Impact of Alcohol Intake on Measures of Lipid Metabolism Depends on Context Defined by Gender, Body Mass Index, Cigarette Smoking, and Apolipoprotein E Genotype
Hyperlipidemia, smoking, and obesity are well-known risk factors for cardiovascular disease. Conversely, moderate alcohol intake is associated with lower atherosclerosis risk. However, the influence of taking alcohol on the interrelationships of these factors in a particular context has not been thoroughly investigated. In this study, we asked whether the association between plasma measures of lipid metabolism and alcohol intake is dependent on context defined by gender, age, body mass index (BMI), smoking, and apolipoprotein E (APOE) genotype. Data were obtained in a sample of 869 women and 824 men who participated in the Quebec Heart Health Survey. There was no evidence that variation among APOE genotypes influenced the association between LDL cholesterol (LDL-C) or HDL cholesterol (HDL)-C and alcohol, after adjustment for age and BMI. Further, the positive (LDL-C and BMI) and the negative (HDL-C and BMI) associations that were observed in men and women with the ε3/2 and ε3/3 genotypes were not modified by alcohol intake. However, in women with the ε4/3 genotype only, we found a significant influence of an alcohol by BMI interaction on the prediction of total cholesterol, LDL-C, HDL-C, apoA-I, and apoB, and this interaction was influenced by the status of smoking. Whereas the influence of an alcohol by BMI interaction on total cholesterol and LDL-C was significant in smokers, its influence on HDL-C was significant only in non-smokers. This study emphasizes the context dependency of the influence of alcohol on lipid metabolism and demonstrates how biological, environmental, and genetic factors interact to determine cardiovascular disease risk.
Hyperlipidemia is a major risk factor for cardiovascular disease (CVD).1,2⇓ Plasma lipid, lipoprotein, and apolipoprotein levels are influenced by biological indices (ie, age and weight), environmental (diet, alcohol, smoking, and physical activity), and genetic variations. Although the separate influence of each of these measures of individuality on the prediction of plasma levels of lipid traits has been broadly studied, their combined effects are less understood.
Several studies have evaluated the influence of alcohol intake, an index of lifestyle, on plasma lipid and lipoprotein concentrations (reviewed by Savolainen and Kesäniemi3). A moderate intake of alcohol is associated with protection against coronary heart disease, probably due in part to a dose-dependent increase in HDL cholesterol (HDL-C).4 A decrease in LDL cholesterol (LDL-C) with increased alcohol intake has also been reported in some studies, but this effect is less consistent and probably depends on the combination of one or more unmeasured factors.3 Until recently, most cross-sectional studies of the influence of alcohol have been performed in men or in pooled samples of women and men. Furthermore, most studies have been conducted without control for potential genetic or environmental factors.
In studies of the effects of alcohol on lipoprotein metabolism, smoking has been identified as a key interacting factor.5,6⇓ Cigarette smoking generally increases total cholesterol (TC) and triglyceride (TG), VLDL cholesterol (VLDL-C) and LDL-C levels and decreases HDL-C and apolipoprotein (apo) A-I levels.7 Thus, alcohol intake and cigarette smoking exert opposite effects on LDL and HDL metabolism. A few studies, conducted mostly in men, have evaluated the interaction of these habits on blood lipids. Generally, it is expected that the effects of one could be modified by the influence of the other.8,9⇓
The influence of variation in the gene coding for apoE on interindividual variation in the regulation of lipid metabolism has been abundantly studied.10 Individuals who carry the ε4 allele generally have the highest and those with the ε2 allele the lowest levels of LDL-C. The few studies that have evaluated the impact of the interaction between measures of lifestyle and the apoE polymorphism on lipid traits11–13⇓⇓ have provided inconsistent results. Work from our group has shown that the influence of an interaction between age and anthropometric status on interindividual variation in blood lipid levels is dependent on APOE genotype.14–17⇓⇓⇓ The aim of the present study was to extend these studies to determine the degree to which the association between plasma lipids, lipoproteins and apolipoproteins and alcohol intake is dependent on the context of the individual defined by gender, age, body mass index (BMI), smoking status, and APOE genotype.
Details of the sampling design have been previously published.17 Briefly, subjects were women and men aged 18 to 74 years participating in two complementary studies in the Province of Quebec (Canada), the Heart Health Survey and the Nutrition Survey, during the fall of 1990. During an auxiliary study of plasma apolipoproteins, apoA-I and apoB levels were determined, and apoE phenotyping was performed. Finally, data from the 3 studies were combined into a sample of 2010 subjects (1025 women and 985 men). Subjects with missing lipid measures or lifestyle data, as well as those presenting with the less common apoE phenotypes E2/2, E4/2, and E4/4 (≈5%) were excluded from the study, leaving a sample of 869 women and 824 men.
Anthropometric data (height and weight, waist and hip circumferences) and information on lifestyle habits were obtained during the clinic visit and the interview as part of the Heart Health Survey. BMI was calculated as weight in kilograms divided by height in meters squared. Plasma TC, total TG, HDL-C, apoA-I, and apoB concentrations were determined from the fasting blood sample. All of these assays were carried out under standardized conditions at the Lipid Research Laboratory of the University of Toronto, as previously described.18,19⇓ LDL-C was calculated by using the equation of Friedewald et al,20 and VLDL-C was derived by subtracting LDL-C and HDL-C from TC. Thus, subjects with triglyceride levels ≥5.0 mmol/L were excluded in our study. ApoE phenotyping was performed at the Clinical Research Institute of Montreal, according to the method described by Hill and Pritchard.21 Six APOE genotypes, ε2/2, ε3/2, ε3/3, ε4/3, ε4/2, and ε4/4, defined by the three polymorphic alleles, ε2, ε3, and ε4, were inferred from the isoform phenotypes E2/2, E3/2, E3/3, E4/3, E4/2, and E4/4.
In the Heart Health Survey, subjects were asked to provide information on their daily alcohol intake (number of drinks) during the 7 days preceding their clinic visit. In our sample, as a large proportion of subjects were non-drinkers (44% of women and 30% of men), we chose to study the influence of alcohol intake on lipid traits by dividing subjects into 2 categories: non-drinkers (number of drinks/d=0) and drinkers (number of drinks/d >0). Information on cigarette smoking habits was also collected during the clinic visit. Subjects were asked to report the number of cigarettes usually smoked in a day. Occasional cigarette smokers and pipe or cigar smokers were considered non-smokers. Because a large number of subjects were non-smokers (70% of women and 73% of men), we also chose to evaluate the influence of smoking as a discrete variable: non-smokers (number of cigarettes/d=0) and smokers (number of cigarettes/d >0).
All analyses were performed separately for women and men by using the SAS software (SAS Institute Inc, 1993, Version 6). To reduce positive skewness, the log (base 10) transformation of triglyceride and VLDL-C values was used in all analyses. The means of all variables were first calculated on the raw data, for each of the 3 most common APOE genotype subgroups (ε3/2,ε3/3, and ε4/3). By using the t test, statistical significance of the difference between non-drinkers and drinkers was evaluated separately for each APOE genotype. The F statistic was used to test the hypothesis of equality of subgroup variances. Satterthwaite’s approximation of the t test was applied when subgroup variances were significantly different.22 The χ2 contingency test was used to assess differences in the relative frequencies of smokers between non-drinkers and drinkers.
Separate regression equations were estimated for women and men, as the distributions of lipid traits and relationships with most other traits are known to be gender specific.14,17,23,24⇓⇓⇓ Four regression models (denoted heretoforth as models A–D) were used to evaluate the contribution of each variable (age, BMI, alcohol) and interactions to lipid trait variability in each APOE genotype subgroup separately. First, model A evaluated the contribution of age (to the third order) and BMI to lipid trait variability. In model B|A, the additional contribution of alcohol intake (as a discrete variable) to lipid trait variation was evaluated, after adjustment for age and BMI. Next, in model C|AB, the additional contribution to lipid trait variation of the interaction between age and alcohol intake was estimated, after adjustment for age, BMI, and alcohol intake. In model D|ABC, the additional contribution to lipid trait variation of the interaction between BMI and alcohol intake was considered, after adjustment for age, BMI, alcohol intake, and the interaction of age with alcohol intake. Model D, the complete model, was also applied separately in non-smokers and smokers, to evaluate the combined effects of drinking and smoking habits on lipid traits.
Finally, to determine whether results of the regression analyses were significantly different among APOE genotype groups, a test of homogeneity was performed for each regression model by using the F ratio.25 The F ratio statistic, as defined by F=[SSR (complete model, including interaction terms between variables and APOE)−SSR (reduced model, excluding interaction terms with APOE)/degrees of freedom associated with the difference]/MSE (complete model), where SSR and MSE are sums of squares due to regression and mean square error, respectively, is simply a test of the interaction between APOE genotypes and variables in the regression model considered. Whenever statistically significant heterogeneity was detected, pairwise comparisons of the ε3/2 or the ε4/3 genotype with the most common genotype ε3/3 were carried out. In this study, the 0.05 level of probability was considered the criterion for significance of a test statistic.
For the characteristics of women and men according to APOE genotype and drinking status, please see Table I (available at www.atvb.ahajournals.org). Subjects consuming alcohol were generally younger and female drinkers had a lower BMI than non-drinkers. The number of daily drinks consumed was comparable in the 3 APOE genotype subgroups in both women and men. However, the percentage of smokers among drinkers, compared with the percentage among non-drinkers, was significantly greater in women with the ε4/3 genotype. The lipid traits, TC, LDL-C, apoB, VLDL-C, and TG were significantly lower in female drinkers than in non-drinkers in the ε3/3 and ε4/3 subgroups, but there were no differences in men. In the 3 APOE genotype subgroups, HDL-C and apoA-I levels were significantly higher in women who consumed alcohol than in those who abstained, whereas HDL-C in men was significantly increased in drinkers compared with non-drinkers only in those with the ε3/3 genotype, while apoA-I levels were significantly higher in drinkers in all 3 APOE genotype subgroups.
The influence of age, BMI, and alcohol intake was estimated by using multiple regression models in women (Table 1) and men (Table 2). These tables show the percentage of sample variability in lipid traits associated with each regression model (R2×100) and the corresponding level of significance (P value) in each of the 3 APOE genotypes. In women and men, age and BMI (model A) made the largest contribution to trait variability. The R2×100 values for model A varied from 6% (apoA-I) to 42% (TC) in women (Table 1) and from 2% (apoA-I) to 28% (TC) in men (Table 2). This influence of age and BMI was significantly heterogeneous among APOE genotype subgroups in men for TC, LDL-C, and apoB (ε3/2 versus ε3/3, Table 3), but not in women.
In model B‖A, the influence of alcohol intake on lipid traits was determined after adjustment for age and BMI. In women (Table 1), alcohol consumption contributed significantly to the variability in HDL-C (R2×100=1% to 4%) and apoA-I (R2×100=1% to 11%) in each of the 3 APOE genotype subgroups. The influence of alcohol consumption on the variability of apoA-I was significantly greater in women with the ε3/2 genotype (R2×100=11%) compared with ε3/3 (R2×100=1%) (test of homogeneity, Table 3). In men (Table 2), the contribution of alcohol intake to HDL-C variability was significant in the ε3/3 subgroup only (R2×100=3%), whereas the impact on apoA-I was significant in all 3 APOE genotypes (R2×100 was ≈4% in each). There was no significant heterogeneity for these effects among the APOE genotypes in men (Table 3). Except for LDL-C and apoB in women with the ε3/3 genotype, the influence of alcohol intake was not significant on any of the other traits. In both cases, the contribution was less than 1%.
Our analysis strategy also included testing for the contribution of interactions between alcohol intake and age or BMI to variation in lipid traits within each genotype subgroup. There were no significant statistical interactions between age and alcohol (model C‖AB), in women (Table 1) or men (Table 2). However, the influence of a statistically significant interaction between alcohol and BMI (model D‖ABC) was observed for TC, LDL-C, HDL-C, apoA-I and apoB in women with the ε4/3 genotype, for log VLDL-C and log TG in the ε3/3 women subgroup (Table 1), and for apoA-I and apoB in men with the ε3/3 and the ε4/3 genotypes, respectively (Table 2). The tests of homogeneity revealed significant differences in women between the ε4/3 and the ε3/3 genotypes in the influence of the alcohol by BMI interaction on interindividual variation in TC, LDL-C, HDL-C, and apoB. None of these tests were statistically significant in men (Model D‖ABC, Table 3).
Figure 1 illustrates the heterogeneity among APOE genotypes, in women, of the influence of the interaction between alcohol and BMI on LDL-C. No significant interaction was observed for women with the ε3/2 and ε3/3 genotypes, but in women with the ε4/3 genotype, increasing LDL-C concentrations were strongly associated with increasing BMI in drinkers but not in non-drinkers. To minimize the possible influence of outliers on this relationship, values of any of the traits exceeding 4 SDs above and below the mean were removed and the analyses of women repeated. Thirteen subjects were removed because of outlying values for one or more of the traits. The censored and balanced sample included 856 instead of 869 women. For TC, LDL-C, and apoB, exclusion of these women lowered the observed level of significance (P value) of the differences, but those statistically significant differences observed in the larger data set for LDL-C, HDL-C, and apoB were retained, and apoA-I became significant at the 0.05 level of probability. Henceforth, while uncensored data are shown in Table 3, censored data are used in Figures 2 and 3⇓.
Next, we asked whether smoking had an impact on the influence of the interaction between alcohol intake and BMI on lipid traits. The contribution of the interaction between alcohol and BMI to the variability of TC and LDL-C in women was found to be significantly different among genotype subgroups in smokers only (Table 3). Figure 2 shows that in ε4/3 women who smoked, higher LDL-C levels were not associated with higher BMI values in the absence of alcohol, in contrast to the strong positive relationship observed in female smokers who consumed alcohol. However, although similar trends were observed in women who did not smoke, taking or not taking alcohol did not significantly influence the statistical relationship between LDL-C and BMI. Furthermore, the heterogeneity in the association between apoB and BMI observed between the ε4/3 and the ε3/3 genotypes for female non-smokers in the total sample (Table 3) was no longer significant in the censored sample (not shown). For HDL-C, the contribution of the interaction between alcohol and BMI to trait variability was found to be statistically different among APOE genotypes only in non-smokers (Table 3). Specifically, as illustrated in Figure 3, a significant decrease in HDL-C levels was associated with increasing BMI only in ε4/3 female non-smokers who consumed alcohol, in contrast to the negative association between HDL-C and BMI observed in women who smoked, consuming alcohol or not. Similar results were observed in the analysis of apoA-I (not shown).
We previously reported that the relationships between lipid traits and age, BMI, or waist-to-hip ratio are influenced by APOE genotype.17 In the current study of the same sample, our objective was to evaluate the extent to which the associations between lipid traits and alcohol intake are influenced by the apoE polymorphism, while taking into account other measures of context, ie, gender, age, body size, and cigarette smoking. Although this statistical work does not provide any information about the molecular and physiological mechanisms involved in determining the complex biological relationships among these traits, it provides valuable insight into how these established risk factors might be best used to improve the accuracy of predicting an individual’s risk of CVD. In this regard, we observed the widely documented elevating effect of alcohol on HDL-C and apoA-I levels.3 We found that after adjusting for age and BMI, this effect was not dependent on APOE genotype when other measures of context were ignored (model B). However, further subgroup analyses provided convincing evidence for heterogeneity among APOE genotypes of the influence of the interactions between alcohol, BMI, and smoking on variation in levels of both LDL-C and HDL-C. Specifically, we found that the increase in LDL-C with increasing BMI was significantly greater in drinkers than in non-drinkers and this complex interaction was observed only in women with the ε4/3 genotype (Figure 1). This apparent synergistic effect of alcohol intake and increasing BMI on LDL-C levels showed even more complexity when the influence of smoking was considered. The interaction was stronger in smokers than in non-smokers (Figure 2). In these same women with the ε4/3 genotype, the decrease in HDL-C with increasing BMI was greater in drinkers compared with non-drinkers, especially in non-smokers (Figure 3). We conclude that in ε4/3 women only, alcohol consumption intensifies the expected increase in LDL-C and decrease in HDL-C associated with increasing BMI. The influence of the alcohol by BMI interaction on LDL-C is greatest in smokers while the effect on HDL-C is greatest in non-smokers.
The consideration of the role of age, BMI, smoking, physical activity, social economic status, energy intake, or menopausal status on the effect of alcohol has not been consistently addressed previously. Only a few studies have evaluated the influence of the apoE polymorphism on the contribution of alcohol intake to variation in measures of lipid metabolism and they differ greatly in their objectives and the analytical methods that were used.11,12,26,27⇓⇓⇓ Statistical approaches have included multiple linear regression analyses11,26⇓ or adjusted correlation coefficients12 in separate, or pooled,27 APOE subgroups. One recent study27 found that increased levels of LDL-C in carriers of the ε4 allele and decreased levels in the carriers of the ε2 allele were present in both drinkers and non-drinkers in women but were observed only in drinkers in men. This gender-specific effect of APOE genotype variation remained statistically significant after adjustment for age, BMI, smoking status, and fat and energy intake. Such studies serve to further document the importance of considering the role of context in evaluating the role that genetic variation can play in predicting and understanding the distribution of CVD risk in the population at large. To explore the possibility that the gender-alcohol-smoking-genotype–dependent relationships between LDL-C or HDL-C and BMI are influenced by additional measures of lifestyle, we repeated the regression analyses associated with Figures 1 to 3⇑⇑ and included physical activity (2 levels), family income and education level (3 categories each), and hypertension (yes or no, based on elevated diastolic blood pressure at interview or antihypertensive therapy) as concomitants in model A. The observed influence of an alcohol by BMI interaction effect on LDL-C and HDL-C variation in the ε4/3 subgroups of women remained statistically significant after adjustment (not shown). This result argues that, in our study, the influence of alcohol was not a consequence of factors that are correlated with these important predictors of human health.
Gender-specific effects of genetic variation on measures of lipid metabolism have been repeatedly demonstrated in studies of the APOE gene.28–31⇓⇓⇓ The ability of environmental indices to predict interindividual variation in lipid and other biological traits also differs between women and men.17,32,33⇓⇓ Favorable changes in lipid and lipoprotein levels have been associated with moderate levels of alcohol intake in women.34,35⇓ The recent increase in smoking prevalence in women further emphasizes the gender-specific roles that components of lifestyle36 may take in determining risk of CVD. Our finding that the APOE genotype–specific influence of alcohol and smoking on the relationship between LDL-C and HDL-C and BMI was observed only in women serves to further reinforce the relevance of considering women and men separately when interpreting the contribution of genetic, environmental, and lifestyle variations to predicting and understanding CVD. Gender differences in the distribution of CVD risk factors, and the thoroughly documented differences in the life history of CVD between men and women,37 document the need for separate consideration of the genders in CVD studies. Knowledge of how biological differences between genders influence health can benefit both women and men. The artificial reality created by pooling genders and the reporting of gender-adjusted statistical relationships between variation in measures of health and genetic variation has become a deterrent to progress in all aspects of medical research.38
In summary, we have shown that the influence of alcohol and cigarette smoking on variation in measures of CVD risk are not independent and the effects of non-additivity of the influences of alcohol, smoking, and BMI on the prediction of levels of both the “bad” LDL-C and the “ good” HDL-C are dependent on context defined by gender and APOE genotype. Our statistical findings are consistent with the biological reality that interactions among genetic and environmental agents play an important role in the determination of interindividual variation in the onset, progression, and severity of a common disease. Accepting this reality is especially relevant for those researchers seeking to establish the functional effects of gene variations. Indeed, the influence of variation in a gene or interindividual variation in any measure of health cannot be realistically evaluated unless the concomitant factors that define the context in which the variations occur are considered in the design, analysis, and interpretation of the study.31,39–41⇓⇓⇓ Such knowledge offers the possibility of targeting subgroups of individuals who will benefit most from therapeutic programs while minimizing the potential for unexpected negative effects. Finally, this study validates the reality that neither genes nor environments, but their interactions,41–43⇓⇓ are the primary causes of variation in measures of health that have a complex multifactorial etiology.
This study was supported by the Fonds de la Recherche en Santé du Québec (FRSQ) within the joint FRSQ-Santé Québec program. The apolipoprotein data were acquired under a grant from Health and Welfare Canada (#6606 to 4543-H) to P.W. Connelly and colleagues (Jean Davignon, Bruce Reeder, Robert Hegele, Richard Lessard, Adele Csima and Suzanne Cacan).
Received January 23, 2002; revision accepted February 22, 2002.
- ↵Castelli WP. Lipids, risk factors and ischaemic heart disease. Atherosclerosis. 1996; 124: S1–S9.
- ↵Davignon J, Cohn JS. Triglycerides: a risk factor for coronary heart disease. Atherosclerosis. 1996; 124: S57–S64.
- ↵De Oliveira e Silva ER, Foster D, Harper MM, Seidman CE, Smith JD, Breslow JL, Brinton EA. Alcohol consumption raises HDL cholesterol levels by increasing the transport rate of apolipoproteins A-I and A-II. Circulation. 2000; 102: 2347–2352.
- ↵Phillips NR, Havel RJ, Kane JP. Levels and interrelationships of serum and lipoprotein cholesterol and triglycerides: association with adiposity and the consumption of ethanol, tobacco, and beverages containing caffeine. Arteriosclerosis. 1981; 1: 13–24.
- ↵Whitehead TP, Robinson D, Allaway SL. The effects of cigarette smoking and alcohol consumption on blood lipids: a dose-related study on men. Ann Clin Biochem. 1996; 33(Pt 2): 99–106.
- ↵Davignon J. Apolipoprotein E polymorphism and atherosclerosis.In: Born GVR, Schwartz CJ, eds. New Horizons in Coronary Heart Disease. London: Current Science; 1993: 5.1–5.21.
- ↵Robertson FW, Cumming AM. Effects of apoprotein E polymorphism on serum lipoprotein concentration. Arteriosclerosis. 1985; 5: 283–292.
- ↵Boer JMA, Ehnholm C, Menzel HJ, Havekes LM, Rosseneu M, O’Reilly DS, Tiret L. Interactions between lifestyle-related factors and the ApoE polymorphism on plasma lipids and apolipoproteins: The EARS study. Arterioscler Thromb Vasc Biol. 1997; 17: 1675–1681.
- ↵Zerba KE, Ferrell RE, Sing CF. Genotype-environment interaction: apolipoprotein E (apoE) gene effects and age as an index of time and spatial context in the human. Genetics. 1996; 143: 463–478.
- ↵Connelly PW, MacLean DR, Horlick L, O’Connor B, Petrasovits A, Little JA, Canadian Heart Health Survey Research Group. Plasma lipids and lipoproteins and the prevalence of risk for coronary heart disease in Canadian adults. Can Med Assoc J. 1992; 146: 1977–1987.
- ↵Connelly PW, Poapst M, Davignon J, Lussier-Cacan S, Reeder BA, Lessard R, Hegele RA, Csima A. Reference values of plasma apolipoproteins A-I and B, and association with nonlipid risk factors in the populations of two Canadian provinces: Quebec and Saskatchewan– Canadian Heart Health Surveys Research Group. Can J Cardiol. 1999; 15: 409–418.
- ↵Friedewald WT, Levy RI, Fredrickson DS. Estimation of the concentration of low-density lipoprotein cholesterol in plasma, without use of the preparative ultracentrifuge. Clin Chem. 1972; 18: 499–502.
- ↵Hill JS, Pritchard PH. Improved phenotyping of apolipoprotein E: application to population frequency distribution. Clin Chem. 1990; 36: 1871–1874.
- ↵Snedecor GW, Cochran WG. Statistical Methods. Ames, Iowa: Iowa State University Press; 1989.
- ↵Sokal RR, Rohlf EJ. Biometry: the Principles and Practice of Statistics in Biological Research. New York: WH Freeman Co.; 1995.
- ↵Sonnenberg LM, Quatromoni PA, Gagnon DR, Cupples LA, Franz MM, Ordovas JM, Wilson PWF, Schaefer EJ, Millen BE. Diet and plasma lipids in women, 2: macronutrients and plasma triglycerides, high-density lipoprotein, and the ratio of total to high density lipoprotein cholesterol in women: The Framingham nutrition studies. J Clin Epidemiol. 1996; 49: 665–672.
- ↵Corella D, Tucker K, Lahoz C, Coltell O, Cupples LA, Wilson PW, Schaefer EJ, Ordovas JM. Alcohol drinking determines the effect of the APOE locus on LDL-cholesterol concentrations in men: the Framingham Offspring Study. Am J Clin Nutr. 2001; 73: 736–745.
- ↵Xhignesse M, Lussier-Cacan S, Sing CF, Kessling AM, Davignon J. Influences of common variants of apolipoprotein E on measures of lipid metabolism in a sample selected for health. Arterioscler Thromb. 1991; 11: 1100–1110.
- ↵Kaprio J, Ferrell RE, Kottke BA, Kamboh MI, Sing CF. Effects of polymorphisms in apolipoproteins E, A-IV, and H on quantitative traits related to risk for cardiovascular disease. Arterioscler Thromb. 1991; 11: 1330–1348.
- ↵Schaefer EJ, Lamon-Fava S, Johnson S, Ordovas JM, Schaefer MM, Castelli WP, Wilson PWF. Effects of gender and menopausal status on the association of apolipoprotein E phenotype with plasma lipoprotein levels: results from the Framingham Offspring Study. Arterioscler Thromb. 1994; 14: 1105–1113.
- ↵Frikke-Schmidt R, Nordestgaard BG, Agerholm-Larsen B, Schnohr P, Tybjærg-Hansen A. Context-dependent and invariant associations between lipids, lipoproteins, and apolipoproteins and apolipoprotein E genotype. J Lipid Res. 2000; 41: 1812–1822.
- ↵Lussier-Cacan S, Xhignesse M, Piolot A, Selhub J, Davignon J, Genest J Jr. Plasma total homocysteine in healthy subjects: sex-specific relation with biological traits. Am J Clin Nutr. 1996; 64: 587–593.
- ↵Lussier-Cacan S, Xhignesse M, Kessling AM, Davignon J, Sing CF. Sources of variation in plasma lipid and lipoprotein traits in a sample selected for health. Am J Epidemiol. 1999; 150: 1229–1237.
- ↵Clevidence BA, Reichman ME, Judd JT, Muesing RA, Schatzkin A, Schaefer EJ, Li Z, Jenner J, Brown CC, Sunkin M, Campbell WS, Taylor PR. Effects of alcohol consumption on lipoproteins of premenopausal women: a controlled diet study. Arterioscler Thromb. 1995; 15: 179–184.
- ↵Nanchahal K, Ashton WD, Wood DA. Alcohol consumption, metabolic cardiovascular risk factors and hypertension in women. Int J Epidemiol. 2000; 29: 57–64.
- ↵Canadian Tobacco Use Monitoring Survey (CTUMS), 5: Youth and Young Adults. Wave 1. 1999. Health Canada, Gov. of Canada, Ottawa. www.hc-sc-gc.ca/hecs-sesc/tobacco/research/ctums/trends/
- ↵Barrett-Connor E. Sex differences in coronary heart disease: why are women so superior? The 1995 Ancel Keys Lecture. Circulation. 1997; 95: 52–264.
- ↵Pardue ML. Studying differences between the sexes may spur improvements in medicine. The Scientist. 2001; 15: 6–7.
- ↵Turner ST, Boerwinkle E, Sing CF. Context-dependent associations of the ACE I/D polymorphism with blood pressure. Hypertension. 1999; 34: 773–778.
- ↵Holdrege C. In Genetics and the Manipulation of Life: the Forgotten Factor of Context. Renewal in Science. Hudson, NY: Lindisfarne Press; 1996.
- ↵Lewontin RC. The Triple Helix: Gene, Organism, and Environment. Cambridge, Mass: Harvard University Press; 2000.