SCOPE CLARIFICATION BEFORE NUMERICAL SYNTHESIS — 2026-11-07 12:05 UTC The primary original contribution will be a complete source-level audit of the seven adjusted cohorts included in Majed et al., medRxiv v1 (2026-11-05), not a new comprehensive systematic review of all moderate-AS literature. This fixed external census avoids selecting favorable cohorts, adds an independently checkable estimand audit, and avoids duplicating a newly available extensive meta-analysis. The broad Europe PMC search and title inspection are contextual discovery, not a completed independent systematic review; retain its records privately and report that limit. Additional sources (including prognostic, post-progression, mixed-valve and concomitant-operation studies) provide context rather than entering the seven-cohort denominator. No new pooled estimate or weight has yet been computed. Procedure-time coding will distinguish: explicit echocardiographic moderate stage immediately before treatment (Hariri, Adrichem); early window with moderate-at-entry and incompletely reported procedure-time reclassification (Moon, Lim; preserve Moon's stated moderate-stage intention); mixed procedure-time severity/progression (Samad, Angellotti); CT-defined pseudo-severe low-gradient phenotype (Ludwig). The broader reading of Moon's declared moderate-stage treatment will be a classification sensitivity. These are audit categories, not accusations of errors or evidence that unknown patients were severe. Analyze reported adjusted mortality inputs as a descriptive replication, then remove mixed-severity cohorts and pseudo-severe cohorts, and report the two explicit-moderate cohorts separately. Hariri and Adrichem are geographically distinct, all-EF native moderate-AS cohorts; their pooled association is a descriptive sensitivity despite different procedure mix, follow-up and time-zero handling. It is not a causal estimate or proof that early intervention while moderate improves survival. Provide common-effect and REML with modified Hartung-Knapp intervals, alternative adjustment choices and leave-one-out. Audit the 'AVR versus no AVR' generalization against each control strategy; delayed intervention in Adrichem is a concrete counterexample, not proof that confounding has been solved. Trial endpoints remain separate; mortality arithmetic using PROGRESS inputs that are available only in the preprint must be labeled secondary-source replication rather than independent primary verification.