The statin decision in primary prevention is a baseline-risk decision. The relative effect of a statin is roughly constant; the absolute benefit scales with how high the risk was to begin with (Baseline Risk and the Relative-Absolute Split). Two sources set the decision: USPSTF 2022 (US Preventive Services Task Force, 2022) gives the efficacy magnitudes and a risk threshold; Nasir/MESA 2015 shows a zero coronary-calcium score re-stratifies about half of statin-eligible people into a range where the benefit is small.
Statin efficacy in primary prevention (USPSTF 2022)
A systematic review of 22 trials (mean follow-up 3.3 years, all enrolling ≥1 risk factor):
| Outcome | RR (95% CI) | Absolute risk difference |
|---|---|---|
| All-cause mortality | «0.92 [95% CI, 0.87 to 0.98]» | «−0.35%» |
| Composite CVD | «0.72 [95% CI, 0.64 to 0.81]» | «−1.28%» |
| Myocardial infarction | «0.67 [95% CI, 0.60 to 0.75]» | «−0.89%» |
| Stroke | «0.78 [95% CI, 0.68 to 0.90]» | «−0.39%» |
| CV mortality | «0.91 [95% CI, 0.81 to 1.02]» (NS) | «−0.13%» |
The relative effects are real and consistent; the absolute differences are already small in these higher-risk trial populations. Harms were reassuring — no increase in serious adverse events, myalgia, or (except high-intensity JUPITER, confined to those with diabetes risk factors) new diabetes.
(US Preventive Services Task Force, 2022)
The USPSTF decision rule — and where a low-risk person falls
The threshold is explicit, and it is a 10-year-risk threshold, because «the magnitude of benefit of statin use is proportional to a person’s estimated 10-year CVD risk»:
- 40–75 yr + ≥1 risk factor + 10-yr risk ≥10% → prescribe a statin (Grade B, «at least a moderate net benefit»).
- 7.5% to <10% → «selectively offer» (Grade C, «at least a small net benefit»); «The likelihood of benefit is smaller in this group».
- <7.5%, or ≥76 yr → no recommendation / insufficient evidence.
So a primary-prevention person well below 7.5% 10-year risk is not a statin candidate by USPSTF at all — the drug question is closed by the baseline risk before any imaging is considered. (Two carve- outs USPSTF names: this does «not pertain to adults with familial hypercholesterolemia or an LDL-C level greater than 190 mg/dL» — those are treated regardless.) USPSTF scores risk with the Pooled Cohort Equations, which it concedes overpredict in many populations.
USPSTF does not use CAC: it «addressed the use of coronary artery calcium score for CVD risk assessment in a separate recommendation» — i.e. deliberately kept it out of the statin rule.
(inferred from US Preventive Services Task Force, 2022)
The power of zero (Nasir / MESA 2015)
Where a body does allow CAC, a zero score moves the decision. In 4,758 MESA adults (45–84 yr, median 10.3-yr follow-up), applying the 2013 ACC/AHA thresholds recommended 50% for statins and considered 12% more — and CAC re-stratified them sharply:
- 58% of the whole cohort, and 44% of the statin-eligible, had CAC = 0. Among statin-eligible people with CAC=0 the ASCVD rate was «4.2 per 1,000 person-years», versus «11.2 per 1,000 person-years» with any CAC. In the recommended group specifically, CAC=0 carried «5.2 ASCVD events/1,000 person- years»; in the considered group, «1.5 per 1,000 person- years».
- Estimated 10-year NNT (applying an assumed 30% relative reduction to the observed rates): in the recommended group, 64 for CAC=0 vs 28 for CAC>100; in the considered group, 223 vs 46 (Nasir et al., 2015).
- The headline: «the absence of CAC reclassifies approximately one-half of candidates as not eligible for statin therapy», resting on the principle that «the net benefit from treatment is directly proportional to the absolute risk».
The premise underneath — CTT supplies the constant-relative-effect backbone [2026-08-05]
This page’s opening move — the relative effect of a statin is roughly constant; the absolute benefit scales with baseline risk — is asserted from USPSTF’s composite. CTT 2010 (IPD meta-analysis, 26 trials) is the direct evidence for the constancy: per 1.0 mmol/L LDL-C reduction the RR is ~0.78 and «did not depend on the baseline LDL cholesterol concentration», holding across prior-CHD/none, diabetes, age >75, BP, BMI and HDL baseline-risk strata with no material heterogeneity (a nominal sex difference aside) — including a significant «proportional risk reduction of 25% (99% CI 18-31; p<0·0001) per 1·0 mmol/L reduction in LDL cholesterol in participants with no previous history of vascular disease» (the primary-prevention stratum this page is about). (Cholesterol Treatment Trialists’ Collaboration, 2010) So the constant-relative / absolute-scales-with-baseline logic that drives the whole CAC-deferral argument is the cleanest route-(a) case the corpus holds -> Baseline Risk and the Relative-Absolute Split, LDL Lowering and Cardiovascular Events.
Units, kept distinct (parameter discipline). CTT’s RR ~0.78 is per 1.0 mmol/L LDL-C reduction; USPSTF’s composite RR 0.72 and Nasir’s assumed 0.70 are per statin regimen (a fixed treatment contrast, not per-mmol). They are complementary, not the same quantity — CTT gives the dose-scaled backbone, USPSTF the per-regimen effect a given statin delivers. (inferred from Cholesterol Treatment Trialists’ Collaboration, 2010)
Over 75 without known vascular disease — the one cell the constancy premise does not reach [2026-10-06]
CTT 2010’s age >75 line above pooled primary and secondary prevention. The CTT 2019 age analysis (28 trials, 186,854 people, 14,483 of them >75 at entry) splits that line, and the split matters for this page, because this page is about people without known vascular disease.
- With vascular disease, the relative effect holds at every age: >75 RR 0.85 (99% CI 0.73-0.98) per 1.0 mmol/L, age trend p=0.2. (Cholesterol Treatment Trialists’ Collaboration, 2019)
- Without known vascular disease, >75: RR 0.92 (99% CI 0.73-1.16) on 295 vs 308 events (2.7% vs 2.8%/yr), and the 71-75 group was 0.84 (0.70-1.01). (Cholesterol Treatment Trialists’ Collaboration, 2019) The age trend in this arm did not depend on the heart-failure/dialysis trials: «However, among participants with no history of vascular disease, we observed a significant trend towards smaller proportional risk reductions with increasing age (ptrend=0·05; figure 4), which persisted after exclusion of the heart failure and dialysis trials (ptrend=0·03; appendix).» (Cholesterol Treatment Trialists’ Collaboration, 2019) In the Research-in-context panel the authors call it «a weak trend towards smaller relative risk reductions with increasing age in the primary prevention setting (although there were too few such older participants for reliable assessment in that group alone).» (Cholesterol Treatment Trialists’ Collaboration, 2019)
- Their own conclusion keeps both halves: «There is less definitive direct evidence of benefit in the primary prevention setting among patients older than 75 years, but evidence supports the use of statin therapy in older people considered to have a sufficiently high risk of occlusive vascular events.» (Cholesterol Treatment Trialists’ Collaboration, 2019)
Parameter table — four sources on the >75 cell.
| Parameter | CTT 2010 | CTT 2019 | USPSTF 2022 | ACC-AHA 2026 | Same quantity? |
|---|---|---|---|---|---|
| Object | MVE RR per 1.0 mmol/L | MVE RR per 1.0 mmol/L | recommendation grade | recommendation (COR 2b, LOE B-NR) | 2010 vs 2019: yes. USPSTF, ACC-AHA: no, judgements, not effects |
| Stratum | >75 at entry, prior vascular disease pooled in | >75 at entry, split by prior vascular disease | >=76, primary prevention, initiating | >75, primary prevention, initiating, life expectancy >=2.5 y | 2019 primary-prevention row roughly matches the guidance rows (>75 vs >=76) |
| Value | 16% reduction (99% CI 3-27), i.e. RR ~0.84 | 0.87 (0.77-0.99) all; 0.92 (0.73-1.16) no vascular disease | I statement (insufficient evidence) | «may be reasonable» after a clinician-patient discussion; supporting text cites CTT 2019’s pooled >=75 figure (13% RRR, coronary-driven) | — |
| Trials | 26 | 28; 2010 set nested inside, plus CORONA and HOPE-3 | own review; the held recommendation statement does not cite CTT | cites CTT 2019 (pooled >75, not the primary-prevention arm) | 2010 vs 2019 not independent |
Loci: CTT 2010 «16% (99% CI 3–27; p=0·002) in people older than 75 years at entry» (Cholesterol Treatment Trialists’ Collaboration, 2010); USPSTF «The USPSTF concludes that the current evidence is insufficient to assess the balance of benefits and harms of initiating a statin for the primary prevention of CVD events and mortality in adults 76 years or older.» (US Preventive Services Task Force, 2022); ACC-AHA «RCT data for adults aged >75 years are limited but support primary prevention with statin therapy to reduce ASCVD events, although relative reductions in risk may be lower than in younger populations, and there is no evidence of overall mortality benefit.» (Blumenthal et al., 2026)
What the table supports. The 2019 paper refines the 2010 line (type-F, same collaboration and nested trials, not independent backing). Pooled, the >75 effect looks like everyone else’s; split, the primary-prevention half carries a 99% CI running from a 27% reduction to a 16% increase. That is the insufficient-evidence state, not a no-effect finding. On the expectancy test, silence here tells us little: about a fifth of the over-70 events came from people without vascular disease, too few for the authors to call the >75 estimate reliable (no power calculation is given). Against that, the attenuation trend in this arm is nominally significant and survived removal of the heart-failure/dialysis trials (p=0.05, then 0.03), so a smaller relative effect at older ages is a live possibility, not noise to wave away; it is one of many subgroup tests, which is why the paper reports subgroup CIs at 99%. (inferred from Cholesterol Treatment Trialists’ Collaboration, 2019)
The two guidance bodies lean differently, both at their weakest grade. USPSTF gives an I statement for starting a statin at >=76. ACC-AHA 2026 says starting a moderate-intensity statin «may be reasonable» (COR 2b) for adults >75 with life expectancy of at least 2.5 years, after a clinician-patient discussion: «In adults aged >75 y with an estimated life expectancy of at least 2.5 y, it may be rea sonable to initiate moderate-intensity statin therapy after a clinician–patient discussion of potential benefits and risks to reduce ASCVD risk.4–6,7» (Blumenthal et al., 2026) Its supporting text leans on CTT 2019’s pooled >=75 result (primary and secondary prevention together), not on the primary-prevention arm (0.92, 0.73-1.16 (Cholesterol Treatment Trialists’ Collaboration, 2019)): «statin therapy in this group produced a 13% RRR in major vascular events per 1.0 mmol/L (38.7 mg/dL) reduction in LDL-C, with reductions driven by fewer major coronary events but not overall mortality, vascular death, or stroke.» (Blumenthal et al., 2026) CTT itself concludes the evidence supports statins in older people at sufficiently high risk. CTT and USPSTF agree on the evidence state (thin, imprecise), not on a verdict. So the difference between the bodies is a direction of lean at the lowest grades, not a firm split, and part of ACC-AHA’s lean rests on a pooled figure that the primary-prevention split does not reproduce. A 99% CI that includes both the all-age effect and no effect leaves room for both leans. (Blumenthal et al., 2026; inferred from Cholesterol Treatment Trialists’ Collaboration, 2019; US Preventive Services Task Force, 2022)
Decision consequence for a >75 person with no vascular disease. The relative benefit is not shown to be absent, and absolute risk is higher at this age, which is why CTT models ~80 vs ~50 events avoided per 10,000 per year at 78 vs 63. But that model assumes the relative effect carries over to this exact stratum, which is what the trial data cannot confirm. The honest frame is a shared decision in which the expected benefit is uncertain and the higher baseline risk is the reason to discuss it at all. Continuing a statin someone already takes is a different question from starting one: USPSTF’s I statement is about initiating. CTT does not test stopping; ACC-AHA cites one RCT, in people with estimated life expectancy under 1 year (primary and secondary prevention mixed), in which stopping was safe and quality of life improved, and it says a CAC score of 0 in older adults suggests benefit is unlikely to outweigh risk (a COR 2b recommendation that also covers CAC 1-10, which its own text says «remains to be tested in prospective clinical trials») (Blumenthal et al., 2026). That brings this page’s CAC lever into the >75 cell, as an untested strategy. Ongoing trials in this stratum are named as a gap on the source page. (Blumenthal et al., 2026; inferred from Cholesterol Treatment Trialists’ Collaboration, 2019; US Preventive Services Task Force, 2022)
Confidence scope. The page’s confidence: high covers the constant-relative / baseline-risk logic
for primary prevention up to about 70 and for secondary prevention at any age. It does not extend
past 70 in primary prevention: CTT found no independently significant reduction in either over-70 group
(71-75: 0.84, 0.70-1.01; >75: 0.92, 0.73-1.16), and the >75 cell sits in the insufficient-evidence state.
The honesty on both sides — this is not a settled rule
-
Nasir’s NNT is modeled, not measured. It multiplies observed CAC-stratified event rates by a borrowed 30% relative reduction (from a Cochrane statin meta-analysis), not by a statin effect measured within a CAC-randomized trial. The authors say so: whether a «CAC-based strategy versus guideline-based recommendations for statin selection… will have a favorable impact on outcomes… needs to be critically tested by well-designed comparative-effectiveness clinical studies.» No RCT has randomized statins by CAC.
-
CAC=0 is low risk, not no risk. 4.2/1,000 py is not zero, and NNT 64 is not infinity — a CAC=0 statin candidate still has some modeled benefit. In diabetes, CAC=0 still carried 4.9/1,000 py (NNT 69), so the de-risking is weaker there.
-
The 30% figure ties the two sources together and they agree. Nasir’s assumed relative reduction and USPSTF’s measured composite-CVD effect are the same quantity, and they match:
Parameter Nasir (assumed) USPSTF (measured) Same quantity? Statin relative reduction, composite CVD 30% (RR 0.70) «0.72 [95% CI, 0.64 to 0.81]» (28%) Yes — Nasir’s input is consistent with USPSTF’s trial estimate -
Population bounds: MESA is 45–84 (so it says nothing about adults <45) and may select healthier volunteers; USPSTF’s ARDs come from trial populations enriched for risk factors.
The unresolved divergence — lodged, not adjudicated
Whether CAC should guide the statin decision is contested across guidance families, and the evidence does not settle it (no RCT):
- USPSTF keeps CAC out of the statin rule (insufficient evidence, separate recommendation).
- ESC allows CAC as a Class IIb modifier around thresholds (Risk Modifiers - When Extra Information Changes a Risk Estimate).
- ACC/AHA and the MESA community use CAC=0 to defer statins in the grey zone (this page’s Nasir data is that case).
This is a genuine joined-issue candidate (risk-score-threshold vs imaging-guided allocation) but is not adjudicated here — it AWAITS a source that either randomizes CAC-guided treatment or is an independent institutional appraisal of the imaging-vs-score question.
ACC-AHA 2026 goes furthest — CAC mapped directly onto LDL-C goals, and a lower start threshold [2026-08-06]
The 2026 US dyslipidaemia guideline (superseding the 2018 cholesterol guideline USPSTF sits alongside) operationalizes CAC further than any body above: it maps the Agatston score directly onto a treatment goal, not merely onto a defer/treat call. CAC screening is recommended in «men at least 40 years of age and women at least 45 years of age».
| CAC (Agatston) | ACC-AHA 2026 action | LDL-C goal | COR |
|---|---|---|---|
| >=1000 AU | statin first-line, >=50% reduction | <55 mg/dL (1.4 mmol/L) | 1 |
| >=300-999 AU | >=50% reduction (reasonable to intensify to <55) | <70 mg/dL (1.8 mmol/L) | 1/2a |
| >=100-299 AU or >=75th pctile | treat | <70 mg/dL (1.8 mmol/L) | 1 |
| 1-99 AU and <75th pctile | mod-intensity statin, >=30-49% reduction | <100 mg/dL (2.6 mmol/L) | 2a |
(Blumenthal et al., 2026) So the same CAC=0 → defer logic this page holds (Nasir) is joined, at the top end, by CAC≥1000 → treat-as-very-high (goal <55) — the symmetric upward move, and CAC>300 ≈ secondary-prevention risk, which ESC-EAS 2025 also notes.
And ACC-AHA 2026 starts LLT LOWER than USPSTF — a within-US divergence traced to the calculator. USPSTF prescribes at ≥10% 10-year risk (Pooled Cohort Equations, which it concedes overpredict) and selectively offers at 7.5-<10%. ACC-AHA 2026 uses the newer PREVENT-ASCVD equations and can act well below that: LLT «can be considered» at borderline (3-<5%) and «should be considered» at intermediate (5-<10%), ages 30-79.
| Parameter | USPSTF 2022 | ACC-AHA 2026 | Same quantity? |
|---|---|---|---|
| Calculator | Pooled Cohort Equations (overpredicts) | PREVENT-ASCVD | No |
| ”Prescribe” threshold | ≥10% 10-y risk | high (≥10%); treat-to-goal | partial |
| Lower action band | 7.5-<10% “selectively offer” | 3-<5% consider / 5-<10% should-consider | No — PREVENT reads lower than PCE for the same person |
So the two US bodies’ thresholds are not directly comparable — PREVENT and the Pooled Cohort Equations give different numbers for the same person (PREVENT was built partly because PCE overpredicts), so a «3-<5% PREVENT» is not a looser «7.5% PCE». This is the same different-instrument distinction as PREVENT-vs-SCORE2 -> SCORE2 Baseline Risk and the ESC Treatment Thresholds, not a genuine evidence disagreement. The direction of the change is still notable: ACC-AHA 2026 is willing to start pharmacotherapy at lower estimated risk than USPSTF, and to name a numeric goal once it does. (inferred from Blumenthal et al., 2026; US Preventive Services Task Force, 2022)
The common statin fears — muscle and memory [2026-08-22]
The two most-cited reasons people avoid or stop a statin — muscle damage and memory loss — are BOTH addressed by USPSTF’s trial evidence, and both point the same way in the general adult population: «Evidence from trials in the general adult population shows that statins are not associated with an increased risk of myalgia, elevated alanine aminotransferase level, or cognitive harms compared with placebo». (US Preventive Services Task Force, 2022) So the muscle-symptom and the memory halves of the belief are contradicted by the randomized evidence, not merely unstudied — a REFUTE from held fabric, not an insufficient-evidence gap.
- Serious muscle harm is real but narrow. Rhabdomyolysis — the dangerous end of the muscle spectrum, distinct from ordinary aches — was rare and confined to the highest dose (80 mg simvastatin), quantified on LDL Lowering and Cardiovascular Events. It is a dose-specific tail, not a general property of statin therapy.
- Scope caveat: USPSTF flags that harms evidence specifically in adults 76 years or older is limited, so the reassurance is strongest for the general adult population it was measured in.
The nocebo mechanism — now held, not a gap [2026-08-23]. No association in blinded trials is
compatible with real-world muscle complaints being a nocebo effect (symptoms driven by the
expectation of harm, not the drug), and SAMSON now supplies the dedicated blinded evidence that had
been routed to acquire — the earlier [belief-harvest: statin-nocebo] gap is CASHED. SAMSON is a
blinded three-group n-of-1 trial (n=60) in exactly the hard stratum: people who had «previously
discontinued statins because of side effects» within 2 weeks of starting. Each patient rotated
month-by-month through atorvastatin 20 mg, placebo, and no-tablet, scoring symptoms daily 0-100.
- The symptoms are real but not statin-specific. Mean symptom intensity was 8.0 on no-tablet months, «15.4 during placebo months (95% CI, 12.1 to 18.7; P<0.001 for the comparison with no-tablet months), and 16.3 during statin months (95% CI, 13.0 to 19.6; P<0.001 for the comparison with no-tablet months and P = 0.39 for the comparison with placebo months)». (Wood et al., 2020) Taking a tablet at all roughly doubled symptom burden over no-tablet (both P<0.001); statin vs placebo was statistically indistinguishable (P=0.39).
- The magnitude — the nocebo ratio. «In patients who had discontinued statin therapy because of side effects, 90% of the symptom burden elicited by a statin challenge was also elicited by placebo.» (Wood et al., 2020) The pooled nocebo ratio was 0.90 (the per-patient analysis gave 2.2 with a wide CI −62.3 to 66.7, driven by small/negative denominators; the pooled reanalysis is the reported estimate).
- Decision-relevant downstream. «Half the trial patients were able to successfully restart statins» — 30 of 60 (50%) six months after the trial. (Wood et al., 2020) This is a naturalistic restart rate, not a tested intervention: SAMSON did not randomize showing-patients-their-own-data, so the 50% is a downstream observation, and any rescue-by-feedback reading is a plausible inference the trial does not isolate.
Scope guard — SAMSON does NOT contradict the general-adult null; it explains it. SAMSON’s stratum is the symptomatic-intolerant (already discontinued for side effects), not statin-naive adults, so it answers why these people report symptoms — the piece USPSTF’s population-level null left open — rather than restating the null. The two compose: statins do not raise myalgia over placebo in the general population (USPSTF), and where an intolerant person nonetheless reports muscle symptoms, ~90% of that burden is reproduced by an inert pill (SAMSON). One small n-of-1 trial in a selected stratum: the transferable finding is the mechanism, not a population effect size.
The randomised magnitude sizes the fear — and it is front-loaded [2026-09-07]. The CTT
Collaboration’s IPD meta-analysis of ~155,000 people in double-blind statin RCTs puts a number on the
real causal excess: any muscle pain or weakness ran «rate ratio [RR] 1·03; 95% CI 1·01–1·06», with
the excess «during year 1 … an absolute excess rate of 11 (6–16) events per 1000 person-years,
which indicates that only one in 15 ([1·07–1·00]/1·07) of these muscle-related reports … were
actually due to the statin. After year 1, there was no significant excess». (Reith et al., 2022)
So the muscle-symptom fear that drives primary-prevention non-adherence is small in size, mostly
nocebo, and confined to year 1 — reinforcing the USPSTF null with a randomised magnitude rather than
just a no-association. The full tolerability/deprescribing decision (who should stop, the
year-1-vs-later split, the dose-de-escalation lever) lives on
Statin Muscle Symptoms and the Nocebo Effect.
(Reith et al., 2022; inferred from US Preventive Services Task Force, 2022; Wood et al., 2020)
Decision relevance
- Get the 10-year risk first; the threshold is the decision. Below ~7.5%, USPSTF says not a statin candidate, and CAC only reinforces deferral (a low-risk person is very likely CAC=0 anyway).
- In the grey zone (≈7.5–20%), a zero CAC roughly halves-to-quarters the absolute benefit (NNT 64 vs 28) and is a legitimate deferral input where a body allows it — but the deferral rests on modeled, not trial, benefit, so it belongs in shared decision-making, not an automatic rule.
- CAC=0 de-risks the DECISION, not the biology. ApoB/LDL remain causal; a high or rising ApoB, a high Lp(a), or familial hypercholesterolemia can still warrant treatment regardless of a zero score, and FH is explicitly carved out of the whole framework.
- The statin is the pharmacotherapy comparator to the lifestyle lipid lever — the lipid-axis analogue of Semaglutide for Cardiovascular Risk in Obesity on the weight axis. For a low-risk, CAC=0 person the realistic alternative to a statin is lifestyle plus monitoring, not nothing and not a drug with a fraction-of-a-percentage-point absolute benefit.
(Nasir et al., 2015; inferred from US Preventive Services Task Force, 2022)
Limits
- The CAC-guided-statin question has no RCT — the standing open-loop (R1) problem in a sharp form.
- Nasir is one observational cohort with a modeled NNT; USPSTF is a strong systematic review but its ARDs are population-specific and it declines to rule on CAC.
- Neither source quantifies the CAC=0 negative predictive value by age band below 45 — the gap Risk Modifiers - When Extra Information Changes a Risk Estimate already flagged.