Chapter 5 — Further Reading

This chapter's reading list is the most useful one in the book. Everything here is free, and several items will change how you read health news permanently.

Tier 1 — Verified canonical

Free tools you can use today

ClinicalTrials.gov (U.S. National Library of Medicine). The largest trial registry. Search any compound and see what trials exist, their phase, status, enrollment, and — critically — their registered primary endpoint. Comparing that to what a paper reported is the single highest-yield check available to a non-specialist, and it takes about two minutes.

The WHO International Clinical Trials Registry Platform links national registries worldwide, which matters for compounds developed outside the United States (Chapter 23's Semax and Selank, for example).

PubMed (National Library of Medicine). Free indexing of the biomedical literature, with abstracts. Search a compound name; sort by publication type; filter for "Randomized Controlled Trial" and see how many there actually are. For several compounds in this book, that filter returns nothing, and seeing it yourself is more persuasive than being told.

Cochrane Library. Systematic reviews conducted to an unusually rigorous and transparent standard. Abstracts and plain-language summaries are free. The plain-language summaries are genuinely written for non-specialists and are a good model of how to state an uncertain conclusion honestly.

DailyMed and Drugs@FDA — approved labels, which contain the trial data regulators actually evaluated. See Chapter 4's reading list.

On method

The ICMJE (International Committee of Medical Journal Editors) publishes its recommendations freely, including the trial registration requirement described in Case Study 2.

The CONSORT statement is the standard checklist for reporting randomized trials, published free. Reading the checklist is an efficient way to learn what a well-reported trial contains — and to notice what a poorly reported one omits. PRISMA is the equivalent for systematic reviews.

The Cardiac Arrhythmia Suppression Trial (CAST) was publicly funded and reported in the late 1980s; it remains one of the most-cited negative trials in medicine and is discussed in essentially every clinical epidemiology textbook.

Books

Ben Goldacre, Bad Science and Bad Pharma. Bad Science is the better introduction — accessible, funny, and unusually good on how a plausible study design fails. Bad Pharma is specifically about publication bias, missing trials, and the structural problems Case Study 2 describes. Both are polemical, and reading them alongside an industry perspective is worthwhile.

Trisha Greenhalgh, How to Read a Paper. The standard practical guide, written for clinicians and readable by anyone. If you want one book that teaches §5.5 through §5.7 properly, this is it.

Goodman & Gilman, again, for the pharmacological context.

Tier 2 — Attributed, specifics unverified

On attrition in drug development. The proportion of compounds entering human trials that never reach approval is commonly cited as roughly nine in ten, varying substantially by therapeutic area. This book states it as a proportion rather than a decimal.

On Phase II optimism. That Phase II effect sizes systematically exceed subsequent Phase III results is well described in the drug development literature, attributed to small samples, favorable populations, selection of the best-performing dose, and regression to the mean. Magnitudes vary by field.

On publication bias magnitudes. Studies auditing trial registries against subsequent publications have repeatedly found substantial non-publication rates and meaningful rates of discrepancy between registered and reported primary endpoints. Specific percentages vary considerably by cohort, field, and era, and none is quoted here.

On the STEP program. Semaglutide 2.4 mg weekly in adults with overweight or obesity without diabetes produced mean weight change of about −15% from baseline at 68 weeks, versus roughly −2.4% on placebo, with both arms receiving lifestyle intervention. In adults with type 2 diabetes, the corresponding figure was roughly −10% — a population difference that popular coverage routinely elides.

On SELECT. Semaglutide 2.4 mg in adults with established cardiovascular disease and overweight or obesity but without diabetes, roughly 17,000 participants followed about three years: a 20% relative reduction in major adverse cardiovascular events, hazard ratio about 0.80; in absolute terms roughly 8% to 6.5%, about 1.5 percentage points. Manufacturer-sponsored.

On SURMOUNT-1. Tirzepatide 15 mg in adults with obesity without diabetes over 72 weeks: about −21% on the treatment-regimen estimand and about −22.5% on the efficacy estimand. Both figures are correct and describe the same trial.

On animal study registration. Preclinical animal research is not systematically registered, and initiatives to introduce preclinical registration exist but have not achieved the coverage that human trial registration has. This asymmetry is the basis of §5.3's fifth reason and of Case Study 2's central point.

Tier 3 — Illustrative and constructed

• Figure 5.4 ("the rat tendon") is explicitly labeled a composite — real pattern, constructed specifics. The study design and result described are typical of the genuine preclinical literature on healing peptides; the specific numbers are constructed for teaching.
• The three-baseline-risk table in §5.8 is a constructed teaching example demonstrating arithmetic, not data from any trial.
• The evidence ladder, the surrogate-failure diagram, and the estimand box are this book's own teaching devices.
• The red-flag checklist in §5.10 is assembled by this book from standard critical-appraisal practice.

If you only do one thing

Pick one compound from your dossier. Search ClinicalTrials.gov. Then search PubMed with the "Randomized Controlled Trial" filter on.

Write down two numbers: how many registered trials exist, and how many completed randomized controlled human trials have published.

For an approved drug, both numbers will be large and the picture will be reassuring. For several compounds discussed confidently across the internet, the second number is zero.

Nobody has to tell you that. You can see it yourself, free, in about five minutes, and once you have seen it for one compound you will never again accept "there are hundreds of studies" without asking which kind.

Looking ahead

Chapter 6 covers the hype cycle — how a real preclinical result becomes a sales page in five predictable stages. No preparation needed. But if you want to arrive with something concrete, take the compound you just looked up and search for it outside the scientific literature: forums, product pages, video content. Hold the two pictures side by side. The gap between them is Chapter 6's entire subject.