European Heart Journal (1991) 12,869-874
The relationship between coffee consumption and lipid levels in young and older people in the Heidelberg-Michelstadt-Berlin study L. KOHLMEIER, G.
MENSINK AND M. KOHLMEIER
Institutefor Social Medicine and Epidemiology of the Federal Health Office, Federal Republic ofGermany
KEY WORDS: Cholesterol, coffee, lipoproteins. The relationship between coffee consumption and serum lipid levels was studied in a randomly selected sample of395 young and 385 elderly adults in the Federal Republic of Germany. Analyses were done separately for men and women andfor young and olderpeople. After adjustmentfor body mass index, activity level, smoking, total energy intake, dietary fat.fish, milk, tea and alcohol consumption and oral contraceptive use a statistically significant increase in total serum- and LDLcholesterol levels with increased coffee consumption was observed in the group of young men only. In this subgroup a difference of 0·11 mmol .1- J serum cholesterol and of 0·10 mmol. 1- J LDL cholesterol for each additional cup of coffee consumed daily was calculated. This is one of the few studies that includes elderly people. No significant trend was seen between coffee consumption and serum- or LDL-cholesterollevels in the elderly.
Introduction In 1983 we reported a relationship between serum cholesterollevels and coffee consumption in German men'". In agreement with Scandinavian findings coffee consumption was positively correlated with serum cholesterol level!", The effect of coffee consumption, however, may possibly be attributed to confounding variables such as smoking, dietary habits, physical activity and body mass. This paper presents a closer look at coffee in perspective to other determinants of serum lipid and lipoprotein levels in younger and older people in the Federal Republic of Germany, drawn from regional surveys.
Subjectsandmethods The data are drawn from the Heidelberg-MichelstadtBerlin study on nutrition, lifestyle and health. For this study 395 young (aged 18-24 years) and 385 older (aged 65-74 years) men and women were selected randomly from resident registration lists of three areas of the Federal Republic of Germany'v", A participation rate of 53% was achieved from the original random sample. The major reason given for non-participation was the short stay (2-6 weeks) of the examination unit at one location, at which time many residents were not reachable. The subjects participated in an extensive clinical examination, had blood and urine samples taken for laboratory analyses and completed a detailed questionnaire about their physical activity, nutritional habits, lifestyle and medical history. Additionally, a 7-day dietary and physical activity protocol was recorded. The dietary protocols were self recorded and prospectively conducted. They Submitted for publication on 26 April 1990. and in revised form 2 July 1990. Correspondence: Dr Lenore Kohlmeier, Institut fur Sozialmedizin und Epidemiologie des Bundesgesundheitsamtes, General-Pape-StraBe 62-66 1000 Berlin 42, West Germany.
0195-668Xj91j080869 +06 $03.00/0
were coded with a food code with approximately 1200 foods and food groups'". These protocols formed the basis of the analyses for the effects of dietary fat, alcohol, fish and milk consumption on lipid levels. Coffee consumption (cups of coffee usually consumed per day) was obtained from the questionnaire. Activity level was based on a question about leisure physical activity and was divided into four categories (1. no sporting activities; 2. less than one hour per week; 3. at least one hour per week; 4. competitive sport in a club). For the measurement of lipids and lipoproteins, blood was obtained after an overnight fast, allowed to clot and the serum separated after 45 exactly min. On the same day, all lipid and lipoprotein analyses were completed. Cholesterol and triglycerides in whole serum and in lipoprotein fractions were measured enzymatically. Lipoproteins were prepared quantitatively by micro-ultracentrifugationl'", followed by polyanionic precipitation, comparable to Lipid Research Clinic methodology. Quality control of lipids was carried out according to the requirements of the German 'Bundesarztekammer' and the specifications of the German Clinical Chemistry Society. All external measurements were accurate, with less than a 3% variation. Blind controls were carried out through duplicate samples. Differences of less than 3% on average were found in the double samples from the same venepuncture. Lipoprotein concentrations were calculated from the lipid concentrations in the lipoprotein fractions!". ANALYTICAL MODEL
In a first exploration of lifestyle differences, comparisons were made between heavy (more than three cups per day) and lighter or non coffee drinkers using Student's ttest. A multiple regression modelf ofdeterminants oftotal serum cholesterol, LDL-cholesterol, HDL-cholesterol and apolipoprotein B was designed. This included independent variables such as tea consumption, body mass © 1991 The European Society of Cardiology
870
L. Kohlmeier et al.
Table 1
Frequency distribution of coffee consumption per day Juniors
Seniors
>6cups 5 {j cups 3 -4 cups 1-2 cups < 1 cup Never Missing Total
Men
Women
Men
Women
n
%
n
%
n
%
n
%
7 18 22 52 50 28 7
4·0 10·2 12-4 29·4 28·2 15·8
5 14 58 70 29 30 5
2-4 6·8 28·2 34·0 14·[ 14·6
5 13 68 52 6 22 3
3·0 7-8 41·0 31·3 3-6 13·3
5 20 73 85 13 17 3
2·3 9-4 34·3 39·9 6·1 8·0
184
100
211
100
169
100
216
100
index, alcohol intake, cigarette smoking, physical activity level, fat intake per day in grams, total energy intake and fish and milk consumption, together with the parameter of interest, coffee consumption. The models were run separately for men and women, and for young and older people. In the regression analyses for young women, oral contraceptive use was included as a determinant. Regression equations were calculated but individuals with missing values were not included. All included data were available for 570 of the participants. A high correlation between energy intake and fat intake (in grams) was seen. Regression analysis assumes a linear relationship between coffee consumption and cholesterol, so an additional one way analysis of covariance was conducted using body mass index, physical activity level and (for young women) oral contraceptive use as covariates. Statistical analyses were carried out using the SPSS x software package'?',
Results
In Table I the frequency distribution of coffee consumption of the study population is given, divided by sex and age groups. An examination of the differences in food intake between heavy coffee drinkers (more than three cups per day) and lighter or non coffee drinkers showed significant differences in pork, salad, fruit and schnaps intake among the young men. Heavy coffee drinkers consumed more pork and schnaps but less salad and fruit. It was also noted that heavy coffee drinkers were shorter than the lighter drinkers. Their total fat and total energy intake did not differ from the lighter coffee drinkers. Among young women, milk and soft drink consumption was lower, but wine and champagne intake higher among the heavy coffee drinkers. These women had a lower energy intake, they consumed fewer carbohydrates, minerals, calcium and carbohydrates per day, drank more alcohol and had higher body mass indices than their peers. In Tables 2 and 3 the results of the multiple linear regressions are presented for both gender and age groups. Total energy intake was not included in these equations since they remained insignificant in the earlier models of
cholesterol and energy intake, and are correlated closely with fat intake. It was found that coffee plays a significant unfavourable role in young men only. In this group, coffee consumption remains a substantial determinant ofcholesterol and LO Lcholesterol levels, together with body mass index and lack of sport participation. For young men a difference of 0·109 mmol . 1- I (SO 0,044) of total cholesterol and 0·104 mmol . I-I (SO 0'040) LOL-cholesterol per cup of coffee was seen. The apolipoprotein B levels were also associated with coffee consumption, with an increase of 0·029 g . 1 I for each cup of coffee consumed per day (data not presented). The extent of the effect on this group of young men was surprisingly high. The results of the analysis of covariance show a dose-response relationship between coffee intake and serum cholesterol and LOL-cholesterollevels (Figs I and 2). Among young women, an effect of coffee consumption was not seen in relation to cholesterol and LOL cholesterol levels. The major determinants of cholesterol and LOL cholesterol among young women were body mass index, sport participation, and oral contraceptive use; all except sport operated unfavourably. Mean serum cholesterol and LOL cholesterol levels in women at all levels of coffee intake were as high as those of the men consuming more than four cups of coffee a day (Fig. I). Similarly, among older people, there was no evident effect between current coffee intake and lipid levels. Coffee consumption was negatively correlated (although not statistically significant) with lipid levels, as was tea consumption. Tea consumption is associated with healthier levels of both cholesterol, LOL cholesterol and apolipoprotein B in elderly men. The magnitude of effect corresponds to a reduction of O'240 mmol . 1- I of total serum cholesterol (0·220 mmol l I LOL cholesterol) per cup of tea. Alcohol intake was the only other major determinant of serum lipids and lipoproteins in both genders. i
Discussion
Numerous studies have examined, in experimental or cross sectional design, possible effects of coffee consumption on serum lipid levels. In cross sectional studies an
Relationship between coffee consumption and lipid levels 871
Table 2
Determinants a/serum cholesterol (mmol. I J)
Regr. coelT. ± SE (Tvvalue)
Regr. coelT. ± SE (T-value) Men
Independent varia bles Juniors (N= 124)
Seniors (N= (26)
0·109± 0·044 (2'49)* 0·078 ± 0·032 (2AW -0·159±0·073 (2'18)* 0·004 ± 0·004 -I x 10 '±0·0003 - 0·0002 ± 0·0022 0·042 ± 0·056 0·022 ± 0·035 0·003 ± 0·004 -0·090±0·175 R'=0·18
Coffee (cups. day ') BMI (kg. m ') Sports (categorical) Fish (g. day') Milk (g. day') Fat (g. day') Tea (cups. day') Age Alcohol (g . day') Smoking (dichotomous)
- 0·088 ± 0·063 - 0·045 ± 0·032 -0·010±0·107 - 0·002 ± 0·004 - 0·0005 ± 0·0008 - 0·002 ± 0·004 - O'240 ± 0·083 (-2·90)" 0·023 ± 0·040 -0·01210·005 (- 2'25)* -0·245±0·241 R'=0·13 Women
Juniors (N = (52) Coffee (cups. day') BMI(kg.m ') Sports (categorical) Fish (g . day') Milk (g. day') Fat(g. day') Tea (cups. day ') Age Alcohol (g. day') Smoking (dichotomous) Oral contraceptive use (dichotomous)
Seniors (N= 139)
0·031 ± 0·045 0·036 ± 0·028 - 0·048 ± 0·079 - 0·009 ± 0·006 - 0·0005 ± 0·0006 0·005 ± 0·003 - 0·069 ± 0·054 0·028 ± 0·036 - 0·003 ± 0·009 - 0·128 ± 0·151 0·350 ± 0·142 (2'46)* R'=O·1O
- 0·008 ± 0·071 - 0·0 1010·027 (2'57)* 0·320 ± 0·124 - 0·005 ± 0·006 - 0·002 ± 0·0009 (-2'16)* 0·003 ± 0·004 -0,061 ±0·080 0·026 ± 0·044 - 0·025 ± 0·0 13 0·094 ± O'362
R'=O'1O
*Values presented are significant at a level of P < 0·05. **Values presented are significant at a level of P< 0·0 I. 5'0
4'0
r
4'8 t-'
'~
';=
.-3·5 f-
4·6 f-
'~
c;
c;
E E 4-4
r-
E E
2'5 1-.
4'2 f-.
4·0
r-
3'0
L..-
L...--
< 1 cup
1-2 cups
3-4 cups
>4 cups
Figure I Adjusted mean levels of serum total cholesterol in men and women (aged 18-24 years) according to coffee consumption adjusted for BMI, sporting activity groups and oral contraceptive Use (shaded bar: men; white bar: women).
effect was seen in both sexes!'? 121, in women only'!", in men only[l4.'5) or in neither'!", Experimental findings Support the hypothesis that coffee consumption affects lipid levels'!".
2·0 '---< 1 cup
'---
1-2 cups
3-4 cups
>4 cups
Figure 2 Adjusted mean levels of serum LDL cholesterol in men and women (aged 18-24 years) according to coffee consumption adjusted for BMI, sporting activity groups and oral contraceptive use (shaded bar: men; white bar: women).
Differences in findings between studies could be explained by differences in dietary assessment methodology, study design, population characteristics and age, and differences in quantity of coffee consumed, strength,
872
L. Kohlmeier ei al.
Table 3
Determinants ofserum LDL cholesterol (mmol.1 ') Regr. coeff. ± SE (T-value)
Regr. coeff. ± SE (T-value) Men
Independent variables Juniors (N= 124) Coffee (cups. day') BMI(kg.m ') Sports (categorical) Fish (g. day') Milk (g. day') Fat (g. day') Tea (cups. day') Age Alcohol (g . day') Smoking (dichotomous)
0·104 ± 0'040 0·070 ± 0·029 -0·126±0·066 0·004 ± 0·003 0·0001 ±0·0003 0·0005 ± 0'002 0-035 ± 0·051 0·015±0·031 - 0·0002 ± 0'004 -0·071 ±0'158 R'=0·15
Seniors (N= 126) (2-61)* (2040)*
- 0·094 ± 0·055 - 0·036 ± 0·029 - 0·0009 ± 0·094 - 0·003 ± 0·004 - 0·0002 ± 0·0007 - 0·002 ± 0·004 -0·220±0·073 (-3·01)** 0·0 I0 ± 0·036 - 0·017 ± 0·005 (- 3'54)*" -0,155 ± 0·213 R'=0·18
Women Juniors (N = 152) Coffee (cups. day') BMI(kg.m ') Sports (categorical) Fish (g. day') Milk (g. day') Fat(g.day') Tea (cups. day') Age Alcohol (g . day ') Smoking (dichotomous) Oral contraceptive use (dichotomous)
0·010 ± 0·039 0·045 ± 0·024 -0·147 ± 0·069 (- 2'13)* - 0·006 ± 0·005 0·0002 ± 0·0005 0·003 ± 0·003 - 0·031 ± 0·047 0·008 ± 0·032 - 0·002 ± 0·008 -0,151 ±0·13\ 0·1321 0·124 R'=0'08
Seniors N = 139) 0·025 ± 0·069 0·001 ±0'026 0·27210·119 0·008 ± 0·006 - 0·002 ± 0·0009 0·003 ± 0·004 - 0·005 ± 0·076 0·020 ± 0·042 -0,019 ±0·013 0·162±0·347
(2'28)*
R'=0'09
*Values presented are significant at a level of P
The relationship between coffee consumption and lipid levels in young and older people in the Heidelberg-Michelstadt-Berlin study L. KOHLMEIER, G.
MENSINK AND M. KOHLMEIER
Institutefor Social Medicine and Epidemiology of the Federal Health Office, Federal Republic ofGermany
KEY WORDS: Cholesterol, coffee, lipoproteins. The relationship between coffee consumption and serum lipid levels was studied in a randomly selected sample of395 young and 385 elderly adults in the Federal Republic of Germany. Analyses were done separately for men and women andfor young and olderpeople. After adjustmentfor body mass index, activity level, smoking, total energy intake, dietary fat.fish, milk, tea and alcohol consumption and oral contraceptive use a statistically significant increase in total serum- and LDLcholesterol levels with increased coffee consumption was observed in the group of young men only. In this subgroup a difference of 0·11 mmol .1- J serum cholesterol and of 0·10 mmol. 1- J LDL cholesterol for each additional cup of coffee consumed daily was calculated. This is one of the few studies that includes elderly people. No significant trend was seen between coffee consumption and serum- or LDL-cholesterollevels in the elderly.
Introduction In 1983 we reported a relationship between serum cholesterollevels and coffee consumption in German men'". In agreement with Scandinavian findings coffee consumption was positively correlated with serum cholesterol level!", The effect of coffee consumption, however, may possibly be attributed to confounding variables such as smoking, dietary habits, physical activity and body mass. This paper presents a closer look at coffee in perspective to other determinants of serum lipid and lipoprotein levels in younger and older people in the Federal Republic of Germany, drawn from regional surveys.
Subjectsandmethods The data are drawn from the Heidelberg-MichelstadtBerlin study on nutrition, lifestyle and health. For this study 395 young (aged 18-24 years) and 385 older (aged 65-74 years) men and women were selected randomly from resident registration lists of three areas of the Federal Republic of Germany'v", A participation rate of 53% was achieved from the original random sample. The major reason given for non-participation was the short stay (2-6 weeks) of the examination unit at one location, at which time many residents were not reachable. The subjects participated in an extensive clinical examination, had blood and urine samples taken for laboratory analyses and completed a detailed questionnaire about their physical activity, nutritional habits, lifestyle and medical history. Additionally, a 7-day dietary and physical activity protocol was recorded. The dietary protocols were self recorded and prospectively conducted. They Submitted for publication on 26 April 1990. and in revised form 2 July 1990. Correspondence: Dr Lenore Kohlmeier, Institut fur Sozialmedizin und Epidemiologie des Bundesgesundheitsamtes, General-Pape-StraBe 62-66 1000 Berlin 42, West Germany.
0195-668Xj91j080869 +06 $03.00/0
were coded with a food code with approximately 1200 foods and food groups'". These protocols formed the basis of the analyses for the effects of dietary fat, alcohol, fish and milk consumption on lipid levels. Coffee consumption (cups of coffee usually consumed per day) was obtained from the questionnaire. Activity level was based on a question about leisure physical activity and was divided into four categories (1. no sporting activities; 2. less than one hour per week; 3. at least one hour per week; 4. competitive sport in a club). For the measurement of lipids and lipoproteins, blood was obtained after an overnight fast, allowed to clot and the serum separated after 45 exactly min. On the same day, all lipid and lipoprotein analyses were completed. Cholesterol and triglycerides in whole serum and in lipoprotein fractions were measured enzymatically. Lipoproteins were prepared quantitatively by micro-ultracentrifugationl'", followed by polyanionic precipitation, comparable to Lipid Research Clinic methodology. Quality control of lipids was carried out according to the requirements of the German 'Bundesarztekammer' and the specifications of the German Clinical Chemistry Society. All external measurements were accurate, with less than a 3% variation. Blind controls were carried out through duplicate samples. Differences of less than 3% on average were found in the double samples from the same venepuncture. Lipoprotein concentrations were calculated from the lipid concentrations in the lipoprotein fractions!". ANALYTICAL MODEL
In a first exploration of lifestyle differences, comparisons were made between heavy (more than three cups per day) and lighter or non coffee drinkers using Student's ttest. A multiple regression modelf ofdeterminants oftotal serum cholesterol, LDL-cholesterol, HDL-cholesterol and apolipoprotein B was designed. This included independent variables such as tea consumption, body mass © 1991 The European Society of Cardiology
870
L. Kohlmeier et al.
Table 1
Frequency distribution of coffee consumption per day Juniors
Seniors
>6cups 5 {j cups 3 -4 cups 1-2 cups < 1 cup Never Missing Total
Men
Women
Men
Women
n
%
n
%
n
%
n
%
7 18 22 52 50 28 7
4·0 10·2 12-4 29·4 28·2 15·8
5 14 58 70 29 30 5
2-4 6·8 28·2 34·0 14·[ 14·6
5 13 68 52 6 22 3
3·0 7-8 41·0 31·3 3-6 13·3
5 20 73 85 13 17 3
2·3 9-4 34·3 39·9 6·1 8·0
184
100
211
100
169
100
216
100
index, alcohol intake, cigarette smoking, physical activity level, fat intake per day in grams, total energy intake and fish and milk consumption, together with the parameter of interest, coffee consumption. The models were run separately for men and women, and for young and older people. In the regression analyses for young women, oral contraceptive use was included as a determinant. Regression equations were calculated but individuals with missing values were not included. All included data were available for 570 of the participants. A high correlation between energy intake and fat intake (in grams) was seen. Regression analysis assumes a linear relationship between coffee consumption and cholesterol, so an additional one way analysis of covariance was conducted using body mass index, physical activity level and (for young women) oral contraceptive use as covariates. Statistical analyses were carried out using the SPSS x software package'?',
Results
In Table I the frequency distribution of coffee consumption of the study population is given, divided by sex and age groups. An examination of the differences in food intake between heavy coffee drinkers (more than three cups per day) and lighter or non coffee drinkers showed significant differences in pork, salad, fruit and schnaps intake among the young men. Heavy coffee drinkers consumed more pork and schnaps but less salad and fruit. It was also noted that heavy coffee drinkers were shorter than the lighter drinkers. Their total fat and total energy intake did not differ from the lighter coffee drinkers. Among young women, milk and soft drink consumption was lower, but wine and champagne intake higher among the heavy coffee drinkers. These women had a lower energy intake, they consumed fewer carbohydrates, minerals, calcium and carbohydrates per day, drank more alcohol and had higher body mass indices than their peers. In Tables 2 and 3 the results of the multiple linear regressions are presented for both gender and age groups. Total energy intake was not included in these equations since they remained insignificant in the earlier models of
cholesterol and energy intake, and are correlated closely with fat intake. It was found that coffee plays a significant unfavourable role in young men only. In this group, coffee consumption remains a substantial determinant ofcholesterol and LO Lcholesterol levels, together with body mass index and lack of sport participation. For young men a difference of 0·109 mmol . 1- I (SO 0,044) of total cholesterol and 0·104 mmol . I-I (SO 0'040) LOL-cholesterol per cup of coffee was seen. The apolipoprotein B levels were also associated with coffee consumption, with an increase of 0·029 g . 1 I for each cup of coffee consumed per day (data not presented). The extent of the effect on this group of young men was surprisingly high. The results of the analysis of covariance show a dose-response relationship between coffee intake and serum cholesterol and LOL-cholesterollevels (Figs I and 2). Among young women, an effect of coffee consumption was not seen in relation to cholesterol and LOL cholesterol levels. The major determinants of cholesterol and LOL cholesterol among young women were body mass index, sport participation, and oral contraceptive use; all except sport operated unfavourably. Mean serum cholesterol and LOL cholesterol levels in women at all levels of coffee intake were as high as those of the men consuming more than four cups of coffee a day (Fig. I). Similarly, among older people, there was no evident effect between current coffee intake and lipid levels. Coffee consumption was negatively correlated (although not statistically significant) with lipid levels, as was tea consumption. Tea consumption is associated with healthier levels of both cholesterol, LOL cholesterol and apolipoprotein B in elderly men. The magnitude of effect corresponds to a reduction of O'240 mmol . 1- I of total serum cholesterol (0·220 mmol l I LOL cholesterol) per cup of tea. Alcohol intake was the only other major determinant of serum lipids and lipoproteins in both genders. i
Discussion
Numerous studies have examined, in experimental or cross sectional design, possible effects of coffee consumption on serum lipid levels. In cross sectional studies an
Relationship between coffee consumption and lipid levels 871
Table 2
Determinants a/serum cholesterol (mmol. I J)
Regr. coelT. ± SE (Tvvalue)
Regr. coelT. ± SE (T-value) Men
Independent varia bles Juniors (N= 124)
Seniors (N= (26)
0·109± 0·044 (2'49)* 0·078 ± 0·032 (2AW -0·159±0·073 (2'18)* 0·004 ± 0·004 -I x 10 '±0·0003 - 0·0002 ± 0·0022 0·042 ± 0·056 0·022 ± 0·035 0·003 ± 0·004 -0·090±0·175 R'=0·18
Coffee (cups. day ') BMI (kg. m ') Sports (categorical) Fish (g. day') Milk (g. day') Fat (g. day') Tea (cups. day') Age Alcohol (g . day') Smoking (dichotomous)
- 0·088 ± 0·063 - 0·045 ± 0·032 -0·010±0·107 - 0·002 ± 0·004 - 0·0005 ± 0·0008 - 0·002 ± 0·004 - O'240 ± 0·083 (-2·90)" 0·023 ± 0·040 -0·01210·005 (- 2'25)* -0·245±0·241 R'=0·13 Women
Juniors (N = (52) Coffee (cups. day') BMI(kg.m ') Sports (categorical) Fish (g . day') Milk (g. day') Fat(g. day') Tea (cups. day ') Age Alcohol (g. day') Smoking (dichotomous) Oral contraceptive use (dichotomous)
Seniors (N= 139)
0·031 ± 0·045 0·036 ± 0·028 - 0·048 ± 0·079 - 0·009 ± 0·006 - 0·0005 ± 0·0006 0·005 ± 0·003 - 0·069 ± 0·054 0·028 ± 0·036 - 0·003 ± 0·009 - 0·128 ± 0·151 0·350 ± 0·142 (2'46)* R'=O·1O
- 0·008 ± 0·071 - 0·0 1010·027 (2'57)* 0·320 ± 0·124 - 0·005 ± 0·006 - 0·002 ± 0·0009 (-2'16)* 0·003 ± 0·004 -0,061 ±0·080 0·026 ± 0·044 - 0·025 ± 0·0 13 0·094 ± O'362
R'=O'1O
*Values presented are significant at a level of P < 0·05. **Values presented are significant at a level of P< 0·0 I. 5'0
4'0
r
4'8 t-'
'~
';=
.-3·5 f-
4·6 f-
'~
c;
c;
E E 4-4
r-
E E
2'5 1-.
4'2 f-.
4·0
r-
3'0
L..-
L...--
< 1 cup
1-2 cups
3-4 cups
>4 cups
Figure I Adjusted mean levels of serum total cholesterol in men and women (aged 18-24 years) according to coffee consumption adjusted for BMI, sporting activity groups and oral contraceptive Use (shaded bar: men; white bar: women).
effect was seen in both sexes!'? 121, in women only'!", in men only[l4.'5) or in neither'!", Experimental findings Support the hypothesis that coffee consumption affects lipid levels'!".
2·0 '---< 1 cup
'---
1-2 cups
3-4 cups
>4 cups
Figure 2 Adjusted mean levels of serum LDL cholesterol in men and women (aged 18-24 years) according to coffee consumption adjusted for BMI, sporting activity groups and oral contraceptive use (shaded bar: men; white bar: women).
Differences in findings between studies could be explained by differences in dietary assessment methodology, study design, population characteristics and age, and differences in quantity of coffee consumed, strength,
872
L. Kohlmeier ei al.
Table 3
Determinants ofserum LDL cholesterol (mmol.1 ') Regr. coeff. ± SE (T-value)
Regr. coeff. ± SE (T-value) Men
Independent variables Juniors (N= 124) Coffee (cups. day') BMI(kg.m ') Sports (categorical) Fish (g. day') Milk (g. day') Fat (g. day') Tea (cups. day') Age Alcohol (g . day') Smoking (dichotomous)
0·104 ± 0'040 0·070 ± 0·029 -0·126±0·066 0·004 ± 0·003 0·0001 ±0·0003 0·0005 ± 0'002 0-035 ± 0·051 0·015±0·031 - 0·0002 ± 0'004 -0·071 ±0'158 R'=0·15
Seniors (N= 126) (2-61)* (2040)*
- 0·094 ± 0·055 - 0·036 ± 0·029 - 0·0009 ± 0·094 - 0·003 ± 0·004 - 0·0002 ± 0·0007 - 0·002 ± 0·004 -0·220±0·073 (-3·01)** 0·0 I0 ± 0·036 - 0·017 ± 0·005 (- 3'54)*" -0,155 ± 0·213 R'=0·18
Women Juniors (N = 152) Coffee (cups. day') BMI(kg.m ') Sports (categorical) Fish (g. day') Milk (g. day') Fat(g.day') Tea (cups. day') Age Alcohol (g . day ') Smoking (dichotomous) Oral contraceptive use (dichotomous)
0·010 ± 0·039 0·045 ± 0·024 -0·147 ± 0·069 (- 2'13)* - 0·006 ± 0·005 0·0002 ± 0·0005 0·003 ± 0·003 - 0·031 ± 0·047 0·008 ± 0·032 - 0·002 ± 0·008 -0,151 ±0·13\ 0·1321 0·124 R'=0'08
Seniors N = 139) 0·025 ± 0·069 0·001 ±0'026 0·27210·119 0·008 ± 0·006 - 0·002 ± 0·0009 0·003 ± 0·004 - 0·005 ± 0·076 0·020 ± 0·042 -0,019 ±0·013 0·162±0·347
(2'28)*
R'=0'09
*Values presented are significant at a level of P
