What Actually Prevents Disease

TL;DR. After forty-odd chapters of mechanisms and treatments, the honest summary is that a small number of things account for most of the preventable disease in this book, and they are boring. Do not smoke. Keep blood pressure down. Stay physically active. Keep weight in a reasonable range with food that is mostly not ultra-processed. Drink little or no alcohol. Sleep. Get vaccinated. Attend the handful of screening programmes that have proven mortality benefit. Almost everything sold as prevention beyond that list has either no evidence or evidence of harm. The other half of the answer is not individual at all: the largest gains in human health came from clean water, sewers, food safety, road design, and tobacco policy, not from anything a person chose for themselves.

Key takeaways

• Smoking cessation is the highest-value individual action in medicine. Quitting before 40 avoids almost all of the excess mortality; quitting at any age helps.
• Roughly 30 to 50 percent of cancers are attributable to modifiable risk factors, and around 80 percent of premature heart disease and stroke is considered preventable.
• Geoffrey Rose's prevention paradox: a small shift in the whole population's risk prevents more disease than a large shift in the high-risk minority, and the population approach offers each individual very little, which is why it is politically hard.
• Screening is only worth it when a trial has shown reduced disease-specific mortality, and only a handful of programmes clear that bar.
• Most supplements do nothing, and several have caused harm in trials: beta-carotene increased lung cancer in smokers, vitamin E increased prostate cancer, high-dose antioxidants have repeatedly failed.
• The interventions with the largest population effect are regulatory: tobacco taxation, salt reformulation, clean air standards, seatbelts, and water treatment.

The short list, in rough order of expected benefit

In short: Eleven actions account for most of the preventable disease in this book, and the first four dwarf everything else.

That is most of it. The remainder of this chapter explains why the list is short, why the things not on it are not on it, and what the population-level version looks like.

Rose's insight: sick individuals and sick populations

In short: Most cases arise in the large majority at moderate risk, so shifting the whole population slightly prevents more disease than treating the high-risk few.

Geoffrey Rose observed in the 1980s that most cases of most diseases arise not in the high-risk minority but in the large majority at moderate risk, simply because there are so many more of them. Most babies with Down syndrome are born to younger mothers, because younger women have most of the babies. Most heart attacks occur in people with only moderately raised cholesterol.

Two strategies follow:

The high-risk strategy finds the people at greatest risk and treats them. It is efficient per person treated, well-matched to clinical medicine, motivating for both doctor and patient, and it misses most of the cases.

The population strategy shifts the whole distribution slightly: a little less salt in bread, a little less sugar in drinks, a slightly higher tobacco price, slightly better air. It prevents far more disease in total.

This produces the prevention paradox: a population measure brings large benefits to the community and offers little to each participating individual. Almost nobody who wears a seatbelt is saved by it, and seatbelts save enormous numbers of lives. That asymmetry explains why population prevention is politically difficult (people resent restrictions from which they perceive no personal gain) and why it is nonetheless where the biggest wins are.

Smoking, the single largest lever

In short: Quitting before 40 avoids around 90 percent of the excess risk, and taxation is the most effective population measure ever tested.

Tobacco kills over 7 million people a year through direct use and more than a million through secondhand smoke. Roughly half of long-term smokers die of a smoking-related disease.

The British Doctors Study, following 34,000 male doctors from 1951 for fifty years, provided the numbers that still anchor the field: lifelong smokers lost about 10 years of life expectancy, and the benefit of stopping was strongly age-dependent. Stopping at 60, 50, 40, or 30 gained roughly 3, 6, 9, or 10 years respectively.

What works to help people stop (Chapter 43): varenicline, combination nicotine replacement, cytisine, bupropion, e-cigarettes, and behavioural support, each roughly doubling to tripling success, and working best in combination. Willpower alone succeeds a few percent of the time, which is why treating smoking as a character test rather than a treatable dependence wastes most of the available benefit.

What works at population level, in descending order of evidence: taxation (the single most effective measure, with consumption falling roughly 4 percent for every 10 percent price rise in high-income countries and more in lower-income ones), smoke-free public places, plain packaging and graphic warnings, advertising bans, and free cessation services. The WHO Framework Convention on Tobacco Control has been ratified by most countries, and the measurable declines in smoking prevalence in countries that implemented it fully are among the clearest population health results available.

Movement

In short: The curve is steepest at the bottom, so going from nothing to something matters more than going from moderate to hard.

The dose-response curve for physical activity is steep at the bottom and flattens at the top. Moving from sedentary to light activity produces a bigger mortality reduction than moving from moderate to high activity. The guideline targets (150 minutes a week of moderate activity or 75 of vigorous, plus two resistance sessions) are a reasonable destination and a poor starting instruction.

Recent accelerometer-based studies have found substantial mortality reductions at step counts well below the folkloric 10,000, with benefit accruing from roughly 4,000 steps a day and continuing to improve up to around 8,000 to 10,000. The 10,000 figure originated in a 1960s Japanese pedometer marketing campaign rather than in physiology.

Resistance training deserves separate emphasis because it is chronically neglected. It preserves muscle and bone, improves insulin sensitivity, reduces falls, and is associated with reduced mortality independently of aerobic exercise.

Sitting is a partially separate risk: prolonged uninterrupted sitting is associated with worse metabolic outcomes even in people who meet activity guidelines, though the effect is attenuated at higher activity levels.

Food

In short: Dietary patterns beat individual nutrients, and most single-nutrient advice has failed when tested properly.

The evidence here is weaker and noisier than for smoking or blood pressure, because nutrition research relies heavily on observational data with substantial confounding, and because randomised diet trials are difficult and short. What survives that scrutiny:

• Dietary patterns beat individual nutrients. Decades of single-nutrient trials have mostly disappointed, while pattern-based evidence, especially for Mediterranean-style eating, has held up, including in a randomised trial with hard cardiovascular endpoints.
• Reduce sodium. Population salt reduction lowers blood pressure and, in modelling and in observed national programmes, cardiovascular deaths. The UK's voluntary reformulation programme reduced average intake measurably.
• Reduce sugary drinks. The most consistently implicated single dietary item for weight, diabetes, and dental disease, and the easiest to displace.
• Reduce ultra-processed food. The controlled feeding trial showing people ate about 500 more calories a day on an ultra-processed diet at matched nutrients gives this a causal basis (Chapter 19).
• More fibre, legumes, nuts, whole grains, vegetables, and fruit. The most consistent positive associations across cohorts and the most plausible mechanisms.
• Less processed meat. Classified as a group 1 carcinogen by IARC for colorectal cancer, with a modest absolute effect that is regularly misreported as equivalent in magnitude to smoking. It is not; the classification refers to the strength of evidence that it causes cancer, not to the size of the risk.
• Alcohol. The apparent benefit of moderate drinking has largely dissolved under better comparison groups and Mendelian randomisation methods. Cancer risk rises from low levels of intake. Several countries have revised guidance downward substantially.

What is not supported: most supplements in people without deficiency, detox regimens, alkaline diets, and the great majority of specific "superfood" claims.

Vaccination across the lifespan

Vaccination is the most cost-effective clinical prevention available and is thought of as a childhood matter, which leaves most adults under-protected.

Screening: which ones earn their place

In short: Only a handful of programmes have shown reduced deaths from the disease, and several popular ones have not.

A screening programme is worth running only if it reduces deaths from the disease, at acceptable harm. Judged by that standard rather than by intuition:

The recurring lesson from Chapter 17: finding disease early is intuitively good and empirically variable, and the only way to know is a trial with mortality as the endpoint.

Things that sound like prevention and are not

In short: Several supplements have caused harm in trials, and general annual check-ups have not reduced mortality.

• Antioxidant supplements. Beta-carotene increased lung cancer incidence and mortality in smokers in two large trials, both stopped early. Vitamin E increased prostate cancer risk in the SELECT trial. High-dose antioxidants have never delivered the benefits observational data suggested.
• Multivitamins in well-nourished people. No reduction in cardiovascular disease, cancer, or mortality in large trials.
• Vitamin D in the general population. Clearly worth treating in deficiency, and large trials in unselected adults showed no reduction in cancer or cardiovascular events.
• Fish oil supplements for primary prevention. Repeatedly negative in large trials, in contrast to eating fish.
• Routine annual "check-ups" as a general practice. Randomised trials of general health checks have not found reductions in mortality, though targeted screening embedded within them does work. The value of a check-up lies in the specific proven components, not the ritual.
• Detoxification and cleanses. The liver and kidneys already do this, continuously, and no commercial product improves on them.
• Most microbiome supplements. Probiotics have specific evidence for a few narrow indications and are marketed for everything.

Prevention that is not individual at all

In short: Clean water, sewers, vaccination, salt iodisation, road safety law, and tobacco tax saved more lives than every operating theatre in history.

The largest health gains in history came from things no one chose personally.

Two features recur: these work on everyone regardless of motivation, and they reduce inequality, whereas information campaigns aimed at individual behaviour tend to widen it because better-off people respond first.

A framework for judging any prevention claim

Five questions, which dispose of most of what you will encounter:

1. What is the outcome? A change in a blood test is not the same as fewer deaths, fewer strokes, or fewer fractures (Chapter 17).
2. Randomised or observational? Nutritional and lifestyle claims are overwhelmingly observational and therefore confounded by the fact that people who do healthy thing X also do healthy things Y and Z.
3. What is the absolute benefit? Ask for the number needed to treat, or the change in a 1-in-100 risk, not a percentage.
4. What are the harms? Screening has harms. Supplements have harms. Preventive drugs have harms. A prevention discussion without them is incomplete.
5. Who benefits from you believing it? Not disqualifying, and informative.

A realistic personal plan by decade

In short: What to actually do in your twenties, forties, sixties, and seventies, which is a different list at each stage.

Not a prescription, and a reasonable default shape.

20s and 30s: do not start smoking, and stop if you have. Establish an activity habit including resistance training. Get HPV vaccination if eligible. Know your blood pressure. Contraception and sexual health testing as relevant. Address mental health early, since most lifetime psychiatric illness begins in this window. Do not accumulate concussions.

40s: blood pressure, lipids, and glucose checked. Weight trajectory matters more than weight at any single point. Begin colorectal screening at 45 in most guidelines. Alcohol review. Treat sleep apnoea if snoring with daytime sleepiness.

50s and 60s: continue the above. Cancer screening programmes as offered. Shingles and pneumococcal vaccination at the recommended ages. Bone health assessment if there are risk factors or a fracture. Hearing tested, both for its own sake and because untreated hearing loss is the largest modifiable midlife dementia risk factor. Resistance training becomes more important, not less.

70s and beyond: falls prevention (strength and balance work, medication review, vision, home hazards) becomes the single highest-value activity. Annual vaccination. Medication review and deprescribing. Maintain social connection deliberately, since isolation carries mortality risk comparable to established physical risk factors. Keep moving.

The honest caveat

Nothing on this list guarantees anything. Fit non-smokers get cancer, and people who did everything wrong live to 95. Prevention shifts probabilities across populations; it does not write individual futures. The reason to do it anyway is the same reason to wear a seatbelt: over a life and over a population, small shifts in probability accumulate into a great many years.

And the second honest caveat: a large share of what determines your health was set by where you were born, what your parents earned, what your childhood nutrition was, what your air is like, and whether your work is secure. Individual advice matters and is not a substitute for the conditions that make it possible.

Sources and notes

Smoking mortality and cessation gains: Doll et al., BMJ, 2004 (British Doctors Study, 50-year follow-up), and Jha et al., NEJM, 2013. Tobacco deaths: WHO. Blood pressure effect sizes: Ettehad et al., The Lancet, 2016. Physical activity dose-response: Arem et al., JAMA Internal Medicine, 2015, and accelerometer-based meta-analyses of step counts (Paluch et al., Lancet Public Health, 2022). PREDIMED: Estruch et al., NEJM, 2018. Ultra-processed food trial: Hall et al., Cell Metabolism, 2019. Alcohol and Mendelian randomisation: Millwood et al., The Lancet, 2019. Rose's prevention paradox: Rose, International Journal of Epidemiology, 1985, and The Strategy of Preventive Medicine, 1992. Vaccination deaths averted: Shattock et al., The Lancet, 2024. Beta-carotene harm: ATBC (NEJM, 1994) and CARET (NEJM, 1996). Vitamin E and prostate cancer: SELECT, JAMA, 2011. Vitamin D: VITAL, NEJM, 2019. General health checks: Krogsbøll et al., Cochrane review, 2019. Ovarian cancer screening: UKCTOCS, The Lancet, 2021. Screening programme effect sizes as cited in Chapter 25. Folic acid fortification: CDC and national programme evaluations. Tobacco taxation elasticity: World Bank and WHO analyses.

Open questions. Optimal dietary patterns remain contested in detail even where the broad shape is agreed. Whether population-wide sodium reduction benefits everyone equally is debated. The net value of prostate and breast screening continues to divide serious experts.

Next: the operational version of everything above, with the numbers, ages, and checklists. 👉