25.8 — Clinical Trials, Phase by Phase

Out of a hundred drugs that get as far as the first human dose, about forty-seven finish Phase I. Of those, only about twenty-eight in every hundred that start Phase II finish it. Of the survivors, about fifty-five in every hundred get through Phase III. Multiply it through and roughly six or seven of the original hundred are ever approved.

Look at where the slaughter happens. It is not the first-in-human study, where you would expect the danger to be. It is Phase II — the point at which a company finds out whether the disease actually cares about the target it chose in Chapter 25.5.

Understanding what each phase is asking, and why it asks it in that order, is the single most useful piece of industry knowledge you can carry, because every clinical system, every timeline and every client crisis is anchored to this sequence.

The logic behind the order

Each phase asks one primary question, and the sequence is built to spend the least money possible before the most likely killer is tested.

Phase I asks: is it tolerable in a human, and what does the body do to it? Phase II asks: does it help patients, and at what dose? Phase III asks: is the benefit real and large enough, in the population who will actually receive it? Phase IV asks: what do we learn once hundreds of thousands of ordinary people take it?

A phase is not a legal category and the boundaries blur constantly. You will hear Phase Ib, Phase IIa and IIb, and "seamless Phase I/II" studies that flow into each other. The phase describes the question being asked, not a fixed protocol template — which is why a client's "Phase IIb" may look like somebody else's Phase III.

Phase I: does a human tolerate it

Typically 20 to 100 people, over several months to a year, most often in a dedicated unit where participants stay overnight and are monitored intensively.

For most drugs, the participants are healthy volunteers, paid for their time and inconvenience rather than for the risk. They are used because a healthy body without other diseases and other medicines gives the cleanest read of what the drug itself does.

For cytotoxic cancer drugs the design is different and the reason is ethical. A drug intended to kill dividing cells will harm a healthy person with no possible benefit, so oncology Phase I studies enrol patients with advanced cancer who have exhausted standard treatment. This single difference cascades through every part of oncology trial design and is the most common source of confusion when someone moves between therapy areas.

Two study shapes do most of the work.

Single ascending dose. A small group receives the starting dose calculated in Chapter 25.7. If it is tolerated, the next group receives a higher dose. Repeat upward with predefined stopping rules. Since 2006, first-in-human groups are dosed one participant at a time with a waiting interval, so that a catastrophic reaction affects one person rather than six.

Multiple ascending dose. Once single doses are understood, groups take the drug repeatedly for one to four weeks, to see whether it accumulates and whether tolerance changes with repeated exposure.

Alongside these run the studies that fill in the label: what food does to absorption, what happens in people with impaired kidneys or liver, and dedicated interaction studies with drugs likely to be taken at the same time. There is also a specific cardiac safety study to check whether the drug lengthens the heart's electrical recovery interval, since that can cause a fatal rhythm (Chapter 18.5).

What Phase I produces is a safe dose range, a full picture of blood levels over time, the side effects that appear first, and — where possible — an early biological signal that the drug is doing something in a human.

Phase II: does it actually work

Typically 100 to 300 patients who have the disease, over one to three years. This is where most programmes die, and it is worth being precise about why.

A Phase II study is the first real test of the hypothesis that has been carried since target selection: that interfering with this molecule changes this disease in real people. Everything before it tested whether the drug hits the target. Phase II tests whether hitting the target matters, and the answer is usually no.

The phase is usually split. Phase IIa is proof of concept, often small, often with a biological or short-term measure rather than a hard clinical outcome, designed to give the company a fast and cheap read. Phase IIb is larger and its main job is choosing the dose, comparing several doses against placebo or standard treatment to find the one with the best balance of benefit and side effects.

Getting the dose right here is the most underrated decision in drug development. Choose too low and Phase III fails for lack of effect; choose too high and Phase III fails on side effects or produces a label so restricted that doctors avoid the drug. A regulator's most common criticism of a failed programme is that the dose was never properly characterised.

The word most often heard around this phase is futility. Trials are frequently designed with an interim look, and if the results so far make eventual success implausible the study stops early. Stopping early is a success of the process, not a failure of it — the money saved funds the next programme — but it never feels that way inside the company, and this is often when a services contract is abruptly cancelled.

Phase III: proving it, at scale, to a regulator's standard

Typically 300 to 3,000 patients or more, over two to five years, across dozens or hundreds of hospitals in many countries. This is where most of the cost of drug development sits.

Phase III exists to produce evidence that will survive a regulator's statistician. The American law requires "adequate and well-controlled investigations", which for decades meant two independent trials both showing benefit. Since 1997 one convincing trial plus confirmatory evidence can be enough in defined circumstances, and in practice most new drugs still run two.

Four features define a Phase III design, and Chapter 25.10 takes each apart properly. Patients are randomised so that the groups differ only by chance. The trial is blinded wherever possible, so neither patient nor investigator knows the assignment. There is a control, either placebo or the existing standard treatment. And there is a primary endpoint fixed in advance, on which success or failure is declared.

The reason it takes so long is usually not the treatment period. It is recruitment. A trial needing 1,200 patients with a specific disease, within a defined severity range, without disqualifying other conditions, willing to be randomised, near a participating hospital, is genuinely hard to fill. Recruitment is the number one cause of delay in the industry, and it is the reason patient-matching software has been an active area for two decades.

Phase III also carries the burden of the safety database. Rare side effects only appear in large numbers of people. A side effect occurring in one patient in a thousand needs roughly three thousand exposed patients before you can be reasonably confident of seeing at least one case — the reason regulators specify minimum exposure numbers for chronic-use drugs rather than accepting whatever the efficacy trials happened to enrol.

Phase IV and the things that only appear later

Phase IV is everything after approval: studies the regulator required as a condition, studies the company runs to support use in new groups, registries, and safety surveillance.

The genuine scientific point of Phase IV is that pre-approval trials are, by design, unrepresentative. They exclude the very old, the pregnant, people with several other diseases, and people on many other medicines — precisely the patients who make up much of real-world use. A drug's true safety profile is learned after launch, and Chapter 25.22 is the machinery built to learn it deliberately rather than by accident.

Phase 0, and what it is for

Occasionally you will see a Phase 0 or microdose study: a tiny dose, well below any therapeutic level, given to a handful of people to see where the compound goes in the body and whether it reaches the target tissue at all.

Its purpose is purely economic — kill a compound with a study costing a small fraction of a full Phase I, before spending real money. It is uncommon, but knowing what the term means prevents confusion when it appears on a pipeline chart.

What varies by therapy area

The same phase means quite different things depending on the disease, and a client will assume you know this.

Area	What is unusual
Oncology	Patients in Phase I; response rates as endpoints
Vaccines	Tens of thousands in Phase III; healthy people
Rare disease	Tiny trials; sometimes no placebo group
Psychiatry	Very large placebo response
Antibiotics	Short treatment; hard to recruit fast
Devices	Often no randomised trial at all

Two rows deserve a sentence.

In rare disease there may be a few hundred patients worldwide, so a conventional trial is impossible. Regulators accept smaller studies, comparison against the disease's known natural history rather than a placebo group, and designs where every patient eventually receives the treatment. The evidence standard is not lowered so much as reshaped, and this is a large and growing area of work.

In psychiatry, patients given placebo often improve substantially, which means a drug must beat a moving target. Many genuinely useful psychiatric drugs have failed individual trials for this reason, and it is why those programmes run more trials than most.

Speeding things up: the designations, plainly

Four American mechanisms get confused with each other constantly. They do different things and can be held at the same time.

Fast Track gives more frequent meetings with the agency and the ability to submit the application in pieces as they are ready.

Breakthrough Therapy, created in 2012, requires preliminary clinical evidence of substantial improvement over available treatment, and brings intensive agency guidance including senior staff involvement.

Accelerated Approval, created in 1992 during the HIV emergency, allows approval based on a surrogate endpoint likely to predict benefit — a tumour shrinking rather than a patient living longer — with confirmatory trials required afterwards. This is where the surrogate-endpoint warning from Chapter 25.5 becomes a live and continuing controversy, because some confirmatory trials have taken many years or failed outright.

Priority Review shortens the agency's review clock from a goal of about ten months to about six, counted from the point the application is accepted for filing.

Only accelerated approval changes the evidence needed. The other three change speed and access to the regulator, not the standard. Getting that distinction right in a meeting is a small thing that signals a great deal.

What all of this means for the systems you will build

The phases determine the shape of every clinical system. Early phases are small, intensive, rich in measurements per patient, and often run in one country under one set of rules. Late phases are enormous, spread across many countries and languages, with hundreds of sites needing training and access management, and years of continuous data flow.

The tolerance for downtime differs too. A Phase I unit dosing a volunteer at seven in the morning cannot wait for a system to come back; a Phase III database can absorb a short outage but cannot absorb a data loss, because the affected patient visits cannot be repeated.

And every one of these studies is registered publicly before it starts, with its design and primary endpoint recorded — in the United States on ClinicalTrials.gov, with equivalents elsewhere. That registration exists to stop companies from quietly abandoning trials that came out badly or redefining success after seeing the data, and it is the single most useful public source for understanding a client's pipeline before you walk into their office.

Next: Chapter 25.9, how a trial actually runs day to day — the sponsor, the CRO, the sites, every role by name, and the systems that hold it together.