Widespread demand for accessible mental health support has intensified interest in interventions that can be delivered without face-to-face contact. Standalone smartphone apps are positioned to reach people on waiting lists or in under-resourced settings, offering immediate, self-guided tools grounded in psychological approaches. A large synthesis of randomised controlled trials (RCTs) assessed whether such apps, used without adjunct therapy, outperform inactive comparators. The analysis focused on adults with heightened symptom severity or diagnosed conditions and evaluated outcomes at post-intervention against waitlist, informational material, sham or other inactive controls. Findings indicate clinically relevant benefits in several symptom domains, alongside meaningful uncertainty driven by heterogeneity and study quality.
Efficacy Across Symptom Domains
Across 72 RCTs with 21 702 participants, standalone apps achieved significant post-intervention effects for depression, anxiety and sleep problems, with additional signals for post-traumatic stress disorder (PTSD), eating disorders and body dysmorphic disorder (BDD). Pooled effects favoured apps for depression, anxiety and sleep, with the largest effects observed for sleep problems and BDD. For PTSD, eating disorders and BDD, effects were positive though more modest or based on fewer comparisons. Two small trials targeting obsessive-compulsive disorder (OCD) each reported significant benefits. By contrast, pooled effects were non-significant for smoking, self-injury, suicidal ideation and alcohol misuse, indicating little to no symptom improvement in these domains when apps were used without adjunctive treatment. A single comparison in schizophrenia found no significant difference.
Publication bias analyses identified asymmetry for depression and anxiety. After trim-and-fill adjustment, effects for both outcomes attenuated to small magnitudes while remaining positive, whereas sleep effects were stable. Substantial heterogeneity characterised most pooled outcomes, reflecting variability in samples, measures, intervention components and control conditions. In light of these factors, the evidence supports cautious use of standalone apps for depression, anxiety and sleep problems, while applications to smoking, alcohol misuse, self-injury, suicidal ideation and schizophrenia should not be recommended on current data.
Must Read: Setting Safe Boundaries for AI in Mental Health
Study Quality and Bias Considerations
Risk-of-bias assessments indicated moderate to high concerns overall. Randomisation processes were frequently adequate, and most trials reported minimal deviations from intended interventions. Handling of missing data was generally appropriate in intention-to-treat analyses, yet outcome measurement quality and selective reporting posed recurring limitations. Considerable heterogeneity was common, with I² values often in the substantial to considerable range. Leave-one-out sensitivity checks and prediction intervals underscored between-study variability, limiting generalisability.
The evidence base included diverse target populations, from individuals with subthreshold symptoms to those with diagnoses confirmed by structured interviews. Post-assessment typically occurred within weeks to a few months, and fewer than two thirds of trials included follow-up. Dropout rates varied widely, with some trials retaining large cohorts and others experiencing pronounced attrition. These features, combined with publication bias signals for key outcomes, require conservative interpretation of pooled effects and careful attention to trial design when extrapolating to routine practice.
Intervention Features and Trial Context
Standalone apps were predominantly grounded in cognitive-behavioural therapy, with additional frameworks spanning mindfulness, behavioural activation, cognitive bias modification, exposure therapy, acceptance-based approaches and positive psychology. Common components included engagement features, symptom monitoring, reminders to support adherence and tailoring of content. Interventions were delivered via installed mobile apps, smartphone-optimised apps or mobile web applications. Control conditions were typically waitlist, control apps or informational materials, aligning with the focus on scenarios where no active treatment is available.
Across trials, many participants were explicitly allowed to continue medication, and most analyses followed an intention-to-treat approach. Recruitment spanned high-income and middle-income countries, with notable representation of students, working adults, caregivers, people with comorbid conditions and trauma-exposed groups. Number needed to treat estimates reported for significant outcomes favoured sleep and BDD, consistent with the larger pooled effects in these domains. However, the diversity of populations, measures and app designs contributes to heterogeneity and complicates identification of specific components that drive benefit.
Evidence indicates that standalone smartphone apps can reduce symptoms of depression, anxiety and sleep problems compared with inactive alternatives, with smaller or uncertain benefits across other domains and no reliable effects for smoking, alcohol misuse, self-injury, suicidal ideation or schizophrenia. High heterogeneity, moderate study quality and publication bias for some outcomes warrant cautious implementation and reinforce the primacy of established first-line treatments where available. For settings with limited access to care, these apps might serve as an entry point, provided selection and use are guided by the symptom domain and the evidential strengths and limits identified in the current literature.
Source: The Lancet Digital Health
Image Credit: iStock
