International Nutrition Consultant Association

After stratifying by age, gender, and ethnic group, and adjusting for age, education, region, and health status, infrequent attenders (<24 times/year) were much more likely to be smokers than frequent attenders; odds ratios (95% confidence limits) ranged from 1.74 (1.45–2.10) to 3.06 (1.86–5.03). Among current smokers, frequent attenders smoked an average of 1–5 fewer cigarettes per day. Using serum cotinine ≥14 ng/mL as the gold standard for current smoking, under-reporting of smoking did not vary appreciably with frequency of attendance: false negative percentage for never smokers 3.1% in frequent attenders, 4.2% in others.

Conclusions.

Introduction

Cigarette smoking, a prevalent and well-established cardiovascular, cancer, and pulmonary risk factor in the US, is associated with social and cultural factors [1], [2] and [3]. Religious affiliation, attendance at religious services, and other religious behavior are more prevalent in the US than in any other industrialized nation [4]. Yet few population-based data have been published on the association of smoking with religiousness in the US population [5], [6], [7], [8], [9], [10], [11], [12] and [13]. In the four-city Coronary Artery Risk Development in Young Adults Study (CARDIA), 23% of participants aged 20–32 years who attended religious services at least monthly in 1987–1988 reported current smoking compared to 34% of those attending less frequently or never [5]. After adjusting for ethnicity, gender, region, and other confounders, infrequent or never attenders were still 50% (95% CI 30–80%) more likely to report current smoking and, among smokers, smoked more than frequent attenders. In the Duke (North Carolina) Populations for Epidemiologic Studies of the Elderly (EPESE) survey, similar findings were reported for African and European Americans aged 65 and older [6]. Population-based data for ages 35–64 or for Hispanic Americans were not found. Self-reported data underestimate the smoking population by 1–4.2% compared to serum cotinine levels [14]. It is not known whether degree of religiousness affects the amount of underreporting of smoking in surveys. In order to test the hypothesis that frequent attendance at religious services is associated with lower prevalence, later initiation, and lesser intensity of cigarette smoking among adults of all ages independent of age, gender, ethnicity, education, region, and health status in the American population, data on a large, multi-ethnic sample of adults from the Third National Health and Nutrition Examination Survey (NHANES III) were examined. An additional hypothesis was that frequency of attendance at religious services did not affect the rate of underreporting of cigarette smoking as determined by comparison of data from self-report and serum cotinine, a principal metabolite of nicotine used to measure tobacco exposure in epidemiologic studies.

Methods

The Third National Health and Nutrition Examination Survey (NHANES III) was conducted in 1988–1994 on a nationwide multi-stage probability sample of 39,695 persons from the civilian, non-institutionalized population aged 2 months and over of the United States excluding reservation lands of American Indians. Details of the survey plan, sample design, operations, and response rates have been published as have procedures used to obtain informed consent and to maintain confidentiality of information obtained [15] and [16].

Demographic data including race category and Mexican American (MA) ethnicity, years of education completed, medical history including self-assessed health status and behavioral information including frequency of attending church or religious services and smoking history were collected by household interview [15]. Persons were asked, “How often do you attend church or religious services?” Values ranged from 0 to 1825 times per year. Values in excess of 365 (n = 7) were coded missing as well as four “don’t know” responses. Smoking status was determined by responses to the following questions: “Have you smoked at least 100 cigarettes during your entire life?” Those responding “yes” were asked, “How old were you when you first started smoking cigarettes fairly regularly?” and “Do you smoke cigarettes now?” Current smokers who responded “yes” were asked, “About how many cigarettes do you smoke per day.” Persons reporting “less than 1 cigarette per day” were assigned a value of 0.5 cigarettes per day in this report, the minimum value; the maximum value was 140. Those responding “don’t know” or “varies/varied” were coded missing. Among 33,994 interviewed persons (85.6% of 39,695 sample persons), 15,169 were under age 20 and therefore excluded. Another 16 were excluded for missing ever-smoker status and 35 for missing frequency of church attendance. Complete data were available for 18,774 persons, 99.7% of 18,825 interviewed persons aged 20 and over, forming the present analysis sample.

Examinations were carried out in a mobile examination center. Blood samples were obtained at the examination center. Blood in a red-top Vacutainer tube was allowed to stand for 45 min at room temperature to allow complete clotting and clot retraction. Samples were centrifuged at 1500 × g for 30 min at 4°C. Serum cotinine levels were measured by high-performance liquid chromatography and atmospheric-pressure chemical ionization tandem mass spectrometry [14], [15], [16] and [17]. Previous researchers have used the serum cotinine value of <14 ng/mL to identify nonsmokers. Serum cotinine is directly proportional to absorbed nicotine and has a half-life of approximately 20 h [14]. It is considered a better marker of smoking status than self-reported tobacco use. It has been used in a number of reports to document underreporting of smoking in various groups [14]. Cotinine data were available for 15,512 (83%) of the 18,774 persons for the validity sub-analysis.

Statistical analysis

Detailed weighted descriptive statistics including percentiles and inter-quartile range (IQR) were computed using the Statistical Analysis System (SAS) [18]. Multivariate logistic regression analysis was used to develop models to control for confounding of the association of smoking status with frequency of attendance at religious services [18]. Multivariate linear regression was used for the outcome of number of cigarettes per day among current smokers. Because of documented age, gender, and ethnic differences in religiousness and smoking, interaction terms were initially included for age, gender, and ethnicity with attendance. Because these were significant, separate analyses were conducted for women and men of each major ethnic group aged 20–59 and 60+. All models controlled for age in years within these groups. Regression analyses were performed using SUDAAN procedures, techniques that incorporated sampling weights and design features of the survey [19] and [20].

Results

Table 1 shows the distribution of frequency of attendance at religious services (times per year) in persons aged 20 and over by gender. For all persons 20 and over, the median number of times/year was 8 with 26% reporting weekly attendance (52 times/year). About 5% attended twice a week or more. There was marked digit preference for multiples of 12. Over one-third (39.6%) reported never attending. Persons reporting attending 24 or more times per year (43.1%) were considered frequent attenders: 40.5% of 20–59 year olds, 52.0% of 60+ year olds; 49.3% of women, 36.3% of men; 41.4% of European Americans (EA), 51.3% of African Americans (AA), 47.7% of Mexican Americans (MA), and 45.5% of others. Forty-six percent of men and 34% of women never attended services. Percentages never attending were EA 41.6%, AA 28.6%, MA 31.8%, and other races 40.4%. Among gender, ethnic groups median times/year attending in descending order were as follows: AA, MA women and other women 24, and EA women, AA and MA men 12, EA men 2, and other men 1.

Table 1 also shows the distribution of selected smoking-related variables by gender. Among all persons 20 and over, 54.3% reported smoking at least 100 cigarettes (5 packs) in their lives. Among all ever smokers, the median age at which they started smoking regularly was 17 (IQR 15–19), with 90% starting by age 22. Only 5.9% had started by age 10. Only 52.8% of ever smokers were still smoking at the time of interview. The prevalence (%) of current smoking in men and women by ethnicity was 31.5 in EA men, 39.8 in AA men, 30.6 in MA men, 32.9 in other men, 26.7 in EA women, 28.2 in AA women, 14.8 in MA women, 12.7 in other women. Prevalence was 32.4% at 20–59 years old and 15.7% at 60+ years old. Among current smokers, the median number of cigarettes per day smoked was 20 (IQR 10–25) at ages 20+ and at ages 20–59 and 60+. Median numbers were 20 (IQR 12–30) among EA, 10 (IQR 3–20) among AA, 6 (IQR 2–15) among MA, and 10 (IQR 6–20) among others.

Among ever smokers, median age (years) at initiation of smoking was 17 (IQR 15–20) among frequent attenders and 16 (IQR 15–19) among others. Among current smokers, frequent attenders smoked a median number of 16 (IQR 3–20) cigarettes/day compared to others who smoked a median of 20 (IQR 10–30) cigarettes/day. In tables stratified by gender, the difference was limited to women; in tables stratified by age, the difference was limited to persons over 60 years of age (i.e., apparent effect modification by gender and age). In women, the prevalence of never smokers was higher among women who were frequent-attenders than other women, with prevalence of former smokers being similar between groups. In men, frequent attenders were more likely to be never smokers as well as former smokers compared to other men.

Given the effect modification by ethnicity (not shown), age, and gender, a logistic regression model was fit within gender, age, and ethnic strata, with smoking status (current smoker vs. non-smoker) as the dependent variable and frequent-attender status (yes, no) as the exposure variable in persons aged 20+ years adjusting for age. Therefore, Table 2 shows odds ratios (OR) and 95% Wald confidence limits within strata. Infrequent attenders were about twice as likely to be smokers even after controlling for age, education, region, and health status. Adjustment for multiple confounders produced little change in OR in most groups; however, there was negative confounding (OR increased) in older MA and AA men.

Among current smokers, a linear regression model was fit with number of cigarettes smoked per day as the dependent variable and frequent attender status (yes/no) as the exposure variable. Frequent attenders smoked on average 3.57–3.85 fewer cigarettes/day than others among men and from 1.00 to 4.65 fewer per day among women after controlling for age, education, and region (Table 3). The association was strongest in EA men, older EA women, and MA men. Relative lack of variation in attender status and smoking intensity in MA women may explain the lack of association observed in this group.

Serum cotinine concentration was available for 83% of the persons with smoking history data. Among male frequent attenders the median concentration was 0.307 ng/mL (IQR 0.103–17.4) and among male infrequent attenders the median was 2.17 ng/mL (IQR 0.224–227). Among female frequent attenders the median concentration was 0.176 ng/mL (IQR 0.066–1.210) and among female infrequent attenders the median was 0.549 ng/mL (IQR 0.103–144). Among frequent attenders the percent with cotinine ≥14 ng/mL (current smokers by cotinine) was 18.9 compared to 33.7 in others, consistent with results above for self-reported current smoking. The percentages of self-reported never and former smokers with serum cotinine ≥14 ng/mL, i.e., false negatives for smoking, are shown in Table 4. Rates of false negatives by self-report were higher in former than never smokers by self-report and in men than women. However, there was no evidence for important or systematic variation in rate of under-reporting by frequency of attendance at religious services.

Discussion

Cigarette smoking is one of the leading preventable causes of death in the United States and most industrialized countries [21]. Despite decades of anti-smoking campaigns and declines in prevalence among middle-aged and older adults, prevalence rates remain high and trends among adolescents are unfavorable [1], [2] and [3]. Taking advantage of high rates of religious affiliation and attendance at religious services in the US [4], public and private health agencies have recently sought to cooperate with religious institutions in faith-based disease prevention and health promotion activities such as smoking prevention and cessation [22]. Further, physicians and other members of the health care team have been advised to assess the role religious beliefs and practices play in patients' risk behaviors and susceptibility and response to illness [23]. To design and assess the likely effectiveness of such strategies, a better understanding of the effect of religious behaviors such as attendance at services on risk factor prevalence, incidence, and response to intervention is desirable.

In this national sample of Americans aged 20 and over, smoking was much less prevalent among frequent attenders of religious services than among infrequent or never attenders independent of Hispanic ethnicity, age, and gender. Further, among smokers, frequent attenders smoked significantly fewer cigarettes per day. Contrary to expectation, among ever smokers, reported age at initiation of regular smoking did not vary by current frequency of attendance. In this first report with biochemical validation of smoking history, the null hypothesis that frequent attenders did not under-report smoking more than infrequent attenders was supported. Thus, the current study in a national sample confirms the few previous findings in African and European Americans that suggested a possible beneficial effect of religiousness on cigarette smoking initiation and prevalence and extends them to Mexican Americans.

Mechanisms

Possible mechanisms by which high frequency of attendance at religious services may cause decreased prevalence of cigarette smoking most likely derive from the correlation of this single behavior with other dimensions (e.g., intrinsic and extrinsic) of religiousness. High religiousness may reduce smoking prevalence by simple direct effects such as the prohibition of smoking among Seventh Day Adventists and Mormons and the more recent phenomenon of church-sponsored health promotion activities including anti-smoking messages for members and surrounding communities [13]. Because of its association with other disreputable behavior such as excessive drinking and gambling, many Protestant Christian denominations strongly discouraged smoking prior to the 1940s, especially among women and youths [24]. After World War II, smoking became more widely acceptable. However, as public recognition of the health hazards of smoking grew since the 1964 Surgeon General's Report, leaders and members of religious institutions shared in that recognition. Today, smoking is not countenanced in many religious buildings or functions as in many other public places. Unhealthful habits such as use of tobacco and other drugs are generally discouraged in religious teaching, because they do not honor the body as the temple of the spirit [13] and [25]. These swings over the last century are typical of organized religion's ambivalence towards tobacco use ever since Caribbean Indians introduced the substance to Europeans and others at first contact, repeatedly cycling from condemnation and outright banning to tolerance to active use by clergy and laity, largely due to external influences [24].

Indirect mechanisms for an effect of religiousness on smoking include religion's ability to reduce the emotional impact of stressful life situations, prevent depression, and enhance coping, thus reducing the need to use smoking or other drugs to relieve the effects of stress [26]. Religious communities provide social support for healthy behaviors such as tobacco avoidance. Available evidence suggests that religiousness reduces smoking prevalence by reducing initiation among adolescents and young adults [5], [6], [7], [8], [9], [10], [11], [12] and [13]. The data on religiousness and smoking cessation among adults are conflicting among the few longitudinal studies, the two largest showing no effect [5] and [6] and one suggesting a positive effect [11]. In the latter in Alameda County, CA, women attending services weekly were over twice as likely and men 78% more likely to quit cigarette smoking as other women and men independent of several covariates [11]. Weekly attenders were also 45% less likely to start smoking, although the number starting was small. A large study in college students suggested that the protective effect of religiousness may be mediated by use of religion and problem-focused coping strategies [9]. However, in cross-sectional observational studies, non-causal mechanisms for the observed association of frequency of attendance and smoking must also be considered, e.g., non-smokers may feel more comfortable than smokers at non-smoking events such as religious services and hence are more likely to attend.

Comparisons with previous reports

Previously published studies of frequency of attendance at religious services or church membership and smoking have generally found a negative association [7]. NHANES III is the first study to provide population-based data on the association of attendance at religious services and cigarette smoking in nation-wide representative samples of US Mexican Americans and non-Hispanic European and African Americans. No other reports in Mexican Americans or other Hispanics were found. The CARDIA Study of over 4000 young adults in four US cities reported similar findings in European and African Americans [5]. Infrequent attendance at religious services was also associated with higher smoking initiation rates, but not with different cessation rates compared to frequent attenders. In nearly 4000 persons aged over 65 years in North Carolina, persons who attended religious services at least once per week were 25% more likely not to be current smokers than those who attended less frequently even after controlling for age, gender, race, education, alcohol use, and health status [6]. In African American men in Virginia, frequent religious attendance was the strongest correlate of non-smoking status [8].

Other dimensions of religiousness, not assessed in NHANES III, have also been found to be negatively associated with smoking in most studies [6], [7], [8], [9], [10], [11], [12], [27], [28], [29], [30], [31], [32], [33] and [34]. These include frequency of private religious activities such as private prayer and Bible study, and personal religiousness or religious commitment [7]. Persons who both attended religious services and prayed or studied the Bible frequently were almost 90% more likely to be nonsmokers than persons who did neither [6]. In studies among smokers, both infrequent attendance and/or infrequent private devotions were associated with heavier smoking but were not associated with cessation rate. In a study of female twins in Virginia, higher levels of several measures of religiousness were associated with lower smoking prevalence and lifetime risk for nicotine addiction [7]. Several dimensions of religiousness were negatively associated with smoking initiation and frequency among college students [9], [26], [27] and [28]. In a sample of African American women in Virginia, however, variables on a religiousness scale were not associated with smoking [27]. In Nigeria, high religious commitment was associated with less smoking in both Muslim and Christian college students [7]. In nearly 2000 female Virginia twins, personal devotion and personal and institutional conservatism were inversely related to current smoking and lifetime risk for nicotine dependence [29].

Limitations of the present study include possible bias arising from survey non-response and from missing values for some variables. Special studies of NHANES data have indicated little bias due to non-response [35]. At least 12 dimensions of religiousness have been defined [36]. Attendance at religious services is an indicator of organizational religiousness. Since data on multiple dimensions were unavailable in NHANES III, it was used because it is correlated with other dimensions of public and private religiousness and provided data that are directly comparable with a body of research data on this variable spanning many decades. Modest over-reporting of religious attendance is likely [37]; however, the NHANES III variable should serve well to separate frequent from infrequent attenders.

Adequate reliability has been documented for self-reported smoking in a number of populations [14]. Validity for current smoking prevalence and intensity is good and is supported in NHANES III by the consistency of findings of self-report and serum cotinine analysis. Adequate reliability has been demonstrated for serum cotinine [14] and [16]. Day-to-day variability in serum cotinine would tend to bias reported associations towards the null. Blood collection conditions in NHANES III were standardized [15].

The lack of a single, generally accepted measurement protocol for religiousness in epidemiologic studies remains a problem for inter-study comparisons, perhaps explaining in part inconsistencies among studies [37]. Confounding by variables not controlled for cannot be excluded. However, given the uncertainty about the nature of the association, it is unclear which other variables should be controlled for as confounders. Despite the large overall sample size in NHANES III, statistical power was limited for some subgroups, such as elderly Mexican American women. The number of tests was restricted to those of regression models. The representativeness of the sample and the use of sample weights provide generalizability of the results to United States non-institutionalized population of the same ages, but not necessarily to other nations or smaller US ethnic groups such as American Indians.

Future research should include large surveys of multiple dimensions of religiousness/spirituality and longitudinal studies of smoking initiation, duration, and cessation in population-based samples of non-Hispanic European American and African American and Hispanic American men and women to determine the temporal sequence of the relationship. For example, future surveys could include a multi-dimensional instrument, such as those developed by the National Institute on Ageing and Fetzer Institute or Duke University, to measure religiousness and/or spirituality for study as an exposure, confounder, or effect modifier in epidemiologic studies [37] and [38]. Serum cotinine should be used to confirm the validity of self-reported smoking data in populations with high rates of misclassification based on self-report.

Background.

Methods.

Results.