Association of Baseline Statin Use Among Older Adults Without Clinical Cardiovascular Disease in the SPRINT Trial

Debate continues about the benefits and risks of statins for primary prevention, especially among older adults.^1,2 Recently, older adults with moderate hyperlipidemia and hypertension who were randomized to initiate statin therapy for primary prevention, instead of receiving usual care, experienced no benefit.³

However, patients already taking statins were excluded from that trial.³ This could matter if such patients were systematically at higher risk. To understand the impact of statins without such a potential exclusion bias, I reanalyzed the Systolic Blood Pressure Intervention Trial (SPRINT; NCT01206062).⁴ SPRINT sought to test whether a treatment program aimed at reducing systolic blood pressure (BP) to a lower goal than currently recommended would reduce cardiovascular disease (CVD) risk among patients without diabetes.

Statin use was not randomized in this trial. Accordingly, inverse probability weighted (IPW) methods were used in treatment effects and survival models to adjust for nonrandom statin use.

I obtained SPRINT blood pressure trial data from the National Heart, Lung, and Blood Institute with institutional review board approval from Penn State College of Medicine.

Participants with baseline diagnosed clinical CVD as defined by the trial were excluded. The main results also exclude those younger than 70 years; sensitivity analysis considers 65 years and older. Main results pool BP management arms; sensitivity analyses consider each arm separately.

All analyses were performed in Stata/MP release 15.0 (StataCorp LLC) and controlled for baseline age, sex, race, smoking status, systolic and diastolic BP, number of BP agents, aspirin use, glucose, cholesterol, high-density cholesterol, triglycerides, serum creatinine, estimated glomerular filtration rate, ratio of urine albumin to creatinine, and BP management randomization.

In treatment effects IPW models, the outcome was the primary composite event rate over the trial, comprising myocardial infarction, other acute coronary syndrome, stroke, heart failure, or cardiovascular-related death. Statin assignment was specified as a logistic function of all baseline characteristics above.

In survival time IPW models, the outcome was the timing of the primary composite event. Statin assignment was specified as described; censoring, as an exponential function of all baseline characteristics. Adequate covariate balance was attained in all models.

Of total 9361 patients in SPRINT, 3054 were 70 years or older and met the inclusion criteria, of whom 1350 (44.2%) were taking statins at baseline. Those on statins, compared with those not on statins, had significantly lower Framingham risk scores but were more likely to be male, have chronic kidney disease, higher body mass index, and higher glucose (Table 1). There was no significant difference in primary event rates with or without statin use at baseline: 0.064 vs 0.078 over trial (absolute rate reduction, 0.014; P = .14).

In Table 2, the results of models adjusting for nonrandom assignment of statin use are shown. In the treatment effects IPW model, among patients 70 years or older, the average event rates with and without statin use at baseline were 0.063 vs 0.081, equivalent to an absolute risk reduction of 0.018 over the trial (P = .13). In the survival IPW model, the average time to event with and without statin use at baseline were 669.2 vs 735.8 days, a decrease of 84.6 days (P = .22) and thus an increased hazard rate.

Changing age thresholds to 65 years or older did not materially alter these results (Table 2), nor did restricting analyses to BP management arms, using a Weibull censoring function in IPW models, a propensity score matched treatment effects model, logistic regression, or simpler Cox models that did not adjust for nonrandom statin use.

Participants in the SPRINT trial 65 or 70 years or older without diagnosed CVD who were taking statins at baseline had no significant differences in primary outcomes compared with those not taking statins with or without adjustment for nonrandom statin use. These nonrandomized results contradict randomized trials of statins such as JUPITER⁵ and HOPE-3.⁶

This study is limited by its reliance on a nonrandomized comparison and secondary analysis for which the original trial was not powered, a lack of time-varying data on statin use, and the potential inadequacy of conditioning only on observed variables to model the nonrandom assignment of statin use. Moreover, given low event rates, even in a 1-sided test of the crude primary event rates, smaller differences than 3% cannot be distinguished; while for a 1-sided test of times to primary event, reductions in the hazard ratio less substantial than 0.77 cannot be distinguished.

We still do not have sufficient numbers of primary prevention trials to make strong recommendations about statins in intermediate-risk populations (6%-12% 10-year risk), at least on the basis of survival. Yet even in this relatively high-risk older adult population (22%-25% 10-year risk), significant reductions in cardiovascular events were not found.

Accordingly, until the Australian STAREE randomized trial (NCT02099123) of statin use among older adults concludes,¹ this study lends some support to the concerns increasingly raised about benefits and harms of statins among older adults at higher risk of CVD.^1,3

Corresponding Author: Marco D. Huesch, MBBS, PhD, Department of Radiology, Penn State Milton S. Hershey Medical Center, 500 University Dr, Hershey, PA 17033-0850 (mhuesch@pennstatehealth.psu.edu).

Accepted for Publication: November 13, 2017.

Published Online: January 22, 2018. doi:10.1001/jamainternmed.2017.7844

Conflict of Interest Disclosures: None reported.