How a molecule becomes a medicine
From lab hit to pharmacy shelf: the trial phases, the failure rates, and the patent clock that makes the whole industry run the way it does.
The question
A promising molecule is found in a lab. Years later it is at a pharmacy. What happens in between — and why does it cost so much and fail so often?
What's happening
The pipeline is not one process but a funnel, and the thing that shapes the entire industry is not the discovery; it is the attrition.
How it actually works
The path from molecule to medicine is a staged filter, where each stage is larger, longer and more expensive than the last:
| Stage | Purpose | The brutal part |
|---|---|---|
| Discovery | find a "hit" that binds the target | thousands of candidates, mostly dead |
| Preclinical | test in cells/animals for safety | toxicity kills many here |
| Phase I | safety in ~tens of healthy people | dosing, side effects |
| Phase II | does it work in patients? | most failures happen here |
| Phase III | large trials, vs. standard care | the expensive, decisive stage |
| Approval | regulator reviews the evidence | years of review |
| Market | manufacture, price, distribute | the patent clock is already running |
Three mechanisms explain the industry's behavior:
- Attrition drives cost. Because only a tiny fraction of candidates survive, the cost of the winners must carry the cost of the thousands of losers. This is why a single approved drug can cost billions to develop.
- Phase II is the graveyard. A molecule that is safe is not necessarily effective, and proving efficacy in real patients is where most programs die.
- The patent clock is the whole business model. A patent is filed early, but the drug only earns after approval — so the effective period of exclusivity is short. The industry runs on racing to approval, then maximizing revenue in the remaining window before generics arrive.
Everything about pharma strategy — the urgency, the pricing, the mergers — follows from two facts: most molecules fail, and successful ones have a countdown.
What's unknown
- Whether AI-driven discovery meaningfully raises the survival rate, or mostly speeds up the earliest stage while leaving the Phase II graveyard intact.
- How pricing is justified against R&D cost in a way that stays stable across countries.
- Open: the exact shape of attrition by therapeutic area. Failure rates are wildly different for, say, oncology versus a new painkiller, and I only have rough figures.
Reading
- Clinical-trial phase methodology and success-rate studies.
- Pharmaceutical patent-term and exclusivity explanations.