Current practices in modelling dose-response relationships in network meta-analyses of depression treatments: a meta-research study.

OBJECTIVES: Dose can meaningfully affect the comparative effectiveness of depression treatments, yet how network meta-analyses (NMAs) in this field handle dose is not well characterised. We conducted a meta-research study to examine how current depression NMAs incorporate dosing information and identify methodological gaps in current practice. DESIGN: Meta-research study following guidance for meta-epidemiological research and incorporating relevant elements of Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020. DATA SOURCES: We searched PubMed, Embase, the Cochrane Library, PsycINFO, Web of Science and Scopus for NMAs published between January 2020 and May 2025. ELIGIBILITY CRITERIA: NMAs comparing interventions, or different doses of a single intervention, in the context of depression-where an antidepressant was evaluated for depression-related indications and/or depressive symptoms were the primary clinical focus-and in which dose variation was incorporated into the evidence synthesis. DATA EXTRACTION AND SYNTHESIS: We extracted data on dose-handling strategies, analysis methods, statistical frameworks and reporting quality, and synthesised these narratively to characterise methodological practice across studies. RESULTS: Twenty studies met inclusion criteria, evaluating pharmacologic interventions (n=11), non-pharmacologic interventions (n=7) or both (n=2). Dose handling varied substantially: in primary analyses, 50% treated dosages as separate nodes (split approach), 32% lumped doses together and only 15% used model-based methods incorporating dose-response relationships. Nine studies employed the R package MBNMAdose to conduct a model-based NMA for secondary analyses, primarily examining exercise interventions with varied dose-response functions. Reporting of priors, consistency assessments and justifications for dose categorisation or operationalisation was inconsistent. CONCLUSIONS: Current depression NMAs demonstrate broad variability in handling dose-related heterogeneity, with dose-response modelling often absent or inconsistently applied-particularly for behavioural interventions lacking standardised dose metrics. Greater methodological clarity, practical guidance and reporting standards are needed to integrate dose-response modelling into NMAs of depression treatments, where both pharmacologic and behavioural interventions pose distinct dosing challenges.