- Email: [email protected]
Cigarette Smoking, Environmental Tobacco Smoke Exposure and Insulin Sensitivity
Cigarette Smoking, Environmental Tobacco Smoke Exposure and Insulin Sensitivity The Insulin Resistance Atherosclerosis Study LEORA HENKIN, MPH, MEd, DANIEL ZACCARO, MS, STEVEN HAFFNER, MD, ANDREW KARTER, PhD, MARIAN REWERS, MD, PhD, PHYLISS SHOLINSKY, MSPH, AND LYNNE WAGENKNECHT, DrPH
PURPOSE: To investigate whether active smoking and/or exposure to environmental tobacco smoke (ETS) is associated with insulin sensitivity. METHODS: Insulin sensitivity and tobacco use history were measured in 1481 participants in the Insulin Resistance Atherosclerosis Study (IRAS). IRAS is a large mulitcenter epidemiologic study designed to explore the cross-sectional relationships among insulin resistance, cardiovascular disease risk factors and behaviors, and disease in African-American, Hispanic, and non-Hispanic white men and women, aged 40–69 years, selected to represent a broad range of glucose tolerance. Multiple linear regression models and linear contrasts were employed to describe the association between smoking history, as assessed via structured interview, and insulin sensitivity, as assessed by an insulin modified frequently sampled intravenous glucose tolerance test (FSIGT) with minimal model analysis. RESULTS: Active smoking was not associated with insulin sensitivity. Exposure to ETS was associated with lower insulin sensitivity. Specifically, for all participants combined, levels of SI were lower, indicating reduced insulin sensitivity, for those exposed to ETS when compared to those who were not exposed (p 5 0.019). This association persisted for female participants (p 5 0.013) and exhibited the same trend for males but failed to achieve statistical significance (p 5 0.264). CONCLUSIONS: Our study did not reveal an association between active smoking and insulin sensitivity, as has been shown previously. The association between ETS exposure and insulin sensitivity is a puzzling finding which deserves further investigation in the longitudinal data from IRAS as well as in other populations. Ann Epidemiol 1999;9:290–296. Published by Elsevier Science Inc. Smoking, Tobacco, Smoke Pollution, Insulin, Insulin Resistance, Atherosclerosis, Epidemiologic Study.
KEY WORDS:
INTRODUCTION Cigarette smoking, both active and passive, is a major risk factor for cardiovascular disease (CVD) and increased carotid wall thickness (1–3). The mechanism(s) for the effects of smoking on CVD are not clearly understood; however,
From the Department of Public Health Sciences, Wake Forest University School of Medicine, Winston-Salem, NC (L.H., D.Z., L.W.); Department of Medicine, University of Texas Health Science Center at San Antonio, San Antonio, TX (S.H.); Kaiser Permanente Division of Research, Oakland , CA (A.K.); Department of Preventive Medicine/Biometrics, University of Colorado Health Science Center, Denver, CO (M.R.); and National Heart, Lung and Blood Institute, Division of Epidemiology and Clinical Applications, National Institutes of Health, Bethesda, MD (P.S.) Reprint requests to Leora Henkin, Department of Public Health Sciences, Wake Forest University School of Medicine, Medical Center Blvd., Winston-Salem, NC 27157–1063. Received April 20, 1998; revised December 30, 1998; accepted January 7, 1999. Published by Elsevier Science Inc. 655 Avenue of the Americas, New York, NY 10010
a number of atherogenic characteristics have been reported among chronic smokers. These include impaired fibrinolysis and elevated Plasminogen Activator Inhibitor-1 (4), low concentrations of high-density lipoprotein and high concentrations of very-low-density lipoprotein-triglycerides (5), impaired glucose tolerance and diabetes (6). Many of these atherogenic characteristics are also consistent with decreased insulin sensitivity (7–9). Recent clinical studies using the insulin suppression test (9) and the euglycemic clamp technique (5) in small numbers of subjects have suggested that smokers exhibit reduced insulin sensitivity compared to nonsmokers. The former study demonstrated an acute effect of smoking on insulin sensitivity, although the sample size was small (n 5 70). In the latter study, the plasma insulin response of smokers was significantly higher than that of nonsmokers (n 5 20 matched pairs). Ro¨nnemaa et al. (10) have recently reported that smoking is associated with higher fasting plasma insulin 1047-2797/99/$–see front matter PII S1047-2797(99)00003-4
AEP Vol. 9, No. 5 July 1999: 290–296
levels in a large population-based cohort of non-diabetic men (n 5 616). Smokers, in that study, had higher BMIadjusted fasting plasma insulin levels, independent of other factors that affect insulin sensitivity. The higher insulin levels and higher insulin responses found in these studies are indicative of reduced insulin sensitivity. It has been speculated (5) that impaired insulin sensitivity among smokers may be due to direct effects of nicotine, carbon monoxide, or other agents in tobacco smoke. Alternatively, prolonged smoking may lead to vascular changes that reduce blood flow to skeletal muscles, and, therefore, cause a decrease in insulin-mediated glucose uptake. Using data from the Insulin Resistance Atherosclerosis Study (IRAS), the present study will investigate the cross-sectional association between cigarette smoking and insulin sensitivity as measured by the frequently sampled intravenous glucose tolerance test (FSIGT) across a broad range of glucose tolerance. This study goes beyond existing literature by directly assessing insulin sensitivity in a large multiethnic cohort, by including individuals across a broad range of glucose tolerance (diabetics and non-diabetics), and by including persons exposed and unexposed to active smoking as well as to environmental tobacco smoke (ETS).
METHODS IRAS is a large multicenter epidemiologic study designed to explore the cross-sectional relationships among insulin resistance, cardiovascular disease risk factors and behaviors, and disease in African-American, Hispanic, and non-Hispanic white men and women, aged 40–69 years, selected to represent a broad range of glucose tolerance. A full description of the IRAS study is provided elsewhere (11). Briefly, the study recruited 1625 participants and was conducted at four clinical centers. The San Antonio, Texas, and San Luis Valley, Colorado, centers recruited Hispanic and nonHispanic white participants from two ongoing populationbased studies [San Antonio Heart Study (12) and the San Luis Valley Diabetes study (13), respectively]. Clinical centers in Los Angeles and Oakland, California recruited nonHispanic white and African-American participants from Kaiser Permanente, a non-profit HMO. The final study sample included 719 individuals with normal glucose tolerance, 369 with impaired glucose tolerance and 537 with diabetes. Individuals who had taken insulin within the last five years were excluded. The IRAS examination required two visits conducted approximately one week apart. The oral glucose tolerance test and the frequently sampled intravenous glucose tolerance test (FSIGT) were performed during the first and second visits, respectively. Other physical measurements included anthropometric measures, B-mode ultrasound of the carotid arteries, blood pressure, resting electrocardiogram,
Henkin et al. SMOKING, ETS AND INSULIN SENSITIVITY
291
and a variety of blood and urine measures. Data collected via participant interview during one of the two visits included socio-demographic information, medical history and medication use, and behavioral characteristics such as diet, alcohol consumption, physical activity, perceived stress, and tobacco use. Participants were asked to refrain from tobacco use on the day of each of the two IRAS visits. Insulin sensitivity was assessed by an insulin-modified FSIGT with minimal model analysis (14, 15). From the minimal model, insulin sensitivity is expressed as the parameter SI. In IRAS, glucose (0.3 g/kg) and insulin (0.03 U/kg) were infused intravenously at 0 and 20 minutes, respectively. Blood samples were collected at 25, 2, 4, 8, 19, 22, 30, 40, 50, 70, 100, and 180 minutes for determination of glucose and insulin levels. Insulin and glucose levels were assayed by the IRAS central laboratory. These insulin and glucose levels were then used to estimate the parameters of the minimal model. It has been shown previously that SI correlates with measures obtained with the more difficult glucose clamp procedure (16). The modified protocol used in IRAS has also been shown to be a valid and reliable index of insulin sensitivity compared with the gold standard euglycemic clamp method (17), both overall (r 5 0.55) and in subsets of persons with normal glucose tolerance (r 5 0.53), impaired glucose tolerance (r 5 0.58) and diabetes (r 5 0.30). Cigarette smoking history and habits were assessed by a structured interview conducted at one of the two IRAS visits. Based on this interview, participants were divided into three categories. Individuals who had smoked less than 100 cigarettes in their lifetime were classified as nonsmokers (n 5 707). Individuals who had smoked more than 100 cigarettes in their lifetime, but who do not smoke currently were classified as past smokers (n 5 653). Individuals who currently smoke cigarettes were classified as current smokers (n 5 270). For current and past smokers, pack-years were defined as the average number of packs smoked per day times the number of years as a smoker. Participants were classified as exposed to ETS if they live with a spouse or significant other who currently smokes cigarettes. This report includes the 1481 IRAS participants who have complete tobacco use and SI data. For this analysis, race and ethnicity were assessed by selfreport. Glucose tolerance status (diabetic status) was based on the World Health Organization Criteria (18). Height, weight, and girths (minimum waist, waist at the umbilicus and hip) were measured following a standardized protocol (19). Body mass index (kg/m2) was used as a measure of overall adiposity. Total energy expenditure (Kcal/kg/yr) was estimated from a one-year recall of physical activities. Statistical Methods Smoking status was divided into the following five groups according to the participant’s active smoking status and
292
Henkin et al. SMOKING, ETS AND INSULIN SENSITIVITY
exposure to ETS: 1) never smoked 1 not exposed to ETS (never/no ETS); 2) never smoked 1 exposed to ETS (never/ yes ETS); 3) past smoker 1 not exposed to ETS (past/no ETS); 4) past smoker 1 exposed to ETS (past/yes ETS); and 5) current (active) smoker. Linear regression was used to estimate the relationship between smoking status (independent variable) and insulin sensitivity as measured by MINMOD SI (dependent variable). For all analyses, SI was transformed by log (SI 1 1) to satisfy statistical assumptions of normality. In reporting SI means and standard errors from linear models, the values obtained were back-transformed into their original units. Statistical adjustments were made for potential confounding variables including age, gender, ethnicity, clinic, glucose tolerance status, adiposity, energy expenditure, and socioeconomic status. Models were fit in the following order: 1) adjusted for age, gender, ethnicity, and clinic (demographic model); 2) adjusted for demographics plus glucose tolerance status; 3) adjusted for demographics, glucose tolerance status, and behavioral factors including adiposity (as measured by BMI and waist-to-hip ratio), energy expenditure, alcohol consumption, income, and education. The relationship between smoking and/or exposure to ETS with SI were estimated by forming two linear contrasts of the estimated means for each of the smoking groups. The first linear contrast was designed to answer the question: Is current smoking related to insulin sensitivity? This contrast compares the mean SI of current smokers with the average of the mean SI of the never smoked 1 no ETS (never/no ETS) and past smoker 1 no ETS (past/no ETS). The second linear contrast was designed to answer the question: Is ETS alone (i.e., among nonsmokers) related to insulin sensitivity? This contrast tests whether the average of never smoked 1 ETS (never/yes ETS) and past smoker 1 ETS (past/yes ETS) was significantly different from the average of never smoked 1 no ETS (never/no ETS) and past smoker 1 no ETS (past/no ETS). No differences were found in the association of ETS with SI between never and past smokers (p . 0.10), thus validating the combination of these groups for the purpose of the linear contrasts. Since gender differences in ETS exposure were anticipated and an interaction test was significant (p , 0.10), analyses were also performed separately for males and females. The interactions of smoking status with glucose tolerance status, smoking status with body mass index, and smoking status with ethnicity were tested; none were significant (p . 0.10). In order to investigate any potential dose-response relationship for active smoking, we examined the relationship between pack-years of smoking and SI. The interaction of smoking status (current vs. past-smoker) with pack-years was also tested. Again, this finding was not significant. Finally, we tested for a relationship between smoking and fasting insulin. All analyses were performed using PROC GLM in SASt statistical software (version 6.09).
AEP Vol. 9, No. 5 July 1999: 290–296
RESULTS Characteristis of the study population by smoking status group are presented in Table 1. Current smokers comprise 16.7% of the cohort. The percentage of females is higher than males in both never smoking groups. This gender difference is most pronounced in the never smokers exposed to environmental tobacco smoke. Current smokers are most likely to be Hispanic, while nonsmokers are most likely to be non-Hispanic white. BMI is lowest among current smokers and highest among never smokers. Never smokers not exposed to ETS are most likely to have a college education. For both never smokers and past smokers, individuals exposed to ETS are less likely to have a college education than individuals not exposed to ETS. Results of modeling of the association of active smoking and of environmental tobacco smoke exposure (ETS) with SI are presented in Tables 2–5. Examination of adjusted means reveal no discernable association of active smoking with SI levels for both genders combined (Table 2) nor for either gender considered separately (Tables 3 and 4). This finding persisted when the population was stratified by glucose tolerance status. Current smokers had comparable levels of SI with nonsmokers and past smokers never exposed to ETS. Similarly, there was also no main effect of pack-years on SI (p 5 0.37 for the fully adjusted model; not shown). The association of ETS with SI was significant for all models considered for the combined sample (Table 2). Levels of SI were lower, indicating reduced insulin sensitivity, for those exposed to ETS when compared to those who were not exposed. When analyses were performed separately for males and females, this ETS association persisted for female IRAS participants, but not for males. Women with ETS exposure had lower levels of SI than those without exposure (Table 3). The association of ETS with SI in males, although in the same direction as the combined sample and the females, was not statistically significant once adjusted for glucose tolerance status (Table 4). The same trend of an ETS association persisted when analyses were stratified by glucose tolerance status, albeit with reduced statistical significance (and statistical power) due to the reduced sample sizes (Table 5). The trend persisted when analyses were further restricted only to non-Hispanic white normal OGTT subjects (p 5 0.076 for the fully adjusted model; not shown). The observed pattern of reduced insulin sensitivity among nonsmokers exposed to ETS was not noted among smokers exposed to ETS (data not shown). Finally, there was no association of smoking status or ETS with fasting insulin (not shown).
DISCUSSION The results of this cross-sectional analysis suggest a relationship between ETS exposure and SI, but no significant rela-
AEP Vol. 9, No. 5 July 1999: 290–296
293
Henkin et al. SMOKING, ETS AND INSULIN SENSITIVITY
TABLE 1. Descriptive statistics for variables included in analysis by smoking status Smoking Status Group All Subjects (n 5 1481) Age (years) Gender (% female) Ethnicity (%) NHW AA Hispanic Clinic (%) SA SLV Oakland LA Income (% , 30K) Education % , HS % HS % College Diabetic status Normal IGT Diabetic BMI, kg/m2 WHR Energy expenditure (Kcal/kg/yr) Sl (X1024min21m21ml21) (Fasting insulin) (mU/ml) (2 Hour insulin) (mU/ml)
Never/No ETS (n 5 559) 37.7%
Never/Yes ETS (n 5 85) 5.7%
Past/No ETS (n 5 520) 35.1%
Past/Yes ETS (n 5 69) 4.7%
Current (n 5 248) 16.7%
56 (8)
55 (9)
53 (9)
57 (8)
56 (8)
54 (9)
54.8
65.3
81.2
41.2
56.5
50.0
38.1 27.9 34.0
41.0 27.2 31.8
22.4 25.9 51.8
42.3 29.6 28.1
37.7 26.1 36.2
28.2 27.0 44.8
25.4 25.3 22.9 26.5
26.3 23.6 24.2 25.9
31.8 38.8 10.6 18.8
22.3 21.4 26.5 29.8
26.1 31.9 15.9 26.1
27.4 30.7 18.6 23.4
42.3
37.9
63.1
38.7
40.9
52.5
17 26 57
12 24 64
22 34 44
15 25 60
28 28 45
27 30 43
45.3 22.4 32.3 29.4 (5.9) .88 (.09)
47.9 24.2 27.9 29.6 (6.3) .86 (.08)
35.3 23.5 41.2 31.3 (6.4) .85 (.08)
45.8 20.4 33.9 29.2 (5.3) .89 (.08)
36.2 24.6 39.1 28.8 (4.2) .89 (.08)
44.4 21.8 33.9 28.6 (5.9) .90 (.09)
14.6 (2.6)
14.5 (2.5)
14.8 (2.6)
14.5 (2.6)
14.5 (2.5)
14.9 (3.1)
1.70 (1.81) 17 (13) 102 (93)
1.09 (1.57) 21 (16) 120 (119)
1.65 (1.89) 18 (16) 99 (92)
1.72 (2.02) 19 (16) 94 (87)
1.18 (1.23) 18 (11) 111 (85)
1.69 (2.00) 19 (21) 93 (89)
Values are mean (SD) or n (%); NHW, non-Hispanic white; AA, African-American; SA, San Antonio, Texas; SLV, San Luis Valley, Colorado; Oakland, Oakland, California; LA, Los Angeles, California; IGT, impaired glucose tolerance; BMI, body mass index; WHR, waist-hip ratio.
tionship between active smoking and SI. These results were consistent across glucose tolerance groups. It has been hypothesized that active smoking has a direct negative effect upon insulin action and that this link may provide an expla-
nation for the increased cardiovascular risk associated with smoking. The findings of previous studies exploring the relationship between smoking and insulin sensitivity have been mixed. Attvall et al. (9) concluded that smoking
TABLE 2. Adjusted means from linear models, SI vs. smoking status, all IRAS participants (n 5 1481)a p-values for Linear Contrasts
Smoking Group Model Age, gender, ethnicity, clinic Age, gender, ethnicity, clinic, OGTT status Age, gender, ethnicity, clinic, OGTT status, BMI, W/H ratio, energy expenditure, alcohol, Income level, education a
Never/No ETS (n 5 559)
Never/Yes ETS (n 5 85)
Past/No ETS (n 5 520)
Past/Yes ETS (n 5 69)
Current (n 5 248)
Active Smokingb
ETSc
1.30 (0.04) 1.10 (0.03)
0.83 (0.07) 0.89 (0.06)
1.31 (0.05) 1.16 (0.03)
0.96 (0.09) 0.97 (0.07)
1.25 (0.07) 1.16 (0.05)
0.594 0.613
, 0.001 0.015
1.13 (0.03)
0.97 (0.06)
1.21 (0.04)
1.01 (0.07)
1.21 (0.05)
0.564
0.019
Adjusted (back-transformed) means (SEM) are reported for each smoking status group for different models with successive levels of adjustment for potential confounding variables. b p-values for the active smoking contrast are based on current vs the average of never/no ETS, past/no ETS. c p-values for the ETS contrast are based on the average of never/yes ETS, past/yes ETS vs the average of never/no ETS, past/no ETS.
294
Henkin et al. SMOKING, ETS AND INSULIN SENSITIVITY
AEP Vol. 9, No. 5 July 1999: 290–296
TABLE 3. Adjusted means from linear models, SI vs. smoking status, females only (n 5 811)a p-values for Linear Contrasts
Smoking Group Model Age, ethnicity, clinic Age, ethnicity, clinic, OGTT status Age, gender, ethnicity, clinic, OGTT status, BMI, W/H ratio, energy expenditure, Alcohol, income level, education
Never/No ETS (n 5 559)
Never/Yes ETS (n 5 85)
Past/No ETS (n 5 520)
Past/Yes ETS (n 5 69)
Current (n 5 248)
Active Smokingb
ETSc
1.36 (0.06) 1.15 (0.04)
0.90 (0.09) 0.96 (0.07)
1.40 (0.08) 1.27 (0.06)
0.89 (0.11) 1.01 (0.10)
1.19 (0.09) 1.16 (0.07)
0.149 0.637
, 0.001 0.031
1.19 (0.04)
1.07 (0.07)
1.33 (0.06)
0.97 (0.08)
1.16 (0.06)
0.272
0.013
a
Adjusted (back-transformed) means (SEM) are reported for each smoking status group for different models with successive levels of adjustment for potential confounding variables. b p-values for the active smoking contrast are based on current vs the average of never/no ETS, past/no ETS. c p-values for the ETS contrast are based on the average of never/yes ETS, past/yes ETS vs the average of never/no ETS, past/no ETS.
acutely impairs insulin action and leads to reduced insulin sensitivity. Ro¨nnemaa et al. (10) found that chronic smokers have higher levels of fasting insulin compared with nonsmokers independent of other factors that affect insulin sensitivity. In contrast to these results, Helve et al. (20) failed to demonstrate any influence of either acute or chronic smoking on insulin-mediated glucose disposal in diabetics as determined by the insulin clamp technique. Similarly, our data do not show a significant association between active smoking and insulin sensitivity as measured by MINMOD SI. In understanding the present findings, it must be noted that the smokers in our study population may not be representative of smokers in the general population. Only 16.7% of the study population self reported as current smokers; yet, the estimated prevalence of smoking for this age group in the U.S. in 1990 was 28% (21). Alternatively, previous studies finding an association between smoking and insulin sensitivity may have been to small for adequate adjustment for differences associated with insulin sensitivity. IRAS is the first study of sufficient size to measure insulin sensitivity directly and to allow for appropriate adjustment for potential
confounders. Therefore, our results may be more reflective of the true (lack of) association between active smoking and insulin sensitivity. Ours is the first study to examine a possible association of ETS exposure and insulin sensitivity. Although findings were not entirely consistent, they suggest a moderately strong association of ETS with lower insulin sensitivity among current nonsmokers. This finding was observed in both never smokers and past smokers and persisted, particularly in the larger group of female nonsmokers, after adjustment for a variety of potentially confounding demographic and physiological variables as well as variables measuring socio-economic status. Differences in the physical and chemical properties of mainstream and sidestream smoke have been documented (22–24). Thus, the finding of an ETS association in the absence of an active smoking association could be explained by higher rates of release of one or more of a large number of toxic compounds in sidestream than in mainstream smoke (25). The apparent difference in the strength of association between ETS and SI between women and men could have either a statistical or a biological explanation. The direction
TABLE 4. Adjusted means from linear models, SI vs. smoking status, males only (n 5 670)a p-values for Linear Contrasts
Smoking Group Model Age, ethnicity, clinic Age, ethnicity, clinic, OGTT status Age, gender, ethnicity, clinic, OGTT status, BMI, W/H ratio, energy expenditure, Alcohol, income level, education a
Never/No ETS (n 5 559)
Never/Yes ETS (n 5 85)
Past/No ETS (n 5 520)
Past/Yes ETS (n 5 69)
Current (n 5 248)
Active Smokingb
ETSc
1.22 (0.07) 1.04 (0.05)
0.59 (0.11) 0.75 (0.11)
1.23 (0.05) 1.06 (0.04)
1.03 (0.14) 0.91 (0.10)
1.32 (0.10) 1.18 (0.07)
0.472 0.198
, 0.020 0.128
1.04 (0.05)
0.78 (0.10)
1.12 (0.06)
1.07 (0.12)
1.27 (0.09)
0.040
0.264
Adjusted (back-transformed) means (SEM) are reported for each smoking status group for different models with successive levels of adjustment for potential confounding variables. b p-values for the active smoking contrast are based on current vs the average of never/no ETS, past/no ETS. c p-values for the ETS contrast are based on the average of never/yes ETS, past/yes ETS vs the average of never/no ETS, past/no ETS.
AEP Vol. 9, No. 5 July 1999: 290–296
Henkin et al. SMOKING, ETS AND INSULIN SENSITIVITY
295
TABLE 5. Adjusted means from linear models, SI vs. smoking status, all IRAS participants by glucose tolerance status, fully adjusted modela p-values for Linear Contrasts Smoking Group Glucose Tolerance Status Normal (n 5 671) IGT (n 5 332) Diabetic (n 5 478)
Never/No ETS
Never/Yes ETS
Past/No ETS
Past/Yes ETS
Current
Active Smokingb
ETSc
2.12 (0.08) 1.03 (0.05) 0.43 (0.02)
1.67 (0.17) 0.98 (0.12) 0.34 (0.02)
2.17 (0.09) 1.22 (0.07) 0.50 (0.02)
1.95 (0.20) 0.83 (0.12) 0.37 (0.03)
2.36 (0.15) 1.14 (0.09) 0.36 (0.02)
0.16 0.82 0.15
0.07 0.18 0.16
a
Back-transformed (SEM) adjusted for age, gender, ethnicity, clinic, BMI, waist-hip ratio, total energy expenditure, income, education and alcohol consumption. b p-values for the active smoking contrast are based on current vs the average of never/no ETS, past/no ETS. c p-values for the ETS contrast are based on the average of never/yes ETS, past/yes ETS vs the average of never/no ETS, past/no ETS.
of association was similar in men and in women and sample size was somewhat smaller for men. However, it is also plausible that women living with men who smoke have more intense ETS exposures than men exposed to ETS at home. Alternatively, it is possible that some component of ETS has an anti-estrogenic effect (6, 26, 27). Most puzzling was the absence of any association of ETS with SI in current smokers. If ETS is causally associated with alterations in SI, it is difficult to conceive how active smoking could block this association. Without a biological rationale for variation in these relationships, the possibility that the ETS/SI association represents a chance finding appears to be increased. Certainly, the association deserves additional study in other populations. The present study has several potential limitations. As mentioned above, our sample may not be representative of the population at large. Furthermore, not only is the analysis based on cross-sectional data, but tobacco use information is derived from self-report. Therefore, underreporting is a possibility, but thought to be minimal (28). It is also possible that other relevant, but unknown confounders were not considered. Thus, residual confounding or effect modification may be present. In conclusion, despite the large sample size and direct measure of the key outcome variable, data from the initial IRAS examination reveal no significant relationship between active smoking and insulin sensitivity as has been reported previously. However, a significant relationship was found between ETS exposure and insulin sensitivity. The biological plausibility of the contradiction between the potential effects of active and passive smoking, while explored above, is difficult to interpret. Longitudinal data from the IRAS study, which are currently being collected, may be helpful in clarifying these relationships. This study was supported by National Heart, Lung and Blood Institute contracts UO1-HL47887, UO1-HL47889, UO1-HL47890, UO1-HL47902, and DK-29867. The authors would like to thank the men and women who participated in this study.
REFERENCES 1. Kannel WB, D’Agostino RB, Belanger AJ. Fibrinogen, cigarette smoking, and risk of cardiovascular disease: insights from the Framingham study. Am Heart J. 1987;113:1006–1010. 2. Kannel WB. Update on the role of cigarette smoking in coronary artery disease. Am Heart J. 1981;101:319–328. 3. Howard G, Burke GL, Szklo M, Tell GS, Eckfeldt J, Evans G, Heiss G. Active and passive smoking are associated with increased carotid wall thickness: The Atherosclerosis Risk in Communities Study. Arch Intern Med. 1994;154:1277–1282. 4. Olofson B, Dahlen G, Nilsson T. Evidence for increased levels of plasminogen activator inhibitor and tissue plasminogen activator in plasma of patients with angiographically verified coronary artery disease. Eur Heart J. 1989;10:77–82. 5. Facchini FS, Hollenbeck CB, Jeppesen J, Chen Y-DI, Reaven GM. Insulin resistance and cigarette smoking. Lancet. 1992;339:1128–1130. 6. Rimm EB, Manson JE, Stampfer MJ, Colditz GA, Willett WC, Rosner B, Hennekens CH, Speizer FE. Cigarette smoking and the risk of diabetes in women. Am J Public Health. 1993;83:211–214. 7. Haffner SM, Stern MP, Hazuda HP, Pugh JA, Patterson JK. Hyperinsulinemia in a population at high risk for non-insulin-dependent diabetes mellitus. N Engl J Med. 1986;315:220–224. 8. Howard B, et al. Relationships between insulin resistance and lipoproteins in nondiabetic African Americans, Hispanics and non-Hispanic whites: The Insulin Resistance Atherosclerosis Study. Metabolism. 1998;47:1174–1179. 9. Attvall S, Fowelin J, Lager I, von Schenck H, Smith U. Smoking induces insulin resistance: a potential link with the insulin resistance syndrome. J Int Med. 1993;233:327–332. 10. Ro¨nnemaa T, Ro¨nnemaa EM, Puukka P, Pyo¨ra¨la¨ K, Laakso M. Smoking is independently associated with high plasma insulin levels in nondiabetic men. Diab Care. 1996;19:1229–1232. 11. Wagenknecht LE, Mayer EJ, Rewers M, Haffner S, Selby J, Borok GM, et al. The Insulin Resistance Atherosclerosis Study (IRAS) Objectives, Design and Recruitment Results. Ann Epidemiol. 1995;5:464–471. 12. Stern MP, Rosenthal M, Haffner SM, Hazuda HP, Franco LJ. Sex difference in the effects of sociocultural status on diabetes and cardiovascular risk factors in Mexican Americans: the San Antonio Heart Study. Am J Epidemiol. 1984;120:834–851. 13. Hamman RF, Marshall JA, Baxter J, Kahn LB, Mayer EJ, Orleans M, et al. Methods and prevalence of non-insulin-dependent diabetes mellitus in a biethnic Colorado population. Am J Epidemiol. 1989;129:295–311. 14. Bergman RN, Finegood DT, Ader M. Assessment of insulin sensitivity in vivo. End Rev. 1985;6:45–86.
296
Henkin et al. SMOKING, ETS AND INSULIN SENSITIVITY
15. Yang Y, Youn JH, Bergman RN. Modified protocols improve insulin sensitivity estimation using the minimal model. Am J Physiol. 1987;253:E595–602. 16. Pacini G, Bergman RN. MINMOD: A computer program to calculate insulin sensitivity and pancreatic responsivity from the frequently sampled intravenous glucose tolerance test. Computer Methods Programs Biomed. 1986;23:113–122. 17. Saad MF, Anderson RL, Laws A, Watanabe RM, Kades WW, Chen YDI, et al. A comparison between the minimal model and the glucose clamp in the assessment of insulin sensitivity across the spectrum of glucose tolerance. Diabetes. 1994;43:1114–1121. 18. World Health Organization. Diabetes Mellitus: Report of a WHO Study Group. Technical Report Series No. 727. Geneva, Switzerland: World Health Organization; 1985. 19. Lohman TG, Roche AF, Mortorell R, eds. Anthropometric Standardization Reference Manual (abridged edition). Champaign, IL: Human Kinetics Publishers, Inc.; 1991.
AEP Vol. 9, No. 5 July 1999: 290–296
23.
24.
25.
26.
27.
20. Helve E, Yki-Jarvinen H, Koivisto V. Smoking and insulin sensitivity in type 1 diabetic patients. Metabolism. 1986;35:874–877. 21. Surveillance for smoking-attributable mortality and years of potential life lost, by state—United States, 1990. MMWR. 1994;43:1–8. 22. National Research Council. Environmental Tobacco Smoke: Measur-
28.
ing Exposures and Assessing Health Effects. Washington, DC: National Academy Press; 1986. U.S. Department of Health and Human Services. The Health Consequences of Involuntary Smoking. A Report of the Surgeon General. DHHS Pub. No. (PHS) 87-8398. U.S. DHHS, Public Health Service, Office of the Assistant Secretary for Health, Office of Smoking and Health, Washington, DC; 1986. Guerin MR, Jenkins RA, Tomkins BA. The Chemistry of Environmental Tobacco Smoke: Composition and Measurement. Chelsea, MI: Lewis Publishers; 1992. U.S. Environmental Protection Agency. Respiratory Health Effects of Passive Smoking; Lung Cancer and Other Disorders. EPA/600/6-90006F; Office of Health and Environmental Assessment Office of Research and Development, U.S. Environmental Protection Agency, Washington, DC; 1992. Friedman AJ, Ravnikar VA, Barbieri RL. Serum steroid hormone profiles in postmenopausal smokers and nonsmokers. Fertil Steril. 1987;47:398–401. Khaw KT, Tazuke S, Barret-Connor E. Cigarette smoking and level of adrenal androgens in post-menopausal women. N Engl J Med. 1988:1705–1709. Wagenknecht LE, Burke GL, Perkins LL, Haley NJ, Friedman GD. Misclassification of smoking status in the CARDIA study: A comparison of self-report with serum cotinine levels. Am J Public Health. 1992;82:33–36.
PURPOSE: To investigate whether active smoking and/or exposure to environmental tobacco smoke (ETS) is associated with insulin sensitivity. METHODS: Insulin sensitivity and tobacco use history were measured in 1481 participants in the Insulin Resistance Atherosclerosis Study (IRAS). IRAS is a large mulitcenter epidemiologic study designed to explore the cross-sectional relationships among insulin resistance, cardiovascular disease risk factors and behaviors, and disease in African-American, Hispanic, and non-Hispanic white men and women, aged 40–69 years, selected to represent a broad range of glucose tolerance. Multiple linear regression models and linear contrasts were employed to describe the association between smoking history, as assessed via structured interview, and insulin sensitivity, as assessed by an insulin modified frequently sampled intravenous glucose tolerance test (FSIGT) with minimal model analysis. RESULTS: Active smoking was not associated with insulin sensitivity. Exposure to ETS was associated with lower insulin sensitivity. Specifically, for all participants combined, levels of SI were lower, indicating reduced insulin sensitivity, for those exposed to ETS when compared to those who were not exposed (p 5 0.019). This association persisted for female participants (p 5 0.013) and exhibited the same trend for males but failed to achieve statistical significance (p 5 0.264). CONCLUSIONS: Our study did not reveal an association between active smoking and insulin sensitivity, as has been shown previously. The association between ETS exposure and insulin sensitivity is a puzzling finding which deserves further investigation in the longitudinal data from IRAS as well as in other populations. Ann Epidemiol 1999;9:290–296. Published by Elsevier Science Inc. Smoking, Tobacco, Smoke Pollution, Insulin, Insulin Resistance, Atherosclerosis, Epidemiologic Study.
KEY WORDS:
INTRODUCTION Cigarette smoking, both active and passive, is a major risk factor for cardiovascular disease (CVD) and increased carotid wall thickness (1–3). The mechanism(s) for the effects of smoking on CVD are not clearly understood; however,
From the Department of Public Health Sciences, Wake Forest University School of Medicine, Winston-Salem, NC (L.H., D.Z., L.W.); Department of Medicine, University of Texas Health Science Center at San Antonio, San Antonio, TX (S.H.); Kaiser Permanente Division of Research, Oakland , CA (A.K.); Department of Preventive Medicine/Biometrics, University of Colorado Health Science Center, Denver, CO (M.R.); and National Heart, Lung and Blood Institute, Division of Epidemiology and Clinical Applications, National Institutes of Health, Bethesda, MD (P.S.) Reprint requests to Leora Henkin, Department of Public Health Sciences, Wake Forest University School of Medicine, Medical Center Blvd., Winston-Salem, NC 27157–1063. Received April 20, 1998; revised December 30, 1998; accepted January 7, 1999. Published by Elsevier Science Inc. 655 Avenue of the Americas, New York, NY 10010
a number of atherogenic characteristics have been reported among chronic smokers. These include impaired fibrinolysis and elevated Plasminogen Activator Inhibitor-1 (4), low concentrations of high-density lipoprotein and high concentrations of very-low-density lipoprotein-triglycerides (5), impaired glucose tolerance and diabetes (6). Many of these atherogenic characteristics are also consistent with decreased insulin sensitivity (7–9). Recent clinical studies using the insulin suppression test (9) and the euglycemic clamp technique (5) in small numbers of subjects have suggested that smokers exhibit reduced insulin sensitivity compared to nonsmokers. The former study demonstrated an acute effect of smoking on insulin sensitivity, although the sample size was small (n 5 70). In the latter study, the plasma insulin response of smokers was significantly higher than that of nonsmokers (n 5 20 matched pairs). Ro¨nnemaa et al. (10) have recently reported that smoking is associated with higher fasting plasma insulin 1047-2797/99/$–see front matter PII S1047-2797(99)00003-4
AEP Vol. 9, No. 5 July 1999: 290–296
levels in a large population-based cohort of non-diabetic men (n 5 616). Smokers, in that study, had higher BMIadjusted fasting plasma insulin levels, independent of other factors that affect insulin sensitivity. The higher insulin levels and higher insulin responses found in these studies are indicative of reduced insulin sensitivity. It has been speculated (5) that impaired insulin sensitivity among smokers may be due to direct effects of nicotine, carbon monoxide, or other agents in tobacco smoke. Alternatively, prolonged smoking may lead to vascular changes that reduce blood flow to skeletal muscles, and, therefore, cause a decrease in insulin-mediated glucose uptake. Using data from the Insulin Resistance Atherosclerosis Study (IRAS), the present study will investigate the cross-sectional association between cigarette smoking and insulin sensitivity as measured by the frequently sampled intravenous glucose tolerance test (FSIGT) across a broad range of glucose tolerance. This study goes beyond existing literature by directly assessing insulin sensitivity in a large multiethnic cohort, by including individuals across a broad range of glucose tolerance (diabetics and non-diabetics), and by including persons exposed and unexposed to active smoking as well as to environmental tobacco smoke (ETS).
METHODS IRAS is a large multicenter epidemiologic study designed to explore the cross-sectional relationships among insulin resistance, cardiovascular disease risk factors and behaviors, and disease in African-American, Hispanic, and non-Hispanic white men and women, aged 40–69 years, selected to represent a broad range of glucose tolerance. A full description of the IRAS study is provided elsewhere (11). Briefly, the study recruited 1625 participants and was conducted at four clinical centers. The San Antonio, Texas, and San Luis Valley, Colorado, centers recruited Hispanic and nonHispanic white participants from two ongoing populationbased studies [San Antonio Heart Study (12) and the San Luis Valley Diabetes study (13), respectively]. Clinical centers in Los Angeles and Oakland, California recruited nonHispanic white and African-American participants from Kaiser Permanente, a non-profit HMO. The final study sample included 719 individuals with normal glucose tolerance, 369 with impaired glucose tolerance and 537 with diabetes. Individuals who had taken insulin within the last five years were excluded. The IRAS examination required two visits conducted approximately one week apart. The oral glucose tolerance test and the frequently sampled intravenous glucose tolerance test (FSIGT) were performed during the first and second visits, respectively. Other physical measurements included anthropometric measures, B-mode ultrasound of the carotid arteries, blood pressure, resting electrocardiogram,
Henkin et al. SMOKING, ETS AND INSULIN SENSITIVITY
291
and a variety of blood and urine measures. Data collected via participant interview during one of the two visits included socio-demographic information, medical history and medication use, and behavioral characteristics such as diet, alcohol consumption, physical activity, perceived stress, and tobacco use. Participants were asked to refrain from tobacco use on the day of each of the two IRAS visits. Insulin sensitivity was assessed by an insulin-modified FSIGT with minimal model analysis (14, 15). From the minimal model, insulin sensitivity is expressed as the parameter SI. In IRAS, glucose (0.3 g/kg) and insulin (0.03 U/kg) were infused intravenously at 0 and 20 minutes, respectively. Blood samples were collected at 25, 2, 4, 8, 19, 22, 30, 40, 50, 70, 100, and 180 minutes for determination of glucose and insulin levels. Insulin and glucose levels were assayed by the IRAS central laboratory. These insulin and glucose levels were then used to estimate the parameters of the minimal model. It has been shown previously that SI correlates with measures obtained with the more difficult glucose clamp procedure (16). The modified protocol used in IRAS has also been shown to be a valid and reliable index of insulin sensitivity compared with the gold standard euglycemic clamp method (17), both overall (r 5 0.55) and in subsets of persons with normal glucose tolerance (r 5 0.53), impaired glucose tolerance (r 5 0.58) and diabetes (r 5 0.30). Cigarette smoking history and habits were assessed by a structured interview conducted at one of the two IRAS visits. Based on this interview, participants were divided into three categories. Individuals who had smoked less than 100 cigarettes in their lifetime were classified as nonsmokers (n 5 707). Individuals who had smoked more than 100 cigarettes in their lifetime, but who do not smoke currently were classified as past smokers (n 5 653). Individuals who currently smoke cigarettes were classified as current smokers (n 5 270). For current and past smokers, pack-years were defined as the average number of packs smoked per day times the number of years as a smoker. Participants were classified as exposed to ETS if they live with a spouse or significant other who currently smokes cigarettes. This report includes the 1481 IRAS participants who have complete tobacco use and SI data. For this analysis, race and ethnicity were assessed by selfreport. Glucose tolerance status (diabetic status) was based on the World Health Organization Criteria (18). Height, weight, and girths (minimum waist, waist at the umbilicus and hip) were measured following a standardized protocol (19). Body mass index (kg/m2) was used as a measure of overall adiposity. Total energy expenditure (Kcal/kg/yr) was estimated from a one-year recall of physical activities. Statistical Methods Smoking status was divided into the following five groups according to the participant’s active smoking status and
292
Henkin et al. SMOKING, ETS AND INSULIN SENSITIVITY
exposure to ETS: 1) never smoked 1 not exposed to ETS (never/no ETS); 2) never smoked 1 exposed to ETS (never/ yes ETS); 3) past smoker 1 not exposed to ETS (past/no ETS); 4) past smoker 1 exposed to ETS (past/yes ETS); and 5) current (active) smoker. Linear regression was used to estimate the relationship between smoking status (independent variable) and insulin sensitivity as measured by MINMOD SI (dependent variable). For all analyses, SI was transformed by log (SI 1 1) to satisfy statistical assumptions of normality. In reporting SI means and standard errors from linear models, the values obtained were back-transformed into their original units. Statistical adjustments were made for potential confounding variables including age, gender, ethnicity, clinic, glucose tolerance status, adiposity, energy expenditure, and socioeconomic status. Models were fit in the following order: 1) adjusted for age, gender, ethnicity, and clinic (demographic model); 2) adjusted for demographics plus glucose tolerance status; 3) adjusted for demographics, glucose tolerance status, and behavioral factors including adiposity (as measured by BMI and waist-to-hip ratio), energy expenditure, alcohol consumption, income, and education. The relationship between smoking and/or exposure to ETS with SI were estimated by forming two linear contrasts of the estimated means for each of the smoking groups. The first linear contrast was designed to answer the question: Is current smoking related to insulin sensitivity? This contrast compares the mean SI of current smokers with the average of the mean SI of the never smoked 1 no ETS (never/no ETS) and past smoker 1 no ETS (past/no ETS). The second linear contrast was designed to answer the question: Is ETS alone (i.e., among nonsmokers) related to insulin sensitivity? This contrast tests whether the average of never smoked 1 ETS (never/yes ETS) and past smoker 1 ETS (past/yes ETS) was significantly different from the average of never smoked 1 no ETS (never/no ETS) and past smoker 1 no ETS (past/no ETS). No differences were found in the association of ETS with SI between never and past smokers (p . 0.10), thus validating the combination of these groups for the purpose of the linear contrasts. Since gender differences in ETS exposure were anticipated and an interaction test was significant (p , 0.10), analyses were also performed separately for males and females. The interactions of smoking status with glucose tolerance status, smoking status with body mass index, and smoking status with ethnicity were tested; none were significant (p . 0.10). In order to investigate any potential dose-response relationship for active smoking, we examined the relationship between pack-years of smoking and SI. The interaction of smoking status (current vs. past-smoker) with pack-years was also tested. Again, this finding was not significant. Finally, we tested for a relationship between smoking and fasting insulin. All analyses were performed using PROC GLM in SASt statistical software (version 6.09).
AEP Vol. 9, No. 5 July 1999: 290–296
RESULTS Characteristis of the study population by smoking status group are presented in Table 1. Current smokers comprise 16.7% of the cohort. The percentage of females is higher than males in both never smoking groups. This gender difference is most pronounced in the never smokers exposed to environmental tobacco smoke. Current smokers are most likely to be Hispanic, while nonsmokers are most likely to be non-Hispanic white. BMI is lowest among current smokers and highest among never smokers. Never smokers not exposed to ETS are most likely to have a college education. For both never smokers and past smokers, individuals exposed to ETS are less likely to have a college education than individuals not exposed to ETS. Results of modeling of the association of active smoking and of environmental tobacco smoke exposure (ETS) with SI are presented in Tables 2–5. Examination of adjusted means reveal no discernable association of active smoking with SI levels for both genders combined (Table 2) nor for either gender considered separately (Tables 3 and 4). This finding persisted when the population was stratified by glucose tolerance status. Current smokers had comparable levels of SI with nonsmokers and past smokers never exposed to ETS. Similarly, there was also no main effect of pack-years on SI (p 5 0.37 for the fully adjusted model; not shown). The association of ETS with SI was significant for all models considered for the combined sample (Table 2). Levels of SI were lower, indicating reduced insulin sensitivity, for those exposed to ETS when compared to those who were not exposed. When analyses were performed separately for males and females, this ETS association persisted for female IRAS participants, but not for males. Women with ETS exposure had lower levels of SI than those without exposure (Table 3). The association of ETS with SI in males, although in the same direction as the combined sample and the females, was not statistically significant once adjusted for glucose tolerance status (Table 4). The same trend of an ETS association persisted when analyses were stratified by glucose tolerance status, albeit with reduced statistical significance (and statistical power) due to the reduced sample sizes (Table 5). The trend persisted when analyses were further restricted only to non-Hispanic white normal OGTT subjects (p 5 0.076 for the fully adjusted model; not shown). The observed pattern of reduced insulin sensitivity among nonsmokers exposed to ETS was not noted among smokers exposed to ETS (data not shown). Finally, there was no association of smoking status or ETS with fasting insulin (not shown).
DISCUSSION The results of this cross-sectional analysis suggest a relationship between ETS exposure and SI, but no significant rela-
AEP Vol. 9, No. 5 July 1999: 290–296
293
Henkin et al. SMOKING, ETS AND INSULIN SENSITIVITY
TABLE 1. Descriptive statistics for variables included in analysis by smoking status Smoking Status Group All Subjects (n 5 1481) Age (years) Gender (% female) Ethnicity (%) NHW AA Hispanic Clinic (%) SA SLV Oakland LA Income (% , 30K) Education % , HS % HS % College Diabetic status Normal IGT Diabetic BMI, kg/m2 WHR Energy expenditure (Kcal/kg/yr) Sl (X1024min21m21ml21) (Fasting insulin) (mU/ml) (2 Hour insulin) (mU/ml)
Never/No ETS (n 5 559) 37.7%
Never/Yes ETS (n 5 85) 5.7%
Past/No ETS (n 5 520) 35.1%
Past/Yes ETS (n 5 69) 4.7%
Current (n 5 248) 16.7%
56 (8)
55 (9)
53 (9)
57 (8)
56 (8)
54 (9)
54.8
65.3
81.2
41.2
56.5
50.0
38.1 27.9 34.0
41.0 27.2 31.8
22.4 25.9 51.8
42.3 29.6 28.1
37.7 26.1 36.2
28.2 27.0 44.8
25.4 25.3 22.9 26.5
26.3 23.6 24.2 25.9
31.8 38.8 10.6 18.8
22.3 21.4 26.5 29.8
26.1 31.9 15.9 26.1
27.4 30.7 18.6 23.4
42.3
37.9
63.1
38.7
40.9
52.5
17 26 57
12 24 64
22 34 44
15 25 60
28 28 45
27 30 43
45.3 22.4 32.3 29.4 (5.9) .88 (.09)
47.9 24.2 27.9 29.6 (6.3) .86 (.08)
35.3 23.5 41.2 31.3 (6.4) .85 (.08)
45.8 20.4 33.9 29.2 (5.3) .89 (.08)
36.2 24.6 39.1 28.8 (4.2) .89 (.08)
44.4 21.8 33.9 28.6 (5.9) .90 (.09)
14.6 (2.6)
14.5 (2.5)
14.8 (2.6)
14.5 (2.6)
14.5 (2.5)
14.9 (3.1)
1.70 (1.81) 17 (13) 102 (93)
1.09 (1.57) 21 (16) 120 (119)
1.65 (1.89) 18 (16) 99 (92)
1.72 (2.02) 19 (16) 94 (87)
1.18 (1.23) 18 (11) 111 (85)
1.69 (2.00) 19 (21) 93 (89)
Values are mean (SD) or n (%); NHW, non-Hispanic white; AA, African-American; SA, San Antonio, Texas; SLV, San Luis Valley, Colorado; Oakland, Oakland, California; LA, Los Angeles, California; IGT, impaired glucose tolerance; BMI, body mass index; WHR, waist-hip ratio.
tionship between active smoking and SI. These results were consistent across glucose tolerance groups. It has been hypothesized that active smoking has a direct negative effect upon insulin action and that this link may provide an expla-
nation for the increased cardiovascular risk associated with smoking. The findings of previous studies exploring the relationship between smoking and insulin sensitivity have been mixed. Attvall et al. (9) concluded that smoking
TABLE 2. Adjusted means from linear models, SI vs. smoking status, all IRAS participants (n 5 1481)a p-values for Linear Contrasts
Smoking Group Model Age, gender, ethnicity, clinic Age, gender, ethnicity, clinic, OGTT status Age, gender, ethnicity, clinic, OGTT status, BMI, W/H ratio, energy expenditure, alcohol, Income level, education a
Never/No ETS (n 5 559)
Never/Yes ETS (n 5 85)
Past/No ETS (n 5 520)
Past/Yes ETS (n 5 69)
Current (n 5 248)
Active Smokingb
ETSc
1.30 (0.04) 1.10 (0.03)
0.83 (0.07) 0.89 (0.06)
1.31 (0.05) 1.16 (0.03)
0.96 (0.09) 0.97 (0.07)
1.25 (0.07) 1.16 (0.05)
0.594 0.613
, 0.001 0.015
1.13 (0.03)
0.97 (0.06)
1.21 (0.04)
1.01 (0.07)
1.21 (0.05)
0.564
0.019
Adjusted (back-transformed) means (SEM) are reported for each smoking status group for different models with successive levels of adjustment for potential confounding variables. b p-values for the active smoking contrast are based on current vs the average of never/no ETS, past/no ETS. c p-values for the ETS contrast are based on the average of never/yes ETS, past/yes ETS vs the average of never/no ETS, past/no ETS.
294
Henkin et al. SMOKING, ETS AND INSULIN SENSITIVITY
AEP Vol. 9, No. 5 July 1999: 290–296
TABLE 3. Adjusted means from linear models, SI vs. smoking status, females only (n 5 811)a p-values for Linear Contrasts
Smoking Group Model Age, ethnicity, clinic Age, ethnicity, clinic, OGTT status Age, gender, ethnicity, clinic, OGTT status, BMI, W/H ratio, energy expenditure, Alcohol, income level, education
Never/No ETS (n 5 559)
Never/Yes ETS (n 5 85)
Past/No ETS (n 5 520)
Past/Yes ETS (n 5 69)
Current (n 5 248)
Active Smokingb
ETSc
1.36 (0.06) 1.15 (0.04)
0.90 (0.09) 0.96 (0.07)
1.40 (0.08) 1.27 (0.06)
0.89 (0.11) 1.01 (0.10)
1.19 (0.09) 1.16 (0.07)
0.149 0.637
, 0.001 0.031
1.19 (0.04)
1.07 (0.07)
1.33 (0.06)
0.97 (0.08)
1.16 (0.06)
0.272
0.013
a
Adjusted (back-transformed) means (SEM) are reported for each smoking status group for different models with successive levels of adjustment for potential confounding variables. b p-values for the active smoking contrast are based on current vs the average of never/no ETS, past/no ETS. c p-values for the ETS contrast are based on the average of never/yes ETS, past/yes ETS vs the average of never/no ETS, past/no ETS.
acutely impairs insulin action and leads to reduced insulin sensitivity. Ro¨nnemaa et al. (10) found that chronic smokers have higher levels of fasting insulin compared with nonsmokers independent of other factors that affect insulin sensitivity. In contrast to these results, Helve et al. (20) failed to demonstrate any influence of either acute or chronic smoking on insulin-mediated glucose disposal in diabetics as determined by the insulin clamp technique. Similarly, our data do not show a significant association between active smoking and insulin sensitivity as measured by MINMOD SI. In understanding the present findings, it must be noted that the smokers in our study population may not be representative of smokers in the general population. Only 16.7% of the study population self reported as current smokers; yet, the estimated prevalence of smoking for this age group in the U.S. in 1990 was 28% (21). Alternatively, previous studies finding an association between smoking and insulin sensitivity may have been to small for adequate adjustment for differences associated with insulin sensitivity. IRAS is the first study of sufficient size to measure insulin sensitivity directly and to allow for appropriate adjustment for potential
confounders. Therefore, our results may be more reflective of the true (lack of) association between active smoking and insulin sensitivity. Ours is the first study to examine a possible association of ETS exposure and insulin sensitivity. Although findings were not entirely consistent, they suggest a moderately strong association of ETS with lower insulin sensitivity among current nonsmokers. This finding was observed in both never smokers and past smokers and persisted, particularly in the larger group of female nonsmokers, after adjustment for a variety of potentially confounding demographic and physiological variables as well as variables measuring socio-economic status. Differences in the physical and chemical properties of mainstream and sidestream smoke have been documented (22–24). Thus, the finding of an ETS association in the absence of an active smoking association could be explained by higher rates of release of one or more of a large number of toxic compounds in sidestream than in mainstream smoke (25). The apparent difference in the strength of association between ETS and SI between women and men could have either a statistical or a biological explanation. The direction
TABLE 4. Adjusted means from linear models, SI vs. smoking status, males only (n 5 670)a p-values for Linear Contrasts
Smoking Group Model Age, ethnicity, clinic Age, ethnicity, clinic, OGTT status Age, gender, ethnicity, clinic, OGTT status, BMI, W/H ratio, energy expenditure, Alcohol, income level, education a
Never/No ETS (n 5 559)
Never/Yes ETS (n 5 85)
Past/No ETS (n 5 520)
Past/Yes ETS (n 5 69)
Current (n 5 248)
Active Smokingb
ETSc
1.22 (0.07) 1.04 (0.05)
0.59 (0.11) 0.75 (0.11)
1.23 (0.05) 1.06 (0.04)
1.03 (0.14) 0.91 (0.10)
1.32 (0.10) 1.18 (0.07)
0.472 0.198
, 0.020 0.128
1.04 (0.05)
0.78 (0.10)
1.12 (0.06)
1.07 (0.12)
1.27 (0.09)
0.040
0.264
Adjusted (back-transformed) means (SEM) are reported for each smoking status group for different models with successive levels of adjustment for potential confounding variables. b p-values for the active smoking contrast are based on current vs the average of never/no ETS, past/no ETS. c p-values for the ETS contrast are based on the average of never/yes ETS, past/yes ETS vs the average of never/no ETS, past/no ETS.
AEP Vol. 9, No. 5 July 1999: 290–296
Henkin et al. SMOKING, ETS AND INSULIN SENSITIVITY
295
TABLE 5. Adjusted means from linear models, SI vs. smoking status, all IRAS participants by glucose tolerance status, fully adjusted modela p-values for Linear Contrasts Smoking Group Glucose Tolerance Status Normal (n 5 671) IGT (n 5 332) Diabetic (n 5 478)
Never/No ETS
Never/Yes ETS
Past/No ETS
Past/Yes ETS
Current
Active Smokingb
ETSc
2.12 (0.08) 1.03 (0.05) 0.43 (0.02)
1.67 (0.17) 0.98 (0.12) 0.34 (0.02)
2.17 (0.09) 1.22 (0.07) 0.50 (0.02)
1.95 (0.20) 0.83 (0.12) 0.37 (0.03)
2.36 (0.15) 1.14 (0.09) 0.36 (0.02)
0.16 0.82 0.15
0.07 0.18 0.16
a
Back-transformed (SEM) adjusted for age, gender, ethnicity, clinic, BMI, waist-hip ratio, total energy expenditure, income, education and alcohol consumption. b p-values for the active smoking contrast are based on current vs the average of never/no ETS, past/no ETS. c p-values for the ETS contrast are based on the average of never/yes ETS, past/yes ETS vs the average of never/no ETS, past/no ETS.
of association was similar in men and in women and sample size was somewhat smaller for men. However, it is also plausible that women living with men who smoke have more intense ETS exposures than men exposed to ETS at home. Alternatively, it is possible that some component of ETS has an anti-estrogenic effect (6, 26, 27). Most puzzling was the absence of any association of ETS with SI in current smokers. If ETS is causally associated with alterations in SI, it is difficult to conceive how active smoking could block this association. Without a biological rationale for variation in these relationships, the possibility that the ETS/SI association represents a chance finding appears to be increased. Certainly, the association deserves additional study in other populations. The present study has several potential limitations. As mentioned above, our sample may not be representative of the population at large. Furthermore, not only is the analysis based on cross-sectional data, but tobacco use information is derived from self-report. Therefore, underreporting is a possibility, but thought to be minimal (28). It is also possible that other relevant, but unknown confounders were not considered. Thus, residual confounding or effect modification may be present. In conclusion, despite the large sample size and direct measure of the key outcome variable, data from the initial IRAS examination reveal no significant relationship between active smoking and insulin sensitivity as has been reported previously. However, a significant relationship was found between ETS exposure and insulin sensitivity. The biological plausibility of the contradiction between the potential effects of active and passive smoking, while explored above, is difficult to interpret. Longitudinal data from the IRAS study, which are currently being collected, may be helpful in clarifying these relationships. This study was supported by National Heart, Lung and Blood Institute contracts UO1-HL47887, UO1-HL47889, UO1-HL47890, UO1-HL47902, and DK-29867. The authors would like to thank the men and women who participated in this study.
REFERENCES 1. Kannel WB, D’Agostino RB, Belanger AJ. Fibrinogen, cigarette smoking, and risk of cardiovascular disease: insights from the Framingham study. Am Heart J. 1987;113:1006–1010. 2. Kannel WB. Update on the role of cigarette smoking in coronary artery disease. Am Heart J. 1981;101:319–328. 3. Howard G, Burke GL, Szklo M, Tell GS, Eckfeldt J, Evans G, Heiss G. Active and passive smoking are associated with increased carotid wall thickness: The Atherosclerosis Risk in Communities Study. Arch Intern Med. 1994;154:1277–1282. 4. Olofson B, Dahlen G, Nilsson T. Evidence for increased levels of plasminogen activator inhibitor and tissue plasminogen activator in plasma of patients with angiographically verified coronary artery disease. Eur Heart J. 1989;10:77–82. 5. Facchini FS, Hollenbeck CB, Jeppesen J, Chen Y-DI, Reaven GM. Insulin resistance and cigarette smoking. Lancet. 1992;339:1128–1130. 6. Rimm EB, Manson JE, Stampfer MJ, Colditz GA, Willett WC, Rosner B, Hennekens CH, Speizer FE. Cigarette smoking and the risk of diabetes in women. Am J Public Health. 1993;83:211–214. 7. Haffner SM, Stern MP, Hazuda HP, Pugh JA, Patterson JK. Hyperinsulinemia in a population at high risk for non-insulin-dependent diabetes mellitus. N Engl J Med. 1986;315:220–224. 8. Howard B, et al. Relationships between insulin resistance and lipoproteins in nondiabetic African Americans, Hispanics and non-Hispanic whites: The Insulin Resistance Atherosclerosis Study. Metabolism. 1998;47:1174–1179. 9. Attvall S, Fowelin J, Lager I, von Schenck H, Smith U. Smoking induces insulin resistance: a potential link with the insulin resistance syndrome. J Int Med. 1993;233:327–332. 10. Ro¨nnemaa T, Ro¨nnemaa EM, Puukka P, Pyo¨ra¨la¨ K, Laakso M. Smoking is independently associated with high plasma insulin levels in nondiabetic men. Diab Care. 1996;19:1229–1232. 11. Wagenknecht LE, Mayer EJ, Rewers M, Haffner S, Selby J, Borok GM, et al. The Insulin Resistance Atherosclerosis Study (IRAS) Objectives, Design and Recruitment Results. Ann Epidemiol. 1995;5:464–471. 12. Stern MP, Rosenthal M, Haffner SM, Hazuda HP, Franco LJ. Sex difference in the effects of sociocultural status on diabetes and cardiovascular risk factors in Mexican Americans: the San Antonio Heart Study. Am J Epidemiol. 1984;120:834–851. 13. Hamman RF, Marshall JA, Baxter J, Kahn LB, Mayer EJ, Orleans M, et al. Methods and prevalence of non-insulin-dependent diabetes mellitus in a biethnic Colorado population. Am J Epidemiol. 1989;129:295–311. 14. Bergman RN, Finegood DT, Ader M. Assessment of insulin sensitivity in vivo. End Rev. 1985;6:45–86.
296
Henkin et al. SMOKING, ETS AND INSULIN SENSITIVITY
15. Yang Y, Youn JH, Bergman RN. Modified protocols improve insulin sensitivity estimation using the minimal model. Am J Physiol. 1987;253:E595–602. 16. Pacini G, Bergman RN. MINMOD: A computer program to calculate insulin sensitivity and pancreatic responsivity from the frequently sampled intravenous glucose tolerance test. Computer Methods Programs Biomed. 1986;23:113–122. 17. Saad MF, Anderson RL, Laws A, Watanabe RM, Kades WW, Chen YDI, et al. A comparison between the minimal model and the glucose clamp in the assessment of insulin sensitivity across the spectrum of glucose tolerance. Diabetes. 1994;43:1114–1121. 18. World Health Organization. Diabetes Mellitus: Report of a WHO Study Group. Technical Report Series No. 727. Geneva, Switzerland: World Health Organization; 1985. 19. Lohman TG, Roche AF, Mortorell R, eds. Anthropometric Standardization Reference Manual (abridged edition). Champaign, IL: Human Kinetics Publishers, Inc.; 1991.
AEP Vol. 9, No. 5 July 1999: 290–296
23.
24.
25.
26.
27.
20. Helve E, Yki-Jarvinen H, Koivisto V. Smoking and insulin sensitivity in type 1 diabetic patients. Metabolism. 1986;35:874–877. 21. Surveillance for smoking-attributable mortality and years of potential life lost, by state—United States, 1990. MMWR. 1994;43:1–8. 22. National Research Council. Environmental Tobacco Smoke: Measur-
28.
ing Exposures and Assessing Health Effects. Washington, DC: National Academy Press; 1986. U.S. Department of Health and Human Services. The Health Consequences of Involuntary Smoking. A Report of the Surgeon General. DHHS Pub. No. (PHS) 87-8398. U.S. DHHS, Public Health Service, Office of the Assistant Secretary for Health, Office of Smoking and Health, Washington, DC; 1986. Guerin MR, Jenkins RA, Tomkins BA. The Chemistry of Environmental Tobacco Smoke: Composition and Measurement. Chelsea, MI: Lewis Publishers; 1992. U.S. Environmental Protection Agency. Respiratory Health Effects of Passive Smoking; Lung Cancer and Other Disorders. EPA/600/6-90006F; Office of Health and Environmental Assessment Office of Research and Development, U.S. Environmental Protection Agency, Washington, DC; 1992. Friedman AJ, Ravnikar VA, Barbieri RL. Serum steroid hormone profiles in postmenopausal smokers and nonsmokers. Fertil Steril. 1987;47:398–401. Khaw KT, Tazuke S, Barret-Connor E. Cigarette smoking and level of adrenal androgens in post-menopausal women. N Engl J Med. 1988:1705–1709. Wagenknecht LE, Burke GL, Perkins LL, Haley NJ, Friedman GD. Misclassification of smoking status in the CARDIA study: A comparison of self-report with serum cotinine levels. Am J Public Health. 1992;82:33–36.