17
Atherosclerosis, 30 (1978) 17-25 Scientific 0 Elsevier/North-Holland
CIGARETTE
Publishing,
Ltd.
SMOKING AND HDL CHOLESTEROL
The Framingham
Offspring
Study
R.J. GARRISON, W.B. KANNEL, and S.J. PADGETT
M. FEINLEIB,
W.P. CASTELLI,
P.M. McNAMARA
Epidemiology Branch, Division of Heart and Vascular Diseases, National Heart, Lung, and Blood Institute, National Institutes of Health, Bethesda, Md. 20014 (U.S.A.) (Received 31 October, 1977) (Revised, received 26 January, 1978) (Accepted 27 January, 1978)
Summary High density lipoproteins were measured on fasting blood samples from 4107 men and women. Control for reported alcohol consumption and subscapular skinfold thickness using multiple regression analysis allowed an examination of the relationship between cigarette smoking and HDL cholesterol. Cigarette smoking was found to be associated with an average difference in HDL cholesterol of about 4 mg/dl in men and 6 mg/dl in women. Furthermore, when heavy alcohol drinkers were eliminated a significant negative association between number of cigarettes smoked and HDL cholesterol was demonstrable in both men and women. There was no evidence that former cigarette smokers, with the exception of those who switched to cigars or pipes or had quit less than one year, had lower HDL levels. Cigar or pipe smokers who had never smoked cigarettes had alcohol- and skinfold-adjusted HDL cholesterol comparable to the non-smoker. These observations indicate another possible link between inhaled tobacco smoke and the atherosclerotic process and suggest the need for further studies and experiments that might clarify the mutual relationship of HDL cholesterol, cigarette smoking and the atherosclerotic process. Key words:
Alcohol - Atherosclerosis
- Cigarette smoking - HDL cholesterol
- Obesity
Introduction Improved laboratory techniques for measurement of the cholesterol carried by high density lipoproteins (HDL) have led to wide use of this measurement in
18
epidemiological studies and increased attention to HDL as a potential coronary heart disease (CHD) inverse risk factor. Barr [l] first suggested that coronary patients have lower levels of HDL and recent evidence [ 2-41 indicates a strong inverse association between HDL levels and CHD incidence. Attempts to find factors that are associated with HDL level are a natural next step in efforts to describe the etiology of atherosclerosis. Strong negative associations with obesity [ 51 and positive associations between HDL and alcohol consumption [6] have previously been demonstrated. This report assesses the relationship of cigarette smoking to HDL, after adjustment for obesity and alcohol intake, in a sample of 4,107 men and women aged 20-49 years. Methods Children of participants in the Framingham Heart Study were invited to a physical examination at the Framingham Heart Study facilities. For this report all children of the cohort and their spouses were considered, including those with only one parent in the cohort. A detailed description of the study design is available elsewhere [ 71. Blood samples were drawn after an overnight 14-h fast. HDL cholesterol was measured by heparin/manganese-chloride precipitation of the infranate from the blood sample spun in the preparative ultracentrifuge. The protocol established by the Lipid Research Clinics [8] was followed and duplicates run at LRC laboratories indicate very similar results. Each individual’s height, body weight and subscapular skinfolds were measured by a nurse. Subscapular skinfolds were measured at the inferior angle of the scapula using the Lange caliper. Alcohol consumption quantities and smoking histories were obtained in a standardized manner. The physician recorded the reported usual consumption of cans of beer, glasses of wine, and cocktails per week. Smoking history included questions on age cigarette smoking was started and stopped. Smokers were considered to be former smokers only if they had not smoked for more than a year. Questions about filter use were included as well as those asking about cigar and pipe usage. The association between alcohol consumption and cigarette use together with a positive association between alcohol intake and HDL levels make control for alcohol consumption essential when exploring the relationship between HDL and smoking. While the negative association of obesity and smoking is not large, the strong negative association between HDL and obesity indicates a need to control for obesity as well. Step-down multiple regression analysis was used to test for significant relationships between smoking habit and HDL cholesterol, after adjustment for the associations between HDL and alcohol, obesity and age. In step-down multiple regression analysis, the independent variable with the smallest partial correlation coefficient is the first to be eliminated. The multiple regression is then recomputed without this variable and variables are eliminated sequentially in this manner until all remaining independent variables have significant (P < 0.05) partial correlations. The removal of an independent variable after this step would result in a statistically significant reduction in the predictive power of the regression equation.
19
Results Results are reported for 2149 women and 1958 men between the age of 20 and 49 years who had complete information on smoking history, alcohol consumption and obesity. Simple correlation coefficients between HDL cholesterol, cigarette smoking, measures of alcohol consumption and obesity are given for all ages in Table 1. No large deviations from these values were found in any of the lo-year age groups. These correlations indicate the need to control for alcohol consumption and obesity when examining the relationship between cigarette smoking and HDL levels. Graphical displays of the univariate relationships between HDL and cigarette smoking, alcohol intake and skinfold thickness are shown in Figs. 1, 2 and 3, respectively. In order to show a more realistic picture of these associations heavy drinkers (16+ alcoholic beverages/ wk) have been eliminated from Fig. 1 while in Figs. 2 and 3 only individuals who do not smoke cigarettes are considered. Tables 2 and 3 list partial regression coefficients and t-statistics for the final step of the analysis in which the listed variables entered the first step. For this analysis, current cigarette smokers were assigned a value of one and non-smokers a value of zero. For all 6 agesex groups cigarette smoking has a significant partial regression coefficient. These partial regression coefficients show the average difference in HDL level associated with cigarette smoking to be lower by about 3-4 mg/dl in men and 5-6 mg/dl in women. When heavy drinkers (16+ drinks/wk) are eliminated, the average difference is 4.5 mg/dl and 6.5 mg/dl in men and women, respectively. Notice also that for each analysis at least one and most often two measures of alcohol intake stay in the regression model with the coefficients showing no large heterogeneity within sex. In men, beer consumption seems to be most strongly associated with HDL. In addition, at least one measure of obesity always stays in the model. Subscapular skinfold thickness and Wt/Ht2 show nearly equal ability to predict HDL level. The correlation of these two measures with each other is 0.77 and 0.84 for men and women, respectively. To simplify the further analyses, beer, wine and weekly cocktail consumption were added to give one measure of alcoholic beverage intake and skinfold thickness was used as the obesity measure.
TABLE
1
CORRELATION
COEFFICIENTS
FOR
MEN
AND
WOMEN
AGED
20-49
Men (n = 1958) HDL
(mg/dl)
Women cigarette smoking
Beer
(12
Wine
(glasses/wk)
oz.
Cocktails Skinfold Wt/ht2 Age
cans/wk)
(per wk) (lbs/in*)
(years)
a Individuals
who
do
smoke
cigarette
0.086
0.096
0.110
0.098
0.085
(mg/dl)
smoking
-0.041
0.092
(mm)
HDL
(n = 2149)
a
0.107
0.190
YEARS
-0.016
0.105
0.138
-0.225
-0.098
-0.240
-0.057
-0.216
-0.057
-0.198
-0.064
4.013
-0.079
cigarettes
are coded
= 1 and
those
0.094 who
do not
are coded
4.055 = 0.
a
20
CIGARETTES
SMOKED
PER DAY
NONE
Atherosclerosis, 30 (1978) 17-25 Scientific 0 Elsevier/North-Holland
CIGARETTE
Publishing,
Ltd.
SMOKING AND HDL CHOLESTEROL
The Framingham
Offspring
Study
R.J. GARRISON, W.B. KANNEL, and S.J. PADGETT
M. FEINLEIB,
W.P. CASTELLI,
P.M. McNAMARA
Epidemiology Branch, Division of Heart and Vascular Diseases, National Heart, Lung, and Blood Institute, National Institutes of Health, Bethesda, Md. 20014 (U.S.A.) (Received 31 October, 1977) (Revised, received 26 January, 1978) (Accepted 27 January, 1978)
Summary High density lipoproteins were measured on fasting blood samples from 4107 men and women. Control for reported alcohol consumption and subscapular skinfold thickness using multiple regression analysis allowed an examination of the relationship between cigarette smoking and HDL cholesterol. Cigarette smoking was found to be associated with an average difference in HDL cholesterol of about 4 mg/dl in men and 6 mg/dl in women. Furthermore, when heavy alcohol drinkers were eliminated a significant negative association between number of cigarettes smoked and HDL cholesterol was demonstrable in both men and women. There was no evidence that former cigarette smokers, with the exception of those who switched to cigars or pipes or had quit less than one year, had lower HDL levels. Cigar or pipe smokers who had never smoked cigarettes had alcohol- and skinfold-adjusted HDL cholesterol comparable to the non-smoker. These observations indicate another possible link between inhaled tobacco smoke and the atherosclerotic process and suggest the need for further studies and experiments that might clarify the mutual relationship of HDL cholesterol, cigarette smoking and the atherosclerotic process. Key words:
Alcohol - Atherosclerosis
- Cigarette smoking - HDL cholesterol
- Obesity
Introduction Improved laboratory techniques for measurement of the cholesterol carried by high density lipoproteins (HDL) have led to wide use of this measurement in
18
epidemiological studies and increased attention to HDL as a potential coronary heart disease (CHD) inverse risk factor. Barr [l] first suggested that coronary patients have lower levels of HDL and recent evidence [ 2-41 indicates a strong inverse association between HDL levels and CHD incidence. Attempts to find factors that are associated with HDL level are a natural next step in efforts to describe the etiology of atherosclerosis. Strong negative associations with obesity [ 51 and positive associations between HDL and alcohol consumption [6] have previously been demonstrated. This report assesses the relationship of cigarette smoking to HDL, after adjustment for obesity and alcohol intake, in a sample of 4,107 men and women aged 20-49 years. Methods Children of participants in the Framingham Heart Study were invited to a physical examination at the Framingham Heart Study facilities. For this report all children of the cohort and their spouses were considered, including those with only one parent in the cohort. A detailed description of the study design is available elsewhere [ 71. Blood samples were drawn after an overnight 14-h fast. HDL cholesterol was measured by heparin/manganese-chloride precipitation of the infranate from the blood sample spun in the preparative ultracentrifuge. The protocol established by the Lipid Research Clinics [8] was followed and duplicates run at LRC laboratories indicate very similar results. Each individual’s height, body weight and subscapular skinfolds were measured by a nurse. Subscapular skinfolds were measured at the inferior angle of the scapula using the Lange caliper. Alcohol consumption quantities and smoking histories were obtained in a standardized manner. The physician recorded the reported usual consumption of cans of beer, glasses of wine, and cocktails per week. Smoking history included questions on age cigarette smoking was started and stopped. Smokers were considered to be former smokers only if they had not smoked for more than a year. Questions about filter use were included as well as those asking about cigar and pipe usage. The association between alcohol consumption and cigarette use together with a positive association between alcohol intake and HDL levels make control for alcohol consumption essential when exploring the relationship between HDL and smoking. While the negative association of obesity and smoking is not large, the strong negative association between HDL and obesity indicates a need to control for obesity as well. Step-down multiple regression analysis was used to test for significant relationships between smoking habit and HDL cholesterol, after adjustment for the associations between HDL and alcohol, obesity and age. In step-down multiple regression analysis, the independent variable with the smallest partial correlation coefficient is the first to be eliminated. The multiple regression is then recomputed without this variable and variables are eliminated sequentially in this manner until all remaining independent variables have significant (P < 0.05) partial correlations. The removal of an independent variable after this step would result in a statistically significant reduction in the predictive power of the regression equation.
19
Results Results are reported for 2149 women and 1958 men between the age of 20 and 49 years who had complete information on smoking history, alcohol consumption and obesity. Simple correlation coefficients between HDL cholesterol, cigarette smoking, measures of alcohol consumption and obesity are given for all ages in Table 1. No large deviations from these values were found in any of the lo-year age groups. These correlations indicate the need to control for alcohol consumption and obesity when examining the relationship between cigarette smoking and HDL levels. Graphical displays of the univariate relationships between HDL and cigarette smoking, alcohol intake and skinfold thickness are shown in Figs. 1, 2 and 3, respectively. In order to show a more realistic picture of these associations heavy drinkers (16+ alcoholic beverages/ wk) have been eliminated from Fig. 1 while in Figs. 2 and 3 only individuals who do not smoke cigarettes are considered. Tables 2 and 3 list partial regression coefficients and t-statistics for the final step of the analysis in which the listed variables entered the first step. For this analysis, current cigarette smokers were assigned a value of one and non-smokers a value of zero. For all 6 agesex groups cigarette smoking has a significant partial regression coefficient. These partial regression coefficients show the average difference in HDL level associated with cigarette smoking to be lower by about 3-4 mg/dl in men and 5-6 mg/dl in women. When heavy drinkers (16+ drinks/wk) are eliminated, the average difference is 4.5 mg/dl and 6.5 mg/dl in men and women, respectively. Notice also that for each analysis at least one and most often two measures of alcohol intake stay in the regression model with the coefficients showing no large heterogeneity within sex. In men, beer consumption seems to be most strongly associated with HDL. In addition, at least one measure of obesity always stays in the model. Subscapular skinfold thickness and Wt/Ht2 show nearly equal ability to predict HDL level. The correlation of these two measures with each other is 0.77 and 0.84 for men and women, respectively. To simplify the further analyses, beer, wine and weekly cocktail consumption were added to give one measure of alcoholic beverage intake and skinfold thickness was used as the obesity measure.
TABLE
1
CORRELATION
COEFFICIENTS
FOR
MEN
AND
WOMEN
AGED
20-49
Men (n = 1958) HDL
(mg/dl)
Women cigarette smoking
Beer
(12
Wine
(glasses/wk)
oz.
Cocktails Skinfold Wt/ht2 Age
cans/wk)
(per wk) (lbs/in*)
(years)
a Individuals
who
do
smoke
cigarette
0.086
0.096
0.110
0.098
0.085
(mg/dl)
smoking
-0.041
0.092
(mm)
HDL
(n = 2149)
a
0.107
0.190
YEARS
-0.016
0.105
0.138
-0.225
-0.098
-0.240
-0.057
-0.216
-0.057
-0.198
-0.064
4.013
-0.079
cigarettes
are coded
= 1 and
those
0.094 who
do not
are coded
4.055 = 0.
a
20
CIGARETTES
SMOKED
PER DAY
NONE
