Early Statin Initiation and Outcomes in Patients With Acute Coronary Syndromes FREE

The benefit of 3-hydroxy-3-methylglutaryl coenzyme A (HMG-CoA) reductase inhibitors (statins) for cardiovascular disease secondary prevention has been clearly demonstrated.^1- 3 However, patients who had experienced acute coronary syndromes (ACS) in the previous 3 to 4 months were excluded from these studies. After patients are discharged from the hospital, the American College of Cardiology and the American Heart Association clinical practice guidelines for acute myocardial infarction (MI) and unstable angina recommend lipid-lowering therapy and a low-cholesterol diet for patients whose low-density lipoprotein cholesterol (LDL-C) levels are higher than 130 mg/dL (3.4 mmol/L), but the guidelines do not recommend initiating lipid-lowering therapy before hospital discharge.^4- 5 Similarly, the National Cholesterol Education Program's Adult Treatment Panel III report recommends that, in general, ACS patients with LDL-C levels of 130 mg/dL (3.4 mmol/L) or higher should be discharged taking drug therapy but recommends clinical judgment for lower levels and does not address initiating lipid-lowering therapy earlier.⁶

Given that patients are at higher cardiac risk soon after experiencing ACS and that benefit from statin therapy in secondary prevention trials was evident only after 12 to 18 months,^1- 3 earlier initiation of statins—within a few days of an ACS event—could potentially improve outcomes and shorten the time to benefit. Long-term compliance also may be improved if statins are included in the discharge medical regimen.^7- 8 Thus, interest in earlier statin initiation is high.⁹

The Myocardial Ischemia Reduction with Aggressive Cholesterol Lowering (MIRACL) randomized trial attempted to address these considerations.¹⁰ Although the MIRACL trial showed modest benefit of early statin treatment on a multicomponent 16-week end point, there were no differences in mortality or the composite end point of death or MI. In contrast, observational studies have suggested more favorable outcomes among patients discharged taking lipid-lowering therapy.^{7- 8,11- 12}

To better understand the discordance of available data, we evaluated the association of early statin initiation with 90-day and 1-year outcomes among patients with ACS in the Sibrafiban vs Aspirin to Yield Maximum Protection from Ischemic Heart Events Post-acute Coronary Syndromes (SYMPHONY) and 2nd SYMPHONY trials.

Our study population comprised 15 904 patients in the SYMPHONY and 2nd SYMPHONY trials conducted at 931 clinical centers in 37 countries between August 1997 and August 1999. The complete trial methods and results have been published.^13- 15 Briefly, SYMPHONY randomly assigned 9233 patients, stabilized for at least 12 hours and within 7 days after ACS (ST-segment elevation, non–ST-segment elevation MI, or unstable angina), to receive either aspirin, 80 mg twice daily, or to receive 1 of 2-dose regimens (low or high) of sibrafiban, an oral platelet glycoprotein IIb/IIIa inhibitor, twice daily for 90 days. In the 2nd SYMPHONY trial, 6671 patients meeting similar inclusion criteria were randomly assigned to receive aspirin 80 mg, low-dose sibrafiban plus aspirin 80 mg, or high-dose sibrafiban alone twice daily. Median treatment duration was 90 (interquartile range [IQR], 36-139) days.

Information on baseline characteristics and concomitant medication use was collected by study site personnel from medical record review and patient interview and was entered on the main SYMPHONY and 2nd SYMPHONY case report forms. Information on concomitant medication use was updated at 2 weeks, 1 month, and 3 months in SYMPHONY and at 1-month and at 3-month intervals until the 2nd SYMPHONY trial ended. All case-report form data were double entered, and concomitant medications were queried as part of safety reconciliation at the end of the studies and were monitored as part of the study monitoring plans. Medications were initially reported by class; statins represented as "HMG-CoA reductase inhibitors." Approximately half-way through SYMPHONY and for the entire 2nd SYMPHONY trials, data on concomitant medications were collected using generic name.

Detailed laboratory data were systematically obtained in a subset of SYMPHONY patients. As part of this effort, baseline and 30-day lipoprotein profiles were assayed prospectively in a blinded core laboratory for the first 3573 patients. Results were merged with the main database at study completion.

The primary SYMPHONY end point was the 90-day composite of death (all cause), recurrent MI, or severe recurrent ischemia (SRI). For the 2nd SYMPHONY trial, the time to this composite was the primary end point. Myocardial infarction and SRI were adjudicated by blinded clinical events classification committees for each study.¹³ Severe recurrent ischemia required recurrent ischemic symptoms for 20 minutes or more resulting in unplanned or unscheduled revascularization. Myocardial infarction required creatine kinase–MB (CK-MB) higher than the upper limit of normal (ULN) for clinical events, higher than 3 × ULN after percutaneous coronary intervention (PCI), and higher than 5 × ULN after coronary artery bypass graft surgery, or development of new Q-waves in 2 or more contiguous electrocardiographic leads.

SAS release 8.2 statistical software (SAS Institute Inc, Cary, NC) was used for all statistical analyses. Descriptive statistics (medians, IQR, and percentages) were generated for baseline demographic and clinical characteristics, medication use, and lipid levels. Compliance with statin use was determined as the proportions of patients reported to be using statins at 90 days among those reported to have initiated statin therapy early.

Our primary assessment was the association of early statin initiation (≤7 days after ACS, n = 3952) with 90-day death; death or MI; and death, MI, or SRI and with 1-year death. Secondary outcomes included 90-day death, MI, or rehospitalization; death, MI, or stroke; and stroke alone. The no-statin reference group (n = 8413) included patients surviving 5 or more days from the qualifying ACS event with no recorded statin use prior to or during the study period. Kaplan-Meier estimates were generated for 90-day and 1-year outcome rates. Myocardial infarction and SRI were used as adjudicated for the main studies. Unadjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for early statin vs no statin were determined.

To adjust for differences in key outcome predictors, we used covariates from Cox proportional hazards models previously developed for each 90-day end point using the combined SYMPHONY and 2nd SYMPHONY databases. An additional model was developed for death at 1 year. For each model, satisfaction of the proportional hazards assumption was verified for each retained variable. We also developed a propensity model using stepwise logistic regression to identify factors associated with early statin use. Using this model, patients were classified by quintile of increasing probability for early statin initiation. Patients within a quintile were similar in their likelihood to receive a statin. The propensity score was also included in multivariable adjustment. P values are 2-tailed and presented for descriptive purposes, with no adjustments for multiple comparisons. P<.05 was considered significant.

In analyses restricted to the subset of SYMPHONY patients with core laboratory lipoprotein profiles, we assessed the association of baseline lipid levels with the relation between early statin use and outcomes.

Lipid-lowering therapy was used in 8217 of 15 904 patients (52%); 7469 (47%) were treated with a statin, including 1787 patients who received atorvastatin; 66, cerivastatin; 165, fluvastatin; 195, lovastatin; 888, pravastatin; 2307, simvastatin; and 2492 patients whose statin type was unknown. Four hundred eight patients received 2 or more different statin agents over the course of the studies. Figure 1 summarizes use and timing of statin therapy. Statins were known to be initiated early (early statin) in 3952 patients; median time from qualifying event was 2.0 (IQR, 1.0-3.1) days. Among patients who did not receive statins, 22 died within 5 days after ACS, leaving 8413 patients who did not receive statins (no statin). In early-statin patients, 90-day adherence to therapy was 91%.

Baseline characteristics of early-statin and no-statin patients are displayed in Table 1. Of characteristics associated with 90-day mortality, early-statin patients were younger, had better renal function, and more often had qualifying event MI. They less often had hypertension, prior angina or MI, history of heart failure or heart failure between qualifying event and randomization, or chronic obstructive pulmonary disease. Time from qualifying event to randomization was slightly shorter in early statin-treated patients, and coronary angiography and PCI between qualifying event and randomization were more frequent.

Propensity modeling identified 30 covariates that independently predicted the likelihood of early statin. The model c statistic was 0.774. The distribution of these baseline characteristics by propensity quintile is displayed in Table 2. The relationship of age with propensity was nonlinear, but, in general, older patients were less likely to receive early statin. A history of hypercholesterolemia, higher qualifying event blood pressure, qualifying event MI, and use of β-blockers, heparin, intravenous glycoprotein IIb/IIIa antagonists, and low-molecular-weight heparin were associated with higher probability of early statin. Early statin use was less likely in all geographic regions vs North America. It was also lower among white patients, those with an S₃ gallop or atrial fibrillation, a history of hypertension, stroke, or prior angiography; those receiving aspirin or other oral antiplatelet agents, calcium channel blockers, nitrates or diuretics; and those undergoing coronary artery bypass graft surgery or PCI. Propensity scores by quintile were well-matched (Table 3).

Kaplan-Meier 90-day and 1-year event rate estimates along with unadjusted and adjusted HRs are shown in Table 4. In unadjusted analyses, both 90-day and 1-year mortality were lower among early-statin than no-statin patients, but no significant differences were observed in unadjusted analyses for death or MI or for death, MI, or SRI. After adjustment for statin propensity and covariates, there were no significant differences for death (HR, 1.08; 95% CI, 0.75-1.56); death or MI (HR, 1.08; 95% CI, 0.91-1.29); or death, MI, or SRI (HR, 1.15; 95% CI, 0.99-1.34). One-year mortality was also similar between groups after propensity and covariate adjustment (HR, 0.99; 95% CI, 0.73-1.33). Secondary analyses revealed no associations of early statin with death, MI, or rehospitalization or with death, MI, or stroke although for stroke alone early statin was associated with lower adjusted risk (HR, 0.54; 95% CI, 0.30-0.97).

Exploratory analyses revealed associations of early statin use with outcomes similar to the overall comparisons within major subgroups treated with atorvastatin, pravastatin, or simvastatin and after stratifying by qualifying ACS type and sibrafiban vs aspirin assignment. Including late statin patients and/or those with unknown statin initiation time with no statin patients in a "no early statin strategy" also yielded results similar to the prespecified comparisons. Our findings were also similar in analyses restricted to propensity-matched early statin and no statin cohorts derived using a published macro (http://www2.sas.com/proceedings/sugi26/p214-26.pdf).

Of patients with core laboratory lipoprotein profiles, 770 received early statin and 1941 did not receive statins. Only small differences in baseline lipid levels were observed between early-statin and no-statin groups (Table 5). Substantial 30-day reductions in total cholesterol and LDL-C levels occurred in the early-statin group. Although absolute and relative reductions were greater among early statin patients with baseline LDL-C levels higher than 130 mg/dL (3.4 mmol/L), compared with no-statin patients there were substantial reductions in both LDL-C cohorts.

Figure 2 shows covariate- and propensity-adjusted HRs for 90-day death and death or MI for patients receiving statins early vs those who received no statins as a function of baseline total cholesterol and LDL-C. There were significant interactions of total cholesterol level with early statin use (for interaction term, P = .02 for death; and P = .06 for death or MI) such that early treatment was more favorable as baseline total cholesterol levels increased. For death, the HR for early statin vs no statin was 1.0 at a total cholesterol level of 190 mg/dL (4.9 mmol/L) and for death or MI at 237 mg/dL (6.1 mmol/L); below these levels early statin was associated with higher risk than no statin. At levels less than 150 mg/dL (3.9 mmol/L), the lower 95% CI for risk for death excluded an HR of 1.0 although the CIs were wide. For death or MI this occurred at levels less than 195 mg/dL (5.1 mmol/L).

A similar pattern was observed for LDL-C. The LDL-C level by early statin interaction term P value was .10 for death and P = .38 for death or MI. The 90-day HR by LDL-C was 1.0 at an LDL-C level of 124 mg/dL (3.2 mmol/L) for death and 207 mg/dL (5.3 mmol/L) for death or MI. Confidence intervals were wide, but for baseline LDL-C levels lower than 112 mg/dL (2.9 mmol/L), the lower 95% CI excluded an HR for death of 1.0. For LDL-C levels lower than 125 mg/dL (3.2 mmol/L) an HR for death or MI of 1.0 was below the lower 95% CI. Relationships were similar for 1-year death by total cholesterol and LDL-C levels (Figure 2).

Despite evidence of lipid-lowering effect and 91% adherence to therapy at 90-day follow-up, we could not confirm in a cohort of 12 365 patients the mortality benefits recently associated with early lipid-lowering therapy in other post-ACS databases.^11- 12 Observations from the subset of patients with baseline lipoprotein profiles suggest underlying this is the possibility of a risk-benefit gradient for early statin therapy by total and LDL-C levels such that early statin initiation is associated with worse outcomes at levels below currently recommended treatment levels. Since the analysis subset was small and the CIs wide, cautious consideration of these findings is warranted. However, they should also raise a flag of caution to clinicians about starting statins early in patients having experienced ACS if the LDL-C or total cholesterol level is low. Ultimately, the best timing of statin initiation and a firm estimate of short-term and long-term effects after early initiation await the results of ongoing, adequately powered, randomized clinical trials.

Nonlipid effects of statins, such as anti-inflammatory, antiplatelet, and antithrombotic properties,^16- 28 and improvements in endothelial function^{16,19,29- 32} might confer an advantage early after experiencing an ACS event. However, the immediacy of these effects and their translation to clinical outcome has not been proved; many have only been shown after several weeks to months of therapy. Conversely, smooth muscle cell inhibitory properties of statins have been demonstrated in vitro and in vivo.^33- 35 Although these effects may be beneficial during long-term treatment, statins used for acute events have the potential for further plaque destabilization and impaired early healing and could adversely affect clinical course.

Although small, randomized trials investigated the effect of early statin on surrogate end points,^36- 37 the only randomized clinical outcomes trial to test early treatment in non–ST-segment elevation ACS was the MIRACL trial, and no trials to date have evaluated early initiation after ST-segment elevation MI. In the MIRACL trial the 16-week relative risk (RR) of death, resuscitated sudden death, MI, or recurrent ischemia leading to emergency rehospitalization was 0.84 (95% CI, 0.70-1.00; P = .05), among 3086 patients randomly assigned within 96 hours to receive atorvastatin, 80 mg/d, vs placebo.¹⁰ The composite end point was driven by recurrent ischemia leading to emergency rehospitalization (RR, 0.74; 95% CI, 0.57-0.95). Total rehospitalization was not reported, and the primary analysis used stratification by country, which was not prespecified. There was no effect on death (RR, 0.94; 95% CI, 0.67-1.31) or death or MI (RR, 0.92; 95% CI, 0.75-1.13). The clinical significance of the MIRACL findings in the absence of an effect on death, resuscitated sudden death, or death or MI is unclear.

Two recently published observational studies suggest more favorable associations of discharge lipid-lowering therapy with mortality. In 14 071 unselected post-MI patients in Sweden, investigators demonstrated a significant relationship between discharge statin and reduced 1-year mortality.¹¹ After adjusting for covariates and propensity for statin use, the RR for 1-year mortality was 0.75 (95% CI, 0.63-0.89). A second study, using 2 ACS clinical trials databases, evaluated the association between any lipid-lowering therapy and 6-month mortality among 3653 patients prescribed discharge lipid-lowering therapy and 17 156 who were not.¹² After adjustment for propensity and other confounders, the 6-month HR for mortality was 0.67 (95% CI, 0.48-0.95; P = .02). We observed significantly lower 90-day and 1-year mortality in unadjusted analyses, but in contrast to previous observational analyses, after either propensity or covariate adjustment or both, these differences were no longer apparent. However, upper CIs of previous observational studies and lower CIs of the current study and MIRACL overlap.

Our study is unique in 2 respects. First, in the SYMPHONY studies it was possible to evaluate concomitant medication use over time. Several studies suggest that long-term statin use is improved if statin therapy is included among discharge medications and have reported 1-year rates of medication adherence of 77% to 91%.^7- 9 In our current study, 91% of early-statin patients continued therapy at 90 days. Second, in a subset of patients, prospectively measured baseline and follow-up lipoprotein levels were available, which were not in previous observational studies. This allowed confirmation that statin therapy was associated with a substantial lowering of total cholesterol and LDL-C levels after only 30 days. Thus, it is difficult to explain that our findings resulted from patients' noncompliance with therapy or that therapy failed to affect lipid levels.

More important, it was possible to assess the interaction between cholesterol levels and the relationship of early statin with outcomes. Although the MIRACL trial demonstrated no evidence for harm in ACS patients with a broad range of cholesterol levels treated with statins, data from that study are inadequate to support treating patients with LDL-C levels below 130 mg/dL (3.4 mmol/L) in the acute phase of ACS. Despite wide CIs, the results of our study raise the possibility that early statin initiation in ACS patients with cholesterol values lower than currently recommended treatment levels may be associated with an increased risk for cardiac events compared with patients not receiving statin treatment. The results of ongoing randomized trials of statin initiation in the acute phase of ACS will definitively address this issue, and the result may ultimately be different when those data are available.

Results of our study may differ from those of previous studies for several reasons. Statin use rates differed, possibly reflecting differences in use criteria. If propensity adjustment did not account for all differences in selection of patients for statin use, disparate findings may have resulted. The current population was overall at lower risk, potentially rendering these patients less likely to benefit from aggressive statin therapy or for a benefit to be detected. However, our study had approximately 80% power to detect a mortality difference of the magnitude suggested previously. Furthermore, there were no differences in the association of early statin use with outcome by severity of ACS presentation (MI or unstable angina). Lipid-lowering therapies varied from nonspecific,¹² to multiple statins (our study), to predominantly one statin.^10- 11 Although it is possible that individual statins perform differently, we observed no difference among agents with regard to association of early initiation with outcomes.

Methodological differences including population selection (including vs excluding patients already receiving lipid-lowering therapy) and inception time (survivors to hospital discharge vs time of early statin initiation) also may have contributed to observed differences. Also, because patients were randomly assigned in the SYMPHONY and 2nd SYMPHONY trials several days after experiencing ACS, clinical outcome and propensity modeling could take advantage of clinical and treatment-use variables (such as PCI use) from the period between qualifying ACS and randomization. Therefore, these models may have adjusted better for important differences between early-statin and no statin groups than possible in previous studies. Still, remaining unmeasured confounders may have influenced our results. Issues such as these highlight the problems inherent in all observational analyses and emphasize the need for randomized clinical outcomes trials to guide treatment decisions.

In conclusion, in the absence of evidence for harm and with demonstrated improvement in long-term compliance, initiation of statin therapy during hospitalization should be considered. However, until the observed interaction between cholesterol levels and the relationship of early treatment with outcomes can be confirmed or refuted, clinicians should use caution in starting statin therapy during the acute phase of ACS in patients who do not meet current treatment guidelines (LDL-C levels >130 mg/dL [>3.4 mmol/L]). This study should not be considered evidence for changing practice; rather, it should be considered evidence for sustaining the practice of not starting statins in the acute phase of ACS if LDL-C is lower than current treatment guidelines. The results of ongoing, randomized clinical trials of early statin initiation, powered to assess the effect on death or death or MI and the relationship with lipid levels, will be required to determine if earlier initiation translates into incremental benefit or earlier realization of benefit in secondary prevention.