Phases in clinical trials: set phases to run!
Even if you are not working directly in clinical research, you have probably encountered the concept of trial phases in news coverage, particularly during periods when new treatments are under development. But what do the phases actually mean, how did they come about, and when do they apply?
Clinical trials became central to evidence-based medicine in the 20th century. The phase structure we now take for granted evolved gradually, driven by a growing recognition that participant safety had to be the first priority before any question of efficacy could be meaningfully answered.
Phase 0: Proof of concept
Phase 0 is the newest addition to the framework, emerging in the 1990s and 2000s. The idea is simple: before exposing a meaningful number of people to a new drug, gather early data on how it behaves in humans at sub-therapeutic doses. This gives researchers early pharmacokinetic and pharmacodynamic data with minimal risk, and can prevent moving forward with compounds that are unlikely to work.
Phase I: Safety first
Phase I trials focus primarily on safety rather than efficacy. They typically involve a small number of participants and are designed to identify the maximum tolerated dose. The FDA and other regulatory bodies formalised Phase I requirements as part of a broader shift toward rigorous ethical standards in human research over the second half of the 20th century.
Phase II: Exploring efficacy
With a safety profile established, Phase II studies shift attention toward whether the intervention actually does what it is supposed to do. These trials involve larger cohorts than Phase I, focus on a specific patient or participant population, and provide the evidence base for decisions about further development. Safety monitoring continues throughout.
Phase III: Confirmation at scale
Phase III trials are the confirmatory stage. Smaller earlier-phase studies can produce results that do not translate to broader populations, whether because of statistical noise, narrow eligibility criteria, or demographic homogeneity. Randomised controlled trials at Phase III scale address this by enrolling diverse populations in sufficient numbers to produce statistically meaningful conclusions.
Phase IV: After market launch
Phase IV trials look at drugs or interventions that are already in use. Post-marketing surveillance of this kind is essential for detecting long-term safety signals that would not have been visible in earlier trial populations, particularly for products that become widely prescribed or available over the counter.
Why the attrition matters
The phase system only makes sense once you see how much gets filtered out along the way. A widely cited analysis of investigational drug development, tracking compounds from the 50 largest pharmaceutical companies as they moved through clinical testing, found an overall clinical approval success rate of 19% across the full study period, and just 16% for drugs originated entirely in-house rather than licensed in from elsewhere. Put differently: for every five or six compounds that begin Phase I, roughly one is likely to eventually reach approval. That attrition rate varied meaningfully by drug type too, with large-molecule biologics succeeding at close to double the rate of small-molecule compounds in the same analysis.
| Phase | Primary question | What typically happens next |
|---|---|---|
| Phase 0 | Does this behave as expected in humans at all? | Most compounds that fail here never reach Phase I |
| Phase I | Is this safe, and at what dose? | A meaningful share fail on safety or tolerability grounds |
| Phase II | Does this actually work? | This is where many compounds are dropped for lack of efficacy |
| Phase III | Does it hold up at scale, in a diverse population? | Fewer compounds fail here, but the cost of failure is highest |
| Phase IV | What happens once it's in widespread real-world use? | Rare long-term signals can still surface post-approval |
That attrition isn't a flaw in the system. It's the entire point of having four (or five, counting Phase 0) distinct gates rather than one. Each phase exists specifically because the previous one wasn't designed to answer the question the next one asks, and the low overall success rate is what happens when a system is deliberately built to let unpromising or unsafe compounds fail early and cheaply, rather than letting every candidate run the full gauntlet before anyone finds out it doesn't work.
The phase structure has always had participant safety as its underlying logic. Each phase exists because the previous one was not enough to fully answer the questions that needed answering before taking the next step, and the sobering approval statistics above are really just a description of that logic working as intended.