Lesson 04 · 4 min · 6 things to do
What each phase actually asks
Match a trial stage to the question it can answer.
A drug passes its first human trial in 30 healthy volunteers. What has been established?
- Yes.The first stage asks about safety and dosing, usually in healthy people who cannot benefit from it. Reporting it as a promising treatment result mistakes the question being asked.
- Not quite.That it appears tolerated at those doses in 30 people. Rare harms need thousands.
- Not quite.There is nothing to work on — the volunteers do not have the condition.
Each phase asks a different question, and answers only that one. Most failures happen when a later phase asks the question the earlier one could not.
Match each question to the stage that can answer it.
What dose can be tolerated?
Does it reduce deaths compared with current treatment?
What does the body do with the drug?
Does it cause a rare, serious harm?
Is there any sign of an effect worth testing properly?
Yes.Small trials can find common effects and gross problems. Anything rare, or any comparison against an existing treatment, needs numbers that only a large trial can supply.Slide through the stages and watch what survives.
LaboratoryFirst in humansMid-stageLarge trial100 of 100 · candidates still alive
LaboratoryThousands of candidates show promise in cells and animals.
60 of 100 · candidates still alive
First in humansA fraction are tolerated well enough to continue.
25 of 100 · candidates still alive
Mid-stageNow tested for an actual effect. Most fail here — the effect was not there.
10 of 100 · candidates still alive
Large trialCompared against real alternatives, at scale. Roughly one in ten of the originals arrives.
Nine in ten promising candidates fail. What does that tell you about a laboratory finding in the news?
- Yes."Scientists find compound that kills cancer cells in a dish" is a real result and a very early one. Bleach also kills cancer cells in a dish.
- Not quite.It is doing its job — generating candidates. The filter exists because most candidates do not survive contact with a body.
- Not quite.The failures are mostly drugs that turned out not to work, which is what the trials are for.
Move the control to see what changes.
Of 100 candidates entering human testing, about 10 reach approval. What percentage fail somewhere along the way?
%Yes.Nine in ten. It is why drug development is expensive, and why a company's excitement about an early result should be read as excitement about a lottery ticket with better odds than most.Why do serious harms sometimes appear only after approval?
- Yes.It is arithmetic rather than negligence. This is why monitoring continues after approval, and why the first years of a new medicine's use are themselves a kind of study.
- Not quite.Selective reporting is a real and separate problem. The rare-event arithmetic holds even with perfectly honest reporting.
- Not quite.Raising them means detecting rarer harms, which means far larger and longer trials — a genuine trade with costs on both sides.
A press release says a drug "showed promise in early trials". What is the useful next question?
- Yes.Both halves decide what the sentence is worth. Promise in a dish, promise in mice, and promise against an existing treatment in 2,000 patients are three different claims that share a word.
- Not quite.Useful, and it does not tell you which question the trial was asking.
- Not quite.Worth knowing and much less informative here than the stage.
Lesson complete
Each trial phase answers one question and cannot answer the next one.
Next: Measuring what you can reach →