Data and analysis populations

Finalize data and derive populations

Analysis populations can overlap. Define assignment-based, safety and per-protocol populations without treating them as disjoint partitions. Excluding nonadherent participants can introduce selection and does not by itself estimate a causal per-protocol effect.

Flow diagram

Data and analysis populationsF1.S1Review data quality and queriesF1.S2Review deviations before unblindingF1.S3Derive endpoints and covariatesF1.S4Specify analysis populationsF1.S5Finalize and release datasetsF1.N1Analysis population definitionsPopulationEligibilityruleAssignment/receipt basisDeviationsEndpoint useCountF1.N2Dataset lineageDatasetSourceversionDerivationChecksOwnerReleaseF1.S0 Start flowData and analysis populationsF1.S999 End flowEnd flow

Steps

Review data quality and queries

Reconcile source data and document adjudication.

Review deviations before unblinding

Prespecify handling and decision authority.

Derive endpoints and covariates

Version all transformations and analysis programs.

Specify analysis populations

Define assignment-based, safety and per-protocol populations with explicit rules. These populations may overlap. Excluding nonadherent participants alone does not identify a causal per-protocol effect.

Finalize and release datasets

Document locking, authorized access and retained versions.

Data and analysis populations

End flow

Tables and supporting information

Analysis population definitions

Analysis population definitionsPopulationEligibilityruleAssignment/receipt basisDeviationsEndpoint useCount

Overlapping populations are not an arithmetic partition.

Dataset lineage

Dataset lineageDatasetSourceversionDerivationChecksOwnerRelease

Keep analysis inputs reproducible.

Sources