Cholesterol Reduction: Weighing the Benefits and Risks

Cholesterol Levels and Risk for Coronary Heart Disease

There is little doubt that elevated cholesterol levels increase the risk for coronary heart disease [5]. Observational research indicates that a linear relation exists: A 20% increase in risk for coronary heart disease is associated with a 10% increase in serum cholesterol levels [6]. This dose-response effect occurs at any cholesterol level [7] and is apparent in both men and women and in both black and white persons.

These estimates, made from prospective data, may underestimate the true risk associated with lipoprotein abnormalities because the relation between risk for coronary heart disease and serum cholesterol is generally based on a single measure of the total cholesterol level [6]. The underestimation is caused by two forms of bias. The first is regression dilution bias, which results from the random fluctuation of cholesterol levels over time in any person and from inaccuracies in the assays used to measure cholesterol levels. This form of bias would introduce random misclassification that would tend to underestimate the true association. The second is surrogate dilution bias, which results from less-than-perfect correlation between LDL cholesterol and total cholesterol levels. Because the LDL cholesterol level tends to be a stronger predictor of coronary heart disease than the total cholesterol level, this second form of bias would also tend to underestimate the true association. Correction for these two forms of bias involves using repeated measures of the LDL cholesterol level and yields an increase of approximately 27% in risk for coronary heart disease for each 10% increase in serum cholesterol level [8].

Findings from randomized trials are consistent with evidence from human observational trials. Data from individual randomized trials and meta-analyses of randomized trials consistently show a reduction in risk for both fatal and nonfatal coronary heart disease in both primary and secondary prevention [9-16]. A recent comprehensive overview by Law and colleagues [15] incorporated data from 28 trials of cholesterol reduction, including 6 multiple intervention trials that each had a cholesterol-reducing arm. This overview indicated that a 10% reduction in serum cholesterol level resulted in highly significant reductions of 10% for death from coronary heart disease and 18% for coronary events. These data from randomized trials are consistent with observational data when treatment lasts 5 years or more. A 10% reduction in cholesterol levels was associated with a 25% reduction in coronary events (95% CI, 15% to 35%) among persons treated for more than 5 years [15]. These findings from meta-analyses are also supported by recent reports from the Scandinavian Simvastatin Survival Study [17] and the West of Scotland Coronary Prevention Study [18], two of the first large-scale trials of HMG-CoA reductase inhibitors, which indicate that intensive cholesterol lowering (25% and 20%, respectively) substantially decreases risk for coronary events (by 34% and 31%, respectively).

The strength and consistency of the association, as well as dose-response and duration effects, strongly support a judgment of cause and effect. Thus, the available evidence clearly indicates that higher cholesterol levels increase the risk for coronary heart disease and that reducing cholesterol levels results in corresponding decreases in risk [5].

Cholesterol Reduction and Total Mortality

The hypothesis that a potential hazard is associated with cholesterol reduction is based primarily on two observations: 1) that an association exists between low serum cholesterol levels and an increased risk for nonvascular mortality in observational epidemiologic studies and 2) that randomized trials do not show that cholesterol reduction has a clear benefit on total mortality, even though they show a clear reduction in risk for coronary heart disease. For several reasons, we feel that the currently available evidence does not support the hypothesis that cholesterol reduction increases overall nonvascular mortality or mortality from any specific cause.

First, although data from randomized trials support the association between cholesterol and risk for coronary heart disease, they do not reinforce the concerns raised by observational data. The observational studies do not suggest a linear relation between cholesterol level and nonvascular disease but rather suggest that persons with the lowest cholesterol levels in a given population are at increased risk for nonvascular disease. These low cholesterol levels have never been achieved in randomized trials. It has been postulated that the rapid reductions in blood cholesterol observed in randomized trials may lead to nonvascular events [19], but it is unlikely that the persons in observational studies who have the lowest cholesterol levels achieved those levels as the result of a rapid decrease. Thus, the data from observational studies and data from randomized trials tend to support two causally distinct hypotheses.

Second, although persons with very low cholesterol levels have higher rates of nonvascular mortality [6], it is unclear whether low cholesterol levels are a marker for or a cause of death from nonvascular causes. The association shown in observational studies between an increase in the rate of nonvascular deaths and lower cholesterol levels may be the result of confounding by the existence of subclinical disease. This is suggested by the attenuation of the relation if deaths within the first 5 years are excluded [6]. Given the long latency period for some nonvascular causes of death, such as liver disease and lung cancer, exclusion of events that occur in the first 5 years of follow-up may not eliminate all confounding.

This is further illustrated by the observation from prospective cohort studies that the association of increased mortality from causes other than heart disease varies depending on whether the samples studied were employment or community based [16]. In employment-based cohorts, among persons with low cholesterol levels (< 5 mmol/L [193 mg/dL]), no evidence of increased mortality from causes other than coronary heart disease was seen (relative risk, 1.00 [CI, 0.94 to 1.06]). In addition to the point estimate indicating a null result, the upper bound of the CI (1.06) excludes any substantially increased risk. Similar results were obtained for specific causes of mortality, including vascular disease other than coronary heart disease (relative risk, 0.92 [CI, 0.80 to 1.05]); cancer (relative risk, 1.00 [CI, 0.91 to 1.10]); trauma, including deaths due to accidents or suicide (relative risk, 1.08 [CI, 0.70 to 1.30]); and other causes (relative risk, 1.08 [CI, 0.97 to 1.21]). In contrast, the risk for death from causes other than coronary heart disease apparently increased among persons in community-based cohorts (relative risk, 1.20 [CI, 1.15 to 1.24]), as did the risks for death from cancer (relative risk, 1.23 [CI, 1.17 to 1.30]), death from trauma (relative risk, 1.29 [CI, 1.13 to 1.47]), and death from other causes (relative risk, 1.26 [CI, 1.15 to 1.24]). If low cholesterol levels lead to mortality from causes other than coronary heart disease, these differing results are difficult to explain. An alternative explanation for the inverse relation in the less healthy community-based populations may be the higher prevalence of premorbid disease, such as cancer, liver disease, and depression, all of which may contribute to lower cholesterol levels. Thus, any relation between low cholesterol levels and death from causes other than coronary heart disease seems more likely to be due to confounding than to a causal relation.

Third, although most randomized cholesterol reduction trials do not show a statistically significantly lower total mortality rate, it is important to distinguish between the finding of no effect and the inability to detect an effect. Specifically, the randomized trials of cholesterol reduction that have been done to date have been designed to test for differences in the incidence of fatal plus nonfatal coronary heart disease events. Generally, trials have not been designed to test for the effects of cholesterol reduction on either total or nonvascular mortality, much less cause-specific nonvascular mortality. Most overviews of the randomized trials of lipid level reduction, whether they include primary prevention trials, secondary prevention trials, or both, have reported a small, statistically nonsignificant reduction in total mortality [9-16]. In the recent comprehensive overview by Law and coworkers [16], the observed reduction in total mortality from all randomized trials was 4% (CI, 10% reduction to 2% increase). This is consistent with what would have been expected assuming a 10% reduction in coronary heart disease mortality and no increase in mortality from causes other than coronary heart disease, which would yield a 6% reduction in overall mortality. To reliably detect a 6% reduction in total mortality, very large randomized trials of cholesterol reduction are necessary. Collins and coworkers [13] estimated the number of events needed to provide a good chance of detecting beneficial effects on total mortality (that is, a 95% probability of P < 0.01), assuming that a 25% average reduction in cholesterol levels (which is larger than that observed in most previous trials) yields a reduction of 25% in fatal coronary heart disease. In secondary prevention trials, in which about 80% of deaths are due to coronary heart disease, 1200 deaths would be needed. In primary prevention trials, in which fewer than half of deaths are due to coronary heart disease, 2800 deaths would be needed. Thus, the inability of a trial to show an effect on total mortality cannot be interpreted as evidence of no benefit. Recent reports strongly suggest that intensive cholesterol reduction results in reduced total mortality in both primary and secondary prevention. The Scandinavian Simvastatin Survival Study, a secondary prevention trial, reported a 30% reduction in total mortality [17], and in the West of Scotland Study, a primary prevention trial with higher-risk patients, there was an apparent 22% reduction.

Fourth, in contrast to the relation between cholesterol levels and heart disease, evidence for a dose-response relation between low cholesterol levels and nonvascular mortality is limited. The lack of this relation raises the possibility of a threshold effect. However, when studies are compared cross-culturally, it is the relative rather than the absolute cholesterol level that is associated with an increased risk for nonvascular death. For example, the mean cholesterol level in the lowest quartile in Finland, where cholesterol levels tend to be high, corresponds to average values in the United States, whereas average levels in Japan would be considered low in the United States. This argues against a causal relation between an absolute cholesterol level and nonvascular disease. A plausible explanation for these cross-cultural differences is that chronic diseases, which lead to death, reduce cholesterol levels regardless of the absolute value, which varies from population to population.

This lack of a clear dose-response relation is also apparent in randomized trials. In a recent metaanalysis, Gordon [20] analyzed data according to the extent of cholesterol reduction (11 studies with cholesterol reductions of 12% or more compared with 11 studies with reductions of less than 12%). As expected, a greater reduction in cholesterol level resulted in a greater reduction in the incidence of coronary heart disease (31% compared with 11%) and mortality (27% compared with 2%). In contrast, greater reductions in cholesterol levels were associated with a smaller increase in rates of nonvascular mortality (11% compared with 30%). The lack of dose response is further confirmed by the recent data from the Scandinavian Simvastatin Survival Study [17], in which drug therapy reduced total cholesterol level by 25% (a reduction larger than that seen in any previous trial), and yet deaths from nonvascular causes did not increase (49 nonvascular deaths in the placebo group compared with 46 nonvascular deaths in the treatment group).

Finally, there is no consistent association with any specific cause of death. Although it remains possible that some underlying mechanism, such as impaired immune function or altered membrane characteristics, may contribute to several disease states, no well-developed universal mechanism has been identified to explain the diversity of causes of death attributed to lower cholesterol levels. The two most comprehensive overviews of randomized trials summarize data according to cause of death [14-16]. Previous overviews that have examined the relation of nonvascular mortality and cholesterol level have not combined data from both primary and secondary prevention trials. Although the presence of underlying cardiovascular disease increases the risk for death from cardiovascular causes and therefore affects total mortality, there is no reason to expect any difference in the hypothesized effect on nonvascular death among persons with and persons without coronary heart disease. Combining primary and secondary prevention trials has the advantage of increasing the power to detect an effect on nonvascular mortality. Overall, there is about a 10% excess in the number of deaths from causes other than coronary heart disease among treated patients (P = 0.02), which is only of marginal statistical significance, especially given that these deaths were not a prespecified end point. When these deaths from non-coronary heart disease were categorized as deaths from non-coronary vascular disease, cancer, trauma, and other causes, none of the individual differences between the treatment groups achieved statistical significance. For cancer and trauma deaths, apparent but nonsignificant increases were associated with cholesterol-reducing treatments.

Any apparent increases in cause-specific mortality are confined to relatively few studies. The trend for cancer is primarily accounted for by one relatively large drug trial, the World Health Organization Clofibrate Trial [21-23], and one large dietary trial, the Los Angeles Veterans Administration Study [24]. In these trials, cholesterol levels were reduced by 9% and 13%, respectively. In both, the differences in mortality from cancer in the intervention and control groups (which, if real, might have been expected to persist or even increase with extended follow-up because of the long latency period of cancer) were, in fact, diminished. The apparent excess number of deaths from trauma among treated persons stands out most clearly in the Helsinki trial of gemfibrozil (15 compared with 4 trauma deaths [25]) and the Lipid Research Clinics trial of cholestyramine (11 compared with 4 trauma deaths [26]), in which the net cholesterol reductions were 9% and 10%, respectively. This raises the possibility that increases in cancer or trauma mortality may be intervention-specific and not related to cholesterol reduction in itself. The issue of drug-specific effects was recently raised in a meta-analysis that explored differences in deaths from causes other than coronary heart disease by treatment type [27]. Eight trials that used fibric acid derivatives showed increased mortality from causes other than coronary heart disease and increased total mortality, but combined data from trials using other agents showed no such increases.

Conclusion

The totality of evidence supports the idea of a causal relation between elevated serum cholesterol levels and risk for coronary heart disease. Specifically, a 10% increase in serum cholesterol level is associated with a 20% to 30% increase in risk for coronary heart disease. Elevations earlier in life seem to be associated with greater increases in risk. Treatment to reduce cholesterol levels by 10% has been shown to reduce the risk for death from coronary heart disease by 10% and for events by 18%, and treatment for more than 5 years yields a 25% reduction in coronary heart disease events. In contrast, the existing data do not support the idea that cholesterol reduction causes increases in any specific nonvascular cause of death. The potential for confounding by the presence of subclinical disease, the lack of a dose-response or threshold effect, the lack of a consistent effect for any specific cause of death, the appearance of adverse effects in just a few trials, and the lack of evidence from basic research, all argue against a causal relation between cholesterol reduction and risk for nonvascular mortality.

Results from overviews of randomized trials are most consistent with a reduction in total mortality associated with cholesterol reduction among those with known cardiovascular disease and support a reduction in primary prevention among those at higher risk. The ongoing and planned large-scale mortality trials, in which larger cholesterol reductions are anticipated, will provide more reliable data on total, nonvascular, and cause-specific nonvascular mortality.

At present, existing data support a recommendation for treatment. Mean age-adjusted cholesterol levels have decreased in the United States since the 1960s [28], and it has been estimated that approximately 30% of the decline in coronary heart disease mortality in the United States has been due to cholesterol reduction [29]. Therefore, public health measures to further reduce the mean cholesterol level of the population are safe, effective, and inexpensive ways to prevent coronary heart disease. For the 30% of the adult population in the United States for whom nonpharmacologic intervention is recommended, lifestyle modification avoids costs and any potential risks associated with drug therapy. Recommendations for drug therapy have focused on persons at highest risk for coronary heart disease events, including those with existing atherosclerotic disease and those with multiple risk factors. Approximately 7% of the adult population meet National Cholesterol Education Program Adult Treatment Panel criteria for pharmacologic intervention. The outcome of ongoing trials will enable existing guidelines to be refined. At present, however, current recommendations for screening and treatment seem both justified and warranted.