Effect of Combination Therapy with Lipid-Reducing Drugs in Patients with Coronary Heart Disease and “Normal” Cholesterol Levels

Methods

Patients

The design of the Harvard Atherosclerosis Reversibility Project (HARP) has been described in detail [34]. Patients were eligible for the project if they were nonsmokers aged 30 to 75 years and had 1) angiographically documented stenosis [≥ 30%] of a major coronary artery, 2) a total cholesterol level between 4.7 and 6.5 mmol/L [180 and 250 mg/dL], and 3) a ratio of total cholesterol to HDL cholesterol greater than 4.0. Both men and women were eligible, but women had to be postmenopausal or surgically sterile. Patients were excluded if they were taking hypolipidemic drugs; had insulin-dependent diabetes mellitus; had other endocrine diseases or secondary causes of hyperlipidemia; had symptomatic congestive heart failure or an ejection fraction less than 30%; or had other debilitating, chronic, noncardiac diseases that could have interfered with full participation in the study.

Screening and Randomization

All patients who had cardiac catheterization were screened for eligibility by study coordinators at either Beth Israel Hospital or Brigham and Women's Hospital, Boston, Massachusetts. A registered dietitian instructed all patients about the American Heart Association and the NCEP step I diet [13]. Each patient participated in two screening visits; the first of these occurred at least 3 weeks after discharge from the hospital for cardiac catheterization or coronary angioplasty or at least 6 weeks after coronary artery bypass graft surgery. The screening visits were 1 to 2 weeks apart. At each screening visit, blood lipid levels and safety were measured, a 7-day dietary record was collected, and dietary instruction was reinforced.

Eligible patients returned 1 to 2 weeks later for another visit (the randomization visit), during which they received a medical evaluation by a study physician, received an assessment of physical activity, and had final measurement of prerandomization blood lipid level. In patients who had had coronary artery bypass graft surgery at baseline, therapy with study medications was initiated 12 weeks or more after surgery. Patients were randomly assigned in a blinded manner to receive drug therapy or matching placebo. Randomization was stratified according to whether patients had received medical or surgical management for their coronary disease and according to whether their ratio of total cholesterol to HDL cholesterol was greater than 6.0 or was 6.0 or less.

Stepped-Care Drug Algorithm

The stepped-care therapy program was designed by the study investigators and approved by the Food and Drug Administration. The organizations that funded the study did not participate in the collection, analysis, or interpretation of data. The six-step algorithm (Table 2) used standard dosages of four agents: pravastatin, 40 mg/d; nicotinic acid, 1.5 to 3.0 g/d (slow-release tablets, Endurance Products, Tigard, Oregon); cholestyramine resin, 4 to 24 g/d (Questran powder, Bristol-Myers Squibb, Princeton, New Jersey); and gemfibrozil, 600 to 1200 mg/d (Lopid tablets, Parke-Davis, Morris Plains, New Jersey) (Table 2). The goal was to achieve a total cholesterol level of 4.1 mmol/L (160 mg/dL) or less and a ratio of LDL cholesterol to HDL cholesterol of less than 2.0. If the LDL cholesterol level decreased to less than 2.1 mmol/L (80 mg/dL), additional treatment steps were not introduced, even if the other criteria were not met. All patients who received drug or matching placebo were seen every 6 weeks for measurement of lipid levels and evaluation of side effects. Patients began receiving the next treatment step after 6 weeks if they had not reached the target goal, as confirmed by two lipid profiles done at least 1 week apart. After reaching the target goal, patients were seen every 3 months. If adverse effects occurred, therapy with the agent deemed responsible was decreased or discontinued as specified by the protocol. If the LDL cholesterol level was 1.3 mmol/L (50 mg/dL) or less on two consecutive measurements, therapy with the most recently added drug was decreased or discontinued. Use of lipid-reducing agents other than those provided in accordance with the protocol was prohibited. The strategy for managing the side effects of therapy was also standardized (Table 2). Patients were specifically instructed to take the cholestyramine resin at times when its use would not interfere with the absorption of other study or nonstudy drugs.

If the total cholesterol level of a patient in the control group increased to 6.5 mmol/L (250 mg/dL) or more on two consecutive measurements, intensified dietary instruction was provided using the NCEP step II regimen [13]. If the total cholesterol level remained greater than 6.5 mmol/L (250 mg/dL), monotherapy with cholestyramine resin (4 to 8 g/d), nicotinic acid (500 to 1000 mg/dL), or lovastatin (10 mg) was started to decrease the total cholesterol level to less than 6.5 mmol/L.

Study medications were dispensed in bottles or packages at the time of randomization and at each quarterly visit. Adherence to the prescribed dosages was assessed every 3 months on the basis of pill and package counts.

An interval medical history, including information on the use of concomitant medications and an assessment of physical activity, was obtained every 12 weeks. Physical examinations were done every 6 months.

Laboratory and Safety Analyses

Levels of total cholesterol, total triglycerides, and HDL cholesterol were measured on fresh plasma at the Lipid Research Laboratory, Brigham and Women's Hospital, which participated in the Lipid Standardization Program of the Centers for Disease Control and Prevention (CDC) (Atlanta, Georgia). The mean coefficients of variation were 1.7% for total cholesterol level and 2.7% for HDL cholesterol level. All blood samples were obtained after patients had been fasting for at least 10 hours. Laboratory personnel were blinded to treatment assignments. Levels of total cholesterol, HDL cholesterol, and total triglycerides were measured by enzymatic reagents (Boehringer-Mannheim, Indianapolis, Indiana). We separated HDL cholesterol by precipitating LDL cholesterol and very-low-density cholesterol with dextran sulfate and magnesium chloride. The LDL cholesterol level was calculated using the Friedewald formula: LDL cholesterol level = (total cholesterol level − HDL cholesterol level − triglyceride level)/5. Apolipoprotein levels were measured in plasma specimens that were stored at −80°C for 2 to 4 years; these specimens were collected from a subsample of 79 patients who completed coronary angiography after treatment [34]. Plasma apolipoprotein B levels were measured by immunonephelometry (Inkstar, Stillwater, Minnesota) on an autoanalyzer (Cobas Mira Plus; Roche Diagnostics, Belleville, New Jersey) using calibration standards provided by the CDC. These values were corrected to the recently validated standard of the World Health Organization [35]. Lipoprotein(a) levels were measured by enzyme-linked immunosorbent sandwich assay using polyclonal antilipoprotein(a) for both capture and detection (Terumo, Strategic Diagnostics, Newark, Delaware). The rate of color development in each well was used to calculate concentrations.

Safety measurements, done every 12 weeks, comprised a complete blood count with a differential leukocyte count; a serum chemistry panel that included measurement of alkaline phosphatase, total bilirubin, aspartate aminotransferase, alanine aminotransferase, γ-glutamyltranspeptidase, and creatine phosphokinase levels; and measurement of thyroxine levels, urinalysis, and a test for occult blood. In addition, a limited chemistry panel for liver enzymes was done every 6 weeks. Testing was done by Metpath (Teterboro, New Jersey) on specimens shipped by overnight carrier. Persons with amino transferase levels more than twice but less than three times the upper limit of normal were monitored every 2 weeks until the elevations resolved or were recognized as stable. Aminotransferase levels more than three times the upper limit of normal on repeated measurement necessitated discontinuation of therapy with or reduction in the dosage of the most recently added study medication.

Statistical Analysis

Baseline values for lipid levels were defined as the mean of three measurements obtained at the two screening visits and the randomization visit. The values for lipid levels at each step of the treatment algorithm were the average of measurements made at all visits during that step. Baseline characteristics of the treatment and control groups were compared using the chi-square test (categorical data) or the two-sample t-test (continuous data). The mean changes in plasma lipid levels over time within the two study groups were tested for statistical significance using the paired t-test. The differences in the mean change in blood lipid levels between the two study groups were tested at 12 weeks and at the end of follow-up using the two-sample t-test. The proportions of patients reaching the lipid reduction goal in each study group were compared using the Fisher exact test. The mean change in lipid levels at each step in the algorithm was compared with that at the preceding step using the paired t-test. All analyses were done using SAS statistical software (SAS Institute, Cary, North Carolina). The logarithmic transformation was used to approximately normalize the distribution of the triglyceride and lipoprotein(a) levels. The tests of treatment effects on lipid levels were two sided; P values less than 0.05 were considered statistically significant. The primary study end point was the proportion of patients reaching the target goal.

Results

We randomly assigned 91 patients (80 men and 11 women) to receive stepped-care pharmacologic therapy (n = 44) or placebo (n = 47). The clinical characteristics of the two groups were similar (Table 3). The mean age was 58 years, and the mean body mass index was 28 kg/m². Of the 91 patients, 41 (45%) were being treated for hypertension and 9 (10%) had non–insulin-dependent diabetes mellitus. The average length of follow-up was 125 weeks (range, 4 to 190 weeks). Seventy patients (77%) completed at least 2 years of treatment. The trial ended after a mean duration of treatment of 2.5 years. Nine patients did not complete their scheduled treatment: Five were withdrawn by their physicians so that they could receive open-label treatment for lipids (1 patient in the treatment group and 4 controls), 1 control was withdrawn because of serious noncardiac illness, and 3 participants died (1 patient in the treatment group and 1 control died of cardiac causes, and 1 control died of cancer). For all patients who were randomly assigned to treatment, all data until study completion or withdrawal were included in the analyses.

Changes Seen with Diet

Both groups had similar compliance with dietary therapy. At baseline and during treatment, the average percentage of calories contributed to the diet by saturated fat was 8% in both groups. In the treatment group, mean intake of dietary cholesterol was 223 mg/d at baseline and 185 mg/d during treatment; in the control group, mean intake was 221 mg/d at baseline and during treatment (P < 0.05 for comparison between the groups). The change in intake of dietary cholesterol that occurred in the treatment group would be expected to decrease plasma levels of total cholesterol by 0.01 mmol/L (0.4 mg/dL) [36].

Drug Therapy in Controls

Twelve of the 47 controls needed drug therapy for total cholesterol levels greater than 6.5 mmol/L (250 mg/dL). Cholestyramine (mean dosage, 8 g/d) was prescribed for 11 of these 12 controls for a mean of 37 weeks; nicotinic acid (mean dosage, 1 g/d) was prescribed for 4 of the 12 for a mean of 26 weeks; gemfibrozil (mean dosage, 900 mg/d) was prescribed for 4 of the 12 for a mean of 26 weeks; and lovastatin (dosage, 10 mg/d) was prescribed for 1 of the 12 for 90 weeks. Eight of the 12 controls (67%) developed intolerable side effects from either nicotinic acid or cholestyramine and were prescribed one of the other drugs.

Changes in Plasma Lipid and Apolipoprotein Levels over Time

Figure 1 shows the changes that were seen in plasma lipid and apolipoprotein levels over time. Our entry criteria produced a study sample that had “normal” plasma lipid levels, as intended; lipid and apolipoprotein levels did not differ significantly between the two groups at baseline. By 12 weeks, patients in the treatment group had had substantial and significant decreases (P < 0.001 for all comparisons) in mean total cholesterol level (from 5.5 to 4.1 mmol/L [214 to 159 mg/dL]; −25%),LDL cholesterol level (from 3.6 to 2.3 mmol/L [140 to 90 mg/dL]; −36%),and triglyceride level (from 1.8 to 1.4 mmol/L [159 to 122 mg/dL]; −23%).As additional agents were used, LDL cholesterol levels continued to decline, whereas total cholesterol and triglyceride levels remained relatively stable. Mean plasma apolipoprotein B levels decreased from 116 mg/dL to 93 mg/dL ( −20%)at 6 weeks, 79 mg/dL ( −32%)at 48 weeks, and 87 mg/dL ( −25%)at 96 weeks. In the control group, levels of cholesterol, apolipoprotein B, and triglycerides gradually increased over time.

In the treatment group, a modest but significant (P < 0.001) increase was seen in the mean HDL cholesterol level (from 1.1 to 1.2 mmol/L [42 to 46 mg/dL]; +10%) within the first 12 weeks of drug therapy, and this level steadily increased over time with the addition of nicotinic acid and gemfibrozil. By late in the follow-up period, the mean HDL cholesterol level was 1.4 mmol/L (53 mg/dL), a change of +26% (Figure 1B). In the control group, HDL cholesterol levels were unchanged. Geometric mean plasma lipoprotein(a) levels at baseline were 4 mg/dL (range, 0.3 to 61 mg/dL) in both groups and did not significantly change during the study (Figure 1F). In a subgroup of 11 patients in the treatment group whose baseline lipoprotein(a) levels were greater than 10 mg/dL, the geometric mean lipoprotein(a) level was 23 mg/dL at baseline and 21 mg/dL during drug treatment (P > 0.05).

Forty-one patients in the treatment group (93%) reached the predetermined goal: a total cholesterol level of 4.1 mmol/L (160 mg/dL) or less and either a ratio of LDL cholesterol to HDL cholesterol less than 2.0 or an LDL cholesterol level less than 2.1 mmol/L (80 mg/dL). The mean time to achievement of this goal was 21 weeks. No controls reached the goal. The difference between the proportion of patients in the treatment group and the proportion of controls who reached the goal was highly significant (P < 0.001).

Changes at Each Drug Treatment Step

Figure 2 shows the changes that were seen at each drug treatment step.

Pravastatin

Pravastatin, 40 mg/d, decreased total cholesterol levels by 22%, decreased LDL cholesterol levels by 32%, decreased triglyceride levels by 15%, and increased HDL cholesterol levels by 8% (P < 0.001 for all comparisons) (Table 4). Pravastatin also decreased the ratio of total cholesterol to HDL cholesterol by 28% (P < 0.001) and the ratio of LDL cholesterol to HDL cholesterol by 37% (P < 0.001).

Nicotinic Acid, 1.5 g

Nicotinic acid, 1.5 g/d, was added to pravastatin for 40 of the 44 patients in the treatment group. Nicotinic acid caused additional mean reductions of 6% in total cholesterol levels (P < 0.002), 11% in LDL cholesterol levels (P < 0.001), and 10% in triglyceride levels (P < 0.001), and it caused an additional increase in HDL cholesterol levels of 8% (P < 0.001). Nicotinic acid (1.5 g/d) and pravastatin (40 mg/d) combined produced changes of −26% in total cholesterol levels, −39% in LDL cholesterol levels, −23% in triglyceride levels, +17% in HDL cholesterol levels, −35% in the ratio of total cholesterol to HDL cholesterol, and −46% in the ratio of LDL cholesterol to HDL cholesterol (P < 0.001 for all comparisons).

Nicotinic Acid, 2.25 to 3 g

When the dose of nicotinic acid was increased to 2.25 to 3 g/d in combination with pravastatin in 23 patients, further reductions were seen. Total cholesterol levels were reduced by an additional 7% (P = 0.007), LDL cholesterol levels were reduced by an additional 14% (P < 0.001), and triglyceride levels were reduced by an additional 13% (P = 0.03). Levels of HDL cholesterol increased by an additional 6% (P = 0.02). The total effect of nicotinic acid (2.25 to 3 g/d) and pravastatin (40 mg/d) was to decrease total cholesterol levels by 29%, to decrease LDL cholesterol levels by 44%, to decrease triglyceride levels by 26%, and to increase HDL cholesterol levels by 20% (P < 0.001 for all comparisons). The ratio of total cholesterol to HDL cholesterol decreased by 41%, and the ratio of LDL cholesterol to HDL cholesterol decreased by 53%. The 23 patients who required the higher dose of nicotinic acid had almost the same response to pravastatin as did the overall group. The LDL cholesterol levels of this subgroup at baseline (3.88 mmol/L [150 mg/dL]) tended to be higher, but not significantly higher, than those of the overall group (3.62 mmol/L [140 mg/dL]). However, the levels of HDL cholesterol (1.14 mmol/L [44 mg/dL] in the treatment group and 1.09 mmol/L [42 mg/dL] in the control group) and triglycerides (1.75 mmol/L [155 mg/dL] in the treatment group and 1.80 mmol/L [159 mg/dL] in the control group) were similar in the two groups.

Cholestyramine

Cholestyramine, 4 to 24 g, was added to pravastatin and nicotinic acid in 21 patients, according to the study protocol. The addition of this resin partially reversed the action of the previous steps. The mean triglyceride level increased by 46% (P < 0.001), and the mean HDL cholesterol level decreased by 8% (P = 0.03) relative to the levels achieved at the previous step. The LDL cholesterol level did not change. In the group receiving cholestyramine, the responses to pravastatin and nicotinic acid were similar to those in the overall treatment group (changes in LDL cholesterol were −35% in the group receiving cholestyramine and −39% in the overall treatment group; changes in HDL cholesterol were +16 in the group receiving cholestyramine and +17% in the overall treatment group). The lipid levels at baseline for these patients and for the overall treatment group were 5.77 mmol/L and 5.53 mmol/L ([223 mg/dL and 214 mg/dL], respectively, for total cholesterol levels; 3.85 mmol/L and 3.62 mmol/L (149 mg/dL and 140 mg/dL), respectively, for LDL cholesterol levels; 1.09 mmol/L and 1.09 mmol/L (42 mg/dL and 42 mg/dL), respectively, for HDL cholesterol levels; and 1.85 mmol/L and 1.80 mmol/L (164 mg/dL and 159 mg/dL), respectively, for triglyceride levels (P > 0.05 for all comparisons).

Gemfibrozil

Twelve patients required the final treatment step—gemfibrozil, 600 to 1200 mg/d. In contrast to the lack of overall benefit produced by the addition of cholestyramine, the addition of gemfibrozil increased HDL cholesterol levels by 12% (P = 0.03) and markedly decreased triglyceride levels by 37% (P < 0.001) relative to the levels achieved at the previous step. The LDL cholesterol level increased by 12% (P = 0.09). The baseline lipid levels of the subgroup that received gemfibrozil were similar to those of the group that began receiving the previous treatment step.

National Cholesterol Education Program Guidelines

In our study, 35 of 44 patients in the treatment group (79%) had LDL cholesterol levels at baseline that were higher than the current NCEP threshold for treatment (3.35 mmol/L). These patients had a mean LDL cholesterol level of 3.9 mmol/L (149 mg/dL) (range, 3.4 to 4.9 mmol/L [131 to 188 mg/dL]). Pravastatin, 40 mg, decreased the mean LDL cholesterol level to 2.6 mmol/L in 18 of 35 patients (50%) by the 6-week visit. The addition of nicotinic acid was needed to reduce the LDL cholesterol level to less than 2.6 mmol/L in 15 of the 35 patients (43%). One of the 35 patients (3%) required the addition of both nicotinic acid and cholestyramine. The LDL cholesterol level was never less than 2.6 mmol/L in 1 of the 35 patients (LDL cholesterol level at baseline, 3.6 mmol/L [140 mg/dL]) despite the addition of gemfibrozil. Monotherapy with pravastatin maintained LDL cholesterol levels at less than 2.6 mmol/L at most follow-up visits for the full 2.5 years of the study in 11 of these 35 patients (30%). In 4 of the 5 patients (80%) whose baseline LDL cholesterol levels were 2.6 to 3.35 mmol/L (mean, 3.0 mmol/L [118 mg/dL]), pravastatin reduced LDL cholesterol levels to less than 2.6 mmol/L and maintained the reduction.

Adverse Effects

Table 5 shows the adverse effects that occurred in our study. Fifty-nine percent of patients in the treatment group and 6% of controls reported gastrointestinal symptoms. Dermatologic symptoms were common during treatment with nicotinic acid. One patient discontinued monotherapy with pravastatin because of symptoms of peptic ulcer; the other gastrointestinal symptoms reported were cases of mild nausea that did not require changes in dose. During combination therapy with pravastatin and nicotinic acid (1.5 g/d), 4 of 40 patients (10%) required a decrease in the dose of nicotinic acid to 0.75 to 1 g/d, and 3 of 40 patients (7.5%) discontinued therapy with nicotinic acid because of gastrointestinal or dermatologic symptoms. When the dose of nicotinic acid was increased to 2.25 to 3 g/d, 12 of the 23 patients (52%) reported clinical symptoms (usually an intensification of symptoms tolerated at the lower dose) that required a decrease in dose in 6 patients and discontinuation of therapy in 2 patients. When cholestyramine was added to the therapy regimen in 21 patients, 7 of these patients (33%) reported new gastrointestinal symptoms and 11 (52%) discontinued therapy with the drug. With gemfibrozil, 3 of 12 patients (25%) reported gastrointestinal side effects, and 2 of these patients discontinued therapy with the drug. One patient was hospitalized for a bleeding gastric ulcer that occurred during therapy with pravastatin and nicotinic acid and was attributed to the nicotinic acid. One patient developed a severe skin rash that waxed and waned during therapy with multiple hypolipidemic and antianginal medications; it was impossible to attribute this to a particular medication.

No patient had confirmed elevation of a serum liver enzyme level to more than three times the upper limit of normal while receiving study medication. Sporadic elevations in creatine phosphokinase levels occurred in one patient who was receiving pravastatin, nicotinic acid, and cholestyramine; these elevations were attributed to episodic heavy exercise and athletic trauma. However, they necessitated discontinuation of therapy, according to the protocol, at 96 weeks. Blood glucose levels increased in two diabetic patients after they began receiving nicotinic acid, and this required a reduction in dose or discontinuation of therapy. Serum thyroxine levels decreased in two patients who were receiving nicotinic acid, but no change in dose was necessary. No arrhythmias were documented as a side effect of therapy.

Discussion

Our goal was to evaluate the efficacy and tolerability of a clinical algorithm for the use of stepped-care drug therapy in patients with documented coronary artery disease who did not have hyperlipidemia. We have shown that marked improvement in levels of total cholesterol, LDL cholesterol, HDL cholesterol, apolipoprotein B, and triglycerides can be produced and maintained during a 2- to 3-year period. The combination of pravastatin and nicotinic acid produced the substantial improvement seen in LDL and HDL cholesterol levels, and it decreased the ratio of LDL cholesterol to HDL cholesterol by 50%. Furthermore, we have shown that stepped-care therapy with pravastatin and nicotinic acid can decrease LDL cholesterol levels in normolipidemic patients who have heart disease to the stringent NCEP goal of 2.6 mmol/L (100 mg/dL) or less [13]. Combination therapy was needed in 70% of the patients whose levels of LDL cholesterol before treatment were greater than 3.35 mmol/L (130 mg/dL), despite the considerable effectiveness of monotherapy with pravastatin in reducing these levels. Cholestyramine, as a third therapeutic agent, did not improve lipid levels for unknown reasons, but gemfibrozil produced a substantial decrease in triglyceride levels, a mild but significant increase in HDL cholesterol levels, and a nonsignificant increase in LDL cholesterol levels.

Previous Studies

Various combination therapy regimens have been used successfully in hypercholesterolemic patients. Using lovastatin and the bile-acid binding resin colestipol, Grundy and coworkers [19] reduced LDL cholesterol levels by 52% and increased HDL cholesterol levels by 30% in a small group of patients with heterozygous familial hypercholesterolemia. In other studies [25, 26], pravastatin combined with a bile-acid binding resin reduced LDL cholesterol levels by 45% to 56% and increased HDL cholesterol levels by 11% to 18%. These apparently additive effects of pravastatin and resin are consistent with the effects seen in other studies of hypercholesterolemic patients treated with other HMG CoA reductase inhibitors and a resin [20-23]. In contrast, in a large study of hypercholesterolemic patients, Stein and associates [24] found that simvastatin combined with cholestyramine decreased total and LDL cholesterol levels by only 15% and 21%, respectively. The combination of lovastatin, colestipol, and nicotinic acid was evaluated in 17 patients with familial hypercholesterolemia [18]. In the 6 patients who were treated for more than 2 years, total cholesterol levels decreased by 56%, LDL cholesterol levels decreased by 65% to 68%, and HDL cholesterol levels increased by 28% to 32%. A trial of nicotinic acid and colestipol [10] found that the combination of these drugs produced decreases in LDL cholesterol levels that were greater than the decreases expected from either drug alone. The combination of a statin with either nicotinic acid [18, 28, 29] or a fibric acid [27, 30-32, 37] appears to be effective and safe under careful monitoring (in agreement with our findings), despite previous concerns [38-42].

We were surprised that cholestyramine had no beneficial effect on LDL cholesterol levels when used in combination with pravastatin and nicotinic acid. As we have already noted, the combination of a resin and an HMG CoA reductase inhibitor generally produced additive benefit [20-2325, 26]. Bile-acid-sequestering resins impair drug absorption and can alter the efficacy of study drugs used during the early stages of our stepped-care regimen. Although we carefully instructed (and frequently reminded) patients to take cholestyramine at least 1 hour after or 4 hours before taking other medications, as recommended by the drug manufacturer and the NCEP guidelines, the interference of cholestyramine with other medications cannot be excluded. The unpleasant taste and the gastrointestinal side effects of cholestyramine can reduce compliance, but it is likely that our patients were taking the resin because their triglyceride levels increased; this is known to be an effect of cholestyramine. On the basis of patient interviews and the return of unused drug supplies, we estimated that patients were 75% compliant with cholestyramine therapy. The LDL cholesterol levels of these patients responded typically to pravastatin and nicotinic acid; thus, the apparent hyporesponsiveness of LDL cholesterol was specific to cholestyramine. Combinations of three hypolipidemic medications have seldom been studied. It cannot be assumed that the additive benefit shown with two drugs can be extended to therapy with three drugs. For example, in hypercholesterolemic patients, adding nicotinic acid to lovastatin and cholestyramine decreased HDL cholesterol levels and increased triglyceride levels [18]. This was the opposite of what was expected from nicotinic acid as monotherapy or in two-drug combinations. In contrast to the unfavorable effect caused by cholestyramine, adding gemfibrozil to pravastatin and nicotinic acid produced the expected changes: marked decreases in triglyceride levels and mild increases in HDL and LDL cholesterol levels. Although it was not significant, the slight increase in LDL cholesterol levels (which has previously been shown by others [30, 39]) is of concern in some patients. We believe that gemfibrozil should be reserved for patients with marked hypertriglyceridemia.

Study Limitations

If each step in the treatment algorithm were considered to be a separate trial with its own “treatment group,” then one limitation of our study would be the relatively small sizes of the groups that used three or four drugs. However, because it was our intention to test a stepped-care treatment strategy, the size of our study group was appropriate. Our results do not provide definitive information on the efficacy or safety of each separate treatment step, but they strongly suggest certain conclusions about the value and tolerability of the overall strategy. Our conclusions are only valid for the strategy tested. The drugs were administered in order of their efficacy for the reduction of LDL cholesterol levels (Table 2). If the order were different, the results might be different.

Because of the relatively small size of our study, no conclusions can be drawn about subgroups that differ according to sex, age, history of coronary artery bypass surgery, or extent of coronary artery disease. However, none of these factors has been shown to affect responses in other, large, single-drug studies [3, 5, 6, 10, 16].

Side Effects

The safety of the treatment algorithm was a concern during the design of the algorithm, and we monitored side effects closely. The design of the algorithm proved to be practical for clinicians, who could deal with side effects as they occurred by moving patients back a step in the treatment algorithm.

Therapy was discontinued because of adverse effects in 2% of patients who received monotherapy with pravastatin for an average duration of 42 weeks, in 16% of patients who received nicotinic acid for a mean of 40 weeks, in 52% for patients who received cholestyramine resin for 33 weeks, and in 17% of patients who were treated with gemfibrozil for an average duration of 43 weeks. These rates are similar to summary estimates of rates of discontinuation from randomized trials of at least 1 year's duration: 16% for lovastatin, 4% for nicotinic acid, 31% for bile-acid-sequestering resins, and 15% for gemfibrozil [42]. Rates of discontinuation for randomized trials may not provide an accurate estimate of the tolerability of hyperlipidemic therapy in the general population because intensive attention is given to compliance in such trials. Moreover, users of secondary agents may be more likely to discontinue therapy.

The combination of an HMG CoA reductase inhibitor and gemfibrozil or nicotinic acid is associated with an increased risk for myositis. The Food and Drug Administration has cautioned against the routine use of these combinations [38]. However, the data on which these recommendations have been based are relatively scant and come from one published trial [39] and several anecdotal reports [38, 40, 42]. In the study by Illingworth and Bacon [39], 12 patients were given lovastatin and gemfibrozil. One patient developed muscle pain and a markedly elevated creatine phosphokinase level, but this patient had marked elevation of enzyme levels (without muscle pain) while receiving lovastatin alone. In our study, none of the 12 patients who received pravastatin and gemfibrozil, all of whom were also receiving nicotinic acid, developed symptomatic elevation of creatine phosphokinase levels, and only 1 developed a confirmed elevation of creatine phosphokinase levels. Our favorable experience with pravastatin and gemfibrozil is similar to that reported by Wiklund and colleagues [27]. The usefulness and safety of these combinations should continue to be cautiously explored.

Clinical Implications

The efficacy of therapy for lipid levels in normolipidemic patients has become particularly important as national trends show that plasma cholesterol levels are decreasing [44], and the proportion of patients with coronary artery disease who are normolipidemic is substantial [14, 15, 45-47]. The algorithm for our study was practical and potent. Structured to bring patients to a target total cholesterol level of 4.1 mmol/L (160 mg/dL) or less and a ratio of LDL cholesterol to HDL cholesterol of 2.0 or more, it took into account the reduction of LDL cholesterol levels and the increase of HDL cholesterol levels. It was intentionally made flexible enough to allow investigators to continue to push patients to the target goal, even when adverse or unpleasant side effects were encountered, by facilitating decisions about changes in drug or drug dosages. We found that the combination of pravastatin, 40 mg, and nicotinic acid, 1.5 to 3.0 g, was particularly effective. It reduced LDL cholesterol levels by 39% to 44%, increased HDL cholesterol levels by 17% to 20%, and decreased the ratio of LDL cholesterol to HDL cholesterol by 46% to 53%.

The NCEP guidelines for the secondary prevention of cardiovascular disease recommend the use of drug therapy 1) to decrease LDL cholesterol levels to 2.6 mmol/L (100 mg/dL) or less in patients whose LDL cholesterol levels are greater than 3.35 mmol/L [130 mg/dL] and 2) as an option for patients with LDL cholesterol levels greater than 2.6 mmol/L [13]. Although HARP was designed before the NCEP guidelines were released, we nevertheless determined the number of medications that would have been needed to achieve this goal. Although pravastatin promptly reduced LDL cholesterol levels to less than 2.6 mmol/L in 50% of patients whose levels had been greater than 3.35 mmol/L, the levels eventually increased to more than 2.6 mmol/L in many of these patients during the 2.5 years of follow-up. Therefore, 70% of patients who had only mildly elevated LDL cholesterol levels needed combination therapy to reach the NCEP goal, even when a potent LDL cholesterol-decreasing medication was used for initial therapy. The patients who required additional therapy were not hyporesponsive to pravastatin or low-dose nicotinic acid but rather tended to have higher initial LDL cholesterol levels. Combination drug therapy for lipids is not simple, and careful clinical management is needed because of the frequent, minor side effects that reduce adherence and the unusual major adverse reactions that can occur. We think that this stepped-care algorithm is a practical clinical approach that, when used systematically, would lead to rational drug selection and the use of a minimum of drugs at tolerable doses, as necessary to achieve the NCEP target goal. Thus, this and other stepped-care approaches should continue to be studied and may prove useful for patients with established coronary heart disease, many of whom should have their lipid levels treated aggressively.

Ms. Brown: CARE Coordinating Center, 350 Longwood Avenue, Galleria—Lower Level, Boston, MA 02215.

Dr. Stone: Cardiovascular Division, Brigham and Women's Hospital, 75 Francis Street, Boston, MA 02115.

Dr. Silverman: University of Connecticut Health Center, 263 Farmington Avenue, Farmington, CT 06030-1305.

Dr. Gibson: Wext Roxbury Veterans Affairs Medical Center, 1400 VFW Parkway, West Roxbury, MA 02132.

Dr. Sacks: Department of Nutrition, Harvard School of Public Health, 665 Huntington Avenue, Boston, MA 02115.