Protocol Ledger

WhyStopped Taxonomy & Accrual Forensics · Empirical Biostatistics Study

Phase II and Phase III Clinical Trial Terminations: Accrual Failure, Futility and Safety Signals in ClinicalTrials.gov Filings

Reconciling 54,890 interventional Phase II and Phase III protocols registered on clinicaltrials.gov from 2018 through 2026 identifies 6,796 prematurely stopped trials (12.38% overall; 30.22% of closed studies), where operational slow accrual (1,629 stops, 23.97%) outnumbers clinical safety (735 stops) and efficacy futility (416 stops) combined (1,151 stops).

Published: 2026-11-18 Publication ID: 47c4426df7e9d67d5bfe Primary Sources: clinicaltrials.gov & fda.gov

Operational Enrollment Collapse vs Biological Hypothesis Failure

Public discourse around biopharma R&D attrition routinely assumes that Phase II and Phase III clinical trials fail predominantly when an investigational molecule proves toxic or pharmacologically inactive in human subjects. Reconciling 54,890 interventional protocols across publication 47c4426df7e9d67d5bfe refutes that assumption at scale. When sponsors terminate, withdraw, or suspend a mid-to-late-stage protocol and file a narrative explanation inside the ClinicalTrials.gov WhyStopped module, patient recruitment failure emerges as the single largest classified cause of premature study closure.

Specifically, 1,629 Phase II and Phase III studies stopped explicitly due to insufficient or slow patient accrual (code WS1, 23.97% of all 6,796 stopped protocols). By contrast, safety, toxicity, or adverse-event terminations accounted for 735 protocols (WS4, 10.82%), and formal interim futility or lack-of-efficacy terminations accounted for just 416 protocols (WS5, 6.12%). Even when strategic portfolio reprioritizations (WS2, 1,001 protocols) and funding or operational disruptions (WS3, 918 protocols) are evaluated separately, operational feasibility rather than clinical biology terminates the plurality of closed protocols that fail to finish.

Table 1. Reconciled ClinicalTrials.gov Phase II and Phase III Termination Taxonomy (2018 to 2026 Cohort)
Code Structured WhyStopped Classification Stopped Trials Share of Stops CA Site Trials Median Enrolled Median Days Site Disclosure
WS1 Insufficient / Slow Accrual 1,629 23.97% 221 10.5 715 86.4%
WS2 Strategic / Business / Sponsor Decision 1,001 14.73% 372 34.0 720 84.2%
WS3 Funding / COVID / Operational 918 13.51% 100 14.0 677 92.6%
WS4 Safety / Toxicity / Adverse Events 735 10.82% 321 40.0 758 73.3%
WS5 Lack of Efficacy / Futility 416 6.12% 186 95.0 821 83.8%
WS6 Unspecified 2,097 30.86% 320 20.0 659 84.1%

Enrollment Depth and Time-to-Stop Signatures by Failure Mode

Inspecting median actual patient enrollment and elapsed calendar days prior to study closure reveals distinct operational signatures across the six termination classes in Table 1:

  • Accrual Failures Burn Two Years for Ten Patients: Protocols stopped for insufficient accrual (WS1) enrolled a median of just 10.5 actual participants despite remaining open for a median of 715 calendar days (nearly two full years of IRB renewals, site monitoring contracts, and investigational pharmacy overhead without generating statistically powered efficacy readouts).
  • Futility Stops Require Mature Interim Cohorts: Protocols halted by Data and Safety Monitoring Boards (DSMBs) for lack of efficacy or futility (WS5) reached a median actual enrollment of 95.0 participants across a median of 821 days, reflecting pre-specified Group Sequential or O'Brien-Fleming interim alpha-spending boundaries that require substantial event accumulation before declaring biological futility.
  • Safety Stops Occur Mid-Ramp With Lower Site Transparency: Toxicity and adverse-event terminations (WS4) closed at a median of 40.0 enrolled subjects and 758 days, while exhibiting the lowest physical site disclosure rate among all stop categories (73.3% vs 86.4% for slow-accrual stops), driven heavily by commercial multi-center sponsors using anonymized CRO site slots.

Correcting Right-Censoring Bias: Overall vs Closed-Cohort Termination Rates

Analysts benchmarking trial risk across calendar start years must distinguish between the overall cohort termination rate (12.38%, dividing 6,796 stopped protocols by all 54,890 protocols including 24,052 still-active studies) and the closed-trial termination rate (30.22%, dividing 6,796 stopped studies by the 22,486 protocols that have reached a terminal completed or stopped state). As Table 2 demonstrates, protocols initiated between 2018 and 2021 have largely matured, posting ~1,000 to 1,178 stops per start-year cohort against 2,495 to 2,836 completions, whereas 2024 through 2026 cohorts remain predominantly active in patient recruitment or follow-up.

Table 2. Annual Phase II and Phase III Protocol Vintages, Terminal Outcomes, Stop Causes, and FDA NDA/BLA Approvals
Start Year Protocols Phase II / III Active Load Completed Stopped Slow Accrual Safety + Futility FDA NDA/BLA
2018 5,337 3,385 / 1,952 513 2,836 998 269 173 152
2019 5,466 3,595 / 1,871 739 2,567 1,001 270 183 130
2020 6,143 3,996 / 2,147 926 2,668 1,165 318 188 130
2021 6,576 4,318 / 2,258 1,389 2,495 1,178 282 223 119
2022 5,909 3,937 / 1,972 1,669 1,980 921 202 154 123
2023 5,894 3,932 / 1,962 2,489 1,538 621 128 125 136
2024 6,161 4,173 / 1,988 3,976 1,071 419 73 53 134
2025 6,534 4,489 / 2,045 5,729 477 303 63 34 150
2026 6,870 4,715 / 2,155 6,622 58 190 24 18 110

Designing Pre-Activation Accrual Feasibility Gates

Preventing WS1 slow-accrual collapses requires auditing site-level competing protocols before Institutional Review Board (IRB) submission rather than waiting for six-month recruitment rescue amendments. When clinical operations teams model protocol eligibility criteria against electronic health record (EHR) cohorts and verify that selected academic medical centers are not already running five or more overlapping biomarker-matched trials in the same line of therapy, early zero-enrolling sites drop sharply. Equally important, building adaptive sample-size re-estimation and pre-specified DSMB futility interim checks ensures that genuinely inactive compounds stop cleanly under WS5 futility rules rather than continuing as prolonged low-accrual studies.

Methodology and Unspecified Narrative Caveats

Clinical operations teams inspecting sponsor risk inside the 3D Site Capacity & Execution Flow Map should note two statutory constraints on ClinicalTrials.gov WhyStopped text. First, free-text stop explanations are limited to 160 characters and are omitted or written generically (for example, "Sponsor decision" or left blank on withdrawn studies) in 2,097 stopped protocols (WS6, 30.86%). Second, sponsors occasionally record a compound business and clinical stop as a strategic reprioritization when a competing therapy shifts standard-of-care benchmarks mid-trial. Protocol Ledger applies deterministic lexical rules without speculative imputation, keeping unspecified closures explicitly auditable.