Analysis · Public data

How long enrollment really takes, by phase and therapeutic area

Feasibility decks assume a pace. The registry records what actually happened in 122,063 completed studies. The gap between those two numbers is where rescue budgets come from.

Rock Enroll editorial · ClinicalTrials.gov, retrieved September 2026

We took every interventional study on ClinicalTrials.gov with a status of Completed and a start date between January 2015 and December 2023, and measured the interval from study start to primary completion — the enrollment-and-follow-up window a sponsor actually lived through. After discarding records with missing or implausible dates, 122,063 studies remained.

By phase

Two things stand out. First, Phase 3 is faster than Phase 2 at the median — 21 months versus 23 — while enrolling roughly four times as many patients. That is not because Phase 3 recruitment is easier; it is because Phase 3 studies are funded to run many more sites and to buy demand. Pace is bought.

Second, look at the spread rather than the median. A Phase 2 at the 75th percentile takes 39 months against a median of 23. The distribution has a long right tail, and every study in that tail was planned against something closer to the median. Nobody budgets for the 75th percentile.

By sponsor type (Phase 2 and 3)

An academic Phase 2/3 takes nine months longer than an industry one to enroll roughly a third as many patients, typically from a single site. NIH-led studies take longest of all. The mechanism is not mysterious — one site means one catchment, one PI's referral network and no paid demand generation — but it is worth naming, because single-site feasibility estimates are routinely built from the investigator's clinic volume rather than from the share of that volume that will actually consent.

By therapeutic area (Phase 2 and 3)

Oncology is the outlier, by a wide margin

A median oncology Phase 2/3 takes 37 months to enrol 60 patients across 3 sites. At the 75th percentile it takes 54 months — four and a half years. Compare immunology: 18 months, 178 patients, 27 sites. Same registry, same period, an order of magnitude apart in productivity per month.

Part of that is biology and eligibility narrowness. A biomarker-defined, line-of-therapy restricted, prior-treatment-excluded oncology cohort genuinely is rare. But part of it is structural: oncology recruits through referral rather than demand, from a handful of academic centres, in a population that is being competed for by every other trial at the same institution. The rare-disease row is the interesting rebuttal — 22 months at a median of 14 sites, faster than cardiovascular, because rare-disease sponsors accept from day one that they must go and find patients rather than wait for a clinic to produce them.

How to use this

Three practical uses. Use the 75th-percentile column, not the median, when you stress-test a timeline — half of the studies in your comparator set will land past the median by definition. Compare your planned patients-per-month against the therapeutic-area row rather than against your CRO's reference set, which is selected on their wins. And if your study is already tracking past the 75th percentile for its area, the problem is no longer the plan; it is an operating problem, and adding contingency months will not fix it.

If you are at that point, diagnose which funnel stage is failing before approving more spend. The companion piece on site count versus enrollment pace covers the usual first instinct — adding sites — and what the registry says it actually buys.