← allmeta

Component NMA

Additive model for multi-component interventions (Welton et al., Am J Epidemiol 2009). Each intervention is a set of components (e.g., "drug + exercise + education"). The random-effects model uses a generalized DerSimonian–Laird estimator and normal 95% confidence intervals. It estimates one effect per component; any combination's effect is the sum of its components.

Multi-arm correction off. Contrast rows are treated as independent, so a multi-arm trial entered as several rows double-counts its shared comparator (no τ²/2 off-diagonal). The bus importer refuses multi-arm trials; pasted rows must be independent two-arm contrasts. For multi-arm trials with the correct correlation, use NMA Pro.

Input

Select the component with no effect, or None if all components are active.

Component effects

ComponentEffectSE95% CI

Implied intervention effects vs control

InterventionEffect95% CI

Forest

Assumption: component effects are additive on the outcome scale you used. If interactions between components matter (synergy / antagonism), add interaction terms (e.g., include drug&exercise as a separate component) or use a full interaction model.