Abstract
Dear Editor:
T
Noninferiority
A noninferiority study states that the primary end point for the experimental treatment is worse than that for the positive control treatment by a prespecified margin, and rejects the null hypothesis at the prespecified level of statistical significance to support a claim of noninferiority. The p value was 0.36. The trial primary outcome is “inconclusive” and did not prove that withdrawing statins was safe.
Mortality
Sixty-day mortality was not a reasonable end point. Patients were excluded with the acute coronary syndrome. Only 58% had a history of cardiovascular disease (CAD). Statins are important to start with an acute coronary syndrome and reduce 30-day mortality. 2 For those with stable CAD, 23-month mortality was 4.9% for those adherent to statins and 14.9% for those nonadherent. 3 Sixty-day mortality was not influenced in this study that was six times larger than the statin study. 2 Absolute risk reduction in cardiovascular events or death was 8.7/1000 at 4 years between adherent and nonadherent patients in a study of 97,575 patients with stable CAD. 4 This difference could not possibly be detected in <400 patients. In patients without CAD, statins do not reduce cardiovascular events or deaths. The trial was very much underpowered and why the statisticians recommended >30,000 participants.
Ongoing Benefits from Statins After Withdrawal
Cardiovascular events and deaths do not occur the day after statin withdrawal, benefits are treatment duration dependent. The 60-day mortality outcome would, by design, likely “prove” noninferiority.
Noninferiority and Superiority Nonsignificance
The “inability to discern a difference in survival between patients” is to change the trial focus from noninferiority to a nonsignificant “superiority” trial and a null hypothesis of no difference as evidence of equivalent survival. It is a misconception to state that a nonsignificant difference means there are no differences.
So Why Was the Primary Outcome Not Reached?
The statin study used a one-sided alpha = 0.05 and 90% confidence interval for the differences to establish margins. Traditionally, confidence intervals for noninferiority trials are 97.5% for one-sided and 95% for two-sided margins. The margin for the statin study was biased toward noninferiority. Failure to meet end points was likely due to differences between groups. Individuals on the withdrawal arm had greater cognitive impairment (p = 0.02, 27% vs. 17%), suggesting greater disease severity.
Multiple Comparisons Need Correction
Secondary outcomes and group differences were assessed using a two-sided alpha of 0.05. Multiple outcomes were not corrected for multiple comparisons, which risks a type 1 error. Corrections could be done by a Bonferroni type method or gatekeeping methods.
Does Withdrawal of Statins Improve Quality of Life?
The McGill Quality of Life Questionnaire has established meaningful clinical important differences (MCIDs) of 0.66 for total score. The statin study had a mean difference in quality of life of 0.26. Although statistically significant (p = 0.04), this did not meet MCIDs. Just as important was the lack of difference quality of life between groups on the global question.
Conclusion
We believe there is enough evidence in the cardiovascular literature for a patient-centered approach to decisions regarding statins. Discussions should be informed by risks, CAD history, duration of statin therapy, prognosis, and goals of care. Individuals with acute coronary syndromes should be on statins. Individuals on long-term statins but short-term survival can safely discontinue statins. Individuals with CAD and one to more years of life expectancy should have a risk–benefit discussion. Individuals without CAD should discontinue statins. We believe that this stratified approach is a better approach when discussing withdrawal from statin therapy.