Objective: Attrition, or loss to follow-up, is a common problem in studies of type 2 diabetes remission following roux-en-Y gastric bypass (RYGB) and is often correlated with weight loss. Thus, reported rates of remission may be inflated by attrition bias. We investigate the effect of attrition bias on reported diabetes remission rates following RYGB.
Methods: Using sensitivity analyses, we identified sets of attrition and remission rates that produced simulated outcomes within 95% confidence intervals of the reported outcomes from five studies of diabetes remission following RYGB.
Results: Potential attrition bias varied greatly, yielding possible remission rates of diabetes ranging from 20 to 40% at 1 year. For studies with the attrition greater than ~ 20%, estimates that ignored attrition overestimated diabetes remission rates. Kaplan-Meier estimates were less affected by attrition. Potential for bias was most evident in the study with the largest sample size.
Conclusion: Researchers, clinicians, and policymakers can measure potential attrition bias in clinical studies. In the case of remission of diabetes following RYGB, the potential bias in reported remission rates is generally less than 10%, varies considerably among studies, and is primarily driven by attrition rate and study size. Studies with very large sample sizes may provide a narrow confidence interval around a biased estimate.
Keywords: Bariatric surgery; Bias; Missing data; Retention; Simulation study.