Methodology · v3.2

CTTTransparence des essais cliniques

Clinical trial reporting TRANSPARENCY for {legal_name} (NOT pipeline size — a focused small sponsor that publishes its results on time is fully trustworthy). EXTRACTION RULES: Query clinicaltrials.gov for trials with sponsor={legal_name} (device makers: also EU CTR / clinical investigations under MDR/IVDR), then check EU Clinical Trials Register + AllTrials.net for the same set. Return type="composite_signal" with sub-signals: (a) PRIMARY — results_published_pct — % of COMPLETED trials with results posted within 12mo per FDAAA (target ≥90), bucket <40→25, 40-65→50, 65-85→70, 85-95→85, 95+→95; this is the transparency/compliance signal and MUST drive the score. (b) CONTEXT-ONLY (non-punitive, scale not trustworthiness) — ongoing_trials_count — total active trials; report it but a LOW count is NOT a transparency weakness: if results_published_pct is high, a small pipeline still scores high. Only emit a low band when low count co-occurs with poor reporting. (c) CONTEXT-ONLY (non-punitive) — avg_enrollment_size — median across active trials; report for context, never lower the transparency score on small enrollment alone. ZERO-IS-ABSENT / ANTI-FLOOR (CLAUDE.md #4/#5): DISTINGUISH "sponsor discloses and reporting is genuinely weak" (a real clinicaltrials.gov record showing completed trials with results overdue → score the floor) from "sponsor is not found as a registered trial sponsor / has no completed trials" (absence of data, NOT a measured weakness). If {legal_name} does not appear as a sponsor of any registered trial, OR has no completed trial whose results window has elapsed, return type="absent" reason="no_company_specific_data" for the WHOLE index — do NOT infer 0% / count 0 / a low band from "not found", and never fill floor sub-signals just to satisfy a minimum count. Cite clinicaltrials.gov sponsor URL + AllTrials report. Server applies composite_signal_score.