Adaptive final analysis specification

Account for adaptations in inference

Selection and stopping can affect estimation as well as testing. Prespecify compatible estimators and intervals, and report the rules actually applied rather than analyzing as though the design were fixed.

Flow diagram

Adaptive final analysis specificationF1.S1Define final estimands andpopulationsF1.S2Choose adaptation-aware inferenceF1.S3Account for selection and earlystoppingF1.S4Handle missingness and deviationsF1.S5Report decisions with resultsF1.N1Adaptive final inferenceEstimandAdaptationTest orestimatorInterval methodSensitivityPrimary effectSelectedcomparisonEarly stoppingF1.N2Assumptions and rationaleF1.N3Method and protocol evidenceF1.S0 Start flowAdaptive final analysis specificationF1.S999 End flowEnd flow

Steps

Define final estimands and populations

Preserve objectives through permitted adaptations and stopping events.

Choose adaptation-aware inference

Specify tests, estimators and intervals compatible with the design.

Account for selection and early stopping

Address changed sample sizes, selected arms and decision-dependent bias.

Handle missingness and deviations

Define sensitivity to timing, delayed outcomes and protocol departures.

Report decisions with results

Link estimates to the actual adaptation path and simulation evidence.

Adaptive final analysis specification

End flow

Tables and supporting information

Adaptive final inference

Adaptive final inferenceEstimandAdaptationTest orestimatorInterval methodSensitivityPrimary effectSelectedcomparisonEarly stopping

Do not assume a fixed-design analysis is valid after adaptation.

Assumptions and rationale

Inference must reflect the prespecified design and actual adaptations; simulation supports but does not replace a justified method.

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.

Sources