Importance: Variables such as severe symptoms, comorbidity, and sociodemographic characteristics (eg, low educational attainment or unemployment) are associated with a poorer prognosis in adults treated for depressive symptoms. The exclusion of patients with a poor prognosis from RCTs is negatively associated with the generalizability of research findings.
Objective: To compare the prognostic risk factors (PRFs) in patient samples of RCTs of face-to-face therapy (FTF) and internet-based therapy (IBT) for depression.
Data sources: PsycINFO, Cochrane CENTRAL, and reference lists of published meta-analyses were searched from January 1, 2000, to December 31, 2021.
Study selection: RCTs that compared FTF (individual or group therapy) and IBT (guided or self-guided interventions) against a control (waitlist or treatment as usual) in adults with symptoms of depression were included.
Data extraction and synthesis: Data were extracted by 2 independent observers. The Cochrane revised risk-of-bias tool was used to assess the risk of bias. The study was preregistered with OSF Registries and followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guideline.
Main outcomes and measures: The primary outcome was the standardized mean difference (Hedges g effect size) in depressive symptoms at treatment termination (assessed with standard patient self-report questionnaires), with a positive standardized mean difference indicating larger improvements in the intervention compared with those in the control group. Meta-regression analyses were adjusted for the type of control group. Three preregistered and 2 exploratory sensitivity analyses were conducted. A prognostic risk index (PROG) was created that calculated the sum of 12 predefined individual indicators, with scores ranging from 0 to 12 and higher scores indicating that a sample comprised patients with poorer prognoses.
Results: This systematic review and meta-regression analysis identified 105 eligible RCTs that comprised 18 363 patients. In total, 48 studies (46%) examined FTF, and 57 studies (54%) examined IBT. The PROG was significantly higher in the RCTs of FTF than in the RCTs of IBT (FTF: mean [SD], 3.55 [1.75]; median [IQR], 3.5 [2.0-4.5]; IBT: mean [SD], 2.27 [1.66]; median [IQR], 2.0 [1.0-3.5]; z = -3.68, P < .001; Hedges g = 0.75; 95% CI, 0.36-1.15). A random-effects meta-regression analysis found no association of the PROG with the effect size. Sensitivity analyses with outliers excluded and accounting for risk of bias or small-study effects yielded mixed results on the association between the PROG and effect size.
Conclusions and relevance: The findings of this systematic review and meta-regression analysis suggest that samples of RCTs of FTF vs IBT differ with regard to PRFs. These findings have implications for the generalizability of the current evidence on IBT for depression. More RCTs of internet-based interventions with clinically representative samples are needed, and the reporting of PRFs must be improved.