Incomparable Effectiveness—Apples to Apples?–Reply–I

We thank Drs Winter and Wheelan for their thoughtful comments and respond in detail. We first address the issue of statin dose. Our study adjusted for initial statin dose by matching patients with low, medium, and high doses on the basis of a 1:2 mg potency equivalence for atorvastatin vs simvastatin (see page 1066, lower right, in our article1). Because our data were drawn from real-world practice, a patient's initial statin dose tells us 2 things about the patient. To Drs Winter and Wheelan's point, the initial dose indicates the initial potency for cholesterol lowering. However, the initial dose also reflects physician-assessed cardiovascular (CV) risk because patients with higher risk, possibly due to higher baseline levels of LDL-C, would likely receive higher initial doses. Given these 2 interpretations of initial dose, and a per-milligram dose equivalence of atorvastatin vs atorvastatin between 1:2 and 1:4,2,3 how can we select the appropriate potency equivalents for matching adjustment in an observational study? if we consider only statin potency, then the lower 1:2 mg ratio used in our study appears biased against simvastatin, as Drs Winter and Wheelan describe. However, if we consider the initial statin dose as a marker of risk, the 1:2 mg matching ratio is conservative, and the 1:4 mg ratio would now appear biased against simvastatin because lower-risk patients with an initial dose of 10 mg of atorvastatin, for example, would then be matched to higher-risk patients with an initial dose of 40 mg of simvastatin. This latter consideration seems more relevant to our observational data, in which a patient's initial statin potency could be adjusted over time as his or her physician attempted to control cholesterol. To directly assess how our main conclusions might be affected by the potency ratio used for matching, we reanalyzed our study data using the 1:4 mg dose matching ratio suggested by Drs Winter and Wheelan. In particular, we matched patients whose initial dose of atorvastatin was 10 mg to those whose initial dose of simvastatin was 40 mg and matched patients whose initial dose of atorvastatin was 20 mg to those whose initial dose of simvastatin was 80 mg. Patients whose initial dose was different were necessarily excluded because no other available dose of these drugs can provide a 1:4 mg ratio. In addition to this modified exact matching on dose, all other propensity score and exact factor matching criteria were applied as in our original analysis. This matching process resulted in 4385 matched pairs, 3929 at the lower doses and 456 at the higher doses, which were well-balanced on all measured characteristics (Table 1). The baseline rate of prior inpatient CV events was 8% in both treatment groups (exactly matched by design), and the baseline total costs were numerically higher for the atorvastatin- vs simvastatin-initiated patients ($7984 vs $7605; P=.62) (Table 2). During the 2-year postindex period, atorvastatin-initiated patients experienced a relative 19% lower risk of CV events compared with the matched simvastatin-initiated patients (7.3% vs 9.1%; P=.003) and lower average total costs by $153 per patient ($15,121 vs $15,274; P=.82) (Table 3). TABLE 1. Baseline Characteristics Between Matched Cohorts of Employees Initiating Atorvastatin vs Simvastatina,b TABLE 2. Baseline Cardiovascular (CV) Events, Health Care Resource Use, and Costs for Matched Cohorts of Employees Initiating Atorvastatin or Simvastatina TABLE 3. Comparison of Outcomes Between Matched Cohorts of Employees Initiating Atorvastatin vs Simvastatina These results are markedly more favorable for initiation with atorvastatin vs simvastatin than the results of our original analyses, in which atorvastatin was associated with a 9% reduction in CV risk and higher average total costs by $41 per patient. However, it is likely that the new results described herein reflect confounding due to lower-risk patients initiated with atorvastatin at 10 or 20 mg being matched to higher-risk patients initiated with simvastatin at 40 or 80 mg. We conclude that the results of our original analyses are appropriately conservative in using the lower 1:2 mg dose matching ratio. The letter by Drs Winter and Wheelan raised 2 further points. First, we agree that having cholesterol data would have been ideal. However, our study adjusted for other important risk factors, such as age, diabetes, chronic kidney disease, wage, and prior inpatient admissions with CV diagnoses. Second, although prior studies of secondary prevention have shown no differences among statins, our study population mainly consisted of patients who needed primary prevention, with only 15% having a prior inpatient or outpatient diagnosis of a CV event during the baseline period. As suggested in the editorial by Drs Culler and Weintraub,4 commenting on our article, the large quantity of data available from the employer claims used in our study provides an excellent opportunity to explore subgroup effects. We have recently investigated the effects of atorvastatin vs simvastatin in subgroups of employees stratified by CV risk.5