What 5 Randomized Controlled Trials Tell Us About Guided Digital Self-Help for Depression in LMICs

The evidence base for guided digital self-help programs for depression has grown substantially over the past decade, but no single program has accumulated the depth and geographic breadth of evidence that WHO Step-by-Step now possesses. Five randomized controlled trials across five countries, a combined sample of more than 2,200 participants, and consistent results across widely varying contexts: the program represents a genuine milestone in global mental health research.

This post reviews what those five trials actually show — their designs, their findings, their limitations, and what they collectively tell us about the promise and boundaries of guided digital self-help for depression in low- and middle-income countries.

Background: The Step-by-Step Program

WHO Step-by-Step is a structured five-session program for mild-to-moderate depression, grounded in behavioral activation with elements of problem solving and cognitive restructuring. It was developed by the World Health Organization as part of its mhGAP (Mental Health Gap Action Programme) portfolio — specifically designed for deployment in settings with limited mental health specialist availability.

The program is delivered digitally (initially via a dedicated app, more recently via WhatsApp) and supplemented by weekly 15-minute telephone check-in calls from trained lay counselors. The combination of self-guided digital content and brief non-specialist human support places it in the “guided self-help” (GSH) category, as distinct from unguided self-help (USH), which provides content without human contact.

The theoretical model is consistent with established behavioral activation principles: depression is maintained by behavioral withdrawal and reduced engagement with positive reinforcement; reintroducing rewarding activities and improving problem-solving capacity interrupts the cycle and reduces symptom severity. Sessions address stress psychoeducation, activity scheduling, stress management, cognitive restructuring, and relapse prevention.

Trial 1: Lebanon (Syrian Refugees)

The first large-scale RCT of Step-by-Step was conducted in Lebanon with Syrian refugees — a population facing compounded adversity including forced displacement, poverty, legal insecurity, and significant exposure to conflict-related trauma. This context is particularly demanding for a self-help intervention, given the severity and complexity of psychosocial stressors that participants faced alongside their depression symptoms.

The trial randomized 693 participants to Step-by-Step plus enhanced usual care (EUC) or EUC alone. The primary outcome was depression symptom severity, measured using the PHQ-9 at three months. Secondary outcomes included anxiety, functional impairment, and wellbeing.

Results: Significant between-group differences were found in favor of Step-by-Step on all primary and secondary outcomes. The intervention group showed greater reductions in PHQ-9 scores, lower anxiety, better functional outcomes, and higher wellbeing ratings at follow-up. The effect sizes were in the medium range, consistent with what the literature on guided self-help generally shows in higher-income settings.

This was a notable finding. The success of a self-guided digital program in a refugee population — where one might anticipate significant barriers to engagement including low literacy, limited connectivity, and competing survival priorities — suggested that the model was more robust to difficult conditions than might have been expected.

Trial 2: China

The Chinese trial tested Step-by-Step in a very different context: a middle-income country with increasing digital infrastructure but significant mental health stigma and historically low rates of help-seeking for psychological problems. The trial enrolled participants with mild-to-moderate depression via online recruitment, randomizing them to Step-by-Step plus brief support calls or a waiting list control.

Results demonstrated significant improvements in depression outcomes among Step-by-Step participants relative to controls, with effects maintained at follow-up assessments. The online recruitment and digital delivery modality produced good engagement rates, suggesting that the WhatsApp/app-based delivery model is well-suited to middle-income urban contexts where smartphone penetration is high.

The Chinese trial also explored the role of the lay counselor contact component. Analysis of engagement patterns confirmed that participants who completed more counselor calls showed better outcomes than those who completed fewer — an early signal for what would become a consistent theme across the trial series.

Trial 3: Pakistan

The Pakistan trial extended the evidence base to South Asia, enrolling participants from rural and semi-rural areas with substantially lower digital literacy and access than the Chinese cohort. Cultural adaptation of the Step-by-Step content for Pakistani contexts involved not only translation into Urdu but modification of examples, scenarios, and activity suggestions to reflect local cultural norms and values.

The trial demonstrated significant effects on depression outcomes, consistent with previous trials. Importantly, effects were found across subgroups defined by gender, rural/urban location, and literacy level — suggesting that the program’s core content and delivery model are robust to considerable variation in participant background.

The Pakistan trial also provided useful data on barriers to engagement. Technical difficulties with app access, connectivity interruptions, and competing household demands were identified as significant factors affecting completion rates. These findings informed later iterations of the program’s delivery approach, including the shift toward WhatsApp as a more accessible delivery platform.

Trial 4: Egypt

The Egyptian trial evaluated Step-by-Step in an Arabic-speaking context, testing the program in both urban and peri-urban settings with diverse socioeconomic profiles. Egypt presented a context with high depression prevalence, limited mental health specialist availability, and significant stigma around psychological help-seeking.

Results were consistent with previous trials: significant reductions in depression symptoms in the intervention group relative to controls. The trial also measured anxiety as a secondary outcome and found significant improvements, adding to the evidence that Step-by-Step produces benefits across the emotional disorder spectrum beyond its primary depression focus.

The Egyptian trial included a longer follow-up period than some previous trials, allowing assessment of whether gains were maintained over time. The evidence of sustained effects at extended follow-up is important for cost-effectiveness calculations, which depend on durable rather than transient symptom improvements.

Trial 5: South Africa

The South African trial completed the five-country evidence base, testing Step-by-Step in a Sub-Saharan African context with a population that was predominantly urban, predominantly isiZulu-speaking, and facing significant burden from both depression and comorbid HIV-related health challenges in a subset of participants.

The trial demonstrated significant effects consistent with the overall evidence base. Notably, the Step-by-Step content required substantial cultural adaptation for the South African context — both linguistic translation and modification of illustrative content to reflect local experience. The success of the program despite this adaptation requirement adds to confidence that the core behavioral activation model generalizes across cultural contexts.

Pooled Evidence: What Five Trials Tell Us

Considered individually, each trial provides meaningful evidence. Considered together, they provide something more valuable: evidence of consistent effectiveness across genuinely diverse settings.

The pooled effect size across the five trials is approximately 0.78, placing Step-by-Step in the medium-to-large range for psychological interventions. For reference, meta-analytic estimates for face-to-face cognitive behavioral therapy produce effect sizes around 0.99. The comparison between guided self-help (g = 0.78) and face-to-face therapy (g = 0.99) is not statistically significant — they are not meaningfully different in their impact on depression outcomes.

By contrast, unguided self-help programs — those that provide digital content without any human support — produce average effect sizes around 0.3. The gap between guided and unguided is larger than the gap between guided and in-person therapy. This finding has direct implications for how the field should prioritize digital mental health investment: the human support component is not a luxury but a structural necessity for achieving treatment-level effects.

Completion Rates and Their Implications

One consistent finding across the trials that deserves careful attention is completion rates. Across the five RCTs, approximately 25 percent of participants completed all five sessions of the program. This figure is lower than would be observed in a clinical trial of in-person therapy, where monitoring and accountability structures are more intensive.

How should we interpret this? On one hand, a 25 percent completion rate appears low relative to the program’s potential. On the other hand, several considerations complicate simple interpretation.

First, the trials used intent-to-treat analysis — effects were calculated across all randomized participants, not only completers. The effect size of 0.78 reflects the average benefit across everyone assigned to the intervention, including those who dropped out after one or two sessions. This makes the cost-effectiveness calculation conservative: actual effects among completers are substantially larger.

Second, partial engagement still produces benefits. Analysis of dose-response relationships suggests that even two or three sessions produce measurable improvements, with gains accumulating as participants complete more modules. Non-completion is not equivalent to non-benefit.

Third, completion rates in community-based digital programs are subject to factors — connectivity interruptions, competing responsibilities, stigma — that are substantially different from clinic-based trials. Improving completion rates through program design and counselor engagement strategies is an active area of development.

Limitations and Honest Uncertainties

Rigorous review of this evidence base requires honest acknowledgment of its limitations.

Blinding. As with all psychological intervention trials, blinding of participants to treatment allocation is not possible. Expectancy effects — participants improving partly because they believe they are receiving treatment — cannot be fully ruled out. The use of active control conditions in some trials (rather than waitlist controls) partially addresses this, but placebo effects remain a methodological challenge in the field.

Follow-up duration. Most of the five trials used follow-up periods of three to six months. Long-term durability of effects beyond six months is less well-characterized. Relapse prevention is built into Session 5 of the program, but whether this is sufficient to prevent relapse at one year or beyond requires longer-term studies.

Comorbidity. The trials primarily enrolled participants with mild-to-moderate depression without significant psychiatric comorbidities. The evidence base does not speak directly to the program’s effectiveness in populations with severe depression, active suicidality, trauma histories, or significant psychiatric comorbidities. These populations require different levels of care, and the trials appropriately screened for severe cases.

Publication bias. The five trials are all reporting positive findings. While this is most plausibly explained by a genuinely effective program, publication bias — the tendency for null results to go unpublished — cannot be entirely excluded as a contributing factor. Pre-registration of future trials and systematic effort to publish null results where they occur would strengthen the evidence base.

Generalizability to new contexts. Evidence from Lebanon, China, Pakistan, Egypt, and South Africa provides reasonable grounds for expecting effectiveness in similar contexts. Generalizability to contexts not yet tested — particularly Sub-Saharan Africa outside South Africa, Latin America, and Southeast Asia — is plausible but not yet directly evidenced.

Implications for Research and Practice

Several research priorities emerge from this review.

First, longer follow-up studies are needed. Three-to-six-month outcomes are a reasonable starting point but insufficient for understanding the durability of effects and the program’s impact on depression recurrence.

Second, component studies — dismantling trials that vary the presence and intensity of the lay counselor contact component — would help characterize the minimum effective dose of human support. If briefer or less frequent contact produces comparable outcomes, the cost-effectiveness ratio improves further.

Third, effectiveness trials in new geographic contexts, particularly high-burden countries not yet studied, are needed to extend the evidence base as Kaya Guides expands.

Fourth, research on moderators of treatment response — who benefits most, and who might benefit from a different level of care — would support more efficient participant triage and care-matching at scale.

For clinicians and program designers working in LMICs, the practical implications are clear: guided digital self-help programs, specifically those that include lay counselor contact, should be considered a first-line option for people with mild-to-moderate depression who have limited or no access to specialist care. The evidence supports this recommendation as strongly as it supports many interventions that have already been widely adopted.

The research reviewed here was drawn from the Ambitious Impact / Charity Entrepreneurship report “Treating Depression with Guided Digital Self-Help Programs” (December 2025). For more on Kaya Guides and the scaling of WHO Step-by-Step, visit besidehealth.org.