Magnitude and Pattern of Placebo Response in Clinical Trials of Oral Antihyperglycemic Agents: Data From the U.S. Food and Drug Administration, 1999–2015

Abstract

OBJECTIVE To evaluate the magnitude of placebo response and treatment response patterns in clinical trials of investigational oral antihyperglycemics over time. RESEARCH DESIGN AND METHODS We examined the U.S. Food and Drug Administration medical and statistical reviews for 19 oral antihyperglycemic agents (23,438 patients, 50 trials, and 96 treatment arms) approved between 1999 and 2015. Placebo and medication treatment response (HbA1c reduction) and effect sizes were examined over time (year of approval). Exclusively placebo-controlled and augmented/adjunctive placebo-controlled trials were analyzed separately, and differences were compared. Potential effects of trial and patient characteristics were explored. RESULTS In more recent trials, augmented placebo-controlled arms reduced HbA1c by 0.2% on average and more frequently lowered HbA1c from baseline compared with exclusively placebo-controlled arms (63 vs. 18%; χ2 = 9.93; P = 0.002). In exclusively placebo-controlled trials, placebo response increased significantly over time (β = 0.035; R2 = 0.31; P = 0.0013), reaching ∼0% average change in HbA1c, whereas drug response also increased significantly (β = 0.017; R2 = 0.076; P = 0.0498). In augmented placebo-controlled trials, placebo response (β = 0.33; R2 = 0.407; P < 0.001) showed the same pattern, whereas the growth in drug response was not significant (R2 = 0.031; P = 0.207). Placebo response in both groups increased by 0.5% HbA1c reduction over time, whereas effect sizes remained stable with high success rates (100%; 96 out of 96). Drug response and effect size were not significantly predicted by patient or trial characteristics, but follow-up analysis suggested an inverse relationship of placebo baseline HbA1c with placebo response. CONCLUSIONS Remarkably, placebo-treated patients with diabetes commonly experienced reduction in HbA1c, more markedly in augmented compared with exclusively placebo-controlled treatment arms. Placebo response increased significantly over time without impacting efficacy outcomes. Nonpharmacologic effects measured in the placebo response appear stronger when used with active medication than when implemented in isolation and may be related to the level of HbA1c at baseline.