We recently read the report by Herrmann et al about body mass index (BMI) and its association with patients’ prognosis after acute myocardial infarction with great interest. According to our knowledge, in most similar previous studies, the study population was divided into 4 groups by standard categories of BMI: lean (<18.5 or 20 kg/m ² ), normal weight (18.5 or 20 to 25 kg/m ² ), overweight (25 to 30 kg/m ² ), and obese (≥30 kg/m ² ). This grouping strategy did make sense but led to highly unequal sample sizes between groups. For example, in the report by Li et al, among 1,429 patients recruited, only 15 patients (0.98%) were lean and only 189 patients (13.4%) were obese. Similarly in the report by Kaneko et al, there were only 92 lean patients (7.6%) and 56 obese patients (4.6%) among the study population of 1,205 patients. The significantly smaller sample sizes in lean and obese groups could probably cause selection bias and attenuation of power of test in statistical analyses, which might lead to distortion in results. Herrmann et al tried to overcome this problem by adopting a stratifying method that divided the study population by BMI into quartiles (<24.5, 24.5 to <27.1, 27.1 to 30.1, and >30.1 kg/m ² ). This method made the 4 groups almost the same in sample size (890, 899, 898, and 892 patients in each group). However, an obvious disadvantage is that it made the results difficult to interpret. For example, the first group covered an excessively wide range of BMI from underweight to normal weight, whereas a middle quartile (median) of 27.1 kg/m ² did not mean anything when discussing about BMI. We believed that a better method for solving the problem should be propensity score matching. By calculating the propensity scores, one could perform one-to-many matching between obese and normal-weight groups or between lean and normal-weight groups. This method ensured a meaningful grouping strategy as well as balanced baseline characteristics and comparable sample sizes between groups. Consequently, more reliable results that reveal the truth of obesity paradox would likely be achieved.

In addition, we also noticed that in the report by Herrmann et al, the study population was “multinational.” The cohort was described in the previous published report of Harmonizing Outcomes with Revascularization and Stents in Acute Myocardial Infarction trial as from “123 centers in 11 countries,” but no further details were provided. We would like to indicate that racial differences might possibly play a role in obesity paradox phenomenon. For example, obesity paradox was frequently reported in eastern Asian population, such as in Japanese and Korean. But in China, 3 previous studies declared that obesity paradox was not observed in patients with acute myocardial infarction or those undergoing percutaneous coronary intervention. Therefore, a clear description of racial structure was necessary in a multinational cohort, and subgroup analysis between races might be conducted.