R&D productivity: cost per approval and phase-transition benchmarks

Most candidates die along the way. So the cost of one approval must absorb the cost of every sibling that failed.

The $2.6B figure has two components people often confuse:

1. Out-of-pocket cost: estimated around $1.4B per approved drug (same Tufts study).

2. Capitalized cost: the $2.6B figure, which adds the cost of capital (the return investors could have earned elsewhere) compounded over roughly a decade of development.

The gap between $1.4B and $2.6B is entirely the time value of money. In an industry where trials take years and produce zero revenue, the clock is a huge cost driver.

> Treat $2.6B as a widely cited industry benchmark, not gospel. Estimates range from under $1B to over $2.5B depending on method, therapeutic area, and whether failures are included. Oncology runs far higher; some repurposed or orphan drugs run far lower.

The core calculation: from success rates to cost per approval

Here is the logic that produces a cost-per-approval number. We use Phase I to approval success rates, meaning the probability that a molecule entering Phase I eventually gets approved.

A commonly cited industry benchmark is a Phase I to approval likelihood of approval (LOA) of about 10% (BIO / Informa / QLS "Clinical Development Success Rates" reports, multiple years). In plain terms: roughly 1 in 10 molecules that start human trials ever launch.

Worked example: expected molecules per launch

If the Phase I to approval rate is 10%:

You must fund 10 Phase I starts to expect 1 approval. This single ratio is why cost per approval is so brutal.

Building the cost with phase transitions

Not all failures cost the same. A molecule that dies in Phase I wasted a little; one that dies in Phase III wasted a fortune. We need phase-transition probabilities (the chance of advancing from one phase to the next) and per-phase costs.

Illustrative benchmark transition rates (BIO/Informa, approximate, all-disease):

| Transition | Approx. success rate |

|---|---|

| Phase I to Phase II | ~52% |

| Phase II to Phase III | ~29% |

| Phase III to NDA/BLA filing | ~58% |

| Filing to approval | ~91% |

Multiply them: 0.52 × 0.29 × 0.58 × 0.91 ≈ 0.08, roughly 8%, in line with the ~10% rule of thumb (these numbers drift by report year and therapeutic area, so treat as estimates).

Now attach illustrative per-phase out-of-pocket costs (rough public estimates, US, order of magnitude only):

| Phase | Est. cost per trial |

|---|---|

| Phase I | ~$25M |

| Phase II | ~$60M |

| Phase III | ~$255M |

*(These are widely circulated approximations; actual costs vary enormously by indication. Oncology Phase III trials can exceed $300M.)*

Simple expected-cost stack

Imagine you start 100 molecules in Phase I to see how spend accumulates:

• 100 enter Phase I at $25M each = $2.5B
• 52 advance to Phase II at $60M = $3.12B
• ~15 advance to Phase III at $255M = $3.83B
• End result: roughly 8 approvals

Total out-of-pocket clinical spend ≈ $9.45B for 8 approvals = about $1.18B per approval on an out-of-pocket, pre-capitalization basis.

That lands close to the ~$1.4B out-of-pocket figure once you add preclinical discovery costs and failed programs before Phase I. Add the cost of capital over 10 years and you climb toward the $2.6B capitalized number.

The exercise shows the mechanics: cost per approval is driven far more by the failure rate than by the price of any single trial. Halving your Phase II to Phase III attrition does more for productivity than trimming any budget line.

Reading the number: US vs Europe

The R&D cost drivers are broadly similar across the US and Europe because trials are often global, but two financial nuances matter:

• Pricing at the back end. The US allows freer drug pricing; European systems negotiate via bodies like Germany's G-BA/IQWiG and England's NICE (National Institute for Health and Care Excellence). A given R&D spend earns a higher expected return in the US market, which is why US revenue often justifies global R&D.
• Regulatory pathway timing. The FDA and EMA have broadly comparable review timelines for standard reviews (roughly a year), but incentive schemes differ (FDA Breakthrough Therapy, EMA PRIME). Faster review shortens the capitalization clock and lowers effective cost.

For a free primer on the underlying cost study, see the Tufts CSDD cost of drug development summary.

Benchmarking a real pipelinepipelineAll active sales opportunities across the stages of the sales process, together with their combined potential value and probability of closing.Voir la définition complète →

Now use these benchmarks to judge whether a company's R&D is efficient. Two practical metrics:

1. R&D per approval (trailing)

R&D per approval = Total R&D spend over period / Number of approvals in period

Take a large-cap example structure (illustrative, not exact figures): if a company spent roughly $50B on R&D over five years and won 10 novel approvals, that is $5B per approval, well above the $2.6B industry benchmark. That gap flags either bad luck, expensive therapeutic areas (oncology, CNS), or genuine productivity problems.

Caveat: approvals lag spend by years, so trailing ratios punish companies that are early in a new investment cycle. Always look at a multi-year window.

2. Expected value per Phase I asset

For a pipelinepipelineAll active sales opportunities across the stages of the sales process, together with their combined potential value and probability of closing.Voir la définition complète → with N molecules entering Phase I, expected approvals = N × 0.10.

If a mid-size biotech has 6 molecules entering Phase I, expect ~0.6 approvals from that cohort. If the company's valuation implies 3 future launches from those 6, the market is pricing success far above the industry base rate. That is a red flag to probe: is there a platform advantage, or is it hope?

Adjusting the base rate

Base rates shift by therapeutic area. Rough LOA benchmarks (BIO/Informa, approximate):

• Hematology / oncology: lower, often ~5-8% Phase I to approval.
• Vaccines and infectious disease: higher, sometimes ~15-20%.
• Rare / orphan diseases: typically higher than the all-disease average, partly due to smaller trials and regulatory incentives.

So a purely oncology pipelinepipelineAll active sales opportunities across the stages of the sales process, together with their combined potential value and probability of closing.Voir la définition complète → should be benchmarked against a ~6% rate, not 10%. Applying the wrong base rate is the most common analytical error.

MedTech is different

Do not apply drug benchmarks to medical devices. Device development under the FDA's 510(k) pathway (clearance based on similarity to an existing device) or PMA (Premarket Approval, for higher-risk devices) is far cheaper and faster. A 510(kkThe average number of new users each existing user generates through referrals. Above 1.0, growth compounds on itself and becomes exponential.Voir la définition complète →) clearance can cost single-digit millions and take under a year. Device R&D productivity is measured differently: iteration speed, gross margingross marginGross margin is the share of revenue left after subtracting the direct cost of producing goods or services, expressed as a percentage of revenue.Voir la définition complète →, and installed-base pull-through, not cost per approval.

Key Takeaways

• The ~$2.6B capitalized cost per approved drug is failure-driven. Roughly $1.4B is out-of-pocket; the rest is the cost of capital over ~10 years. Treat all figures as estimates that vary by source and therapy area.
• A ~10% Phase I to approval rate means you fund about 10 Phase I starts per launch. Reducing attrition, especially at the costly Phase II to Phase III jump, improves productivity more than cutting per-trial budgets.
• Benchmark R&D per approval over multi-year windows and adjust the base rate by therapeutic area: ~6% for oncology, higher for vaccines and rare disease.
• Compare a company's implied future launches to the base-rate expectation. If the market prices more successes than attrition rates support, dig into whether a genuine platform edge exists.
• Never apply drug benchmarks to medtech. Device pathways (510(kkThe average number of new users each existing user generates through referrals. Above 1.0, growth compounds on itself and becomes exponential.Voir la définition complète →), PMA) are cheaper and faster, and productivity is measured on speed and margins, not cost per approval.

*This lesson is educational and not investment advice.*