PoS benchmarks

Probability of success by indication and clinical stage: the baselines the model computes with, derived from the disease-area, modality and biomarker rows of BIO/QLS/Informa 2021 (Thomas et al.)¹, which analysed 12,728 phase transitions. Preclinical rates are PhaseFolio assumptions. Multipliers for genetic validation, orphan designation, orphan gene therapy and first-in-class are applied via log-odds transformation. These are reference figures, not a prediction for any asset: they prefill the asset modeler, where every value can be changed, and they are updated when the sources change.

Model: 4.1.0
Data version: 2026-09-27
Matrix: 264 cells
Reference: BIO 2021¹ · 12,728 transitions

Limits of use: the model has not been backtested, these figures are not a prediction for any asset, and nothing here is investment advice. See the model card.

01Attrition by therapeutic area

How many candidates each baseline carries from Phase I to approval, drawn beside the BIO 2021 area rows they are derived from. Baseline cohort: small molecule, no biomarker strategy.

Figure 1. Share of candidates still in development at each stage, from Phase I, on a log scale. Lines are PhaseFolio baselines; open circles mark the highlighted area's BIO 2021 row.¹Oncology (Hematologic): 6.9% reach approval · BIO 2021 7.5%

Therapeutic areaOf 1,000 candidates entering Phase IPhaseFolioBIO 2021¹

Approved
Failed at NDA / BLA
at Phase III
at Phase II
at Phase I
BIO 2021 approved count

Figure 2. Of 1,000 candidates entering Phase I, how many each baseline carries to approval and where the rest stop. The open circle marks where the BIO 2021 area row would stop the approved count. Choose an area to explore every modality and biomarker cell below.

Table 1. Baseline stage-transition probabilities. LoA is the cumulative probability of approval from preclinical, the product of all five stages; BIO 2021 reports no preclinical stage.

Baseline stage-transition probabilities by therapeutic area
Therapeutic area	Preclinical	Phase I	Phase II	Phase III	NDA / BLA	LoA
Oncology (Solid Tumor)	4.0%	49.5%	22.1%	41.3%	91.8%	0.17%
Oncology (Hematologic)	7.0%	50.7%	26.4%	58.4%	88.5%	0.48%
Rare Disease	8.0%	67.9%	42.8%	58.8%	92.6%	1.3%
Neurology	4.0%	48.3%	25.4%	51.5%	84.8%	0.21%
Immunology	6.0%	55.8%	29.8%	63.8%	93.2%	0.59%
Infectious Disease	7.0%	58.4%	36.7%	62.5%	91.8%	0.86%
Cardiovascular	5.0%	50.6%	19.8%	53.6%	80.1%	0.22%
Metabolic	6.0%	62.4%	43.2%	62.1%	85.7%	0.86%
Respiratory	5.0%	56.5%	20.7%	63.0%	94.9%	0.35%
Dermatology	6.0%	64.1%	36.9%	58.4%	86.7%	0.72%
Ophthalmology	6.0%	72.1%	33.8%	49.6%	89.7%	0.65%

Each row is the BIO 2021 area row shifted for small molecules and no biomarker strategy. Table 1 of the PoS methodology shows each row beside its BIO 2021 area row. Every modality and biomarker cell is in the explorer below.

02Explore the matrix

Every indication, modality and biomarker-strategy cell the model defines, with the three indication-class modifiers. Choose a cell to see its stage-by-stage derivation. Where an area has no baseline for a modality, the model uses that area's Other row, and the grid says so.

Indication-class modifiers

First-in-classOrphan designationGenetic validation

Likelihood of approval from Phase I

no modifiers

Likelihood of approval from Phase I for every modality and biomarker strategy in Oncology (hematologic)
Modality	None	Enrichment	Companion diagnostic
Small molecule
Monoclonal antibody
Bispecific antibody
Antibody-drug conjugate
Cell therapy
Gene therapy
Peptide
Other

88 of 88 area–modality pairs have a baseline of their own.

Oncology (hematologic) · Small molecule · None

6.9%

reach approval from Phase I

Cell baseline: 6.9%
Modifiers: None
BIO 2021 area row: 7.5%

Ink: this cell. Circles: BIO 2021 area row.

Stage-by-stage PoS derivation: oncology_hematologic, small_molecule, none
Stage	Base PoS	Adjusted PoS	LoA from stage
PreclinicalPhaseFolio assumption	7.0%	7.0%	0.48%
Phase I	50.7%	50.7%	6.9%
Phase II	26.4%	26.4%	13.6%
Phase III	58.4%	58.4%	51.7%
NDA / BLA	88.5%	88.5%	88.5%

LoA from stage is the probability of approval from the start of that stage. The preclinical row is the model's full LoA; the Phase I row is the figure above, the measure BIO 2021 reports.

Multipliers applied

No modifiers selected.

Model 4.1.0 · Data version 2026-09-27 · 11 areas × 8 modalities × 3 biomarker strategies

03Evidence-based multipliers

Each multiplier is applied to the baseline PoS at the stages listed, via log-odds (logit) transformation. The source estimand (RR or OR) is disclosed alongside how the model applies it.

Evidence-based PoS multipliers
Factor	Value	Applied to	Estimand	Source
Genetic Validation	2.60×	Phase II, Phase III	RR as OR	Minikel et al. 2024 Nature
Orphan Designation	1.50×	Phase II, Phase III	RR as OR	Mullard 2016 Nat. Rev. Drug Disc.
Gene Therapy (Orphan)	1.13×	Phase I, Phase II	OR as OR	Phares, Phillip & Trusheim 2025 Nat. Rev. Drug Disc. (calibrated to 18.5%; replaces Orphan Designation for gene therapy)
First-in-Class	0.85×	Phase II, Phase III	RR as OR	PhaseFolio assumption (no published figure)

04Use the benchmarks

The full matrix with modifier toggles is applied inside each project's rNPV and through an authenticated REST call.

In productProject rNPV →Each project's rNPV looks up its PoS from this matrix and shows the modifier derivation for every stage.
Public RESTGET /benchmarks/pos →The baselines on this page as machine-readable JSON. No authentication; version-stamped.

§Notes and references

Baseline rows are small molecule × no biomarker. Modality- and biomarker-specific values are available via the POST /benchmarks/pos/lookup endpoint (in-product, via MCP, or via authenticated API call).
Expanded modality labels from triage/imported assets normalize to one of the 8 benchmark baselines before lookup; see /methodology/pos-calibration for the current proxy mapping.
Biomarker strategy is built into each matrix cell (BIO 2021 Figure 11), not applied as a multiplier. Model 3.0.0 has no CAR-T multiplier.
Multipliers are applied via log-odds (logit) transformation. See the methodology page for the full derivation.
Preclinical PoS is never adjusted by any multiplier; NDA/BLA approval rates reflect filing quality, not drug-specific clinical attributes.

1Thomas, D., Chancellor, D., Micklus, A., LaFever, S., Hay, M., Chaudhuri, S., Bowden, R., & Lo, A.W. (2021). Clinical Development Success Rates and Contributing Factors 2011–2020. BIO, Informa Pharma Intelligence, QLS Advisors. go.bio.org/rs/490-EHZ-999/images/ClinicalDevelopmentSuccessRates2011_2020.pdf
2PhaseFolio benchmark matrix, model 4.1.0, data version 2026-09-27. backend/app/computation/pos_benchmarks.json

Read the full methodology →