Comparative effects of the restriction method in two large observational studies of body mass index and mortality among adults

Abstract

Background A method applied in some large studies of weight and mortality is to begin with a well-defined analytic cohort and use successive restrictions in order to control for methodologic bias and arrive at final analytic results. Materials and methods Two observational studies of body mass index and mortality allow a comparative assessment of these restrictions in very large data sets. One was a meta-analysis of individual participant data with a sample size of 8 million. The second was a study of a South Korean cohort with a sample size of 12 million. Both presented results for participants without pre-existing disease before and after restricting the sample to never smokers and deleting the first 5 years of follow-up. Results Initial results from both studies were generally similar, with hazard ratios (HRs) below 1 for overweight and above 1 for underweight and obesity. The meta-analysis showed higher HRs for overweight and obesity after the restrictions, including a change in the direction of the HR for overweight from 0.99 (95% CI 0.98–1.01) to 1.11 (95% CI 1.10, 1.11). The South Korean data showed little effect of the restrictions and the HR for overweight changed from 0.85 (95% CI 0.84–0.86) to 0.91 (95% CI 0.90, 0.91). The summary effect size for overweight was 0.90 (95% CI 0.89–0.91) before restrictions and 1.02 (95% CI 1.02, 1.03) after restrictions. Conclusions The effect of the restrictions is not consistent across studies, weakening the argument that analyses without such restrictions lack validity.

Flawed methods and inappropriate conclusions for health policy on overweight and obesity: the Global BMI Mortality Collaboration meta‐analysis

“…Yi et al . in particular still found hazard ratios below 1 for overweight subjects, suggesting that in fact, the GBMC findings are misleading and the applied restrictions introduced bias to the GBMC data . The Yi paper is instead consistent with the findings of Flegal et al .…”

“…Yi et al in particular still found hazard ratios below 1 for overweight subjects, suggesting that in fact, the GBMC findings are misleading and the applied restrictions introduced bias to the GBMC data. 29 The Yi paper is instead consistent with the findings of Flegal et al 30 In particular, the data from Yi et al showed that overweight was not associated with an increased risk of death (hazard ratio of 0.85) and grade 1 obesity (BMI 30-35) was associated with only a very small increased risk of death (hazard ratio of 1.06).…”

The evolution of body composition in oncology—epidemiology, clinical trials, and the future of patient care: facts and numbers

“…cancer-specific survival or overall survival) is less consistent, and for many malignancies, overweight or obesity is associated with a survival advantage ( Figure 1). [17][18][19][20][21][22] Other explanations involve BMI being too crude a measure to be useful at the individual patient level. For example, in a pooled analysis of 22 randomized therapeutic treatment trials that included 11 724 patients with cancer, 67% were overweight/obese (BMI ≥25 kg/m 2 ) at the time of enrolment (e.g.…”

“…However, when these methodological concerns are empirically tested, many are not substantiated and the previously observed associations persist. [17][18][19][20][21][22] Other explanations involve BMI being too crude a measure to be useful at the individual patient level. 23 BMI does not differentiate lean mass from adipose mass, nor does it describe regional adipose tissue deposition (e.g.…”

Anthropometric risk factors for ovarian cancer in the NIH-AARP Diet and Health Study

“…A further threat to the validity of our estimates is potential unobserved confounding by undiagnosed ovarian cancer (often referred to as “reverse causation” [ 34 ]) if these conditions are symptomatic enough to induce a change in body weight. We, therefore, assumed 3 years of minimum latent period required for weight change due to unobserved disease to affect the outcome and excluded events that occurred during this time [ 35 ]. The statistical analysis was performed using Stata 15.1.…”