Cardiovascular mortality was identical in both arms: 2.4 events per 1000 person-years, atorvastatin versus placebo, a hazard ratio of exactly 1.00.
That number sits inside a trial whose headline says statins cut major cardiac events by 30% in healthy adults over 70. Both facts are true. The distance between them is the whole story.
For four decades, the trials that built statin guidelines quietly excluded the people now most likely to be offered a statin: independently living adults over 70. Fewer than a quarter of participants in the pooled analyses that underpin current recommendations were older than 70, and the US Preventive Services Task Force still calls the evidence insufficient to judge risks and benefits in anyone 76 and older.
STAREE closed that gap the hard way. Monash University researchers mailed atorvastatin or placebo to 9,971 Australians aged 70 and up—mean age 74.7, 52% women, more than 40% already past 75—recruited through general practices, with anyone carrying prior cardiovascular disease, diabetes, or dementia screened out. Median follow-up ran 5.9 years.
Then it did something rare: it hit one primary endpoint spectacularly and missed the other entirely.
The win is real and robust. A major cardiovascular event—cardiovascular death, nonfatal MI, stroke, or coronary revascularization—occurred in 297 people on atorvastatin versus 412 on placebo. That is 6.0% versus 8.3%, a hazard ratio of 0.70 (95% CI 0.61–0.82, p<0.001).
The number needed to treat was 37 over 5.9 years to prevent one such event. This is RCT-grade evidence, the highest tier, in exactly the population that never had it.
The benefit held up across the fault lines you'd worry about. It was similar in the 70–74 group and the 75-plus group, and consistent across baseline LDL, blood pressure, BMI, and kidney function. This matters because the sharpest prior objection—"we have no idea if it works past 75"—no longer holds.
Notably, these were not people with alarming cholesterol. Mean baseline LDL was about 127 mg/dL (3.27 mmol/L), and atorvastatin dropped it by roughly 48 mg/dL against 16 mg/dL on placebo.
So far, this reads as a clean mandate. Look harder at what drove the 30%.
The reduction was predominantly in nonfatal events. Myocardial infarction fell sharply (HR 0.57; 95% CI 0.43–0.75), as did coronary revascularization (HR 0.57; 95% CI 0.45–0.72). Stroke moved in the right direction, but the confidence interval crossed 1 (HR 0.87; 95% CI 0.68–1.11).
And cardiovascular mortality did not move at all: 2.4 events per 1000 person-years in both arms, HR 1.00 (95% CI 0.72–1.37). The statin prevented heart attacks and stents. It did not, in this trial, prevent cardiovascular death.
The statin prevented heart attacks and stents. It did not, in this trial, prevent cardiovascular death.
That sets up the miss. STAREE's second primary endpoint was disability-free survival—survival free of death from any cause, dementia, or persistent physical disability. On that, the drug did nothing statistically detectable.
Death, dementia, or persistent disability occurred in 637 atorvastatin recipients and 676 on placebo: 12.8% versus 13.6%, HR 0.94 (95% CI 0.84–1.05, p=0.25). Fewer heart attacks did not translate into more years lived independently.
The investigators offered the arithmetic behind the disconnect plainly: 80% of the deaths in the trial were from noncardiovascular causes. When four in five deaths come from something a statin can't touch, preventing heart attacks barely shifts the composite that combines death, dementia, and disability.
This is the pivot the headline hides. A 30% relative reduction sounds like a life-extending intervention. It is, on this evidence, an event-reducing one—and in a population whose survival is governed mostly by cancers, falls, infections, and the accumulated burden of age, reducing cardiac events is not the same as changing how long or how well someone lives.
Abdulla Damluji, who chairs the American College of Cardiology's Geriatric Cardiology Leadership Council, put the takeaway bluntly: statins in independently living older patients produce "event reduction, but not preserved independence."
His prescribing corollary is the sentence guidelines should be built around. "The curves separate after roughly 2 years," he noted, "so a patient unlikely to survive 5 more years, or who is already frail and carrying substantial polypharmacy, may gain little from a statin."
Read the NNT through that lens. Thirty-seven people treated for nearly six years to prevent one major event, with a two-year lag before any benefit appears, means the entire calculation collapses for someone whose remaining life expectancy is short. You cannot bank a benefit that starts in year two if you won't see year five.
That is why the right guideline verb is authorize, not mandate. The trial's discussant, François Mach of Geneva University Hospital, drew exactly that line: "Older age should no longer be used as a reason to withhold statin therapy for primary prevention," but "STAREE does not mean that every person 70 years old or older should automatically receive a statin treatment."
The decision variable, in Mach's framing, is "life expectancy, competing risk, treatment burden, and patient priority"—not the number on a birth certificate. Age gets you into the conversation; it should not settle it.
On safety, STAREE largely defuses the fears that kept older patients off these drugs. Serious adverse events occurred in 2.7% of participants in both arms—identical. Musculoskeletal, hepatobiliary, and diabetes-related events were more common on atorvastatin, but the muscle signal was modest: 32.0% versus 29.4%, in a cohort where roughly half already reported musculoskeletal complaints at baseline.
Damluji's clinical read is worth anchoring for patients: "most new aches after starting will not be the drug, which is precisely why a planned rechallenge or dose de-escalation beats abrupt discontinuation."
There is also a clean null worth stating precisely, because it functions as the RCT-grade counterpoint to years of observational worry. Dementia occurred at 10.6 versus 10.3 events per 1000 person-years, HR 1.03 (95% CI 0.87–1.21)—no signal of harm, and no signal of benefit. Whatever the mechanistic debates, this trial found nothing.
The limits are real and should temper the enthusiasm. More than 15% of participants stopped their medication during the trial, which biases a real-world effect toward the null and means adherent patients may do somewhat better than the intention-to-treat numbers suggest. The cohort was almost exclusively white with a mean age of 75, so generalizability beyond that profile is genuinely uncertain.
And the trial was not powered for mortality alone—the flat cardiovascular-death and disability-free-survival results are consistent with "no benefit," but a smaller mortality effect can't be fully excluded.
Where that leaves the guidelines is narrower and more honest than the press release. STAREE has earned statins a place on the menu for healthy over-70s; it has not earned them a place on the schedule.
The single number to carry into the exam room is not 30%. It is the two-year lag against the patient's remaining life expectancy—the variable that decides whether an NNT of 37 is a gift or an empty gesture.
Prescribe accordingly. Authorize, don't automate.