Shaomei Yu-Poth, Guixiang Zhao, Terry Etherton, Mary Naglak, Satya Jonnalagadda, and Penny M Kris-Etherton. See corresponding editorial on page 581.

Transcription

1 Effects of the National Cholesterol Education Program s Step I and Step II dietary intervention programs on cardiovascular disease risk factors: a meta-analysis 1,2 Shaomei Yu-Poth, Guixiang Zhao, Terry Etherton, Mary Naglak, Satya Jonnalagadda, and Penny M Kris-Etherton ABSTRACT Background: Plasma lipid and lipoprotein responses have been variable in dietary intervention studies. Objective: The objective of this study was to evaluate the effects of the National Cholesterol Education Program s Step I and Step II dietary interventions on major cardiovascular disease risk factors using meta-analysis. Design: MEDLINE was used to select 37 dietary intervention studies in free-living subjects published from 1981 to1997. Results: Step I and Step II dietary interventions significantly decreased plasma lipids and lipoproteins. Plasma total cholesterol (TC), LDL cholesterol, triacylglycerol, and TC:HDL cholesterol decreased by 0.63 mmol/l (10%), 0.49 mmol/l (12%), 0.17 mmol/l (8%), and 0.50 (10%), respectively, in Step I intervention studies, and by 0.81 mmol/l (13%), 0.65 mmol/l (16%), 0.19 mmol/l (8%), and 0.34 (7%), respectively, in Step II intervention studies (P < 0.01 for all). HDL cholesterol decreased by 7% (P = 0.05) in response to Step II but not to Step I dietary interventions. Positive correlations between changes in dietary total and saturated fatty acids and changes in TC and LDL and HDL cholesterol were observed (r = 0.59, 0.61, and 0.46, respectively; P < 0.001). Multiple regression analyses showed that for every 1% decrease in energy consumed as dietary saturated fatty acid, TC decreased by mmol/l and LDL cholesterol by 0.05 mmol/l. Moreover, for every 1-kg decrease in body weight, triacylglycerol decreased by mmol/l and HDL cholesterol increased by mmol/l. Exercise resulted in greater decreases in TC, LDL cholesterol, and triacylglycerol and prevented the decrease in HDL cholesterol associated with low-fat diets. Conclusion: Step I and Step II dietary interventions have multiple beneficial effects on important cardiovascular disease risk factors. Am J Clin Nutr 1999;69: KEY WORDS National Cholesterol Education Program, NCEP Step I diet, NCEP Step II diet, total cholesterol, LDL cholesterol, HDL cholesterol, triacylglycerol, body weight, risk factors, cardiovascular disease, exercise, meta-analysis, humans INTRODUCTION Diet is the first line of therapy for the management of plasma lipids in the prevention and treatment of cardiovascular disease See corresponding editorial on page 581. (CVD). The goal of dietary therapy is to reduce elevated total cholesterol and LDL-cholesterol concentrations and thereby reduce CVD morbidity and mortality. The National Cholesterol Education Program (NCEP) recommends that dietary therapy be implemented in a stepwise manner. Step I and Step II diets are designed to progressively reduce dietary saturated fatty acids (SFA) and cholesterol and to promote weight loss, if indicated, through diet and exercise. In addition, individuals are encouraged to adopt a healthy lifestyle that includes regular physical activity. Controlled feeding studies have consistently found that a reduction in dietary SFA decreases plasma total and LDL-cholesterol concentrations. In general, a Step I diet decreases plasma total cholesterol and LDL cholesterol by 7 9% compared with the average American diet. A Step II diet has been shown to decrease total cholesterol and LDL cholesterol by 10 20% (1, 2). In controlled feeding studies in which body weight was maintained, low-fat diets often were associated with decreases in HDL cholesterol and increases in triacylglycerol (2 4). A low HDL-cholesterol concentration and an elevated triacylglycerol concentration are both risk factors for CVD (4). In contrast with these potentially adverse effects of low-fat diets on HDL-cholesterol and triacylglycerol concentrations, which have been reported in well-controlled clinical feeding studies, a body of evidence from dietary intervention studies conducted in free-living populations has shown that low-fat diets are typically accompanied by weight loss, and often other risk-factor modifications result in a decrease in plasma total cholesterol, LDL cholesterol, and triacylglycerol, and no change in HDL cholesterol (5 15). Likewise, many free-living populations worldwide consume very-low-fat diets and have a favorable lipid profile, which likely is due to their lifestyle practices, including regular physical activity and maintenance of an ideal body weight (16, 17). 1 From the Graduate Program in Nutrition, The Pennsylvania State University, University Park. 2 Address reprint requests to PM Kris-Etherton, S-126 Henderson Building, Nutrition Department, University Park, PA pmk3@psu.edu. Received May 21, Accepted for publication November 3, Am J Clin Nutr 1999;69: Printed in USA American Society for Clinical Nutrition

2 DIETARY INTERVENTION AND CVD RISK 633 Many primary and secondary intervention studies have evaluated how different intervention strategies to reduce the risk of CVD, including diet modification, affect various CVD risk factors in free-living subjects. In general, the responses in these intervention studies have been quite variable. For example, some studies have shown that dietary intervention and other risk-factor modifications often accompanied by weight loss reduce plasma total cholesterol, LDL cholesterol, as well as triacylglycerol, but increase or have no significant effects on HDL cholesterol (5 15). Other intervention studies, however, found that low-fat diets resulted in an increase in plasma triacylglycerol and a decrease in HDL cholesterol (18 22). Thus, the purpose of the present study was to evaluate the effects of different dietary interventions on major CVD risk factors in healthy and high-risk subjects by conducting a meta-analysis. METHODS Selection of studies MEDLINE (National Library of Medicine, Bethesda, MD) and the references in the papers we identified were used to search all published dietary intervention studies related to cholesterol lowering or reduction of other CVD risk factors in freeliving subjects. Thirty-seven (5 15, 18 43) intervention studies published between 1981 and 1997 were selected for the present meta-analysis. The following criteria were used for inclusion of a dietary intervention trial: 1) the study was designed to lower blood cholesterol concentrations or to decrease body weight for the primary purpose of preventing CVD; 2) the investigators used a randomized design; 3) a Step I diet (in all intervention groups: 30% of total energy as fat, 10% of energy as SFA, and 300 mg dietary cholesterol/d), a Step II diet ( 7% of energy as SFA and 200 mg dietary cholesterol/d), or both were part of the dietary intervention; 4) the subjects were free-living, prepared their own food, and were counseled by dietitians or other professionals about implementing low-fat diets; and 5) the intervention lasted 3 wk to stabilize plasma cholesterol concentrations. Statistical analysis Changes in plasma total cholesterol, LDL cholesterol, HDL cholesterol, and triacylglycerol after Step I and Step II dietary interventions were assessed. Effects of exercise and body weight were evaluated. In addition, effects of baseline plasma total cholesterol, LDL-cholesterol, HDL-cholesterol, and triacylglycerol concentrations on lipid responses were also analyzed. We also examined the relation between changes in body weight and changes in dietary fat and energy consumption. All analyses were done by using the SAS statistical package (44). In each study, plasma lipid concentrations after dietary intervention were compared with lipid concentrations in the control groups as well as with baseline lipid concentrations. Changes in plasma lipid concentrations and in dietary fat or cholesterol were calculated by using the difference between a treatment group and a control group or differences between intervention and baseline values in the intervention groups. Analysis of variance was used to compare the effects of Step I with those of Step II dietary interventions and the effects of interventions including exercise with those not including exercise. Correlations between changes in plasma lipid concentrations (both absolute and percentage changes) and changes in total fat and SFA intakes (as a percentage of total daily energy intake) and changes in dietary cholesterol (mg/d) and changes in body weight (kg) were evaluated by Pearson correlation analysis. Changes in plasma total cholesterol, LDL cholesterol, HDL cholesterol, and triacylglycerol in response to changes in body weight and in dietary total fat, SFA, and cholesterol were evaluated by regression analysis. In each study, the differences in plasma lipid concentrations between intervention and control groups (or between baseline and intervention values) were used as dependent variables and the differences in dietary total fat, SFA, and cholesterol as independent variables. Changes in body weight in the intervention groups were used as a covariable in the regression analysis. Both bivariate and multiple regression analyses were conducted. Bivariate regression analysis included the change ( )in 1 independent variable ( TF, SFA, or cholesterol) and 1 covariable ( BW); multiple regression analysis included the change in 2 independent variables ( TF and cholesterol or SFA and cholesterol) and 1 covariable ( BW). The equations are as follows: Total cholesterol ( LDL cholesterol, HDL cholesterol, or triacylglycerol) = 1 TF ( SFA, or cholesterol) + 2 BW for bivariate regression analysis (Model 1) Total cholesterol ( LDL cholesterol, HDL cholesterol, or triacylglycerol) = 1 TF ( SFA) + 2 cholesterol + 3 BW for multiple regression analysis (Model 2) where BW is body weight and TF is total fat. The coefficients ( 1, 2, and 3 ) were estimated by least-squares regression. Changes in body weight in response to changes in dietary total fat intake were tested by regression analysis and Pearson correlation analysis. In the regression analysis, the change in body weight after intervention was used as a dependent variable and the change in total fat intake was used as an independent variable. The regression equation is as follows: BW = 1 TF (1) Correlation between change in body weight and change in fat was evaluated using the Pearson correlation analysis conducted with and without using subject number as a weight factor from each study. RESULTS The present meta-analysis included 37 intervention studies (5 15, 18 43) in which there were 9276 subjects in intervention groups and 2310 subjects in control groups. The study designs varied remarkably; some were sequential studies but most were randomized, parallel-arm studies. The dietary interventions ranged from vegetarian diets providing <10% of energy as fat, <6% of energy as SFA, and <100 mg cholesterol/d to a Step I diet providing 30% of energy as fat, <10% of energy as SFA, and 300 mg cholesterol/d. The diet compositions and study designs of the interventions are summarized in Table 1; 8 studies evaluated the effects of diet on body weight. Despite the differences in experimental design and populations studied, total cholesterol decreased by 2 25% as a result of dietary and other risk-factor interventions. Lipid concentration data from 30 studies before and after intervention are summarized in Table 2.

3 634 YU-POTH ET AL TABLE 1 Summary of study designs and diet composition in 37 intervention studies 1 Duration Baseline diet Experimental diet Weight Reference and group of study Energy TF SFA Cholesterol Energy TF SFA Cholesterol Exercise change kj % of % of mg/d kj % of % of mg/d kg energy energy energy energy Step I intervention Hjermann et al (5) I (n = 604M) 4 y ND ND ND ND No 3.6 C (n = 628M) 4 y ND ND ND ND No 0.6 Nikolaus et al (6) I (n = 18M) 1 y Yes 4.5 C (n = 27M) 1 y No 0.6 Wood et al (7) I (n = 31F) 1 y No 4.1 I (n = 42F) 1 y Yes 5.1 C (n = 39F) 1 y No 1.3 I (n = 40M) 1 y No 5.1 I (n = 39M) 1 y Yes 8.7 C (n = 40M) 1 y No 1.7 Schuler et al (8) I (n = 56M) 1 y ND ND 135 Yes 4.8 C (n = 57M) 1 y ND ND 232 No 0.6 Singh et al (9) I (n = 204M+F) 1 y No 7.6 C (n = 202M+F) 1 y No 1.1 Singh et al (10) I (n = 231M) 12 wk ND ND ND ND Yes 3.7 C (n = 232M) 12 wk ND ND ND ND No 1.5 Singh et al (11) I (n = 310M+F) 16 wk No 1.7 I (n = 310M+F) 24 wk Yes 3.6 C (n = 311M+F) 16 wk No 1.1 C (n = 311M+F) 24 wk No 1.4 Baer (12) I (n = 31M) 1 y ND ND 304 No 5.0 C (n = 33M) 1 y ND ND 430 No 1.0 Katzel et al (13) I (n = 14M) 3 mo ND ND 220 No 1.0 I (n = 14M) 12 mo ND ND 190 Yes 11.0 McCarron et al (14) I (n = 277M+F) 10 wk No 3.2 Ehnholm et al (18) I (n = 30M) 6 wk No 1.1 I (n = 24F) 6 wk No 1.0 Kuusi et al (19) I (n = 19M) 6 wk No ND I (n = 19M) 12 wk No 1.5 C (n = 19M) 6 wk No ND C (n = 19M) 12 wk No 1.6 I (n = 20F) 6 wk No ND I (n = 20F) 12 wk No 1.5 C (n = 20F) 6 wk No ND C (n = 20F) 12 wk No 1.6 de Lorgeril et al (20) I (n = 219M+F) 2 y No 1.4 C (n = 192M+F) 2 y No 2.3 Knopp et al (21) I (n = 78M) 1 y No 3.0 I (n = 62M) 1 y No 3.0 I (n = 71M) 1 y No 2.0 I (n = 59M) 1 y No 2.0 I (n = 57M) 1 y No 2.0 I (n = 55M) 1 y No 2.0 I (n = 62M) 1 y No 6.0 (Continued)

4 DIETARY INTERVENTION AND CVD RISK 635 TABLE 1 (Continued) Duration Baseline diet Experimental diet Weight Reference and group of study Energy TF SFA Cholesterol Energy TF SFA Cholesterol Exercise change kj % of % of mg/d kj % of % of mg/d kg energy energy energy energy Step I intervention cont. Ehnholm et al (23) I (n = 38M+F) 6 wk ND ND No 0.7 C (n = 36M+F) 6 wk ND ND No ND Boyd et al (24) I (n = 100F) 1 y No 1.0 C (n = 106F) 1 y No 0.1 Denke and Grundy (25) I (n = 44M) 4 mo ND No 0.9 C (n = 39M) 1 mo ND No 0.1 Geil et al (26) I (n = 63F) 8 wk ND ND ND ND No 0.83 I (n = 99M) 8 wk ND ND ND ND No 0.89 Dengel et al (27) I (n = 28M) 3 mo No 1.0 I (n = 14M) 3 mo No 1.0 I (n = 28M) 9 mo No 11.0 I (n = 14M) 9 mo No 0 Davidson et al (28) I (n = 44M+F) 4 wk No 1.5 I (n = 44M+F) 8 wk No 1.8 Bae et al (29) I (n = 87M+F) 6 wk No 0.6 I (n = 87M+F) 12 wk No 1.1 I (n = 87M+F) 18 wk No 2.4 Step II intervention McCarron et al (14) I (n = 282M+F) 10 wk No 4.6 Haskell et al (15) I (n = 118M+F) 4 y Yes 3.0 C (n = 127M + F) 4 y No 0.9 Kasim et al (22) I (n = 34F) 1 y No 3.4 C (n = 38F) 1 y No 0.8 Arntzenius et al (30) I (n = 39M+F) 2 y 8376 ND ND 7 59 No 1.2 Ornish et al (31) I (n = 22M+F) 1 y ND ND 12 Yes 10.1 C (n = 19M+F) 1 y ND ND 190 No 1.4 Barnard (32) I (n = 4587M+F) 3 wk ND AAD ND ND ND <10 ND <25 Yes 4.3 Barnard et al (33) I (n = 72M+F) 3 wk ND AAD ND ND ND <10 ND <25 Yes 4.0 Seim et al (34) I (n = 41M+F) 6 wk ND 36 ND ND ND ND Yes 4.7 Walden et al (35) I (n = 84F) 6 mo No 1.6 I (n = 94F) 6 mo No 2.2 I (n = 123M) 6 mo No 3.4 I (n = 108M) 6 mo No 2.9 Diet + weight-loss intervention 2 Fox et al (36) I (n = 16F) 24 wk ND 43 ND ND ND ND Yes 7.1 I (n = 13F) 24 wk ND 43 ND ND ND ND No 6.6 I (n = 12F) 24 wk ND 41 ND ND ND ND No 5.8 Sheppard et al (37) I (n = 171F) 6 mo ND ND ND ND No 3.2 I (n = 171F) 1 y ND ND ND ND No 3.0 I (n = 158F) 2 y ND ND ND ND No 1.9 C (n = 105F) 6 mo ND ND ND ND No 0.4 (Continued)

5 636 YU-POTH ET AL TABLE 1 (Continued) Duration Baseline diet Experimental diet Weight Reference and group of study Energy TF SFA Cholesterol Energy TF SFA Cholesterol Exercise change kj % of % of mg/d kj % of % of mg/d kg energy energy energy energy Diet + weight-loss intervention cont. 2 C (n = 105F) 1 y ND ND ND ND No 0.4 C (n = 105F) 2 y ND ND ND ND No 0.1 Tremblay et al (38) I (n = 4F) 14 mo ND ND ND ND Yes 4.6 Schlundt et al (39) I (n = 25M+F) wk ND ND ND ND No 4.6 I (n = 24M+F) wk ND ND ND ND No 8.3 Shah et al (40) I (n = 47F) 6 mo ND ND ND ND No 4.4 I (n = 42F) 6 mo ND ND ND ND No 3.8 Siggaard et al (41) I (n = 50M+F) 12 wk ND ND ND ND No 4.2 C (n = 16M+F) 12 wk ND ND ND ND No 0.8 Jeffery et al (42) I (n = 39F) 6 mo ND ND ND ND No 4.6 I (n = 35F) 6 mo ND ND ND ND No 2.1 I (n = 39F) 12 mo ND ND ND ND No 3.7 I (n = 35F) 12 mo ND ND ND ND No 0.5 Raben et al (43) I (n = 24M+F) 11 wk ND ND ND ND No 1.3 C (n = 24M+F) 11 wk ND ND ND ND No 0 1 I, intervention diet group; C, control group (consumed habitual diet); ND, no data; TF, total fat; SFA, saturated fatty acids. 2 Dietary SFA were not reported in these studies. 3 Change in total energy. Twenty-one intervention studies included both men and women, 9 studies included only men, and 7 studies included only women. Nineteen studies included a control group in which subjects maintained their habitual lifestyle and food consumption throughout the study. Dietary information was estimated by using either a 24-h food recall or 3 7-d food records; a food-frequency questionnaire was also used in some studies. Some studies did not report complete dietary information. For example, 3 Step I (5, 10, 26) and 2 Step II (32, 33) dietary interventions did not report baseline energy, total fat, SFA, and cholesterol intakes (Table 1). The length of intervention ranged from 3 wk to 4 y. Intervention intensity was moderate to high and 13 studies included an exercise intervention. Mean baseline total cholesterol and LDL-cholesterol concentrations were between 4.84 and 6.88 mmol/l (x ± SE: 6.04 ± 0.53 mmol/l) and 3.05 and 4.55 mmol/l (4.01 ± 0.46 mmol/l), respectively, except in the study of Hjermann et al (5) in which subjects had higher baseline concentrations of total cholesterol (8.47 mmol/l) and LDL cholesterol (6.78 mmol/l). Mean baseline HDL-cholesterol concentrations were between 0.72 and 1.72 mmol/l (1.24 ± 0.23 mmol/l) and triacylglycerol concentrations were between 0.85 and 2.51 mmol/l (1.67 ± 0.46 mmol/l). Most studies had more than one endpoint blood collection. Some studies (24, 25, 30, 34) did not have complete plasma lipid data. For example, 2 studies (24, 30) did not report plasma LDL-cholesterol, HDL-cholesterol, and triacylglycerol concentrations and 1 study (25) did not report baseline plasma lipid concentrations (Table 2). A total of 59 dietary intervention groups yielded 59 data points for the regression and correlation analyses. Comparison of the effects of Step I and Step II dietary interventions on plasma lipids Plasma total cholesterol, LDL cholesterol, HDL cholesterol, triacylglycerol, and total cholesterol:hdl cholesterol all decreased after both Step I and Step II dietary interventions, by 0.63 ± 0.06 mmol/l (10%), 0.49 ± 0.05 mmol/l (12%), 0.04 ± 0.02 mmol/l (1.5%), 0.17 ± 0.04 mmol/l (8%), and 0.50 ± 0.11 (10%), respectively, after the Step I dietary interventions (P < 0.01 for all, except for HDL cholesterol ) and by 0.81 ± 0.12 mmol/l (13%), 0.65 ± 0.09 mmol/l (16%), 0.09 ± 0.03 mmol/l (7%), 0.19 ± 0.14 mmol/l (8%), and 0.34 ± 0.12 (7%), respectively, after the Step II dietary intervention studies (P < 0.01 for all). The Step II dietary intervention resulted in greater decreases in plasma total cholesterol (P < 0.05), LDL cholesterol (P < 0.05), HDL cholesterol (P = 0.13), triacylglycerol (P = 0.37), and total cholesterol:hdl cholesterol (data not shown) than did the Step I dietary intervention (Figure 1). When analyses were weighted by the number of subjects in each study, plasma total cholesterol, LDL-cholesterol, HDL-cholesterol, and triacylglycerol concentrations decreased by 21%, 21%, 13%, and 33%, respectively, after Step II dietary interventions. Plasma total cholesterol, LDL-cholesterol, and triacylglycerol concentrations decreased by 8%, 8%, and 10%, respectively, after Step I dietary interventions; HDL cholesterol increased by 2%. Decreases in plasma lipids and lipoproteins were much greater after the Step II dietary interventions than after Step I dietary interventions (P < 0.001).

6 DIETARY INTERVENTION AND CVD RISK 637 FIGURE 1. Changes ( ) in plasma lipids and lipoproteins after National Cholesterol Education Program Step I and Step II dietary interventions. * Significantly different from Step I, P < TC, total cholesterol; TG, triacylglycerol. Interestingly, plasma lipid and lipoprotein responses of males and females were comparable after both Step I and Step II dietary interventions (data not shown), with one notable exception. The decrease in HDL cholesterol was greater in women (0.10 mmol/l, 6.8%) than in men (0.03 mmol/l, 2.2%) (P < 0.05) after the Step II intervention. In addition, triacylglycerol concentrations tended to increase in women by 0.01 mmol/l (2.4%) and 0.07 mmol/l (5.4%) and decrease in men by 0.21 mmol/l (10.4%) and 0.03 mmol/l (1.5%), respectively, after Step I and Step II dietary interventions. In addition, most lipid responses were comparable after interventions lasting <6 mo and those after interventions lasting >6 mo (data not shown). The only exception was that HDL-cholesterol FIGURE 2. Correlation between change ( ) in dietary total fat and change in plasma lipids and lipoproteins. Pearson correlation coefficients are significant for plasma total cholesterol (TC, ; r = 0.61, P < ), LDL cholesterol ( ; r = 0.63, P < ), and HDL cholesterol ( ; r = 0.41, P < 0.001), but not for triacylglycerol ( ; r = 0.19, P = 0.47). concentrations decreased by 0.09 mmol/l (6.4%) and by 0.13 mmol/l (9.7%), respectively, after Step I and Step II interventions lasting <6 mo and increased by 0.03 mmol/l (4.7%) after Step I interventions and decreased by only 0.01 mmol/l (0.5%) after Step II interventions lasting >6 mo (P < 0.05). Effects of dietary fat and SFAs and body weight on plasma lipids Changes in plasma total cholesterol, LDL cholesterol, and HDL cholesterol were significantly correlated (by Pearson correlation analyses) with changes in dietary total fat (Figure 2) and SFA (Figu re 3). The correlation between the change in plasma triacylglycerol and the change in dietary fat and SFA was not significant. Changes in total cholesterol, LDL cholesterol, HDL cholesterol, and triacylglycerol (both absolute and percentage changes) were significantly correlated with changes in dietary cholesterol (Table 3). When Pearson correlation analyses were weighted by the number of subjects in each study, all correlation coefficients were significant. The regression coefficients of dietary fat, SFA, and cholesterol were significant for changes in total cholesterol, LDL cholesterol, HDL cholesterol, and triacylglycerol by bivariate regression analysis with the change in body weight as a covariable (Table 4). For example, every 1% decrease in energy from dietary total fat decreased total cholesterol by 0.06 mmol/l (0.9%), LDL cholesterol by mmol/l (1.04%), and HDL cholesterol by 0.01 mmol/l (0.79%). The regression analysis showed that changes in body weight significantly affected plasma HDL-cholesterol and triacylglycerol concentrations. For example, with dietary total fat as a variable and body weight as a covariable, every 1-kg increase in body weight increased plasma triacylglycerol by mmol/l (1.14%) and decreased HDL cholesterol by 0.01 mmol/l (0.83%). The regression coefficients of body weight for changes in total cholesterol and LDL cholesterol were not significant (data not shown). Multiple regression analyses have shown that dietary total fat and SFA had significant effects on plasma total cholesterol, LDL cholesterol, and HDL cholesterol (Table 5). For example, every 1% decrease in energy from dietary SFA resulted in a decrease in

7 638 YU-POTH ET AL FIGURE 3. Correlation between change ( ) in dietary saturated fatty acids and change in plasma lipids and lipoproteins. Pearson correlation coefficients are significant for plasma total cholesterol ( ; r = 0.70, P < ), LDL cholesterol ( ; r = 0.70, P < ), and HDL cholesterol ( ; r = 0.41, P < 0.001), but not for triacylglycerol ( ; r = 0.36, P = 0.06). total cholesterol by mmol/l (0.77%), LDL cholesterol by 0.05 mmol/l (1.07%), and HDL cholesterol by mmol/l (0.6%). Dietary cholesterol had significant effects on total cholesterol, LDL cholesterol, and possibly triacylglycerol, but not on HDL cholesterol. The regression coefficients of dietary total fat and SFA for triacylglycerol were not significant. Body weight change was shown again to have significant effects on HDL cholesterol and triacylglycerol. With every 1-kg decrease in body weight, plasma triacylglycerol concentrations decreased by mmol/l ( %), whereas HDL-cholesterol concentrations increased by mmol/l ( 1%) (Table 5). The regression coefficients for body weight change and changes in total cholesterol and LDL cholesterol were not significant (data not shown). Pearson correlation analyses showed that body weight change was positively correlated with changes in plasma triacylglycerol concentrations (r = 0.35, P < 0.01) and negatively correlated with HDL-cholesterol concentrations (r = 0.38, P < 0.02) in Step I and Step II dietary intervention studies. The correlations between body weight change and changes in total cholesterol and LDL cholesterol were significant only when the analyses were weighted by the number of subjects in each study: r = 0.49 (P = 0.001) and r = 0.49 (P = 0.002), respectively. Effects of exercise on plasma lipids In the present study, we included 14 intervention groups with exercise and 45 intervention groups without exercise. Analysis of variance showed that exercise had significant effects on plasma lipids and lipoproteins. Plasma total cholesterol, LDL-cholesterol, HDL-cholesterol, and triacylglycerol concentrations decreased by 0.60 ± 0.06, 0.47 ± 0.05, 0.06 ± 0.02, and 0.11 ± 0.04 mmol/l, respectively, in intervention groups without exercise and decreased by 0.78± 0.13, 0.56 ± 0.12, 0.01 ± 0.04, and 0.35 ± 0.12 mmol/l in intervention groups with exercise (Figure 4). Exercise groups had a greater decrease than nonexercise groups in plasma total cholesterol (13% compared with 10%), LDL cholesterol (15% compared with 11%), and triacylglycerol (17% compared with 5.2%), but no significant change in HDL cholesterol was observed between exercise and nonexercise groups. When the analyses were weighted by a subject number from each study, the exercise groups had as much as a 3-fold greater decrease in total cholesterol (by 1.27 compared with 0.43 mmol/l, 21% compared with 7%) and LDL cholesterol (by 0.83 compared with 0.29 mmol/l, 21% compared with 7%), a 5-fold greater decrease in triacylglycerol (by 0.77 compared with 0.11 mmol/l, 33% compared with 6%), and a 10-fold smaller decrease in HDL-cholesterol concentrations (by compared with mmol/l, 0.02% compared with 5.1%) (P < for all comparisons) than the nonexercise groups. Correlation between baseline lipids and lipoproteins and responses to intervention Pearson correlation coefficients showed that the changes in LDL cholesterol, HDL cholesterol, and triacylglycerol were significantly correlated with the baseline concentrations of these lipids (Figure 5). The change in total cholesterol was not correlated with the baseline concentration (Figure 5). However, if baseline total cholesterol concentrations were <6.2 mmol/l, the change in total cholesterol was significantly correlated with baseline concentrations (r = 0.591, P < 0.001). In contrast, no relation was observed in subjects with an initial total cholesterol concentration >6.2 mmol/l (data not shown). A similar relation was observed for LDL cholesterol (data not shown). Thus, it appears that individuals with marked elevations in total cholesterol and LDL cholesterol were less responsive to dietary interventions than were mildly to moderately hypercholesterolemic individuals. Effects of dietary fat and energy intake and exercise on body weight Dietary fat had a significant effect on body weight. The change in body weight after intervention was highly correlated

8 DIETARY INTERVENTION AND CVD RISK 639 TABLE 2 Lipid concentrations before and after interventions in 30 studies 1 Baseline lipids Endpoint lipids Percentage change Reference and group TC LDL-C HDL-C TG TC LDL-C HDL-C TG TC LDL-C HDL-C TG mmol/l mmol/l % Step I intervention Hjermann et al (5) I C Nikolaus et al (6) I C Wood et al (7) I I C I I C Schuler et al (8) I C Singh et al (9) I C Singh et al (10) I C Singh et al (11) I I C C Baer (12) I C Katzel et al (13) I I McCarron et al (14) I Ehnholm et al (18) I I 5.74 ND ND ND Kuusi et al (19) I I C C I I C C de Lorgeril et al (20) I C Knopp et al (21) I I I (Continued)

9 640 YU-POTH ET AL TABLE 2 (Continued) Baseline lipids Endpoint lipids Percentage change Reference and group TC LDL-C HDL-C TG TC LDL-C HDL-C TG TC LDL-C HDL-C TG mmol/l mmol/l % I I I I Ehnholm et al (23) I C Boyd et al (24) I 4.83 ND ND ND 4.63 ND ND ND 4.1 ND ND ND C 4.88 ND ND ND 4.88 ND ND ND 0.00 ND ND ND Denke and Grundy (25) I ND ND ND ND ND ND ND ND C ND ND ND ND ND ND ND ND Geil et al (26) I I Dengel et al (27) I I I I Davidson et al (28) I I Bae et al (29) I I I Step II intervention McCarron et al (14) I Haskell et al (15) I C Kasim et al (22) I C Arntzenius et al (30) I 6.90 ND 1.01 ND 6.20 ND 0.98 ND 10.1 ND 3.0 ND Ornish et al (31) I C Barnard (32) I I Barnard et al (33) I Seim and Holtmeier (34) I ND ND ND Walden et al (35) I I I I I, intervention diet group; C, control group (consumed habitual diet); ND, no data; TC, total cholesterol; TG, triacylglycerol; LDL-C, LDL cholesterol; HDL-C, HDL cholesterol.

10 DIETARY INTERVENTION AND CVD RISK 641 FIGURE 4. Effects of exercise on plasma total cholesterol (TC), LDL cholesterol (LDL-C), HDL cholesterol (HDL-C), and triacylglycerol (TG). * Significantly different from no exercise, P < with the change in dietary total fat (Figure 6) as well as with the change in energy intake. The change in dietary fat was related to the change in energy intake. Regression analysis showed that the change in dietary fat had a significant effect on the change in body weight. BW = 0.28 TF (R 2 = 0.57, P < ) (4) The regression equation revealed that for every 1% decrease in energy as total fat, there was a 0.28-kg decrease in body weight. The effect of change in total fat on weight loss explained 57% of the total variance. Diet intervention with exercise resulted in significantly greater weight loss than diet intervention without exercise. Body weight decreased by 5.66 ± 0.77 kg in intervention groups with exercise and by 2.79 ± 0.31 kg in intervention groups without exercise (Figure 7). Furthermore, there was no significant difference in the change in dietary fat between intervention groups with and without exercise ( 11.6 ± 1.9% compared with 10.0 ± 0.7% of FIGURE 5. Correlation between changes ( ) in plasma total cholesterol (TC), LDL cholesterol (LDL-C), HDL cholesterol (HDL-C), and triacylglycerol (TG) and their baseline values. r = Pearson correlation coefficient.

11 642 YU-POTH ET AL TABLE 3 Correlation between changes in dietary total fat (TF), saturated fatty acids (SFA), and cholesterol and changes in plasma lipids and lipoproteins 1 total energy, P > 0.05). Thus, the effect of change in dietary fat on body weight was independent of the effect of exercise. DISCUSSION Short-term controlled-feeding studies have shown that Step I and Step II diets typically decrease total cholesterol and LDL cholesterol by 7 9% and 10 20%, respectively (1). In addition, these diets decrease HDL-cholesterol and increase triacylglycerol concentrations (2, 4, 45, 46). Many intervention studies conducted in free-living populations have not observed these potentially adverse effects on HDL cholesterol and triacylglycerol (5, 8 11, 14, 20, 21). Results of the present meta-analysis of dietary interventions in free living populations (n = >9000) showed that consumption of Step I and Step II diets, respectively, significantly (P < 0.001) decreased plasma total cholesterol (by 10% and 13%), LDL cholesterol (by 12% and 16%), and triacylglycerol (by 8% and 8%). Plasma HDL-cholesterol concentrations did not decrease significantly (by 1.5%) after Step I dietary intervention studies but did decrease significantly (by 7%) after Step II dietary interventions; total cholesterol:hdl cholesterol decreased significantly after both Step I and Step II dietary interventions. Average plasma lipid and Intervention versus control group TF SFA Cholesterol r P r P r P Absolute change (mmol/l) TC 0.59 (0.89) < (0.91) < (0.89) < LDL-C 0.61 (0.86) < (0.89) < (0.84) <0.001 HDL-C 0.46 (0.81) < (0.78) < (0.73) TG 0.15 (0.88) (0.89) (0.86) Percentage change (%) TC 0.57 (0.88) < (0.90) < (0.89) < LDL-C 0.61 (0.87) < (0.89) < (0.84) <0.001 HDL-C 0.43 (0.85) (0.82) (0.78) TG 0.07 (0.85) (0.88) (0.86) r, Pearson correlation coefficient; TC, total cholesterol; LDL-C, LDL cholesterol; HDL-C, HDL cholesterol; TG, triacylglycerol. Values in parentheses are weighted by the number of subjects in each study; P < for all values. lipoprotein responses were comparable with those observed in controlled feeding studies. However, there was an appreciable range in the response to the dietary interventions with the maximal effect being more than twice the average response reported in controlled feeding studies with Step I dietary interventions (total cholesterol concentrations decreased by 23.5% compared with 9%). Moreover, the maximal response of total cholesterol ( 24.4%) to the Step II dietary intervention programs implemented was marked. The marked hypocholesterolemic response to Step I and Step II dietary interventions in some free-living subjects likely reflected the effects of diet, weight loss, and exercise. The attenuated cholesterol lowering response in individuals with pronounced hypercholesterolemia raises important questions about the biological basis of hypercholesterolemia as well as what the ideal interventions should be. Thus, overall, Step I and Step II dietary interventions have beneficial effects on plasma lipid profiles in free-living populations, especially in individuals who are not markedly hypercholesterolemic. These findings support current dietary guidelines and recommendations to exercise regularly to reduce the risk of CVD. Although the results of the present meta-analysis clearly indicate the benefits of a low-fat diet on CVD risk factors, especially those of Step I dietary interventions (because they did not decrease HDL TABLE 4 Bivariate regression analysis: plasma lipids and lipoproteins in response to changes ( ) in dietary total fat (TF), saturated fatty acids (SFA), or cholesterol, with the change in body weight (BW) as a covariable 1 Coefficient P Coefficient P Coefficient P TF, BW TF, BW R 2 SFA, BW TF, BW R 2 Cholesterol, BW Cholesterol, BW R 2 Absolute change (mmol/l) TC (0.060, NS) (0.0001, NS) 0.77 (0.107, NS) (0.0001, NS) 0.84 (0.049, NS) (0.0001, NS) 0.84 LDL-C (0.042, NS) (0.0001, NS) 0.57 (0.078, NS) (0.0001, NS) 0.84 (0.033, NS) (0.0001, NS) 0.78 HDL-C (0.010, 0.010) (0.0001, 0.001) 0.50 (0.018, 0.009) (0.0001, 0.005) 0.50 (0.008, 0.011) (0.0001, 0.001) 0.46 TG (0.012, 0.018) (0.04, 0.05) 0.49 (0.027, 0.019) (0.001, 0.001) 0.60 (0.013, 0.011) (0.004, 0.24) 0.46 Percentage change (%) TC (0.90, NS) (0.0001, NS) 0.77 (1.73, NS) (0.0001, NS) 0.82 (0.79, NS) (0.0001, NS) 0.84 LDL-C (1.04, NS) (0.0001, NS) 0.78 (1.89, NS) (0.0001, NS) 0.82 (0.85, NS) (0.0001, NS) 0.78 HDL-C (0.79, 0.83) (0.0001, 0.001) 0.46 (1.36, 0.78) (0.0001, 0.005) 0.43 (0.60, 0.96) (0.0001, 0.001) 0.44 TG (0.50, 1.14) (0.07, 0.008) 0.45 (1.32, 1.22) (0.001, 0.001) 0.66 (0.70, 0.75) (0.002, 0.08) TF and SFA, change in percentage of energy per day from TF and SFA; BW, change in BW in kilograms; cholesterol, change in dietary cholesterol per day divided by 10; TC, total cholesterol; LDL-C, LDL cholesterol; HDL-C, HDL cholesterol; TG, triacylglycerol.

12 DIETARY INTERVENTION AND CVD RISK 643 FIGURE 6. Correlation between changes ( ) in body weight and changes in dietary total fat and energy intake and between changes in energy intake and dietary total fat. Pearson correlation coefficients (r) are significant for all correlations. FIGURE 7. Effects of diet and diet plus exercise interventions on weight loss. * Significantly different from diet alone, P < cholesterol ), there is compelling emerging data that a diet high in monounsaturated fatty acids but low in SFA and cholesterol may result in a more favorable lipid profile (ie, higher HDLcholesterol and lower triacylglycerol concentrations) (47). An alternative dietary intervention may be most appropriate for women because they had lower HDL-cholesterol and higher triacylglycerol concentrations than did men who were following a Step II diet. Because of this provocative evidence, there is a need to conduct intervention studies with diets high in monounsaturated fatty acids to evaluate their efficacy in free-living subjects and to establish which dietary intervention is superior for reducing CVD risk. Another beneficial effect of low-fat diets on CVD risk was that these diets often result in weight loss. Obesity, defined as a body weight 20% above ideal, is also a risk factor for CVD (3). Mild-to-moderate overweight has been associated with an increased risk of CVD (4). Decreases in body weight and body fat are associated with numerous favorable changes in CVD risk factors, including increased HDL-cholesterol concentrations; decreased total cholesterol, LDL-cholesterol, VLDLcholesterol, and triacylglycerol concentrations; and decreased factor VII and plasminogen activator inhibitor 1 concentrations (14, 48 51). In these studies, for every 1-kg decrease in body weight, plasma HDL-cholesterol concentrations increased by mmol/l (48 51) and triacylglycerol concentrations decreased by mmol/l (49). In agreement with earlier reports (49 51), results of the present meta-analysis showed that every 1-kg decrease in body weight resulted in a mmol/L increase in HDL cholesterol (P < 0.001) and a mmol/L decrease in plasma triacylglycerol. A metaanalysis conducted by Dattilo and Kris-Etherton (48) also showed that every 1-kg decrease in body weight was associated with a 0.05-mmol/L decrease in plasma total cholesterol and a 0.02-mmol/L decrease in LDL cholesterol. The present metaanalysis showed a significant correlation between weight loss and a decrease in total cholesterol and LDL cholesterol when analyses were weighted by the number of subjects in each study. Diet in combination with exercise can effectively reduce multiple risk factors for CVD (eg, high plasma LDL, VLDL, and triacylglycerol concentrations; low plasma HDL concentrations; high blood pressure; and excess body weight). Epidemiologic studies have shown that some populations who consume very-low-fat diets have a very low incidence of CVD; this may be because these populations have a very low incidence of overweight and obesity and a high level of physical

13 644 YU-POTH ET AL TABLE 5 Multiple regression analysis: plasma lipids and lipoproteins in response to changes in dietary total fat (TF), saturated fatty acids (SFA), and cholesterol with the change in body weight (BW) as a covariable 1 Variables Coefficient P R 2 Model 1 Absolute change (mmol/l) TC 0.86 TF Cholesterol LDL-C 0.83 TF Cholesterol HDL-C 0.51 TF Cholesterol BW TG 0.46 TF Cholesterol BW Percentage change (%) TC 0.86 TF Cholesterol LDL-C 0.83 TF Cholesterol HDL-C 0.49 TF Cholesterol BW TG 0.56 TF Cholesterol BW Model 2 Absolute change (mmol/l) TC 0.88 SFA Cholesterol LDL-C 0.86 SFA Cholesterol HDL-C 0.51 SFA Cholesterol BW TG 0.64 SFA Cholesterol BW Percentage change (%) TC 0.88 SFA Cholesterol LDL-C 0.85 SFA Cholesterol HDL-C 0.47 SFA Cholesterol BW TG 0.70 SFA Cholesterol BW TF and SFA, change in percentage of energy per day from TF and SFA; BW, change in BW in kilograms; cholesterol, change in dietary cholesterol per day divided by 10; LDL-C, LDL cholesterol; HDL-C, HDL cholesterol. activity (16, 17). In dietary intervention studies conducted in free-living populations, subjects are encouraged to adopt a healthy lifestyle that includes exercise. The present metaanalysis found that exercise had significant effects on plasma lipids and lipoproteins that were independent of the effects of diet. Both dietary modification and body weight loss had independent beneficial effects on plasma lipid profiles and the effects were additive. Diet in combination with exercise resulted in a further significant decrease in plasma total cholesterol, LDL cholesterol, and triacylglycerol compared with diet alone. Furthermore, exercise offset the adverse effects of low-fat diets on plasma HDL-cholesterol concentrations. The effects of exercise on plasma HDL-cholesterol concentrations may have been due to weight loss. Wood et al (7) showed that a Step I dietary intervention with exercise resulted in a greater decrease in body weight and less of a decrease (in females) or an increase (in males) in HDL-cholesterol than did the intervention with no exercise. An earlier meta-analysis (48) found that plasma HDL cholesterol significantly increased only when weight reduction was maintained. In the present meta-analysis, exercise resulted in significant weight loss additional to that of decreased dietary fat and energy intake alone, which prevented a decrease in HDL-cholesterol concentrations. Both intervention and within-population epidemiologic studies have shown that a high fat intake plays a role in weight gain (16, 17, 52, 53). A significant positive correlation between the change in body weight and the change in dietary fat was found in the present meta-analysis (r = 0.46, P < 0.001). A reduction in dietary fat results in weight loss (54, 55), likely because of a decrease in energy intake. Some studies have reported that a high fat intake is related to high energy consumption (54 57). Free-living subjects consuming high-fat diets tend to consume more energy than those consuming highcarbohydrate diets, which results in weight gain (54 58). The present meta-analysis showed that weight loss after dietary intervention was significantly correlated with the change in energy intake (r = 0.54, P < 0.001), which was related to the change in dietary fat (r = 0.47, P < 0.001). Thus, a decrease in dietary fat facilitates a reduction in energy intake, thereby promoting weight loss. In summary, an exhaustive survey of the literature of dietary interventions showed that a reduction in dietary fat and SFA has beneficial effects on CVD risk factors in free-living subjects. Plasma total cholesterol, LDL-cholesterol, and triacylglycerol concentrations and the ratio of total cholesterol to HDL cholesterol significantly decreased after both Step I (by 10%, 12%, 8%, and 10%, respectively) and Step II (by 13%, 16%, 8%, and 7%, respectively) dietary interventions. In many of these interventions, subjects lost weight ( kg, x : 3.38 kg). Weight loss and exercise resulted in a decrease in plasma triacylglycerol and an increase in HDL-cholesterol concentrations. Both exercise and a reduction in dietary fat (related to a decrease in energy intake) increased weight loss (2.8 kg weight loss from exercise and 0.28 kg weight loss for every 1% decrease in energy from total fat, respectively) and the effects were additive. The results of this study provide a good benchmark of the extent to which Step I and Step II intervention programs can affect CVD risk status. In addition, it is clear that exercise as well as weight loss can markedly potentiate the effects of diet. Thus, effective intervention programs should target healthy lifestyle practices that include diet modification, exercise, and