Study design and population

IMPROVE-IT was a randomized, double-blind, placebo-controlled trial that enrolled 18,144 patients hospitalized within the preceding 10 days with ACS with an LDL-C level of 50-100 mg/dL (if chronically taking prior prescription LLT) or 50-125 mg/dL (otherwise) (6,8). Patients were randomized to receive either ezetimibe 10 mg plus simvastatin 40 mg (ezetimibe/simvastatin) or matching placebo plus simvastatin 40 mg once daily (placebo/simvastatin), and were followed up for a median of 6 years (interquartile range [IQR]: 4.3-7.1 years). During the trial, patients in either treatment group who had LDL-C levels >79 mg/dL on 2 consecutive measurements had the simvastatin dose increased to 80 mg. In 2011 (after enrollment completed), in accordance with the U.S. Food and Drug Administration guidance for limiting the use of simvastatin 80 mg, patients who had been receiving simvastatin 80 mg for <1 year were dose-reduced to 40 mg, and future uptitration to 80 mg simvastatin was not permitted. Key exclusion criteria were baseline ezetimibe use in combination with a statin, treatment with a lipid-lowering regimen more potent than simvastatin 40 mg daily, clinical instability, creatinine clearance <30 mL/min, or active liver disease. Other aspects of study design and methods were reported previously (6,8). All patients provided written informed consent, and the protocol was approved by the ethics committee at each participating center.

Endpoints

In this analysis, the prespecified efficacy endpoints of the trial were used (6,8). The primary efficacy endpoint was a composite of death from CV disease, a major coronary event (nonfatal myocardial infarction [MI], documented unstable angina requiring hospital admission, or coronary revascularization occurring at least 30 days after randomization), or stroke (ischemic or hemorrhagic). The 3 secondary efficacy endpoints were as follows: 1) a composite of death from any cause, major coronary event, or stroke; 2) a composite of death from coronary heart disease, MI, or urgent coronary revascularization at least 30 days after randomization; and 3) a composite of CV death, MI, hospitalization for unstable angina, all arterial revascularization at least 30 days after randomization, or stroke.

Prespecified safety endpoints included abnormal elevations of liver enzyme and creatine kinase levels, myopathy, rhabdomyolysis, hepatobiliary events, cancer, and adverse events leading to study drug discontinuation. Neurocognitive events were identified in a post hoc manner as previously described (9). All safety endpoints, except for adverse events leading to study drug discontinuation and neurocognitive events, were independently adjudicated by a clinical events committee blinded to treatment assignment.

Statistical analysis

In this analysis from the IMPROVE-IT trial, patients were categorized into 3 subgroups according to LDL-C concentration at the time of the qualifying event, hereafter “baseline LDL-C.” If an LDL-C was measured within 24 hours after admission, this value was used as the baseline. However, if no measurement was available, then an LDL-C measurement obtained within the prior 6 months was used as the baseline, provided the patient had no changes in LLT and had been clinically stable until the qualifying ACS event. The baseline LDL-C concentration differed from the value at randomization, which was collected at a median of 5 days (IQR: 3-8 days) after admission for the qualifying ACS. LDL-C level at 4 months was used for acquired LDL-C value on treatment, because LDL-C was stable from month 4 through the 6-year median follow-up, as reported previously (9).

Cut points for baseline LDL-C concentration (50-<70, 70-<100, or 100-125 mg/dL) were based on the Adult Treatment Panel III Guideline Update 2004 (10) and were not prespecified. Baseline characteristics are presented as frequencies and percentages for categorical variables and as medians and IQRs for continuous variables. Comparisons between patients by LDL-C concentration at baseline were made using the chi-square test for categorical variables and the Kruskal-Wallis test for continuous variables. Time to first event efficacy analyses were performed by intention-to-treat using Cox proportional hazards modeling with randomized treatment (ezetimibe/simvastatin vs placebo/simvastatin) and randomization stratification factors (prior use of LLT, type of ACS, and status with respect to enrollment in the concurrent EARLY ACS [Early Glycoprotein IIb/IIIa Inhibition in Non–ST-Segment Elevation Acute Coronary Syndrome] trial) as covariates (8,11). The proportional hazard assumption was assessed using Schoenfeld residuals and was not violated. All presented event rates are 7-year Kaplan-Meier estimates. Effect modification of ezetimibe treatment effect by subgroup was tested by incorporating interaction terms into the Cox models. For the analysis of HR normalized for the change in LDL-C level on treatment, we used the approach previously reported by the CTT collaborators in which missing values for LDL-C at 4 months were imputed by using the baseline LDL-C in the calculation of normalized HR per 1-mmol/L difference in achieved LDL-C at 4 months between treatment arms (5,6). A competing-risk analysis based on Fine and Gray's proportional hazards model for the subdistribution of competing risks of non-CV death was performed. Negative binomial regression models were performed to compare the total number of endpoints between patients in the ezetimibe/simvastatin and placebo/simvastatin groups using methods previously described (12). Landmark analyses were performed using the primary outcome after 4 months for the ezetimibe/simvastatin vs placebo/simvastatin comparison in which the patients who experienced events prior to 4 months were excluded. All tests were 2-sided, and a P value <0.05 was considered significant. Analyses were performed using SAS version 9.4 (SAS Institute). Values of P < 0.05 were considered to signify nominal statistical significance. No adjustments were made for multiple comparisons.

LDL-C at baseline and after 4 months

The median baseline LDL-C levels were 62 mg/dL (IQR: 57-66 mg/dL), 86 mg/dL (IQR: 79-93 mg/dL), and 113 mg/dL (IQR: 106-120 mg/dL) in the 3 groups stratified by baseline LDL-C (Table 1, Supplemental Figures 1 and 2). A significant and progressively greater absolute reduction in LDL-C occurred from baseline to month 4 as the baseline LDL-C increased in both treatment arms (Supplemental Figures 1 and 2). However, the absolute differences in median LDL-C achieved at 4 months between treatment arms were similar (17-20 mg/dL), regardless of the baseline LDL-C group (P = 0.64) (Figure 1). Using the approach that was used by the CTT collaborators (5), the absolute differences in LDL-C achieved at 4 months between treatment arms were also similar (14-16 mg/dL) across the 3 baseline LDL-C groups (P = 0.19) (Figure 2).

Clinical efficacy of ezetimibe/simvastatin vs placebo/simvastatin by baseline LDL-C

In the overall trial, the HR for the primary endpoint at 7 years with ezetimibe/simvastatin vs placebo/simvastatin was 0.93 (95% CI: 0.89-0.99). The relative risk reductions in the primary endpoint with ezetimibe/simvastatin vs placebo/simvastatin were consistent among the 3 baseline LDL-C categories (P = 0.95) (Central Illustration, Supplemental Figure 3). In the group with baseline LDL-C of 50-<70 mg/dL, the primary endpoint occurred in 38.3% vs 42.2% (HR: 0.92; 95% CI: 0.80-1.05) in the ezetimibe/simvastatin vs placebo/simvastatin arms. The absolute risk reductions in the primary endpoint were also consistent among the 3 baseline LDL-C categories (P = 0.63) (Supplemental Figure 4). A sensitivity analysis using a competing risk model for non-CV death (Supplemental Figure 5) and analyzing total events (Supplemental Table 2) showed similar findings (P interaction = 0.91 and 0.98, respectively). Effects of ezetimibe/simvastatin vs placebo/simvastatin on individual components of primary endpoint by 3 baseline LDL-C categories were shown in Supplemental Figure 6.

When baseline LDL-C was analyzed as a continuous variable, the benefit of ezetimibe/simvastatin over placebo/simvastatin was also similar across the range of baseline LDL values from 50 to 125 mg/dL (P interaction = 0.69) (Figure 3). Using the approach that was used by the CTT collaborators, normalized relative risk reductions in the primary endpoint per 1-mmol/L difference in achieved LDL-C at 4 months between treatment arms were 16% in the overall trial, 21% in patients with baseline LDL-C of 50-<70 mg/dL, 16% in those with 70-<100 mg/dL, and 13% in those with 100-125 mg/dL (P interaction = 0.91) (Figure 2). Landmark analysis after 4 months also showed consistent normalized relative risk reductions across the 3 baseline LDL-C groups (P interaction = 0.71) (Supplemental Figure 7).

Intensive LLT with ezetimibe/simvastatin compared with placebo/simvastatin also consistently reduced the 7-year event rates of each of the 3 secondary endpoints regardless of baseline LDL-C group (Supplemental Figure 8). Normalized relative risk reductions in the 3 secondary endpoints per 1-mmol/L difference in achieved LDL-C between treatment arms were shown in Supplemental Figure 9.

Safety endpoints by baseline LDL-C categories

Rates of discontinuation caused by adverse events were similar with ezetimibe/simvastatin compared with placebo/simvastatin across baseline LDL-C categories (7.2% vs 7.8%; P = 0.53; 7.7% vs 7.2%; P = 0.39; 8.6% vs 8.0%; P = 0.43 for baseline LDL-C of 50-<70, 70-<100, and 100-125 mg/dL, respectively, with a treatment-subgroup P interaction of 0.60) (Supplemental Table 3). There were no significant treatment interactions by baseline LDL-C categories for any of the other 10 safety endpoints.

Study limitations

First, although the IMPROVE-IT was a large trial with long follow-up, and hence a large number of patients experienced the primary endpoint, the trial was not designed to specifically detect differences in treatment effect as a function of baseline LDL-C. Second, the cut points of 70 and 100 mg/dL were post hoc, but were selected given the treatment targets and thresholds cited in guidelines (7,10,20). Third, we only studied patients hospitalized with recent ACS with a minimum LDL-C of 50 mg/dL. Fourth, the background statin was 40 mg simvastatin (with uptitration to 80 mg in selected patients), which does not represent a maximally potent statin regimen, as is recommended in current guidelines. Fifth, the missing LDL-C values at 4 months in 2,569 patients (14.3%) were imputed by using the baseline values in the calculation of normalized HR per 1-mmol/L difference in achieved LDL-C to reproduce the CTTC approach, as is standard in the lipid field (5). Sixth, although we present some analyses with relative risk reductions normalized per 1-mmol/L reduction in LDL-C, it is important to recognize that for a given patient, the benefit of LDL-C reduction will be a function of the absolute degree of LDL-C lowering achieved in that patient. Seventh, our data are from a randomized controlled trial with strict inclusion/exclusion criteria, including that the patients have high-risk ACS, and thus may not apply to all patients seen in clinical practice. Finally, the population with baseline LDL-C of 50-<70 mg/dL was relatively smaller and at increased risk at baseline compared with those with baseline LDL-C of 70-<100 and 100-125 mg/dL.