# Statistical Analysis Plan Shell — v0.8

**Status:** Must be completed and signed by a qualified statistician before independent extraction results are unmasked.

## 1. Analysis populations

- Product A direct randomized MDD stratum.
- Product A prespecified indirect strata.
- Product B monitored-exposure inventory strata.

## 2. Denominators

Define rules for randomized, dosed, safety, and observed-at-window denominators; participant-level first event; withdrawals; and missing follow-up.

## 3. Primary estimands and effect measures

- Product A: arm risk and risk difference for each compatible comparator/time-window class.
- Product B: descriptive participant risk within each design/population/ascertainment stratum.

## 4. Zero-event handling

Specify exact study-level intervals, rules for combined bounds, treatment of zero-zero trials, and prohibition of arbitrary continuity corrections.

## 5. Pooling decision

Predefine the minimum clinical and methodological compatibility required for pooling. State what will be reported when pooling is not justified.

## 6. Sparse-data model

If a pooled model is authorized, predefine the model, estimator, software, convergence checks, prior distributions if Bayesian, and sensitivity analyses. Do not select the model after viewing the independent results.

## 7. Missing data

Specify outcome nonreporting, incomplete follow-up, unavailable denominators, author contact, and best/worst plausible sensitivity analyses.

## 8. Multiplicity

Identify primary and secondary outcomes and how claims will be controlled or described without selective emphasis.

## 9. Subgroups and indirectness

MDD/TRD, comparator, sponsor, ascertainment intensity, exact/overlapping window, and single/repeated regimen remain prespecified strata. Avoid underpowered significance tests for subgroup differences.

## 10. Reproducibility

Specify software versions, package lock, random seeds where applicable, machine-readable input schema, and independent code review.

## Sign-off

Statistician: ____________________  Date: __________  Version/hash: ____________________