The Anatomy of Hiddenness: A Multi-Axis Audit of Structural Non-Disclosure in ClinicalTrials.gov

Abstract

Most systematic reviews treat ClinicalTrials.gov as a near-complete census of registered trials. However, a large number of registered studies never post results and never link to a publication. Using a single 29 March 2026 full-registry snapshot — 578,109 records, of which 441,191 were interventional and 290,524 were closed interventional — we used 249,507 eligible older closed interventional studies as the analytic core and audited structural non-disclosure along six axes. Two-year no-results rates ran from 67.0% in oncology to 76.2% in metabolic medicine; 39–43% of studies in each major disease family showed neither posted results nor a linked publication. The missingness is structured: it tracks phase (76.7% for phase I vs 45.5% for phase III), architecture (one-arm 72.8% vs ≥10-arm 55.3%), and outcome density (100% no-results when zero primary outcomes are recorded). Geographic footprint matters too (single-site 79.5% vs ≥20-site 31.7%), as does completion speed (same-year 85.7% vs 6–10-year lag 57.6%). It is concentrated — the top 1% of lead sponsors hold 39.6% of the backlog (Gini 0.818). A strict NCT-indexed PubMed audit rescued only 1.2% of no-link studies, but broadening to Europe PMC lifted weighted any-match visibility to 40.8% — meaning much, though not all, of the apparent silence is an indexing-and-linkage gap rather than true absence. Hiddenness is a predictable, measurable property of the registry, and should be priced into downstream synthesis.