Design: Systematic review and meta-analysis of placebo arms in blinded RCTs
Population: 30 trials, 5,680 participants in digital placebo arms
Primary outcome: Change in PHQ-9 within digital placebo arms
Effect: Hedges g 0.44 (95% CI 0.29-0.59), I² 93.2%

A meta-analysis in the Journal of Medical Internet Research (9 October 2026) pooled the control arms of 30 blinded randomised trials, with 5,680 participants, in which a sham app or other digital placebo was the comparator and depression was measured with the PHQ-9.

Within those placebo arms, PHQ-9 scores improved with a pooled Hedges g of 0.44 (95% CI 0.29-0.59), with very high heterogeneity (I² 93%). Trials in people with a primary psychiatric disorder, and those with higher baseline scores, showed bigger placebo improvements; together these explained about half the variation.

This changes how to read the evidence for mental health apps. A trial comparing an app with a waiting list folds this sizeable placebo response into the app's apparent benefit. The useful question about any digital therapeutic is how much it beats a credible sham.

Check what the control arm was before trusting a mental health app trial
Discount app benefits measured against waiting lists or usual care alone
Expect larger placebo responses in more severe depression and in clinical populations
Tell patients that engagement itself may help, but is not the same as a specific treatment effect

Why it matters

Much of the claimed benefit of mental health apps may be the response to any credible digital contact.

The statistics, in plain English

Hedges g of 0.44 is a within-arm, moderate-sized improvement, similar in scale to some active treatments. The I² of 93% means the trials varied enormously, so the pooled number is an average across very different situations rather than a figure to expect in any one setting.