- Email: [email protected]
Phenotypic Predictors of Response to Simvastatin Therapy Among African-Americans and Caucasians: The Cholesterol and Pharmacogenetics (CAP) Study
Phenotypic Predictors of Response to Simvastatin Therapy Among African-Americans and Caucasians: The Cholesterol and Pharmacogenetics (CAP) Study Joel A. Simon, MD, MPHa,b,*, Feng Lin, MSb, Stephen B. Hulley, MD, MPHb, Patricia J. Blanche, MSc, David Waters, MDd, Stephen Shiboski, PhDb, Jerome I. Rotter, MDe,f, Deborah A. Nickerson, PhDg, Huiying Yang, MD, PhDe,f, Mohammed Saad, MDh, and Ronald M. Krauss, MDc Although statins are effective lipid-lowering agents, the phenotypic and demographic predictors of such lowering have been less well examined. We enrolled 944 AfricanAmerican and white men and women who completed an open-label, 6-week pharmacogenetics trial of 40 mg of simvastatin. The phenotypic and demographic variables were examined as predictors of the change in lipids and lipoproteins using linear regression analysis. On average, treatment with simvastatin lowered low-density lipoprotein (LDL) cholesterol by 54 mg/dl and increased high-density lipoprotein (HDL) cholesterol by 2 mg/dl. Compared with African-Americans, whites had a 3-mg/dl greater LDL reduction and a 1-mg/dl higher HDL elevation, independent of other variables, including baseline lipoprotein levels (p <0.01). Multivariate analyses revealed moderate subgroup differences, with older participants having a larger decrease in LDL cholesterol and apolipoprotein B levels compared with younger participants (p <0.001), women having larger increases in HDL than men (p <0.01), nonsmokers having larger decreases in LDL and triglyceride levels compared with smokers (p <0.05), those with hypertension having smaller decreases in apolipoprotein B than those without hypertension (p <0.05), and those with a larger waist circumference having a diminished lowering of triglycerides in response to treatment with simvastatin (p <0.01). In conclusion, treatment with simvastatin produced favorable lipid and lipoprotein changes among all participants. The magnitude of the lipid and lipoprotein responses, however, differed among participants according to a number of phenotypic and demographic characteristics. © 2006 Elsevier Inc. All rights reserved. (Am J Cardiol 2006;97:843– 850)
Many of the clinical trials of statin therapy, particularly those conducted in Europe, enrolled few, if any, nonwhite participants. Because substantial variability exists in the drug response among patients and because African-Americans have, in general, been under-represented in clinical trials of
a General Internal Medicine Section, Medical Service, Veterans Affairs Medical Center, San Francisco, California; bDepartment of Epidemiology and Biostatistics, University of California, San Francisco, School of Medicine, San Francisco, California; cChildren’s Hospital Oakland Research Institute, Oakland, California; dSan Francisco General Hospital, Department of Medicine, University of California, San Francisco, School of Medicine, San Francisco, California; eDepartment of Medicine, University of California, Los Angeles, School of Medicine, Los Angeles, California; f Cedars-Sinai Medical Center, Los Angeles, California; gDepartment of Genome Sciences, University of Washington, Seattle, Washington; and h Department of Preventive Medicine, State University of New York Health Sciences Center, Stony Brook, New York. Manuscript received July 18, 2005; revised manuscript received and accepted September 23, 2005. CAP was supported by Grant U01-HL69757 from the National Institutes of Health, Bethesda, Maryland. Additional support was provided by Pfizer, Incorporated, New York, New York. * Corresponding author: Tel: 415-750-2093; fax: 415-379-5573. E-mail address: [email protected] (J.A. Simon).
0002-9149/06/$ – see front matter © 2006 Elsevier Inc. All rights reserved. doi:10.1016/j.amjcard.2005.09.134
statins, we undertook the Cholesterol and Pharmacogenetics (CAP) Study (1 of 2 pharmacogenetics studies conducted by the Pharmacogenetics and Risk of Cardiovascular disease [PARC] Study Group) to examine the genetic factors affecting the response to therapy among African-Americans and whites. For this report, we examined the nongenetic, phenotypic, and demographic predictors of the lipid-lowering response to statins.
Methods Study population: Between March 2002 and October 2004, we enrolled 1,007 African-American and white men and women, aged ⱖ30 years, with a baseline total serum cholesterol level of 160 to 400 mg/dl in the PARC Study. Using clinic-based flyers, public service advertisements, and community outreach, we recruited and enrolled participants at 2 clinical centers located at San Francisco General Hospital (San Francisco, California) and from the University of California, Los Angeles, School of Medicine (Los Angeles, California). Baseline data were collected during the screening and enrollment visits and included demowww.AJConline.org
844
The American Journal of Cardiology (www.AJConline.org)
graphic characteristics, medical history, risk factors for coronary heart disease, physical examination findings, and laboratory data. Participants were followed for a total of 6 weeks on simvastatin therapy (40 mg at bedtime) and were seen at clinic visits conducted at 2-week intervals. Participants were considered African-American if ⱖ3 of their grandparents were reported to be African-American and were considered white if ⱖ3 of their grandparents were reported to be white. Potential participants from other racial/ ethnic groups were not enrolled. This study was 1 of 2 pharmacogenetics studies, CAP and the Hypertension and Ramipril Pharmacogenetics Study (HARP), conducted by the PARC Study Group. The genetic predictors of response to simvastatin will be reported separately. We conducted a 2-week run in using a placebo and excluded any potential participant who was unable to maintain ⱖ90% compliance with the study medication. Other exclusionary criteria included (1) the use of lipid-lowering medication (or over-the-counter products containing sterol or stanol esters or fish oil fatty acids); (2) a recent or planned change in dietary intake or a weight change of ⱖ4.5 kg; (3) the use of corticosteroids, immunosuppressive drugs, or drugs affecting the CYP3A4 system; (4) known liver disease or elevated transaminase levels more than twice the upper limit of normal; (5) elevated creatine phosphokinase levels ⬎10 times the upper limits of normal; (6) uncontrolled hypertriglyceridemia, blood pressure, or diabetes mellitus; (7) abnormal renal or thyroid function; (8) current alcohol or drug abuse; (9) recent major illness in the preceding 3 months; (10) current pregnancy; and (11) known intolerance to statins. The institutional review board at the clinical centers, laboratory centers, and coordinating center approved the PARC Study, and all participants provided informed consent before enrollment. A total of 63 participants dropped out of the study, leaving 944 participants (335 African-Americans and 609 whites) with complete data available for analysis. Of the 63 participants who dropped out, 11 discontinued because of side effects (including 2 participants who experienced either fatigue or muscle aches) and 19 discontinued because of meeting an exclusionary criterion (e.g., intercurrent use of certain medications such as corticosteroids and protease inhibitors). No CAP participant had marked creatine phosphokinase elevations or rhabdomyolysis. Measurements: Baseline data included self-reported information on the participants’ age, ethnicity, marital status, highest grade or year of school completed, alcohol consumption, smoking habits, level of physical activity, and family history of diabetes mellitus, hypercholesterolemia, coronary heart disease, and hypertension. We obtained a self-reported history of coronary disease, stroke, diabetes, and use of estrogen-containing medications. Participants with a history of gestational diabetes were not considered to have diabetes mellitus. Data were also obtained on all current prescription and nonprescription medications. Participants underwent a brief physical examination that
recorded blood pressure using an electronic manometer and a standardized protocol. Using a standard balance beam scale or an electronic scale, participants were weighed to the nearest 0.1 kg after removing their shoes and outdoor clothing. Height was measured to the nearest 0.1 cm using the height rod of a standard beam scale or, where available, a wall-mounted stadiometer. We calculated the body mass index as the weight in kilograms divided by the height in meters squared. Hip and waist circumference were also obtained using standardized protocols. After an overnight fast, baseline laboratory tests were performed by local clinic laboratories from specimens obtained at the screening visit. These tests included a complete blood count, serum electrolytes, standard lipid panel (i.e., total serum cholesterol, low-density lipoprotein [LDL] cholesterol, high-density lipoprotein [HDL] cholesterol, and triglycerides) plus apolipoprotein (apo)-B and apo-AI, liver panel (total serum bilirubin, transaminases, alkaline phosphatase, albumin), blood urea nitrogen, serum creatinine, serum thyroid-stimulating hormone, serum glucose, serum creatine phosphokinase, and a urine pregnancy test for premenopausal women. At enrollment and weeks 4 and 6 of follow-up on study medication, we obtained the following analytes: plasma lipids and lipoproteins, fasting serum glucose, C-reactive protein, and white blood cells for transformation and genotyping. All lipid and lipoprotein measurements were conducted at the Children’s Hospital Oakland Research Institute (Oakland, California) in a laboratory certified by the Centers for Disease Control and Prevention (Atlanta, Georgia) for the measurement of serum lipids and lipoproteins. LDL cholesterol was calculated using the Friedewald equation.1 High compliance with the study protocol was strictly maintained; participants had to demonstrate ⱖ90% pill compliance determined by pill counts performed at each 2-week clinic visit. We monitored participants for adverse events, side effects, and on-study hospitalizations. Hepatotoxicity (serum transaminase levels more than twice the upper limit of normal) and muscle toxicity or rhabdomyolysis (serum creatine phosphokinase levels ⬎10 times the upper limit of normal) resulted in discontinuance of the study medication and immediate medical follow-up. Statistical analysis: We compared the baseline characteristics of the African-American and white study participants using 2-tailed t tests for continuous variables and chi-square tests for categorical variables. Changes in serum lipid and lipoprotein levels from baseline to week 6 were symmetrically distributed in the exploratory analyses and were subsequently examined using linear regression models. The baseline lipid levels were represented as the mean of the levels obtained at the screening and enrollment visits. Similarly, the on-medication levels were represented as the mean of levels obtained at 4 and 6 weeks of treatment. The demographic and phenotypic predictors of observed changes in LDL cholesterol, HDL cholesterol, triglycerides,
Preventive Cardiology/Predictors of Simvastatin Response
845
Table 1 Baseline characteristics of participants in Cholesterol and Pharmacogenetics (CAP) study (n ⫽ 944) Characteristic Age (yrs) Men Married Level of education (yrs) Body mass index (kg/m2) Waist circumference (cm) Hypertension* Current cigarette smoker Past cigarette smoker Exercise level More active than peers Less active than peers About the same Alcohol intake None ⱕ2/wk ⱖ3/wk Previous coronary heart disease Previous stroke Previous diabetes mellitus Postmenopausal estrogen use† Total cholesterol level (mg/dl) LDL cholesterol level (mg/dl) HDL cholesterol level (mg/dl) Triglyceride level (mg/dl) Apo A-I (mg/dl) Apo B (mg/dl)
African-American (n ⫽ 335)
White (n ⫽ 609)
p Value
54.4 ⫾ 12.2 48% 29% 13.8 ⫾ 2.5 30.1 ⫾ 6.5 98.7 ⫾ 15.1 81% 31% 26%
54.4 ⫾ 12.6 53% 41% 15.6 ⫾ 2.6 27.7 ⫾ 5.5 94.5 ⫾ 14.7 70% 14% 33%
0.98 0.17 ⬍0.001 ⬍0.001 ⬍0.001 ⬍0.001 ⬍0.001 ⬍0.001 0.02 0.04
42% 21% 37%
43% 15% 43%
45% 42% 13% 2%
27% 45% 27% 0.5%
⬍0.01
3% 8% 10%
0.7% 2% 19%
⬍0.01 ⬍0.001 0.02
207 ⫾ 34
212 ⫾ 35
0.02
131 ⫾ 34
133 ⫾ 32
0.26
55 ⫾ 16
54 ⫾ 16
0.18
107 ⫾ 56 129 ⫾ 27 92 ⫾ 19
129 ⫾ 67 127 ⫾ 26 96 ⫾ 23
⬍0.001 0.15 ⬍0.01
Figure 1. Distribution of change in LDL cholesterol levels on 40 mg/day simvastatin.
⬍0.001
Data are presented as numbers (percentages) or means ⫾ SDs. * Defined as mean systolic blood pressure ⱖ140 mm Hg or mean diastolic blood pressure ⱖ90 mm Hg. † Based on responses from 459 women, including women using combination estrogen plus progestin products.
apo-AI, and apo-B levels were initially entered singly into the regression models. The variables associated with the outcome measures at p ⬍0.20 were considered for inclusion in the multivariate models. These models were also adjusted for baseline levels of LDL, HDL, triglycerides, apo-AI, and apo-B. Additional models were fitted to screen for the presence of 2-way interactions between selected predictors. We considered 2-tailed p values of ⬍0.05 to be statistically significant. All analyses were performed using Statistical Analysis Systems, version 8.02 (SAS Institute, Cary, North Carolina).2
Results The baseline characteristics of the CAP participants are presented in Table 1. On average, we enrolled middle-age African-American and white men and women who were
Figure 2. Distribution of change in HDL cholesterol levels on 40 mg/day simvastatin.
overweight or obese and had high blood pressure (Table 1). The African-American and white participants differed in a number of demographic and medical variables. Compared with the African-Americans, the whites were more likely to be married, more educated, and leaner, and were less likely to have high blood pressure or be current smokers (all p ⬍0.001). With the exception of serum triglyceride and apo-B levels, which were higher among whites than African-Americans, the lipid and lipoprotein levels were similar between the 2 racial/ethnic groups. The distributions of LDL and HDL responses to simvastatin are presented in Figures 1 and 2. Most participants experienced a 30% to 60% decrease in LDL levels. Approximately 66% of participants experienced an increase in HDL levels, and 34% experienced a modest decrease. LDL and apo-B responses to simvastatin: On average, for all participants, treatment with 40 mg/day of simvastatin lowered LDL levels by 54 mg/dl (interquartile range 41 to 68) or 41%. In unadjusted models, 5 variables were significantly associated with the LDL response to simvastatin: race/ethnicity, increasing age, marital status, education level, and smoking status (Table 2). After multivariate adjustment, 3 variables remained significant independent predictors of the LDL response to simvastatin: race/ethnicity, age, and smoking status. African-Americans had a 3-mg/dl smaller decrease in LDL
846
The American Journal of Cardiology (www.AJConline.org)
Table 2 Simvastatin-associated changes in serum low-density lipoprotein (LDL) cholesterol levels (milligrams per deciliter) in Cholesterol and Pharmacogenetics (CAP) study from baseline to week 6* Variable Race White African-American Age (10 yrs) Sex Women Men Marital status Nonmarried Married Education (yrs) Body index mass (5 kg/m2) Waist circumference (5 cm) Hypertension No Yes Current cigarette smoker No Yes Exercise level More active than peers Less active than peers About the same Alcohol intake None ⱕ2/wk ⱖ3/wk Previous coronary heart disease No Yes Previous stroke No Yes Previous diabetes mellitus No Yes Postmenopausal estrogen use§ No Yes
Baseline
Change From Baseline
Difference or Change From Baseline (95% CI)†
133 ⫾ 32 131 ⫾ 34 132 ⫾ 33
⫺56 ⫾ 21 ⫺51 ⫾ 24
133 ⫾ 33 131 ⫾ 32
⫺55 ⫾ 23 ⫺53 ⫾ 22
130 ⫾ 34 136 ⫾ 30 132 ⫾ 32 132 ⫾ 32 132 ⫾ 32
⫺52 ⫾ 23 ⫺57 ⫾ 21
127 ⫾ 28 134 ⫾ 34
⫺52 ⫾ 20 ⫺55 ⫾ 23
⫺2 (⫺5 to 1)
133 ⫾ 31 128 ⫾ 37
⫺56 ⫾ 21 ⫺47 ⫾ 25
9 (5 to 12)储
131 ⫾ 31 135 ⫾ 35 132 ⫾ 33
⫺53 ⫾ 22 ⫺55 ⫾ 22 ⫺55 ⫾ 23
⫺2 (⫺6 to 2) ⫺2 (⫺5 to 2)
134 ⫾ 34 129 ⫾ 29 135 ⫾ 36
⫺54 ⫾ 23 ⫺53 ⫾ 22 ⫺56 ⫾ 22
1 (⫺2 to 5) ⫺1 (⫺5 to 3)
132 ⫾ 32 127 ⫾ 34
⫺54 ⫾ 22 ⫺47 ⫾ 28
7 (⫺7 to 20)
132 ⫾ 33 131 ⫾ 29
⫺54 ⫾ 22 ⫺56 ⫾ 17
⫺2 (⫺14 to 9)
132 ⫾ 32 130 ⫾ 32
⫺54 ⫾ 22 ⫺57 ⫾ 23
⫺3 (⫺10 to 4)
140 ⫾ 31 130 ⫾ 25
⫺60 ⫾ 22 ⫺56 ⫾ 21
4 (⫺2 to 11)
Adjusted Difference or Change From Baseline (95% CI)‡
5 (2 to 8)储 ⫺4 (⫺5 to ⫺3)储
3 (1 to 5)¶ ⫺2 (⫺3 to ⫺1)储
3 (⫺0.2 to 6)
1 (⫺1 to 4)
⫺5 (⫺8 to ⫺2)¶ ⫺0.5 (⫺1 to 0)** ⫺0.5 (⫺2 to 1) ⫺0.2 (⫺1 to 0)
⫺1 (⫺3 to 2) ⫺0.2 (⫺1 to 0.2)
1 (⫺1 to 4)
3 (0.4 to 6)**
* In Tables 2 to 5, values in parentheses denote clinically relevant increment in age, years of education, body mass index, and waist circumference, respectively, associated with change in lipid or lipoprotein level. For example, for every 10-year increase in age, there was an associated 2-mg/dl greater decrease in LDL cholesterol level. † Absolute differences or change from baseline values for LDL are unadjusted and denote change per unit increment for continuous variables and difference in change from baseline to end of follow-up for categorical variables. ‡ Adjusted for the following variables associated with change in LDL cholesterol at p ⱕ0.20: race/ethnicity, age, gender, marital status, level of education, hypertension, smoking, and baseline LDL level. § Among 303 women ⱖ50 years old. 储 p ⬍0.001; ¶ p ⬍0.01; ** p ⬍0.05. CI ⫽ confidence interval.
compared with white participants (p ⬍0.01). Older participants also responded in a more robust fashion than younger participants. For each 10-year increment in age, there was an associated 2-mg/dl larger decrease in LDL (p ⬍0.001). Within the age range of participants enrolled in CAP (aged 30 to 88 years), differences in age accounted for approximately 12 mg/dl of the predicted difference in response to simvastatin. On average, smokers had a 4-mg/dl smaller decrease in LDL
compared with nonsmokers (p ⬍0.05). We examined additional models to test for possible interactions between age ⫻ gender and race/ethnicity ⫻ smoking on the response to LDL lowering. We found no evidence for such interactions in the multivariate models (p ⬎0.16). Apo-B, the protein moiety found in LDL and VLDL, decreased a mean of 28 mg/dl with simvastatin treatment (Table 3). As was observed for LDL, race/ethnicity, in-
Preventive Cardiology/Predictors of Simvastatin Response
847
Table 3 Simvastatin-associated change in serum apo-B levels (milligrams per deciliter) in Cholesterol and Pharmacogenetics (CAP) study from baseline to week 6 Variable* Race White African-American Age (10 yrs) Sex Women Men Marital status Non-married Married Education (yrs) Body index mass (5 kg/m2) Waist circumference (5 cm) Hypertension No Yes Current cigarette smoker No Yes Exercise level More active than peers Less active than peers About the same Alcohol intake None ⱕ2/wk ⱖ3/wk Previous coronary heart disease No Yes Previous stroke No Yes Previous diabetes mellitus No Yes Postmenopausal estrogen use¶ No Yes
Baseline
Change From Baseline
Difference or Change From Baseline (95% CI)*
96 ⫾ 23 92 ⫾ 19 94 ⫾ 22
⫺30 ⫾ 16 ⫺26 ⫾ 15
93 ⫾ 21 96 ⫾ 23
⫺29 ⫾ 15 ⫺28 ⫾ 16
93 ⫾ 23 96 ⫾ 20 94 ⫾ 22 94 ⫾ 22 94 ⫾ 22
⫺27 ⫾ 15 ⫺30 ⫾ 16
89 ⫾ 21 96 ⫾ 22
⫺27 ⫾ 15 ⫺29 ⫾ 16
⫺2 (⫺4 to 1)
95 ⫾ 21 92 ⫾ 24
⫺30 ⫾ 15 ⫺23 ⫾ 17
6 (4 to 9)§
93 ⫾ 20 96 ⫾ 23 96 ⫾ 22
⫺28 ⫾ 15 ⫺29 ⫾ 16 ⫺29 ⫾ 16
⫺2 (⫺4 to 1) ⫺1 (⫺4 to 1)
96 ⫾ 22 93 ⫾ 22 95 ⫾ 22
⫺28 ⫾ 16 ⫺29 ⫾ 15 ⫺28 ⫾ 16
⫺1 (⫺3 to 1) 0.2 (⫺3 to 3)
94 ⫾ 22 89 ⫾ 20
⫺28 ⫾ 16 ⫺24 ⫾ 11
5 (⫺5 to 14)
94 ⫾ 22 98 ⫾ 25
⫺28 ⫾ 16 ⫺33 ⫾ 13
⫺5 (⫺12 to 3)
94 ⫾ 22 95 ⫾ 19
⫺28 ⫾ 15 ⫺30 ⫾ 16
⫺2 (⫺6 to 3)
96 ⫾ 20 99 ⫾ 18
⫺31 ⫾ 13 ⫺32 ⫾ 15
⫺0.6 (⫺5 to 3)
3 (1 to 5)‡ ⫺2 (⫺3 to ⫺1)§
Adjusted Difference or Change From Baseline (95% CI)†
.01 (⫺2 to 2) ⫺1 (⫺2 to ⫺0.6)§
0.7 (⫺1 to 3)
⫺3 (⫺5 to ⫺1)‡ ⫺0.3 (⫺0.7 to 0.0) ⫺0.2 (⫺1 to 0.7) ⫺0.2 (⫺0.5 to 0.1)
⫺0.4 (⫺2 to 1) ⫺0.3 (⫺0.6 to 0.1)
2 (0.2 to 4)储
4 (2 to 6)§
* Absolute differences or change from baseline values for serum apo-B levels are unadjusted and denote change per unit increment for continuous variables and difference in the change from baseline to end of follow-up for categorical variables. † Adjusted for the following variables associated with change in serum apo-B level at p ⱕ0.20: race/ethnicity, age, marital status, level of education, history of hypertension, smoking, and baseline apo-B level. ‡ p ⬍0.01; § p ⬍0.001; 储 p ⬍0.05. ¶ Among 303 women ⱖ50 years old. Abbreviation as in Table 2.
creasing age, marital status, and smoking status were associated in the unadjusted models with a differential response to treatment with simvastatin (all p ⬍0.01). After adjustment for baseline apo-B levels and other potential confounders, only increasing age, hypertension, and smoking were independently associated with a change in apo-B levels (all p ⬍0.05). For each 10-year increment in age, there was an associated 1-mg/dl larger decrease in apo-B levels (p ⬍0.001). Participants with hypertension had a 2-mg/dl smaller decrease in apo-B levels compared with normotensive participants, and cur-
rent smokers had a 4-mg/dl smaller decrease compared with nonsmokers. HDL and apo-AI responses to simvastatin: On average, for all CAP participants, simvastatin treatment increased HDL cholesterol by 2 mg/dl (interquartile range ⫺1 to 5 mg/dl) or approximately 4%. When examining the effect of simvastatin on HDL levels in the univariate models (Table 4), we observed differential effects involving 2 variables. There was a small, but statistically significant, differential response by race/ethnicity such that, on average,
848
The American Journal of Cardiology (www.AJConline.org)
Table 4 Simvastatin-associated change in serum high-density lipoprotein (HDL) cholesterol levels (milligrams per deciliter) in Cholesterol and Pharmacogenetics (CAP) study from baseline to week 6 Variable Race White African-American Age (10 yrs) Sex Women Men Marital status Nonmarried Married Education (yrs) Body index mass (5 kg/m2) Waist circumference (5 cm) Hypertension No Yes Current cigarette smoker No Yes Exercise level More active than peers Less active than peers About the same Alcohol intake None ⱕ2/wk ⱖ3/wk Previous coronary heart disease No Yes Previous stroke No Yes Previous diabetes mellitus No Yes Postmenopausal estrogen use储 No Yes
Baseline
Change From Baseline
Difference or Change From Baseline (95% CI)*
54 ⫾ 16 55 ⫾ 16 54 ⫾ 16
2⫾6 1⫾6
61 ⫾ 17 48 ⫾ 13
2⫾6 1⫾5
55 ⫾ 16 53 ⫾ 16 54 ⫾ 16 54 ⫾ 16 54 ⫾ 16
2⫾6 2⫾5
58 ⫾ 18 53 ⫾ 15
2⫾5 2⫾6
⫺0.4 (⫺1 to 0.4)
55 ⫾ 16 52 ⫾ 17
2⫾5 1⫾7
⫺0.7 (⫺2 to 0.2)
57 ⫾ 17 49 ⫾ 13 53 ⫾ 16
1⫾6 2⫾6 2⫾6
1.0 (⫺0 to 2) 0.8 (⫺0 to 2)
51 ⫾ 14 54 ⫾ 16 58 ⫾ 18
2⫾6 2⫾6 2⫾6
0.2 (⫺1 to 1) 0.6 (⫺0.4 to 2)
54 ⫾ 16 59 ⫾ 30
2⫾6 4 ⫾ 11
2 (⫺1 to 6)
54 ⫾ 16 55 ⫾ 25
2⫾6 1⫾6
⫺0.4 (⫺3 to 3)
54 ⫾ 16 49 ⫾ 12
2⫾6 2⫾4
⫺0.2 (⫺2 to 2)
61 ⫾ 17 65 ⫾ 16
2⫾7 4⫾6
Adjusted Difference or Change From Baseline (95% CI)†
⫺1 (⫺2 to ⫺0.5)‡ 0.2 (⫺0 to 0.5)
⫺1 (⫺2 to ⫺0.5)‡ 0.2 (⫺0 to 0.5)
⫺1 (⫺2 to ⫺0.1)§
⫺1 (⫺2 to ⫺0.5)‡
⫺0.5 (⫺1 to 0.3) ⫺0.0 (⫺0.1 to 0.1) ⫺0.2 (⫺0.5 to 0.1) ⫺0.1 (⫺0.2 to 0.1)
⫺0.5 (⫺1 to 0.5)
0.8 (⫺0.3 to 2) 0.5 (⫺0.3 to 1)
3 (⫺0.4 to 6)
2 (⫺0.1 to 4)
* Absolute differences or change from baseline values for HDL are unadjusted and denote change per unit increment for continuous variables and difference in change from baseline to end of follow-up for categorical variables. † Adjusted for the following variables associated with change in HDL cholesterol at p ⱕ0.20: race/ethnicity, age, gender, marital status, level of physical activity, history of coronary disease, smoking, and baseline HDL level. ‡ p ⬍0.01; § p ⬍0.05. 储 Among 303 women ⱖ50 years old. Abbreviation as in Table 2.
whites had a 1-mg/dl larger increase on treatment compared with African-Americans (p ⬍0.01). Women had, on average, a 2-mg/dl increase in HDL compared with a 1-mg/dl increase in men (p ⬍0.05). These small, but statistically significant, race/ethnicity and gender-based differences in HDL response continued to be observed after multivariate adjustment. Apo-AI, a protein moiety found in HDL and chylomicrons, increased a mean of 1 mg/dl (1.4%) with simvastatin treatment. In the univariate models, only the level of physical activity appeared associated with the apo-AI response
to treatment with simvastatin. After adjustment, however, no variable was independently associated with a differential response to changes in apo-AI levels. Triglyceride responses to simvastatin: On average, participants treated with simvastatin for 6 weeks had a 22-mg/dl decrease in serum triglyceride levels (Table 5). In the univariate models, 2 variables were associated with differential changes in serum triglyceride levels on simvastatin: race/ethnicity and waist circumference (p ⬍0.05; Table 5). After ad-
Preventive Cardiology/Predictors of Simvastatin Response
849
Table 5 Simvastatin-associated change in serum triglyceride levels (milligrams per deciliter) in Cholesterol and Pharmacogenetics (CAP) study from baseline to week 6 Variable Race White African-American Age (10 yrs) Sex Women Men Marital status Non-married Married Education (yrs) Body index mass (5 kg/m2) Waist circumference (5 cm) Hypertension No Yes Current cigarette smoker No Yes Exercise level More active than peers Less active than peers About the same Alcohol intake None ⱕ2/wk ⱖ3/wk Previous coronary heart disease No Yes Previous stroke No Yes Previous diabetes mellitus No Yes Postmenopausal estrogen use储 No Yes
Baseline
Change From Baseline
Difference or Change From Baseline (95% CI)*
129 ⫾ 67 107 ⫾ 56 121 ⫾ 64
⫺24 ⫾ 45 ⫺18 ⫾ 37
112 ⫾ 55 130 ⫾ 71
⫺21 ⫾ 29 ⫺22 ⫾ 52
121 ⫾ 67 122 ⫾ 60 121 ⫾ 64 121 ⫾ 64 121 ⫾ 64
⫺22 ⫾ 40 ⫺21 ⫾ 47
107 ⫾ 58 126 ⫾ 66
⫺21 ⫾ 35 ⫺22 ⫾ 45
120 ⫾ 65 124 ⫾ 62
⫺23 ⫾ 43 ⫺17 ⫾ 39
112 ⫾ 62 126 ⫾ 64 129 ⫾ 66
⫺20 ⫾ 40 ⫺25 ⫾ 37 ⫺22 ⫾ 48
⫺5 (⫺13 to 3) ⫺2 (⫺8 to 4)
123 ⫾ 62 118 ⫾ 66 124 ⫾ 64
⫺21 ⫾ 41 ⫺23 ⫾ 38 ⫺20 ⫾ 53
⫺2 (⫺8 to 5) 2 (⫺6 to 9)
121 ⫾ 64 113 ⫾ 58
⫺22 ⫾ 43 ⫺13 ⫾ 43
121 ⫾ 64 137 ⫾ 78
⫺21 ⫾ 43 ⫺34 ⫾ 41
⫺13 (⫺34 to 9)
121 ⫾ 64 128 ⫾ 67
⫺21 ⫾ 43 ⫺31 ⫾ 40
⫺10 (⫺23 to 4)
112 ⫾ 54 138 ⫾ 58
⫺22 ⫾ 31 ⫺29 ⫾ 31
⫺7 (⫺16 to 2)
6 (1 to 12)‡ ⫺2 (⫺4 to ⫺0.4)‡
Adjusted Difference or Change From Baseline (95% CI)†
⫺2 (⫺7 to 4) ⫺1 (⫺3 to 0.7)
⫺2 (⫺7 to 4)
1 (⫺5 to 7) ⫺0.4 (⫺1 to 0.7) ⫺1 (⫺3 to 1) ⫺1 (⫺2 to ⫺0.1)‡
1 (0.3 to 2)§
0.6 (⫺7 to 6)
6 (⫺0.4 to 13)
8 (1 to 14)‡
⫺5 (⫺13 to 2) 0.6 (⫺5 to 6)
9 (⫺17 to 34)
⫺8 (⫺20 to 4)
* Absolute differences or change from baseline values for serum triglyceride levels are unadjusted and denote change per unit increment for continuous variables and difference in change from baseline to end of follow-up for categorical variables. † Adjusted for the following variables associated with change in serum triglyceride level at p ⱕ0.20: race/ethnicity, age, waist circumference, level of physical activity, smoking, diabetes, and baseline serum triglyceride level. ‡ p ⬍0.05; § p ⬍0.01. 储 Among 303 women ⱖ50 years old. Abbreviation as in Table 2.
justment, however, waist circumference and smoking were the only factors independently associated with a change in triglyceride levels. Leaner participants had a more robust triglyceride lowering response with simvastatin, such that for each 5-cm decrease in waist girth, a 1-mg/dl larger decrease in serum triglyceride levels was observed (p ⬍0.01). Smoking blunted the triglyceride-lowering effects of simvastatin. Smokers had an adjusted 8-mg/dl higher serum triglyceride level compared with nonsmokers (p ⬍0.05), independent of baseline triglyceride levels.
Discussion Most studies of the effects of statins have reported that the beneficial effect of statins on clinical outcomes, such as coronary heart disease events, occurs across all subgroups. Relatively few studies have examined whether statins affect lipid and lipoprotein levels differentially when stratified by demographic and phenotypic characteristics. The Expanded Clinical Evaluation of Lovastatin (EXCEL) study reported on the differential effects of lovastatin by patient characteristics.3 In the
850
The American Journal of Cardiology (www.AJConline.org)
EXCEL study, white participants had 3% to 6% greater decreases in LDL relative to African-Americans, similar to what we observed in CAP. Shear et al3 also found that older participants had a greater response than younger participants, especially older women. Also in agreement with our findings, Hunt et al4 reported that treatment with pravastatin resulted in larger decreases of lipid levels among older patients compared with younger patients, and Kannel et al5 likewise found a greater absolute lowering of LDL with lovastatin as age increased. In contrast, the effect of pravastatin on lipid lowering was similar among patients ⬍65 and ⱖ65 years in the Cholesterol and Recurrent Events (CARE) trial, although the older participants did have a nonsignificant 2% greater lowering of LDL.6 We did not find that women had a greater LDL-lowering response compared with men, in contrast to Ose et al7 who reported a markedly larger LDL-lowering response among women using cerivastatin. We did, however, observe a slightly larger statinrelated increase in HDL among women compared with men. Our HDL findings agree with those of some other investigators,3,5 but not others.8 We found that smokers had less lowering of LDL, apo-B, and triglyceride levels than did nonsmokers. Most studies have not examined whether smoking modifies the effect of statin treatment. Shear et al3 and Downs et al,8 however, did test for such a statin-related interaction with smoking, but found no such evidence. It is well known that smoking lowers HDL levels and has other potentially deleterious effects on the fatty acid composition of cholesterol esters and phospholipids.9 Clinical trials of statins have found that statin therapy in smokers reduces events, but to a lesser degree than in nonsmokers.10 The diminished effect on LDL and triglyceride lowering that we found in smokers may account in part for the findings from the clinical trials. We also observed that participants with hypertension had higher apo-B levels than did normotensive participants. To our knowledge, blood pressure-related effects on the apo-B response to statins have not been previously examined. However, others have tested for blood pressure ⫻ statin interactions on LDL and HDL responses and have been unable to detect such effect modifications.5,8 The only other variable that was associated with a differential response to statins was waist circumference. Participants with increased waist girth had a diminished triglyceride lowering response. Increased visceral adiposity, as reflected by increased waist girth, is associated with insulin resistance, higher triglyceride levels, and lower HDL levels, and it is possible that these metabolic changes affect the simvastatin response. We are unaware of other reports examining the effect of statins on the basis of the differences in waist circumference. A number of limitations of the CAP Study deserve comment. The overall effect of statins on LDL cholesterol levels in our study was slightly greater than that observed by others,7 which may be a consequence of regression to the mean (our participants were selected for having a baseline total serum cholesterol level of ⱖ160 mg/dl). The failure to observe other phenotypic subgroup differences may reflect inadequate statistical power. Conversely, we could not exclude the possibility
that some of the differences that were found resulted from chance, especially because many comparisons were tested. We chose not to control for multiple comparisons because not adjusting leads to fewer errors of interpretation when the data are not random numbers, but actual observations of nature.11 Furthermore, other statins and other doses of simvastatin might have had different effects. We did not collect data on dietary intake and, hence, could not be certain that some of our findings might have reflected, in part, the confounding effects of dietary factors. However, for differences by race/ethnicity, gender, and age, other investigators, albeit inconsistently, have reported similar findings. In conclusion, we found significant differential effects of simvastatin on LDL based on race/ethnicity, age, and smoking status. Factors responsible for differences in lipoprotein responsiveness to statin treatment may require consideration in optimizing treatment for specific patients.
Acknowledgment: Simvastatin was donated by Merck and Company, Incorporated, Horsham, Pennsylvania. 1. Friedewald WT, Levy RI, Fredrickson DS. Estimation of the concentration of low-density lipoprotein cholesterol in plasma, without use of the preparative ultracentrifuge. Clin Chem 1972;18:499 –502. 2. SAS Institute. Statistical analysis system, version 8.02. SAS Institute: Cary, North Carolina, 2001. 3. Shear CL, Franklin FA, Stinnett S, Hurley DP, Bradford RH, Chremos AN, Nash DT, Langendorfer A. Expanded Clinical Evaluation of Lovastatin (EXCEL) study results: effect of patient characteristics on lovastatin-induced changes in plasma concentrations of lipids and lipoproteins. Circulation 1992;85:1293–1303. 4. Hunt D, Young P, Simes J, Hague W, Mann S, Owensby D, Lane G, Tonkin A. Benefits of pravastatin on cardiovascular events and mortality in older patients with coronary disease are equal to or exceed those seen in younger patients: results from the LIPID trial. Ann Intern Med 2001;134:931–940. 5. Kannel WB, D’Agostino RB, Stepanians M, D’Agostino LC. Efficacy and tolerability of lovastatin in a six-month study: analysis by gender, age and hypertensive status. Am J Cardiol 1990;66(suppl):1B–10B. 6. Lewis SJ, Moye LA, Sacks FM, Johnstone DE, Timmis G, Mitchell J, Limacher M, Kell S, Glasser SP, Grant J, et al. Effect of pravastatin on cardiovascular events in older patients with myocardial infarction and cholesterol levels in the average range: results of the Cholesterol and Recurrent Events (CARE) trial. Ann Intern Med 1998;129:681– 689. 7. Ose L, Luurila O, Eriksson J, Olsson A, Lithell H, Widgren B. Efficacy and safety of cerivastatin, 0.2 mg and 0.4 mg, in patients with primary hypercholesterolaemia: a multinational, randomised, double-blind study. Curr Med Res Opin 1999;15:228 –240. 8. Downs JR, Clearfield M, Weis S, Whitney E, Shapiro DR, Beere PA, Langendorfer A, Stein EA, Kruyer W, Gotto AM Jr. Primary prevention of acute coronary events with lovastatin in men and women with average cholesterol levels: results of AFCAPS/TexCAPS. JAMA 1998; 279:1615–1622. 9. Simon JA, Fong J, Bernert JTJ, Browner WS. Relation of smoking and alcohol consumption to serum fatty acids. Am J Epidemiol 1996;144: 325–334. 10. Milionis HJ, Rizos E, Mikhailidis DP. Smoking diminishes the beneficial effect of statins: observations from the landmark trials. Angiology 2001;52:575–587. 11. Rothman KJ. No adjustments are needed for multiple comparisons. Epidemiology 1990;1:43– 46.
Many of the clinical trials of statin therapy, particularly those conducted in Europe, enrolled few, if any, nonwhite participants. Because substantial variability exists in the drug response among patients and because African-Americans have, in general, been under-represented in clinical trials of
a General Internal Medicine Section, Medical Service, Veterans Affairs Medical Center, San Francisco, California; bDepartment of Epidemiology and Biostatistics, University of California, San Francisco, School of Medicine, San Francisco, California; cChildren’s Hospital Oakland Research Institute, Oakland, California; dSan Francisco General Hospital, Department of Medicine, University of California, San Francisco, School of Medicine, San Francisco, California; eDepartment of Medicine, University of California, Los Angeles, School of Medicine, Los Angeles, California; f Cedars-Sinai Medical Center, Los Angeles, California; gDepartment of Genome Sciences, University of Washington, Seattle, Washington; and h Department of Preventive Medicine, State University of New York Health Sciences Center, Stony Brook, New York. Manuscript received July 18, 2005; revised manuscript received and accepted September 23, 2005. CAP was supported by Grant U01-HL69757 from the National Institutes of Health, Bethesda, Maryland. Additional support was provided by Pfizer, Incorporated, New York, New York. * Corresponding author: Tel: 415-750-2093; fax: 415-379-5573. E-mail address: [email protected] (J.A. Simon).
0002-9149/06/$ – see front matter © 2006 Elsevier Inc. All rights reserved. doi:10.1016/j.amjcard.2005.09.134
statins, we undertook the Cholesterol and Pharmacogenetics (CAP) Study (1 of 2 pharmacogenetics studies conducted by the Pharmacogenetics and Risk of Cardiovascular disease [PARC] Study Group) to examine the genetic factors affecting the response to therapy among African-Americans and whites. For this report, we examined the nongenetic, phenotypic, and demographic predictors of the lipid-lowering response to statins.
Methods Study population: Between March 2002 and October 2004, we enrolled 1,007 African-American and white men and women, aged ⱖ30 years, with a baseline total serum cholesterol level of 160 to 400 mg/dl in the PARC Study. Using clinic-based flyers, public service advertisements, and community outreach, we recruited and enrolled participants at 2 clinical centers located at San Francisco General Hospital (San Francisco, California) and from the University of California, Los Angeles, School of Medicine (Los Angeles, California). Baseline data were collected during the screening and enrollment visits and included demowww.AJConline.org
844
The American Journal of Cardiology (www.AJConline.org)
graphic characteristics, medical history, risk factors for coronary heart disease, physical examination findings, and laboratory data. Participants were followed for a total of 6 weeks on simvastatin therapy (40 mg at bedtime) and were seen at clinic visits conducted at 2-week intervals. Participants were considered African-American if ⱖ3 of their grandparents were reported to be African-American and were considered white if ⱖ3 of their grandparents were reported to be white. Potential participants from other racial/ ethnic groups were not enrolled. This study was 1 of 2 pharmacogenetics studies, CAP and the Hypertension and Ramipril Pharmacogenetics Study (HARP), conducted by the PARC Study Group. The genetic predictors of response to simvastatin will be reported separately. We conducted a 2-week run in using a placebo and excluded any potential participant who was unable to maintain ⱖ90% compliance with the study medication. Other exclusionary criteria included (1) the use of lipid-lowering medication (or over-the-counter products containing sterol or stanol esters or fish oil fatty acids); (2) a recent or planned change in dietary intake or a weight change of ⱖ4.5 kg; (3) the use of corticosteroids, immunosuppressive drugs, or drugs affecting the CYP3A4 system; (4) known liver disease or elevated transaminase levels more than twice the upper limit of normal; (5) elevated creatine phosphokinase levels ⬎10 times the upper limits of normal; (6) uncontrolled hypertriglyceridemia, blood pressure, or diabetes mellitus; (7) abnormal renal or thyroid function; (8) current alcohol or drug abuse; (9) recent major illness in the preceding 3 months; (10) current pregnancy; and (11) known intolerance to statins. The institutional review board at the clinical centers, laboratory centers, and coordinating center approved the PARC Study, and all participants provided informed consent before enrollment. A total of 63 participants dropped out of the study, leaving 944 participants (335 African-Americans and 609 whites) with complete data available for analysis. Of the 63 participants who dropped out, 11 discontinued because of side effects (including 2 participants who experienced either fatigue or muscle aches) and 19 discontinued because of meeting an exclusionary criterion (e.g., intercurrent use of certain medications such as corticosteroids and protease inhibitors). No CAP participant had marked creatine phosphokinase elevations or rhabdomyolysis. Measurements: Baseline data included self-reported information on the participants’ age, ethnicity, marital status, highest grade or year of school completed, alcohol consumption, smoking habits, level of physical activity, and family history of diabetes mellitus, hypercholesterolemia, coronary heart disease, and hypertension. We obtained a self-reported history of coronary disease, stroke, diabetes, and use of estrogen-containing medications. Participants with a history of gestational diabetes were not considered to have diabetes mellitus. Data were also obtained on all current prescription and nonprescription medications. Participants underwent a brief physical examination that
recorded blood pressure using an electronic manometer and a standardized protocol. Using a standard balance beam scale or an electronic scale, participants were weighed to the nearest 0.1 kg after removing their shoes and outdoor clothing. Height was measured to the nearest 0.1 cm using the height rod of a standard beam scale or, where available, a wall-mounted stadiometer. We calculated the body mass index as the weight in kilograms divided by the height in meters squared. Hip and waist circumference were also obtained using standardized protocols. After an overnight fast, baseline laboratory tests were performed by local clinic laboratories from specimens obtained at the screening visit. These tests included a complete blood count, serum electrolytes, standard lipid panel (i.e., total serum cholesterol, low-density lipoprotein [LDL] cholesterol, high-density lipoprotein [HDL] cholesterol, and triglycerides) plus apolipoprotein (apo)-B and apo-AI, liver panel (total serum bilirubin, transaminases, alkaline phosphatase, albumin), blood urea nitrogen, serum creatinine, serum thyroid-stimulating hormone, serum glucose, serum creatine phosphokinase, and a urine pregnancy test for premenopausal women. At enrollment and weeks 4 and 6 of follow-up on study medication, we obtained the following analytes: plasma lipids and lipoproteins, fasting serum glucose, C-reactive protein, and white blood cells for transformation and genotyping. All lipid and lipoprotein measurements were conducted at the Children’s Hospital Oakland Research Institute (Oakland, California) in a laboratory certified by the Centers for Disease Control and Prevention (Atlanta, Georgia) for the measurement of serum lipids and lipoproteins. LDL cholesterol was calculated using the Friedewald equation.1 High compliance with the study protocol was strictly maintained; participants had to demonstrate ⱖ90% pill compliance determined by pill counts performed at each 2-week clinic visit. We monitored participants for adverse events, side effects, and on-study hospitalizations. Hepatotoxicity (serum transaminase levels more than twice the upper limit of normal) and muscle toxicity or rhabdomyolysis (serum creatine phosphokinase levels ⬎10 times the upper limit of normal) resulted in discontinuance of the study medication and immediate medical follow-up. Statistical analysis: We compared the baseline characteristics of the African-American and white study participants using 2-tailed t tests for continuous variables and chi-square tests for categorical variables. Changes in serum lipid and lipoprotein levels from baseline to week 6 were symmetrically distributed in the exploratory analyses and were subsequently examined using linear regression models. The baseline lipid levels were represented as the mean of the levels obtained at the screening and enrollment visits. Similarly, the on-medication levels were represented as the mean of levels obtained at 4 and 6 weeks of treatment. The demographic and phenotypic predictors of observed changes in LDL cholesterol, HDL cholesterol, triglycerides,
Preventive Cardiology/Predictors of Simvastatin Response
845
Table 1 Baseline characteristics of participants in Cholesterol and Pharmacogenetics (CAP) study (n ⫽ 944) Characteristic Age (yrs) Men Married Level of education (yrs) Body mass index (kg/m2) Waist circumference (cm) Hypertension* Current cigarette smoker Past cigarette smoker Exercise level More active than peers Less active than peers About the same Alcohol intake None ⱕ2/wk ⱖ3/wk Previous coronary heart disease Previous stroke Previous diabetes mellitus Postmenopausal estrogen use† Total cholesterol level (mg/dl) LDL cholesterol level (mg/dl) HDL cholesterol level (mg/dl) Triglyceride level (mg/dl) Apo A-I (mg/dl) Apo B (mg/dl)
African-American (n ⫽ 335)
White (n ⫽ 609)
p Value
54.4 ⫾ 12.2 48% 29% 13.8 ⫾ 2.5 30.1 ⫾ 6.5 98.7 ⫾ 15.1 81% 31% 26%
54.4 ⫾ 12.6 53% 41% 15.6 ⫾ 2.6 27.7 ⫾ 5.5 94.5 ⫾ 14.7 70% 14% 33%
0.98 0.17 ⬍0.001 ⬍0.001 ⬍0.001 ⬍0.001 ⬍0.001 ⬍0.001 0.02 0.04
42% 21% 37%
43% 15% 43%
45% 42% 13% 2%
27% 45% 27% 0.5%
⬍0.01
3% 8% 10%
0.7% 2% 19%
⬍0.01 ⬍0.001 0.02
207 ⫾ 34
212 ⫾ 35
0.02
131 ⫾ 34
133 ⫾ 32
0.26
55 ⫾ 16
54 ⫾ 16
0.18
107 ⫾ 56 129 ⫾ 27 92 ⫾ 19
129 ⫾ 67 127 ⫾ 26 96 ⫾ 23
⬍0.001 0.15 ⬍0.01
Figure 1. Distribution of change in LDL cholesterol levels on 40 mg/day simvastatin.
⬍0.001
Data are presented as numbers (percentages) or means ⫾ SDs. * Defined as mean systolic blood pressure ⱖ140 mm Hg or mean diastolic blood pressure ⱖ90 mm Hg. † Based on responses from 459 women, including women using combination estrogen plus progestin products.
apo-AI, and apo-B levels were initially entered singly into the regression models. The variables associated with the outcome measures at p ⬍0.20 were considered for inclusion in the multivariate models. These models were also adjusted for baseline levels of LDL, HDL, triglycerides, apo-AI, and apo-B. Additional models were fitted to screen for the presence of 2-way interactions between selected predictors. We considered 2-tailed p values of ⬍0.05 to be statistically significant. All analyses were performed using Statistical Analysis Systems, version 8.02 (SAS Institute, Cary, North Carolina).2
Results The baseline characteristics of the CAP participants are presented in Table 1. On average, we enrolled middle-age African-American and white men and women who were
Figure 2. Distribution of change in HDL cholesterol levels on 40 mg/day simvastatin.
overweight or obese and had high blood pressure (Table 1). The African-American and white participants differed in a number of demographic and medical variables. Compared with the African-Americans, the whites were more likely to be married, more educated, and leaner, and were less likely to have high blood pressure or be current smokers (all p ⬍0.001). With the exception of serum triglyceride and apo-B levels, which were higher among whites than African-Americans, the lipid and lipoprotein levels were similar between the 2 racial/ethnic groups. The distributions of LDL and HDL responses to simvastatin are presented in Figures 1 and 2. Most participants experienced a 30% to 60% decrease in LDL levels. Approximately 66% of participants experienced an increase in HDL levels, and 34% experienced a modest decrease. LDL and apo-B responses to simvastatin: On average, for all participants, treatment with 40 mg/day of simvastatin lowered LDL levels by 54 mg/dl (interquartile range 41 to 68) or 41%. In unadjusted models, 5 variables were significantly associated with the LDL response to simvastatin: race/ethnicity, increasing age, marital status, education level, and smoking status (Table 2). After multivariate adjustment, 3 variables remained significant independent predictors of the LDL response to simvastatin: race/ethnicity, age, and smoking status. African-Americans had a 3-mg/dl smaller decrease in LDL
846
The American Journal of Cardiology (www.AJConline.org)
Table 2 Simvastatin-associated changes in serum low-density lipoprotein (LDL) cholesterol levels (milligrams per deciliter) in Cholesterol and Pharmacogenetics (CAP) study from baseline to week 6* Variable Race White African-American Age (10 yrs) Sex Women Men Marital status Nonmarried Married Education (yrs) Body index mass (5 kg/m2) Waist circumference (5 cm) Hypertension No Yes Current cigarette smoker No Yes Exercise level More active than peers Less active than peers About the same Alcohol intake None ⱕ2/wk ⱖ3/wk Previous coronary heart disease No Yes Previous stroke No Yes Previous diabetes mellitus No Yes Postmenopausal estrogen use§ No Yes
Baseline
Change From Baseline
Difference or Change From Baseline (95% CI)†
133 ⫾ 32 131 ⫾ 34 132 ⫾ 33
⫺56 ⫾ 21 ⫺51 ⫾ 24
133 ⫾ 33 131 ⫾ 32
⫺55 ⫾ 23 ⫺53 ⫾ 22
130 ⫾ 34 136 ⫾ 30 132 ⫾ 32 132 ⫾ 32 132 ⫾ 32
⫺52 ⫾ 23 ⫺57 ⫾ 21
127 ⫾ 28 134 ⫾ 34
⫺52 ⫾ 20 ⫺55 ⫾ 23
⫺2 (⫺5 to 1)
133 ⫾ 31 128 ⫾ 37
⫺56 ⫾ 21 ⫺47 ⫾ 25
9 (5 to 12)储
131 ⫾ 31 135 ⫾ 35 132 ⫾ 33
⫺53 ⫾ 22 ⫺55 ⫾ 22 ⫺55 ⫾ 23
⫺2 (⫺6 to 2) ⫺2 (⫺5 to 2)
134 ⫾ 34 129 ⫾ 29 135 ⫾ 36
⫺54 ⫾ 23 ⫺53 ⫾ 22 ⫺56 ⫾ 22
1 (⫺2 to 5) ⫺1 (⫺5 to 3)
132 ⫾ 32 127 ⫾ 34
⫺54 ⫾ 22 ⫺47 ⫾ 28
7 (⫺7 to 20)
132 ⫾ 33 131 ⫾ 29
⫺54 ⫾ 22 ⫺56 ⫾ 17
⫺2 (⫺14 to 9)
132 ⫾ 32 130 ⫾ 32
⫺54 ⫾ 22 ⫺57 ⫾ 23
⫺3 (⫺10 to 4)
140 ⫾ 31 130 ⫾ 25
⫺60 ⫾ 22 ⫺56 ⫾ 21
4 (⫺2 to 11)
Adjusted Difference or Change From Baseline (95% CI)‡
5 (2 to 8)储 ⫺4 (⫺5 to ⫺3)储
3 (1 to 5)¶ ⫺2 (⫺3 to ⫺1)储
3 (⫺0.2 to 6)
1 (⫺1 to 4)
⫺5 (⫺8 to ⫺2)¶ ⫺0.5 (⫺1 to 0)** ⫺0.5 (⫺2 to 1) ⫺0.2 (⫺1 to 0)
⫺1 (⫺3 to 2) ⫺0.2 (⫺1 to 0.2)
1 (⫺1 to 4)
3 (0.4 to 6)**
* In Tables 2 to 5, values in parentheses denote clinically relevant increment in age, years of education, body mass index, and waist circumference, respectively, associated with change in lipid or lipoprotein level. For example, for every 10-year increase in age, there was an associated 2-mg/dl greater decrease in LDL cholesterol level. † Absolute differences or change from baseline values for LDL are unadjusted and denote change per unit increment for continuous variables and difference in change from baseline to end of follow-up for categorical variables. ‡ Adjusted for the following variables associated with change in LDL cholesterol at p ⱕ0.20: race/ethnicity, age, gender, marital status, level of education, hypertension, smoking, and baseline LDL level. § Among 303 women ⱖ50 years old. 储 p ⬍0.001; ¶ p ⬍0.01; ** p ⬍0.05. CI ⫽ confidence interval.
compared with white participants (p ⬍0.01). Older participants also responded in a more robust fashion than younger participants. For each 10-year increment in age, there was an associated 2-mg/dl larger decrease in LDL (p ⬍0.001). Within the age range of participants enrolled in CAP (aged 30 to 88 years), differences in age accounted for approximately 12 mg/dl of the predicted difference in response to simvastatin. On average, smokers had a 4-mg/dl smaller decrease in LDL
compared with nonsmokers (p ⬍0.05). We examined additional models to test for possible interactions between age ⫻ gender and race/ethnicity ⫻ smoking on the response to LDL lowering. We found no evidence for such interactions in the multivariate models (p ⬎0.16). Apo-B, the protein moiety found in LDL and VLDL, decreased a mean of 28 mg/dl with simvastatin treatment (Table 3). As was observed for LDL, race/ethnicity, in-
Preventive Cardiology/Predictors of Simvastatin Response
847
Table 3 Simvastatin-associated change in serum apo-B levels (milligrams per deciliter) in Cholesterol and Pharmacogenetics (CAP) study from baseline to week 6 Variable* Race White African-American Age (10 yrs) Sex Women Men Marital status Non-married Married Education (yrs) Body index mass (5 kg/m2) Waist circumference (5 cm) Hypertension No Yes Current cigarette smoker No Yes Exercise level More active than peers Less active than peers About the same Alcohol intake None ⱕ2/wk ⱖ3/wk Previous coronary heart disease No Yes Previous stroke No Yes Previous diabetes mellitus No Yes Postmenopausal estrogen use¶ No Yes
Baseline
Change From Baseline
Difference or Change From Baseline (95% CI)*
96 ⫾ 23 92 ⫾ 19 94 ⫾ 22
⫺30 ⫾ 16 ⫺26 ⫾ 15
93 ⫾ 21 96 ⫾ 23
⫺29 ⫾ 15 ⫺28 ⫾ 16
93 ⫾ 23 96 ⫾ 20 94 ⫾ 22 94 ⫾ 22 94 ⫾ 22
⫺27 ⫾ 15 ⫺30 ⫾ 16
89 ⫾ 21 96 ⫾ 22
⫺27 ⫾ 15 ⫺29 ⫾ 16
⫺2 (⫺4 to 1)
95 ⫾ 21 92 ⫾ 24
⫺30 ⫾ 15 ⫺23 ⫾ 17
6 (4 to 9)§
93 ⫾ 20 96 ⫾ 23 96 ⫾ 22
⫺28 ⫾ 15 ⫺29 ⫾ 16 ⫺29 ⫾ 16
⫺2 (⫺4 to 1) ⫺1 (⫺4 to 1)
96 ⫾ 22 93 ⫾ 22 95 ⫾ 22
⫺28 ⫾ 16 ⫺29 ⫾ 15 ⫺28 ⫾ 16
⫺1 (⫺3 to 1) 0.2 (⫺3 to 3)
94 ⫾ 22 89 ⫾ 20
⫺28 ⫾ 16 ⫺24 ⫾ 11
5 (⫺5 to 14)
94 ⫾ 22 98 ⫾ 25
⫺28 ⫾ 16 ⫺33 ⫾ 13
⫺5 (⫺12 to 3)
94 ⫾ 22 95 ⫾ 19
⫺28 ⫾ 15 ⫺30 ⫾ 16
⫺2 (⫺6 to 3)
96 ⫾ 20 99 ⫾ 18
⫺31 ⫾ 13 ⫺32 ⫾ 15
⫺0.6 (⫺5 to 3)
3 (1 to 5)‡ ⫺2 (⫺3 to ⫺1)§
Adjusted Difference or Change From Baseline (95% CI)†
.01 (⫺2 to 2) ⫺1 (⫺2 to ⫺0.6)§
0.7 (⫺1 to 3)
⫺3 (⫺5 to ⫺1)‡ ⫺0.3 (⫺0.7 to 0.0) ⫺0.2 (⫺1 to 0.7) ⫺0.2 (⫺0.5 to 0.1)
⫺0.4 (⫺2 to 1) ⫺0.3 (⫺0.6 to 0.1)
2 (0.2 to 4)储
4 (2 to 6)§
* Absolute differences or change from baseline values for serum apo-B levels are unadjusted and denote change per unit increment for continuous variables and difference in the change from baseline to end of follow-up for categorical variables. † Adjusted for the following variables associated with change in serum apo-B level at p ⱕ0.20: race/ethnicity, age, marital status, level of education, history of hypertension, smoking, and baseline apo-B level. ‡ p ⬍0.01; § p ⬍0.001; 储 p ⬍0.05. ¶ Among 303 women ⱖ50 years old. Abbreviation as in Table 2.
creasing age, marital status, and smoking status were associated in the unadjusted models with a differential response to treatment with simvastatin (all p ⬍0.01). After adjustment for baseline apo-B levels and other potential confounders, only increasing age, hypertension, and smoking were independently associated with a change in apo-B levels (all p ⬍0.05). For each 10-year increment in age, there was an associated 1-mg/dl larger decrease in apo-B levels (p ⬍0.001). Participants with hypertension had a 2-mg/dl smaller decrease in apo-B levels compared with normotensive participants, and cur-
rent smokers had a 4-mg/dl smaller decrease compared with nonsmokers. HDL and apo-AI responses to simvastatin: On average, for all CAP participants, simvastatin treatment increased HDL cholesterol by 2 mg/dl (interquartile range ⫺1 to 5 mg/dl) or approximately 4%. When examining the effect of simvastatin on HDL levels in the univariate models (Table 4), we observed differential effects involving 2 variables. There was a small, but statistically significant, differential response by race/ethnicity such that, on average,
848
The American Journal of Cardiology (www.AJConline.org)
Table 4 Simvastatin-associated change in serum high-density lipoprotein (HDL) cholesterol levels (milligrams per deciliter) in Cholesterol and Pharmacogenetics (CAP) study from baseline to week 6 Variable Race White African-American Age (10 yrs) Sex Women Men Marital status Nonmarried Married Education (yrs) Body index mass (5 kg/m2) Waist circumference (5 cm) Hypertension No Yes Current cigarette smoker No Yes Exercise level More active than peers Less active than peers About the same Alcohol intake None ⱕ2/wk ⱖ3/wk Previous coronary heart disease No Yes Previous stroke No Yes Previous diabetes mellitus No Yes Postmenopausal estrogen use储 No Yes
Baseline
Change From Baseline
Difference or Change From Baseline (95% CI)*
54 ⫾ 16 55 ⫾ 16 54 ⫾ 16
2⫾6 1⫾6
61 ⫾ 17 48 ⫾ 13
2⫾6 1⫾5
55 ⫾ 16 53 ⫾ 16 54 ⫾ 16 54 ⫾ 16 54 ⫾ 16
2⫾6 2⫾5
58 ⫾ 18 53 ⫾ 15
2⫾5 2⫾6
⫺0.4 (⫺1 to 0.4)
55 ⫾ 16 52 ⫾ 17
2⫾5 1⫾7
⫺0.7 (⫺2 to 0.2)
57 ⫾ 17 49 ⫾ 13 53 ⫾ 16
1⫾6 2⫾6 2⫾6
1.0 (⫺0 to 2) 0.8 (⫺0 to 2)
51 ⫾ 14 54 ⫾ 16 58 ⫾ 18
2⫾6 2⫾6 2⫾6
0.2 (⫺1 to 1) 0.6 (⫺0.4 to 2)
54 ⫾ 16 59 ⫾ 30
2⫾6 4 ⫾ 11
2 (⫺1 to 6)
54 ⫾ 16 55 ⫾ 25
2⫾6 1⫾6
⫺0.4 (⫺3 to 3)
54 ⫾ 16 49 ⫾ 12
2⫾6 2⫾4
⫺0.2 (⫺2 to 2)
61 ⫾ 17 65 ⫾ 16
2⫾7 4⫾6
Adjusted Difference or Change From Baseline (95% CI)†
⫺1 (⫺2 to ⫺0.5)‡ 0.2 (⫺0 to 0.5)
⫺1 (⫺2 to ⫺0.5)‡ 0.2 (⫺0 to 0.5)
⫺1 (⫺2 to ⫺0.1)§
⫺1 (⫺2 to ⫺0.5)‡
⫺0.5 (⫺1 to 0.3) ⫺0.0 (⫺0.1 to 0.1) ⫺0.2 (⫺0.5 to 0.1) ⫺0.1 (⫺0.2 to 0.1)
⫺0.5 (⫺1 to 0.5)
0.8 (⫺0.3 to 2) 0.5 (⫺0.3 to 1)
3 (⫺0.4 to 6)
2 (⫺0.1 to 4)
* Absolute differences or change from baseline values for HDL are unadjusted and denote change per unit increment for continuous variables and difference in change from baseline to end of follow-up for categorical variables. † Adjusted for the following variables associated with change in HDL cholesterol at p ⱕ0.20: race/ethnicity, age, gender, marital status, level of physical activity, history of coronary disease, smoking, and baseline HDL level. ‡ p ⬍0.01; § p ⬍0.05. 储 Among 303 women ⱖ50 years old. Abbreviation as in Table 2.
whites had a 1-mg/dl larger increase on treatment compared with African-Americans (p ⬍0.01). Women had, on average, a 2-mg/dl increase in HDL compared with a 1-mg/dl increase in men (p ⬍0.05). These small, but statistically significant, race/ethnicity and gender-based differences in HDL response continued to be observed after multivariate adjustment. Apo-AI, a protein moiety found in HDL and chylomicrons, increased a mean of 1 mg/dl (1.4%) with simvastatin treatment. In the univariate models, only the level of physical activity appeared associated with the apo-AI response
to treatment with simvastatin. After adjustment, however, no variable was independently associated with a differential response to changes in apo-AI levels. Triglyceride responses to simvastatin: On average, participants treated with simvastatin for 6 weeks had a 22-mg/dl decrease in serum triglyceride levels (Table 5). In the univariate models, 2 variables were associated with differential changes in serum triglyceride levels on simvastatin: race/ethnicity and waist circumference (p ⬍0.05; Table 5). After ad-
Preventive Cardiology/Predictors of Simvastatin Response
849
Table 5 Simvastatin-associated change in serum triglyceride levels (milligrams per deciliter) in Cholesterol and Pharmacogenetics (CAP) study from baseline to week 6 Variable Race White African-American Age (10 yrs) Sex Women Men Marital status Non-married Married Education (yrs) Body index mass (5 kg/m2) Waist circumference (5 cm) Hypertension No Yes Current cigarette smoker No Yes Exercise level More active than peers Less active than peers About the same Alcohol intake None ⱕ2/wk ⱖ3/wk Previous coronary heart disease No Yes Previous stroke No Yes Previous diabetes mellitus No Yes Postmenopausal estrogen use储 No Yes
Baseline
Change From Baseline
Difference or Change From Baseline (95% CI)*
129 ⫾ 67 107 ⫾ 56 121 ⫾ 64
⫺24 ⫾ 45 ⫺18 ⫾ 37
112 ⫾ 55 130 ⫾ 71
⫺21 ⫾ 29 ⫺22 ⫾ 52
121 ⫾ 67 122 ⫾ 60 121 ⫾ 64 121 ⫾ 64 121 ⫾ 64
⫺22 ⫾ 40 ⫺21 ⫾ 47
107 ⫾ 58 126 ⫾ 66
⫺21 ⫾ 35 ⫺22 ⫾ 45
120 ⫾ 65 124 ⫾ 62
⫺23 ⫾ 43 ⫺17 ⫾ 39
112 ⫾ 62 126 ⫾ 64 129 ⫾ 66
⫺20 ⫾ 40 ⫺25 ⫾ 37 ⫺22 ⫾ 48
⫺5 (⫺13 to 3) ⫺2 (⫺8 to 4)
123 ⫾ 62 118 ⫾ 66 124 ⫾ 64
⫺21 ⫾ 41 ⫺23 ⫾ 38 ⫺20 ⫾ 53
⫺2 (⫺8 to 5) 2 (⫺6 to 9)
121 ⫾ 64 113 ⫾ 58
⫺22 ⫾ 43 ⫺13 ⫾ 43
121 ⫾ 64 137 ⫾ 78
⫺21 ⫾ 43 ⫺34 ⫾ 41
⫺13 (⫺34 to 9)
121 ⫾ 64 128 ⫾ 67
⫺21 ⫾ 43 ⫺31 ⫾ 40
⫺10 (⫺23 to 4)
112 ⫾ 54 138 ⫾ 58
⫺22 ⫾ 31 ⫺29 ⫾ 31
⫺7 (⫺16 to 2)
6 (1 to 12)‡ ⫺2 (⫺4 to ⫺0.4)‡
Adjusted Difference or Change From Baseline (95% CI)†
⫺2 (⫺7 to 4) ⫺1 (⫺3 to 0.7)
⫺2 (⫺7 to 4)
1 (⫺5 to 7) ⫺0.4 (⫺1 to 0.7) ⫺1 (⫺3 to 1) ⫺1 (⫺2 to ⫺0.1)‡
1 (0.3 to 2)§
0.6 (⫺7 to 6)
6 (⫺0.4 to 13)
8 (1 to 14)‡
⫺5 (⫺13 to 2) 0.6 (⫺5 to 6)
9 (⫺17 to 34)
⫺8 (⫺20 to 4)
* Absolute differences or change from baseline values for serum triglyceride levels are unadjusted and denote change per unit increment for continuous variables and difference in change from baseline to end of follow-up for categorical variables. † Adjusted for the following variables associated with change in serum triglyceride level at p ⱕ0.20: race/ethnicity, age, waist circumference, level of physical activity, smoking, diabetes, and baseline serum triglyceride level. ‡ p ⬍0.05; § p ⬍0.01. 储 Among 303 women ⱖ50 years old. Abbreviation as in Table 2.
justment, however, waist circumference and smoking were the only factors independently associated with a change in triglyceride levels. Leaner participants had a more robust triglyceride lowering response with simvastatin, such that for each 5-cm decrease in waist girth, a 1-mg/dl larger decrease in serum triglyceride levels was observed (p ⬍0.01). Smoking blunted the triglyceride-lowering effects of simvastatin. Smokers had an adjusted 8-mg/dl higher serum triglyceride level compared with nonsmokers (p ⬍0.05), independent of baseline triglyceride levels.
Discussion Most studies of the effects of statins have reported that the beneficial effect of statins on clinical outcomes, such as coronary heart disease events, occurs across all subgroups. Relatively few studies have examined whether statins affect lipid and lipoprotein levels differentially when stratified by demographic and phenotypic characteristics. The Expanded Clinical Evaluation of Lovastatin (EXCEL) study reported on the differential effects of lovastatin by patient characteristics.3 In the
850
The American Journal of Cardiology (www.AJConline.org)
EXCEL study, white participants had 3% to 6% greater decreases in LDL relative to African-Americans, similar to what we observed in CAP. Shear et al3 also found that older participants had a greater response than younger participants, especially older women. Also in agreement with our findings, Hunt et al4 reported that treatment with pravastatin resulted in larger decreases of lipid levels among older patients compared with younger patients, and Kannel et al5 likewise found a greater absolute lowering of LDL with lovastatin as age increased. In contrast, the effect of pravastatin on lipid lowering was similar among patients ⬍65 and ⱖ65 years in the Cholesterol and Recurrent Events (CARE) trial, although the older participants did have a nonsignificant 2% greater lowering of LDL.6 We did not find that women had a greater LDL-lowering response compared with men, in contrast to Ose et al7 who reported a markedly larger LDL-lowering response among women using cerivastatin. We did, however, observe a slightly larger statinrelated increase in HDL among women compared with men. Our HDL findings agree with those of some other investigators,3,5 but not others.8 We found that smokers had less lowering of LDL, apo-B, and triglyceride levels than did nonsmokers. Most studies have not examined whether smoking modifies the effect of statin treatment. Shear et al3 and Downs et al,8 however, did test for such a statin-related interaction with smoking, but found no such evidence. It is well known that smoking lowers HDL levels and has other potentially deleterious effects on the fatty acid composition of cholesterol esters and phospholipids.9 Clinical trials of statins have found that statin therapy in smokers reduces events, but to a lesser degree than in nonsmokers.10 The diminished effect on LDL and triglyceride lowering that we found in smokers may account in part for the findings from the clinical trials. We also observed that participants with hypertension had higher apo-B levels than did normotensive participants. To our knowledge, blood pressure-related effects on the apo-B response to statins have not been previously examined. However, others have tested for blood pressure ⫻ statin interactions on LDL and HDL responses and have been unable to detect such effect modifications.5,8 The only other variable that was associated with a differential response to statins was waist circumference. Participants with increased waist girth had a diminished triglyceride lowering response. Increased visceral adiposity, as reflected by increased waist girth, is associated with insulin resistance, higher triglyceride levels, and lower HDL levels, and it is possible that these metabolic changes affect the simvastatin response. We are unaware of other reports examining the effect of statins on the basis of the differences in waist circumference. A number of limitations of the CAP Study deserve comment. The overall effect of statins on LDL cholesterol levels in our study was slightly greater than that observed by others,7 which may be a consequence of regression to the mean (our participants were selected for having a baseline total serum cholesterol level of ⱖ160 mg/dl). The failure to observe other phenotypic subgroup differences may reflect inadequate statistical power. Conversely, we could not exclude the possibility
that some of the differences that were found resulted from chance, especially because many comparisons were tested. We chose not to control for multiple comparisons because not adjusting leads to fewer errors of interpretation when the data are not random numbers, but actual observations of nature.11 Furthermore, other statins and other doses of simvastatin might have had different effects. We did not collect data on dietary intake and, hence, could not be certain that some of our findings might have reflected, in part, the confounding effects of dietary factors. However, for differences by race/ethnicity, gender, and age, other investigators, albeit inconsistently, have reported similar findings. In conclusion, we found significant differential effects of simvastatin on LDL based on race/ethnicity, age, and smoking status. Factors responsible for differences in lipoprotein responsiveness to statin treatment may require consideration in optimizing treatment for specific patients.
Acknowledgment: Simvastatin was donated by Merck and Company, Incorporated, Horsham, Pennsylvania. 1. Friedewald WT, Levy RI, Fredrickson DS. Estimation of the concentration of low-density lipoprotein cholesterol in plasma, without use of the preparative ultracentrifuge. Clin Chem 1972;18:499 –502. 2. SAS Institute. Statistical analysis system, version 8.02. SAS Institute: Cary, North Carolina, 2001. 3. Shear CL, Franklin FA, Stinnett S, Hurley DP, Bradford RH, Chremos AN, Nash DT, Langendorfer A. Expanded Clinical Evaluation of Lovastatin (EXCEL) study results: effect of patient characteristics on lovastatin-induced changes in plasma concentrations of lipids and lipoproteins. Circulation 1992;85:1293–1303. 4. Hunt D, Young P, Simes J, Hague W, Mann S, Owensby D, Lane G, Tonkin A. Benefits of pravastatin on cardiovascular events and mortality in older patients with coronary disease are equal to or exceed those seen in younger patients: results from the LIPID trial. Ann Intern Med 2001;134:931–940. 5. Kannel WB, D’Agostino RB, Stepanians M, D’Agostino LC. Efficacy and tolerability of lovastatin in a six-month study: analysis by gender, age and hypertensive status. Am J Cardiol 1990;66(suppl):1B–10B. 6. Lewis SJ, Moye LA, Sacks FM, Johnstone DE, Timmis G, Mitchell J, Limacher M, Kell S, Glasser SP, Grant J, et al. Effect of pravastatin on cardiovascular events in older patients with myocardial infarction and cholesterol levels in the average range: results of the Cholesterol and Recurrent Events (CARE) trial. Ann Intern Med 1998;129:681– 689. 7. Ose L, Luurila O, Eriksson J, Olsson A, Lithell H, Widgren B. Efficacy and safety of cerivastatin, 0.2 mg and 0.4 mg, in patients with primary hypercholesterolaemia: a multinational, randomised, double-blind study. Curr Med Res Opin 1999;15:228 –240. 8. Downs JR, Clearfield M, Weis S, Whitney E, Shapiro DR, Beere PA, Langendorfer A, Stein EA, Kruyer W, Gotto AM Jr. Primary prevention of acute coronary events with lovastatin in men and women with average cholesterol levels: results of AFCAPS/TexCAPS. JAMA 1998; 279:1615–1622. 9. Simon JA, Fong J, Bernert JTJ, Browner WS. Relation of smoking and alcohol consumption to serum fatty acids. Am J Epidemiol 1996;144: 325–334. 10. Milionis HJ, Rizos E, Mikhailidis DP. Smoking diminishes the beneficial effect of statins: observations from the landmark trials. Angiology 2001;52:575–587. 11. Rothman KJ. No adjustments are needed for multiple comparisons. Epidemiology 1990;1:43– 46.