Ethnicity, heart failure, atrial fibrillation and diabetes: collider bias

In the accompanying paper,1 Tan et al conclude that in people with heart failure (HF) there is a strikingly lower prevalence of atrial fibrillation (AF) among Asian people in Singapore compared with European people in New Zealand (NZ). When seeking an explanation for this, they found that Asian patients with HF with diabetes had less AF than Asian patients without diabetes. This lower rate of diabetes is unexpected and could be called a paradox since previous cohort studies had shown diabetes to be associated with a higher risk of AF.2 3 Both of these findings, if correct, are of importance. Ethnic differences might point to an underlying genetic cause of AF; if diabetes does not cause AF (in Asians) then lifestyle changes to prevent diabetes4 might not prevent AF, as has been suggested.5 Despite the potential interest of the findings in this work, it is important to consider that the design of the study has an important bearing on the results and their interpretation. The prevalence of AF may be lower in Asian people in Singapore, as the authors suggest, but this may be due to untestable differences in the make-up of the peoples in the study or because of a differential survivor bias, prerecruitment. Here we consider the problem of collider bias and how the structure of this study may have influenced the results and their interpretation. The paper1 is an observational study looking at data from a previous prospective multicentre longitudinal study6 based in NZ and Singapore. In this original study, patients with HF were assessed at baseline and followed over 2 years in two countries, NZ and Singapore. People were selected from 9598 presenting with HF to hospital or attending a hospital clinic for management of HF within 6 months of decompensated HF. …