Secondary Logo

Journal Logo


Can Collider Bias Explain Paradoxical Associations?

Stensrud, Mats Julius; Valberg, Morten; Aalen, Odd O.

Author Information
doi: 10.1097/EDE.0000000000000653
  • Free

To the Editor:

Limiting the study population to diseased subjects may influence the effect estimates,1–4 because collider bias is introduced. Sperrin et al.5 recently suggested that “the bias is small relative to the causal relationships between the variables.” Furthermore, they stated that collider bias fully explains obesity paradoxes only if “the true causal effect is small, and therefore unlikely to be important; or the effect of the unmeasured confounder on both the mediator and the outcome is very large, therefore unlikely to be missed from the analysis.”

We agree with Sperrin et al.5 that the unknown effects of U must be large to fully explain the paradoxical associations. However, we disagree with the statement that a large, unmeasured effect is implausible. We rather believe that some diseases are characterized by large unknown effects. In this letter, we will consider the effect of obesity in patients with rheumatoid arthritis (RA), a frequently mentioned example of an obesity paradox. We will assess the bias by using a frailty model, and we argue that this model has particularly nice properties to explore selection bias.

Obesity is associated with increased mortality in the general population, and it is also associated with increased risk of developing RA.6 However, in patients with RA, obesity seems to be associated with lower mortality rates. For example, a hazard ratio of 0.93 (95% confidence interval [CI]: 0.89, 0.98) for a unit increase in body mass index (BMI) has been reported in patients with BMI above 20.7 This estimate, however, may be influenced by selection bias due to conditioning on a collider.

In a parallel publication, we suggested methods to evaluate selection bias.8 These methods are based on frailty models, and they are naturally applicable to time-to-event outcomes. We will now apply these methods to assess obesity paradoxes, using published data on RA in a crude example.

We consider the same causal structure as Sperrin et al.5 Let us study subjects at a particular age, say t = 50 years, which is similar to that of Escalante et al.7 Rather than assuming a binary U, we let U (hereby denoted the frailty) be drawn from a probability distribution, in this example the gamma distribution. Without loss of generality, we assume that the distribution has mean equal to 1. Assuming that the probability of developing RA is equal among first-degree siblings, we can now estimate the variance δ from the sibling recurrence risk (familial relative risk, FRR), when the disease is rare9: δ = FRR − 1

For RA, the FRR is reported to be 4.64.10 In contrast to Sperrin et al.,5 our approach therefore allows us to estimate the effect of U on the exposure M using published data. The effect of U on the outcome Y is still unknown, but we let this effect be the same as the effect of U on M Heuristically, we thereby assume that some people with a high risk of developing RA also have a higher mortality rate, for example, due to an aggressive form of RA or general “frailty.” Then, we use the approach described in the appendix of Stensrud et al.8 Consistent with published data,6 we assume that obese people experience an additional hazard rate of RA that is 0.4 times the baseline hazard rate. To get a notion about the size of collider bias, we first assume no causal effect of obesity on mortality, that is, a causal hazard ratio HRc = 1. Then, we would observe a hazard ratio HRo = 0.77 immediately after RA is diagnosed.

Given RA, we may also assume that obesity has a causal, harmful effect on mortality. For example, we introduce an additional hazard component in obese people, and let this effect be 0.5 times the baseline hazard rate. We assume a baseline hazard rate of 0.036, coherent with the results by Escalante et al.7 In a 10-year period after RA was diagnosed, we calculated that the observed HR would lie between 0.87 and 0.88. Hence, even if obesity increases disease risk causally, it may appear protective due to collider bias.

It has been claimed that an unknown confounder must be unrealistically large to explain obesity paradoxes.5 By using a frailty model supplied with published data on RA, we suggest that large effects could be present in real life. RA is a disease with a strong familial risk component, but the causes of RA are to a large extent unknown.11 Indeed, this may suggest that considerable, unknown factors contribute to the disease process. If these factors also influence mortality, the magnitude of collider bias may be large.

Frailty models allow us to use published data when the size of an unknown confounder is evaluated. Furthermore, these models are naturally compatible with time-to-event outcomes. Inevitably, the frailty approach also relies on a parametric, causal structure. However, the frailty model assumptions are relatively easy to evaluate, and a range of frailty models may be fitted to assess the stability of the bias estimates.

Mats Julius Stensrud

Department of Biostatistics

Institute of Basic Medical Sciences

University of Oslo

Oslo, Norway

Diakonhjemmet Hospital

Oslo, Norway

Morten Valberg

Odd O. Aalen

Department of Biostatistics

Institute of Basic Medical Sciences

University of Oslo

Oslo, Norway


1. Flanders WD, Eldridge RC, McClellan W. A nearly unavoidable mechanism for collider bias with index-event studies. Epidemiology. 2014;25:762–764.
2. Glymour MM, Vittinghoff E. Commentary: selection bias as an explanation for the obesity paradox: just because it’s possible doesn’t mean it’s plausible. Epidemiology. 2014;25:4–6.
3. Banack HR, Kaufman JS. The “obesity paradox” explained. Epidemiology. 2013;24:461–462.
4. Banack HR, Kaufman JS. From bad to worse: collider stratification amplifies confounding bias in the “obesity paradox”. Eur J Epidemiol. 2015;30:1111–1114.
5. Sperrin M, Candlish J, Badrick E, Renehan A, Buchan I. Collider bias is only a partial explanation for the obesity paradox. Epidemiology. 2016;27:525–530.
6. Voigt LF, Koepsell TD, Nelson JL, Dugowson CE, Daling JR. Smoking, obesity, alcohol consumption, and the risk of rheumatoid arthritis. Epidemiology. 1994;5:525–532.
7. Escalante A, Haas RW, del Rincón I. Paradoxical effect of body mass index on survival in rheumatoid arthritis: role of comorbidity and systemic inflammation. Arch Intern Med. 2005;165:1624–1629.
8. Stensrud MJ, Valberg M, Røysland K, Aalen OO. Exploring selection bias by causal frailty models: the magnitude matters. Epidemiology. 2017;.28:379–386
9. Moger TA, Aalen OO, Heimdal K, Gjessing HK. Analysis of testicular cancer data using a frailty model with familial dependence. Stat Med. 2004;23:617–632.
10. Hemminki K, Li X, Sundquist J, Sundquist K. Familial associations of rheumatoid arthritis with autoimmune diseases and related conditions. Arthritis Rheum. 2009;60:661–668.
11. Schur P, Gabriel S, Crowson C. Epidemiology of, risk factors for, and possible causes of rheumatoid arthritis. UpToDate, Waltham. 2014. Available at: Accessed April 10, 2017.
Copyright © 2017 Wolters Kluwer Health, Inc. All rights reserved.