Moderators of placebo effect in Generalized Anxiety Disorder (GAD) trials: a meta-regression analysis

INTRODUCTION: Several meta-analyses have studied the moderators of placebo response in psychiatry clinical trials for schizophrenia, depression, ADHD and OCD amongst others. These analyses have identified that number of sites, publication date, sample size and baseline severity are moderators of placebo response. To our knowledge, no meta-regression analysis has been conducted for Generalized Anxiety Disorder (GAD)

AIM: The objective of the present study is to characterize the moderators of placebo response in GAD trials.

METHODS: We conducted a systematic literature search in PubMed to identify randomized, placebo-controlled trials of pharmacological treatments for GAD using the Hamilton Anxiety Rating Scale (HAM-A) as a primary endpoint. The search was limited to studies published from 1994 onwards given that the diagnostic criteria changed substantially in 1994 with the publication of the DSM-IV.

A mixed-effects meta-regression model was conducted using a restricted maximum likelihood (REML) estimator to examine the association between study-level covariates and the observed effect sizes. The model included the following moderators: sample size, number of study arms, baseline scores, treatment duration, number of sites, and publication year. Residual heterogeneity was quantified using τ² and I² statistics. The significance of moderators was assessed using Wald-type tests, and an omnibus test of moderators (QM) was used to evaluate the joint contribution of covariates. The analysis was performed using R and the package metafor.

RESULTS: The search yielded 334 publications. Of these, 268 articles were excluded for not meeting the inclusion criteria (not randomized, not placebo-controlled, not conducted in GAD populations, or not reporting outcomes using the Hamilton Anxiety Rating Scale). A total of 66 full-text articles were assessed for eligibility, of which 23 were excluded because data needed for analysis was not available. Ultimately, 41 studies were included in the quantitative meta-analysis.

Substantial residual heterogeneity was observed (τ² = 3.78; I² = 89.35%), indicating that most of the variability in effect sizes remained unexplained. The set of moderators was not jointly significant (QM(df = 6) = 9.37, p = 0.154), suggesting limited overall explanatory power, with only 7.75% of heterogeneity accounted for by the model. Among individual predictors, the number of study arms was significantly associated with the outcome (β = −0.92, p = 0.028). Given that more negative values reflect a greater placebo effect, this finding indicates that trials with a higher number of arms tend to exhibit a larger placebo response. No other covariates, including sample size, baseline scores, treatment duration, number of sites, or publication year, showed statistically significant associations.

CONCLUSIONS: The only significant moderator of placebo response in GAD trials was the number of arms of the study, which could affect the placebo effect by expectation bias (higher probability of receiving active treatment) or acting as a proxy of trial complexity. Interestingly, other moderators identified in other psychiatric disorder trials were not significant and most of the variance remains unexplained, suggesting that other factors specific to GAD trials moderate the placebo effect in this disorder.

References

No references