Pravastatin and Coronary Heart Disease
To the Editor:
In the April 21, 1998, edition of Circulation, the WOSCOPS study group (West of Scotland Coronary Prevention Study) report that the benefit achieved by administration of pravastatin in WOSCOPS was essentially independent of the percentage by which LDL cholesterol was lowered by treatment and greater than that which could be attributed to lowering of LDL cholesterol alone.1
With unexpected findings of this nature, it is important to exclude bias, especially when the results are derived from post hoc subgroup analysis. In the case at hand, a potential source of bias lies in the selection procedure. WOSCOPS included only men with an LDL cholesterol level of 174 to 232 mg/dL. However, individuals with a history of severe illness, ECG abnormalities or arrhythmias, or severe arterial hypertension, some of which are statistically associated with raised LDL cholesterol levels, were excluded.2 This may have led to a gradient of non–LDL-related risk of coronary heart disease (CHD), ie, men at the high end of the LDL distribution may have had fewer non-LDL risk factors than men at the low end. This might explain why event rates with pravastatin treatment at a given LDL concentration were less than those with placebo at the same LDL concentration: the treated men in WOSCOPS might have been otherwise healthier than men with an identical LDL cholesterol level at baseline. Such a bias would also explain the lack of correlation between extent of LDL cholesterol lowering and reduction in event rate. This conjecture is supported by the finding that event rates in the upper and lower parts of the LDL distribution in the placebo group of WOSCOPS were almost identical (8.3 events per 100 in 5 years at LDL cholesterol ≥189 mg/dL versus 7.6 events per 100 in 5 years at LDL cholesterol <189 mg/dL).2
Most epidemiological and intervention studies point to a semilogarithmic relationship between CHD risk and LDL cholesterol levels, including those obtained with statin treatment.3 This means that the benefit of cholesterol reduction diminishes at low LDL levels, but it does not mean that there is no relationship between on-treatment LDL and CHD risk. In our view, dissemination of the message that minimal lowering of LDL cholesterol (eg, by 12%) by pravastatin is sufficient for primary prevention of CHD would be potentially hazardous and even ethically problematical. The results of this WOSCOPS analysis should be seen as a preliminary hypothesis-generating exercise and not as a guide for treatment decisions.
- Copyright © 1998 by American Heart Association
We thank Drs Assmann, Schulte, and Cullen for their comments and respond as follows.
It is suggested that selection bias may have confounded our analysis. Specifically, the question is asked whether subjects at the high end of the LDL distribution had fewer non-LDL risk factors. When the subjects were divided into those with LDL >189 or <189 mg/dL at baseline, we found a virtually identical distribution of nonlipid risk factors; eg, age was 55.2 versus 55.2 years, diastolic blood pressure was 84 versus 84 mm Hg, incidence of smoking was 46% versus 42%, and incidence of hypertension was 16% versus 16%, respectively. Minor manifestations of coronary heart disease (CHD) at baseline were, if anything, more prevalent in the group with LDL >189 mg/dL; nitrate use/angina was present in 6.8% versus 4.7% and ECG abnormalities in 8.9% versus 7.3% of the high and low LDL groups, respectively. When baseline LDL was divided into quintiles, an association with CHD risk (as shown in Figure 1) was observed. Thus, it is clear that the subjects with higher LDL levels at baseline were not healthier than those with lower levels.
Furthermore, as stated in Methods, adjustment for all baseline risk factors was made in the multivariate models used to examine the difference in risk for subjects taking placebo versus pravastatin for a given on-treatment LDL level (in fact, no substantive differences were present between the 2 groups).
With respect to the relationship between LDL change and risk reduction, we believe that attention should be drawn to the marked attenuation in absolute risk reduction that accompanies increasing degrees of LDL lowering. We did not say that “minimal lowering of LDL ∞ is sufficient for primary prevention.” Rather, we suggested that the additional clinical benefit obtained by lowering LDL beyond 20% to 25% may be small. We were very careful also to point out that “the results are derived from post hoc analysis and, therefore, must be viewed cautiously.” It was of great interest to learn that the observations of the CARE investigatorsR1 closely paralleled our own findings with pravastatin.
Because clinical benefit is now believed to come largely from plaque stabilization rather than regression, and there is increasing recognition that atherosclerosis is not fully reversible, the result of an intervention cannot be predicted accurately by recourse to epidemiological data alone. The abundance of clinical event data now available for a number of statins can be used to address the questions raised by epidemiology.