Affiliate disclosure
Book titles on this page link to Amazon. As an Amazon Associate, DataField.Dev earns from qualifying purchases — at no additional cost to you.
Chapter 2 — Further Reading
Free tools you should actually have open
PubMed — pubmed.ncbi.nlm.nih.gov
Free abstracts for essentially the entire biomedical literature. This is the single most useful link in this book. You will not always get full text, but you will get species, design, sample size, duration, and outcome — which is enough to place a study on the ladder, and enough to catch most headlines overstating their source. Learn the filters (article type, species, publication date). For: everyone. Twenty minutes learning to use this pays off for life.
Cochrane Library — cochranelibrary.com (plain-language summaries are free)
Systematic reviews conducted to an unusually rigorous and transparent standard, with explicit certainty assessments. Read several nutrition summaries and notice how often the conclusion is "the evidence is of low certainty." That's the honest state of the field, stated without a product attached — and it's the best available calibration for how confident you should be about anything. For: everyone; essential for students and clinicians.
The GRADE working group — gradeworkinggroup.org
The framework at the centre of the Case Study 2 controversy. Understanding what GRADE does — and what it systematically does to observational nutrition evidence — makes a large amount of expert disagreement suddenly legible. For: students, clinicians, and anyone who wants to understand why nutrition experts fight.
On the beta-carotene reversal
ATBC and CARET. Both trials are described in the published literature and summarized on the
National Cancer Institute site (cancer.gov). Read what the investigators themselves said about
the result — the tone of genuine surprise from serious scientists is more instructive than any
secondhand account.
For: everyone. This is the story the whole chapter turns on.
The Women's Health Initiative — whi.org and NIH summaries.
The clearest demonstration of healthy-user bias outside nutrition: large observational evidence of cardioprotection from hormone therapy, not confirmed by randomization. Reading how a whole clinical practice reversed is bracing in a useful way. For: anyone who wants to see the same shape in a different field.
On confounding and study design
Modern Epidemiology (Rothman, Greenland, Lash and colleagues; Wolters Kluwer, multiple editions).
The standard graduate reference. Dense, expensive, and definitive. If you're going into research or clinical practice, this is the book behind the book. For: students and professionals only. Do not start here.
Kenneth Rothman, Epidemiology: An Introduction (Oxford University Press).
The accessible entry point by one of the field's major figures. Short, clear, and genuinely readable for a motivated non-specialist. If §2.3 interested rather than repelled you, read this next. For: anyone who wants the real version of this chapter.
The literature on specification-curve and multiverse analysis. Rather than a single citation, search those terms alongside "epidemiology" or "nutrition." Seeing the full distribution of results from defensible analytical choices on one dataset is genuinely destabilizing in a productive way. For: the statistically curious; also for anyone who thinks one published paper settles anything.
On relative vs. absolute risk, and communicating uncertainty
Gerd Gigerenzer, Reckoning with Risk / Risk Savvy (Penguin; also published as Calculated Risks).
Gigerenzer's central argument is that people are far better at probabilistic reasoning than psychologists claimed — provided you present the numbers as natural frequencies rather than percentages. His worked examples of how relative-risk framing misleads doctors, not just patients, are the best treatment of §2.6 anywhere. For: everyone. Genuinely changes how you read health news.
David Spiegelhalter, The Art of Statistics (Pelican, 2019) and his public writing.
Spiegelhalter is a statistician who has spent a career on public risk communication, and he writes about uncertainty with unusual honesty and unusual clarity. His commentary on nutrition and cancer risk headlines is a model of the tone this book is aiming for. For: everyone; especially anyone who found the arithmetic in §2.6 satisfying.
On Mendelian randomization
The MRC Integrative Epidemiology Unit at Bristol (bristol.ac.uk/integrative-epidemiology)
publishes accessible explanations of MR alongside its research output; the group has been central to
developing the method.
MR is the most important addition to nutritional epidemiology's toolkit in decades and it's worth understanding its assumptions, not just its results — because when MR disagrees with observational data, the interesting question is always which assumption might be violated. For: students, clinicians, and anyone who wants to follow the alcohol argument in Chapter 12.
On the red meat controversy (Case Study 2)
The 2019 systematic reviews and recommendation, published in Annals of Internal Medicine, together with the accompanying editorial and the published correspondence responding to them. Read the reviews, the recommendation, and at least two critical responses. Then decide for yourself where the disagreement actually sits. For: anyone who wants to watch a genuine methodological dispute rather than read about it.
The World Cancer Research Fund / AICR Continuous Update Project (wcrf.org).
An ongoing, systematically updated synthesis of diet and cancer evidence, with explicit grading of how strong each conclusion is. Their assessments of red and processed meat are the mainstream counterweight to the 2019 panel, and the difference in how they weight the same evidence is instructive. For: everyone with a specific cancer-and-diet question.
On PREDIMED and self-correction
The PREDIMED retraction notice and the republished analysis, both in the New England Journal of Medicine, are publicly available. Read the retraction notice — it's short — and note what was corrected and what wasn't. For: anyone who wants to see what scientific self-correction actually looks like in practice, as opposed to how both its defenders and its detractors describe it.
Retraction Watch — retractionwatch.com. Covered this case in detail, and covers the ongoing
correction process across science generally.
One thing to be careful with
Search engines will happily supply you with a large genre of content titled some variant of "Everything You Know About Nutrition Is Wrong." Much of it correctly identifies the problems in §2.2–§2.4 — and then uses them selectively, to demolish findings the author dislikes while treating their own preferred evidence, which is usually weaker, as settled.
That's not evidence literacy. It's §2.8's funding-effects rule applied in one direction only.
The test for whether a critic is being honest: do they ever apply their own standards to a claim they like? If reading them has never once made you more confident about something, you're reading an advocate, not an analyst.