The Pipeline the Headlines Skip
Every drug claim in this series, from the psilocybin trials to the nitazene alerts to the CBD marketing, rests somewhere on the clinical trial pipeline, and the pipeline's structure (the phases, the endpoints, the attrition rates that the headlines skip) is the literacy this series' how-to-read-a-study post recommended in its applied form. The public conversation about drug development proceeds as if the process were a straight line (the discovery, the trials, the approval), when the reality is a funnel of staggering attrition (the roughly one-in-ten of the drug candidates that enter the human trials ever reaching approval, the billions spent on the nine that fail) with distinct phases of distinct purposes (each phase answering a distinct question, with distinct designs and distinct failure modes), and the phases' understanding is the difference between the informed reading of a trial's result (the Phase 2's signal, the Phase 3's confirmation, the pivotal-trial designation) and the marketed misreading (the Phase 1's safety-only data promoted as efficacy, the single-study result promoted as established). This post is the pipeline primer: the phases, the endpoints, the attrition, and the reading rules.
The Phases, Each With Its Question
Phase 0 and the preclinical: the animal and the laboratory work (the safety pharmacology and the toxicology, the efficacy models that justify the human exposure) that the FDA requires before the human trials begin, the phase where the series' preclinical-overclaim critiques apply (the animal findings that the human translation disappoints, the BDNF and the critical-period coverage's discipline).
Phase 1: the first-in-human safety (the small cohorts of the healthy volunteers or the patient volunteers, the dose-escalation designs that establish the safety profile and the maximum tolerated dose, the pharmacokinetic measurements that the dose-selection requires). The Phase 1's reading rule: the phase answers the is-it-safe question, not the does-it-work question, and any efficacy signal at Phase 1 (the early biomarker shifts, the subjective improvements) is hypothesis-generating only, the promotion of Phase 1 signals as efficacy being the biotech-finance and the marketing standard that the informed reader discounts.
Phase 2: the exploratory efficacy (the larger patient cohorts, the randomized and the controlled designs, the dose-ranging, the preliminary efficacy against the placebo or the standard care). The Phase 2's reading rule: the signal, not the proof (the small samples' and the short durations' limitations, the multiplicity and the endpoint-flexibility's vulnerabilities, the positive Phase 2 that the larger Phase 3 fails to confirm being the pipeline's characteristic attrition).
Phase 3: the confirmatory (the large multicenter trials that the registration requires, the definitive endpoints and the pre-registered analyses, the statistical power that the smaller phases lacked). The Phase 3's reading rule: the confirmation, conditional on the design (the single pivotal trial's vulnerability to the chance and the bias, the replication requirement, the series' placebo-coverage's demand for the active placebos and the expectancy analyses in the expectancy-sensitive fields).
Phase 4 and the post-marketing: the surveillance (the adverse-event reporting, the real-world populations that the trials' exclusion criteria excluded, the rare harms that the trial samples could not detect, the withdrawal of the approved drugs that the post-marketing surveillance identifies) and the series' study-reading post's point (the approval is not the truth's endpoint, the post-marketing being where the population-scale reality arrives).
The Endpoints and the Designs
The endpoint question deserves the primer's treatment, because the outcome-choice shapes the trial's meaning more than the phase designation. The primary endpoint (the pre-registered outcome that the trial's power is calculated for, the statistical integrity's anchor), the secondary endpoints (the exploratory additions, the multiplicity's correction requirement), the surrogate endpoints (the proxies that the series' study-reading post covered, the biomarker standing for the outcome, the validation question that the surrogate requires), and the composite endpoints (the combined outcomes that the rare-event trials use, the component-mixing's interpretive hazards) are the vocabulary, and the designs (the superiority and the non-inferiority frameworks, the crossover and the parallel-group, the adaptive designs that the modern trials increasingly use, the active-placebo and the expectancy-manipulation innovations that the series' psychedelic-placebo coverage demanded) are the architecture the informed reader must recognize.
The Attrition and the Economics
The attrition deserves the honest numbers, because they explain the marketing's distortions. The funnel (the thousands of the preclinical candidates narrowing to the hundreds of the Phase 1 entries, to the dozens of the Phase 2 survivors, to the one of the Phase 3 confirmations), the cost accounting (the billion-dollar per-approval estimates that the industry cites, the failures' costs amortized into the successes' prices), and the incentive distortions (the publication bias that the positive-results' publication skews, the pipeline-gaming that the unregistered endpoints and the surrogate-heavy designs enable, the investor-pressure that the public-company era adds, the series' gold-rush coverage's clinical dimension) are the context the reader must hold, and the attrition's corollary deserves stating: most of the exciting results the reader encounters (the positive Phase 2s, the single-trial signals, the company press releases) will not survive the pipeline's narrowing, and the discount that the attrition justifies is the informed reader's default.
The Reading Rules, Applied
The pipeline's rules, applied to this series' recurring examples: the psilocybin-depression field's Phase 2 signals and the Phase 3 confirmations (the COMPASS program's pivotal trials, the registration pathway's demands), the MDMA-PTSD saga's Phase 3-to-FDA arc (the 2024 rejection being the Phase 3-to-approval attrition that the pipeline's final gate produces, the design-integrity questions that the gate enforced), the ketamine's accelerated-approval precedent (the fast-track pathway that the unmet-need designation enables, the post-marketing-confirmation requirement that the acceleration trades), the nitazene and the xylazine coverage's evidence tiers (the case-series and the surveillance data of the preclinical-plus-observational tiers, the trial evidence that the controlled substances' scheduling makes nearly impossible), and the CBD marketing's tier-confusion (the preclinical and the anecdotal promoted as the clinical, the study-reading post's pyramid applied). The pipeline, in short, is the series' epistemology in institutional form: the questions asked in sequence, the evidence's tiers respected, the attrition's honesty preserved, and the approval the final gate rather than the first claim.
The Bottom Line
Clinical trial phases are the pipeline's way of asking the questions in order (is it safe, does it work in the small and the controlled, does it confirm in the large and the definitive, what does the population-scale reality show), and the attrition funnel (the one-in-ten that survives, the billions spent on the nine that fail) is the humility the pipeline imposes on every exciting result the reader encounters. The reading rules are the series' skeptical toolkit in institutional form: the Phase 1 answers safety not efficacy, the Phase 2 signals and does not prove, the Phase 3 confirms conditionally, the Phase 4 delivers the population's truth, and the endpoints and the designs and the registration status are the questions the informed reader asks of every claim. The headlines skip the phases. The pipeline does not. The reader who knows the pipeline reads the headline's claim, locates it on the funnel, and discounts accordingly, which is the only literacy the drug-marketing era has ever required and the one this series has tried, across a hundred and seventy posts, to provide.
