CLAIMBOUND / MEDICAL SOURCE BOUNDARY

What the “27%” Leqembi headline actually measures

This is a source-boundary audit, not medical advice, an individual benefit prediction, or a benefit-risk recommendation.

Claim transformation

CDR-SB mean change at 18 months → relative arithmetic 0.45 / 1.66 = 27.1% → headline shorthand “slows decline by about 27%”. It does not automatically support “memory improves by 27%”, “27% recovered”, “reverses”, or “cures”.

Primary-paper boundary

MED-LEQ-01-D1201 through 04-D1204: BLOCKED_SOURCE. The frozen NEJM URL returned HTTP 403; no easier source was substituted.

FDA-supported boundaries

D1205–D1207 and D1209–D1212: PASSED_UNDER_PROTOCOL for their narrow, qualifier-preserving statements.

Overclaim boundary

D1208: INSUFFICIENT_COVERAGE. Mere absence of stronger public promises was not relabelled as a negative result.

Baseline and R1 reruns preserve access state, hashes, and source drift. See the full campaign record.