Effect of Pravastatin on Coronary Disease Events in Subgroups Defined by Coronary Risk Factors
The Prospective Pravastatin Pooling Project
Background—Previous trials have had insufficient numbers of coronary events to address definitively the effect of lipid-modifying therapy on coronary heart disease in subgroups of patients with varying baseline characteristics.
Methods and Results—The data from 3 large randomized trials with pravastatin 40 mg were pooled and analyzed with the use of a prospectively defined protocol. Included were 19 768 patients, 102 559 person-years of follow-up, 2194 primary end points (coronary death or nonfatal myocardial infarction), and 3717 expanded end points (primary end point, CABG, or PTCA). Pravastatin significantly reduced relative risk in younger (<65 years) and older (≥65 years) patients, men and women, smokers and nonsmokers, and patients with or without diabetes or hypertension. The relative effect was smaller, but absolute risk reduction was similar in patients with hypertension compared with those without hypertension. Relative risk reduction was significant in predefined categories of baseline lipid concentrations. Tests for interaction were not significant between relative risk reduction and baseline total cholesterol (5% to 95% range 177 to 297 mg/dL, 4.6 to 7.7 mmol/L), HDL cholesterol (27 to 58 mg/dL, 0.7 to 1.5 mmol/L), and triglyceride (74 to 302 mg/dL, 0.8 to 3.4 mmol/L) concentrations, analyzed as continuous variables. However, for LDL cholesterol, the probability values for interaction were 0.068 for the prespecified primary end point and 0.019 for the expanded end point. Relative risk reduction was similar throughout most of the baseline LDL cholesterol range (125 to 212 mg/dL, 3.2 to 5.5 mmol/L) with the possible exception of the lowest quintile of CARE/LIPID (<125 mg/dL) (relative risk reduction 5%, 95% CI 19% to −12%).
Conclusions—Pravastatin treatment is effective in reducing coronary heart disease events in patients with high or low risk factor status and across a wide range of pretreatment lipid concentrations.
The hydroxy methyl glutaryl coenzyme A reductase inhibitor pravastatin prevents coronary events and death in patients who have had a myocardial infarction (MI)1 2 or unstable angina,2 as well as in hypercholesterolemic patients without clinical coronary heart disease (CHD).3 However, there remains uncertainty about the benefits of this and other lipid-modifying treatments in groups of patients such as older patients, women, and diabetics because these groups composed only a small proportion of the total population in the major trials. The influence of pretreatment plasma lipid concentrations on risk reduction during treatment also is not fully resolved, again because of an insufficient number of patients in lipid subgroups in the individual studies.
For these reasons, the investigators of the 3 large randomized trials with pravastatin1 2 3 initiated a pooling project shortly after the individual trials started. The present report is a study of relative risk reduction in coronary events with pravastatin in patient groups defined on the basis of baseline characteristics, including CHD risk factors and plasma lipids, in the pooled data from 3 large trials with a total of 19 768 patients, 102 559 person-years of follow-up, 2194 primary coronary end points (CHD death and nonfatal MI), and 3717 expanded coronary events (CHD death, nonfatal MI, CABG, and PTCA).
The Prospective Pravastatin Pooling Project (PPP) commenced in 1992, and a protocol was designed before the results of any of the trials were known.4 The objectives of the project were to report on the pooled results of the 3 trials of treatment effects on death and stroke (reported separately) and on evidence of treatment effects within subgroups. In brief, the data were pooled from the West of Scotland Coronary Prevention Study (WOSCOPS),3 the Cholesterol and Recurrent Events trial (CARE),1 and the Long-term Intervention with Pravastatin in Ischemic Disease study (LIPID).2 The patient eligibility criteria have been reported previously.1 2 3 4 WOSCOPS was a primary prevention trial in Scotland among 6595 men with hypercholesterolemia who had not sustained MI or unstable angina.3 CARE was a secondary prevention trial conducted in the United States and Canada in 4159 men and women who had a documented acute MI and average total cholesterol and LDL cholesterol concentrations.1 LIPID was a secondary prevention trial conducted in Australia and New Zealand in 9014 men and women who had either an acute MI or unstable angina and plasma total and LDL cholesterol concentrations in an average to mildly elevated range.2 All of these trials compared 1 dose of pravastatin, 40 mg/d, and a matching placebo and were conducted in a double-blind manner. The average durations of follow-up were 5 years for WOSCOPS and CARE and 6.0 years for LIPID. Each trial was approved by institutional review boards, all subjects gave their informed consent, and the procedures that were followed were in accordance with institutional guidelines.
The primary outcome variable for analysis of treatment effects in the subgroups was coronary death or nonfatal MI. The secondary outcome, not prespecified in the protocol, was an expanded end point that included the primary end point, CABG or PTCA. This end point provided additional events and hence was considered useful in the investigation of risk reduction in subgroups in which there were few events and in further exploration of possible interactions identified in the prespecified analyses. Coronary death was defined as fatal MI, fatal coronary artery disease, or sudden death. The diagnosis of definite nonfatal MI required either (1) typical symptoms of acute MI and elevation of serum creatine kinase concentration or its MB isoform, (2) unequivocal ECG evidence (eg, development of new pathological Q waves), or (3) typical symptoms with strong ECG evidence.
Statistical analyses were performed at Wake Forest University (R.B.; Winston-Salem, NC) independently of the sponsor. The protocol specified analyses of risk reduction in the 3 trials combined and in CARE and LIPID. The intention-to-treat model of analysis was used. Primary analysis involved use of the Cox proportional hazards model, with trial as a covariate, and was unadjusted for baseline covariates. Participant differences in study duration or underlying expected event rates were adjusted for by including a main effect for “study” in each model. Ancillary analysis included the baseline risk factors of age, sex, smoking, diabetes, hypertension; LDL cholesterol, HDL cholesterol, and triglyceride levels; and qualifying coronary event (MI, unstable angina, no coronary event) as covariates. An examination of effect modification according to patient characteristics involved the use of a Cox proportional hazards model with the “trial” and the patient factor as a covariate and the appropriate treatment interaction terms. The relationship between baseline plasma lipid concentrations and coronary event reduction with pravastatin was investigated with the lipids as continuous variables in the Cox model, as prespecified, with a linear interaction term. Risk reduction was first investigated in the prespecified lipid categories and subsequently explored in quintiles of the lipid distributions. Relationships between follow-up event rates and baseline cholesterol levels were assessed with generalized linear models5 to predict events. A probability value of <0.05, 2-sided, was reported as significant. No formal adjustment was made for multiple comparisons (per protocol), and appropriate allowance for this is needed in interpreting results.
Pravastatin reduced the risk of coronary events significantly and similarly in patients with manifest CHD, either MI or unstable angina, and in those without manifest CHD (Tables 2⇓ and 3⇓). Relative risk reductions in the primary and expanded end points in the pravastatin group were highly statistically significant for patients <55 years, 55 to 64 years, and 65 to 75 years old. Women and patients with diabetes assigned to receive pravastatin experienced a highly significant reduction in risk for the expanded end point. Risk reductions in the primary end point for women and diabetic patients were similar to those in the expanded end point, although the probability value for the primary end point was borderline due to the smaller number of events. In women, after adjustment for baseline risk factors, the risk reduction for the primary end point was 22% (P=0.045). In other instances, the inclusion of baseline coronary risk factors as covariates did not affect the magnitude of the relative risk reductions or the significance of the interaction testing.
There was no evidence for a different relative effect of pravastatin according to age of the patients, in men compared with women, in diabetic compared with nondiabetic patients, and in smokers compared with nonsmokers (Tables 2⇑ and 3⇑).
Pravastatin significantly reduced the risk of the primary and expanded end points in patients who reported a history of hypertension, although the magnitudes of the reductions were significantly less than those for the patients who did not report hypertension (14% versus 33%, P interaction=0.003 for the primary end point, and 16% versus 29%, P interaction=0.013 for the expanded end point) (Tables 2⇑ and 3⇑). These differences were slightly diminished when the analysis was restricted to CARE and LIPID (14% versus 31%, P interaction=0.2 for the primary end point, and 17% versus 26%, P interaction=0.09 for the expanded end point) and were not affected by the inclusion of baseline risk factors and the use of antihypertensive medications as covariates in the multivariate analysis. In further multivariate analysis, the relative risk reductions for patients who did not report a history of hypertension but who were taking antihypertensive medications for indications other than hypertension such as angina pectoris or congestive heart failure (n=5245 CARE/LIPID) were 31% for the primary end point and 24% for the expanded end point. Thus, the use of antihypertensive medications per se appears not to have been responsible for the difference in risk reduction between hypertensive and nonhypertensive patients. There also was no effect of pravastatin on blood pressure levels in these trials.
Risk reductions were highly significant for the primary and expanded end points in all of the categories of total cholesterol, LDL cholesterol, and HDL cholesterol that were prespecified in the protocol (Tables 4⇓ and 5⇓). For triglycerides, risk reductions for the primary and expanded end points were significant in the lower and intermediate prespecified categories (<133 and 133 to 219 mg/dL); in the higher category (≥220 mg/dL), risk reduction was borderline in significance for the primary end point (P=0.057) and significant for the expanded end point (P=0.029). Event rates were lower in the pravastatin group than in the placebo group across the range of total cholesterol (Figure 1⇓, top), HDL cholesterol (Figure 2⇓, top) and triglyceride (Figure 2⇓, bottom) concentrations. There was no significant difference in the effect of pravastatin on the relative reduction of coronary risk according to baseline total cholesterol, HDL cholesterol, or triglyceride levels, as shown by a nonsignificant probability value in tests for interaction with these lipid concentrations treated as continuous variables (Tables 4⇓ and 5⇓). For the expanded end point, the P value for interaction between baseline triglycerides and risk reduction was borderline (P=0.055). However, risk reduction was nearly identical for the lowest quintile of triglycerides (<98 mg/dL [1.1 mmol/L], risk reduction expanded end point 20%) compared with the highest quintile (>207 mg/dL [2.4 mmol/L], risk reduction 18%) (Figure 2⇓). Thus, the evidence is weak, at best, for modification of the effect of pravastatin by triglyceride concentrations.
For LDL cholesterol, risk reduction for the primary end point was 22% in the lowest prespecified category of <135 mg/dL (4.5 mmol/L) (P=0.005), 23% in the middle category of 135 to 174 mg/dL (4.5 to 5.5 mmol/L) (P<0.001), and 32% in the highest category of ≥175 mg/dL (5.5 mmol/L) (P<0.001) and for the expanded end point (17%, 23%, and 30%, respectively) (Tables 4⇑ and 5⇑). Test for a difference in risk reduction according to baseline LDL cholesterol concentration, considered as a continuous variable, was borderline for the primary end point (P=0.068) and significant for the expanded end point (P=0.019). An exploratory analysis of this finding in the CARE/LIPID pooled population showed that event rates increased continuously over the range of LDL cholesterol levels in the placebo group but not in the pravastatin group (Figure 1⇑, bottom). Risk was reduced significantly in all except the lowest quintile of <125 mg/dL (RR for primary end point 0.96 [0.77 to 1.19], RR for expanded end point 0.95 [0.81 to 1.12]) (Figure 1⇑, bottom).
Trials are rarely designed with sufficient patients to reliably investigate effects in portions of the population, because these generally are secondary objectives. This eventuality was anticipated by the investigators of WOSCOPS, CARE, and LIPID, and we formed a prospective collaboration that had a major goal of determining the effects of pravastatin on coronary events in clinically important subgroups of patients who participated in the trials.4 This research approach is strengthened by similarities of the study designs of the individual trials,1 2 3 of which all used the same drug, pravastatin, at the same daily dose, 40 mg; were double-blind with placebo control; and had rigorous uniform definitions of outcome variables that were prespecified, comprehensive ascertainment of events, and meticulous validation of the clinical reports of the events that composed the primary and secondary outcome variables. Moreover, the study populations of the 2 secondary prevention trials, CARE and LIPID, shared many important baseline characteristics; had a similar use of therapies, including aspirin, β-blockers, and prior coronary revascularization; and were both intended to be broadly representative of populations with CHD. A key strength of the pooling project is its prospective design, in which the research questions, the analytic approach, and the definitions of end points were determined in advance of the conclusion of any of the individual trials.
Remaining weaknesses of this pooling project are the relatively small numbers of women and patients with diabetes, which did not provide sufficient statistical power for definitive results on risk reduction for the prespecified primary end point. When the sensitivity of the study was increased by adding revascularization procedures to the primary end point, event reduction due to pravastatin for women and diabetics was highly significant, providing evidence that strongly favored an interpretation that pravastatin therapy prevents coronary events in women and diabetics. A similar situation occurred in the evaluation of whether baseline LDL cholesterol concentrations influenced the effect of pravastatin on the primary end point. The P value for interaction between baseline LDL cholesterol and coronary events was borderline for the primary end point but significant for the expanded end point.
In older patients, aged 65 to 75 years at the beginning of the trial in which they were enrolled and 70 to 80 years at its end (not included in WOSCOPS), relative risk reduction was highly significant, and there was no evidence that relative risk reduction differed according to age. Trials that used clofibrate and niacin6 or simvastatin7 and included patients up to the age of 70 years found that risk reduction was significant in those of age 60 to 70 years and similar to that in the younger patients. Because three fourths of deaths from MI in the United States occur in patients older than 65 years,8 these results should support the use of pravastatin to reduce overall coronary disease in the population.
Pravastatin significantly reduced coronary events in patients with a history of hypertension. However, the relative risk reduction was less than that in patients who did not report a history of hypertension. This finding is unexpected because no individual trial of pravastatin or of other statins showed reduced efficacy in hypertensives. We have identified no explanation for the difference in relative risk reduction between hypertensives and nonhypertensives, and this finding may be due to chance and the testing of multiple subgroups. No formal adjustment of multiple comparisons was made among the 10 prespecified subgroups, and one might expect a false-positive result on average with every 20 independent tests. Furthermore, this interpretation is in part supported by the weakening of the significance level for the difference in relative risk reduction in the expanded end point compared with the primary end point. If the difference were real, then one might anticipate that the statistical significance would increase as the number of events included in the analysis increases. We further emphasize that absolute risk reduction is substantial with pravastatin in the hypertensives in view of their high rate of recurrent coronary events. For example, the number of patients needed to be treated to prevent a coronary event (coronary death, MI, CABG, or PTCA) was 21 among hypertensive patients compared with 20 in the overall CARE/LIPID population.
Baseline total cholesterol, HDL cholesterol, and triglyceride concentrations did not significantly affect relative risk reduction with pravastatin. This is consistent with the individual results of the 3 pravastatin trials2 9 10 ; Scandinavian Simvastatin Survival Study (4S), a secondary prevention trial with simvastatin11 ; and Air Force/Texas Coronary Atherosclerosis Prevention Study (AFCAPS/TexCAPS), a primary prevention trial with lovastatin.12
Relative risk reduction was significant in each prespecified category of baseline LDL cholesterol concentration. However, an interaction test for an effect of baseline LDL cholesterol on relative risk reduction, also prespecified, showed a trend for the primary end point (P=0.068) that increased in significance with the expanded end point (P=0.019). The finding was explored by an analysis of LDL quintiles, suggesting a possible diminished risk reduction only in the lowest quintile of the CARE and LIPID range (ie, <125 mg/dL [3.2 mmol/L]). An LDL cholesterol of 125 mg/dL corresponds to approximately the 30th percentile of the middle-age and older population in the United States.13 Possible explanations include a threshold below which there is no risk reduction, a curvilinear relationship with diminishing risk reduction as baseline LDL cholesterol decreases, or simply chance. A contributing factor could be the effect of the diminished absolute LDL cholesterol reduction observed with lower baseline LDL cholesterol concentration.10 Besides CARE and LIPID,1 2 no other clinical end point trials of statins investigated patients in this range of baseline LDL cholesterol concentrations and thus do not contribute data to this topic.3 11 12 Finally, a conclusive investigation would require a still larger number of patients and events than exist in CARE and LIPID in the lower part of the range of LDL cholesterol concentrations.
In conclusion, the PPP demonstrates remarkable uniformity of relative risk reduction in primary and secondary prevention in men and woman, in older and younger patients, and in diabetic and nondiabetic patients; across the population ranges of total cholesterol, HDL cholesterol, and triglycerides; and in most of the LDL cholesterol range. Absolute risk reduction by pravastatin is then driven by the level of risk in the patient groups and is large in those with high event rates, such as older patients, diabetic patients, smokers, and those with high total or LDL cholesterol concentrations. These considerations should weigh strongly in determination of the appropriateness of this treatment for primary prevention.14 Finally, these findings support a policy of near-universal treatment of patients with clinical evidence of CHD.
The PPP is funded by an investigator-initiated grant from Bristol Myers Squibb to Wake Forest University. We acknowledge the valuable participation of the Pravastatin Pooling Project research group. We also thank Tim Craven, Wake Forest University, for statistical and programming advice and assistance.
- Received January 6, 2000.
- Revision received May 31, 2000.
- Accepted May 31, 2000.
- Copyright © 2000 by American Heart Association
McCullagh P, Nelder JA. Generalized Linear Models, 2nd ed. London, UK: Chapman and Hall; 1989.
Miettinen TA, Pyorala K, Olsson AG, et al. Cholesterol-lowering therapy in women and elderly patients with myocardial infarction or angina pectoris. Circulation. 1997;96:4211–4218.
Graves EJ. Detailed diagnoses and procedures. In: National Hospital Discharge Survey, 1989. National Center for Health Statistics. Centers for Disease Control and Prevention, Atlanta. Vital Health Stat 1991;13; No. 108.
West of Scotland Coronary Prevention Study Group. Influence of pravastatin and plasma lipids on clinical events in the West of Scotland Coronary Prevention Study (WOSCOPS). Circulation. 1998;97:1440–1445.
Pedersen TR, Olsson AG, Faergeman O, et al. Lipoprotein changes and reduction in the incidence of major coronary heart disease events in the Scandinavian Simvastatin Survival Study (4S). Circulation. 1998;97:1453–1460.
Gotto AM, Whitney E, Stein EA, et al. Relation between baseline and on-treatment lipid parameters and first acute major coronary events in the Air Force/Texas Coronary Atherosclerosis Prevention Study (AFCAPS/TexCAPS). Circulation. 2000;101:477–484.
National Cholesterol Education Program. Second Report of the Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults. National Institutes of Health, National Heart, Lung, and Blood Institute. Bethesda, Md: National Institutes of Health; 1993:1A4; NIH publication No. 93-3095.