Exploratory question recorded 2026-11-11T08:52:05.310102+00:00, after inspection of all input counts and submission of the primary study (entry 1237). Not outcome-blind or preregistered. Do not alter that exact signed bundle/archive. Question: Do unique-patient marginal counts in FAME-3 Table S9 constrain the unreported death-or-spontaneous-MI union without access to individual overlap? Use D=death, M=all MI, T=stroke, S=spontaneous MI subset of M, U=D union M union T. Table expressly labels number of patients and crude intention-to-treat percentages. Derive cardinality inequalities; use integer programming with anonymous 4-bit cells to establish feasible endpoints and each intermediate integer for these observed inputs. Independently enumerate all populations through N=6 to check formula on small feasible examples. Retain endpoint witnesses, no fabricated patient-level clinical dataset. Optional hypothetical independent complete-follow-up binomial confidence envelope must be labelled conditional, not an interval for actual time-to-event risks. State that actual censoring and stable-specific missingness are unavailable. Extreme global-loss sensitivity is permissive, not a missingness model. Do not claim clinical significance, mortality equivalence, or a new patient-level clinical effect. This was developed as a local exploratory supplement; it is now packaged separately for verification.