Why registered trials terminate

Sample: 8,000 terminated interventional studies from ClinicalTrials.gov, 30 October 2026 (capped at 8 pages). Registry denominators: 34,631 terminated, 16,941 withdrawn, 333,500 completed, 65,221 recruiting.

By phase (top reasons)

• EARLY_PHASE1 (n=105): other/unspecified 37%, enrollment 28%, efficacy 13%
• NA (n=3242): other/unspecified 42%, enrollment 33%, efficacy 9%
• PHASE1 (n=922): other/unspecified 41%, enrollment 17%, business 13%
• PHASE1/PHASE2 (n=426): other/unspecified 34%, enrollment 25%, efficacy 14%
• PHASE2 (n=1732): enrollment 33%, other/unspecified 32%, efficacy 16%
• PHASE2/PHASE3 (n=151): other/unspecified 40%, efficacy 19%, enrollment 19%
• PHASE3 (n=819): other/unspecified 40%, efficacy 22%, enrollment 22%
• PHASE4 (n=603): enrollment 41%, other/unspecified 38%, efficacy 11%

By sponsor class

• FED (n=59): other/unspecified 44%, enrollment 20%, efficacy 14%
• INDIV (n=12): efficacy 50%, other/unspecified 25%, enrollment 25%
• INDUSTRY (n=2746): other/unspecified 39%, enrollment 17%, efficacy 15%
• NETWORK (n=97): enrollment 44%, other/unspecified 38%, efficacy 7%
• NIH (n=231): other/unspecified 55%, enrollment 24%, business 10%
• OTHER (n=4760): enrollment 37%, other/unspecified 37%, efficacy 12%
• OTHER_GOV (n=94): other/unspecified 45%, enrollment 34%, efficacy 9%
• UNKNOWN (n=1): safety 100%

Method note

whyStopped is free text, so reasons are assigned by keyword regex into six categories plus 'other'. Categories can overlap in the source text; the first match wins, in the order efficacy, safety, enrollment, funding, business, covid. This is a coarse but reproducible classification.

Limits

• 8-page cap: the sample is terminated studies in registry order, not a random sample.
• Free-text reasons are inconsistently written; 'other/unspecified' is large.
• Withdrawn studies are excluded here (separate status), so pre-start cancellations are missed.
• Registration practice varies by sponsor and country.