Author + information
- Received April 29, 1998
- Revision received August 5, 1998
- Accepted September 10, 1998
- Published online January 1, 1999.
- Marc A Pfeffer, MD, PhD, FACCa,* (, )
- Frank M Sacks, MDa,
- Lemuel A Moyé, MD, PhD∗,
- Cara East, MD†,
- Steven Goldman, MD, FACC‡,
- David T Nash, MD, FACC§,
- Jacques R Rouleau, MD, FACC∥,
- Jean Lucien Rouleau, MD, FACC¶,
- Bruce A Sussex, MD, FACC#,
- Pierre Theroux, MD, FACC¶,
- Ron J Vanden Belt, MD, FACC∗∗ and
- Eugene Braunwald, MD, FACCa
- ↵*Address for correspondence: Dr. Marc Pfeffer, Cardiovascular Division, Brigham and Women’s Hospital, 75 Francis Street, Boston, Massachusetts 02115
Objectives. We sought to assess the influence of baseline lipid levels on coronary event rates and the effectiveness of pravastatin therapy in the Cholesterol And Recurrent Events (CARE) study.
Background. The CARE study cohort provided a relatively unique opportunity to examine the relation between lipid levels and clinical events in a post–myocardial infarction (MI) population with relatively low cholesterol and low density lipoprotein (LDL) cholesterol values.
Methods. There were 4,159 patients with a previous infarct and a total cholesterol level <240 mg/dl, LDL cholesterol level 115 to 174 mg/dl and triglyceride level <350 mg/dl randomly allocated to placebo (n = 2,078) or pravastatin 40 mg/day (n = 2,081). Time to either coronary death or nonfatal MI (primary end point) or to the secondary end point, which included undergoing a coronary revascularization procedure, was determined as a function of baseline lipids (total, LDL, high density lipoprotein [HDL] cholesterol and triglyceride levels).
Results. Quartile analysis indicated important effects for LDL cholesterol, in which a higher LDL was associated with greater cardiac event rates (in the placebo group, every 25-mg/dl increment in LDL was associated with a 28% increased risk [5% to 56%, p = 0.015]) in the primary event. The differential event rates with respect to baseline LDL cholesterol for placebo and pravastatin groups reduced the difference in clinical outcomes at lower LDL cholesterol levels. In both the placebo and pravastatin groups, an inverse relation between baseline HDL cholesterol and cardiac events was observed (10 mg/dl lower baseline HDL cholesterol level was associated with a 10% [0% to 19%, p = 0.046] increase in coronary death or nonfatal MI).
Conclusions. Within the LDL cholesterol levels in CARE (115 to 174 mg/dl), baseline values influenced both the risk of events in the placebo group as well as the clinical effectiveness of pravastatin therapy.
Epidemiologic studies have quantified the importance of elevated plasma cholesterol as a risk factor for the development of first (1–4)as well as recurrent coronary heart disease events (3–6). This relation between cholesterol, particularly low density lipoprotein (LDL) cholesterol, and coronary risk is nonlinear; the absolute and relative increases in risk are greater at the higher end of the cholesterol distribution (1,7,8). Before the development of 3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitors (HMGs), there was evidence that a cholesterol reduction with either dietary measures, an intestinal bypass procedure or treatment with lipid-lowering agents of moderate efficacy could reduce this risk of coronary heart disease events in hypercholesterolemic patients with existing coronary disease (9,10). However, offsetting questions regarding the magnitude of this effect, compliance issues and concerns of possible adverse effects of long-term pharmacologic therapies limited their implementation (11,12).
Randomized, controlled clinical trials with HMGs in hypercholesterolemic patients without (13)and with (14)overt coronary disease have proved definitively that these agents reduce coronary heart disease events (i.e., deaths, myocardial infarction [MI] and the need for coronary revascularization procedures) without an offsetting increase in noncardiovascular events. Further analysis of the Scandinavian Simvastatin Survival Study (4S) demonstrated that within the range of total cholesterol (5.5 to 8.0 mmol/liter, 212 to 306 mg/dl) the relative risk reduction of these major coronary events attributed to simvastatin was not influenced by the baseline cholesterol value (15).
The CARE trial was a randomized, placebo-controlled trial testing the effectiveness of pravastatin (40 mg) on the incidence of coronary heart disease events in a broad group of survivors of MI with a total cholesterol level <240 mg/dl (mean 209) (16). The trial demonstrated an overall benefit of pravastatin therapy with a 24% reduction in fatal coronary heart disease deaths and nonfatal MI (95% confidence interval [CI] 9% to 36%) and a 24% (95% CI 13% to 33%) reduction in the end point of major coronary events used for prespecified subgroup analysis (16). Thus, the CARE study extended the benefits of cholesterol-lowering therapy to an even broader group of patients with average cholesterol values. The present analysis of CARE data was conducted to determine the influence of baseline lipid values on the coronary event rates of the groups assigned to placebo and pravastatin, with particular emphasis on the patients with lower total and LDL cholesterol values, a cohort underrepresented in previous trials.
The CARE study design and results have been reported previously (16,17). In brief, 4,159 survivors of MI (average 3 to 20 months) with a total cholesterol level <240 mg/dl, an LDL cholesterol level between 115 and 174 mg/dl and a fasting triglyceride level <350 mg/dl were randomized to conventional therapy plus either placebo or pravastatin 40 mg/day. Lipid levels were determined by a central core laboratory from an average of at least two samples. Additional inclusion and exclusion criteria have been delineated (17). The study was designed to have a median follow-up period of 5 years and had as its primary end point death attributed to coronary heart disease or a nonfatal MI. A secondary end point of major coronary events (fatal coronary heart disease, nonfatal MI, coronary artery bypass surgery or angioplasty) was constructed to assess therapeutic efficacy within subgroups (16).
Quartile ranges were constructed based on the distribution of prerandomization values for total, LDL and high density lipoprotein (HDL) cholesterol levels as well as triglyceride levels. For the purpose of comparison, a post-hoc analysis was conducted which utilized the lipid eligibility criteria employed in 4S (total cholesterol ≥212 mg/dl and triglyceride <220 mg/dl) to designate CARE patients as either “4S lipid-eligible” or “4S lipid-ineligible.”
All analyses were performed on an intention-to-treat basis, and p values were two-sided. The comparability of baseline characteristics across baseline lipid levels (lower vs. highest quartile) was ascertained using standard normal tests for continuous variables and the chi-square test for categoric variables. All hypothesis testing and all risk reductions with their confidence intervals were assessed using the Cox proportional hazards model. Kaplan-Meier survival curves for the control and pravastatin groups were calculated (18).
The impact of the baseline lipid level on the primary and major coronary event end points in CARE was assessed using a Cox proportional hazards model, with time to the event as the dependent variable. The role of baseline lipids as a risk factor for the occurrence of either the primary or secondary end point was obtained through interaction models. The models attempted to identify interactions in two ways. The first procedure included the baseline lipid level as an independent variable, which allowed an assessment of the linear effect that baseline lipids would have on the end point. The second interaction analysis examined the impact of lipids on the end point by including three indicator variables (the lowest quartile is the referent quartile) simultaneously in the Cox proportional hazards model, which allowed not only the effect of lipids to be different from quartile to quartile, but also the effect of therapy in adjusting these quartile effects to vary across lipid quartiles. Relations between baseline lipid values and event rates and therefore treatment effect were also estimated using a regression model.
During the course of the trial, the pravastatin group had a reduction of total cholesterol by 20%, LDL cholesterol by 28% and triglyceride level by 14%, and HDL cholesterol increased by 5% as compared with the placebo group. The frequency of the primary end point was 13.2% in the placebo group and 10.2% in the pravastatin group (relative risk reduction 24%, p = 0.003). Major coronary events (primary end point or myocardial revascularization) occurred in 26.4% of the placebo group and 20.1% in the pravastatin group (relative risk reduction 24%, p < 0.001).
Within the cholesterol range of the CARE patients (159 to 239 mg/dl), the baseline value did not significantly influence the risk of experiencing a coronary event subsequent to randomization (Fig. 1). Even when the analysis was limited to the 2,078 placebo-assigned patients a significant relation between baseline cholesterol and the primary study end point was not detected (11% relative risk for coronary heart disease death or nonfatal MI for a 25-mg/dl increase in baseline cholesterol, 95% CI −7 to 32, p = 0.246). Similarly, expanding this coronary end point to also include revascularization procedures did not result in a significant influence of baseline cholesterol and coronary event risk in the placebo group (6% relative risk for every 25-mg/dl cholesterol increment, 95% CI −6% to 20%, p = 0.315) (Fig. 1).
However, there was a significant relation between the baseline LDL cholesterol level (range 115 to 174 mg/dl) and the risk of experiencing a coronary event. For the entire cohort (i.e., patients in both treatment arms combined into a single group), for every 25-mg/dl increment of baseline LDL cholesterol, there was an 18% increase in the relative risk of coronary death or MI (95% CI 1% to 37%, p = 0.034). In the overall group, the risk of experiencing the expanded coronary end point for a 25-mg/dl increase in baseline LDL cholesterol was 22% (95% CI 2% to 46%, p = 0.028). This increase in risk with higher baseline LDL cholesterol was mainly attributed to the placebo-assigned patients. In the placebo group, a 25-mg/dl increment of baseline LDL cholesterol was associated with a 28% increased risk of coronary death or MI (95% CI 5% to 56%, p = 0.015). Similarly, for the major coronary event end points (already defined), there was a significant influence of baseline LDL cholesterol on coronary events only in the placebo group, with every 25-mg/dl increment in LDL cholesterol associated with an 18% increased risk (95% CI 3% to 36%, p = 0.021) (Fig. 2). For pravastatin-assigned patients, a 25-mg/dl increase in baseline LDL cholesterol was not associated with a significant increase in the risk (6%) of either coronary death or nonfatal MI (95% CI −16% to 34%, p = NS) or the risk (2%) of the expanded end point, which includes coronary revascularization procedures (95% CI −13% to 20%, p = NS). This trend toward a differential influence of baseline LDL cholesterol in the placebo and pravastatin groups on coronary event rates (interaction p = 0.226) resulted in a convergence of event rates between the placebo and pravastatin groups at the lower LDL cholesterol levels.
Designating by quartiles for higher LDL cholesterol levels, as anticipated, resulted in a concomitantly higher total cholesterol level. However, HDL cholesterol did not vary across LDL cholesterol quartiles, and fasting triglyceride levels decreased with increasing LDL cholesterol quartiles (Table 1). The patient characteristics were, for the most part, comparable across LDL cholesterol quartiles, except that the lowest quartile patients were slightly older, had a higher prevalence of diabetes and were predominantly men than the higher LDL quartile patients (Table 1). At the lowest LDL quartile (n = 1,046, range 115 to 127 mg/dl, median 121), the risk reduction with pravastatin for the primary end point was only 9% (95% CI −38% to 33%) and, similarly, for all major coronary events it was 7% (95% CI −29% to 20%) (Fig. 2).
In the combined treatment arms, there was an important inverse relation between baseline values for HDL cholesterol and coronary event rates. For each 10-mg/dl decrement in the baseline HDL cholesterol value, there was an 11% greater likelihood of experiencing a coronary death or nonfatal MI (95% CI 0% to 23%, p = 0.049). When major coronary events were considered as the end point, the influence of baseline HDL cholesterol was similar, with a 10% increase in the risk of these coronary events (95% CI 3% to 19%, p = 0.007) for each 10-mg/dl decrement in baseline HDL cholesterol of the combined groups. This influence of baseline HDL cholesterol on coronary events was similar in the placebo and pravastatin groups, with no trend for an interaction. In each quartile of HDL cholesterol, the benefits of pravastatin were comparable (Fig. 3).
Within the range of fasting triglyceride levels (90 to 350 mg/dl), no statistically significant linear relation between baseline levels and the primary or expanded end points was observed in either the overall cohort or within the treatment groups. Although the risk reduction for coronary events was not significantly altered by baseline triglycerides, the lowest two quartiles demonstrated the greater risk reduction with pravastatin (risk reduction 30.3%, 28.8%, 14.1% and 18.5% for expanded end points, lowest to highest triglyceride quartile, respectively) (Fig. 4).
Utilizing 4S lipid criteria, 34% of the CARE population (n = 1,409) could have been considered as eligible and the majority (66%, n = 2,750) would have had either cholesterol values too low or triglyceride values too high, or both, to have been eligible for that study. This construction for 4S lipid-eligible and lipid-ineligible subgroups within CARE produced groups that were comparable with respect to age, ejection fraction, history of hypertension, previous coronary artery bypass graft surgery or angioplasty and use of aspirin (Table 2). However, the 4S lipid-eligible group had a lower percentage of men and diabetics (Table 2). By design, the cholesterol values of these two groups within CARE differed, with total cholesterol levels higher (224 ± 7.7 vs. 201 ± 15 mg/dl) and triglyceride levels lower (146 ± 40 vs. 161 ± 69 mg/dl) in the CARE patients who were 4S lipid-eligible as compared lipid-ineligible.
For the placebo group, this selection of 4S lipid criteria resulted in a greater likelihood of experiencing either coronary death or nonfatal MI (15.4% [109 of 710 patients] vs. 12.0% [116 of 1,368 patients], 4S lipid-eligible vs. ineligible). In the pravastatin group the event rates in both the 4S lipid-eligible (9.8% [136 of 1,382 patients]) and lipid-ineligible (10.7% [75 of 699 patients]) subgroups were both lower than that in the placebo group. The expanded end point rates, which included coronary revascularization procedures, were also more frequent in placebo-assigned 4S lipid-eligible (28.2% [200 of 710 patients]) and lipid-ineligible groups (25.4% [348 of 1,368 patients]) than in their respective pravastatin groups (21.2% [148 of 699] vs. 20.4% [282 of 1,382]). Therefore, in CARE patients, the risk reduction of coronary deaths and MI with pravastatin therapy tended to be of a greater magnitude in the 4S lipid-eligible (risk reduction 32% [9% to 49%]) than in the 4S lipid-ineligible patients (risk reduction 18% [−2 to 35]) (Fig. 5). The risk reduction for the expanded coronary end point was 26% (95% CI 9% to 40%, p = 0.005) and 21% (95% CI 8% to 33%, p = 0.003) for the 4S lipid-eligible and lipid-ineligible groups, respectively. Indeed, in a smaller (n = 574) CARE cohort meeting 4S lipid and other inclusion criteria (age <70 years, no MI in past 6 months, no past cerebral vascular event or current antiarrhythmic therapy), a striking 48% risk reduction (95% CI 11% to 70%) was observed for coronary death or MI, with a 20% risk reduction (95% CI 3% to 34%) observed in the remaining 3,555 non-4S eligible CARE patients.
CARE and 4S
Taken together, the 4S and CARE studies definitively show the value of pravastatin and simvastatin for an extremely broad range of patients with prevalent coronary artery disease. Because the risk attributed to cholesterol is not linear, but increases sharply over the higher ranges, a greater benefit of cholesterol reduction would be anticipated in patients with the most severe elevations in cholesterol. In 4S, the patients who received placebo who were within the highest quartiles for either total cholesterol or LDL cholesterol (≥282 and 207 mg/dl, respectively) experienced ∼20% more major coronary events than the patients who received placebo in the lowest quartiles (total cholesterol <241 and LDL cholesterol ≤170 mg/dl). A quartile analysis for baseline total and LDL cholesterol within 4S for the effectiveness of simvastatin showed a relatively uniform clinical benefit of therapy within those subgroups over the range of values in their study (15).
Approximately one-third of CARE patients had lipid values within the 4S range. The analysis of the overlapping lipid-eligible patients in CARE and 4S underscores that in patients with documented coronary artery disease and total cholesterol >212 mg/dl, substantial clinical benefits can be anticipated with the use of either simvastatin or pravastatin. In the two-thirds of CARE patients whose total cholesterol values were lower than those of 4S patients, a clinically important 18% reduction in coronary deaths and nonfatal MI was demonstrated with pravastatin therapy in survivors of MI who had modern management.
Patients with LDL cholesterol
However, our analysis suggests that the efficacy of HMG therapies in patients with coronary artery disease with low baseline LDL cholesterol values may be more limited. The quartile analysis for LDL cholesterol revealed only a modest, nonsignificant reduction in coronary events in the lowest quartile (115 < 127 mg/dl). As previously presented in the lowest quintile (LDL cholesterol <125 mg/dl) of CARE, no beneficial trend was observed with pravastatin (16). It should be emphasized that with the declining serum cholesterol levels in many patients (19)and with aging (20)that many coronary events occur in individuals with “desirable” total cholesterol and LDL values (21). Indeed, a frequent exclusion for CARE was an LDL value ≤114 mg/dl (17). This finding of an apparent reduction in the efficacy of HMG therapies in patients with lower LDL levels and coronary artery disease has major public health implications concerning the appropriate therapies for a substantial group of patients.
Patients with coronary artery disease despite low total cholesterol and LDL cholesterol tend to have relatively low HDL cholesterol and are more likely to be diabetic (17,21). Indeed, the mean HDL cholesterol level in CARE of 39 mg/dl was lower than the mean value of 46 mg/dl in 4S. Other less well-characterized risk factors are likely to have a more pronounced role in the pathogenesis of their coronary atherosclerosis. Our data, although hypothesis generating and not definitive, indicate that further lowering of LDL cholesterol in this understudied group with low LDL cholesterol and coronary artery disease may not have a major clinical impact. Meta-analyses of secondary prevention trials of cholesterol-lowering therapies suggest that there is less benefit with lower baseline cholesterol values (22). The Long-term Intervention with Pravastatin in Ischemic Disease (LIPID) study, with 25% of its 9,014 patients having a total cholesterol and LDL cholesterol level of 196 and 131 mg/dl or less, respectively, will add a great deal of information on the effectiveness of statins in patients in this range (23). Major clinical trials exploring the alternative and not mutually exclusive approaches of pharmacologically raising HDL cholesterol in this patient group are under way (24,25). The other classic risk factors and new potentially modifiable risk factors, such as elevated homocysteine levels (26)and other yet to be identified atherosclerotic risk factors, may have a greater influence in this group with low LDL cholesterol and coronary disease.
The additive and interactive nature of atherosclerotic risk factors complicates the use of arbitrary threshold values for the initiation of therapy. In patients with clinically apparent coronary artery disease, the greater likelihood of a recurrent cardiovascular event justifies an intensive individual approach to their risk factor management. Although the CARE study demonstrates that therapy would be anticipated to reduce clinical events in a broad group of patients with coronary artery disease who receive modern management, the study also indicates that the cohort with atherosclerosis and low LDL cholesterol levels appears to benefit less with this form of therapy and constitutes an important focus for new investigative efforts.
☆ The Cholesterol And Recurrent Events (CARE) study was supported by an investigator-initiated grant from Bristol-Myers Squibb Research Institute, Princeton, New Jersey.
- Cholesterol And Recurrent Events
- confidence interval
- high density lipoprotein
- 3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitors
- low density lipoprotein
- myocardial infarction
- Scandinavian Simvastatin Survival Study
- Received April 29, 1998.
- Revision received August 5, 1998.
- Accepted September 10, 1998.
- American College of Cardiology
- Stamler J, Wentworth D, Neaton JD, for the MRFIT Research Group
- Task Force on Cholesterol Issues, American Heart Association,
- LaRosa J.C,
- Hunninghake D,
- Bush D,
- et al.
- Rosengren A,
- Hagman M,
- Wedel H,
- Wihelmsen L
- Gordon D.J
- Oliver M.F
- Davey Smith G,
- Pekkanen J
- Miller M,
- Seidler A,
- Kwiterovich P.O,
- Pearson T.A
- Holme I
- The LIPID Study Group
- Rubins H.B,
- Robins S.J,
- Iwane M.K,
- et al.
- The Bezafibrate Infarction Prevention (BIP) Study Group