The evidence

Show the work.

Every figure on this site traces to a named study. This page shows the strongest of them, and what they do not say.

The population finding

A short list of factors
carries most of the risk.

1 INTERHEART, 52 countries, 29,972 participants (Yusuf, Lancet, 2004): nine modifiable factors accounted for over 90% of population-attributable first-MI risk: 90% in men, 94% in women. INTERSTROKE, 32 countries (O’Donnell, Lancet, 2016): ten factors, ~90.7% of stroke risk.

In the largest case-control study of its kind, nine modifiable factors (lipids, smoking, blood pressure, diabetes, abdominal obesity, diet, exercise, alcohol, and psychosocial stress) accounted for over 90% of the population-attributable risk of a first heart attack, across 52 countries.1

The stroke literature reached the same conclusion from a different direction: ten modifiable factors accounted for ~90.7% of population-attributable stroke risk across 32 countries.

2 Population-attributable risk estimates the share of cases statistically tied to a factor if that factor were removed entirely. It describes populations. It cannot promise any one person an outcome.

Then the honest counterpart. A 2023 prospective pooled cohort (measuring five of these factors forward in time, rather than backward from events) put the attributable share at 52.6% in men and 57.2% in women. Case-control designs run higher; cohort designs run conservative. Both figures are printed here, because the finding is about populations, not a promise about any one person, and either way, the risk is largely measurable and modifiable.

90%

of population-attributable first heart-attack risk: nine modifiable factors, INTERHEART, 52 countries

90.7%

of population-attributable stroke risk: ten modifiable factors, INTERSTROKE, 32 countries

57%

the conservative prospective-cohort estimate, women (52.6% in men), printed because you deserve both numbers

The therapeutic record

What treatment has
actually delivered.

The levers that have moved events in randomized outcome trials are few: lower the burden of atherogenic particles,2 and quiet the arterial inflammation that remains after lipids are controlled.3 The record is short enough to print in full.

Lipids · the CTT meta-analyses

≈20–25% fewer major vascular events per 1 mmol/L of LDL-C lowering.

The Cholesterol Treatment Trialists’ meta-analyses pooled the major trials and found the same relationship however the reduction was achieved: the relative benefit tracks the absolute LDL-C reduction. It is the most dose-consistent finding in cardiovascular therapeutics.

Inflammation · CANTOS · 2017

15% fewer major events, with no lipid change.

In patients whose inflammation persisted after lipid control, an anti-inflammatory antibody reduced major adverse cardiovascular events: the first randomized proof that inflammation is a treatable axis, not merely a marker.

COLCOT · 2019

4,745 patients after heart attack: 23% reduction.

Low-dose colchicine, a generic anti-inflammatory, reduced major events when started soon after infarction.

LoDoCo2 · 2020

5,522 patients with chronic coronary disease: 31% relative reduction.

The same generic drug, in stable disease, produced the largest relative reduction of the three trials. Inflammation is not a residual detail; it is a target.

The absolute-value check

36%1.1%

A worked example from the corpus: a 36% relative reduction can translate to a 1.1% absolute one

63vs217

Statin NNT (number needed to treat) per year: secondary prevention versus primary

The rule on this site: relative and absolute figures travel together. A relative number quoted without its absolute counterpart is marketing, and you have been marketed at long enough.4

Positions we do not take

What we decline to claim.

Prevention has an internet, and the internet sells certainty. Four popular positions fail our bar, not because they are unfashionable, but because the evidence does not carry them.

  1. Routine prolonged fasting as therapy.

    The human evidence is short-horizon and built on surrogate markers; no outcome trial has tied extended fasting to fewer events, and for some patients it carries real risk.

  2. High-saturated-fat ketogenic diets as heart-safe.

    Losing weight and being heart-safe are different claims; a diet that raises atherogenic particle counts owes long-term outcome data that does not yet exist.

  3. Vitamin K2 for calcium scores.

    Tested directly in a randomized trial, and it did not slow calcium-score progression; the evidence we have offers no reason to prescribe it.

  4. The claim that LDL does not matter.

    The chain from particle number to plaque to events is the most replicated causal relationship in cardiovascular medicine; we are not interested in relitigating it.

What we do not know yet

The frontier is measured,
not promised.

Some of what we propose to measure sits ahead of the guidelines. The inflammasome-marker panel is best understood as a service with an embedded research registry, not a settled standard of care. It is measured to build the evidence, under consent, and it is labeled that way.

Decisions today run on guideline-grade markers. The proposed assessment inventory (ApoB, a one-time Lp(a), inflammatory markers, advanced lipids, and coronary calcium scoring) is ordered against a single rule: only when the result can change what we do. Final selection is an individual clinical decision, made with you.5

If the registry earns a place in routine care, this page will say so, with figures. If it does not, this page will say that instead.

Engraved pressure waveforms stacked in fine pale lines on a black field, a single trace picked out in arterial red, a study of the pressure wave.
Plate VI · The Capillary FieldEngraving, 2026

Trust, measured

Why this page exists.

Trust in medicine has fallen far enough to measure. We are not asking you to reverse that on our say-so; this page (its figures, its caveats, its refusals) is the argument.

71.5%40.1%

Trust in physicians and hospitals · April 2020 to January 2024 (Perlis et al., JAMA Network Open)6

93%85%

Share who trust their own physician · 93% in June 2023, 85% at the most recent measure (KFF)

The answer to a trust deficit is not branding. It is showing the work: every claim named, every source cited, every caveat printed beside the claim it qualifies.

References · this page

The full ledger.

  1. INTERHEART: Yusuf S, et al. Lancet, 2004. Nine modifiable factors accounted for over 90% of population-attributable first myocardial-infarction risk across 52 countries (90% men, 94% women). INTERSTROKE: O’Donnell MJ, et al. Lancet, 2016. Ten factors, ~90.7% of stroke risk across 32 countries. Conservative prospective counterpart: five factors, 52.6% of cases in men and 57.2% in women (NEJM, 2023). Population-attributable risk is a population finding, not an individual guarantee.
  2. Cholesterol Treatment Trialists’ meta-analyses: ≈20–25% relative reduction in major vascular events per 1 mmol/L of LDL-C lowering.
  3. CANTOS (Ridker, 2017): anti-inflammatory therapy reduced major events 15% with no lipid change. COLCOT (Tardif, 2019): 4,745 patients post-MI, 23% reduction. LoDoCo2 (Nidorf, 2020): 5,522 patients with chronic coronary disease, 31% relative reduction.
  4. Absolute and relative figures are reported together throughout this site. Worked corpus examples: a 36% relative reduction can translate to 1.1% absolute; statin NNT is 63 per year in secondary prevention versus 217 in primary prevention.
  5. The proposed assessment inventory (ApoB, Lp(a), inflammatory markers, advanced lipids, coronary calcium scoring) describes a practice design pending launch; final test selection is an individual clinical decision.
  6. Perlis RH, et al. JAMA Network Open: trust in physicians and hospitals, 71.5% (April 2020) to 40.1% (January 2024). KFF: trust in one’s own physician, 93% (June 2023) to 85%.