858
Lipoprotein Lipids in Older People Results From the Cardiovascular Health Study Walter H. Ettinger, MD; Patricia W. Wahl, PhD; Lewis H. Kuller, MD, DPH; Trudy L. Bush, PhD, MPH; Russell P. Tracy, PhD; Teri A. Manolio, MD, MHS; Nemat 0. Borhani, MD, MPH; Nathan D. Wong, PhD; and Daniel H. O'Leary, MD; for the CHS Collaborative Research Group
Background. Cardiovascular disease is the leading cause of death and disability in older people. There is little information about the distributions of risk factors in older populations. This article describes the distribution and correlates of lipoprotein lipids in people >65 years old. Methods and Resuls. Lipoprotein lipid concentrations were measured in 2,106 men (M) and 2,732 women (F) who were participants in the Cardiovascular Health Study, a population-based epidemiological study. Distributions of lipids by age and sex and bivariate and multivariate relations among lipids and other variables were determined in cross-sectional analyses. Mean concentrations of lipids were cholesterol: M, 5.20±0.93 mmol/l (201±36 mg/dl) and F, 5.81±0.98 mmol/l (225+±38 mg/dl); triglyceride (TG): M, 1.58+0.85 mmol/l (140±75 mg/dl) and F, 1.57±0.78 mmol/1 (139±69 mg/dl); high density lipoprotein cholesterol (HDL-C): M, 1.23±0.33 mmol/l (48±16 mgldl), and F, 1.53+±0.41 mmol/l (59±+16 mg/dl); low density lipoprotein cholesterol (LDL-C): M, 3.27±0.85 mmol/l (127±33 mg/dl) and F, 3.57±0.93 mmol/l (138+36 mg/dl). The total cholesterol to HDL-C ratios were M, 4.49±+1.29 and F, 4.05± 1.22. TG, total cholesterol, and LDL-C concentrations were lower with increasing age, the last more evident in men than in women. TG concentration was positively associated with obesity (in women), central fat patterning, glucose intolerance, use of (-blockers (in men), and use of estrogens (in women) and negatively associated with age, renal function, alcohol use, and socioeconomic status. In general, HDL-C had opposite relations with these variables, except that estrogen use was associated with higher HDL-C concentrations. LDL-C concentration was associated with far fewer variables than the other lipids but was negatively associated with age in men and women and positively correlated with obesity and central fat patterning and negatively correlated with renal function and estrogen use in women. There were no differences in total cholesterol and LDL-C concentrations among participants with and without prevalent coronary heart disease and stroke, but TG concentration was higher and HDL-C lower in men with both coronary heart disease and stroke and in women with coronary heart disease. Conclusions. Cholesterol and cholesterol/HDL-C ratio were lower and HDL-C higher than previously reported values in older people, suggesting that lipid risk profiles may be improving in older Americans. TG and HDL-C concentrations, and to a lesser extent LDL-C, were associated with potentially important modifiable factors such as obesity, glucose intolerance, renal function, and medication use. (Circulation 1992;86:858-869) KEY WoRDs * cholesterol * aging * risk factors yslipidemia is an important, modifiable risk factor for coronary heart disease in young and middle-aged adults.1-7 However, the importance of lipid levels in older adults is controversial. Several studies have suggested that the association D
From the Departments of Internal Medicine and Public Health Sciences (W.H.E.), Wake Forest University, Winston-Salem, N.C.; the Department of Biostatistics (P.W.W.), University of Washington, Seattle; the Department of Epidemiology (L.H.K.), University of Pittsburgh, Pittsburgh, Pa.; the Department of Epidemiology (T.L.B.), School of Hygiene and Public Health, Johns Hopkins University, Baltimore, Md.; the Departnents of Pathology and Biochemistry (R.P.T.), University of Vermont, Burlington; the Division of Epidemiology and Clinical Applications (T.A.M.), National Heart, Lung, and Blood Institute, Bethesda, Md.; the CHS Coordinating Center (N.O.B.), Seattle, Wash.; the Department of Medicine (N.D.W.), University of California Irvine; and the Department of Radiology (D.H.O'L.), Harvard University, Boston.
between cholesterol concentrations and atherosclerotic coronary artery disease weakens with age and that there is little potential benefit from screening and treating older adults for dyslipidemia.8-11 In contrast, other reports suggest that lipoprotein levels remain a significant risk factor for coronary heart disease in older people, and treatment of dyslipidemia in older people may have a greater impact on coronary heart disease Supported by National Heart, Lung, and Blood Institute contracts N01-HC-87079, N01-HC-87080, N01-HC-87081, N01-HC87082, N01-HC-87083, N01-HC-87084, N01-HC-87085, and N01-HC-87086. Address for reprints: CHS Coordinating Center, University of Washington, JD-36, 1107 NE 45th Street, Suite 530, Seattle, WA 98105. Address for correspondence: Walter H. Ettinger Jr., MD, Medical Center Boulevard, Winston-Salem, NC 27157. Received September 30, 1991; revision accepted May 27, 1992.
Downloaded from http://circ.ahajournals.org/ by guest on March 10, 2015
Ettinger et al Lipid Levels in Older People
859
mortality than in younger people because the total attributable risk from dyslipidemia is greater in the older age group.'2-'5 One reason for the controversy about screening and treating hyperlipidemia in older adults is the small number of published studies that have examined the power of lipid levels as risk factors for cardiovascular disease in older adults." Furthermore, few studies describe the distribution of lipoprotein concentrations in older adults. Most studies have relatively small numbers of participants, especially among the very old, and often do not give details about the demographic characteristics and health status of the sample."",16-27 In this article, we report the lipoprotein lipid levels in 4,838 adults aged .65 years from the Cardiovascular Health Study (CHS). The CHS is a population-based, longitudinal study of adults .65 years old, designed to identify factors related to the onset and course of coronary heart disease and stroke. Data from the CHS baseline examination were analyzed for the following purposes: 1) to determine the distribution of plasma lipoprotein lipids in men and women .65 years old; 2) to identify associations between plasma lipid values and demographic characteristics, health habits, and physiological measurements made at the baseline examination; and 3) to determine the association between lipoprotein lipid levels and prevalent cardiovascular disease.
isolated by centrifugation at 3,000 rpm for 30 minutes at 4°C. Serum samples for measurement of insulin, glucose, and creatinine were drawn into tubes without an anticoagulant, and the blood was allowed to coagulate at room temperature for 30 minutes. The plasma and serum samples were frozen at -70°C; samples were shipped weekly, on dry ice, to the CHS Central Blood Laboratory at the University of Vermont, where all analyses were performed. Plasma total cholesterol and triglyceride (TG) were measured by enzymatic methods on an Olympus Demand System (Olympus Corp., Lake Success, N.Y.). High density lipoprotein cholesterol (HDL-C) was measured by an enzymatic method after precipitation of apolipoprotein B-containing lipoproteins with dextran sulfate/magnesium sulfate. The CHS central blood laboratory is standardized by the Centers for Disease Control for lipid measurements. Coefficients of variation using reference standards for total cholesterol, TG, and HDL-C were 1.66%, 1.78%, and 2.15%, respectively. Low density lipoprotein cholesterol (LDL-C) was calculated according to the Friedewald equation for individuals whose serum TG was c4.51 mmol/1.29 Creatinine and fasting glucose were measured in serum with the Kodak Ektachem E-700 (Kodak, Rochester, N.Y.), and insulin was measured with a competitive radioimmunoassay from Diagnostics Products (Los Angeles, Calif.). Two-hour serum insulin and glucose were measured after ingestion of a 75-g glucose load.30
Methods
Measurement of Other Variables Standard questionnaires were used to obtain information regarding race, education, and income and use of alcohol, tobacco, and medication.28 The amount and type of physical activity were determined by a questionnaire asking about exercise and recreational activity, and results were expressed as total kilocalories per week. Medications used in the preceding 2 weeks were identified from prescription bottles. Anthropometric measurements were obtained according to a standardized protocol and included weight, height, and waist and hip circumferences.28 The presence of cardiovascular disease was determined from self-report and confirmed by medical record review. Definite coronary heart disease was defined as self-reported myocardial infarction or angina pectoris confirmed by medication use and hospital and/or physician records, self-reported coronary bypass surgery or coronary angioplasty confirmed by medical record review, or definite myocardial infarction on resting ECG by Minnesota code. Stroke was defined as self-report of stroke confirmed by medical record review. Hypertension was defined as an average (two measurements) systolic blood pressure .160 mm Hg, an average diastolic blood pressure of :95 mm Hg, or a self-report of hypertension and concomitant use of antihypertensive medications. Blood pressure was measured in the right arm with a random-zero sphygmomanometer with the participant in the sitting position. Diabetes was defined as a self-report of diabetes and use of insulin or oral hypoglycemics or a fasting blood sugar >7.8 mmol/l and/or 2-hour postglucose load blood sugar of >11.0 mmol/1.31 Impaired glucose tolerance was defined as 2-hour post-glucose load blood sugar of 7.8-11.0 mmol/1.31 Functional status was determined by assessing whether or not a person
Study Design Details of the CHS study design have been published.28 Briefly, the CHS cohort of 5,201 men and women was recruited during the period from May 1989 to May 1990 from a random sample of the Health Care Finance Administration's Medicare eligibility list in four U.S. communities: Forsyth County, North Carolina; Sacramento County, California; Washington County, Maryland; and Pittsburgh, Pennsylvania. All people >65 years old who were residents of an enumerated household were asked to participate in the study. People were excluded from participating in the study if they expected to leave the study area in the next 3 years, resided in a nursing home, were wheelchair-bound in the home, or were currently under treatment for cancer. Of persons who were contacted, 34.4% refused participation and 9.5% were ineligible. Among those enrolled, 3,654 (70.3%) were from the Medicare sampling list and 1,547 (29.7%) were recruited as other household members. Compared with those who were ineligible or those who refused, enrolled participants were younger, more highly educated, more likely to be married, and less likely to report limitations in activity. Participants were enrolled in a two-stage process in which initial eligibility was ascertained and part of the questionnaire information was collected at home, and an examination was performed in the clinic within 2 weeks.
Laboratory Methods Subjects came to the clinic after an overnight fast, and blood was obtained on their arrival at the clinic. Samples for lipoprotein lipids were drawn in tubes containing EDTA (final concentration, 1 mg/ml). Plasma was
Downloaded from http://circ.ahajournals.org/ by guest on March 10, 2015
860
Circulation Vol 86, No 3 September 1992
reported difficulty walking 0.5 mile or any difficulty with activities of daily living (bathing, dressing, using the toilet, transferring from bed to chair, or eating).32
Statistical Analysis There were 5,201 participants who completed the CHS baseline examination. Two hundred seventy-two participants who reported taking lipid-lowering medications within 2 weeks preceding the baseline examination and 91 participants who fasted less than 9 hours before phlebotomy were excluded from the analyses. Analysis of the data was based on the remaining 4,838 participants. Additionally, information was unavailable for 24 cholesterol, 27 TG, 87 LDL-C, and 30 HDL-C measurements. Cholesterol and TG are expressed both as millimoles per liter and milligrams per deciliter. Values can be converted from millimoles per liter to milligrams per deciliter by multiplication of the cholesterol value in millimoles per liter by 38.67 and the TG value by 88.6. Descriptive statistics were calculated for all lipid variables and stratified by age and sex and included the mean, SD, and percentile values. TG values displayed a positively skewed distribution. Hence, a logarithmic transformation was used for TG values in tests of significance, in computation of Pearson's correlation coefficient, or in multivariate analysis. Significant differences in lipids for all categorical variables were determined by ANOVA. Linear trends in lipid values by age were determined by ANOVA. After stratification by sex, stepwise multiple regression analysis was used to determine the linear relation between lipids and other variables, which were found to be statistically associated in bivariate analysis for either men or women. Adjusted means were calculated by ANCOVA when associations between lipid values and cardiovascular disease were determined.
Results The characteristics of the CHS population are presented in Table 1. The cohort was predominantly white. The socioeconomic profile of the cohort was diverse; however, 35% of the cohort completed grade 12 or higher, and 37% had a household income of more than $25,000/yr. Prevalent cardiovascular disease was common. Twenty-five percent of participants had definite coronary heart disease, 40% definite hypertension, and 5% a definite or probable stroke. Reflecting the high prevalence of cardiovascular disease, 26% were taking diuretics and 12% 13-blockers. Twenty-three percent had confirmed diabetes or met the World Health Organization's (WHO) criteria for diabetes from the glucose tolerance test.31 An additional 28% had impaired glucose tolerance by the WHO criteria. Despite the high prevalence of chronic diseases, the functional status of the cohort was high. Eighty percent reported no difficulty walking 0.5 mile, and only 7% reported any difficulty in an activity of daily living. Plasma lipid concentrations in CHS participants, stratified by age and sex, are shown in Table 2. (Percentiles for lipid concentrations are presented in Table 3.) Women had higher concentrations of total cholesterol, LDL-C, and HDL-C than men at all ages. In contrast, TG concentrations were similar in men and women, except that women .85 years old had significantly higher concentrations than men. Women had
lower cholesterol/HDL-C ratios except in people >85 old, in whom men had the lower ratio. TG concentrations were lower with increasing age in men. Cholesterol and LDL-C concentrations were lower among older men, particularly in men >85 years old. There was little change in HDL-C concentrations by age, except that men >85 years old had a significantly higher HDL-C than the other age groups. Among women, age-associated changes in lipid concentrations were less apparent. TG concentrations were lower in women >85 years old, and cholesterol and LDL-C concentrations were slightly lower in women .80 years old compared with younger women, but there were no statistically significant differences in HDL-C with age. The cholesterol/HDL-C ratios declined with age in men but not women. There were significant differences in lipoprotein lipid values by race (Table 4). African-American men had lower TG and higher HDL-C concentrations than whites, but there were no statistically significant differences in total cholesterol or LDL-C concentrations. African-American women had significantly lower TG, total cholesterol, and LDL-C concentrations than whites, but there were no differences in HDL-C concentrations. We examined the associations of lipoprotein lipid values and other factors that have been shown in other studies to be associated with lipid values and are thought to be biologically important determinants of lipoprotein metabolism.23,25,33-44Table 5 shows bivariate associations for TG, LDL-C, and HDL-C. TG concentration was inversely related to socioeconomic level, alcohol consumption, and physical activity level. TG concentration was positively associated with body mass index, waist circumference, measures of glucose intolerance (insulin and glucose concentrations), and creatinine (which is inversely related to glomerular filtration rate, a measure of renal function). Subjects taking either diuretics or 13-blockers had higher TG values, and estrogen use was associated with a higher TG value in women. The correlations of HDL-C with these variables were opposite, for the most part, to those for TG, including socioeconomic status, alcohol use, physical activity level, body weight, body fat distribution (as measured by waist and hip circumference), glucose intolerance, renal function, and medication use. Estrogen use was an exception to the latter, as HDL-C was higher in women taking sex hormones. In general, the associations between TG and HDL-C concentrations and other variables were similar and of the same magnitude in both men and women. There were few significant correlations between LDL-C and other variables, but the associations were more significant in women. Body mass index, waist circumference, glucose intolerance, and creatinine were positively associated with LDL-C in women. Estrogen use was inversely related to LDL-C concentrations in women. None of the lipid values were related to functional status or cigarette use. Table 6 presents stepwise, multivariate models that include the variables that were significantly associated with each lipid in bivariate analysis for either men or women. For the most part, variables that were significant in bivariate analyses remained significant in multivariate analyses. In men, however, body mass index, the
years
Downloaded from http://circ.ahajournals.org/ by guest on March 10, 2015
Ettinger et al Lipid Levels in Older People
861
TABLE 1. Demographic, Clinical, and Metabolic Characteristics of Cardiovascular Health Study Participants Stratified by Sex Men Women (n=2,106) (n=2,732) Demographic characteristics Age (years) (mean±SD) 73.4±5.8 72.5+5.5 Race White, n (%) 2,571 (94.4) 1,999 (95.0) Black, n (%) 94 (4.5) 134 (4.9) 12 (0.6) Other, n (%) 18 (0.7) Education (highest grade completed) Grade 12, n (%) 833 (39.6) 861 (31.7) Income (annual, total household) 310 (15.5) 755 (30.1) 85 Total
98
2,097
p
1.64+0.90 (145±80) 1.59±0.79 (141±70) 1.52±0.82 (135±73) 1.40±0.59 (124±52) 1.30±0.81 (115±72) 1.58±0.85 (140±75) '0.0001
Total cholesterol/ HDL-C ratio
5.31+0.90
3.36±0.84
(205±35)
(130±32)
(48±13)
5.20±0.89 (201±34) 5.14±0.95 (199±37) 5.14+1.1 (199±43) 4.87±0.85 (188±33) 5.20±0.93 (201±36)
3.27±0.82 (127±32) 3.22±0.84 (125±32) 3.23±1.0 (125±39) 2.95±0.75 (114±29) 3.27±0.85 (127±33)
1.20±0.32 (46±12) 1.22±0.34 (47±13) 1.27±0.32 (49±12) 1.32±0.36 (51±14) 1.23±0.33 (48±13)
.0.0001
85 Total
548 233 88
2,717
1.61±0.83 (143±74) 1.59±0.75 (141±66) 1.55±0.76 (137±37) 1.48+0.74 (131 ±66) 1.50+0.64t
5.84+1.Ot (226±39) 5.87±0.98t (227±38) 5.78±0.96t (224±37) 5.64±0.94t (218±36) 5.66±1.1t
(133±57)
(219±43)
3.59±0.92t (139+36) 3.55±0.89t (138±34)
1.52±0.41t
3.43±0.88t
1.54±0.44t (60±17)
3.50±1.Ot (135±39)
1.46±0.35t (57± 16) 1.53+0.41t
1.57±0.78
5.81±0.98t
3.57±0.93t
(225±38)
(138±36)
(59±16) 0.15 .0.01 0.13 0.15 0.15 p* LDL-C, low density lipoprotein cholesterol; HDL-C, high density lipoprotein cholesterol. Values are mean±SD. Numbers in parentheses are concentrations in mg/dl. *Statistical comparison for linear change across age groups. tStatistically significantly different from men, p240 mg/dl) by National Cholesterol Education Program guidelines. Forty percent of subjects had borderline high LDL-C levels (130-159 mg/dl), and 46% had high levels (>160 mg/ dl). The proportion of subjects with hyperlipidemia was substantial even among people >80 years old. These data indicate the need to quantify the risk associated with lipoprotein lipids in older adults and to determine the effectiveness of screening and treatment of dyslipidemia in the older age group.11,14 As in other studies, cholesterol concentrations were lower in the older age groups, with the effect more predominant in men than women (Figure 1).16222327 This finding can be explained, in part, by the earlier death of people with an atherogenic lipid profile. Consistent with this hypothesis, the negative association of cholesterol with age appeared to result from a decline in both LDL-C and very low density lipoprotein (VLDL) cholesterol (estimated in this study as TG/5), both of which are risk factors for coronary heart disease.49-51 However, some longitudinal studies support a true age-related fall in plasma cholesterol levels.52,53 The reasons for the decline in plasma cholesterol with age are unknown but may be related to changes in nutrient intake, body weight, and health status. Longitudinal
TABLE 4. Plasma Lipid Concentrations in Cardiovascular Health Study Participants Stratified by Race and Sex Cholesterol LDL-C Triglycerides (mmOIJ/) Race Number of subjects (mmOI/1) (mmolll) Men White 1.58+0.91 5.19+0.90 3.25+0.85 1,987 (126±33) (201±35) (140±81) 93 5.21±1.05 3.17±0.93 African-American 1.39±1.1 (123±36) (202±41) (123±101)
Lipoprotein Lipids in Older People Results From the Cardiovascular Health Study Walter H. Ettinger, MD; Patricia W. Wahl, PhD; Lewis H. Kuller, MD, DPH; Trudy L. Bush, PhD, MPH; Russell P. Tracy, PhD; Teri A. Manolio, MD, MHS; Nemat 0. Borhani, MD, MPH; Nathan D. Wong, PhD; and Daniel H. O'Leary, MD; for the CHS Collaborative Research Group
Background. Cardiovascular disease is the leading cause of death and disability in older people. There is little information about the distributions of risk factors in older populations. This article describes the distribution and correlates of lipoprotein lipids in people >65 years old. Methods and Resuls. Lipoprotein lipid concentrations were measured in 2,106 men (M) and 2,732 women (F) who were participants in the Cardiovascular Health Study, a population-based epidemiological study. Distributions of lipids by age and sex and bivariate and multivariate relations among lipids and other variables were determined in cross-sectional analyses. Mean concentrations of lipids were cholesterol: M, 5.20±0.93 mmol/l (201±36 mg/dl) and F, 5.81±0.98 mmol/l (225+±38 mg/dl); triglyceride (TG): M, 1.58+0.85 mmol/l (140±75 mg/dl) and F, 1.57±0.78 mmol/1 (139±69 mg/dl); high density lipoprotein cholesterol (HDL-C): M, 1.23±0.33 mmol/l (48±16 mgldl), and F, 1.53+±0.41 mmol/l (59±+16 mg/dl); low density lipoprotein cholesterol (LDL-C): M, 3.27±0.85 mmol/l (127±33 mg/dl) and F, 3.57±0.93 mmol/l (138+36 mg/dl). The total cholesterol to HDL-C ratios were M, 4.49±+1.29 and F, 4.05± 1.22. TG, total cholesterol, and LDL-C concentrations were lower with increasing age, the last more evident in men than in women. TG concentration was positively associated with obesity (in women), central fat patterning, glucose intolerance, use of (-blockers (in men), and use of estrogens (in women) and negatively associated with age, renal function, alcohol use, and socioeconomic status. In general, HDL-C had opposite relations with these variables, except that estrogen use was associated with higher HDL-C concentrations. LDL-C concentration was associated with far fewer variables than the other lipids but was negatively associated with age in men and women and positively correlated with obesity and central fat patterning and negatively correlated with renal function and estrogen use in women. There were no differences in total cholesterol and LDL-C concentrations among participants with and without prevalent coronary heart disease and stroke, but TG concentration was higher and HDL-C lower in men with both coronary heart disease and stroke and in women with coronary heart disease. Conclusions. Cholesterol and cholesterol/HDL-C ratio were lower and HDL-C higher than previously reported values in older people, suggesting that lipid risk profiles may be improving in older Americans. TG and HDL-C concentrations, and to a lesser extent LDL-C, were associated with potentially important modifiable factors such as obesity, glucose intolerance, renal function, and medication use. (Circulation 1992;86:858-869) KEY WoRDs * cholesterol * aging * risk factors yslipidemia is an important, modifiable risk factor for coronary heart disease in young and middle-aged adults.1-7 However, the importance of lipid levels in older adults is controversial. Several studies have suggested that the association D
From the Departments of Internal Medicine and Public Health Sciences (W.H.E.), Wake Forest University, Winston-Salem, N.C.; the Department of Biostatistics (P.W.W.), University of Washington, Seattle; the Department of Epidemiology (L.H.K.), University of Pittsburgh, Pittsburgh, Pa.; the Department of Epidemiology (T.L.B.), School of Hygiene and Public Health, Johns Hopkins University, Baltimore, Md.; the Departnents of Pathology and Biochemistry (R.P.T.), University of Vermont, Burlington; the Division of Epidemiology and Clinical Applications (T.A.M.), National Heart, Lung, and Blood Institute, Bethesda, Md.; the CHS Coordinating Center (N.O.B.), Seattle, Wash.; the Department of Medicine (N.D.W.), University of California Irvine; and the Department of Radiology (D.H.O'L.), Harvard University, Boston.
between cholesterol concentrations and atherosclerotic coronary artery disease weakens with age and that there is little potential benefit from screening and treating older adults for dyslipidemia.8-11 In contrast, other reports suggest that lipoprotein levels remain a significant risk factor for coronary heart disease in older people, and treatment of dyslipidemia in older people may have a greater impact on coronary heart disease Supported by National Heart, Lung, and Blood Institute contracts N01-HC-87079, N01-HC-87080, N01-HC-87081, N01-HC87082, N01-HC-87083, N01-HC-87084, N01-HC-87085, and N01-HC-87086. Address for reprints: CHS Coordinating Center, University of Washington, JD-36, 1107 NE 45th Street, Suite 530, Seattle, WA 98105. Address for correspondence: Walter H. Ettinger Jr., MD, Medical Center Boulevard, Winston-Salem, NC 27157. Received September 30, 1991; revision accepted May 27, 1992.
Downloaded from http://circ.ahajournals.org/ by guest on March 10, 2015
Ettinger et al Lipid Levels in Older People
859
mortality than in younger people because the total attributable risk from dyslipidemia is greater in the older age group.'2-'5 One reason for the controversy about screening and treating hyperlipidemia in older adults is the small number of published studies that have examined the power of lipid levels as risk factors for cardiovascular disease in older adults." Furthermore, few studies describe the distribution of lipoprotein concentrations in older adults. Most studies have relatively small numbers of participants, especially among the very old, and often do not give details about the demographic characteristics and health status of the sample."",16-27 In this article, we report the lipoprotein lipid levels in 4,838 adults aged .65 years from the Cardiovascular Health Study (CHS). The CHS is a population-based, longitudinal study of adults .65 years old, designed to identify factors related to the onset and course of coronary heart disease and stroke. Data from the CHS baseline examination were analyzed for the following purposes: 1) to determine the distribution of plasma lipoprotein lipids in men and women .65 years old; 2) to identify associations between plasma lipid values and demographic characteristics, health habits, and physiological measurements made at the baseline examination; and 3) to determine the association between lipoprotein lipid levels and prevalent cardiovascular disease.
isolated by centrifugation at 3,000 rpm for 30 minutes at 4°C. Serum samples for measurement of insulin, glucose, and creatinine were drawn into tubes without an anticoagulant, and the blood was allowed to coagulate at room temperature for 30 minutes. The plasma and serum samples were frozen at -70°C; samples were shipped weekly, on dry ice, to the CHS Central Blood Laboratory at the University of Vermont, where all analyses were performed. Plasma total cholesterol and triglyceride (TG) were measured by enzymatic methods on an Olympus Demand System (Olympus Corp., Lake Success, N.Y.). High density lipoprotein cholesterol (HDL-C) was measured by an enzymatic method after precipitation of apolipoprotein B-containing lipoproteins with dextran sulfate/magnesium sulfate. The CHS central blood laboratory is standardized by the Centers for Disease Control for lipid measurements. Coefficients of variation using reference standards for total cholesterol, TG, and HDL-C were 1.66%, 1.78%, and 2.15%, respectively. Low density lipoprotein cholesterol (LDL-C) was calculated according to the Friedewald equation for individuals whose serum TG was c4.51 mmol/1.29 Creatinine and fasting glucose were measured in serum with the Kodak Ektachem E-700 (Kodak, Rochester, N.Y.), and insulin was measured with a competitive radioimmunoassay from Diagnostics Products (Los Angeles, Calif.). Two-hour serum insulin and glucose were measured after ingestion of a 75-g glucose load.30
Methods
Measurement of Other Variables Standard questionnaires were used to obtain information regarding race, education, and income and use of alcohol, tobacco, and medication.28 The amount and type of physical activity were determined by a questionnaire asking about exercise and recreational activity, and results were expressed as total kilocalories per week. Medications used in the preceding 2 weeks were identified from prescription bottles. Anthropometric measurements were obtained according to a standardized protocol and included weight, height, and waist and hip circumferences.28 The presence of cardiovascular disease was determined from self-report and confirmed by medical record review. Definite coronary heart disease was defined as self-reported myocardial infarction or angina pectoris confirmed by medication use and hospital and/or physician records, self-reported coronary bypass surgery or coronary angioplasty confirmed by medical record review, or definite myocardial infarction on resting ECG by Minnesota code. Stroke was defined as self-report of stroke confirmed by medical record review. Hypertension was defined as an average (two measurements) systolic blood pressure .160 mm Hg, an average diastolic blood pressure of :95 mm Hg, or a self-report of hypertension and concomitant use of antihypertensive medications. Blood pressure was measured in the right arm with a random-zero sphygmomanometer with the participant in the sitting position. Diabetes was defined as a self-report of diabetes and use of insulin or oral hypoglycemics or a fasting blood sugar >7.8 mmol/l and/or 2-hour postglucose load blood sugar of >11.0 mmol/1.31 Impaired glucose tolerance was defined as 2-hour post-glucose load blood sugar of 7.8-11.0 mmol/1.31 Functional status was determined by assessing whether or not a person
Study Design Details of the CHS study design have been published.28 Briefly, the CHS cohort of 5,201 men and women was recruited during the period from May 1989 to May 1990 from a random sample of the Health Care Finance Administration's Medicare eligibility list in four U.S. communities: Forsyth County, North Carolina; Sacramento County, California; Washington County, Maryland; and Pittsburgh, Pennsylvania. All people >65 years old who were residents of an enumerated household were asked to participate in the study. People were excluded from participating in the study if they expected to leave the study area in the next 3 years, resided in a nursing home, were wheelchair-bound in the home, or were currently under treatment for cancer. Of persons who were contacted, 34.4% refused participation and 9.5% were ineligible. Among those enrolled, 3,654 (70.3%) were from the Medicare sampling list and 1,547 (29.7%) were recruited as other household members. Compared with those who were ineligible or those who refused, enrolled participants were younger, more highly educated, more likely to be married, and less likely to report limitations in activity. Participants were enrolled in a two-stage process in which initial eligibility was ascertained and part of the questionnaire information was collected at home, and an examination was performed in the clinic within 2 weeks.
Laboratory Methods Subjects came to the clinic after an overnight fast, and blood was obtained on their arrival at the clinic. Samples for lipoprotein lipids were drawn in tubes containing EDTA (final concentration, 1 mg/ml). Plasma was
Downloaded from http://circ.ahajournals.org/ by guest on March 10, 2015
860
Circulation Vol 86, No 3 September 1992
reported difficulty walking 0.5 mile or any difficulty with activities of daily living (bathing, dressing, using the toilet, transferring from bed to chair, or eating).32
Statistical Analysis There were 5,201 participants who completed the CHS baseline examination. Two hundred seventy-two participants who reported taking lipid-lowering medications within 2 weeks preceding the baseline examination and 91 participants who fasted less than 9 hours before phlebotomy were excluded from the analyses. Analysis of the data was based on the remaining 4,838 participants. Additionally, information was unavailable for 24 cholesterol, 27 TG, 87 LDL-C, and 30 HDL-C measurements. Cholesterol and TG are expressed both as millimoles per liter and milligrams per deciliter. Values can be converted from millimoles per liter to milligrams per deciliter by multiplication of the cholesterol value in millimoles per liter by 38.67 and the TG value by 88.6. Descriptive statistics were calculated for all lipid variables and stratified by age and sex and included the mean, SD, and percentile values. TG values displayed a positively skewed distribution. Hence, a logarithmic transformation was used for TG values in tests of significance, in computation of Pearson's correlation coefficient, or in multivariate analysis. Significant differences in lipids for all categorical variables were determined by ANOVA. Linear trends in lipid values by age were determined by ANOVA. After stratification by sex, stepwise multiple regression analysis was used to determine the linear relation between lipids and other variables, which were found to be statistically associated in bivariate analysis for either men or women. Adjusted means were calculated by ANCOVA when associations between lipid values and cardiovascular disease were determined.
Results The characteristics of the CHS population are presented in Table 1. The cohort was predominantly white. The socioeconomic profile of the cohort was diverse; however, 35% of the cohort completed grade 12 or higher, and 37% had a household income of more than $25,000/yr. Prevalent cardiovascular disease was common. Twenty-five percent of participants had definite coronary heart disease, 40% definite hypertension, and 5% a definite or probable stroke. Reflecting the high prevalence of cardiovascular disease, 26% were taking diuretics and 12% 13-blockers. Twenty-three percent had confirmed diabetes or met the World Health Organization's (WHO) criteria for diabetes from the glucose tolerance test.31 An additional 28% had impaired glucose tolerance by the WHO criteria. Despite the high prevalence of chronic diseases, the functional status of the cohort was high. Eighty percent reported no difficulty walking 0.5 mile, and only 7% reported any difficulty in an activity of daily living. Plasma lipid concentrations in CHS participants, stratified by age and sex, are shown in Table 2. (Percentiles for lipid concentrations are presented in Table 3.) Women had higher concentrations of total cholesterol, LDL-C, and HDL-C than men at all ages. In contrast, TG concentrations were similar in men and women, except that women .85 years old had significantly higher concentrations than men. Women had
lower cholesterol/HDL-C ratios except in people >85 old, in whom men had the lower ratio. TG concentrations were lower with increasing age in men. Cholesterol and LDL-C concentrations were lower among older men, particularly in men >85 years old. There was little change in HDL-C concentrations by age, except that men >85 years old had a significantly higher HDL-C than the other age groups. Among women, age-associated changes in lipid concentrations were less apparent. TG concentrations were lower in women >85 years old, and cholesterol and LDL-C concentrations were slightly lower in women .80 years old compared with younger women, but there were no statistically significant differences in HDL-C with age. The cholesterol/HDL-C ratios declined with age in men but not women. There were significant differences in lipoprotein lipid values by race (Table 4). African-American men had lower TG and higher HDL-C concentrations than whites, but there were no statistically significant differences in total cholesterol or LDL-C concentrations. African-American women had significantly lower TG, total cholesterol, and LDL-C concentrations than whites, but there were no differences in HDL-C concentrations. We examined the associations of lipoprotein lipid values and other factors that have been shown in other studies to be associated with lipid values and are thought to be biologically important determinants of lipoprotein metabolism.23,25,33-44Table 5 shows bivariate associations for TG, LDL-C, and HDL-C. TG concentration was inversely related to socioeconomic level, alcohol consumption, and physical activity level. TG concentration was positively associated with body mass index, waist circumference, measures of glucose intolerance (insulin and glucose concentrations), and creatinine (which is inversely related to glomerular filtration rate, a measure of renal function). Subjects taking either diuretics or 13-blockers had higher TG values, and estrogen use was associated with a higher TG value in women. The correlations of HDL-C with these variables were opposite, for the most part, to those for TG, including socioeconomic status, alcohol use, physical activity level, body weight, body fat distribution (as measured by waist and hip circumference), glucose intolerance, renal function, and medication use. Estrogen use was an exception to the latter, as HDL-C was higher in women taking sex hormones. In general, the associations between TG and HDL-C concentrations and other variables were similar and of the same magnitude in both men and women. There were few significant correlations between LDL-C and other variables, but the associations were more significant in women. Body mass index, waist circumference, glucose intolerance, and creatinine were positively associated with LDL-C in women. Estrogen use was inversely related to LDL-C concentrations in women. None of the lipid values were related to functional status or cigarette use. Table 6 presents stepwise, multivariate models that include the variables that were significantly associated with each lipid in bivariate analysis for either men or women. For the most part, variables that were significant in bivariate analyses remained significant in multivariate analyses. In men, however, body mass index, the
years
Downloaded from http://circ.ahajournals.org/ by guest on March 10, 2015
Ettinger et al Lipid Levels in Older People
861
TABLE 1. Demographic, Clinical, and Metabolic Characteristics of Cardiovascular Health Study Participants Stratified by Sex Men Women (n=2,106) (n=2,732) Demographic characteristics Age (years) (mean±SD) 73.4±5.8 72.5+5.5 Race White, n (%) 2,571 (94.4) 1,999 (95.0) Black, n (%) 94 (4.5) 134 (4.9) 12 (0.6) Other, n (%) 18 (0.7) Education (highest grade completed) Grade 12, n (%) 833 (39.6) 861 (31.7) Income (annual, total household) 310 (15.5) 755 (30.1) 85 Total
98
2,097
p
1.64+0.90 (145±80) 1.59±0.79 (141±70) 1.52±0.82 (135±73) 1.40±0.59 (124±52) 1.30±0.81 (115±72) 1.58±0.85 (140±75) '0.0001
Total cholesterol/ HDL-C ratio
5.31+0.90
3.36±0.84
(205±35)
(130±32)
(48±13)
5.20±0.89 (201±34) 5.14±0.95 (199±37) 5.14+1.1 (199±43) 4.87±0.85 (188±33) 5.20±0.93 (201±36)
3.27±0.82 (127±32) 3.22±0.84 (125±32) 3.23±1.0 (125±39) 2.95±0.75 (114±29) 3.27±0.85 (127±33)
1.20±0.32 (46±12) 1.22±0.34 (47±13) 1.27±0.32 (49±12) 1.32±0.36 (51±14) 1.23±0.33 (48±13)
.0.0001
85 Total
548 233 88
2,717
1.61±0.83 (143±74) 1.59±0.75 (141±66) 1.55±0.76 (137±37) 1.48+0.74 (131 ±66) 1.50+0.64t
5.84+1.Ot (226±39) 5.87±0.98t (227±38) 5.78±0.96t (224±37) 5.64±0.94t (218±36) 5.66±1.1t
(133±57)
(219±43)
3.59±0.92t (139+36) 3.55±0.89t (138±34)
1.52±0.41t
3.43±0.88t
1.54±0.44t (60±17)
3.50±1.Ot (135±39)
1.46±0.35t (57± 16) 1.53+0.41t
1.57±0.78
5.81±0.98t
3.57±0.93t
(225±38)
(138±36)
(59±16) 0.15 .0.01 0.13 0.15 0.15 p* LDL-C, low density lipoprotein cholesterol; HDL-C, high density lipoprotein cholesterol. Values are mean±SD. Numbers in parentheses are concentrations in mg/dl. *Statistical comparison for linear change across age groups. tStatistically significantly different from men, p240 mg/dl) by National Cholesterol Education Program guidelines. Forty percent of subjects had borderline high LDL-C levels (130-159 mg/dl), and 46% had high levels (>160 mg/ dl). The proportion of subjects with hyperlipidemia was substantial even among people >80 years old. These data indicate the need to quantify the risk associated with lipoprotein lipids in older adults and to determine the effectiveness of screening and treatment of dyslipidemia in the older age group.11,14 As in other studies, cholesterol concentrations were lower in the older age groups, with the effect more predominant in men than women (Figure 1).16222327 This finding can be explained, in part, by the earlier death of people with an atherogenic lipid profile. Consistent with this hypothesis, the negative association of cholesterol with age appeared to result from a decline in both LDL-C and very low density lipoprotein (VLDL) cholesterol (estimated in this study as TG/5), both of which are risk factors for coronary heart disease.49-51 However, some longitudinal studies support a true age-related fall in plasma cholesterol levels.52,53 The reasons for the decline in plasma cholesterol with age are unknown but may be related to changes in nutrient intake, body weight, and health status. Longitudinal
TABLE 4. Plasma Lipid Concentrations in Cardiovascular Health Study Participants Stratified by Race and Sex Cholesterol LDL-C Triglycerides (mmOIJ/) Race Number of subjects (mmOI/1) (mmolll) Men White 1.58+0.91 5.19+0.90 3.25+0.85 1,987 (126±33) (201±35) (140±81) 93 5.21±1.05 3.17±0.93 African-American 1.39±1.1 (123±36) (202±41) (123±101)
