American Journal of Epidemiology Copynght C 1992 by The Johns Hopkins University School of Hyojene and Pubic Health All rights reserved
Vol 136, No. 5 Printed in U.S.A.
Postprandial Lipemia: Reliability in an Epidemiologic Field Study
Ten subjects from the Forsyth County, North Carolina, and Washington County, Maryland, field centers in the Atherosclerosis Risk in Communities Study had two fat tolerance tests within a 10-day period from September 1988 to February 1989 to determine the reproducibility of markers for postprandial lipemia. No significant differences between visits were found in fasting mean plasma lipids, lipoproteins, and apolipoproteins. Postprandial triglycerides and retinyl palmitate were measured at 3.5 and 9.0 hours after the test meal in whole plasma. There were no significant differences in the mean levels of these analytes between visits. The correlation of triglycerides between repeat visits at 9.0 hours (r = 0.87) was stronger than in fasting samples {r = 0.67) or at 3.5 hours (r = 0.69). The mean plasma retinyl palmitate level at 3.5 hours was 15% higher than at the 9.0-hour level. The correlation of repeat measures of retinyl palmitate at 9.0 hours (r = 0.94) was much stronger than at 3.5 hours (r = 0.79). In conclusion, estimates of reliability in postprandial measurements of 9.0-hour triglycerides and retinyl palmitate levels were as strong as fasting lipid measurements of total cholesterol, high density lipoprotein cholesterol, low density lipoprotein cholesterol, and high density lipoprotein3 cholesterol, and both postprandial triglyceride measurements exceeded that of fasting triglyceride (r = 0.67). Am J Epidemiol 1992;136:538-45. lipoproteins, HDL cholesterol; lipoproteins, LDL cholesterol; triglycerides
used to control for either total cholesterol or low density lipoprotein cholesterol (LDL cholesterol), this association between triglycerides and coronary heart disease remained (6-9). However, prospective studies have shown that this association is not maintained upon correction for high density lipoprotein cholesterol (HDL cholesterol) because of a strong inverse relation of fasting triglyceride levels with HDL cholesterol levels (9-12). Although HDL cholesterol is a more powerful indicator of coronary heart disease risk than are triglycerides, other evidence indicates that the metabolism of plasma triglycerides influences HDL cholesterol levels (13). A relation between postprandial lipemia and coronary heart disease was extensively demonstrated as early as 1950 (14) and has subsequently been confirmed (15-21). In these studies, postprandial triglyceride levels and other markers of triglyceride-rich lipo-
Plasma triglyceride levels have been shown to be associated with the risk of coronary heart disease (1). In a number of studies, a univariate association between triglycerides and coronary heart disease was found (2-5). When multivariate analysis was
Received for publication November 8,1991, and in final form April 2, 1992. Abbreviations: ARC, Atherosclerosis Risk in Communities; HDL cholesterol, high density lipoprotein cholesterol, LDL cholesterol, low density lipoprotein cholesterol. ' Department of Medicine, Baylor College of Medicine, and Trie Methodist Hospital, Houston, TX. 2 Department of Biostatistics, University of North Carolina, Chapel Hill, NC. 3 Epidemiology and Biometry Program, National Heart, Lung, and Blood Institute, Bethesda, MD. Reprint requests to Dr. Spencer A Brown, The Methodist Hospital, Mail Stop A-601, 6565 Fannin Avenue, Houston, TX 77030. This work was supported in part by Atherosclerosis Risk in Communities (ARIC) Study grant HC NO1-HC55016 and grant 27341 from the National Heart, Lung, and Blood Institute.
538
Downloaded from https://academic.oup.com/aje/article-abstract/136/5/538/98759 by Insead user on 11 October 2018
Spencer A. Brown,1 Uoyd E. ChamWess,2 A. Richey Sharrett,3 Antonio M. Gotto, Jr.,1 and Wolfgang Patsch1
Reliability in an Epidemiologic Field Study
MATERIALS AND METHODS Subjects
Ten volunteer subjects, seven women and three men aged 45-64 years, were selected by using inclusion criteria identical to that of the ARIC Study (24). Subjects were selected without regard to lipid levels and were thought to be broadly representative of the residents of their communities. There were five subjects from each of two field centers, Forsyth County, North Carolina, and Washington County, Maryland. Subjects were scheduled for two visits to their respective field center within a 10-day time period. Fat tolerance test
On the day of the fat tolerance test, subjects arrived at the field centers after fasting for 12 hours. A fasting blood specimen was taken, with blood collected into ethylenediaminetetraacetic acid (EDTA)-containing tubes (1.5 mg/ml of blood). The test meal consisted of 1,374 kcal, 10.6 g of protein, 49.5 g of carbohydrate, 481 mg of cholesterol, 130.3 of g fat, and 100,000 IU of Aquasol A drops, Vitamin A (Armour Pharmaceutical Co., Kankakee, Illinois) (25). The test meal was administered per 2 square meters of body surface area. The volume of the entire test meal, 377 ml, was reduced in proportion to the participant's body surface area, which was determined from weight and height measurements. After the ingestion of the test meal, the participant was instructed not to take anything by mouth except water, sugarless drinks, or coffee for a period of 9.0 hours. Two postprandial blood specimens were collected, one at 3.5 hours and one at 9.0 hours after the ingestion of the test meal. Plasma was separated by centrifugation (1,500 x g, 20 minutes at 4°C) and stored at 4°C for 1-3 days. All plasma specimens were shipped on crushed ice to the laboratory by an overnight carrier. All laboratory measurements were performed within a week in the Atherosclerosis Clinical Laboratory of Baylor College of Medicine, Houston, Texas.
Downloaded from https://academic.oup.com/aje/article-abstract/136/5/538/98759 by Insead user on 11 October 2018
protein metabolism were higher in patients with clinically documented atherosclerosis compared with controls. However, in none of these early studies was HDL cholesterol taken into account. Two recent studies suggest that postprandial triglycerides are at least as good as HDL cholesterol for discriminating cases from controls as verified by angiography (22, 23). To examine a possible causal relation between postprandial triglycerides and the risk of atherosclerosis, two experimental designs can be used. One design utilizes prospective epidemiologic studies. A second design uses a cross-sectional study in which a causal relation may be demonstrated in asymptomatic individuals with documented arterial disease. In such a design, confounding effects introduced by medical intervention or resulting from potential effects of myocardial compromise on triglyceride metabolism may be minimized. The purpose of a recent National Heart, Lung, and Blood Institute research initiative was to determine whether there is a possible causal relation between postprandial lipemia and atherosclerosis. One of two populations to be investigated was selected from the Atherosclerosis Risk in Communities (ARIC) Study (24). The postprandial protocol of this study has taken advantage of the results from two recent clinical trials in which postprandial triglycerides and other markers of triglyceride-rich lipoprotein measured 6-12 hours after the fat tolerance test meal were found to be most discriminatory for angiographically assessed coronary heart disease (22, 23). Since ARIC is a multicenter study, the level of complexity, i.e., preparing of test meals and processing and transporting specimens, is increased, and standardization of postprandial protocols is required. Therefore, a pilot study was conducted to determine the reproducibility of plasma lipid markers during the postprandial state in a multicenter setting. We report here that the postprandial measurements were more reproducible than fasting triglyceride determinations. Thus, reliable postprandial studies can be implemented in epidemiologic studies.
539
540
Brown et al.
Laboratory measurements
Statistical analysis
Mean differences of laboratory measures between visits were compared by a paired / test (35). In order to assess the repeatability of the laboratory measures made on individuals at repeat visits, a one-factor nested random effects model was used to estimate between-person and combined withinperson plus method variances (36). From these components, the reliability coefficient R was estimated as the proportion of total variance in the between-person component. R is also interpretable as the correlation between measures on an individual made at visits approximately 1 week apart. An alternate estimate of this correlation is the Spearman rank correlation (35), which is reported along with scatter diagrams of the data. RESULTS
Mean fasting plasma lipid, lipoprotein, and apolipoprotein levels at each visit are shown in table 1. All subjects except one (180 mg/100 ml) had LDL cholesterol levels of less than 160 mg/100 ml, and triglyceride levels ranged from 64-187 mg/100 ml. All subjects except one (33 mg/100 ml) had HDL cholesterol levels of greater than 35
TABLE 1. Fasting lipids, lipoproteins, and apolipoproteins in two visits in 10 individuals in the Atherosclerosis Risk in Communities Study, September 1988 to February 1989 Vlsit2
Visit 1 Analyte
Cholesterol* Triglyceride HDL cholesterolt HDU cholesterol HDLj cholesterol LDL cholesterolt Apolipoprotein A-ljApolipoprotein B%
Mean
Standard deviation
Mean
Standard deviation
202.7 118.4 56.8 42.4 14.4 122.1 134.2 78.8
34.84 42.36 12.15 8.72 5.44 36.52 25.20 18.79
196.5 108.6 55.5 40.5 14.9 118.0 137.8 70.3
29.35 31.12 18.03 12.12 7.83 33.94 25.60 14.57
* All results are expressed as mg/100 ml. t HDL cholesterol, high density lipoprotein cholesterol; LDL cholesterol, low density Ipoproteln cholesterol. | n •=> 10 for all analytes except when noted, when n = 9
Downloaded from https://academic.oup.com/aje/article-abstract/136/5/538/98759 by Insead user on 11 October 2018
Total plasma cholesterol (26) and triglycerides (27) were measured enzymatically on a Cobas-Bio centrifugal analyzer (Roche Diagnostics, Montclair, New Jersey) with the respective enzymatic kits (catalog nos. 236691 and 701912; Boehringer Mannheim Diagnostics, Indianapolis, Indiana). HDL cholesterol was determined by measuring cholesterol in the supernate after precipitation of plasma with MgCl2 and dextran sulfate (28). Concentrations of HDL3 cholesterol were determined after reprecipitation of the total HDL supernate with different concentrations of MgCl2 and dextran sulfate (29, 30). HDL2 cholesterol concentration was calculated by subtracting the HDL3 cholesterol value from the value for total HDL cholesterol. LDL cholesterol was calculated according to the method of Friedewald et al. (31). Apolipoprotein A-I (32) and apolipoprotein B levels (33) were determined by radioimmunoassay methods. The coefficients of laboratory variation were calculated from both Centers for Disease Control serum pools and internal quality control plasma pools at multiple analyte levels. The laboratory coefficients of variation were 2.5 percent for cholesterol, 2.7 percent for triglycerides, 3.7 percent for HDL cholesterol, 5.2 percent for LDL cholesterol, 9.0 percent for apolipoprotein A-I, and 9.0 percent for apolipoprotein B. Retinyl palmitate levels were measured by high pressure liquid chro-
matography (SP8000; Spectra-Physics, San Jose, California) using a C-18 reverse phase column (3.9 x 300 mm uBondapak C18; Waters Instrument Co., Milford, Massachusetts) (34).
Reliability in an Epidemiologic Field Study
retinyl palmitate levels. The range of levels at this time point increased dramatically to 19.5-fold (range, 198-4,060 fig/100 ml), and the correlation coefficient for retinyl palmitate exceeded that of triglyceride at the late time point. Estimates of reliability coefficients in fasting and postprandial measurements are shown in table 3. As expected, fasting total plasma cholesterol, HDL cholesterol, HDL3 cholesterol, and LDL cholesterol measurements were highly reproducible between the two visits. The apolipoproteins were somewhat less reliable, and triglycerides had the lowest reliability. During the postprandial state, the reliability of postprandial triglyceride measurements was increased at 9.0 hours compared with the 3.5-hour measurements. Retinyl palmitate levels also increased in reliability from the 3.5-hour to the 9.0-hour measurements. In comparison, the reliability of 9.0-hour postprandial triglycerides and retinyl palmitate levels was similar to that of fasting total cholesterol, HDL cholesterol, LDL cholesterol, and HDL3 cholesterol measurements. DISCUSSION
A large number of studies have shown an association between high triglyceride levels and coronary heart disease, even when adjustment for other markers of plasma lipid transport is performed (1). Previous studies, including those with angiographically assessed subjects, suggest that postprandial
TABLE 2. Fasting and postprandial plasma retinyl palmitate and triglyceride levels in two visits In 10 individuals In the Atherosclerosis Risk in Communities Study, September 1988 to February 1989. Visit 1
Vlslt2
An alyte (hours) Mean
Triglycerides* Fasting 3.5 9.0 Retinyl palmitatet Fasting 3.5 9.0
Standard deviation
Mean
Standard deviation
118 237 113
42.4 83.1 56.3
109 242 103
31.1 86.0 46.9
Vol 136, No. 5 Printed in U.S.A.
Postprandial Lipemia: Reliability in an Epidemiologic Field Study
Ten subjects from the Forsyth County, North Carolina, and Washington County, Maryland, field centers in the Atherosclerosis Risk in Communities Study had two fat tolerance tests within a 10-day period from September 1988 to February 1989 to determine the reproducibility of markers for postprandial lipemia. No significant differences between visits were found in fasting mean plasma lipids, lipoproteins, and apolipoproteins. Postprandial triglycerides and retinyl palmitate were measured at 3.5 and 9.0 hours after the test meal in whole plasma. There were no significant differences in the mean levels of these analytes between visits. The correlation of triglycerides between repeat visits at 9.0 hours (r = 0.87) was stronger than in fasting samples {r = 0.67) or at 3.5 hours (r = 0.69). The mean plasma retinyl palmitate level at 3.5 hours was 15% higher than at the 9.0-hour level. The correlation of repeat measures of retinyl palmitate at 9.0 hours (r = 0.94) was much stronger than at 3.5 hours (r = 0.79). In conclusion, estimates of reliability in postprandial measurements of 9.0-hour triglycerides and retinyl palmitate levels were as strong as fasting lipid measurements of total cholesterol, high density lipoprotein cholesterol, low density lipoprotein cholesterol, and high density lipoprotein3 cholesterol, and both postprandial triglyceride measurements exceeded that of fasting triglyceride (r = 0.67). Am J Epidemiol 1992;136:538-45. lipoproteins, HDL cholesterol; lipoproteins, LDL cholesterol; triglycerides
used to control for either total cholesterol or low density lipoprotein cholesterol (LDL cholesterol), this association between triglycerides and coronary heart disease remained (6-9). However, prospective studies have shown that this association is not maintained upon correction for high density lipoprotein cholesterol (HDL cholesterol) because of a strong inverse relation of fasting triglyceride levels with HDL cholesterol levels (9-12). Although HDL cholesterol is a more powerful indicator of coronary heart disease risk than are triglycerides, other evidence indicates that the metabolism of plasma triglycerides influences HDL cholesterol levels (13). A relation between postprandial lipemia and coronary heart disease was extensively demonstrated as early as 1950 (14) and has subsequently been confirmed (15-21). In these studies, postprandial triglyceride levels and other markers of triglyceride-rich lipo-
Plasma triglyceride levels have been shown to be associated with the risk of coronary heart disease (1). In a number of studies, a univariate association between triglycerides and coronary heart disease was found (2-5). When multivariate analysis was
Received for publication November 8,1991, and in final form April 2, 1992. Abbreviations: ARC, Atherosclerosis Risk in Communities; HDL cholesterol, high density lipoprotein cholesterol, LDL cholesterol, low density lipoprotein cholesterol. ' Department of Medicine, Baylor College of Medicine, and Trie Methodist Hospital, Houston, TX. 2 Department of Biostatistics, University of North Carolina, Chapel Hill, NC. 3 Epidemiology and Biometry Program, National Heart, Lung, and Blood Institute, Bethesda, MD. Reprint requests to Dr. Spencer A Brown, The Methodist Hospital, Mail Stop A-601, 6565 Fannin Avenue, Houston, TX 77030. This work was supported in part by Atherosclerosis Risk in Communities (ARIC) Study grant HC NO1-HC55016 and grant 27341 from the National Heart, Lung, and Blood Institute.
538
Downloaded from https://academic.oup.com/aje/article-abstract/136/5/538/98759 by Insead user on 11 October 2018
Spencer A. Brown,1 Uoyd E. ChamWess,2 A. Richey Sharrett,3 Antonio M. Gotto, Jr.,1 and Wolfgang Patsch1
Reliability in an Epidemiologic Field Study
MATERIALS AND METHODS Subjects
Ten volunteer subjects, seven women and three men aged 45-64 years, were selected by using inclusion criteria identical to that of the ARIC Study (24). Subjects were selected without regard to lipid levels and were thought to be broadly representative of the residents of their communities. There were five subjects from each of two field centers, Forsyth County, North Carolina, and Washington County, Maryland. Subjects were scheduled for two visits to their respective field center within a 10-day time period. Fat tolerance test
On the day of the fat tolerance test, subjects arrived at the field centers after fasting for 12 hours. A fasting blood specimen was taken, with blood collected into ethylenediaminetetraacetic acid (EDTA)-containing tubes (1.5 mg/ml of blood). The test meal consisted of 1,374 kcal, 10.6 g of protein, 49.5 g of carbohydrate, 481 mg of cholesterol, 130.3 of g fat, and 100,000 IU of Aquasol A drops, Vitamin A (Armour Pharmaceutical Co., Kankakee, Illinois) (25). The test meal was administered per 2 square meters of body surface area. The volume of the entire test meal, 377 ml, was reduced in proportion to the participant's body surface area, which was determined from weight and height measurements. After the ingestion of the test meal, the participant was instructed not to take anything by mouth except water, sugarless drinks, or coffee for a period of 9.0 hours. Two postprandial blood specimens were collected, one at 3.5 hours and one at 9.0 hours after the ingestion of the test meal. Plasma was separated by centrifugation (1,500 x g, 20 minutes at 4°C) and stored at 4°C for 1-3 days. All plasma specimens were shipped on crushed ice to the laboratory by an overnight carrier. All laboratory measurements were performed within a week in the Atherosclerosis Clinical Laboratory of Baylor College of Medicine, Houston, Texas.
Downloaded from https://academic.oup.com/aje/article-abstract/136/5/538/98759 by Insead user on 11 October 2018
protein metabolism were higher in patients with clinically documented atherosclerosis compared with controls. However, in none of these early studies was HDL cholesterol taken into account. Two recent studies suggest that postprandial triglycerides are at least as good as HDL cholesterol for discriminating cases from controls as verified by angiography (22, 23). To examine a possible causal relation between postprandial triglycerides and the risk of atherosclerosis, two experimental designs can be used. One design utilizes prospective epidemiologic studies. A second design uses a cross-sectional study in which a causal relation may be demonstrated in asymptomatic individuals with documented arterial disease. In such a design, confounding effects introduced by medical intervention or resulting from potential effects of myocardial compromise on triglyceride metabolism may be minimized. The purpose of a recent National Heart, Lung, and Blood Institute research initiative was to determine whether there is a possible causal relation between postprandial lipemia and atherosclerosis. One of two populations to be investigated was selected from the Atherosclerosis Risk in Communities (ARIC) Study (24). The postprandial protocol of this study has taken advantage of the results from two recent clinical trials in which postprandial triglycerides and other markers of triglyceride-rich lipoprotein measured 6-12 hours after the fat tolerance test meal were found to be most discriminatory for angiographically assessed coronary heart disease (22, 23). Since ARIC is a multicenter study, the level of complexity, i.e., preparing of test meals and processing and transporting specimens, is increased, and standardization of postprandial protocols is required. Therefore, a pilot study was conducted to determine the reproducibility of plasma lipid markers during the postprandial state in a multicenter setting. We report here that the postprandial measurements were more reproducible than fasting triglyceride determinations. Thus, reliable postprandial studies can be implemented in epidemiologic studies.
539
540
Brown et al.
Laboratory measurements
Statistical analysis
Mean differences of laboratory measures between visits were compared by a paired / test (35). In order to assess the repeatability of the laboratory measures made on individuals at repeat visits, a one-factor nested random effects model was used to estimate between-person and combined withinperson plus method variances (36). From these components, the reliability coefficient R was estimated as the proportion of total variance in the between-person component. R is also interpretable as the correlation between measures on an individual made at visits approximately 1 week apart. An alternate estimate of this correlation is the Spearman rank correlation (35), which is reported along with scatter diagrams of the data. RESULTS
Mean fasting plasma lipid, lipoprotein, and apolipoprotein levels at each visit are shown in table 1. All subjects except one (180 mg/100 ml) had LDL cholesterol levels of less than 160 mg/100 ml, and triglyceride levels ranged from 64-187 mg/100 ml. All subjects except one (33 mg/100 ml) had HDL cholesterol levels of greater than 35
TABLE 1. Fasting lipids, lipoproteins, and apolipoproteins in two visits in 10 individuals in the Atherosclerosis Risk in Communities Study, September 1988 to February 1989 Vlsit2
Visit 1 Analyte
Cholesterol* Triglyceride HDL cholesterolt HDU cholesterol HDLj cholesterol LDL cholesterolt Apolipoprotein A-ljApolipoprotein B%
Mean
Standard deviation
Mean
Standard deviation
202.7 118.4 56.8 42.4 14.4 122.1 134.2 78.8
34.84 42.36 12.15 8.72 5.44 36.52 25.20 18.79
196.5 108.6 55.5 40.5 14.9 118.0 137.8 70.3
29.35 31.12 18.03 12.12 7.83 33.94 25.60 14.57
* All results are expressed as mg/100 ml. t HDL cholesterol, high density lipoprotein cholesterol; LDL cholesterol, low density Ipoproteln cholesterol. | n •=> 10 for all analytes except when noted, when n = 9
Downloaded from https://academic.oup.com/aje/article-abstract/136/5/538/98759 by Insead user on 11 October 2018
Total plasma cholesterol (26) and triglycerides (27) were measured enzymatically on a Cobas-Bio centrifugal analyzer (Roche Diagnostics, Montclair, New Jersey) with the respective enzymatic kits (catalog nos. 236691 and 701912; Boehringer Mannheim Diagnostics, Indianapolis, Indiana). HDL cholesterol was determined by measuring cholesterol in the supernate after precipitation of plasma with MgCl2 and dextran sulfate (28). Concentrations of HDL3 cholesterol were determined after reprecipitation of the total HDL supernate with different concentrations of MgCl2 and dextran sulfate (29, 30). HDL2 cholesterol concentration was calculated by subtracting the HDL3 cholesterol value from the value for total HDL cholesterol. LDL cholesterol was calculated according to the method of Friedewald et al. (31). Apolipoprotein A-I (32) and apolipoprotein B levels (33) were determined by radioimmunoassay methods. The coefficients of laboratory variation were calculated from both Centers for Disease Control serum pools and internal quality control plasma pools at multiple analyte levels. The laboratory coefficients of variation were 2.5 percent for cholesterol, 2.7 percent for triglycerides, 3.7 percent for HDL cholesterol, 5.2 percent for LDL cholesterol, 9.0 percent for apolipoprotein A-I, and 9.0 percent for apolipoprotein B. Retinyl palmitate levels were measured by high pressure liquid chro-
matography (SP8000; Spectra-Physics, San Jose, California) using a C-18 reverse phase column (3.9 x 300 mm uBondapak C18; Waters Instrument Co., Milford, Massachusetts) (34).
Reliability in an Epidemiologic Field Study
retinyl palmitate levels. The range of levels at this time point increased dramatically to 19.5-fold (range, 198-4,060 fig/100 ml), and the correlation coefficient for retinyl palmitate exceeded that of triglyceride at the late time point. Estimates of reliability coefficients in fasting and postprandial measurements are shown in table 3. As expected, fasting total plasma cholesterol, HDL cholesterol, HDL3 cholesterol, and LDL cholesterol measurements were highly reproducible between the two visits. The apolipoproteins were somewhat less reliable, and triglycerides had the lowest reliability. During the postprandial state, the reliability of postprandial triglyceride measurements was increased at 9.0 hours compared with the 3.5-hour measurements. Retinyl palmitate levels also increased in reliability from the 3.5-hour to the 9.0-hour measurements. In comparison, the reliability of 9.0-hour postprandial triglycerides and retinyl palmitate levels was similar to that of fasting total cholesterol, HDL cholesterol, LDL cholesterol, and HDL3 cholesterol measurements. DISCUSSION
A large number of studies have shown an association between high triglyceride levels and coronary heart disease, even when adjustment for other markers of plasma lipid transport is performed (1). Previous studies, including those with angiographically assessed subjects, suggest that postprandial
TABLE 2. Fasting and postprandial plasma retinyl palmitate and triglyceride levels in two visits In 10 individuals In the Atherosclerosis Risk in Communities Study, September 1988 to February 1989. Visit 1
Vlslt2
An alyte (hours) Mean
Triglycerides* Fasting 3.5 9.0 Retinyl palmitatet Fasting 3.5 9.0
Standard deviation
Mean
Standard deviation
118 237 113
42.4 83.1 56.3
109 242 103
31.1 86.0 46.9
