Introduction
The probability of success (PoS) at each clinical phase is the most important input in an rNPV model. It determines how much of a projected cash flow stream is probability-weighted value and how much is aspiration. Getting the PoS assumption wrong by even 10 percentage points can swing a pipeline asset's value by 50-100%, which is why it is the first number a buyer, investor or board will challenge. Healthcare bankers start from historical success rate benchmarks, then adjust for therapeutic area and program-specific factors.
Overall Success Rates by Phase
The most widely cited benchmark, the BIO, Informa Pharma Intelligence and QLS Advisors study of 2011-2020 clinical development, analyzed 12,728 phase transitions and found that only 7.9% of drugs entering Phase I went on to FDA approval.
| Phase Transition | Transition Success Rate | Cumulative PoS (from Phase I) |
|---|---|---|
| Phase I to Phase II | 52.0% | 52.0% |
| Phase II to Phase III | 28.9% | 15.0% |
| Phase III to NDA/BLA filing | 57.8% | 8.7% |
| NDA/BLA to Approval | 90.6% | 7.9% |
Transition Rates vs. Likelihood of Approval
Each row in the table is a transition probability: the chance of clearing that one stage. A valuation needs the probability of reaching the market from where the asset sits today.
- Likelihood of Approval (LOA)
The probability that a drug candidate at a given stage eventually wins regulatory approval, calculated by multiplying the success rates of every remaining phase transition. On 2011-2020 data, LOA is about 7.9% from Phase I, 15.1% from Phase II, 52.4% from Phase III and 90.6% from NDA/BLA submission. It is the probability applied to post-approval revenue in an rNPV model.
A common modeling error is to apply a single phase transition rate to launch revenue. Weighting a Phase II asset's sales by the 28.9% chance of reaching Phase III ignores the Phase III and regulatory risk still ahead and overstates probability-weighted revenue by almost 2x. The costs follow different logic: each future trial's spend is weighted by the probability of reaching that trial, not by the LOA, because a program that fails in Phase II never pays for Phase III.
Why Published Benchmarks Disagree
Not every source lands on 7.9%, which is why the overall figure is usually quoted as a range of roughly 8-14%. BIO's 2006-2015 study put Phase I LOA near 10%, with Phase I and Phase II transition rates of about 63% and 30%. An MIT analysis by Wong, Siah and Lo in Biostatistics studied 406,038 trial entries covering more than 21,000 compounds from 2000 to 2015 and found that 13.8% of development programs reached approval.
The gap is mostly methodology. The BIO studies multiply the success rates of observed phase transitions, while the MIT team traced each drug's full development path and imputed missing trials. Neither is wrong, but a model should name its source and stay consistent: mixing transition rates from one dataset with LOAs from another produces probabilities that do not reconcile.
Success Rates by Therapeutic Area
Probability of success varies dramatically by therapeutic area, reflecting differences in disease biology, endpoint clarity and how high the regulatory and competitive bar sits. The figures below come from the 2011-2020 BIO dataset; the rare disease row excludes oncology indications.
| Therapeutic Area | LOA from Phase I | Phase II to III | Key Factors |
|---|---|---|---|
| Hematology | 23.9% | 48.1% | Well-defined targets, measurable blood endpoints |
| Rare Disease (non-oncology) | 17.0% | 44.6% | Genetic drivers, small focused trials, FDA flexibility |
| Metabolic | 15.5% | 45.0% | Established biomarkers (HbA1c, LDL-C) |
| Infectious Disease | 13.2% | 38.4% | Objective endpoints (viral load, cure rates) |
| Autoimmune | 10.7% | 31.4% | Validated pathways, rising competitive bar |
| Psychiatry | 7.3% | 26.8% | Subjective rating scales, high placebo response |
| Neurology | 5.9% | 26.8% | Heterogeneous diseases, slow progression |
| Oncology | 5.3% | 24.6% | Tumor heterogeneity, high bar versus standard of care |
| Cardiovascular | 4.8% | 21.0% | Large, long outcomes trials against cheap generics |
| All Indications | 7.9% | 28.9% | Benchmark average |
The spread is the point: hematology programs reach approval about five times as often as cardiovascular programs, and the single biggest gap opens up in Phase II. Rare disease benefits from the same dynamic, which is a large part of why orphan drug biotechs command premium valuations; chronic, high-prevalence diseases came in at just 5.9% in the same study.
Oncology Is Not One Number
Oncology accounts for a third of all transitions in the BIO dataset, so its low LOA drags down the industry average, and the MIT study measured it at just 3.4%. Within oncology, the sub-category matters as much as the therapeutic area: drugs for solid tumors reached approval 4.6% of the time from Phase I, versus 7.5% for hematologic cancers and 12.4% for immuno-oncology programs. A Phase I solid tumor asset and a Phase I immuno-oncology asset can carry probabilities more than 2x apart, even though both sit in the same therapeutic area.
Program-Specific Adjustments
Within a therapeutic area, several factors can increase or decrease the probability of success relative to historical averages.
Biomarker Selection
- Biomarker-Selected Clinical Trial
A trial that enrolls only patients whose tumors (or disease) express a specific biomarker or genetic mutation that the drug targets. Biomarker selection enriches the trial population for patients most likely to respond, increasing the observed treatment effect and the probability of the trial succeeding. Programs using biomarker-based patient selection have roughly twice the likelihood of approval of unselected programs.
In the BIO data, programs using patient preselection biomarkers reached approval 15.9% of the time from Phase I, against 7.6% for those without, and nearly the whole gap opens in Phase II, where 46.3% of biomarker programs advanced versus 28.3% of the rest. The effect is sharpest in oncology: the MIT study found that oncology trials using biomarkers for patient selection succeeded 10.7% of the time, against 1.6% for those without. This is one of the key drivers of the nichebuster strategy in pharma, where companies target smaller but better-defined patient populations.
Mechanism of Action Validation
If a drug targets a mechanism that has been validated by other approved drugs (e.g., a new PD-1 inhibitor when several are already approved), the probability of success is higher than for first-in-class mechanisms with no prior validation. The logic is straightforward: if the mechanism works for one drug, it is more likely to work for similar drugs. Immuno-oncology shows the effect clearly: programs against the PD-1/PD-L1 axis reached approval 23.9% of the time from Phase I in the 2011-2020 data, against 1.9% for other immunotherapy approaches. Conversely, if multiple drugs targeting the same mechanism have failed (e.g., amyloid-beta in Alzheimer's prior to aducanumab, or CETP inhibitors in cardiovascular disease), the probability is lower because the failures suggest the mechanism itself may be flawed.
Prior Clinical Data Quality
For assets entering later phases, the quality of earlier-phase data affects PoS. A Phase III program with robust Phase II data (large effect size, narrow confidence intervals, consistent across subgroups) has a higher Phase III PoS than one entering Phase III based on marginal Phase II data. Analysts adjust PoS upward for programs with strong Phase II signals and downward for programs advancing on borderline data. These adjustments are analyst judgment rather than published benchmarks, so they should be tied to specific evidence in the data and shown as a sensitivity range around the therapeutic area baseline.
Endpoint Selection
The choice of clinical endpoint also affects PoS. Trials using surrogate endpoints (biomarkers that correlate with clinical benefit, such as tumor response rates in oncology or HbA1c in diabetes) tend to have higher success rates than trials using hard clinical endpoints (overall survival, major adverse cardiovascular events). This is partly because surrogate endpoints are more sensitive to treatment effects and require smaller patient populations and shorter trial durations. However, success on a surrogate does not guarantee ultimate regulatory success. Since the Consolidated Appropriations Act, 2023, the FDA can require confirmatory trials to be underway before granting accelerated approval, and its draft guidance on when a confirmatory trial is underway expects enrollment to have started. See surrogate endpoints and confirmatory trials for how that withdrawal risk feeds into valuation.
The next article covers pipeline SOTP and peak sales estimation, which brings together PoS data with market sizing to produce the aggregate biotech valuation.
