020:1–10.

Rosenthal PJ, et al. COVID-19: shining the light on Africa. Am J Trop Med Hyg. 2020;102(6):1145–8.

Kraemer MUG, et al. The effect of human mobility and control measures on the COVID-19 epidemic in China. Science. 2020;368:493–7.

Dalziel BD, Kissler S, Gog JR, Viboud C, Bjørnstad ON, Metcalf CJE, et al. Urbanization and humidity shape the intensity of influenza epidemics in U.S. cities. Science. 2018;362:75–9.

Sahoo PK, Powell MA, Mittal S, Garg VK. Is the transmission of novel coronavirus disease (COVID-19) weather dependent? J Air Waste Manage Assoc. 2020;70(11):1061–4.

Selcuk M, Gormus S, Guven M. Impact of weather parameters and population density on the COVID-19 transmission: evidence from 81 provinces of Turkey. Earth Syst Environ. 2021;5(1):87–100.

Abraham J, Turville C, Dowling K, Florentine S. Does climate play any role in covid-19 spreading?—an Australian perspective. Int J Environ Res Public Health. 2021;18(17)

Sehra ST, Salciccioli JD, Wiebe DJ, Fundin S, Baker JF. Maximum daily temperature, precipitation, ultraviolet light, and rates of transmission of severe acute respiratory syndrome coronavirus 2 in the United States. Clin Infect Dis. 2020;71(9):2482–7.

CAS   PubMed   Google Scholar  

Rubin D, Huang J, Fisher BT, Gasparrini A, Tam V, Song L, et al. Association of Social Distancing, population density, and temperature with the instantaneous reproduction number of SARS-CoV-2 in counties across the United States. JAMA Netw Open. 2020;3(7):1–12.

Comunian S, Dongo D, Milani C, Palestini P. Air pollution and covid-19: The role of particulate matter in the spread and increase of covid-19’s morbidity and mortality. Int J Environ Res Public Health. 2020;17(12):1–22.

Tosepu R, Gunawan J, Effendy DS, Ahmad LOAI, Lestari H, Bahar H, et al. Correlation between weather and Covid-19 pandemic in Jakarta, Indonesia. Sci Total Environ. 2020;725

Chen and Shakhnovich. Thermal adaptation of viruses and Bacteria. Biophys J. 2010;98:1109–18.

Martins LD, da Silva I, Batista WV, Andrade MF, Freitas ED, Jorge Alberto Martins JA. How socio-economic and atmospheric variables impact COVID-19 and influenza outbreaks in tropical and subtropical regions of Brazil. Environ Res. 2020;191:110184.

Kulkarni H, Khandait H, Narlawar UW, Rathod P, Mamtani M. Independent association of meteorological characteristics with initial spread of Covid-19 in India. Sci Total Environ. 2021;764:142801.

Menebo MM. Science of the Total environment temperature and precipitation associate with Covid-19 new daily cases : A correlation study between weather and Covid-19 pandemic in. Sci Total Environ. 2020;737:139659.

Gupta S, Patel KK. Global Epidemiology of First 90 Days into COVID-19 Pandemic :Disease Incidence , Prevalence , Case Fatality Population Density, Urbanisation. J Health Manag. 2020;22(2):117–28.

Haque SE, Rahman M. Association between temperature, humidity, and COVID-19 outbreaks in Bangladesh. Environ Sci Pol. 2020;114:253–5.

Sharma P, Singh AK, Agrawal B, Sharma A. Correlation between weather and COVID-19 pandemic in India: an empirical investigation. J Public Aff. 2020;20(4)

Fu S, Wang B, Zhou J, Xu X, Liu J, Ma Y, et al. Meteorological factors, governmental responses and COVID-19: evidence from four European countries. Environ Res. 2021;194:110596.

Mecenas P, Bastos RT, Vallinoto AC, Normando D. Effects of temperature and humidity on the spread of COVID-19: A systematic review. PLoS One. 2020;15:1–21.

Malki Z, Atlam ES, Hassanien AE, Dagnew G, Elhosseini MA, Gad I. Association between weather data and COVID-19 pandemic predicting mortality rate: machine learning approaches. Chaos, Solitons Fractals. 2020;138:110137.

Sasikumar K, Nath D, Nath R, Chen W. Impact of extreme hot climate on COVID-19 outbreak in India. GeoHealth. 2020;4(12)

Kodera S, Rashed EA, Hirata A. Correlation between COVID-19 morbidity and mortality rates in Japan and local population density, temperature, and absolute humidity. Int J Environ Res Public Health. 2020;17(15):1–14.

Sobral MFF, Duarte GB, da Penha Sobral AIG, Marinho MLM, de Souza Melo A. Association between climate variables and global transmission oF SARS-CoV-2. Sci Total Environ. 2020;729:138997.

Patel SK, Pathak M, Tiwari R, Yatoo MI, Malik YS. A vaccine is not too far for COVID-19 coronavirus pandemic A vaccine is not too far for COVID-19. May; 2020.

Nicola M, Neill NO, Sohrabi C, Khan M, Agha M, Agha R. Evidence based management guideline for the COVID-19 pandemic - review article. Int J Surg. 2020;77:206–16.

Atangana E, Atangana A. Facemasks simple but powerful weapons to protect against COVID-19 spread: can they have sides effects? Results in Physics. 2020;19:103425.

Sarmadi M, Moghanddam VK, Dickerson AS, Martelletti L. Association of COVID-19 distribution with air quality, sociodemographic factors, and comorbidities: an ecological study of US states. Air Qual Atmos Health. 2021;14(4):455–65.

Chung CJ, Nazif TM, Wolbinski M, Hakemi E, Lebehn M, Brandwein R, et al. The restructuring of structural heart disease practice during The Covid-19 pandemic. J Am Coll Cardiol. 2020; InPress

Bukhari Q, Massaro JM, D’agostino RB, Khan S. Effects of weather on coronavirus pandemic. Int J Environ Res Public Health. 2020;17(15):1–12.

Sun C, Zhai Z. The efficacy of social distance and ventilation effectiveness in preventing COVID-19 transmission. Sustain Cities Soc. 2020;62:102390.

Nair N, Selvaraj P. Using a cultural and social identity lens to understand pandemic responses in the US and India. Int J Cross-cult Manag. 2021;21(3):545–68.

Cetron M, Landwirth J. Public health and ethical considerations in planning for quarantine. Yale J Biol Med. 2005;78(5):325–30.

Jernigan DB. Update: public health response to the coronavirus disease 2019 outbreak—United States, February 24, 2020. MMWR. Morbidity and mortality weekly report, 69. 2020.

Stasi C, Fallani S, Voller F, Silvestri C. Treatment for COVID-19: an overview. Eur J Pharmacol. 2020;889:173644.

Foad CAKK, Xun N, Pejman J, Nataraj RC. Nonlinear dynamic analysis of an epidemiological model for COVID-19 including public behavior and government action. Nonlinear Dynamics. 2020;101(3):1545–59.

Sun Z, Zhang H, Yang Y, Wan H, Wang Y. Science of the Total environment impacts of geographic factors and population density on the COVID-19 spreading under the lockdown policies of China. Sci Total Environ. 2020;746(666):141347.

Abdullah S, Mansor AA, Napi NNLM, Mansor WNW, Ahmed AN, Ismail M, et al. Air quality status during 2020 Malaysia movement control order (MCO) due to 2019 novel coronavirus (2019-nCoV) pandemic. Sci Total Environ. 2020;729:139022.

Menhat M, Yusuf Y. Factors influencing the choice of performance measures for the oil and gas supply chain - exploratory study. IOP Conference Series: Materials Science and Engineering. 2018;342(1)

Ćirkovića MM, Vukotića B. Long-term prospects: mitigation of supernova and gamma-ray burst threat to intelligent beings. Acta Astronautica. 2016;129:438–46.

Asyary A, Veruswati M. Science of the Total environment sunlight exposure increased Covid-19 recovery rates : A study in the central pandemic area of Indonesia. Sci Total Environ. 2020;729:139016.

Jayaweeraa M, Pererab H, Gunawardanaa B, Manatungea J. Transmission of COVID-19 virus by droplets and aerosols: A critical review on the unresolved dichotomy. Environ Res. 2020;188:1–18.

Leung K, Shum MHH, Leung GM, Lam TTY, Wu JT. Early transmissibility assessment of the N501Y mutant strains of SARS-CoV-2 in the United Kingdom, October to November 2020. Euro Surveill. 2020;26(1)

Download references

Acknowledgements

The authors would also like to acknowledge the Editors and an anonymous reviewer, who contributed immensely to improving the quality of this publication and a special thanks to Muhammad Hafiy Nauwal Effi Helmy, that contributed an excellent idea through singing during the COVID-19 lockdown period.

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Author information

Authors and affiliations.

Faculty of Maritime Studies, Universiti Malaysia Terengganu, 21030, Kuala Nerus, Terengganu, Malaysia

Masha Menhat

Institute of Oceanography and Environment, Universiti Malaysia Terengganu, 21030, Kuala Nerus, Terengganu, Malaysia

Effi Helmy Ariffin, Junainah Zakaria & Mohd Fadzil Akhir

Faculty of Science and Marine Environment, Universiti Malaysia Terengganu, 21030, Kuala Nerus, Terengganu, Malaysia

Wan Shiao Dong & Aminah Ismailluddin

Institute for Social Science Studies, Universiti Putra Malaysia, 43400, Serdang, Selangor, Malaysia

Hayrol Azril Mohamed Shafril

Social, Environmental and Developmental Sustainability Research Center, Faculty of Social Sciences and Humanities, Universiti Kebangsaan Malaysia, 43600, Bangi, Selangor, Malaysia

Mahazan Muhammad

Institute of Geology Malaysia, Board of Geologists, 62100, Putrajaya, Malaysia

Ahmad Rosli Othman

Executive Office, Proofreading By A UK PhD, 51-1, Biz Avenue II, 63000, Cyberjaya, Malaysia

Thavamaran Kanesan

Faculty of Engineering, Universiti Malaya, 50603, Kuala Lumpur, Malaysia

Suzana Pil Ramli

Faculty of Applied Sciences, Uva Wellassa University, Badulla, 90000, Sri Lanka

Amila Sandaruwan Ratnayake

You can also search for this author in PubMed   Google Scholar

Contributions

All authors have been involved in writing this editorial and contributing to the review of the manuscript. MM and EHA contribute to conceptualization. IA and ARO have made the figure.

Corresponding author

Correspondence to Effi Helmy Ariffin .

Ethics declarations

Ethics approval and consent to participate.

Not Applicable.

Consent for publication

Competing interests.

The authors declare that they have 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.

Menhat, M., Ariffin, E.H., Dong, W.S. et al. Rain, rain, go away, come again another day: do climate variations enhance the spread of COVID-19?. Global Health 20 , 43 (2024). https://doi.org/10.1186/s12992-024-01044-w

Download citation

Received : 27 July 2023

Accepted : 22 April 2024

Published : 14 May 2024

DOI : https://doi.org/10.1186/s12992-024-01044-w

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

• Coronavirus
• Solar radiation
• Temperature
• Social distancing

Globalization and Health

ISSN: 1744-8603

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