Whenever a promising clinical-stage biotech suddenly collapses 80% overnight on a Phase 2 or Phase 3 readout, retail investors and commentators invariably ask the same fundamental question: Are clinical trials failing more often today than they used to, or does it just feel that way because of 24-hour financial media?

Over the past five decades, the biopharmaceutical sector has witnessed an unprecedented technological revolution. We moved from phenotypic animal screening in the 1970s to recombinant DNA, high-throughput automated chemistry, robotic crystallography, human genome sequencing, cryo-electron microscopy, and de novo computational protein design.

Intuitively, one would expect that 50 years of exponential scientific tooling would make drug development vastly more reliable, driving clinical failure rates steadily downward.

The actual empirical data tells the exact opposite story.

Across the 50-year arc of systematic biopharmaceutical tracking (1970s to 2026), clinical trial failure rates have gone UP, and cumulative pipeline success rates have gone DOWN.

  • In the 1970s and 1980s, foundational tracking by the Tufts Center for the Study of Drug Development (CSDD) revealed that roughly 1 in 5 investigational new drugs entering Phase 1 human trials ultimately secured FDA approval—a cumulative success rate of ~20% to 23% (a ~77% to 80% failure rate).
  • By the 2000s and 2010s, landmark industry-wide analyses published in Nature Biotechnology, Biostatistics, and by the Biotechnology Innovation Organization (BIO) demonstrated that the overall likelihood of approval (LOA) from Phase 1 had dropped to ~9.6% to 13.8% (an 86% to 90.4% failure rate).
  • In the 2011–2020 benchmark cohort (BIO / Informa / QLS Advisors), aggregate clinical success fell to just 7.9%—a historic peak cumulative failure rate of 92.1%.
  • In the modern era, single-year composite success rates tracked by the IQVIA Institute hit an unprecedented trough of 5.0% in 2021 (a 95.0% failure rate) before recovering to 10.8% in 2023 and stabilizing near 11% to 12% in 2024–2026.
┌────────────────────────────────────────────────────────────────────────────────────────────────────────┐
│             CUMULATIVE DRUG ATTRITION FROM PHASE 1 TO FDA APPROVAL (50-YEAR HISTORICAL ARC)            │
├───────────────────────┬──────────────────────────────┬────────────────────────────┬────────────────────┤
│ Era & Tracking Study  │ Study Cohort Window          │ Cumulative Success (PoS)   │ Overall Fail Rate  │
├───────────────────────┼──────────────────────────────┼────────────────────────────┼────────────────────┤
│ Tufts CSDD (DiMasi)   │ 1970–1982 IND Filings        │ ~20.0% – 23.0%             │ 77.0% – 80.0%      │
│ Tufts CSDD (DiMasi)   │ 1981–1992 IND Filings        │ ~16.0% – 19.0%             │ 81.0% – 84.0%      │
│ Merck / Nat Rev Drug  │ 1991–2000 Industry Portfolios│ ~11.0%                     │ 89.0%              │
│ Tufts CSDD (DiMasi)   │ 1993–2004 IND Filings        │ ~16.0% (Self) / 19% (All)  │ 81.0% – 84.0%      │
│ Nature Biotech (Hay)  │ 2003–2011 Industry Pipeline  │ 10.4% (All) / 15.3% (Lead) │ 89.6% (All)        │
│ BIO / Biomedtracker   │ 2006–2015 Industry Pipeline  │ 9.6% (All) / 11.9% (ex-Onc)│ 90.4% (All)        │
│ MIT / Wong, Siah, Lo  │ 2000–2015 Citeline Dataset   │ 13.8% (Path) / 6.9% (Phase)│ 86.2% / 93.1%      │
│ BIO / Informa / QLS   │ 2011–2020 Industry Pipeline  │ 7.9% (All Indications)     │ 92.1% (All)        │
│ IQVIA Institute       │ 2021 Global Pipeline Nadir   │ 5.0% (Single-Year Trough)  │ 95.0% (Trough)     │
│ IQVIA Institute       │ 2023–2025 Pipeline Rebound   │ 10.8% – 11.7 (CPPI Index)  │ 88.3% – 89.2%      │
└───────────────────────┴──────────────────────────────┴────────────────────────────┴────────────────────┘

Why did failure rates climb despite revolutionary scientific progress? Where along the developmental pipeline do experimental medicines actually meet their demise? And why is modern drug development bifurcating into two radically different realities—one where trials almost always fail, and another where precision medicine has tripled success rates?

In this deep forensic research analysis, we unpack the history, quantitative datasets, biological hurdles, and economic incentives governing 50 years of clinical trial attrition.

1. The Genesis of Clinical Tracking (1962–1976)

To understand where failure rates come from, one must understand how the clinical trial pipeline was invented.

The Pre-1962 Landscape: Safety Only

Before 1962, drug regulation in the United States was governed by the Federal Food, Drug, and Cosmetic Act of 1938. Under that framework, pharmaceutical sponsors only had to demonstrate that a compound was non-toxic and safe prior to commercial sale. There was no statutory requirement to prove therapeutic efficacy. Clinical testing was largely informal, fragmented, and lacked uniform developmental gates.

The 1962 Kefauver-Harris Drug Amendments

The turning point in pharmaceutical history occurred in the early 1960s with the worldwide thalidomide tragedy, where a sedative prescribed for pregnancy morning sickness caused severe phocomelia (limb malformations) in thousands of infants across Europe and the UK. Although FDA medical officer Frances Kelsey famously blocked U.S. approval due to inadequate safety documentation, the crisis catalyzed immediate legislative action.

In October 1962, President John F. Kennedy signed the Kefauver-Harris Drug Amendments. The amendments fundamentally transformed medical research by mandating three requirements:

  1. Sponsors must file a formal Investigational New Drug (IND) application before administering an investigational agent to human subjects.
  2. Sponsors must obtain documented informed consent from all clinical trial participants.
  3. Crucially, sponsors were required to provide “substantial evidence” of efficacy through “adequate and well-controlled investigations, by experts qualified by scientific training and experience.”

To operationalize these statutory mandates, the FDA structured human experimentation into the tripartite clinical framework that remains the global gold standard today:

  • Phase 1: Initial human pharmacology, safety, tolerability, pharmacokinetics, and dose escalation (20–100 subjects).
  • Phase 2: Preliminary proof-of-concept (PoC), initial efficacy signals, and dose-ranging in target patient populations (100–300 patients).
  • Phase 3: Large-scale, randomized, double-blind, multicenter confirmatory trials against placebo or standard of care (1,000–5,000+ patients).
  • Regulatory Review (NDA / BLA): Comprehensive dossier submission to the FDA or European Medicines Agency (EMA) for commercial market clearance.
┌────────────────────────────────────────────────────────────────────────────────────────────────────────┐
│                               THE TRIPARTITE CLINICAL REGULATORY GATEWAY                               │
├─────────────────┬──────────────────────┬────────────────────────┬──────────────────────────────────────┤
│ Development Gate│ Typical Cohort Size  │ Primary Endpoint Focus │ Dominant Historical Failure Driver    │
├─────────────────┼──────────────────────┼────────────────────────┼──────────────────────────────────────┤
│ Phase 1         │ 20 – 100 Subjects    │ Safety, PK/PD, MTD     │ Unacceptable toxicity, poor PK profile│
│ Phase 2         │ 100 – 300 Patients   │ Proof-of-Concept, Dose │ Lack of clinical efficacy (~50-60%)  │
│ Phase 3         │ 1,000 – 5,000+ Pts   │ Confirmatory Pivotal   │ Failure to beat SOC / Placebo (~50%) │
│ NDA / BLA Review│ Complete Pipeline    │ Benefit/Risk, CMC      │ Manufacturing (CMC), data integrity   │
└─────────────────┴──────────────────────┴────────────────────────┴──────────────────────────────────────┘

The Inception of Attrition Accounting (1975–1976)

Once clinical trials were codified into discrete regulatory milestones, economists and pharmacologists began measuring pipeline dropouts:

  • Wardell & Lasagna (1975): In their landmark monograph Regulation and Drug Development (American Enterprise Institute), William M. Wardell and Louis Lasagna quantified the emergence of the “drug lag” and documented pipeline attrition rates under the post-1962 regime.
  • Tufts CSDD (Founded 1976): Lasagna and Wardell founded the Center for the Study of Drug Development (CSDD) at the University of Rochester in 1976, which subsequently relocated to Tufts University in 1984. Led by health economist Joseph A. DiMasi, Tufts CSDD created the first longitudinal IND tracking databases, tracking cohorts of new molecular entities (NMEs) from 1963 onward.
  • Commercial Databases (1980s–2000s): Asset-level tracking expanded with commercial intelligence engines:
    • Pharmaprojects (founded 1980 by Citeline / Informa)
    • CMR International (founded 1981 by the ABPI, now Clarivate)
    • BioMedTracker (Informa)
    • ClinicalTrials.gov (mandated under the FDA Modernization Act of 1997, launched in 2000, and expanded with mandatory results tracking under FDAAA Section 801 in 2007).

2. The Phase-by-Phase Anatomy of Failure

Clinical trial attrition is highly asymmetric. Each clinical phase presents an entirely different biological, pharmacological, and operational filter.

Phase 1: The Safety & Pharmacokinetic Screen

  • Typical Transition Success Rate: ~52% to 66% (Failure Rate: 34% to 48%)
  • Core Objective: Phase 1 trials evaluate how a drug is absorbed, distributed, metabolized, and excreted (ADME/PK), alongside identifying maximum tolerated doses (MTD) and dose-limiting toxicities (DLTs).
  • Primary Failure Modes: Acute organ toxicity (hepatotoxicity, cardiotoxicity, neurotoxicity), intolerable adverse events, severe off-target binding, or poor pharmacokinetic properties (half-life too short to sustain therapeutic exposure, insolubility, or extensive first-pass hepatic clearance).

Phase 2: The “Valley of Death”

  • Typical Transition Success Rate: ~24% to 35% (Failure Rate: 65% to 76%)
  • Core Objective: Phase 2 is the first time an experimental molecule is administered to patients suffering from the actual target disease. It is designed to establish biological proof-of-concept (PoC), confirm target engagement, and determine the optimal dose-response curve (Phase 2a vs. Phase 2b).
  • Primary Failure Modes: Phase 2 is universally recognized as the single largest bottleneck in the pharmaceutical industry. More than 70% of all Phase 2 trials fail. The overwhelming culprit is lack of clinical efficacy—accounting for 50% to 60% of all Phase 2 terminations. In vitro assays and animal disease models routinely mislead researchers; when exposed to the full heterogeneity of human pathophysiology, the drug simply fails to alter the disease trajectory.

Phase 3: The Confirmatory Catastrophe

  • Typical Transition Success Rate: ~50% to 65% (Failure Rate: 35% to 50%)
  • Core Objective: Phase 3 trials are large, randomized, double-blind, multicenter trials designed to definitively prove statistical superiority or non-inferiority against an existing Standard of Care (SOC) or placebo across diverse patient demographics.
  • Primary Failure Modes: Despite having passed Phase 1 safety and Phase 2 proof-of-concept, roughly 40% to 50% of Phase 3 trials fail. The primary causes are:
    1. Failure on Primary Efficacy Endpoints (~50–55% of failures): The effect size observed in a tightly controlled Phase 2 trial evaporates in large, heterogeneous real-world patient cohorts, or the placebo arm performs far better than anticipated (especially prominent in CNS and psychiatric trials).
    2. Emergent Safety Hurdles (~17–20% of failures): Idiosyncratic toxicities (e.g., drug-induced liver injury, thrombotic microangiopathy, QT prolongation) that occur in 1 out of 1,000 patients will never appear in a 100-patient Phase 2 trial, but become glaringly obvious in a 3,000-patient Phase 3 study.
    3. Commercial Obsolescence (~15–22% of failures): Because Phase 3 trials take 3 to 5 years to complete, a competitor drug frequently reaches the market first or establishes a higher efficacy benchmark, rendering the sponsor’s candidate commercially unviable.

Regulatory Review (NDA / BLA): The Final Gate

  • Typical Clearance Success Rate: ~85% to 92% (Failure Rate: 8% to 15%)
  • Core Objective: Submission of comprehensive clinical, non-clinical, and Chemistry, Manufacturing, and Controls (CMC) dossiers for formal approval.
  • Primary Failure Modes: Complete Response Letters (CRLs) issued by the FDA rarely stem from a complete lack of efficacy if Phase 3 was positive. Instead, CRLs are predominantly triggered by CMC / manufacturing facility violations (contamination, data integrity, batch-to-batch inconsistency), questionable post-hoc subgroup statistical analyses, or labeling impasses regarding black-box warnings.
┌────────────────────────────────────────────────────────────────────────────────────────────────────────┐
│                         WHY DO CLINICAL DRUGS FAIL? (HARRISON, NAT REV DRUG DISCOV)                    │
├───────────────────────────────┬────────────────────────────────────┬───────────────────────────────────┤
│ Failure Category              │ Phase 2 Terminations (% Share)     │ Phase 3 Terminations (% Share)    │
├───────────────────────────────┼────────────────────────────────────┼───────────────────────────────────┤
│ Lack of Clinical Efficacy     │ 52%                                │ 54%                               │
│ Safety & Toxicity             │ 24%                                │ 17%                               │
│ Commercial & Strategic Triage │ 15%                                │ 22%                               │
│ Operational / PK / Formulation│ 9%                                 │ 7%                                │
└───────────────────────────────┴────────────────────────────────────┴───────────────────────────────────┘

3. The 50-Year Longitudinal Master Matrix

To definitively evaluate whether failure rates have increased or decreased, we must examine the multi-decade empirical record across all major benchmark studies:

┌────────────────────────────────────────────────────────────────────────────────────────────────────────┐
│                       50 YEARS OF CLINICAL TRANSITION DATA: THE DEFINITIVE MATRIX                      │
├──────────────────────────────┬────────────┬─────────────┬─────────────┬─────────────┬──────────────────┤
│ Study & Primary Authors      │ Time Span  │ Phase 1 → 2 │ Phase 2 → 3 │ Phase 3 →Reg│ Overall PoS (P1) │
├──────────────────────────────┼────────────┼─────────────┼─────────────┼─────────────┼──────────────────┤
│ DiMasi et al. (1991, 1995)   │ 1970–1982  │ ~75.0%      │ ~50.0%      │ ~65.0%      │ ~20.0% – 23.0%   │
│ DiMasi (2001)                │ 1981–1992  │ ~71.0%      │ ~44.0%      │ ~65.0%      │ ~16.0% – 19.0%   │
│ Kola & Landis (Merck, 2004)  │ 1991–2000  │ ~65.0%      │ ~35.0%      │ ~55.0%      │ ~11.0%           │
│ DiMasi et al. (2010)         │ 1993–2004  │ 71.0%       │ 45.0%       │ 60.0%       │ 16.0% (19% All)  │
│ Hay et al. (Nat Biotech 2014)│ 2003–2011  │ 64.5%       │ 32.4%       │ 60.1%       │ 10.4%            │
│ Thomas et al. (BIO 2016)     │ 2006–2015  │ 63.2%       │ 30.7%       │ 58.1%       │ 9.6%             │
│ Wong, Siah, Lo (MIT 2019)    │ 2000–2015  │ 66.4%       │ 58.3%*      │ 59.0%       │ 13.8% (6.9% P2P) │
│ Thomas et al. (BIO 2021)     │ 2011–2020  │ 52.0%       │ 28.9%       │ 57.8%       │ 7.9%             │
│ IQVIA Institute (2021–2025)  │ 2021–2024  │ Variable    │ Variable    │ Rebound     │ 5.0% → 10.8%     │
└──────────────────────────────┴────────────┴─────────────┴─────────────┴─────────────┴──────────────────┘

The mathematical progression is undeniable:

  • In the 1970s and 1980s, overall clinical approval success sat comfortably between 16% and 23% (meaning failure was 77% to 84%).
  • By the 2000s, success dropped to 10% to 14% (failure reached 86% to 90%).
  • In the 2010s, comprehensive tracking by BIO across 12,728 phase transitions demonstrated that overall success bottomed at 7.9% (failure reached 92.1%).
  • Phase 2 transition success suffered the most violent collapse, falling from ~50% in the 1970s to 28.9% in the 2011–2020 decade.

4. Deconstructing Eroom’s Law: Why Did Failure Rates Rise?

In 2012, Jack W. Scannell and colleagues published a landmark analysis in Nature Reviews Drug Discovery titled “Diagnosing the decline in pharmaceutical R&D efficiency.” They observed that despite monumental leaps in technology, the number of new drugs approved per billion dollars spent on R&D had halved roughly every nine years since 1950. They termed this Eroom’s Law (“Moore’s Law” spelled backwards).

The steady upward climb in clinical trial failure rates is the central statistical engine driving Eroom’s Law. Four structural dynamics explain why failure rates rose over the last 50 years:

┌────────────────────────────────────────────────────────────────────────────────────────────────────────┐
│                               THE FOUR PILLARS OF RISING ATTRITION                                     │
├───────────────────────────────┬────────────────────────────────────────────────────────────────────────┤
│ The "Better Than the Beatles" │ Off-patent, cheap generic blockbusters raise the clinical bar. New     │
│ Problem                       │ drugs must beat entrenched standards of care, not just placebo.        │
├───────────────────────────────┼────────────────────────────────────────────────────────────────────────┤
│ The Biological Complexity     │ The "low-hanging fruit" is gone. R&D shifted to refractory polygenic   │
│ Horizon                       │ diseases: Alzheimer's (>99% fail), solid tumors, autoimmune networks.  │
├───────────────────────────────┼────────────────────────────────────────────────────────────────────────┤
│ The "Fail Fast, Fail Cheap"   │ Big Pharma intentionally spikes Phase 2 attrition via ruthless go/no-go│
│ Strategic Doctrine            │ gates (e.g., AstraZeneca 5R) to protect against $500M+ Phase 3 crashes.│
├───────────────────────────────┼────────────────────────────────────────────────────────────────────────┤
│ Regulatory Ratcheting & Hard  │ Regulators demanded hard clinical outcomes (survival, CVOTs) rather   │
│ Clinical Endpoints            │ than easy surrogate biomarkers (blood pressure, cholesterol).          │
└───────────────────────────────┴────────────────────────────────────────────────────────────────────────┘

1. The “Better than the Beatles” Problem

Imagine being a pop musician today, but copyright law decreed that every song ever written by the Beatles, Queen, or Michael Jackson was permanently free and universally broadcast. To sell a record, you would have to prove that your new song is definitively better than “Hey Jude.”

That is the exact commercial and clinical reality of modern biopharma:

  • In the 1970s and 1980s, medical practice lacked effective treatments for widespread primary conditions. Developing a first-generation beta-blocker, ACE inhibitor, H2 receptor antagonist, or simple antibiotic required proving modest clinical benefit against a placebo.
  • Today, those first- and second-generation miracle drugs are off-patent, universally available generics costing pennies per pill: atorvastatin, amlodipine, omeprazole, metformin, sertraline, lisinopril.
  • A novel drug candidate can no longer be approved merely by proving it works better than an inert sugar pill. It must demonstrate statistically significant non-inferiority or outright superiority against an entrenched, near-perfect standard of care in massive, multi-year clinical trials.

2. The Inexorable Push into Complex Biology

In the first three decades following 1962, biopharma harvested the “low-hanging fruit” of human biology: simple receptor antagonists, enzyme inhibitors, and well-understood monogenic pathways.

Once those targets were exhausted, capital flowed into the most recalcitrant, multifactorial biological territories:

  • Neurodegeneration (Alzheimer’s, Parkinson’s, ALS): Between 1998 and 2021, more than 400 clinical trials for Alzheimer’s disease were conducted. The cumulative failure rate was an astonishing 99.6%. The brain’s blood-brain barrier, extreme cellular heterogeneity, and irreversible synaptic loss before symptom onset created an empirical graveyard for biopharma.
  • Oncology Clonal Resistance: Cancer is not a single disease, but an evolving darwinian ecosystem. Cytotoxic chemotherapies and single-target small molecules routinely induce drug resistance via secondary mutations, target downregulation, and immunosuppressive tumor microenvironments. Unstratified oncology trials routinely post >95% failure rates.

3. The “Fail Fast, Fail Cheap” Paradox

One of the most counterintuitive discoveries in pharmaceutical economics is that higher early-stage failure rates can actually indicate superior corporate strategy.

In the 1990s, pharmaceutical companies routinely pushed questionable molecules out of Phase 2 into Phase 3, hoping that larger sample sizes would rescue borderline efficacy signals. The result was financially catastrophic: a string of massive Phase 3 failures that cost billions of dollars and destroyed enterprise value.

Recognizing this trap, major pharmaceutical firms completely overhauled their internal governance:

  • AstraZeneca’s “5R Framework” (Morgan et al., 2018): In 2011, AstraZeneca established the 5R framework: Right target, Right tissue, Right safety, Right patient, Right commercial potential. They mandated that a compound must hit uncompromising target-engagement and biomarker benchmarks before clearing Phase 2.
  • By instituting a “truth-seeking” corporate culture that actively killed unviable candidates early, AstraZeneca increased its Phase 2 failure rate, but successfully quadrupled its pipeline conversion from candidate nomination to Phase 3 completion from 4% to 19%.
  • In short: Part of the historical increase in Phase 2 failure rates was deliberate. The industry learned that it is far better to kill a drug for $30 million in Phase 2 than to bury it for $500 million in Phase 3.

4. Regulatory Ratcheting and Hard Endpoints

Over the decades, regulatory agencies progressively raised the evidentiary bar.

  • Following the 2007 controversy over rosiglitazone (Avandia) and potential cardiovascular risks, the FDA issued landmark guidance in 2008 mandating that every novel type 2 diabetes drug undergo massive Cardiovascular Outcomes Trials (CVOTs) enrolling 10,000+ patients to prove cardiovascular safety.
  • Regulators increasingly rejected surrogate endpoints (e.g., biomarker lowering, radiographic progression-free survival) in favor of “hard” clinical outcomes (overall survival, hospitalization reduction, all-cause mortality). Hard endpoints require larger trials, longer follow-up periods, and create more opportunities for statistical failure.

5. The Great Bifurcation: Where Failure Rates Are Collapsing

While the macro industry baseline shows failure rates worsening from ~80% to ~92%, examining aggregate numbers alone misses the most important story in modern medicine: the divergence between broad all-comer trials and precision medicine.

Modern clinical development has split into two parallel realities:

                                  ALL CLINICAL TRIALS
                                          │
                 ┌────────────────────────┴────────────────────────┐
                 ▼                                                 ▼
     BROAD / UNSTRATIFIED PATH                         PRECISION MEDICINE PATH
  ───────────────────────────────                   ─────────────────────────────
  • Solid Tumors (All-Comers): ~96% Fail            • Biomarker-Enriched: ~74% Fail
  • CNS / Alzheimer's: ~95-99% Fail                 • Monogenic Rare Diseases: ~75% Fail
  • Broad Metabolic Megatrials                      • Targeted Monoclonal Antibodies
                 │                                                 │
                 ▼                                                 ▼
      CUMULATIVE SUCCESS: 3 - 5%                       CUMULATIVE SUCCESS: 20 - 26%

1. The Biomarker Revolution

The single most powerful lever for reducing clinical failure rates is the integration of patient-selection biomarkers. When a trial prospectively identifies which patients express the exact biological target of the drug, the signal-to-noise ratio improves dramatically:

  • Thomas et al. (BIO 2016 Report):
    • Overall LOA with selection biomarkers: 25.9% (Failure rate: 74.1%)
    • Overall LOA without selection biomarkers: 8.4% (Failure rate: 91.6%)
    • Impact: Selection biomarkers deliver a 3.1-fold increase in clinical success.
  • Wong, Siah, & Lo (MIT 2019 Study):
    • Oncology trials without biomarkers: 1.6% cumulative success (98.4% failure rate)
    • Oncology trials with selection biomarkers: 10.7% cumulative success (89.3% failure rate)
    • Impact: In cancer—the highest-attrition therapeutic area in biopharma—biomarkers reduce the failure barrier by nearly seven-fold.

2. The Rare Disease & Orphan Drug Paradigm

Rare diseases represent another arena where failure rates have defied the broader industry decline:

  • In the BIO 2011–2020 dataset, rare disease programs achieved an overall approval success rate of 17.0% (a 73% failure rate), compared to just 5.9% (a 94.1% failure rate) for non-rare chronic conditions.
  • Why Rare Diseases Succeed: Most rare diseases are severe monogenic disorders (e.g., cystic fibrosis, spinal muscular atrophy, sickle cell disease). The biological link between the genetic lesion and human pathology is unambiguous, preclinical validation is highly translatable, and the FDA provides accelerated approval pathways with smaller patient cohorts.

3. Modality Disparities: Biologics vs. Small Molecules

Throughout the 2000s and 2010s, large-molecule biologics (monoclonal antibodies, recombinant enzymes, fusion proteins) consistently outperformed traditional small molecules:

  • DiMasi et al. (2010): Clinical approval success for recombinant proteins was 32.0% vs. 13.0% for small-molecule chemical entities.
  • Thomas et al. (BIO 2021): Biologics achieved an overall approval rate of 9.1% vs. 5.7% for small-molecule NMEs.
  • Mechanistic Rationale: Small molecules are prone to off-target promiscuity, unexpected CYP450 drug-drug interactions, and idiosyncratic reactive metabolite toxicity. Monoclonal antibodies, by contrast, bind with exquisite nanomolar specificity to their extracellular epitopes, virtually eliminating off-target chemical toxicities.

6. Indications Benchmarked: Where Capital Goes to Die

To evaluate risk across therapeutic portfolios, we inspect the full phase-by-phase breakdown across disease classes from the BIO / Informa / QLS Advisors (2011–2020) dataset covering 12,728 clinical transitions:

┌────────────────────────────────────────────────────────────────────────────────────────────────────────┐
│                   CLINICAL SUCCESS & FAILURE RATES BY THERAPEUTIC AREA (BIO 2011–2020)                 │
├─────────────────────┬─────────────┬─────────────┬─────────────┬─────────────┬────────────┬─────────────┤
│ Therapeutic Class   │ Phase 1 → 2 │ Phase 2 → 3 │ Phase 3 →Reg│ Reg → Appr  │ Overall PoS│ Overall Fail│
├─────────────────────┼─────────────┼─────────────┼─────────────┼─────────────┼────────────┼─────────────┤
│ Hematology          │ 61.3%       │ 49.3%       │ 70.3%       │ 90.9%       │ 19.3%      │ 80.7%       │
│ Metabolic Diseases  │ 56.7%       │ 39.8%       │ 65.6%       │ 88.5%       │ 13.1%      │ 86.9%       │
│ Infectious Disease  │ 56.6%       │ 38.8%       │ 62.6%       │ 91.5%       │ 12.6%      │ 87.4%       │
│ Ophthalmology       │ 59.8%       │ 33.7%       │ 60.0%       │ 90.8%       │ 11.0%      │ 89.0%       │
│ Autoimmune / Allergy│ 55.6%       │ 30.6%       │ 63.8%       │ 94.6%       │ 10.3%      │ 89.7%       │
│ Cardiovascular      │ 59.7%       │ 24.3%       │ 55.4%       │ 88.9%       │ 7.1%       │ 92.9%       │
│ Neurology (CNS)     │ 54.7%       │ 27.6%       │ 53.6%       │ 86.8%       │ 7.0%       │ 93.0%       │
│ Oncology (All)      │ 48.8%       │ 24.6%       │ 47.7%       │ 92.5%       │ 5.3%       │ 94.7%       │
│ Urology             │ 45.4%       │ 18.2%       │ 50.0%       │ 87.5%       │ 3.6%       │ 96.4%       │
├─────────────────────┼─────────────┼─────────────┼─────────────┼─────────────┼────────────┼─────────────┤
│ INDUSTRY AVERAGE    │ 52.0%       │ 28.9%       │ 57.8%       │ 90.6%       │ 7.9%       │ 92.1%       │
└─────────────────────┴─────────────┴─────────────┴─────────────┴─────────────┴────────────┴─────────────┘

The clinical implications are stark:

  • Hematology is the premier performer in biopharma, boasting a 19.3% overall success rate. Hematologic malignancies and blood disorders allow direct sampling of circulating cells, immediate pharmacodynamic verification, and clear surrogate endpoints (complete remission, minimal residual disease).
  • Oncology is the industry’s largest recipient of venture capital and public equity, yet posts an overall success rate of just 5.3% (a 94.7% failure rate). Phase 2 is particularly brutal, with less than 1 in 4 cancer drugs advancing to Phase 3.
  • Neurology (CNS) suffers from a 93.0% failure rate, held back by high placebo response rates, subjective clinical endpoint scoring (e.g., cognitive scales, depression inventories), and blood-brain barrier impermeability.

7. The Contemporary Landscape (2020–2026): Pandemic Shock, Rebound, and AI Realism

The period between 2020 and 2026 witnessed extreme volatility in clinical productivity metrics:

The 2021 Nadir: The 5.0% Composite Low

According to the IQVIA Institute’s annual Global Trends in R&D reports, the composite clinical development success rate fell to a historic low of 5.0% in 2021. This crash was driven by a perfect storm of external and internal factors:

  1. Pandemic Disruptions: Clinical site staffing shortages, delayed patient enrollment, missed monitoring visits, and protocol deviations compromised hundreds of ongoing non-COVID trials.
  2. Oncology Trial Redundancy: A massive speculative bull market in 2020–2021 funded dozens of “me-too” biotech startups pursuing overlapping solid tumor targets (e.g., over 100 clinical trials testing indistinguishable PD-1/PD-L1 combinations or TIGIT antibodies). As these low-probability trials matured simultaneously, a wave of negative readouts flooded the sector.

The 2023–2025 Rebound

Following the 2021 trough, biopharmaceutical productivity began a meaningful recovery:

  • The IQVIA composite success rate rebounded to 10.8% in 2023—the highest level recorded since 2018.
  • IQVIA’s Clinical Program Productivity Index (CPPI)—which synthesizes phase success rates, trial duration, and protocol complexity—rose to 11.7 in 2024 (up from 10.9 in 2023), driven primarily by recovering Phase 3 success rates.
  • Operational Drivers: This rebound was catalyzed by post-pandemic operational adaptations:
    • Adaptive Trial Designs: Bayesian trial architectures that allow sponsors to drop non-performing arms early or adjust sample sizes mid-stream.
    • Decentralized and Digital Monitoring: Remote wearable telemetry, home nursing visits, and electronic patient-reported outcomes (ePRO) dramatically reduced patient dropout rates.
    • Pruning of Marginal Pipelines: Rising interest rates in 2022–2024 forced biotechs to ruthlessly cancel marginal Phase 1/2 programs, focusing capital only on their highest-conviction assets.

The AI Reality Check: Has Computational Biology Bended the Curve?

Between 2020 and 2026, vast sums were poured into artificial intelligence, machine learning, and generative structural biology platforms (AlphaFold, de novo generative chemistry).

Has AI lowered clinical trial failure rates?

The sober scientific consensus—articulated in critical audits by Jack Scannell, Andreas Bender, and industry analysts—is that AI has not yet measurably improved late-stage Phase 2 and Phase 3 clinical failure rates.

AI has successfully collapsed the timeline of preclinical hit-to-lead discovery, reducing candidate nomination from 4–5 years down to 12–18 months. However, once a computational molecule enters a Phase 2 trial in living human patients, it encounters the same intractable biological reality: If the underlying disease hypothesis is wrong, a drug designed by an algorithm fails just as quickly as a drug found by wet-lab screening.

Until AI models can accurately simulate systemic human organ interactions and multicellular disease pathways, Phase 2 efficacy attrition will remain biopharma’s dominant bottleneck.

8. Strategic Takeaways for Biotech Investors

For healthcare portfolio managers, biotech founders, and retail investors, the 50-year history of clinical attrition yields indispensable rules for underwriting pipeline risk:

  1. Phase 1 Safety is Table Stakes, Not a Thesis: Retail investors frequently celebrate “clean” Phase 1 safety data as a major de-risking catalyst. It is not. Historically, two-thirds of all drugs pass Phase 1. Phase 1 merely proves the drug doesn’t acutely poison healthy humans. It tells you almost nothing about whether it will cure the target disease.
  2. Never Underwrite Phase 2 Without a Biomarker: The base-rate failure in Phase 2 across all indications is 71%, and in oncology it exceeds 75%. Investing in an unstratified Phase 2 trial in solid tumors or CNS without a clear mechanistically validated biomarker is statistically akin to buying a lottery ticket.
  3. Beware the “Phase 2 Winner’s Curse”: Roughly 40% to 50% of Phase 3 trials fail. Many of these failures occur because sponsors over-interpret small, unblinded Phase 2 subgroup analyses or surrogate endpoints, only to watch the statistical significance evaporate when powered across 3,000 patients.
  4. Modality and Target Validation Trumps Technology Hype: Large-molecule biologics and monogenic rare disease therapies have consistently doubled or tripled the success rates of small molecules and polygenic chronic targets for over three decades.

9. Primary Literature & Data Citations

  1. DiMasi, J. A., Hansen, R. W., Grabowski, H. G., & Lasagna, L. (1991). Cost of innovation in the pharmaceutical industry. Journal of Health Economics, 10(2), 107–142. DOI: 10.1016/0167-6296(91)90001-4.
  2. DiMasi, J. A. (1995). Success rates for new drugs entering clinical testing in the United States. Clinical Pharmacology & Therapeutics, 58(1), 1–14. PMID: 7628176.
  3. DiMasi, J. A. (2001). Risks in new drug development: approval success rates for investigational drugs. Clinical Pharmacology & Therapeutics, 69(5), 297–307. PMID: 11371997.
  4. Kola, I., & Landis, J. (2004). Can the pharmaceutical industry reduce attrition rates? Nature Reviews Drug Discovery, 3(8), 711–715. PMID: 15286737.
  5. DiMasi, J. A., Feldman, L., Seibring, A., & Cafarella, A. (2010). Trends in risks associated with new drug development: success rates for investigational drugs. Clinical Pharmacology & Therapeutics, 87(3), 272–277. PMID: 20130567.
  6. Pammolli, F., Magazzini, L., & Riccaboni, M. (2011). The productivity crisis in pharmaceutical R&D. Nature Reviews Drug Discovery, 10(6), 428–438. PMID: 21629293.
  7. Scannell, J. W., Blanckley, A., Boldon, H., & Warrington, B. (2012). Diagnosing the decline in pharmaceutical R&D efficiency. Nature Reviews Drug Discovery, 11(3), 191–200. PMID: 22378269.
  8. Hay, M., Thomas, D. W., Craighead, J. L., Economides, C., & Rosenthal, J. (2014). Clinical development success rates for investigational drugs. Nature Biotechnology, 32(1), 40–51. PMID: 24406927.
  9. Thomas, D. W., Burns, J., Audette, J., Carroll, A., Dow-Hygelund, C., & Hay, M. (2016). Clinical Development Success Rates 2006–2015. Biotechnology Innovation Organization (BIO), Biomedtracker, Amplion. BIO Report Archive.
  10. Harrison, R. K. (2016). Phase II and phase III failures: 2013–2015. Nature Reviews Drug Discovery, 15(12), 817–818. PMID: 27811931.
  11. Morgan, P., Brown, D. G., Lennard, S., et al. (2018). Impact of a five-dimensional framework on R&D productivity at AstraZeneca. Nature Reviews Drug Discovery, 17(3), 167–181. PMID: 29348674.
  12. Wong, C. H., Siah, K. W., & Lo, A. W. (2019). Estimation of clinical trial success rates and related parameters. Biostatistics, 20(2), 273–286. Corrigendum in Biostatistics, 20(2), 366. PMID: 29394327 / PMC6409418.
  13. Dowden, H., & Munro, J. (2019). Trends in clinical success rates and therapeutic focus. Nature Reviews Drug Discovery, 18(7), 495–496. PMID: 31267073.
  14. Ringel, M. S., Scannell, J. W., Baedeker, M., & Schulze, U. (2020). Breaking Eroom’s Law. Nature Reviews Drug Discovery, 19(12), 833–834. PMID: 33177699.
  15. Thomas, D. W., et al. (2021). Clinical Development Success Rates and Contributing Factors 2011–2020. Biotechnology Innovation Organization (BIO), Informa Pharma Intelligence, QLS Advisors. BIO Report Archive.
  16. IQVIA Institute for Human Data Science (2022–2025). Global Trends in R&D: Overview through 2021, 2022, 2023, and 2024. IQVIA Institute Annual Reports. IQVIA Institute.