Clinical Trial Phases

Clinical trial phases are the sequential, legally regulated stages of human experimental research—spanning Phase 0, Phase I, Phase II, Phase III, and Phase IV—that evaluate the safety, tolerability, pharmacokinetics, dose–response relationships, and therapeutic efficacy of an investigational drug, biologic, or medical device. Initiated only after an Investigational New Drug (IND) application clears regulatory review based on GLP preclinical toxicology and safety pharmacology, clinical trials operate under the ethical mandates of the Declaration of Helsinki (1964), the Belmont Report (1979: respect for persons, beneficence, and justice), independent Institutional Review Boards (IRBs), and ICH E6 Good Clinical Practice (GCP) standards.

Across the pharmaceutical industry, clinical development is the longest, most capital-intensive, and highest-attrition segment of the drug discovery pipeline. Empirical analyses of over 9,000 clinical programs between 2011 and 2023 demonstrate that the overall Likelihood of Approval (LoA) for an investigational drug entering Phase I to ultimately receive New Drug Application (NDA) or Biologics License Application (BLA) approval averages 7.9% to 10.0% (ranging from ~5.3% in oncology to ~21%–25% in hematology and rare genetic diseases), spanning a median clinical timeline of 6 to 8 years.

Phase 0: Exploratory IND and Human Microdosing

Introduced under the 2006 FDA Exploratory IND Guidance, Phase 0 trials evaluate human ADMET and target engagement prior to traditional Phase I escalation using limited preclinical toxicology packages. A single sub-pharmacological microdose—defined as ≤100 μg of a small molecule (or ≤30 nmol for a protein biologic) and ≤1/100th of the NOAEL or predicted pharmacologically active dose—is administered to a small cohort of 5 to 15 healthy volunteers. Because plasma concentrations at microdoses fall in the femtomolar-to-picomolar range, Phase 0 trials pair ¹⁴C-radiolabeled drug administration with ultrasensitive Accelerator Mass Spectrometry (AMS) or liquid chromatography–tandem mass spectrometry (LC-MS/MS) to measure human clearance, oral bioavailability, and metabolic stability, or utilize Positron Emission Tomography (PET) with short-lived isotopes (¹¹C or ¹⁸F) to confirm blood–brain barrier penetration and target receptor occupancy.

Phase I: First-in-Human Safety, Tolerability, and Pharmacokinetics

Phase I trials represent the formal transition to pharmacologically active human dosing, enrolling 20 to 100 participants over 1 to 1.5 years with a historical 52% to 65% transition probability to Phase II. For non-cytotoxic therapies, Phase I studies are conducted in healthy volunteers at dedicated Clinical Pharmacology Units; however, for cytotoxic oncology agents, narrow-therapeutic-index immunosuppressants, and irreversible gene therapies, Phase I trials enroll patients with advanced, treatment-refractory disease for whom potential therapeutic benefit justifies the toxicity risk.

Standard Phase I programs execute two sequential core protocols alongside specialized clinical pharmacology sub-studies:

1. Single Ascending Dose (SAD) and Multiple Ascending Dose (MAD) Cohorts: Subjects are randomized in consecutive dose cohorts (typically 6 active : 2 placebo per cohort). In the first cohort, sentinel dosing administers the drug to 1 active and 1 placebo subject 24 to 48 hours ahead of the remaining cohort to guard against acute anaphylaxis or cytokine release (anchored to the NOAEL-derived Maximum Recommended Starting Dose [MRSD] or MABEL). If safety laboratory panels, telemetry ECGs, and adverse event monitoring remain acceptable, subsequent cohorts escalate the dose (often following a modified Fibonacci sequence: +100%, +67%, +50%, +40%, +33%) first as single doses (SAD) and then over 7 to 28 days of repeat dosing until steady state is achieved (MAD).
2. Clinical Pharmacology Sub-Studies: Phase I packages characterize the food effect (high-fat, high-calorie 800–1,000 kcal meal vs. fasted state on C_max and AUC), human mass-balance ADME (administering ~100 μCi of ¹⁴C-labeled drug to quantify total fecal and urinary recovery and identify circulating metabolites), Thorough QT (TQT) or concentration–QTc modeling, clinical drug-drug interaction studies, and hepatic/renal impairment cohorts.

Oncology Dose Escalation: 3+3, CRM, BOIN, and Project Optimus

In oncology Phase I trials, dose escalation traditionally seeks the Maximum Tolerated Dose (MTD) based on the incidence of Dose-Limiting Toxicities (DLTs)—defined by the National Cancer Institute Common Terminology Criteria for Adverse Events (NCI CTCAE) as Grade ≥3 non-hematologic or Grade 4 hematologic toxicities occurring during Cycle 1 (typically 21 or 28 days)—to select the Recommended Phase 2 Dose (RP2D).

• Rule-Based 3+3 Design: Enrolls cohorts of 3 patients. If 0/3 experience a DLT, the dose escalates; if 1/3 experiences a DLT, the cohort expands to 6 patients; if ≥2/6 experience a DLT, theMTD is exceeded and the previous lower dose is declared the MTD (targeting a DLT rate <33%).
• Model-Based (CRM) and Model-Assisted (BOIN) Designs: Because the 3+3 design is statistically inefficient and treats many patients at sub-therapeutic doses, modern trials use the Continual Reassessment Method (CRM) (which continuously updates a Bayesian parametric dose–toxicity curve after each cohort to converge on a target toxicity rate, e.g., 25%–30%) or the Bayesian Optimal Interval (BOIN) design (which compares the observed DLT rate at the current dose against prespecified optimal probability boundaries [λ_e, λ_d] with transparent decision rules).
• FDA Project Optimus: Because molecularly targeted kinase inhibitors, PROTACs, and monoclonal antibodies often achieve maximal target saturation well below their toxic MTD, the FDA Oncology Center of Excellence launched Project Optimus in 2021, requiring sponsors to randomize patients across two or more pharmacologically active doses in Phase I/II to select the optimal biological dose based on exposure–response, pharmacodynamics, and long-term tolerability rather than defaulting to the MTD.

Phase II: Proof-of-Concept and Dose-Ranging (The "Valley of Death")

Phase II trials enroll 100 to 300 patients diagnosed with the target disease over 2 to 3 years to test whether modulating the biological target translates into human clinical efficacy. Phase II exhibits the lowest transition success rate in pharmaceutical development (~28% to 35%), as approximately 50% of Phase II failures result from insufficient clinical efficacy and 25% from unexpected safety liabilities. Phase II is commonly divided into two sub-stages:

• Phase IIa (Exploratory Proof-of-Concept): Evaluates biological activity and early clinical response using pharmacodynamic biomarkers, imaging endpoints, or short-term surrogate markers. In single-arm oncology trials, Simon's Two-Stage Design (Optimal or Minimax) is widely employed to minimize the number of patients exposed to an ineffective drug: Stage 1 enrolls n₁ patients, and if r₁ or fewer objective tumor responses occur, the trial terminates early for futility; otherwise, an additional n₂ patients are enrolled.
• Phase IIb (Definitive Dose-Ranging): Randomized, double-blind, placebo- or active-controlled multi-arm trials (testing 3 to 4 dose levels or dosing regimens) analyzed via MCP-Mod (Multiple Comparison Procedure – Modelling) to characterize the E_max dose–response curve, define the minimum effective dose, and lock in the optimal dose for Phase III.

Phase III: Pivotal Confirmatory Trials and Master Protocols

Phase III trials are large-scale, multi-national, adequately powered randomized controlled trials enrolling 300 to 5,000+ patients (and up to 30,000–45,000 participants for preventive vaccines or cardiovascular outcome trials) over 2 to 4 years (~58%–65% transition probability to NDA/BLA filing). Under the 1962 Kefauver–Harris Amendments, the FDA typically requires two adequate and well-controlled Phase III trials (each achieving a two-sided p < 0.05 on a pre-specified primary clinical endpoint) to establish "substantial evidence" of effectiveness, or a single large, internally consistent pivotal trial with statistically persuasive evidence (p < 0.001) in high-unmet-need or orphan diseases.

To accelerate precision medicine, modern Phase II/III development increasingly utilizes Master Protocols (FDA 2022 Guidance):

• Basket Trials: Evaluate a single targeted therapy across multiple tumor histologies or diseases that share a common molecular biomarker (e.g., larotrectinib across NTRK gene-fusion-positive solid tumors regardless of tissue of origin).
• Umbrella Trials: Evaluate multiple targeted therapies within a single disease or tumor histology, stratifying patients into parallel sub-studies based on genomic profiling (e.g., Lung-MAP in non-small-cell lung cancer).
• Platform Trials: Multi-arm, multi-stage perpetual trial infrastructures using a shared master control arm where new experimental therapies enter and ineffective arms drop out dynamically via Bayesian adaptive randomization (e.g., I-SPY 2 in breast cancer and the RECOVERY trial in COVID-19).

Phase IV: Post-Marketing Surveillance and Pharmacovigilance

Following regulatory approval, Phase IV studies evaluate the drug across thousands to millions of diverse real-world patients (including pregnant individuals, elderly patients with comorbidities, and polypharmacy populations excluded from Phase III). Because a Phase III database of 3,000 patients has a 95% probability of detecting only adverse events occurring at an incidence of ≥1 in 1,000 (by the Rule of Three: 3 / N), Phase IV pharmacovigilance—utilizing the FDA Adverse Event Reporting System (FAERS), the Sentinel Initiative electronic health record network, and mandatory Phase IV post-marketing requirements (PMRs, such as cardiovascular outcome trials [CVOTs] for antidiabetic drugs)—is essential to detect rare (<1 in 10,000), delayed, or idiosyncratic adverse drug reactions.

What is the "Rule of Three" in clinical trial safety sample sizes?

In clinical biostatistics, based on the Poisson approximation to the binomial distribution, if zero events of a specific adverse reaction are observed in a sample of N treated patients, the upper 95% confidence bound on the true underlying incidence rate of that event is approximately 3 / N (since −ln(0.05) ≈ 3.00). Thus, to have a 95% probability of observing at least one case of a rare adverse event whose true incidence is 1 in 10,000 (0.01%), a clinical program must expose at least N = 30,000 patients—explaining why rare toxicities are typically discovered during Phase IV post-marketing surveillance.

Why did the FDA establish Project Optimus for oncology Phase I trials?

The classical oncology Phase I paradigm—escalating doses until the Maximum Tolerated Dose (MTD) is reached—was designed in the 1960s for non-selective cytotoxic chemotherapies, where efficacy and toxicity share the same DNA-damaging mechanism. Modern targeted therapies (kinase inhibitors, ADCs, bispecific antibodies) often saturate their molecular target at doses well below the MTD. Escalating to the MTD exposes patients to unnecessary chronic toxicities, frequent dose reductions, and treatment discontinuations without improving survival. Project Optimus mandates randomized dose-optimization comparisons before Phase III.

How does a Basket Trial differ from an Umbrella Trial?

Both are biomarker-driven Master Protocols, but they invert the drug-to-disease mapping. A Basket Trial tests 1 drug across multiple diseases (histologies) that share 1 common genetic mutation (e.g., a BRAF V600E inhibitor tested across melanoma, colorectal, thyroid, and hairy-cell leukemia baskets). An Umbrella Trial studies 1 single disease (e.g., triple-negative breast cancer), sequences every patient's tumor, and assigns patients to multiple different targeted drugs matched to their specific driver mutations.

1. Wong, C. H., Siah, K. W., & Lo, A. W. (2019). "Estimation of clinical trial success rates and related parameters." Biostatistics, 20(2), 273–286. https://doi.org/10.1093/biostatistics/kxx069
2. Simon, R. (1989). "Optimal two-stage designs for phase II clinical trials." Controlled Clinical Trials, 10(1), 1–10. https://doi.org/10.1016/0197-2456(89)90015-990015-9)
3. Woodcock, J., & LaVange, L. M. (2017). "Master protocols to study multiple therapies, multiple diseases, or both." New England Journal of Medicine, 377(1), 62–70. https://doi.org/10.1056/NEJMra1510062
4. Shah, M., Rahman, A., Theoret, M. R., & Pazdur, R. (2021). "The drug-dosing conundrum in oncology — When less is more." New England Journal of Medicine, 385(16), 1445–1447. https://doi.org/10.1056/NEJMp2109826