Auditing the public clinical-trials record: a registry-integrity atlas of registration, reporting and completion gaps

Abstract

Background. Prospective registration and results reporting underpin trustworthy evidence synthesis, yet registration quality, results-reporting compliance and completion status are themselves rarely audited at scale, and the gaps they leave are inherited silently by downstream meta-analyses.

Methods. We audited the ClinicalTrials.gov / AACT full-registry snapshot of 29 March 2026 (578,109 records; 441,191 interventional; 290,524 closed interventional), with a primary cohort of 249,507 eligible older closed interventional studies whose primary completion was at least two years before the snapshot. Across twenty-four pre-specified analyses in eight thematic families we computed two-year no-results rates, ghost-protocol rates (missing results plus missing publication links), hiddenness scores, full visibility, black-box stock and study-mix-adjusted excess, broken down by registration fields, trial architecture, reporting era, intervention type, enrollment, stopped-trial status, geography, outcome density, design purpose, completion delay, site footprint, sponsor class, disease family and named sponsor.

Results. Among closed interventional studies with primary completion at least two years earlier, 72.7 percent had no posted results, and recent eligible cohorts (2021–2024, 77.0 percent) were no cleaner than the FDAAA 801 era (2008–2016, 67.1 percent), with full visibility at 10.8 percent. Structural fields were strong opacity signals: missing actual enrollment 100.0 percent, one-arm studies 72.8 percent, blank primary purpose 86.4 percent and same-year submission-to-completion 85.7 percent. Opacity concentrated by geography and sponsor — single-site studies 79.5 percent versus 31.7 percent for studies with 20 or more sites; any-U.S. 52.1 percent versus 88.7 percent for no-U.S.; Egypt 95.8 percent among large named countries — and by disease family (cardiovascular 75.0 percent, metabolic 76.2 percent, oncology 67.0 percent). Study-mix-adjusted watchlists ranked France (2,187 excess no-results studies), oncology (543) and Assistance Publique – Hôpitaux de Paris (265) highest.

Conclusions. Audited end to end, the public clinical-trials record carries systematic, measurable and structurally patterned gaps that reviewers should screen for before pooling.