Statins Work in Older Adults, Statistics Be Damned

The STREAM trial, presented at the American Heart Association Scientific Sessions, gave patients and clinicians a clear answer to the question of whether stopping statins in older patients has a cost — despite its confusing statistical analysis. The point of trials is to show what treatments work. In many if not most cases, the better therapy or strategy is obvious. Poring over the statistical analyses helps most when the outcomes are close, where the results might be due to chance. Statistics should clarify, not confuse. The results of the STREAM trial are clear; the statistical analysis not so much. The STREAM Statin Trial In the STREAM trial of more than 1800 older adults (mean age 77) with at least 2 comorbidities who were taking statin drugs for primary prevention, a strategy of stopping statins was compared to continuing statins. The thinking behind this clinical question is sound: older adults may have enough competing risks (eg cancer, frailty, etc) to render the typical statin reduction of major adverse cardiovascular events (MACE) meaningless; or, older adults with multimorbidity may be akin to patients with advanced heart failure or end-stage renal disease, two conditions where statins proved ineffective; or, statin side effects in older people may impair quality of life enough to counter the benefit in MACE reduction. The Swiss-led team decided to measure a primary outcome of myocardial infarction (MI), stroke, or death from any cause. They then chose to analyze the difference in outcomes with non-inferiority — meaning success would occur if event rates in the statin discontinuation arm were not that much worse. Quality of life measures, pain, falls, muscle function, and sarcopenia were also measured as counters to the allowance of slightly worse cardiac events. Sadly, the trialists used a complicated way of measuring outcomes. Instead of simply counting MIs, strokes, and deaths, they reported restricted mean survival times (RMST), which is the area under the event-free survival curve up to 2 years in this study. Or put another way, how many of those 104 weeks were spent event-free on average over all patients. The RMST is an absolute measure. ResultsThe RMST for the primary outcome was 98.5 weeks in the continuation arm vs 96.5 weeks in the discontinuation arm. The –2.0 week difference had 95% confidence intervals that went from –3.7 to –0.3 weeks. Since the upper bound is below 0, we can say that patients who discontinued statins had “significantly less” event-free survival. The event rates support that claim: A primary outcome occurred in 12.7% of patients in the discontinuation arm vs 9.2% in the continuation arm; rates of MI, stroke, and death from any cause were all higher with stopping. It is clear that statin discontinuation was associated with significantly more major adverse cardiac outcomes than continuation. The upper bound of the hazard ratio allowed for 81% more events (HR, 1.38; 95% CI, 1.05-1.81).As for the secondary outcomes, there were no differences in quality of life, pain, number of falls, or sarcopenia. After reviewing the report of the last interim analysis, the data and safety monitoring board recommended that patients in the discontinuation arm be advised to restart their statin. Explaining the Claim of Noninferiority And yet the conclusion in the New England Journal of Medicine is that discontinuation of statin therapy was noninferior to continuation. This is misleading. Let me explain: The trialists estimated a primary outcome event rate of 34% at 2 years, based on a trial in similar patients. They chose a noninferiority margin of 5.2 weeks of lost event-free time, which translates to about 8 or 9 extra events per 100 patients. In relative terms, which is important, the 5.2 weeks approximates to a hazard ratio of about 1.3. The problem is that the actual primary outcome event rate was way lower than 34%, at 10%. The fixed margin of 5.2 weeks is now far too permissive — those 8 or 9 extra events would represent a doubling of the 10% baseline rate. And that is how a 38% higher rate of events could achieve something rare in trials: statin discontinuation could be both statistically noninferior and inferior. A thought experiment shows why this trial was all but guaranteed to declare noninferiority. Imagine statins were 100% effective at eliminating MACE. Over the 2-year window, patients who stopped statins averaged 96.5 event-free weeks, so they lost about 7.8 weeks to events. But only about 3.5 of those weeks involved MACE; the rest were mostly noncardiovascular deaths, which statins cannot prevent. So even a perfect statin would have produced a gap of only about 3.5 weeks between the groups, less than the 5.2-week noninferiority margin. The solution to this problem would have been to test noninferiority with both absolute and relative margins — as is conventional. Using the 34% event rate the trialists anticipated, their 5.2-week margin corresponds to a hazard ratio of about 1.3. Had they also required the result to fall within that relative margin, discontinuation would have clearly missed noninferiority. The upper bound of the confidence interval was 1.81, and even the point estimate of 1.38 exceeds 1.3. Other Observations Also Favor Statin Continuation The trial found no other benefits of statins. Quality of life, pain, and sarcopenia were not improved in the statin discontinuation arm. There was no counter-benefit to stopping statins, other than taking one less pill per day. The author’s statement that patients who stopped statins had their events (on average) about 2 weeks earlier invites readers to misread what RMST measures. No one’s heart attack simply arrives 2 weeks early; the 2-week figure is an average spread across everyone. The absolute risk increase for the full follow-up of 2 years is 3.5%, which translates to a number needed to harm from stopping statins of 28. That is a meaningful difference. ConclusionThis trial, along with the positive STAREE trial, finds that the reliably consistent cardiovascular benefit of statin drugs also applies to older patients who have yet to have a cardiac event. This is a new and useful discovery as so few older patients were included in statin trials. Whether the amount of benefit is enough to take a daily pill is up to the patient. To make an informed decision, however, both patient and clinician need an accurate assessment of the evidence. The conduct of the STREAM trial and its findings are fine. The problem comes in its flawed analysis and interpretation of the data. Ignore the claims of noninferiority. We can simply say that statin drugs provide patients like those in the trials a clinically meaningful reduction in the risk of MI, stroke, and death. John Mandrola practices cardiac electrophysiology in Louisville, Kentucky, and is a writer and podcaster for Medscape. He espouses a conservative approach to medical practice. He participates in clinical research and writes often about the state of medical evidence.