More information about text formats
Sallis and colleagues showed that patients who were not consistently meeting physical activity guidelines prior to COVID-19 contamination had a substantially greater risk of hospitalisation, admission in intensive care units, and death than patients who were consistently meeting physical activity guidelines (>150min/week engaging in moderate or strenuous exercise over 2-months).1 Identifying risk factors associated with negative COVID-19 outcomes is timely. COVID-19 has resulted in almost 3,000,000 deaths worldwide by the middle of April 2021 2, and vaccination seems insufficient without health and political behaviour changes. 3
However, we have some concerns about Sallis and colleague’s conclusions. The authors recommended “efforts to promote physical activity” relied on strong assumptions that meeting physical activity guidelines would cause less COVID-19 negative outcomes such as hospitalisation, admission in intensive care units, and deaths. Although exercise has many benefits to individuals, we cannot allow that the urgency of solving problems lead to hasty and imprecise conclusions of causality, as well as unnecessary efforts for implementation.
Consider a “0-10 causality strength scale”, proposed by Pearl (2018) 4, where 0 is weak evidence of causality and 10 is strong evidence of causality. Depending on the assumptions and procedures used in the studies to test the association between variables, we become more or less confident...
Consider a “0-10 causality strength scale”, proposed by Pearl (2018) 4, where 0 is weak evidence of causality and 10 is strong evidence of causality. Depending on the assumptions and procedures used in the studies to test the association between variables, we become more or less confident that an exposure (e.g., physical activity) causes an outcome (e.g., reduction in negative COVID-19 outcomes). Sallis and colleagues conducted a study that tries to understand the relationship of the variables through association, one of the weakest evidence of causation.
Selection bias is a major concern in this study. 5 Firstly, the authors report adjusting for demographics and other risk factors for severe COVID-19 with only a brief mention of the unmeasured confounders. Barbarawi et al. highlighted the importance of unmeasured confounders. Those authors conducted a systematic review that challenged observation evidence to suggest an association between Vitamin D supplementation and cardiovascular disease. 6 The systematic review demonstrates the limitation of using observation data to make causal claims. Even when researchers attempt to emulate the target trial, unmeasured confounding can bias the results. 7
Second, without a causal diagram to understand the assumptions behind the adjustments, it is unclear if, by adjusting for demographics and severe COVID-19 risk factors, the authors have introduced bias into the model, i.e. by adjusting on a collider and opening a back-door path. 8 9 10
We made two recommendations to improve this manuscript. The authors could consider calculating the E-value to assess the sensitivity of results to potential unmeasured confounding. 11 Secondly, we recommend the inclusion of directed acyclic graphs (DAGS). DAGS help depict causal structure to provide a solid theoretical basis on which to base assumptions when considering adjusting for selection bias.12
There is a tendency to accept physical inactivity as a so-called component cause of many illnesses,13 including COVID 19. We understand that this impulse is motivated by a concern for public health. Still, we should not allow author confirmation bias and consistency of these correlational findings to substitute for actual evidence of causality.
Our greatest concern is not that people are advised to increase their physical activity; we are concerned that a misunderstanding of the relation between physical activity and severe COVID-19 leads to an oversimplification of a complex problem that we are just beginning to understand.
We declare no competing interests.
1 Sallis R, Young DR, Tartof SY, et al. Physical inactivity is associated with a higher risk for severe COVID-19 outcomes: a study in 48 440 adult patients. Br J Sports Med 2021;:1–8. doi:10.1136/bjsports-2021-104080
2 Center CR. COVID19 Dashboard by the Center for Systems Science and Engineering (CSSE) at John Hopkins University.
3 Prado B. COVID-19 in Brazil: “So what?” Lancet 2020;395:1461. doi:10.1016/S0140-6736(20)31095-3
4 Pearl J, McKenzie D. The book of why: Thew new science of cause and effect. Basic Books Inc, Division of HarperCollins 10 E.53rd St. New Year, NY 2018.
5 Hernán MA, Cole SR. Invited commentary: Causal diagrams and measurement bias. Am J Epidemiol 2009;170:959–62. doi:10.1093/aje/kwp293
6 Barbarawi M, Kheiri B, Zayed Y, et al. Vitamin D Supplementation and Cardiovascular Disease Risks in More Than 83000 Individuals in 21 Randomized Clinical Trials: A Meta-analysis. JAMA Cardiol 2019;4:765–75. doi:10.1001/jamacardio.2019.1870
7 Hernán MA, Robins JM. Using Big Data to Emulate a Target Trial When a Randomized Trial Is Not Available. Am J Epidemiol 2016;183:758–64. doi:10.1093/aje/kwv254
8 VanderWeele TJ, Hernán MA, Robins JM. Causal directed acyclic graphs and the direction of unmeasured confounding bias. Epidemiology 2008;42:157–62. doi:10.1037/a0030561.Striving
9 VanderWeele TJ. Principles of confounder selection. Eur J Epidemiol 2019;34:211–9. doi:10.1007/s10654-019-00494-6
10 Hernán MA. Invited commentary: Selection bias without colliders. Am J Epidemiol 2017;185:1048–50. doi:10.1093/aje/kwx077
11 Vanderweele TJ, Ding P, Mathur M. Technical Considerations in the Use of the E-value. J Causal Infer 2019;:1–11.
12 Hernán MA, Hernández-Díaz S, Robins JM. A structural approach to selection bias. Epidemiology 2004;15:615–25. doi:10.1097/01.ede.0000135174.63482.43
13 Guthold R, Stevens GA, Riley LM, et al. Articles Worldwide trends in insufficient physical activity from 2001 to 2016 : a pooled analysis of 358 population-based surveys with 1 · 9 million participants. Lancet Glob Heal 2016;6:e1077–86. doi:10.1016/S2214-109X(18)30357-7