Participant-data synthesis specification

Model participants within study structures

Specify study structure in both one-stage and two-stage approaches. The two-stage equation is illustrative; model choice, interactions and missingness need their own specifications.

Flow diagram

Participant-data synthesis specificationF1.S1Define participant-level estimandsF1.S2Choose one-stage or two-stageapproachF1.S3Separate within-study andbetween-study effectsF1.S4Specify missingness and aggregateintegrationF1.S5Plan estimation and outputsF1.N1IPD model specificationQuestionStageapproachStudy effectsInteractionMissingnessAggregatedataMain effectParticipantinteractionSensitivityF1.N2Assumptions and rationaleF1.N3Method and protocol evidenceF1.N4Two-stage synthesis notationEstimate beta_j within study j, then combinecompatible effects with model-specificweights w_j. Preserve study structure andspecify uncertainty; this does not define aone-stage model.F1.S0 Start flowParticipant-data synthesisspecificationF1.S999 End flowEnd flow

Steps

Define participant-level estimands

Specify outcomes, treatment comparisons and covariate interactions.

Choose one-stage or two-stage approach

Justify study clustering, heterogeneity and model structure.

Separate within-study and between-study effects

Preserve randomization and avoid ecological interpretations of participant interactions.

Specify missingness and aggregate integration

Explain whether aggregate-only studies contribute and under which assumptions.

Plan estimation and outputs

Define uncertainty, model diagnostics and study-specific contributions.

Participant-data synthesis specification

End flow

Tables and supporting information

IPD model specification

IPD model specificationQuestionStageapproachStudy effectsInteractionMissingnessAggregatedataMain effectParticipantinteractionSensitivity

Specify models without automatically pooling participant rows.

Assumptions and rationale

Treating all participants as one unclustered study loses study structure and can change the estimand.

Method and protocol evidence

Attach the protocol, methodological sources and versioned software specification supporting these decisions. Record citations and provenance; do not enter participant identifiers.

Two-stage synthesis notation

Estimate beta_j within study j, then combine compatible effects with model-specific weights w_j. Preserve study structure and specify uncertainty; this does not define a one-stage model.

Sources