Why most drugs fail between phase 2 and approval

A company announces that its experimental drug "met the primary endpoint" in a phase 2 trial. The stock jumps, patient forums light up, and a few articles call it a breakthrough. Three years later the phase 3 trial misses, and the program quietly ends.

That sequence is common enough to be the default. The stretch between the start of phase 2 and an FDA decision is where most experimental drugs that reach people are abandoned. Understanding why helps anyone reading trial news, whether they are a patient, a relative or an investor.

The numbers

The most cited recent figures come from "Clinical Development Success Rates 2011-2020," a 2021 report by the Biotechnology Innovation Organization (BIO), Informa Pharma Intelligence and QLS Advisors. It tracked about 9,700 development programs and more than 12,700 transitions from one phase to the next.

Multiply the last three rows and you get a rough rule: of 100 programs that start phase 2, around 15 end up approved. The same report put the average time from phase 1 to approval at about 10.5 years.

The FDA's own patient page on clinical research uses rounder, older numbers: roughly 70 percent of drugs move from phase 1 to phase 2, about 33 percent from phase 2 to phase 3, and 25 to 30 percent from phase 3 onward. The methods differ, but the shape is the same. Phase 2 is the narrowest point.

Averages hide large differences. The BIO report found hematology programs among the most successful and oncology programs among the least, and it found better odds for programs that selected patients using biomarkers and for rare disease programs. A single "1 in 10" figure is a starting point, not a forecast for any particular drug.

Why phase 2 is the filter

Phase 1 asks mainly whether a drug is tolerable at doses that might work. Most compounds that get that far clear that bar. Phase 2 is the first serious test of whether the drug does anything useful in patients with the disease.

It is also a small test. Phase 2 trials typically enroll up to a few hundred people over several months to two years, often with an endpoint that can be measured quickly, such as a blood marker or a scan, rather than survival or hospital admissions. That design is sensible, since nobody wants to spend phase 3 money on a drug with no signal. But small trials with fast endpoints produce results that look better than the truth more often than people expect.

Four reasons stand out:

1. Chance. With a few dozen patients per arm, a real but modest effect and a lucky fluke can look alike. Programs that get lucky in phase 2 are the ones that move on, so phase 3 starts with a built-in overestimate. Statisticians call this regression to the mean, and in drug development it guarantees a certain number of disappointments.
2. Patient selection. Phase 2 trials often enroll carefully chosen patients at specialist centers. Phase 3 casts a wider net, across more sites and countries, and the effect can thin out.
3. The endpoint itself. A phase 2 result based on a marker that is supposed to predict benefit (more on this below) can be real and still not translate into people living longer or feeling better.
4. Dose. Phase 2 is usually where the phase 3 dose gets chosen, often from only two or three options. Too low and the larger trial misses; too high and side effects pile up. The FDA's oncology center started Project Optimus in 2021 to push cancer drug developers to compare doses properly before their large confirmatory trials, because too many had settled on the highest tolerable dose by default.

The reasons programs end

A 2022 review in Acta Pharmaceutica Sinica B, titled "Why 90% of clinical drug development fails and how to improve it?", summarized analyses of clinical trial data from 2010 to 2017. It attributed roughly 40 to 50 percent of clinical failures to lack of efficacy, about 30 percent to unmanageable toxicity, 10 to 15 percent to poor drug-like properties and about 10 percent to lack of commercial need or poor strategic planning.

Efficacy

The most common reason is simple: the drug does not work well enough. Sometimes the biology was wrong, and the target that looked central in cells or mice turns out to be a bystander in human disease. Sometimes the drug works but no better than existing treatment, which, for a regulator weighing risks and for a doctor choosing a prescription, amounts to failure. Placebo response also matters, especially in pain, psychiatry and other conditions where patients report their own symptoms.

Safety

Phase 1 and phase 2 together may include a few hundred people. A side effect that affects 1 person in 1,000 can easily go unseen until phase 3 or later. Liver injury, heart rhythm problems and immune reactions are classic late surprises. Even a moderate safety signal can end a program if the drug's benefit is small or the condition is not life-threatening.

Commercial and strategic reasons

Not every abandoned drug failed scientifically. A competitor may reach the market first with a similar product. The standard of care may change mid-development. A company may be acquired, and the buyer may cut programs that do not fit. A small company may simply run out of money before phase 3, usually the most expensive stage by far. These decisions are less visible, but they account for a real share of programs that stop.

Surrogate endpoints: useful, and a known trap

A surrogate endpoint is a measurement used in place of the outcome patients care about. Blood pressure stands in for strokes, a tumor's shrinkage for survival, a lab value for kidney failure. Surrogates make trials smaller and faster, and some are well established. The FDA keeps a public table of surrogate endpoints that have supported approvals, separating validated surrogates from those only "reasonably likely" to predict benefit. Drugs approved through the accelerated approval program on the second kind of surrogate must confirm their benefit in later trials.

History is full of surrogates that moved in the right direction while patients did worse. In the Cardiac Arrhythmia Suppression Trial, reported in 1989, drugs that suppressed irregular heartbeats on monitors were associated with more deaths than placebo. In 2006, a drug that raised HDL ("good") cholesterol substantially was stopped after its large outcome trial showed excess deaths. Heart research has a cell therapy version of the same story: early trials measured improvements in pumping function on scans, and later, larger trials struggled to show clinical benefit. Our explainer on cell therapy after a heart attack goes through those trials.

When a phase 2 result rests on a surrogate, the question is how good that surrogate's track record is in that disease.

Reading trial news as a patient or family member

Nothing here is medical advice. It is a way to read headlines more calmly.

• Look for the word "randomized" and for what the comparison group received. A single-arm trial with no control can show that something happened, not that the drug caused it.
• Find the number of participants. Results from 30 people are a hint; results from 3,000 are evidence.
• Check whether the primary endpoint is a measurement or an outcome patients feel.
• Look the trial up. Most US phase 2 and phase 3 drug trials are required to be registered on ClinicalTrials.gov, and our guide on how to read a ClinicalTrials.gov listing shows how to find the pre-registered primary outcome and compare it with what the press release emphasizes.

Phrases such as "trend toward significance," "in a subgroup of patients" or "on a key secondary endpoint" usually mean the main question was not answered in the drug's favor.

Reading trial news as an investor

This is general education, not investment advice. For small biotech companies with one or two programs, a phase 2 readout can decide whether the company survives, which makes the base rates above directly relevant. A few habits help:

• Read the 8-K filed with the results, not just the press release, and wait for the full data presentation if it comes later.
• Check whether the endpoint reported as the headline was the one registered in advance.
• Look at how much cash the company has and what a phase 3 would cost. A positive phase 2 at a company with nine months of cash usually means a large share sale is close.
• Notice who pays for phase 3. A small company that licenses its drug to a larger partner after phase 2 gives up much of the upside but also much of the cost and risk, and the deal terms say a lot about how the partner rates the data.

Our reference page on reading a biotech company's filings covers cash runway, going concern language and dilution in more detail.

One number is worth keeping in mind each time a phase 2 headline appears: on the industry's own data, roughly 85 of every 100 drugs that start phase 2 are never approved. A positive result moves a drug into the minority that keeps going, and the next trial is designed to find out whether it belongs there.