{
  "study": {
    "slug": "clinical-trial-discontinuation-2026",
    "title": "Where clinical trials stop: the Phase 2 valley, 2026",
    "standfirst": "Of the 373,998 clinical trials on ClinicalTrials.gov that have reached a final outcome, 51,959 — about 1 in 7 — ended in discontinuation rather than completion. Attrition is uneven: 23.0% of settled Phase 2 trials stop early, against 14.7% at Phase 1, the signature of the efficacy-stage valley where a candidate must first prove it works.",
    "desk": "care-quality",
    "article_type": "Original Research",
    "published": "2026-06-16",
    "issue": 78,
    "doi": "10.5072/fonteum/clinical-trial-discontinuation-2026",
    "url": "https://fonteum.com/research/clinical-trial-discontinuation-2026",
    "methodology_version": "clinicaltrials/v1"
  },
  "data_as_of": "2026-06-14",
  "datasets": [
    {
      "slug": "clinicaltrials-gov",
      "name": "ClinicalTrials.gov",
      "publisher": "NIH — National Library of Medicine",
      "upstream_url": null
    }
  ],
  "key_findings": [
    {
      "number": "13.9%",
      "finding": "of the 373,998 clinical trials that have reached a settled outcome ended in discontinuation — terminated, withdrawn, or suspended — rather than completion. That is 51,959 trials, about 1 in 7; across all 589,453 registered studies the discontinuation share is 8.8%",
      "dataset": "clinicaltrials-gov"
    },
    {
      "number": "23.0%",
      "finding": "of settled Phase 2 trials discontinued — the highest of any single phase — against 15.3% at Phase 3 and 14.7% at Phase 1. Combined Phase 1/Phase 2 designs are worse still at 27.2%. Attrition clusters at the efficacy stage, not the safety stage",
      "dataset": "clinicaltrials-gov"
    },
    {
      "number": "21.0%",
      "finding": "of phased interventional trials run by non-industry sponsors discontinued, against 15.8% for industry — the reverse of the all-trials headline (industry 15.3% vs 13.4% for everyone else). The flip is a phase-mix artifact: industry runs disproportionately many phased drug trials",
      "dataset": "clinicaltrials-gov"
    },
    {
      "number": "98.8%",
      "finding": "of the 16,441 withdrawn trials with a reported enrollment count enrolled zero participants — stopped before the first patient — while terminated trials posted a median enrollment of 21. Withdrawn and terminated are two structurally different kinds of stopping",
      "dataset": "clinicaltrials-gov"
    },
    {
      "number": "589,453",
      "finding": "registered studies across 50,756 distinct lead sponsors make up the snapshot, dated 2026-06-14 and pulled from the ClinicalTrials.gov Data API v2; 76.3% interventional, 23.3% observational. Every figure is a count over published records — no investigator, sponsor, or trial is named, ranked, or scored",
      "dataset": "clinicaltrials-gov"
    }
  ],
  "faqs": [
    {
      "q": "What does it mean for a clinical trial to be 'discontinued'?",
      "a": "It means the study stopped before completing. ClinicalTrials.gov records three discontinued states: terminated (the study halted after enrolling participants), withdrawn (it halted before enrolling its first participant), and suspended (it paused). Together these are 51,959 of the 373,998 trials that have reached a final outcome — 13.9%, about 1 in 7. Discontinuation is a study-lifecycle event, not a quality or conduct signal."
    },
    {
      "q": "Which phase of clinical trial fails most often?",
      "a": "Phase 2. Among settled trials, 23.0% of Phase 2 studies discontinued, the highest of any single phase, against 15.3% at Phase 3 and 14.7% at Phase 1. Combined Phase 1/Phase 2 designs are higher still at 27.2%. Phase 2 is the efficacy stage — the point where a candidate that already cleared initial safety must show it actually works — so attrition concentrates there rather than at the safety-focused Phase 1."
    },
    {
      "q": "Do industry-sponsored trials fail more often than academic ones?",
      "a": "It depends how you count. Across all trials, industry-sponsored studies discontinue slightly more often (15.3% vs 13.4% for all other sponsors). But that is a phase-mix effect — industry runs disproportionately many phased drug trials, which fail more. Hold phase constant and the order reverses: among phased interventional trials, non-industry sponsors discontinue more (21.0% vs 15.8%), and the gap holds inside Phase 2 (24.7% vs 20.6%)."
    },
    {
      "q": "What is the difference between a withdrawn and a terminated trial?",
      "a": "A withdrawn trial stops before enrolling its first participant; a terminated trial stops after enrolling some. The registry data confirm the distinction cleanly: 98.8% of the 16,441 withdrawn trials with a reported enrollment count enrolled zero participants, with a median enrollment of 0, while terminated trials had a median enrollment of 21. One is an abandonment on paper; the other is a study halted mid-stream."
    },
    {
      "q": "Does a discontinued trial mean someone did something wrong?",
      "a": "No. A discontinued status is a lifecycle record, not a disciplinary or quality signal. Trials stop for many reasons — slow recruitment, funding, a business decision, an interim look that changed the calculus, or a safety finding — and the registry status alone does not distinguish them. This study draws no inference about any investigator, sponsor, or provider from a trial's status."
    },
    {
      "q": "Why are so many registered trials still marked 'unknown' or ongoing?",
      "a": "ClinicalTrials.gov holds 589,453 registered studies, but only 373,998 have reached a final outcome. The rest are ongoing (recruiting, active, not yet recruiting) or carry the legacy UNKNOWN status — 93,459 studies whose sponsors stopped updating them past their expected completion date. This study computes discontinuation rates only over settled trials, so unsettled and never-updated records do not distort the denominator."
    },
    {
      "q": "Can I reproduce these figures?",
      "a": "Yes. Every number is a direct count over the public clinical_trials table — the ClinicalTrials.gov Data API v2 registry, snapshot 2026-06-14 — with no modeling. The exact SQL for the discontinuation rate, the phase breakdown, the sponsor-class split, the phase-held-constant reversal, and the withdrawn-versus-terminated enrollment profile is published in the reproducibility block below."
    }
  ],
  "citation": {
    "apa": "Fonteum Research. (2026, June 16). Where clinical trials stop: the Phase 2 valley, 2026. Fonteum Research, Issue 78. https://doi.org/10.5072/fonteum/clinical-trial-discontinuation-2026",
    "url": "https://fonteum.com/research/clinical-trial-discontinuation-2026"
  },
  "reproducible_sql": "-- How often clinical trials STOP before completing — and where the attrition\n-- concentrates. Fully reproducible query.\n--\n-- Question: of the studies registered on ClinicalTrials.gov, how many end in\n-- discontinuation (terminated, withdrawn, or suspended) rather than completion,\n-- and how does that rate vary by trial phase and by who sponsors the trial?\n-- The lead figure: of the 373,998 trials that have reached a settled outcome\n-- (completed OR stopped), 51,959 — 13.9%, about 1 in 7 — ended in\n-- discontinuation rather than completion. Discontinuation is a study-lifecycle\n-- signal, NOT a quality, fraud, or wrongdoing signal of any kind, and says\n-- nothing about any investigator, sponsor, or provider.\n--\n-- Source:\n--   public.clinical_trials — ClinicalTrials.gov Data API v2 (the U.S. National\n--     Library of Medicine registry of interventional and observational\n--     studies). 589,453 registered studies; snapshot 2026-06-14. Public,\n--     read-only. License: ClinicalTrials.gov Terms & Conditions (NLM public\n--     data). methodology_version = 'clinicaltrials/v1'.\n--\n-- Universe: this study reads the published registry AS A WHOLE — every row is a\n--   study CMS/NLM lists on ClinicalTrials.gov as of the 2026-06-14 snapshot.\n--   \"Settled\" = a study that has reached a final lifecycle outcome (COMPLETED,\n--   TERMINATED, WITHDRAWN, or SUSPENDED); ongoing statuses (RECRUITING,\n--   ACTIVE_NOT_RECRUITING, etc.) and the legacy UNKNOWN status are excluded from\n--   rate denominators because they have not yet settled.\n--\n-- Counting note: 589,453 distinct NCT IDs, one row each; no row is duplicated.\n--   No individual investigator, sponsor, or provider is named in the study.\n\n-- ============================================================================\n-- (1) Universe reconciliation — the registry at a glance.\n-- ============================================================================\nSELECT\n  count(*)                                                          AS studies,\n  count(DISTINCT nct_id)                                            AS distinct_nct,\n  count(DISTINCT lead_sponsor_name)                                 AS sponsors,\n  count(*) FILTER (WHERE study_type = 'INTERVENTIONAL')             AS interventional,\n  count(*) FILTER (WHERE study_type = 'OBSERVATIONAL')              AS observational,\n  max(ingested_at)::date                                            AS snapshot\nFROM public.clinical_trials;\n--  studies 589,453 · distinct_nct 589,453 · sponsors 50,756\n--  interventional 449,800 (76.3%) · observational 137,624 (23.3%) · snapshot 2026-06-14\n\n-- ============================================================================\n-- (2) HEADLINE: the discontinuation rate. Of all 589,453 registered studies,\n--     51,959 (8.8%) are discontinued. Restricted to the 373,998 that have\n--     SETTLED (reached a final outcome), the discontinuation rate is 13.9% —\n--     about 1 in 7. Discontinuation = TERMINATED (stopped after starting),\n--     WITHDRAWN (stopped before the first participant), or SUSPENDED (paused).\n-- ============================================================================\nSELECT\n  count(*)                                                          AS all_studies,\n  count(*) FILTER (WHERE overall_status IN\n    ('TERMINATED','WITHDRAWN','SUSPENDED'))                         AS discontinued,\n  round(100.0 * count(*) FILTER (WHERE overall_status IN\n    ('TERMINATED','WITHDRAWN','SUSPENDED')) / count(*), 1)          AS disc_pct_of_all,\n  count(*) FILTER (WHERE overall_status IN\n    ('COMPLETED','TERMINATED','WITHDRAWN','SUSPENDED'))             AS settled,\n  round(100.0 * count(*) FILTER (WHERE overall_status IN\n    ('TERMINATED','WITHDRAWN','SUSPENDED')) / count(*) FILTER\n    (WHERE overall_status IN\n    ('COMPLETED','TERMINATED','WITHDRAWN','SUSPENDED')), 1)         AS disc_pct_of_settled,\n  count(*) FILTER (WHERE overall_status = 'TERMINATED')             AS terminated,\n  count(*) FILTER (WHERE overall_status = 'WITHDRAWN')              AS withdrawn,\n  count(*) FILTER (WHERE overall_status = 'SUSPENDED')              AS suspended\nFROM public.clinical_trials;\n--  all_studies 589,453 · discontinued 51,959 · disc_pct_of_all 8.8%\n--  settled 373,998 · disc_pct_of_settled 13.9%\n--  terminated 33,776 · withdrawn 16,459 · suspended 1,724\n\n-- ============================================================================\n-- (3) WHERE attrition concentrates — discontinuation rate by trial phase,\n--     among settled trials. Phase 2 is the high-attrition phase (23.0%); the\n--     combined Phase 1/Phase 2 designs are worse (27.2%). Phase 3 (15.3%) and\n--     Phase 1 (14.7%) are markedly lower. This is the classic efficacy-stage\n--     \"valley\" — the point where a candidate must show it works.\n-- ============================================================================\nSELECT\n  coalesce(phase, '(none)')                                         AS phase,\n  count(*)                                                          AS settled,\n  count(*) FILTER (WHERE overall_status IN\n    ('TERMINATED','WITHDRAWN','SUSPENDED'))                         AS discontinued,\n  round(100.0 * count(*) FILTER (WHERE overall_status IN\n    ('TERMINATED','WITHDRAWN','SUSPENDED')) / count(*), 1)          AS disc_pct\nFROM public.clinical_trials\nWHERE overall_status IN ('COMPLETED','TERMINATED','WITHDRAWN','SUSPENDED')\nGROUP BY phase\nORDER BY settled DESC;\n--  NA            141,021 · 15,474 · 11.0%\n--  (none)         77,779 ·  7,663 ·  9.9%\n--  PHASE2         43,693 · 10,031 · 23.0%   <- the valley\n--  PHASE1         37,713 ·  5,526 · 14.7%\n--  PHASE3         30,539 ·  4,680 · 15.3%\n--  PHASE4         24,966 ·  4,067 · 16.3%\n--  PHASE1/PHASE2  10,141 ·  2,761 · 27.2%\n--  PHASE2/PHASE3   4,762 ·    986 · 20.7%\n--  EARLY_PHASE1    3,384 ·    771 · 22.8%\n\n-- ============================================================================\n-- (4) WHO stops — discontinuation rate by lead-sponsor class, among settled\n--     trials. Headline: industry-sponsored trials discontinue slightly more\n--     often than everyone else (15.3% vs 13.4%). But that is a phase-mix\n--     artifact: industry runs disproportionately many phased drug trials, which\n--     fail more. Hold phase constant (next query) and the order REVERSES.\n-- ============================================================================\nSELECT\n  coalesce(lead_sponsor_class, '(none)')                            AS sponsor_class,\n  count(*)                                                          AS settled,\n  count(*) FILTER (WHERE overall_status IN\n    ('TERMINATED','WITHDRAWN','SUSPENDED'))                         AS discontinued,\n  round(100.0 * count(*) FILTER (WHERE overall_status IN\n    ('TERMINATED','WITHDRAWN','SUSPENDED')) / count(*), 1)          AS disc_pct\nFROM public.clinical_trials\nWHERE overall_status IN ('COMPLETED','TERMINATED','WITHDRAWN','SUSPENDED')\nGROUP BY sponsor_class\nORDER BY settled DESC;\n--  OTHER     247,117 · 33,364 · 13.5%   INDUSTRY 101,598 · 15,559 · 15.3%\n--  NIH         9,753 ·  1,327 · 13.6%   OTHER_GOV  7,849 ·    633 ·  8.1%\n--  FED         3,776 ·    469 · 12.4%   NETWORK    3,456 ·    541 · 15.7%\n--  industry 15.3% vs all-other 13.4% overall (the phase-mix headline)\n\n-- ============================================================================\n-- (4b) The reversal, holding phase constant: among PHASED interventional\n--      trials (Early Phase 1 .. Phase 4), NON-industry sponsors discontinue\n--      MORE than industry (21.0% vs 15.8%) — and the gap holds inside Phase 2\n--      specifically (24.7% non-industry vs 20.6% industry). Better-resourced\n--      industry monitoring, not a quality verdict on academic science.\n-- ============================================================================\nWITH p AS (\n  SELECT\n    CASE WHEN lead_sponsor_class = 'INDUSTRY' THEN 'industry'\n         ELSE 'non-industry' END                                   AS grp,\n    phase, overall_status\n  FROM public.clinical_trials\n  WHERE overall_status IN ('COMPLETED','TERMINATED','WITHDRAWN','SUSPENDED')\n    AND phase IN ('EARLY_PHASE1','PHASE1','PHASE1/PHASE2','PHASE2',\n                  'PHASE2/PHASE3','PHASE3','PHASE4')\n)\nSELECT\n  grp,\n  count(*)                                                          AS settled,\n  count(*) FILTER (WHERE overall_status IN\n    ('TERMINATED','WITHDRAWN','SUSPENDED'))                         AS discontinued,\n  round(100.0 * count(*) FILTER (WHERE overall_status IN\n    ('TERMINATED','WITHDRAWN','SUSPENDED')) / count(*), 1)          AS disc_pct,\n  count(*) FILTER (WHERE phase = 'PHASE2')                          AS phase2_settled,\n  round(100.0 * count(*) FILTER (WHERE phase = 'PHASE2'\n    AND overall_status IN ('TERMINATED','WITHDRAWN','SUSPENDED'))\n    / nullif(count(*) FILTER (WHERE phase = 'PHASE2'), 0), 1)       AS phase2_disc_pct\nFROM p\nGROUP BY grp\nORDER BY settled DESC;\n--  non-industry  81,817 · 17,202 · 21.0% · phase2 25,509 · 24.7%\n--  industry      73,381 · 11,620 · 15.8% · phase2 18,184 · 20.6%\n\n-- ============================================================================\n-- (5) TWO kinds of stopping — withdrawn vs terminated by enrollment profile.\n--     A WITHDRAWN trial stops before enrolling its first participant; a\n--     TERMINATED trial stops after enrolling some. The data confirm the\n--     distinction: 98.8% of withdrawn trials enrolled zero participants\n--     (median 0), against a median of 21 for terminated trials.\n-- ============================================================================\nSELECT\n  overall_status,\n  count(*)                                                          AS trials,\n  count(*) FILTER (WHERE enrollment_count = 0)                      AS zero_enrolled,\n  round(100.0 * count(*) FILTER (WHERE enrollment_count = 0)\n    / count(*), 1)                                                  AS zero_enroll_pct,\n  percentile_cont(0.5) WITHIN GROUP (ORDER BY enrollment_count)     AS median_enrolled\nFROM public.clinical_trials\nWHERE overall_status IN ('WITHDRAWN','TERMINATED')\n  AND enrollment_count IS NOT NULL\nGROUP BY overall_status\nORDER BY trials DESC;\n--  TERMINATED 33,529 · zero 29 · 0.1% · median_enrolled 21\n--  WITHDRAWN  16,441 · zero 16,237 · 98.8% · median_enrolled 0",
  "license": "U.S. Government Works (federal sources; 17 U.S.C. §105)",
  "generated_by": "Fonteum — https://fonteum.com",
  "notes": "Aggregate, source-traced figures frozen to the snapshot above. Reproduce by running reproducible_sql against the cited federal dataset; no per-entity records are included."
}