Articles

Probability of Success by Clinical Phase: Benchmarks and Adjustments

The 2011-2020 success-rate benchmarks by phase, what they measure, and how to adjust them for a specific asset without double counting.

The probability of success is the single most consequential input in a drug asset valuation, and the one most often pulled from memory. A defensible number starts from published benchmarks, uses the right conditional for the asset's stage, and moves away from the benchmark only for documented reasons. This article sets out the most widely cited dataset, the BIO, Informa Pharma Intelligence and QLS Advisors study of clinical development success rates for 2011-2020, explains what its figures measure, and shows how to condition and adjust them using two assets of Veltrana Therapeutics, a fictional biotech used as a teaching case: VLT-201 in Phase 2b and VLT-305 in Phase 1.

Worked in full in The Biotech Valuation Handbook by Julian R. Sterling, with every figure reproduced in live Excel workbooks.Get the DRM-free PDF, $19.99 →

The benchmark dataset

The study, Clinical Development Success Rates and Contributing Factors 2011-2020, was published in February 2021. It covers 9,704 development programs run by 1,779 companies, with 12,728 phase transitions drawn from Biomedtracker data. Its value lies in size and transparency: the transitions are counted consistently, the method is published, and the headline figures are reported by phase, by disease group and by a few program characteristics. Other datasets exist, including academic work from MIT and proprietary vendor data, and a careful analyst cross-checks against them, but this report is the common starting point for a PoS memo.

For all indications combined, the phase transition rates are:

TransitionAll indicationsOncologyNon-oncology
Phase I to Phase II52.0%48.8%53.9%
Phase II to Phase III28.9%24.6%31.2%
Phase III to NDA/BLA57.8%47.7%61.3%
NDA/BLA to approval90.6%92.0%90.2%
LOA from Phase I7.9%5.3%9.3%

Phase II is the hardest gate everywhere. Fewer than three programs in ten that enter it move on to Phase III, and in oncology fewer than one in four. Once a program has passed Phase III and filed, approval is the norm rather than the exception.

Likelihood of approval from each phase

The likelihood of approval, or LOA, is the product of the remaining transition rates. For all indications, the report gives an LOA of 7.9 percent from Phase I, 15.1 percent from Phase II and 52.4 percent from Phase III. You can rebuild them from the table: 0.289 x 0.578 x 0.906 is about 15.1 percent; 0.578 x 0.906 is about 52.4 percent.

The first discipline is to use the LOA from the phase the asset is actually in. An asset in Phase 2 should never carry the Phase I LOA, which already includes a Phase I failure risk it has passed. Conversely, an asset that has just completed a Phase 2 should not be valued with the Phase II LOA if the readout is already known. The probability is a statement about what is known on the valuation date.

The earlier edition of the same work, covering 2006-2015, put the LOA from Phase I at 9.6 percent. Benchmarks move between editions, so always record which one you used.

Durations: the other half of the benchmark

The same report gives an average of 10.5 years from Phase I to approval: 2.3 years in Phase I, 3.6 in Phase II, 3.3 in Phase III and 1.3 in regulatory review. Duration matters as much as probability, because every year of delay pushes the commercial cash flows further out while patent life keeps running. In the Veltrana case, a one-year delay of VLT-201, with the loss of exclusivity date unchanged, cuts the rNPV from $71.09 million to $50.06 million. Asset-specific plans will differ from the average, but the average is a useful sanity check on a timeline that looks too fast.

What a transition rate actually measures

Three reading points prevent the most common misuses.

What moves the probability

The report isolates a few factors that shift the LOA from Phase I well away from the 7.9 percent average. Rare disease programs show 17.0 percent. Programs using patient preselection biomarkers show 15.9 percent, with Phase II success close to one in two. Chronic high-prevalence diseases sit at 5.9 percent. Among disease groups, hematology is the highest at 23.9 percent and urology the lowest at 3.6 percent.

These drivers overlap. A rare oncology program with a biomarker does not earn all three uplifts at once, because the subgroups share many of the same programs. Pick the most specific base that maps to the asset without ambiguity, usually oncology or non-oncology by phase, and treat the subgroup figures as evidence on the size of a plausible adjustment, not as rates to stack.

How to adjust a benchmark for a specific asset

A practical adjustment takes four steps:

The survivorship trap deserves a special mention. Every program in the benchmark Phase II population had early data good enough to enter Phase II. Positive early data are the entry ticket, not an edge. An uplift needs data stronger than typical: a clear dose response, a validated mechanism, an endpoint close to the one the next phase will use.

VLT-201: a modest, documented uplift

VLT-201 is an oral small molecule in Phase 2b for moderate-to-severe ulcerative colitis, so the non-oncology column is the base.

GateVLT-201Non-oncologyComment
Phase 2 to Phase 335%31.2%Modest uplift
Phase 3 to NDA62%61.3%In line
NDA to approval90%90.2%In line
Cumulative19.53%17.25%Driven by Phase 2

The cumulative likelihood of approval is 0.35 x 0.62 x 0.90 = 19.53 percent. Almost the whole difference from the benchmark sits at the Phase 2 gate, where asset-specific evidence weighs most. Note that the base here is the non-oncology column rather than a disease-specific line: the report's gastroenterology grouping excludes inflammatory bowel disease, so a careful analyst does not borrow it.

What does the uplift cost or add? Sensitivity S6 replaces the case probabilities with the benchmark rates of 31.2, 61.3 and 90.2 percent, for a cumulative 17.25 percent. The rNPV falls from $71.09 million to $57.21 million, a swing of $13.88 million. Sensitivity S2 flexes the Phase 2 to 3 rate alone: at 25 percent the rNPV drops to $38.58 million, at 45 percent it rises to $103.61 million, a swing of $65.03 million, the third largest in the tornado behind peak share and the discount rate.

Two lessons follow. The move from benchmark to case-specific numbers is small compared with the uncertainty around a single gate, which is what a well-calibrated adjustment looks like. And the Phase 2 rate deserves far more diligence than the regulatory rate, because it is both the most uncertain and the most valuable. Once the Phase 2b reads out positively, the remaining probability through approval becomes 55.8 percent (0.62 x 0.90) and the value at January 1, 2028 rises to $360.92 million. You can see each of these effects by changing the probability cells in the free rNPV drug asset valuation model.

VLT-305: one justified departure in oncology

VLT-305 is a bispecific antibody in Phase 1 for a biomarker-selected solid tumor, valued against the oncology column.

GateVLT-305OncologyComment
Phase 1 to Phase 250%48.8%In line
Phase 2 to Phase 330%24.6%Biomarker selection
Phase 3 to BLA48%47.7%In line
BLA to approval92%92.0%In line

The cumulative probability is 6.62 percent, against an oncology LOA from Phase I of 5.3 percent. The single departure sits at Phase 2, the gate where biomarker selection has its strongest documented effect. The model does not move anywhere near the one-in-two Phase II success seen for biomarker-selected programs: that subgroup spans all therapeutic areas, the biomarker hypothesis is untested in patients, and a novel bispecific carries its own early risk. A few points at one gate is the proportionate response. The biomarker is used once, where it belongs, and is not used again to raise peak share, since a selected population is narrower, not broader.

With 6.62 percent and Phase 3 costing $140.00 million a year from 2031 to 2033, VLT-305 developed alone has an rNPV of -$25.16 million despite an unrisked NPV of $1,010.83 million. Low probability combined with heavy late spending is exactly the profile that pushes small companies toward partnering; the guide on how to value a biotech licensing deal shows how a term sheet turns that negative value into $61.01 million for the licensor.

Writing the PoS memo

A one-page PoS memo should state the benchmark edition and column, the rate retained at each gate, the evidence behind any departure, the cumulative probability, and the sensitivity of value to the most uncertain gate. Reviewers rarely object to a judgment that is written down; they object to a number nobody can trace. The probabilities then feed the full build described in how to calculate rNPV step by step, and the complete framework, including the chapter on benchmarks and judgment, is in The Biotech Valuation Handbook.

Questions practitioners ask

What is the probability of success from Phase 2 to Phase 3?

In the BIO, Informa Pharma Intelligence and QLS Advisors 2011-2020 study, the Phase II to Phase III transition rate is 28.9 percent across all indications, 24.6 percent in oncology and 31.2 percent outside oncology. It is the lowest transition rate in development in every column.

What is the likelihood of approval from Phase 1?

The same study reports 7.9 percent for all indications, 5.3 percent for oncology and 9.3 percent for non-oncology. Rare disease programs reach 17.0 percent and programs using patient preselection biomarkers 15.9 percent, while chronic high-prevalence diseases sit at 5.9 percent.

Can I combine the rare disease and biomarker uplifts?

Not by multiplication. The subgroups overlap heavily, so stacking them double counts the same programs. Start from the most specific clean base, usually oncology or non-oncology by phase, and use the subgroup figures to size a documented adjustment at the gate where the evidence applies.

How much does the probability assumption move the valuation?

A lot. For VLT-201, moving the Phase 2 to 3 rate between 25 and 45 percent moves the rNPV between $38.58 million and $103.61 million. Replacing all three case rates with the non-oncology benchmark lowers it from $71.09 million to $57.21 million. Test it in the free rNPV model.

Read the full handbook

This guide is one arc of The Biotech Valuation Handbook. The book values a whole clinical-stage company: probabilities of success, development costs, a patient-based peak sales forecast, gross-to-net, launch curves, exclusivity and the IRA, the asset cash flow, the discount rate, sensitivities, two licensing deals and the sum-of-the-parts.

Paperback PDF edition, $19.99 Book page and companion workbooks

Free download: rNPV Drug Asset Valuation Model (Excel).

Also on this site

Illustrative figures computed by the book's model; the company, its products and every dollar figure are fictional. Rules and benchmarks as read in November 2026; education only, not investment advice.