Four intervention pages — a drug for weight (Semaglutide for Cardiovascular Risk in Obesity), a drug for lipids (Statins for Primary Prevention and the Power of Zero CAC), lifestyle weight loss (Does Weight Loss Reduce Cardiovascular Events), and blood-pressure lowering (Blood Pressure Lowering and Cardiovascular Events) — answer the same decision question, and the pattern is on none of them individually. Assembled, they make one claim that is a decision-change: for a genuinely low-risk primary-prevention person, no cardiometabolic intervention the wiki holds delivers a large absolute reduction in hard cardiovascular events. The first three are unproven there or below the action threshold; the fourth, BP-lowering, is the sharp test — its relative benefit is proven in primary prevention, and the claim survives only because a proven relative effect on a low absolute baseline is still a small absolute gain.
The four, on the one question
| Intervention | Hard-CV-outcome evidence, and where it applies | For a low-risk / primary-prevention person |
|---|---|---|
| Semaglutide (drug→weight) | MACE «hazard ratio, 0.80; 95% confidence interval, 0.72 to 0.90» — but only in secondary prevention (established CVD); primary prevention «were not studied» | Unproven. Weight + surrogates transfer; the CV-event benefit does not |
| Statin (drug→lipids) | Efficacy real and roughly constant per mmol/L; USPSTF acts only at ≥7.5–10% 10-year risk because benefit is proportional to risk; CAC=0 lifts NNT to 64 vs 28 | Small, below the threshold. USPSTF: below ~7.5% risk, not a candidate at all |
| Lifestyle weight loss | Look AHEAD «hazard ratio 0.95; 95% CI 0.83 to 1.09» — null in T2D despite real weight loss and a −42% CRP drop | Unproven / null via this route |
| BP-lowering (drug) | BPLTTC IPD: HR 0.91 (0.89-0.94) even without prior CVD, no heterogeneity down to <120 mmHg — relative benefit proven in primary prevention | Proven relative, small absolute. ~10% per 5 mmHg on a low baseline is a small absolute gain; the authors say treat on absolute risk |
Each per-page cross-source comparison already carries its own parameter table; this page configures those settled findings, it does not re-derive a magnitude.
Why they converge — three shared mechanisms, not a coincidence
- Absolute benefit scales with baseline risk (Baseline Risk and the Relative-Absolute Split). A constant relative effect on a low baseline is a small absolute effect. This is why USPSTF sets a risk threshold, why SELECT’s benefit would shrink ~5-fold at a primary-prevention baseline, and why a CAC=0 person’s statin NNT balloons. The machinery is identical across all three.
- The hard-outcome trials were run where the events are — in secondary prevention or high-risk populations, because that is where a trial can accrue events. So the evidence that exists is systematically about people at higher risk than the low-risk person asking. This is a structural feature of the evidence base, not a property of the drugs.
- Surrogate improvement does not carry the outcome (Surrogate Outcomes; The Certainty-Importance Inversion). Look AHEAD is the worked case: weight, HbA1c, fitness, blood pressure and CRP all moved the right way and the events did not follow. A good-looking risk-factor profile after an intervention is not evidence of an averted event.
The largest levers are missing from the table — because their evidence is observational
The obvious objection to a ceiling finding is “but exercise.” It is worth answering, because the answer sharpens the ceiling rather than breaking it. The levers with the largest associations to hard outcomes are not the four above — they are activity and fitness, and they are absent from the proven-lever table on purpose:
- Aerobic activity: all-cause HR ~0.34 at the top of the dose-response (Physical Activity Dose and Mortality) — the single largest association the vault holds, dwarfing every RCT-proven lever here.
- Cardiorespiratory fitness: RR 0.87 per 1-MET (Cardiorespiratory Fitness and Mortality).
- Muscle-strengthening: all-cause RR 0.85 (Muscle-Strengthening Activity and Mortality).
They are absent because the evidence is a different, weaker kind. All three are observational — no RCT randomizes a lifetime of activity or fitness onto hard outcomes — so they carry the confounding and reverse-causation load the four RCT levers were designed to escape (the healthy-user and sick-quitter signatures; cf. The U-Shaped Association Artifact). This is a distinction of evidence type, not a magnitude contrast: you cannot set HR 0.34 (observational, all-cause) beside HR 0.80 (RCT, MACE, secondary prevention) as commensurable — which is exactly why activity is a predictor here, not a proven causal lever.
The sharpened ceiling: it is not that no large lever exists. It is that the proven levers are small-for-low-risk, and the large lever (activity/fitness) is the one whose causal effect on hard outcomes is structurally unprovable by RCT. So the low-risk person’s honest position is unchanged — no proven large CV-event lever — even after the biggest association in the vault is on the table.
The decision-change — this is a ceiling finding
This is Layer 1 - Ranking Interventions for a Stratum’s the ceiling is itself a finding made concrete for cardiovascular events in a low-risk person: the search for a large, proven CV-event lever comes up empty across drug and lifestyle routes alike, and reporting that is a result, not a failure. It licenses a low-risk person to stop optimizing for a heart-attack reduction that the evidence does not promise them — and to pursue these interventions, if at all, for the outcomes they do move (weight, glycemia, diabetes prevention, MASLD, function, and — for higher-risk people — events).
Three things this does NOT say, each a guard against over-reading:
- Not these interventions do not work. They work where risk is high enough (SELECT in secondary prevention; statins above threshold) and on non-CV outcomes everywhere. The claim is scoped to hard CV events in a low-risk person.
- Not surrogates are worthless. They are the reason to act on many other outcomes, and they raise prior plausibility — they just are not proof of an averted event.
- Not a claim about high-risk people. For established CVD, high 10-year risk, or a high CAC, the same machinery runs the other way: absolute benefit is large and the interventions are strongly indicated. The pattern is a statement about the low-risk tail, where most of the wiki’s healthy strata sit.
What tested it — and why the test sharpened rather than broke it
BP-lowering was the candidate refutation, and it is instructive that it refined the claim instead. Unlike the other three, its relative hard-outcome benefit is directly proven in primary prevention (BPLTTC HR 0.91 without prior CVD, no heterogeneity across baseline BP). Had the synthesis claimed no intervention has a proven relative effect in primary prevention, BP-lowering would have broken it. But the claim is about large absolute benefit for a low-risk person, and there BP-lowering lands exactly where the mechanism predicts: the paper’s own decision rule is treat on absolute cardiovascular risk, not the blood-pressure number, so a proven ~10%-per-5-mmHg relative effect still buys little for someone at low absolute risk. The one intervention that could speak to primary prevention confirmed mechanism 1 rather than escaping it.
PREDIMED is the second confirming test — a HIGH-risk contrast, not a counterexample. A dietary pattern (Mediterranean diet, energy-unrestricted) cut CV events ~30% (HR 0.70) with a real absolute benefit of ~1.7-2.1 percentage points over 5 years (Mediterranean Diet and Cardiovascular Events). That looks like a lifestyle intervention with a proven hard-outcome effect — but it lands exactly where the mechanism predicts, because its population was high absolute risk (~49% T2D, ~82% hypertensive), and its authors say outright that generalization «to persons at lower risk requires further research». Like BP-lowering, it refines the ceiling: an intervention can buy hard-outcome benefit where baseline risk is high; it says nothing about the low-risk person, where the absolute benefit of the same relative effect shrinks toward the mechanism-1 floor. (And PREDIMED’s all-cause mortality was null over 4.8 yr, so even at high risk the event benefit did not extend to death in-window.)
What still would break it: a primary-prevention cardiovascular-outcome trial of an obesity drug (does not yet exist), a weight-loss-mortality review finding benefit beyond T2D, a CAC-randomized statin trial with benefit at CAC=0, or a dietary-pattern RCT showing hard-outcome benefit in a genuinely LOW-risk population — any would move a large-absolute-at-low-risk cell from unproven toward proven. Until then the pattern holds, and its confidence is high: five intervention classes across drug, diet-pattern and lifestyle — including the ones with the strongest primary-prevention evidence — all land in the same place by the same mechanism.