The costs of publication of this article were defrayed, in part, by the payment of page charges. This article must, therefore, be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Department of Public Health and Primary Care, University of Cambridge, Strangeways Site, Worts’ Causeway, Cambridge CB1 8RN, UK. E-mail: firstname.lastname@example.org
Objective: To examine the relationship between cigarette smoking habits and fat distribution in a population-based cohort of men and women.
Research Methods and Procedures: We analyzed cross-sectional data from 21, 828 men and women who were 45 to 79 years of age, residents in Norfolk, United Kingdom, and were recruited between 1993 and 1997. Cigarette smoking habits and other lifestyle factors were assessed using self-reported questionnaires. Anthropometric measures were obtained during a health examination.
Results: Waist-hip ratio was highest among current smokers and least among never smokers after adjusting for age, BMI, alcohol intake, total energy intake, physical activity, and education. Higher waist-hip ratio was directly associated with higher smoking pack-years in current and former smokers and inversely with duration since quitting smoking in former smokers. Adjusting for age, BMI, and other covariates, current smokers had higher waist circumference but lower hip circumference compared with former or never smokers.
Discussion: Cigarette smoking habits seem to influence fat distribution patterns. Although smokers have lower mean BMI compared with nonsmokers, they have a more metabolically adverse fat distribution profile, with higher central adiposity. The explanation for this association may help elucidate the mechanisms underlying the adverse health consequences of cigarette smoking and abdominal obesity.
Obesity is well recognized to be associated with various adverse health outcomes including increased mortality from all causes and from cardiovascular disease (1). Obesity, or excess fat, is generally defined using a weight-related measure such as BMI. Cigarette smokers tend to have lower BMI (2, 3, 4), although they seem to have increased abdominal or central adiposity (2, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16). Central adiposity, which reflects visceral fat deposition, seems to be a better indicator of the adverse metabolic consequences of obesity than overall adiposity (17). Additionally, weight loss associated with smoking may be caused by a reduction in lean mass rather than fat mass, which BMI may not fully differentiate (18).
Several studies have suggested an association between smoking and abdominal patterns of obesity (2, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16). However, increased abdominal fat in smokers may simply reflect underlying differences in the lifestyle habits between smokers and nonsmokers, such as alcohol and caloric intake, physical activity, or educational level. The importance of duration of exposure to or cessation from smoking could also be important in showing the extent of the health effects of cigarette smoking (19), yet there is limited information on the effect of this duration on body fat distribution. We examined the relationship between cigarette smoking habits and fat distributions in a large population-based cohort of British men and women.
Research Methods and Procedures
The European Prospective Investigation into Cancer and Nutrition (EPIC)1 is a multicenter population-based study of diet and cancer in Europe. The cohort study in Norfolk, United Kingdom, expanded its aims to include determinants of other chronic diseases. The EPIC-Norfolk cohort recruitment started in 1993 and ended in 1997. The study was approved by the Norfolk Health District Ethics Committee. Details of procedures and participant recruitment have been described previously (20). Briefly, we recruited participants 45 to 79 years of age using age-sex registers of collaborating general practices in Norfolk. Participants completed the Health and Lifestyle and Food Frequency Questionnaires. They also attended a health check at a clinic where trained nurses obtained anthropometric measures using a standard protocol (21) on individuals standing up and in light clothing without shoes. Height was measured to the nearest 0.1 cm using a free-standing stadiometer. Weight was measured to the nearest 100 grams using digital scales (Salter, Tonbridge, United Kingdom). Waist circumference was measured at the smallest circumference between the ribs and iliac crest to the nearest 0.1 cm, with the abdomen relaxed and at the end of a normal expiration. Where there was no natural waistline, the measurement was taken at the level of the umbilicus. Hip circumference was taken at the maximum circumference between the iliac crest and the crotch and was recorded to the nearest 0.1 cm. A D-loop nonstretch fiberglass tape was used. BMI was calculated as weight divided by height squared and waist-hip ratio (WHR) as the ratio of waist and hip circumferences. We used these circumference measures because these indices have been shown to relate more closely to abdominal and peripheral fat patterns than skinfold measures (22) and have been shown to predict various health outcomes (23).
We obtained information on cigarette smoking habits and other information on health and lifestyle from self-completed questionnaires. Participants were classified as current smokers if they reported to be smoking at present, former smokers if they were not smoking at present but had previously smoked as much as one cigarette a day for as long as a year, and never smokers if they smoked neither at present nor in the past. We also obtained information on the number of cigarettes they smoked during specific periods in the past to the present. One smoking pack-year was equivalent to one cigarette pack (or 20 cigarette sticks) per day for 1 year.
We assessed history of a physician-diagnosed illness, including cancer, heart disease, stroke, and bronchitis/emphysema, from the questionnaires. We categorized participants from sedentary (I) to most active (IV) using a physical activity index that was derived from information on the number of hours spent on the level of activity at work and during recreational and leisure time (24). Educational attainment was defined as follows: I, no qualification; II, O level or equivalent; III, A level or equivalent; IV, degree or equivalent (25). (O level refers to an ordinary level examination taken at the end of the compulsory secondary education, and A level refers to an advanced level examination taken as a pre-degree qualification.) Weekly alcohol intake (units per week) was also assessed. In the Food Frequency Questionnaire (26, 27), participants recorded their average diet over the past year using a list of 160 food items and frequency categories. Food energy intake was calculated by multiplying the frequency of food consumption by standard portion weights to obtain weight of food consumed per day. These data were converted to weekly total energy intake (kilojoules per week) using food tables (28).
There were 25, 623 who attended the baseline health check, of whom 24, 600 were 45 to 79 years of age during clinical assessment. We excluded individuals who reported a history of heart attack or stroke (n = 1099), cancer (except non-melanoma skin cancer; n = 1373), or bronchitis/emphysema (n = 2269) and those who had missing anthropometric measures (n = 86), died within the first year after the baseline visit (n = 129), or had missing data on smoking habits (n = 215). This analysis is based on data from the remaining 9819 men and 12, 009 women.
We estimated sex-specific crude age- and BMI-adjusted and covariate-adjusted mean WHR, waist circumference, and hip circumference by cigarette smoking habits. The covariates we used in the regression model included age (continuous), BMI (continuous), weekly alcohol intake (continuous), physical activity index (I, II III, and IV), education (I, II III, and IV), and weekly total energy intake (continuous). We further determined the covariate-adjusted mean WHR, waist circumference, and hip circumference stratified by cigarette smoking habit as well as by BMI categories based on its tertile distribution (<25, 25 to <27.5, and ≥27.5 kg/m2). We computed mean WHR by cigarette smoking habit and by median pack-years of smoking categories (<10 and ≥10). Former smokers were further stratified according to the median number of years since they quit smoking (<20 and ≥20 years). We also examined the relationship between WHR and cigarette smoking habit in a subgroup after excluding those who had high alcohol intake (weekly intake of alcohol of ≥21 units for men or ≥14 units in women), sedentary lifestyle (physical activity index I), or high or low total energy intake (top or bottom decile on the sex-specific distribution of weekly total energy intake). We used p < 0.05 (two-sided) to determine statistical significance of regression coefficients or mean values compared with never smokers (reference group). We performed analyses using STATA 8 statistical software (College Station, TX).
Characteristics of men and women by cigarette smoking habit are shown in Table 1. Men who were never smokers had the lowest mean WHR, and former smokers had the highest mean WHR, waist circumference, and hip circumference. Male current smokers had lower mean hip circumference than never or former smokers. Waist circumference was almost similar among never and current male smokers. In women, never smokers had the lowest mean WHR, but current smokers had lower mean BMI and waist and hip circumference.
Table 1. . Characteristics of 9819 men and 12, 009 women 45 to 79 years of age without prevalent heart disease, stroke, or cancer in the EPIC-Norfolk cohort, 1993 to 1997, by cigarette smoking habit
Cigarette smoking habit
Values are mean (SD).
All p for differences in means and proportions relative to never smokers were <0.01.
Alcohol intake of ≥21 units in men and ≥14 units in women per week.
Never smokers had the lowest mean WHR and waist circumference compared with former smokers and current smokers in both men and women, even after adjustments for age, BMI, and other covariates (Table 2). Mean hip circumference was lowest among current smokers compared with never and former smokers in both men and women. After excluding individuals who had high weekly intake of alcohol, sedentary lifestyle, or high or low weekly total energy intake, current smokers still had higher mean WHR and waist circumference and lower mean hip circumference compared with never and former smokers.
Table 2. . WHR and waist and hip circumferences of 9819 men and 12, 009 women 45 to 79 years of age without prevalent heart disease, stroke, or cancer in the EPIC-Norfolk cohort, 1993 to 1997, by cigarette smoking habit
Cigarette smoking habit
Data presented as mean (SE). All p < 0.01 except
p < 0.05 and
p > 0.05.
Excluding those who had weekly alcohol intake of ≥21 units for men or ≥14 units in women, physical activity index I, or belonged to the 1st or 10th decile of total energy intake determined separately for each sex (remaining men = 4575 and women = 5667).
Adjusted for age, BMI, alcohol intake, physical activity (I, II, III, and IV), total energy intake, and education (I, II, III, and IV).
Adjusted for age, BMI, alcohol intake, physical activity (II, III, and IV), total energy intake, and education (I, II, III, and IV).
Among current and former smokers, those who had a greater cumulative exposure, as assessed by having a higher number of pack-years smoked, also had higher WHR after adjusting for covariates (Table 3, A and B). The magnitude of effect of current smoking on WHR was similar for men and women in both pack-years of smoking categories. In the population of former smokers, time since quitting smoking was inversely related to WHR. After adjustment for covariates, the mean WHR in the group of ex-smokers who had stopped smoking for >20 years was similar to that in never smokers, regardless of the number of smoking pack-years. Among those with missing data on duration since quitting smoking, covariate-adjusted β coefficients (SE) for WHR in men (n = 1045) among former compared with never smokers were 0.005 (0.002) and 0.010 (0.003) for <10 and ≥10 smoking pack-years, respectively; in women (n = 646), these coefficients were 0.002 (0.003) and 0.011 (0.004) for <10 and ≥10 smoking pack-years, respectively.
Table 3A. . Difference in waist-hip ratio between former or current smokers relative to never smokers in 8774 men 45 to 79 years of age without prevalent heart disease, stroke, or cancer in the EPIC-Norfolk cohort, 1993 to 1997
<20 years since quitting smoking
≥20 years since quitting smoking
Data presented as β coefficients (SE) using never smokers as the reference group in a regression model.
Data on duration of smoking cessation among former smokers were missing for 1045 men and 646 women.
All p < 0.001 except
p < 0.05 and
p > 0.05.
Covariates were age, BMI, alcohol intake, physical activity level (I, II, III, and IV), total energy intake, and education (I, II, III, and IV).
Table 4 shows that current smokers had relatively higher WHR compared with never smokers in both men and women across all BMI categories. In women, relative increase in waist circumference was highest among current smokers compared with never smokers across all BMI categories; in men, similar results were noted except in the lowest BMI categories, in which current smokers had lower waist circumference values than never smokers. Hip circumference coefficients were also generally lower for current smokers across all BMI categories except in the highest BMI category in women, in which similar values were noted between current and never smokers.
Table 4. . Change in WHR, waist circumference, and hip circumference among former and current smokers for each BMI tertile relative to lean (BMI < 25 kg/m2) never smokers in 9819 men and 12, 009 women 45 to 79 years of age without prevalent heart disease, stroke, or cancer in the EPIC-Norfolk cohort, 1993 to 1997
Cigarette smoking habit
Data presented as β-coefficients (SE) adjusted for age, BMI, alcohol intake, physical activity level (I, II, III, and IV), total energy intake, and education (I, II, III, and IV) and using never smokers with BMI < 25 kg/m2 as the reference group in the regression model.
All p < 0.001 except:
p < 0.05 and
p > 0.05.
Number of never, former, and current smokers were BMI < 25 kg/m2 = 1222, 1564, and 500, BMI of 25–27.5 kg/m2 = 1137, 1711, and 367, BMI > 27.5 kg/m2 = 947, 2016, and 355 in men, and BMI < 25 kg/m2 = 3026, 1541, and 707, BMI of 25–27.5 kg/m2 = 1687, 912, and 310, BMI > 27.5 kg/m2 = 2095, 1387, 344 in women, respectively.
In Figure 1, the age-adjusted prevalence of current smoking was higher in the top compared with the bottom quintile of WHR. The prevalence of current smoking linearly increased across WHR quintiles, particularly in women. However, the proportion of current smokers decreased with increasing quintiles of BMI, weight, and waist or hip circumference.
Compared with never smokers, WHR was higher among current smokers and among those with more smoking pack-years in both men and women. Adjusting for age and BMI and other possible confounding factors such as alcohol intake, total energy intake, physical activity, and educational level only strengthened the difference in the mean WHR between never and current smokers. Waist and hip circumferences varied with cigarette smoking status in both men and women. After adjusting for age, BMI, and other covariates, mean waist circumference was higher and hip circumference was lower among current smokers compared with never smokers.
It is possible that those in the higher WHR quintiles underreport their current smoking habits. However, this underreporting should only bring the association toward the null; therefore, our results could only underestimate the true association. Alternatively, the relationship we found could be explained by underlying differences in the background characteristics of smokers from nonsmokers. There may be differences in caloric intake (4), alcohol intake (3, 15, 29, 30, 31, 32, 33), level of physical activity (4, 11, 14, 15, 30, 33), and level of education (11, 14, 33) between smokers and nonsmokers. Adjusting for these variables only slightly attenuated our findings. Nevertheless, we cannot totally rule out the effect of confounding caused by factors that we may not have measured.
Several studies have examined WHR in relation to smoking and have consistently shown similar results in men and women (5, 6, 7, 8, 9, 10, 11, 12, 13, 14). Some studies that have considered the effects of important but difficult to measure confounding factors such as alcohol and food intake, physical activity, and education still showed similar findings (8, 11, 12, 13, 14, 15, 16). However, in other studies, the association between smoking and abdominal adiposity was less conclusive in men (15) and older women (12). Heterogeneity in the results could be caused by differences in sample sizes, because smaller studies are less likely to detect modest effects, variation in reporting smoking variables and other important confounders, and age structure of the cohort population. In the Scottish Health Survey 1998, smoking-associated increases in WHR were not apparent among men 16 to 74 years of age. Nevertheless, higher WHR was noted among the subgroup of men ≥45 years of age (15).
Higher WHR among current smokers could be explained by either the waist or hip circumference or a combination of the two, but data on the separate associations of central and peripheral adiposity in relation to smoking have been shown in fewer studies (9, 11, 12, 14, 15, 16). However, some of these studies have not shown statistically significant differences in circumference measures between smokers and nonsmokers for waist girth in men (14, 15) or women (14) or for hip girth in men (16). In the EPIC-Norfolk cohort, both waist and hip girths were associated with smoking, and these associations varied with BMI level, suggesting that smoking may affect differentially the regional fat mass and/or muscle mass.
Measures of exposure to cigarettes used in earlier studies were generally limited to the daily amount of cigarettes smoked. In our study, we were able to examine the influence of duration of smoking and duration of smoking cessation. Smoking cessation leads to weight gain, and smoking resumption leads to weight loss (34). When moderate smokers have stopped for prolonged periods of time, their weight tends to approximate that of never smokers, whereas those who have stopped for shorter periods of time are more comparable to heavy smokers (35). Our findings show a similar pattern of association for WHR, in that the amount smoked and the duration of smoking were both associated with WHR. Our data suggest that a longer period of smoking cessation may be required to reduce WHR level to that of never smokers for those who have smoked for longer or smoked more heavily. However, in an older Dutch population, higher WHR among recent quitters compared with long-term quitters was observed among men but not among women (12).
The biological mechanism underlying the relationship between smoking and the pattern of regional fat distribution is unclear. Smoking could have an antiestrogenic effect by increasing the 2-hydroxylation of estradiol or inducing an imbalance in androgenic to estrogenic activity in men and women smokers (36, 37, 38). Cigarette smoking could also induce a heightened activity of gluteal adipose tissue lipoprotein lipase (39), upregulating the uptake and storage of triglyceride fatty acids by the adipocytes and consequently increasing fat mass. However, it is not clear whether changes in fat mass vary with its anatomical location. For example, reduced hip circumference may simply reflect a reduction in the lean mass of the gluteal region. As both increased abdominal and decreased peripheral adiposity are separately associated with adverse health effects (40, 41, 42, 43, 44), elucidating the extent of the differential impact of smoking on fat and lean mass in different parts of the body may further reveal additional adverse health effects of smoking.
We recognize the inherent limitations of observational studies. Moreover, the cross-sectional nature of our study does not allow us to establish any definitive temporal association between smoking and adiposity. Moreover, we relied on self-reported measures of smoking habit and used surrogate markers for fat distribution, suggesting the need for further studies using more precise techniques in quantifying both cigarette smoking exposure and fat distribution. Nevertheless, our findings were based on a large population of both men and women who had a relatively detailed smoking history and individual data on a large number of potential confounders.
Although smoking is associated with lower BMI, its relationship with increased abdominal obesity may reflect the metabolic consequences of smoking. The demonstration of these negative effects on body composition may go some way to reversing the belief that smoking is an effective strategy for weight control. Given the association with abdominal obesity, which is closely related to adverse metabolic outcomes, it is clear that smoking cessation and avoidance of smoking commencement remain key public health messages.
The EPIC-Norfolk is supported by research program funding from the Cancer Research Campaign and Medical Research Council, with additional grants from the Stroke Association, British Heart Foundation, Department of Health, Europe Against Cancer Programme Commission of the European Union, Food Standards Agency, and Wellcome Trust.
Nonstandard abbreviations: EPIC, European Prospective Investigation into Cancer and Nutrition; WHR, waist-hip ratio.