Clinical Trials: Phases, Endpoints, and What the Data Means

Introduction

Clinical trial data is the raw material of biotech valuation. Every value inflection point, every M&A decision, and every probability assumption in an rNPV model ultimately traces back to clinical trial results. Healthcare bankers do not need to be clinical scientists, but they must understand the structure of clinical development, the hierarchy of endpoints, the key statistical concepts that determine whether data supports approval, and how trial design choices affect timelines and costs. This article provides the complete framework that healthcare bankers use to evaluate clinical data and translate it into financial analysis.

Clinical development follows a phased structure mandated by the FDA. Each phase increases in size, cost, and rigor, progressively building the evidence package required for approval. The total development timeline from first-in-human dosing to approval averages 8-12 years, though expedited pathways can compress this significantly for drugs addressing serious conditions with unmet medical need.

Phase I: Safety and Dosing (6-12 months, $4-10M)

Phase I trials enroll 20-80 healthy volunteers (or patients, in oncology where healthy volunteer dosing is unethical) to establish the drug's safety profile, appropriate dosing range, and pharmacokinetic properties (how the body absorbs, distributes, metabolizes, and excretes the drug).

The primary objectives of Phase I are: identifying the maximum tolerated dose (MTD), characterizing side effects and their dose-dependence, understanding the drug's pharmacokinetic profile (half-life, bioavailability, metabolism pathway), and selecting the dose for Phase II testing.

Phase I trials rarely fail on safety alone: approximately 60-65% of drugs advance from Phase I to Phase II. Failures at this stage are typically due to unacceptable toxicity at doses needed for efficacy, unfavorable pharmacokinetic profiles (drug is metabolized too quickly, does not achieve adequate tissue concentrations), or unexpected drug interactions.

Maximum Tolerated Dose (MTD) and Recommended Phase II Dose (RP2D)

The MTD is the highest dose level at which a drug can be administered without causing unacceptable side effects (dose-limiting toxicities). The RP2D is the dose selected for Phase II testing, which may be at or below the MTD. Dose selection is critical because it determines the efficacy signal in Phase II: too low and the drug may appear ineffective; too high and toxicity may obscure the benefit. In oncology, where tolerability thresholds are higher due to disease severity, dose escalation studies are particularly important. Modern dose escalation uses Bayesian statistical methods (like the Bayesian Optimal Interval design) rather than the traditional 3+3 approach, allowing more efficient and safer dose finding.

Financial implications for Phase I. The low cost ($4-10 million) and relatively short timeline make Phase I a period of modest cash burn. For pre-revenue biotechs, Phase I is funded by venture capital or early public market proceeds. The primary financial risk is not the Phase I cost itself but the binary outcome: if Phase I reveals safety issues, the entire program's value (potentially hundreds of millions in future development costs avoided) is eliminated, and the company's stock can decline 30-50%.

Phase II: Efficacy Signal (1-3 years, $10-50M)

Phase II is the most important phase for healthcare bankers because it is where the fundamental question gets answered: does the drug work in patients? Phase II trials enroll 100-300 patients with the target disease and measure preliminary efficacy while continuing to monitor safety.

Phase II has the highest attrition rate of any phase: only 30-35% of drugs advance from Phase II to Phase III. This high failure rate reflects the biological reality that promising preclinical and Phase I data frequently does not translate to efficacy in larger patient populations. The reasons for Phase II failure include: insufficient efficacy (the drug works but not well enough to justify further development), safety signals that emerge in larger patient populations, inability to identify a responsive patient subgroup, and biomarker hypotheses that do not hold up in clinical testing.

Phase II is sometimes divided into Phase IIa (dose-finding and initial efficacy signal, often 50-100 patients) and Phase IIb (larger, more definitive efficacy assessment, often 200-400 patients with randomization against a comparator), though not all development programs make this distinction. Phase IIb trials that are randomized and well-powered can sometimes serve as the basis for accelerated approval, blurring the boundary between Phase II and Phase III.

Phase III: Confirmation (2-4 years, $50-150M+ per trial)

Phase III trials are large, randomized, controlled studies designed to provide definitive evidence of efficacy and safety for regulatory submission. They enroll 1,000-5,000+ patients across multiple clinical sites (often 100-300 sites across 20-40 countries for global programs), are typically double-blinded and controlled (patients are randomized to receive the drug or a comparator), and must hit pre-specified primary endpoints with statistical significance.

Phase III trials are the most expensive phase of clinical development, often costing $50-150 million per trial. Many programs require two Phase III trials for regulatory approval (the FDA's "substantial evidence" standard generally requires replication), doubling the cost and timeline. The Phase III success rate is approximately 55-65%, reflecting the stringent statistical requirements for approval.

The endpoint hierarchy determines what data the trial measures and which measurements the FDA uses to make approval decisions. The choice of endpoint has profound financial implications because it determines trial size, duration, and cost.

Endpoint Selection Strategy

The choice of primary endpoint involves strategic tradeoffs that directly affect the financial model:

Overall survival (OS) is the gold standard endpoint: it measures whether patients live longer. OS is unambiguous, universally accepted, and supports the strongest labeling claims. However, OS trials require very large patient populations (to detect statistically significant differences) and long follow-up periods (often 3-5+ years in oncology), making them the most expensive and time-consuming trial design. OS trials in adjuvant oncology settings can cost $200-400 million and take 5-7 years.

Progression-free survival (PFS) measures the time before disease worsens. PFS events occur more frequently than death, so PFS trials require fewer patients and shorter follow-up than OS trials, reducing cost and timeline. PFS is widely accepted as a primary endpoint in many oncology settings, though debate continues about whether PFS gains always translate to OS benefits.

Objective response rate (ORR) measures the percentage of patients whose tumors shrink by a defined amount. ORR trials can be single-arm (no comparator needed), require the fewest patients (often 100-200), and are the fastest to complete. ORR can support accelerated approval but typically requires a confirmatory trial with a harder endpoint.

Healthcare bankers need to understand three statistical concepts that determine whether clinical data is approvable:

P-Value

The probability that the observed treatment effect (or a more extreme effect) would occur by chance if the drug had no real benefit. In clinical trials, the conventional threshold for statistical significance is p < 0.05, meaning there is less than a 5% probability the observed result is due to chance. A p-value of 0.001 provides much stronger evidence than p = 0.04. When evaluating trial results, the first question is always: did the primary endpoint achieve p < 0.05? If not, the trial missed its primary endpoint regardless of how encouraging the numerical results may appear. Context matters: a p-value of 0.051 in a trial with a strong secondary endpoint package may still generate discussion at an FDA advisory committee, but it is formally a miss.

Hazard ratio (HR). Used in time-to-event analyses (survival, progression-free survival), the hazard ratio compares the rate of events (death, disease progression) in the treatment group versus the control group. HR < 1.0 means the treatment reduces the event rate. An HR of 0.65 means the treatment reduces the risk of the event by 35%. Lower HRs indicate stronger treatment effects. In oncology, an HR of 0.70-0.80 is considered clinically meaningful for most indications, while an HR below 0.60 is considered a strong result.

Confidence interval (CI). The range within which the true treatment effect likely falls. A 95% CI that does not cross 1.0 (for hazard ratios) or 0 (for absolute differences) indicates statistical significance. A narrow CI indicates precise estimation; a wide CI indicates uncertainty. The CI provides more information than the p-value alone because it shows both the magnitude and precision of the treatment effect. A drug with HR 0.70 (95% CI: 0.55-0.90) has a well-characterized benefit, while one with HR 0.70 (95% CI: 0.45-1.10) might not be statistically significant despite the same point estimate.

Traditional clinical trials follow a linear path (Phase I, then Phase II, then Phase III). Modern designs compress timelines and improve efficiency, which directly affects biotech cash burn, time to revenue, and valuation:

Adaptive designs allow modifications to the trial (sample size, dose selection, patient population, randomization ratio) based on interim data, without compromising the trial's statistical integrity. Adaptive designs can reduce development time and cost by eliminating unpromising doses or populations earlier. A dose-adaptive trial might start with three dose levels, drop the lowest dose at an interim analysis based on insufficient efficacy, and reallocate patients to the remaining doses, completing enrollment faster with a higher probability of demonstrating efficacy at the optimal dose.

Basket trials test a single drug across multiple cancer types that share a common biomarker or genetic mutation (e.g., testing a BRAF inhibitor in melanoma, colorectal cancer, thyroid cancer, and non-small cell lung cancer simultaneously). This allows simultaneous evaluation in multiple indications, accelerating the identification of responsive patient populations and potentially supporting multiple approval indications from a single trial.

Umbrella trials test multiple drugs against different biomarker-defined subgroups within a single disease. All patients with lung cancer, for example, are screened for multiple biomarkers (EGFR, ALK, ROS1, KRAS, PD-L1) and assigned to the appropriate treatment arm based on their tumor's molecular profile. Umbrella trials create efficiency by sharing a common screening infrastructure across multiple experimental treatments.

Seamless Phase II/III designs combine Phase II and Phase III into a single trial, using an adaptive framework where Phase II patients are enrolled first, an interim analysis determines whether to proceed, and additional patients are enrolled to complete the Phase III portion. If the interim analysis is positive, Phase II patients can be included in the Phase III analysis, saving 12-18 months vs. sequential Phase II and Phase III. This design is increasingly common for breakthrough therapy designated programs where speed is critical.

The next article covers surrogate endpoints and accelerated approval, which create a unique valuation dynamic where drugs can generate revenue while confirmatory evidence is still being collected.