Drug development gets sold as a clean scientific conveyor belt: lab, mice, humans, approval, pharmacy shelf. That’s the brochure. The real machine looks more like a casino with a research budget — most things lose, the house designs the games, and the odds aren’t posted anywhere you’d casually look.
Here’s the whole pipeline, phase by phase, plus the approval mechanics and the parts that never make the press release.
Preclinical: years of work, zero human rules
Everything starts in cells and animals. No consent forms, no ethics board watching your every move the way it will later. Researchers test thousands of compounds, tweak molecules, and try to find something that does the thing without immediately destroying the liver.
This stage takes years and burns enormous money. The overwhelming majority of candidates die here and nobody outside the building ever hears about them. The animal models are also famously mediocre predictors of what happens in humans — a fact the industry knows, acknowledges internally, and politely ignores in public.
What escapes this stage is not the best drug. It’s the drug that survived the gauntlet and looked promotable.
Phase 0: the phase you’ve never heard of
Also called exploratory or microdosing studies. Tiny numbers of people get sub-therapeutic doses — amounts too small to treat anything — purely to see how the body handles the compound. No safety conclusions, no efficacy conclusions, no headlines.
Its whole purpose is triage: figure out which candidate deserves the expensive human trials before you spend the money. If you’ve never seen a Phase 0 trial advertised, that’s working as intended.
Phase I: is this going to kill you?
Phase I is 20 to 100 people, and the question is brutally simple: what does this do to a human body? Doses escalate in small steps. Participants are usually paid healthy volunteers, not patients — because you want to see the drug’s effects without a disease muddying the picture.
This is where a whole subculture exists. Serial participants who treat first-in-human studies as a side income. Clinics that run these studies on tight schedules. The pay varies wildly depending on how long you’re confined and how invasive the protocol is, and the risk is real — this is the phase where catastrophic reactions happen.
Exception: in oncology and other severe diseases, Phase I trials often enroll actual patients who’ve run out of approved options. That’s a very different calculation, and it’s usually a last-resort shot rather than a paid gig.
Phase II: where most drugs go to die
Now you have a few hundred patients with the actual condition. The question shifts to efficacy: does this do anything at all, and at what dose?
This is the graveyard. A huge share of compounds that looked great in animals and passed safety testing simply don’t work in people. Sometimes they work, but not enough. Sometimes they work in a subgroup nobody predicted.
Phase II is also where the language starts getting slippery. Endpoints get defined in ways that can be measured and reported favorably. Timelines get chosen after the fact. It’s not always fraud — it’s a lot of judgment calls made by people whose bonuses depend on the answer.
Phase III: expensive theater with real stakes
Hundreds to thousands of patients. Randomized, controlled, often double-blinded. This is the trial that decides whether the drug gets approved, and it costs enough to sink a mid-sized company if it fails.
Here’s the uncomfortable part: Phase III trials are designed to win, not to answer your question. The exclusion criteria quietly remove anyone complicated — people over a certain age, people on multiple medications, people with more than one condition, pregnant people (always), anyone whose data might muddy the signal.
So the “average patient” in the trial often doesn’t exist in real life. The label gets written for a person who never walked into the clinic.
Phase IV: the part after the money’s made
Post-market surveillance. Sometimes regulators require it, sometimes it’s voluntary. This is where rare side effects surface once millions of people take the drug instead of a few thousand carefully screened ones.
It’s also where seeding trials live — studies whose real function is to get prescribers familiar with the product while generating publishable data. The science is thin; the marketing is not.
How approval actually happens
The sponsor assembles a dossier — often tens of thousands of pages — and hands it to regulators. Key things to understand:
- Regulators don’t read the raw data. They read the sponsor’s analysis of it, then audit. This is a well-known structural weakness.
- Advisory committees of outside experts vote, but they usually get the sponsor’s summary, not the underlying datasets, and their votes aren’t binding.
- Accelerated pathways allow approval based on surrogate markers — a lab number, a scan, something that correlates with benefit rather than proving it. Confirmatory trials are supposed to follow. They sometimes take a decade, and the drug stays on the market the whole time.
- Priority review and special designations shorten timelines for serious conditions and rare diseases. Good policy, occasionally gamed.
- The label is the real product. What’s written there — which patients, which dose, which warnings — is where all the negotiation happened.
The stuff that doesn’t make the brochure
- Roughly half of registered trials never report results publicly within a year of finishing, despite rules requiring it.
- Negative results are buried or reframed. Positive subgroup findings get promoted to headline status.
- Primary endpoints get quietly swapped after unblinding. It’s detectable if you compare the registry entry to the publication — very few people bother.
- Comparators are sometimes chosen because they’re easy to beat.
- Guideline panels and advisory bodies have documented conflicts of interest that are disclosed in fine print nobody reads.
Getting in when you’re not invited
If you’re trying to enroll — for yourself or someone else — most people fail on logistics, not medicine.
- Call the site directly. Trial coordinators answer phones. They know about slots that never hit the public listing.
- Ask what makes people screen-fail. It’s usually paperwork, travel distance, or a stable medication. Sometimes it’s fixable.
- Ask about expanded access and compassionate use programs. Most companies have one and almost none advertise it. You typically need your physician to push.
- Ask who pays for travel, lodging, and the drug after the trial ends. Many trials reimburse travel and nobody mentions it. The post-trial drug question is the trap that ruins people.
- Get the consent document before you commit and read the section on what happens if you’re injured. It’s short for a reason.
Getting the data they’d rather you not read
Regulators publish their own review documents for many approvals — the internal analysis, the disagreements, the stuff the press release skipped. Transparency rules in several jurisdictions let you request the full clinical study report. Public registries are legally required to hold results, and you can compare what was registered against what was published.
None of this is secret. It’s just buried, and burying things works remarkably well.
The bottom line
Drug development is a business with a scientific engine bolted on, not a charity with a lab. Phases aren’t magic gates — they’re risk-transfer checkpoints where the design choices already tilt the odds. Understanding that doesn’t make you cynical. It makes you literate. And literacy is the only leverage that works at every phase, on both sides of the clipboard.