1.1 Information on the potential benefits of treating high blood cholesterol levels comes from various sources. Much of the information is indirect; despite the voluminous literature on the effects of cholesterol reduction, major gaps remain. The long-term safety of the most frequently prescribed class of cholesterol-lowering medications remains unknown. Important populations, such as young men, elderly men, and women of all ages, were excluded from most randomized clinical trials of cholesterol reduction. The lack of direct clinical trial evidence in these populations, some would argue, proscribes lipoprotein screening. We offer a pragmatic alternative: We seek to make the best estimate of the effectiveness of treatment on the basis of results from clinical trials and, for populations not included in the trials, analyses of observational data.

See the related article on pp 505-508 and an editorial comment on pp 513-514.

2.0 Treatment Effectiveness: Evidence from Randomized, Controlled Clinical Trials

2.1 Screening can only be effective if it leads to an action—namely, to an intervention to reduce an elevated total or low-density lipoprotein (LDL) cholesterol level—that prevents or alleviates disease. Evidence about this issue comes from more than 40 published reports of randomized trials of cholesterol reduction, primarily done in middle-aged men Table 1, Table 5, [11-56]. These findings reinforce laboratory, physiologic, and epidemiologic evidence that cholesterol has a causal role in the development of coronary heart disease. The trials also quantify the risks and benefits of therapy and show that cholesterol-lowering therapies can mitigate the effects of high blood cholesterol levels on the illness and death caused by coronary heart disease.

2.3 The major findings of the trials and meta-analyses can be summarized as follows:

Cholesterol reduction diminishes the incidence of and the morbidity and mortality associated with coronary heart disease

2.4 In men and women with coronary heart disease, cholesterol reduction retards or reverses the progression of atherosclerotic plaques and reduces mortality from coronary heart disease [13, 15-17]. When used for primary prevention, cholesterol-lowering therapies reduce the incidence of coronary heart disease [58, 63]. The effects of lowering cholesterol on coronary heart disease depend on the magnitude of cholesterol reduction: Each 10% reduction in cholesterol levels is associated with roughly a 20% to 30% reduction in the incidence of coronary heart disease [60]. The small cholesterol reduction in many trials of low-fat diets may thus account for the modest effects of such diets on coronary heart disease outcomes: Typical diets (such as a type I diet, in which total and saturated fat consumption account for no more than 30% and 10% of all caloric intake, respectively) reduce cholesterol levels by an average of 2% [65, 66]. One widely cited study reported that an intensive dietary intervention reduced total and LDL cholesterol levels by 5% compared with a control group encouraged to increase fat consumption [67] (average high-density lipoprotein [HDL] cholesterol levels also decreased by 5%). Drugs usually decrease LDL and total cholesterol levels more effectively than does diet. The newest class of cholesterol-lowering drugs, the statins, are better tolerated and reduce LDL cholesterol levels more than do the drugs included in most randomized trials [14, 67-70].

Cholesterol reduction has been shown to reduce total mortality only in high-risk populations

2.5 Cholesterol reduction for the secondary prevention of coronary heart disease diminishes mortality from all causes and coronary heart disease [48, 62, 71]. Favorable trends seen in one trial that examined the effects of cholesterol reduction in patients with carotid atherosclerosis but without symptomatic heart disease (the Asymptomatic Carotid Artery Progression Study [72]) suggest that cholesterol reduction will produce similar benefits in other populations at high short-term risk for dying of coronary heart disease.

2.6 Individual trials have generally lacked the statistical power necessary to assess the effect of lowering cholesterol on total mortality in primary prevention settings. In the recent West of Scotland Study [56], the reduction in all-cause mortality was of borderline statistical significance (P = 0.05). The World Health Organization trial [20] unexpectedly found that clofibrate caused a statistically significant increase in total mortality. Pooled analyses of primary prevention trials have found no reductions in all-cause mortality [57, 73, 74]; the difference in net mortality effects between primary and secondary prevention is statistically significant [60, 73]. Several cholesterol-lowering interventions had unexpected harmful effects on mortality from events unrelated to coronary heart disease, including cancer and deaths from accidents and violence [57, 73]. To the extent that these events are not a chance finding, they represent more than an association—they are the effects of randomly assigned treatments administered as part of clinical trials. The events are not likely to correspond directly to the J-shaped relation between cholesterol levels and mortality noted in observational data on middle-aged men. The J-shaped relation refers to markedly elevated rates of noncardiac mortality seen at cholesterol levels well below those typically reached in clinical trials of treatment of high blood cholesterol. The observational association may reflect the consequences of noncardiac diseases that depress cholesterol levels for many years before they end in death rather than a direct result of cholesterol reduction. In contrast, successful randomization ensures that benefits and harms seen in randomized clinical trials are unconfounded by the effects of chronic disease and other unmeasured differences between persons who do and do not receive the treatment being studied.

At what level of risk for death from coronary heart disease do the benefits of therapy outweigh the risks?

2.7 The risks for developing and dying of coronary disease are much lower in primary prevention settings than in secondary prevention settings, even when all trial participants have high cholesterol levels. The relation between underlying risk and potential benefits has crucial implications: A therapy that halves the risk for coronary heart disease would reduce the annual mortality associated with coronary heart disease from, for example, 40 deaths per 1000 patients to 20 deaths per 1000 patients in a secondary prevention setting and from 4 deaths per 1000 patients to 2 deaths per 1000 patients in a primary prevention setting. Noncardiac side effects of the therapy causing 2 deaths per 1000 patients each year would greatly reduce overall mortality rates if therapy was used for secondary, but not primary, prevention. In a population in which the annual rate of deaths from coronary disease decreased to fewer than 2 deaths per 1000 patients, therapy might actually increase the number of deaths.

2.8 Davey Smith and colleagues [60] pooled primary and secondary prevention trials and examined the net effect of treatment on mortality after considering the underlying risk for coronary heart disease. Their meta-analysis shows that the net benefit of cholesterol lowering depends on the underlying risk for death from coronary heart disease. This effect is statistically significant and appears not to have changed in the 20 years represented by the combined clinical trials. Cholesterol lowering probably prolongs the lives of patients in whom the annual rate of death from coronary heart disease is at least 3%. In addition to persons with manifest coronary heart disease, this population includes a few persons who lack symptoms of coronary heart disease (< 1% of middle-aged men and even fewer women) and are at high risk because they have several cardiac risk factors in addition to high blood cholesterol levels. This cutoff is likely to be different for treatments with risks and benefits differing from those represented in the trials included in the meta-analysis by Davey Smith and colleagues [60].

2.9 Because so few patients were enrolled in published studies of any single drug or specific diet, most meta-analyses include diverse cholesterol-lowering interventions. Although it is possible that one or more adverse effects might result from all methods of cholesterol reduction, a key implication of such analyses is that adverse effects and potential net benefits depend on the specific therapy (such as drug class) used. Thus, each type of intervention must be evaluated separately. Harmful effects of cholesterol lowering on mortality unrelated to coronary heart disease are seen with pharmacologic therapies but not diet, and the increase in mortality not related to coronary heart disease may be restricted to specific drug classes (such as gemfibrozil, clofibrate, and other fibrates and estrogen) and appears to be unrelated to the degree of cholesterol reduction [64]. By controlling for the type of therapy or excluding trials with negative results, the authors of some meta-analyses concluded that cholesterol lowering may reduce mortality when such therapy is used for primary prevention [61, 64]. This finding is consistent with the possibility that only a subset of cholesterol-lowering drugs increases mortality not related to coronary heart disease, but it could also be a spurious result: The post hoc exclusion of trials with negative results is necessarily biased toward finding a favorable effect, even when none exists. Thus, the meta-analyses offer hope that some interventions will be found to reduce all-cause mortality in primary prevention settings; however, mortality reduction remains unproven.

2.10 Notably absent from the meta-analyses are statins (lovastatin, simvastatin, fluvastatin, and pravastatin), which are now the most widely used class of cholesterol-lowering drugs. Until 1994, no trials reporting the effects of these drugs on coronary heart disease end points and mortality in primary or secondary prevention settings had been published. In the first secondary prevention trial, published late in 1994 [48], a survival benefit was seen in a population in which the annual risk for death from coronary disease was about 1%. The first primary prevention trial, published late in 1995 [56], showed that in middle-aged men with an annual rate of death from coronary heart disease of about 0.3% to 0.4% (total mortality rate, about 0.8%), pravastatin reduced total mortality and mortality from coronary heart disease (P = 0.04 for death from definite and suspected coronary events; the finding was not significant for definite events alone). Neither trial reported that the statins increased mortality from noncardiovascular factors. These results suggest that statins have a more favorable risk–benefit profile than older cholesterol-lowering drugs. However, the long-term side effects of statins are unknown, and the drugs are carcinogenic in laboratory animals at two to seven times the plasma drug levels achieved with recommended doses in humans [75]. Until further information is available, the long-term safety and effectiveness of statins in lower-risk primary prevention settings cannot be extrapolated from the results of trials of other drugs or of statins used in high-risk settings.

3.0 Using Observational Data To Generalize Results of Randomized Clinical Trials

3.1 Many candidates for cholesterol reduction—young men, elderly men, and women—are not studied in randomized trials. Even for middle-aged men, the principal population studied, clinical trials provide limited information on how other coronary heart disease risk factors modify the benefits of cholesterol reduction. Before clinical trials were completed, observational data were the primary source of information on the magnitude of the benefits of cholesterol reduction among middle-aged men; currently, such data continue to be essential to estimating benefits in populations excluded from clinical trials. In this section, we first review the published information on cholesterol as a risk factor in diverse populations. We then use the association between cholesterol level and the subsequent risk for death from coronary disease or all causes observed in the Framingham Heart Study to calculate the hypothetical change in health outcomes that would result from cholesterol reduction. We then use the results of these analyses to develop practical strategies to identify patients who may benefit from screening. Our approach is an extension of familiar strategies for interpreting results of diagnostic tests: We seek to determine the "pretest" level of risk on the basis of other cardiac risk factors. A high pretest risk makes it more probable that a positive test result (that is, an elevated cholesterol level) will lead to a management decision that improves health outcomes [76].

Effects of Age and Sex on the Association between Cholesterol and Coronary Heart Disease

3.2 In middle-aged men, a continuous and graded association can be seen between blood cholesterol levels and the risk for developing or dying of coronary heart disease. The high incidence of coronary disease and its strong association with cholesterol levels are among the main reasons middle-aged men have been the focus of most primary prevention trials. Cholesterol is also a risk factor for the eventual development of coronary heart disease in young men [77], but the incidence of coronary heart disease in this group is low [78].

3.3 Cholesterol is also a risk factor for coronary heart disease in middle-aged women, but it appears not to be a risk factor for death from cardiovascular causes, including stroke [79, 80]. Furthermore, because the incidence of coronary heart disease is lower in women than in men of the same age [81], the hypothetical reduction in the incidence of and mortality from coronary heart disease with cholesterol-lowering therapies is small [82].

3.4 The effects of cholesterol as a risk factor in both men and women 65 years of age and older are more complex. With age, the incidence of coronary heart disease increases, whereas the relative risk for coronary heart disease caused by an elevated cholesterol level decreases [81, 83, 84]. Until ages in the mid-70s, the former effect predominates; therefore, the absolute increase in annual mortality from coronary heart disease attributable to a high cholesterol level increases with age [83]. Among persons in their late 70s and older, however, mortality from coronary heart disease is not associated with elevated cholesterol levels [84-86].

3.5 The epidemiologic data indicate that the effects of cholesterol reduction probably depend on characteristics such as age, sex, blood pressure, history of cigarette smoking, and glucose intolerance that modify the association between cholesterol level and risk for coronary heart disease.

Cholesterol and All-Cause Mortality

3.6 The association between blood cholesterol level and all-cause mortality has long been controversial. In many observational studies, low cholesterol levels (typically less than 4.14 mmol/L [160 mg/dL]) were associated with increased noncardiovascular and total mortality [79]. Studies based on observational data have not determined whether chronic diseases (such as lung cancer) reduced the cholesterol level or whether low cholesterol levels caused the associated diseases. This finding, however, appears to have little direct relevance for individualized management of hypercholesterolemia. Treatment seldom causes cholesterol levels to decrease from high levels to the levels at which increased mortality from noncardiovascular causes is seen.

3.7 More relevant for treatment is the association between cholesterol and all-cause mortality at higher cholesterol levels: Does it parallel the relation between cholesterol and death from coronary heart disease, as is commonly assumed? This appears to be true in middle-aged and younger men whose cholesterol levels are in the range relevant to treatment decisions [79, 87], but not in women or older men. In a pooled analysis of more than 124 000 women, no relation was found between cholesterol level and death from all causes or death from cardiovascular disease [79, 80]. Mortality is either unassociated with cholesterol level [88] or decreases with increasing cholesterol levels at advanced ages [84, 87, 89]. In a large cohort of middle-aged to elderly Japanese-American men [90], age-adjusted mortality was positively associated with cholesterol level in men who smoked, had untreated hypertension, or consumed large amounts of alcohol, but not in men with none of these risk factors. Thus, an elevated cholesterol level is associated with increased overall mortality in middle-aged men, particularly those with other risk factors, but not in the other populations that have been studied.

Modeling the Effectiveness of Therapy

3.8 We estimated the effects of cholesterol reduction on coronary heart disease and all-cause mortality in different strata of age, sex, and risk for coronary heart disease in order to develop specific criteria to identify the persons likely to benefit from cholesterol reduction. To assess the hypothetical effect of therapy, we assumed that the observational association between blood cholesterol level and risk for coronary heart disease or death is causal and reversible and that cholesterol reduction confers immediate full benefit—that is, the risk for coronary heart disease or death of a person who decreases his or her cholesterol level will be the same as that for a person whose cholesterol was at the reduced level without a specific health intervention. This assumption produces an upper-bound estimate for the effect of therapy, as long as the cholesterol level is measured precisely. Partial validation of such models comes from an important observation in clinical trials: After 2 years of treatment, the magnitude of the reduction in the rate of coronary heart disease end points in the clinical trials of cholesterol reduction is about two thirds the magnitude of the added risk for coronary heart disease associated with elevated cholesterol levels in the observational studies [91]. At about 5 years of treatment, the risk reduction reaches the entire hypothetical magnitude [92]. These findings contradict the belief that cholesterol reduction must be initiated early in the process of plaque formation (decades before the onset of symptomatic coronary heart disease) if it is to be maximally effective.

3.9 However, there are important limitations to using observational data to project the effects of cholesterol-lowering interventions. If such interventions have unexpected benefits that are not related to cholesterol reduction itself (as has been proposed for low-fat diets), they would not be detected in an observational analysis. On the other hand, observational studies provide little information on the potential adverse effects of drugs used to decrease cholesterol levels. Drugs can induce metabolic abnormalities, sometimes producing liver inflammation (niacin) or redirecting the pathways through which cholesterol is metabolized (statins). There is no reason to believe that an observational study in a population that was not exposed to the drugs would detect the effects of such physiologic changes.

Data and Variables

3.10 The data are drawn from the 30-year follow-up of the Framingham Heart Study [93]. At examinations done every 2 years, risk factors were measured, and events that had occurred since the previous examination were recorded. Laboratory errors and fluctuation in the measured values of risk factors can lead to underestimation of their association with health outcomes (regression—dilution bias). To mitigate the effect of such variability, we used the mean of repeated measurements (the mean of three blood pressure measurements obtained at a single examination and the mean of two cholesterol level measurements taken at different examinations) [91, 94].

3.11 The association between cholesterol and overall mortality at low cholesterol levels may be confounded by the effect of chronic disease on cholesterol. Further, for some age groups, the association between death from coronary heart disease and cholesterol is not monotonic; in several cohort studies of elderly men and women, for example, the lowest relative risk for coronary heart disease is not seen for patients with the lowest cholesterol levels [81, 84]. To maximize the estimated benefits of cholesterol reduction, our analyses exclude all observations for which the measured mean cholesterol level is less than 180 mg/dL (4.7 mmol/L).

Analytic Approach

3.12 We estimated the 2-year risk for death from coronary heart disease within each sex and age (< 55 and more than equals to 55 years of age) stratum; we used the standard Framingham risk factors (systolic and diastolic blood pressure, Metropolitan Relative Weight, glucose intolerance, and current cigarette smoking) but excluded cholesterol levels. We used the regression relation to define a predicted risk for death from coronary heart disease for each person. Within strata defined by age, sex, and predicted risk, we then did logistic regressions predicting the 2-year risk for death from coronary heart disease or from all causes, as a function of age and serum cholesterol level. This general approach was supported by our preliminary analyses, which found a statistically significant interaction between cholesterol level and the predicted risk for death from coronary heart disease: The risk attributable to cholesterol level was greater for persons at high predicted risk for dying of coronary heart disease. For each stratum, we then estimated the risk for death from coronary heart disease and from all causes at an initial cholesterol level of 280 mg/dL (7.2 mmol/L) and the absolute change in the risk for death that would result from a 40-mg/dL (1.0 mmol/L) decrease in cholesterol level.

Results

3.13 We used two different cutoffs for the "high-risk" category: a 0.25% and a 1% annual incidence of death from coronary heart disease Table 3 and Table 4. Different risk cutoffs are used because at young ages, particularly in women, a high incidence of coronary heart disease is unusual; the opposite is true at older ages. For example, a 65-year-old woman with a cholesterol level of 240 mg/dL [6.2 mmol/L] has more than 60 times the incidence of death from coronary heart disease of a 35-year-old woman with a similar cholesterol level (Table 3). In the study sample, fewer than 2% of women younger than 55 years of age had a predicted incidence greater than 0.25%, and less than half as many had an incidence greater than 1%. The highest risk category is also rare among young men. In contrast, not a single man 55 years of age or older had a predicted incidence less than 0.25%.

3.14 A reduction in cholesterol levels decreases the predicted incidence of death from coronary heart disease in every category, but for younger age groups and for older persons with a low short-term (within 5 years of cholesterol measurement) risk for death from coronary heart disease, the magnitude of the reduction is small and, in many cases, not statistically significant. The absolute change in the probability of dying of coronary heart disease is greater for men and women at high risk than for those at low risk. If, for example, treatment is considered worthwhile as long as it prevents one death from coronary heart disease each year for every 1000 persons treated, the data in Table 3 and Table 4 imply that only men and women with a predicted incidence of death from coronary heart disease of at least 0.25% qualify for treatment. As noted, few women younger than 55 years of age are in this category.

3.15 The association between cholesterol levels and total mortality is weaker and is consistent with that found in the previous studies described above [84, 95]. Although our analyses exclude persons with cholesterol levels less than 180 mg/dL (4.7 mmol/L), who make up the left limb of the U-shaped relation between cholesterol levels and mortality, the positive association between cholesterol level and total mortality is statistically significant only for men younger than 55 years of age whose baseline risk for coronary heart disease is at least 0.25%. Within this age range, the size of the hypothetical mortality benefit from cholesterol reduction increases with advancing age. Cholesterol is not a risk factor for total mortality in men 55 years of age and older. However, the lack of a positive association at ages younger than 65 years may have been masked by combining all ages older than 55 years in the stratum; a negative association is seen between cholesterol levels and mortality at ages 65 years and older. The association with total mortality is not statistically significant for any category of women; often, the trend is in the wrong direction (that is, risk for death decreases as cholesterol level increases).

Summary

3.16 These findings from our analysis of the Framingham data and from other observational studies imply that 1) cholesterol reduction offers small potential benefits in young men and women, even if they have other cardiac risk factors, because the incidence of coronary heart disease is so low in this group; 2) the potential reduction in mortality from coronary heart disease is only slightly greater in middle-aged women; 3) the reduction in mortality is substantially greater in middle-aged men, particularly if other risk factors are present; and 4) the reduction in mortality decreases with advanced ages.

4.0 Other Screening Considerations

4.1 The preceding discussion addressed screening for the identification of persons who might benefit from treatment of an elevated cholesterol level. We now discuss other aspects of screening. We consider a special high-risk group, persons with familial hypercholesterolemia; evaluate the role of cholesterol testing in promoting behavioral change; examine the potential role of other tests that might be used to screen for high-risk lipid profiles; and address the frequency with which screening tests should be administered. We then describe how clinical trial and observational data can be used to identify groups that might benefit from screening for lipoprotein disorders and discuss the cost-effectiveness of cholesterol screening in these populations.

Familial Hypercholesterolemia

4.2 Heterozygous familial hypercholesterolemia, a hereditary disorder with a prevalence in the general population of about 1 per 500 persons, is associated with high levels of LDL and total cholesterol and an elevated risk for coronary heart disease. Major features of the disorder include its autosomal dominant inheritance pattern and the presence of tendinous xanthomas [96, 97]. Family history may be useful in identifying persons with familial hypercholesterolemia and is the only practical approach for the rare homozygous form of the disease. The relative risk for coronary heart disease associated with heterozygous familial hypercholesterolemia is moderate, and most patients with this condition do not develop coronary heart disease until 40 years of age or older. The relative risk appears to decrease with age, and the disorder may not be associated with an excess risk for coronary heart disease among men or women 60 years of age and older [98].

Measuring Cholesterol Levels To Promote Behavioral Change

4.3 Cholesterol screening programs, like other cardiac risk-factor modification programs, may stimulate and reinforce behavioral change by focusing attention on diet and cardiac risk. This benefit will be realized if screening results in dietary change and if the resulting change in diet decreases cholesterol levels. However, few studies have investigated the effects of an intervention that consists solely of measuring cholesterol levels; some studies [99, 100] reported that persons who are found to have an elevated cholesterol level are more likely than unscreened persons to achieve short-term reductions in cholesterol levels of about 2% to 4%, and another [101] found no effect. A randomized trial [102] and an observational study [103] examining the combined effects of cholesterol screening and instruction in dietary or lifestyle modification reported cholesterol reductions of about 2% and 7%, respectively; in a randomized trial of periodic health checkups that included cholesterol measurement [104], the mean cholesterol level decreased by 3.1%. The duration of the decrease in cholesterol level is unknown, but these results are similar to those of trials of cholesterol-lowering diets in the community and suggest that the long-term reduction in cholesterol levels will be about 2% to 3%.

4.4 Because no evidence suggests that a step 1 diet is harmful, even the small reduction in average cholesterol levels might justify screening to promote behavioral change. Concerns about screening to reinforce dietary compliance, however, extend from the actions that may be deemed necessary if, as will usually be the case, the cholesterol level remains elevated. If failure to achieve targeted cholesterol levels results in a recommendation for drugs, the pertinent issue becomes the safety and effectiveness of the drug therapy that is likely to follow. Furthermore, the potential harm due to "labeling" hypercholesterolemic persons, including worsened subjective health status and a diminished sense of well-being, must also be considered [105].

Roles of Other Lipoprotein Screening Tests

4.5 Prediction of risk for coronary heart disease improves when HDL or LDL cholesterol levels are used in addition to total cholesterol levels [106-108]. One study [88] reported that a low HDL cholesterol level predicts both coronary heart disease and all-cause mortality in the elderly. Limited clinical trial evidence suggests that in a person whose HDL cholesterol level increases over time, the risk for developing coronary heart disease is reduced [109, 110]. Such evidence led the second Adult Treatment Panel [9] to modify earlier recommendations by adding the HDL cholesterol level to initial screening, noting that this level added information about risk for coronary heart disease that would be missed by measuring total cholesterol levels alone. The panel designated an HDL cholesterol level of less than 35 mg/dL (0.9 mmol/L) to be a risk factor for coronary heart disease and a level of 60 mg/dL (1.6 mmol/L) or more to be a "negative risk factor" (that is, off-setting one risk factor).

4.6 In the past, neither HDL nor LDL cholesterol measurements in routine clinical use were as well standardized as total cholesterol assays or other common laboratory tests. Further, the usual approach to measuring the LDL cholesterol level requires a fasting blood specimen for measurement of triglyceride levels. New analyzers that can do direct LDL cholesterol measurement, sometimes combined with HDL cholesterol measurement, may offer a practical way to incorporate lipoprotein fractions into a screening program [111]. Information on their accuracy in routine clinical settings is not yet available. However, LDL cholesterol levels are unlikely to dramatically improve the risk discrimination provided by the total cholesterol level alone because the two measurements are highly correlated.

4.7 The role of the triglyceride level as a screening test and as a risk factor for coronary heart disease has long been controversial [112]. Although many studies have reported that triglyceride levels predict the incidence of coronary heart disease when the effects of other risk factors are ignored, the triglyceride level is usually not a strong independent predictor of risk after adjustment for HDL and LDL cholesterol levels. This finding probably results from the strong negative correlation between HDL cholesterol and triglyceride levels; because single measurements of triglycerides vary more than HDL cholesterol measurements, they are not likely to predict coronary events as precisely as HDL cholesterol [113]. Triglyceride levels might identify persons in whom lipid-lowering treatment is especially effective. Two such "interactions" have been proposed [114, 115]. However, until clinical trials establish the benefit of identifying such interactions [116], evidence is insufficient to support using blood triglyceride measurements in routine screening to prevent coronary heart disease [112, 113].

Frequency of Screening

4.8 How often should cholesterol tests be done in persons in whom cholesterol screening is appropriate? The rule of thumb proposed by the National Cholesterol Education Program and other expert groups, to test every 5 years, is reasonable for many adults; more frequent screening might be appropriate for anyone whose cholesterol level approaches a threshold for initiating treatment. If the cholesterol level is well below a treatment threshold, less frequent measurement may be sufficient; a person who has had a normal cholesterol measurement (< 200 mg/dL [5.2 mmol/L]) is unlikely to have a "high-risk" cholesterol level (> 240 mg/dL [6.2 mmol/L]) in the following 6 years [117]. Thus, the U.S. Preventive Services Task Force [10] recommends measuring cholesterol "at least once in men ages 35-65 and women ages 45-65," with further testing depending on the results of the initial test.

Cost-Effectiveness of Screening

4.9 The cost-effectiveness of cholesterol screening is similar to that of treatment, because treatment is responsible for most of the costs of cholesterol screening programs [118]. The costs of a screening program include direct expenditures for complying with diets and drugs and with associated follow-up care, the savings resulting from preventing myocardial infarction and other manifestations of coronary heart disease, and the costs of any diseases or health conditions that result from treatment. The "net costs" make up the numerator of the cost-effectiveness ratio; the denominator is usually the increase in life expectancy (or in quality-adjusted life-years) that results from treatment. Published studies of the cost-effectiveness of treatment assume that cholesterol reduction decreases the incidence of coronary heart disease and does not increase morbidity or mortality from any other cause [119]. Insofar as this assumption is incorrect, such studies overestimate the beneficial health effects and may underestimate the costs of the intervention.

4.10 Despite these optimistic assumptions, most cost-effectiveness studies show that only in persons at high risk for developing coronary heart disease does the cost-effectiveness of cholesterol reduction compare favorably with that of other interventions. For example, Goldman and colleagues [120] report that using low-dose lovastatin to treat high blood cholesterol levels ( to 300 mg/dL [7.8 mmol/L]) in middle-aged male smokers who had moderate to severe diastolic hypertension resulted in a cost per year of life saved of $28 000 to $48 000. In 35- to 44-year-old women with similar cholesterol levels but no other risk factors, the estimated cost per year of life saved was $1.5 million. Figure 1, an extrapolation of these values, shows that the cost per year of life saved increases markedly at younger ages. A more recent cost-effectiveness study by Hamilton and colleagues [121], which accounted for the effects of lovastatin on the HDL cholesterol level and which assumed that the benefits of cholesterol reduction decreased with age, produced lower cost-effectiveness ratios. However, these ratios were still unfavorable in "low-risk" men with high blood cholesterol levels who were younger than age 40 years and in women with high blood cholesterol levels who were younger than age 60 years.

View larger version (28K): [in this window] [in a new window]   Figure 1. Cost-effectiveness of treating high blood cholesterol levels with lovastatin, 20 mg/d, drawn (solid lines) or extrapolated (dotted lines) from tabular data from Goldman and colleagues [120]. The models assume only favorable effects on mortality, do not include screening costs, and define high blood cholesterol levels as at least 300 mg/dL (7.76 mmol/L) for primary prevention in persons without other risk factors (low-risk) and as at least 250 mg/dL (6.47 mmol/L) for secondary prevention in persons with previous coronary heart disease (CHD). (The cost of treating 35- to 54-year-old men with CHD is zero because savings from future CHD outweigh the costs of treatment.) The ordinate is a log scale; the dashed line is at $100 000 per year of life saved. Reproduced with permission from The Journal of the American Medical Association. 1993; 269:1416-9.

4.11 These findings result from the low rates of death from coronary heart disease, even in the presence of high blood cholesterol levels, in most young and middle-aged persons. Because the hypothetical benefits of treating elevated cholesterol levels in low-risk populations necessarily occur in a small proportion of those treated, the cost per year of life saved is high. Conversely, decreasing cholesterol levels in survivors of myocardial infarction and other high-risk populations is highly cost-effective (or even cost-saving).

5.0 Conclusions

5.1 Any screening program requires strong evidence that benefits exceed harms before widespread adoption can be recommended. Yet, the foundation of scientific data to support cholesterol screening in many segments of the adult population remains incomplete. The main purpose of screening is to identify persons who would benefit from specific cholesterol-lowering treatment. Clinical trials have shown that patients with high blood cholesterol levels and symptomatic coronary heart disease are in this category. It is probable that men and women with other forms of atherosclerotic vascular disease are also in this category. In the much larger population of adults with high blood cholesterol levels but no clinical evidence of atherosclerotic disease, cholesterol reduction probably prevents the manifestations of coronary heart disease; however, no direct evidence suggest that cholesterol reduction prolongs life.

5.2 When used for primary prevention, cholesterol reduction prevents coronary heart disease in middle-aged men with hypercholesterolemia and might prolong life in middle-aged men and women whose cholesterol levels and other risk factors place them at greater short-term risk for coronary heart disease than the men enrolled in most primary prevention trials. A survival benefit might also be found in lower-risk populations if the treatment to be used has a more favorable risk–benefit profile than the treatments studied in published trials.

5.3 Primary prevention in younger populations (for example, men less than 35 years of age and women less than 45 years of age) is untested in clinical trials and has smaller hypothetical benefits. Exceptions are likely to be persons who have a history or physical examination suggestive of familial hypercholesterolemia or who are otherwise at high risk for coronary heart disease. In elderly persons, the association between cholesterol level and risk for coronary heart disease disappears by the late 70s, and neither clinical trial nor epidemiologic evidence suggests that cholesterol reduction is beneficial at older ages.

5.4 Uncertainty about the benefits and potential harms of cholesterol-lowering therapies is greatest for men and women aged 65 to 75 years, when cholesterol remains a risk factor for cardiovascular disease and coronary atherosclerosis is common. Both the risks and benefits of therapy may be heightened; it is impossible to confidently predict which will be greater for any class of lipid-lowering drugs. Randomized, controlled clinical trials are needed to establish who among the elderly are most likely to benefit from cholesterol reduction. Such trials offer the hope of improving our ability to safely and effectively prevent coronary heart disease.

Dr. Hulley: Department of Epidemiology and Biostatistics, University of California, San Francisco, Box 0560, San Francisco, CA 94143-0560.