Relation Between a Health-Conscious Diet and Blood Lipids

Original Communication Giessen Wholesome Nutrition Study: relation between a health-conscious diet and blood lipids

I Hoffmann1,2*, MJ Groeneveld1, H Boeing3, C Koebnick4, S Golf 5, N Katz5 and C Leitzmann1

1Institute of Nutrition Science, University of Giessen, Germany; 2National Research Centre for Nutrition, Karlsruhe, Germany; 3German Institute for Nutrition Research, Potsdam-Rehbru¨cke, Germany; 4Department of Medical Informatics, Biometry and Epidemiology, University of Erlangen-Nuremberg, Germany; and 5Institute of Clinical Chemistry, Department of Medicine, University of Giessen, Germany

Objective: To study in humans the relationship between a diet consistent with most of the current recommendations for the prevention of nutrition-related diseases (Wholesome Nutrition) and the blood lipid profile (total cholesterol, LDL-, HDL-cholesterol, LDL=HDL-ratio, triglycerides). Design: Cross-sectional study with two diet groups. Setting: Former West Germany. Subjects: Healthy women (n ¼ 243, aged 25 – 65 y) adhering to Wholesome Nutrition for at least 5 y (subdivided into 111 ovo-lacto vegetarians and 132 low-meat eaters) and an according control group of 175 women eating an average German mixed diet. They were all recruited through an advertisement campaign and selected on the basis of their food consumption. Results: Considering potential confounders, the Wholesome Nutrition subgroups had higher HDL-cholesterol levels than the control group. No differences were observed for total cholesterol and LDL-cholesterol. For LDL=HDL-ratio and triglycerides the effect of diet was dependent on interaction terms. With increasing risk factors (age or body mass index (BMI)) the Wholesome Nutrition subgroups showed more favourable blood lipids. Conclusions: Women eating a preventive diet on a long-term basis exhibit more favourable blood lipid profiles than women consuming an average mixed diet. This is particularly obvious for HDL-cholesterol in the presence of certain risk factors and when an ovo-lacto vegetarian version is practised. Sponsorship: Eden Foundation, Bad Soden, Germany. Descriptors: wholesome nutrition; cholesterol; triglycerides; vegetarianism; blood lipid profile European Journal of Clinical Nutrition (2001) 55, 887–895

Introduction 1994; Pyo¨ra¨la¨, 1996). High levels of total cholesterol, LDL- cholesterol and triglycerides, a high LDL=HDL-ratio and The risk of coronary heart disease (CHD) is especially high low levels of HDL-cholesterol are predictive for the in industrialised countries. Even though the mortality rate increased risk of CHD and are considered to be the major due to CHD has declined significantly in a number of risk factors, besides smoking and high blood pressure European countries during the last two decades, it still (Pyo¨ra¨la¨, 1996). remains the main cause of death. Apart from genetic As diet is involved in the various CHD-related risk factors, environmental factors such as lifestyle and nutri- factors, dietary recommendations for the prevention of tion are involved in its etiology (EAS, 1992; Pyo¨ra¨la¨ et al, CHD are promoted by experts and public health organisa- tions. These include achieving or maintaining a healthy body weight, choosing a diet low in fat, particularly in *Correspondence: I Hoffmann, Institute of Nutrition Science, University saturated fatty acids and cholesterol, and consuming a of Giessen, Wilhelmstr. 20, D-35392 Giessen, Germany. Guarantor: I Hoffmann. variety of vegetables, fruit and whole grain products. Contributors: IH, MJG and CL contributed to the concept of the study, Moreover, sugar, salt and sodium should be used in recruited and obtained data; IH, HB and CK were responsible for statistical moderation, and alcohol, if at all, should be consumed in analyses and writing the paper. SG and NK were responsible for laboratory moderate amounts (Nutrition Committee, American Heart analyses and quality control procedures of the blood lipids. CL conceived Association, 1996, 2000). and participated in all phases of the study. Received 10 January 2000; revised 23 March 2001; accepted With increasing health awareness, diets that meet 30 March 2001 preventive recommendations are becoming increasingly Giessen Wholesome Nutrition Study I Hoffmann et al 888 popular in Germany. This also applies for wholesome between 301 and 1050 g=week, of unheated vegetables nutrition, which has the following general features: (1) between 170 and 852 g=week and of alcohol below preference of foods of plant origin (a primarily ovo-lacto 140 g=week. The criteria applied in the control group vegetarian diet); (2) preference of foods processed as little avoided overlapping of the diet groups with respect to food as possible; (3) plentiful consumption of unheated fresh choice. For the control group, the criteria ensured a con- food; (4) careful preparation of meals from fresh foods; and sumption according to the average for women reported by (5) the sparse use of fat (Koerber et al, 1999). This diet is, the German National Consumption Study (Projekttra¨- therefore, mainly composed of vegetables, fruit, whole gerschaft Forschung im Dienste der Gesundheit, 1991). grain products, potatoes, legumes and dairy products. If The selection criteria were met by 360 women in the eaten at all, the consumption of meat, fish and eggs is wholesome nutrition group and 332 women in the control limited to one or two portions per week. Persons adhering group. A drop-out in connection with blood collection, to wholesome nutrition do not necessarily meet the recom- pregnancy, illness and a second questionnaire on lifestyle mendations for the prevention of CHD to their full extent. factors reduced the number of study participants to 243 However, they represent a suitable group to study the long- women in the wholesome nutrition group and to 175 in the term effect of a health-conscious diet on blood lipid profile control group. For the examination of blood lipids, all for which the empirical database so far is small. Therefore, smokers (wholesome nutrition group, one; control group, with the Giessen Wholesome Nutrition Study, it was 35) and women taking lipid-lowering medication (control examined whether a group of healthy women adhering to group, two) were also excluded, leaving 242 in the whole- wholesome nutrition for at least 5 y had a more favourable some nutrition group and 138 in the control group. These blood lipid profile than a comparable group of women groups represent 67, respectively 42% of the eligible study eating an average German mixed diet (control group). population. The wholesome nutrition group consisted of the subgroups of 111 ovo-lacto vegetarians and 131 low-meat eaters. Methods

Selection of study participants Dietary assessment For a more homogenous group and better compliance, The dietary intake of the study participants was assessed by women were exclusively selected for study participation. a self-administered 7-day estimated (standardised) food Using different advertisements in newspapers and magazines record. Closed food record categories for 151 food items in 1989 and 1990, healthy women aged 25 – 65 y were with the characteristic portion size (household measures recruited in former West Germany and West Berlin. Poten- and quantity in grams) were provided with a booklet, in tial participants from former East Germany were excluded which the food consumption was noted at the time of because of changes in their dietary habits during that period eating. To increase reporting accuracy and unambiguous of time. Inclusion criteria were that women of the whole- classification of certain food items, photographs and some nutrition group had practised wholesome nutrition for detailed descriptions of portion sizes were additionally at least 5 y and that women of the control group had not provided. Detailed written instructions informed the followed any special diet in the past 5 y. About 1500 women participants how to complete the food record. per group responded to the advertisements. Subsequently This semiquantitative food record was developed espe- they received a questionnaire asking about their nutritional cially for this study. A validation study with a subgroup of habits, frequency of food consumption (using a semiquanti- 67 women (33 in the wholesome nutrition group, 34 in the tative food frequency list), anthropometric and socio-demo- control group) showed that the 7-day estimated food record, graphic data, smoking and physical activity, as well as compared to a 7-day weighed food record, classified on prevalent diseases. In all, 1319 women returned the com- average 55% of the participants into the same tertiles. This pleted questionnaire for the wholesome nutrition group and validation study took nutrients as well as food items into 1324 women for the control group. Based on the semiquan- consideration. However, the estimated intake of energy and titative food frequency list, eligible study participants for this nutrients was slightly higher with the 7-day estimated food study were selected according to their responses to several record. The misclassification rates did not differ between specific indicator food items enquired about in the semi- the diet groups and therefore differential misclassification quantitative food frequency list. A priori defined selection can be excluded (Hoffmann, 1994; Hoffmann et al, 1994). criteria for the wholesome nutrition group were: consump- The nutrient intake was calculated on the basis of the tion of whole grain products above 1100 g=week, of German Food Code and Nutrition Data Base BLS II.2 unheated vegetables above 700 g=week, of highly refined (Federal Institute for Health Protection of Consumers and grain products below 1050 g=week, of meat below Veterinary Medicine, 1998) as well as producer’s data. 300 g=week, of meat products below 100 g=week and of alcohol below 70 g=week. Vegans were not considered as study participants since a vegan diet does not agree with the Blood collection wholesome nutrition concept. For the control group, the The blood samples were collected after overnight fasting following criteria were defined: consumption of meat under standardised conditions between 6.00 and 9.00 am in

European Journal of Clinical Nutrition Giessen Wholesome Nutrition Study I Hoffmann et al 889 a sitting position. The blood samples were immediately nificance level was 0.05. All data analyses were performed cooled on ice and centrifuged at 4000 rpm. Blood lipids with SPSS1 10.0 (SPSS Inc., Chicago, IL, USA). were analysed on the same day with enzymatic-photometric assays (CHOD-PAP method for total cholesterol and GPO- PAP method for triglycerides, with a Hitachi 717, Boeh- Results ringer Mannheim, Germany) at the Institute of Clinical Chemistry, Giessen, Germany. The coefficient of variation was 1.5 for total cholesterol and 2.1 for LDL-cholesterol. Basic characteristics The triglycerides were analysed after hydrolysis of glycerol In the distribution over categories, no differences between and fatty acids by measurement of glycerol using glycerol the three study groups could be observed for age, physical kinase and glycerol phosphate oxidase. The coefficient of activity and per capita income (Table 1). Except for alcohol variation was 1.4 for triglycerides. consumption, there were no significant differences between HDL-cholesterol was determined in the supernatant fluid ovo-lacto vegetarians and low-meat eaters. In addition to a after precipitation of VLDL and LDL using phosphowol- lower alcohol consumption, the two wholesome nutrition framate and magnesium chloride; LDL-cholesterol was subgroups had a lower intake of sex hormones and higher determined by subtraction of total cholesterol minus cho- educational levels than the control group. Moreover, the lesterol in the supernatant fluid after precipitation of LDL by polyvinyl sulfate (Boehringer Mannheim, Germany). Table 1 Characteristics of the diet groups The coefficient of variation was 2.6 for HDL-cholesterol. Control Low-meat Ovo-lacto The ratio of LDL-cholesterol to HDL-cholesterol was group eaters vegtarians calculated. Potential confounders of the association between diet 138 131 111 and blood lipids were assessed by a questionnaire (age in years, physical activity and socio-demographic status such n % n % n % as education and per capita income; number of years Age practising wholesome nutrition), by the 7 day estimated 25–34 322327212724 food record (alcohol consumption, intake of sex hormones 35–44 473434263431 45–54 392835272825 including oral contraceptives and hormone replacement 55–66 201535272220 therapy), by direct weighing (in light clothing, without shoes), and by direct measuring of height. Body mass Body mass indexab index (BMI) was calculated as weight in kg divided by 19 6 4 15111917 2 19–24 644689687164 height in m. 24–30 543924182018 > 30 13 9 2 2 1 1 Alcohol consumptionabc No alcohol 18 13 35 27 49 44 Statistical methods 0 – 15 g=day 91 66 88 67 56 50 The relationship between the diet groups and blood lipids was > 15 g=day 29 21 8 6 6 5 tested at different levels: (a) by comparing the means of the Physical activity groups with one-way ANOVA (post-hoc test: Tukey’s hon- Little=never 15 11 6 5 5 5 estly significant difference (HSD) test), (b) by analysis of Medium 93 67 94 72 84 76 covariance and multiple linear regression to control for Much 30 22 31 24 22 20 potential confounders and interactions. All the above-men- Sex hormonesab tioned confounders were tested. However, in the final model Yes 53 38 26 20 15 14 only those confounders and interactions with a significant No 85 62 105 80 96 86 effect of 0.15 were considered (confounders — for all Educationab parameters age and BMI, for LDL-cholesterol additionally 9 y of schooling 35 25 21 16 19 17 number of years practising wholesome nutrition and for 10 y of schooling 66 48 42 32 37 33 triglycerides additionally intake of oral contraceptives; inter- 13 y of schooling 17 12 20 15 14 13 University degree 20 15 48 37 41 37 actions — for LDL=HDL-ratio dietÂBMI, for triglycerides dietÂage). Missing values of blood lipid concentrations (total Income (DM) cholesterol, one; LDL-cholesterol, 10; HDL-cholesterol, 500 18 13 13 10 14 13 500 – 1000 53 38 35 27 28 25 four; LDL=HDL-ratio, 10) were substituted by the average 1000 – 1500 39 28 48 37 38 34 (median) of all other participants (n ¼ 380). Differences in 1500 – 2000 13 9 16 12 12 11 food consumption between two groups were tested by > 2000 13 9 18 14 16 14 ANOVA (post-hoc test, Tukey’s HSD test). The Pearson w2 Significant difference with Pearson chi-square test between: aovo-lacto test was used to test whether one diet group differed from the vegetarians and control group; blow-meat eaters and control group; covo- other diet groups with respect to the categories of confound- lacto vegetarians and low-meat eaters. ing variables. Results were considered significant when sig- P 0.05.

European Journal of Clinical Nutrition Giessen Wholesome Nutrition Study I Hoffmann et al 890 number of participants of the two wholesome nutrition bread instead of white bread, brown rice instead of white subgroups in the lower categories of BMI was higher. rice); they consumed much more and preferably raw The final participants of this study did not show any vegetables and fruit. Whereas the ovo-lacto vegetarian significant difference in comparison to the eligible group group ate no meat=meat products and fish=fish products, when characteristics such as consumption of convenience the low-meat group ate little of these food items. The food, age, BMI, physical activity or occupation were consumption of sweets and foods containing sugar (eg compared. soft drinks) was lower than in the control group. Moreover, they drank fewer alcoholic beverages and preferred mostly Food consumption herbal teas and mineral water. Of the three study groups, The food consumption of the two wholesome nutrition the ovo-lacto vegetarians consumed the least milk (in subgroups differed significantly from the control group comparison to low-meat eaters, difference not significant) (Table 2), partly due to the selection criteria for further and dairy products as well as eggs, indicating that they participation in the study that was based on dietary habits. replaced meat and fish with plant foods instead of other Both wholesome nutrition subgroups preferred whole grain animal foods. The total carbohydrate intake of the whole- products to highly processed grain products (eg whole grain some nutrition subgroups was higher (containing more

Table 2 Food consumption (in g=day) of the diet groups

Control group Low-meat eaters Ovo-lacto vegetarians Food item n ¼ 138 n ¼ 131 n ¼ 111

Bread and pastries, total median 191 179 179 95% CI 179=202 160=200 157=202 Whole grain productsa,b median 47 140 159 95% CI 41=57 127=160 133=175 From refined floura,b median 136 25 9 95% CI 122=148 18=35 0=16 Vegetables, legumes, totala,b,c median 203 322 405 95% CI 187=229 283=354 353=443 rawa,b,c median 72 168 238 95% CI 61=87 154=195 216=260 heateda median 118 147 142 95% CI 107=143 116=162 127=164 Fruit, totala,b,c median 190 317 373 95% CI 169=213 285=362 314=412 raw fruita,b,c median 164 298 343 95% CI 143=193 252=332 293=375 Potatoesa median 89 66 51 95% CI 77=103 55=77 51=77 Edible fat=cooking oila,b median 15 19 20 95% CI 14=18 17=22 17=24 Milk=dairy products median 223 239 163 95% CI 194=267 205=279 112=213 Eggsa,c median 16 12 6 95% CI 10=16 8=16 2=8 Meat=meat products, totala,b,c median 127 21 0 95% CI 116=136 16=32 0=0 meata,b,c median 70 12 0 95% CI 64=80 11=17 0=0 meat productsa,b,c median 50 5 0 95% CI 46=59 1=70=0 Fish=fish productsa,b,c median 21 11 0 95% CI 11=21 0=15 0=0 Sweeteners median 5 4 3 95% CI 4=63=62=5 Desserts=sweetsa,b median 33 12 9 95% CI 26=36 9=16 6=14 Beverages, total median 1315 1209 1257 95% CI 1237=1426 1122=1309 1129=1453 coffee, black teaa,b,c median 489 343 129 95% CI 450=536 257=407 43=249 alcoholic beveragesa,b median 86 29 14 95% CI 63=114 29=57 0=29

Signiﬁcant differences (P 0.05) with Tukey’s HSD test between: aovo-lacto vegetarians and control group; blow-meat eaters and control group; covo-lacto vegetarians and low-meat eaters.

European Journal of Clinical Nutrition Giessen Wholesome Nutrition Study I Hoffmann et al 891 complex carbohydrates and monosaccharides but less disaccharides), as was that of dietary ﬁbre (water-soluble as well as water-insoluble). The total fat intake was lower, whereas the intake of polyunsaturated fatty acids was higher and that of saturated fatty acids was lower. The intake of cholesterol was about half and the alcohol consumption very low compared to the control group.

Blood lipids Examining the raw data, no crude association between belonging to a diet group and total cholesterol as well as LDL-cholesterol could be detected (Table 3). The LDL=HDL-ratio and triglyceride concentrations of the ovo-lacto vegetarians as well as low-meat eaters were significantly lower; the concentration of HDL-cholesterol of the low-meat eaters was significantly higher than that of the controls. After controlling for potential confounders and interactions, the results were partly different. There was neither an association between diet and total cholesterol nor a difference in total cholesterol levels between the study groups. Figure 1 Association between LDL=HDL-ratio and BMI for the diet groups. The arrow indicates the start of the significant association between Additionally, no association between diet and LDL-choles- LDL=HDL-ratio and diet. terol and no significant differences in LDL-cholesterol concentrations between the diet groups was observed. For HDL-cholesterol, a significant association with diet could be shown (P ¼ 0.026). The mean HDL-cholesterol levels of are no significant differences between low-meat eaters and low-meat eaters (P ¼ 0.009) and ovo-lacto vegetarians control group. (P ¼ 0.050) were significantly higher than those of the Considering an interaction between diet and age, there control group. For the other parameters, multiple analysis was a significant relationship between diet and level of revealed an interaction between diet and one potential triglycerides starting at an age of 45 y (P ¼ 0.046, Figure 2). confounder, meaning that the diet has a different effect With increasing age, the control group had significantly size at different levels of the confounder. For the higher levels of triglycerides than the low-meat eaters LDL=HDL-ratio, an association with diet and an interaction (P ¼ 0.030 for age ¼ 45 y) and ovo-lacto vegetarians between diet and BMI was observed (Figure 1). Starting (P ¼ 0.0031 for age ¼ 45 y). There were no differences in with a BMI of 25 kg=m2, a significant association between the triglyceride levels between the two subgroups of whole- diet and LDL=HDL-ratio has been observed (P < 0.044 for some nutrition. BMI ¼ 25 kg=m2). From this point upwards, the control group showed an increasingly and significantly higher LDL=HDL-ratio compared to the ovo-lacto vegetarians Discussion (P ¼ 0.013 for BMI ¼ 25 kg=m2). Low-meat eaters showed higher LDL=HDL-ratio than ovo-lacto vegetarians The present study shows that women adhering to pre- when BMI was higher than 28 kg=m2 (P ¼ 0.044). There ventive dietary recommendations in their daily diet on a

Table 3 Blood lipids (mmol=l) of the diet groups, raw data

Control group Low-meat eaters Ovo-lacto vegetarians n ¼ 138 n ¼ 131 n ¼ 111

Total cholesterol median 5.53 5.72 5.40 95% CI 5.33=5.77 5.48=5.92 5.20=5.64 LDL cholesterol median 3.52 3.57 3.31 95% CI 3.26=3.80 3.39=3.85 3.05=3.59 HDL cholesterolb median 1.50 1.63 1.66 95% CI 1.45=1.63 1.58=1.78 1.58=1.76 LDL=HDL-ratioa,b median 2.34 2.15 2.03 95% CI 2.16=2.48 1.95=2.41 1.79=2.21 Triglyceridesa,b median 2.12 1.86 1.84 95% CI 1.94=2.46 1.63=2.02 1.68=1.94

Signiﬁcant differences (P 0.05) with Tukey’s HSD test between: aovo-lacto vegetarians and control group; blow-meat eaters and control group.

European Journal of Clinical Nutrition Giessen Wholesome Nutrition Study I Hoffmann et al 892 Study (Kafrissen, 1990) and the Prospective Cardiovascular Mu¨nster Study (PROCAM, Assmann & Schulte, 1992). For triglycerides, the existing data revealed an association with diet for women above 45 y of age. With an increasing risk due to age (Gillum et al, 1998) both subgroups of wholesome nutrition exhibited significantly lower triglyceride levels. A recent meta analysis based on prospective studies revealed that high triglyceride levels are an important risk factor for CHD, especially in women, and independent of HDL-cholesterol (Austin, 1997; Austin et al, 1998). For total cholesterol and LDL-cholesterol, our data did not show differences between the diet groups. Both lipoproteins are risk factors for CHD, whereas LDL-cholesterol shows the strongest and most consistent association with CHD of all lipid fractions. Data of the PROCAM and other prospective studies confirm a log – linear relationship between LDL-cholesterol and risk of CHD (Cullen et al, 1998). The reason why no differences for total cholesterol were detected in our study may partly be due to the fact that this parameter combines several subcomponents, including Figure 2 Association between level of triglycerides and age for the diet the two contrary subfractions of LDL- and HDL-choles- groups. The arrow indicates the start of the significant association between terol. Since most of the cholesterol in plasma is found triglycerides and diet. in LDL particles, total cholesterol and LDL-cholesterol are highly correlated. This may also explain the similar results for total cholesterol and LDL-cholesterol in this investigation. long-term basis have more favourable lipid profiles than Taking the presented data, a lower risk for CHD in women consuming an average diet. This is especially the long run may be expected for the wholesome nutri- obvious for HDL-cholesterol and when certain risk factors tion groups, since important lipid risk factors are more are present and when an ovo-lacto vegetarian version of favourable. such a preventive diet is practised. This positive influence Furthermore, our data support the importance of pre- occurred despite some attenuating factors present in this ventive dietary recommendations since the protective char- study. The facts that the control group showed a somewhat acter becomes more obvious when risk factors are more favourable diet than the German average, that the size elevated — the higher a risk factor the more favourable of the study groups was comparatively small, that the the lipid profile with a preventive diet. Within the ‘normal dietary intake and blood lipids were determined only range’, the effect of a preventive diet seems to be less once, and that no correction for within-person variability decisive than in a range of risk. This is in agreement with could be undertaken indicate that the observed relationships the findings reported by Liu et al (2000, 2001). In the are stronger in reality than empirically observed. Women’s Health Study (Liu et al 2000), the inverse Women practising wholesome nutrition exhibited a association between fruit and vegetable intake and the higher level of HDL-cholesterol. HDL-cholesterol is con- incidence of CVD, even though not significant, also sidered to be a main risk factor for CHD with high HDL- appeared more evident among women with BMI > 25 and cholesterol concentrations being protective against CHD current smokers. This could be confirmed on a statistically and low HDL-cholesterol concentrations being associated significant basis for men in the Physicians’ Health Study with a higher risk of CHD (Expert Panel on Detection, (Liu et al, 2001). Evaluation, and Treatment of High Blood Cholesterol in No significant differences in blood lipids were Adults, 1993). This has been confirmed with the Lipid found between ovo-lacto vegetarians and low-meat eaters. Research Clinics Prevalence Study (Jacobs et al, 1990) and It is possible that no differences exist, but since the whole- the Framingham-Study (Castelli et al, 1986), as well as some nutrition group was divided up into the two sub- other studies. groups of ovo-lacto vegetarians and low-meat eaters they The ovo-lacto vegetarians showed a significantly lower are smaller and show a fairly high similarity in their diets. and therefore more favourable LDL=HDL-ratio than the Therefore, potentially existing differences may be difficult control group for BMI above 25 kg=m2. This is of special to show. importance since a comparison of four blood cholesterol Even though none of the examined blood lipids discri- measurements within large study populations has shown minated between ovo-lacto vegetarians and low-meat that this ratio is a better predictor of CHD risk than the eaters, the advantage of the ovo-lacto vegetarian version single blood lipid measurements (Kinosian et al, 1994). of wholesome nutrition (compared to the average German This fact is supported by studies such as the Bogalusa Heart diet) was more obvious than that of the low-meat eaters.

European Journal of Clinical Nutrition Giessen Wholesome Nutrition Study I Hoffmann et al 893 Other studies with healthy females and males support and a negative correlation with triglycerides (Kimira et al, the advantages of vegetarian diets concerning blood lipids. 1998). Flavonoids and antioxidant vitamins may serve as However, their results are contrary to our findings as they markers for diets rich in plant foods. A study of 34 492 predominantly show lower concentrations of total choles- postmenopausal women revealed a clear inverse associa- terol and LDL-cholesterol for vegetarians in comparison to tion between whole grain intake and risk of death due to the average population, whereas the results for HDL- ischemic heart disease, which was not explainable by cholesterol are not consistent. Besides the differing effect dietary fibre intake or several other constituents of whole of risk factors, the most important finding of our study is grains (Jacobs et al, 1998). The data of the Nurses’ Health the higher concentration of HDL-cholesterol compared to Study with 75 521 middle-aged women supports the pro- the control group. tective effect of an increased whole grain consumption The findings of the Berlin Vegetarian Study with its against CHD that could not be fully explained by the intake univariate statistical analyses show that the female vegetar- of dietary fibre, folate, vitamin B6 and vitamin E (Liu et al, ians had lower total cholesterol and triglyceride levels than 1999). Also, the results of the Scottish Heart Health Study the non-vegetarians, whereas no differences for HDL- led to the conclusion that antioxidant vitamins and dietary cholesterol level were observed (Rottka, 1990). In a lipid fibre may be indicative in the prediction of the cholesterol screening at population level for HDL-cholesterol in the level (Bolton-Smith et al, 1991). All these findings may area of Leipzig (Germany), no significant differences give an explanation as to why the HDL-concentration in the between vegetarians (43 men and 76 women) and the two wholesome nutrition subgroups is higher than in the general population were observed, whereas the vegetarians control group. They also support the necessity to pay more had lower total cholesterol levels and the total cholest- attention to the relationship between lipoproteins and the erol=HDL-cholesterol-ratios were more favourable for complete diet or the overall consumption of fruit and vegetarians (Richter et al, 1993). A British study with vegetables. 1550 ovo-lacto vegetarians, 114 vegans, 415 fish eaters The wholesome nutrition subgroups met the CHD-pre- and 1198 meat eaters found higher total and LDL-choles- ventative recommendations except for the fat intake. For terol levels in meat eaters than in vegans and in vegetarians example, they preferred full-fat to non-fat or low-fat dairy and fish eaters intermediate and similar values. HDL- products. The intake of fat was higher than recommended. cholesterol levels were highest in the fish eaters but did Even though the absolute fat intake of the ovo-lacto not differ among the other groups (Thorogood et al, 1987). vegetarians was lower, the proportion of energy deriving Study groups with special diets such as macrobiotics (Sacks from fat was about the same compared to the control group. et al, 1975), vegans (Burslem et al, 1978), and subjects However, the distribution over the different fatty acids was with a special interest in ‘health food’ (Burr et al, 1981) more favourable (ie higher intake of omega-6 fatty acids also show lower total cholesterol, LDL-cholesterol but and polyunsaturated fatty acids, lower intake of saturated lower or higher HDL-cholesterol levels than the controls fatty acids). Considering the question as to what nutrients consuming an average diet. explain the differences of the blood lipids, we found that A possible explanation for the fact that the results of our only a small part of the variation of the blood lipids could study differ from other studies may be that participants of be explained by the intake of total fat, saturated fatty acids, the wholesome nutrition subgroups had a well planned food polyunsaturated fatty acids, cholesterol, polysaccharides, choice close to the recommendations for the primary monosaccharides, soluble dietary fibre and energy (with prevention of CHD with the exception of the percentage multiple linear regression; data not shown). This was true of fat intake (Nutrition Committee, American Heart Asso- for the raw data, the data corrected by the validation study, ciation, 1996). They ate a favourable variety of food items the energy adjusted data and the data combined to the Keys of all food groups and they ate in moderation (eg adequate score. If significant effects of nutrients on blood lipid energy intake), with a high proportion of vegetables, fruit concentrations were found (mainly for HDL-cholesterol), and whole grain products. Furthermore, they drank little, if it was more the case for the carbohydrate fractions, poly- any, alcohol. There were significant differences between and monosaccharides, and soluble dietary fibre rather than the three study groups concerning the adherence to food- for fat or its fractions and cholesterol. related recommendations. The vegetarians were closest to A British cross-sectional study on the relation between the recommendations, the control group differed most from diet and blood lipids in health conscious groups (vegetar- those, and the low-meat eaters were in-between. This ians, vegans, fish eaters and meat eaters, 52 participants in overall health conscious food choice of the wholesome each group) came to similar conclusions concerning fat nutrition subgroups cannot always be assured for the intake. The subjects in this study did not replace energy participants of other studies, in which the participants derived from fat with energy deriving from carbohydrate. often merely omit certain food items from their diet. Our Instead, they altered the nature of both the fat and carbo- finding of a higher HDL-concentration in blood is sup- hydrate foods consumed. None of the measures of fat intake ported by studies describing a favourable influence of was associated with HDL-cholesterol (Thorogood et al, phytochemicals on this blood parameter. A study with 50 1990). There is good evidence that the replacement of fats women demonstrated a positive correlation of flavonoids by carbohydrates is associated with a decrease of HDL- and antioxidant vitamins with the level of HDL-cholesterol cholesterol (Ve´lez-Carrasco et al, 1999). The effect is

European Journal of Clinical Nutrition Giessen Wholesome Nutrition Study I Hoffmann et al 894 strongest for the replacement of saturated fatty acids by Federal Institute for Health Protection of Consumers and Veterinary carbohydrates but also seen for the replacement of mono- Medicine (1998): The German Food Code and Nutrient Database (BLS II.2): Conception, Structure and Documentation of the Database and polyunsaturated fatty acids (Katan, 1998). A study blsdat. BgVV Publications, Berlin, Germany. from New Zealand with 23 non-vegetarian and 24 vegetar- Gillum RF, Mussolino ME & Madans JH (1998): Coronary heart disease ian Seventh-Day-Adventists supports this finding, since risk factors and attributable risks in African-American women and men: both groups exhibited similar blood lipid levels even NHANES I epidemiologic follow-up study. Am. J. Public Health 88, though the vegetarian group consumed a diet lower in 913 – 917. Harman SK & Parnell WR (1998): The nutritional health of New Zealand total fat, saturated fatty acids and cholesterol (Harman & vegetarian and non-vegetarian Seventh-Day Adventists: selected vita- Parnell, 1998). min, mineral and lipid levels. N.Z. Med. J. 111, 91 – 94. The results suggest that, in a preventive diet, a reduction Hoffmann I (1994): Giebener Vollwert-Ernaehrungs-Studie: Untersuchung of total fat intake may not be as decisive for the blood lipid auf Bias am Beispiel von Fettstoffwechsel-Parametern. Dissertation, University of Giessen. Giessen: Wiss. Fachverlag. (in German). fractions and for the risk of CHD as the proportion of foods Hoffmann I, Kohl M, Groeneveld M & Leitzmann C (1994): Development of plant origin in the diet. For favourable blood lipid and validation of a new instrument to measure food intake. Am. J. Clin. profiles, individuals can apparently choose a diet with a Nutr. 59(Suppl 1), 284S. similar proportion of total fat and total carbohydrate com- Jacobs DR, Mebane IL, Bangdiwala SI, Criqui MH, Tyroler HA (Lipid pared to the general population if the proportion of plant Research Clinics Program) (1990): High density lipoprotein cholesterol as a predictor of cardiovascular disease mortality in men and women: foods is high. Such a diet can fulfil most of the current the follow-up study of the lipid research clinics prevalence study. Am. J. recommendations and may prove to be more acceptable, Epidemiol. 131, 32 – 47. compliant and palatable than low-fat=high-carbohydrate Jacobs DR Jr, Meyer KA, Kushi LH & Folsom AR (1998): Whole-grain diets. Also, this dietary approach assures the intake of intake may reduce the risk of ischemic heart disease death in post- food constituents (eg antioxidants, phytochemicals) that menopausal women: the Iowa Women’s Health Study. Am. J. Clin. Nutr. 68, 248 – 257. are relevant in reducing risk factors for CHD apart from Kafrissen ME (1990): Prevention of cardiovascular risk in women. A new blood lipids. Further investigations are needed to corrobo- concern for the obstetrician=gynecologist. Acta Obstet. Gynecol. Scand. rate these findings. 152(Suppl), 13 – 20. Katan MB (1998): Effect of low-fat diets on plasma high-density lipoprotein concentrations. Am. J. Clin. Nutr. 67(Suppl 3), 573S – Acknowledgements —The authors gratefully acknowledge the financial 576S. support of the Eden Foundation, Bad Soden, Germany. Thanks to P Kimira M, Arai Y, Shimoi K & Watanabe S (1998): Japanese intake of Leitzmann for reading the manuscript. flavonoids and isoflavonoids from foods. J. Epidemiol. 8, 168 – 175. Kinosian B, Glick H & Garland G (1994): Cholesterol and coronary heart disease: predicting risks by levels and ratios. Ann Intern Med. 121, 641 – 647. References Koerber Kv, Maennle T & Leitzmann C (1999): Vollwert-Ernaehrung — Konzeption einer zeitgemaessen Ernaehrungsweise, 9th edn. Heidel- Assmann GS & Schulte G (1992): Relation of high-density lipoprotein berg: Haug (in German). cholesterol and triglycerides to incidence of atherosclerotic coronary Liu S, Stampfer MJ, Hu FB, Giovannucci E, Rimm E, Manson JE, artery disease (the PROCAM experience). Am. J. Cardiol. 70, 733 – Hennekens CH & Willet WC (1999): Whole-grain consumption and 737. risk of coronary heart disease: results from the Nurses’ Health Study. Austin MA (1997): Triacylglycerol and coronary heart disease. Proc. Nutr. Liu S & Manson JE (2000): Fruit and vegetable intake and risk of Soc. 56, 667 – 670. cardiovascular disease: the Women’s Health Study. Am. J. Clin. Nutr. Austin MA, Hokanson JE & Edwards KL (1998): Hypertriglyceridemia as 72, 922 – 928. a cardiovascular risk factor. Am. J. Cardiol. 81, 7B – 12B. Liu S, Lee I, Ajani U, Cole S, Buring J & Manson J (2001): Intake of Bolton-Smith, Woodward M, Smith WCS & Thunstall-Pedoe H (1991): vegetables rich in carotinoids and risk of coronary heart disease in men: Dietary and non-dietary predictors of serum total and HDL-cholesterol the Physicians’ Health Study. Int. J. Epidemiol. 30, 130 – 135. in men and women: results from the Scottish Heart Health Study. Int. J. Nutrition Committee, American Heart Association (1996): Dietary Epidemiol. 20, 95 – 104. Guidelines for healthy American adults. Circulation 94, 1795 – Burr ML, Bates CJ, Fehily AM & St Leger AS (1981): Plasma cholesterol 1800. and blood pressure in vegetarians. J. Hum. Nutr. 35, 437 – 441. Nutrition Committee, American Heart Association (2000): AHA Scientific Burslem J, Schonfeld G, Howald MA, Weidman SW & Miller JP (1978): Statement: AHA Dietary Guidelines. Revision 2000: a statement for Plasma apoprotein and lipoprotein lipid levels in vegetarians. Metabo- healthcare professionals from the Nutrition Committee of the American lism 27, 711 – 719. Heart Association. J. Nutr. 131, 132 – 146. Castelli WP, Garrison RJ, Wilson PWF, Abbott RD, Kalousdian S & Projekttra¨gerschaft Forschung im Dienste der Gesundheit (1991): Die Kannell WB (1986): Incidence of coronary heart disease and lipoprotein Nationale Verzehrsstudie. Ergebnisse der Basisauswertung, Band 18. cholesterol levels. The Framingham Study. J.A.M.A. 256, 2835 – 2838. Bonn: Schriftenreihe zum Programm der Bundesregierung (in German). Cullen P, Funke H, Schulte H & Assmann G (1998): Lipoproteins and Pyo¨ra¨la¨ K (1996): CHD prevention in clinical practice. Lancet 348, S26 – cardiovascular risk — from genetics to CHD prevention. Eur. Heart. J. S28. 19(Suppl C), C5 – C11. Pyo¨ra¨la¨ K, De Backer G, Graham I, Poole-Wilson P & Wood D EAS (European Atherosclerosis Society) (1992): Prevention of coronary (on behalf of the Task Force) (1994): Prevention of coronary heart disease: scientific background and new clinical guidelines. Nutr. heart disease in clinical practice. Recommendations on the Task Metab. Cardiovasc. Dis. 2, 113 – 156. Force of the European Society of Cardiology, European Atherosclerosis Expert Panel on Detection, Evaluation and Treatment of High Blood Society and European Society of Hypertension. Eur. Heart J. 15, Cholesterol in Adults (1993): Summary of the second report of the 1300 – 1331. National Cholesterol Education Program (NCEP) expert panel on Richter V, Rassoul F, Purschwitz K, Hentschel B & Rotzsch W (1993): detection, evaluation and treatment of high blood cholesterol in adults Lipid screening on population basis and of vegetarians. Akt. Ernaehr. (adult treatment panel II). J.A.M.A. 269, 3015 – 3023. Med. 18, 286 – 290.

European Journal of Clinical Nutrition Giessen Wholesome Nutrition Study I Hoffmann et al 895 Rottka H (1990): Health and vegetarian life-style. Bibl. Nutr. Dieta. 45, Thorogood M, Roe L, McPherson K & Mann J (1990): Dietary intake and 176 – 194. plasma lipid levels: lessons from a study of the diet of health conscious Sacks FM, Castelli WP, Donner A & Kass EH (1975): Plasma lipids and groups. Br. Med. J. 300, 1297 – 1301. lipoproteins in vegetarians and controls. New Engl. J. Med. 292, 1148 – Ve´lez-Carrasco W, Lichtenstein Alice H, Welty FK, Li Z, Lamon-Fava S, 1151. Dolnikowski GG & Schaefer EJ (1999): Dietary restriction of saturated Thorogood M, Carter R, Benﬁeld L, McPherson K & Mann JI (1987): fat and cholesterol decreases HDL ApoA-I secretion. Arterioscler. Plasma lipids and lipoprotein cholesterol concentrations in people with Thromb. Vasc. Biol. 19, 918 – 924. different diets in Britain. Br. Med. J. 295, 351 – 353.

European Journal of Clinical Nutrition