- Email: [email protected]
Adolescent Obesity as a Risk Factor for High-Level Nicotine Addiction in Young Women
Journal of Adolescent Health 49 (2011) 511–517
www.jahonline.org Original article
Adolescent Obesity as a Risk Factor for High-Level Nicotine Addiction in Young Women Aliya Esmail Hussaini, M.D., M.Sc.a,b,*, Lisa Marie Nicholson, Ph.D.c, David Shera, Sc.D.d, Nicolas Stettler, M.D., M.S. C.E.e, and Sara Kinsman, M.D., Ph.D.f a
Robert Wood Johnson Clinical Scholars Program, University of Pennsylvania, Philadelphia, Pennsylvania Michael & Susan Dell Foundation, Austin, Texas Institute for Health Research and Policy, Department of Public Health, University of Illinois at Chicago, Chicago, Illinois d Merck Research Laboratories, WestPoint, Pennsylvania e Division of Gastroenterology, Hepatology, and Nutrition, Children’s Hospital of Philadelphia, Philadelphia, Pennsylvania f Division of Adolescent Medicine, Children’s Hospital of Philadelphia, Philadelphia, Pennsylvania b c
Article history: Received September 28, 2010; Accepted April 1, 2011 Keywords: Adolescent obesity; Smoking; Mediation analysis
A B S T R A C T
Purpose: Obesity and cigarette smoking are two of the most frequent and preventable causes of disease and death in the United States; both are often established during youth. We hypothesized that obese, adolescent girls would be at higher risk for nicotine addiction in young adulthood, and that particular individual and social factors would mediate this association. Methods: Students surveyed in the National Longitudinal Study of Adolescent Health (Add Health), a nationally representative school-based and in-home survey conducted in three waves, comprised the sample. More than 4,000 respondents were used for the multivariate linear and logistic regression analyses used to determine the association between obesity and level of nicotine addiction. Potential mediation effects of the association were also examined. Results: Obesity doubled the risk of the highest level of nicotine addiction after controlling for demographic factors, parent and friend smoking, and baseline smoking (OR, 2.12; 95% CI, 1.22–3.68). Family smoking was the strongest predictor of nicotine addiction (OR, 4.72; 95% CI, 2.89 –7.72). Grade point average was a partial mediator of this relationship (OR, .48; 95% CI, .32–.74). Conclusions: Obese, adolescent females are at increased risk for high-level nicotine addiction in young adulthood as compared with their nonobese peers. Grade point average partially mediates the association, and may represent a confluence of factors including increased absenteeism, social marginalization, biases, and lack of confidence in academic ability. Obese, adolescent females may require targeted interventions to address their risk of subsequent high-level nicotine addiction, especially if risk factors such as parental smoking and poor school performance are present. 䉷 2011 Society for Adolescent Health and Medicine. All rights reserved.
Obesity and cigarette smoking are two of the most frequent and preventable causes of disease and death in the United States. Approximately nine million adults, 4.7% of the adult U.S. population, report having both a body mass index (BMI) of ⱖ30 and
* Address correspondence to: Aliya Esmail Hussaini, M.D., M.Sc., Michael & Susan Dell Foundation, PO Box 163867, Austin, TX 78716. E-mail address: [email protected] (A.E. Hussaini).
smoking cigarettes [1]. These two health risks are individually associated with an increased likelihood of mortality and morbidity. The combination of obesity and smoking results in a 2.7- to 3.8-fold increased risk for mortality, as compared with nonobese, nonsmokers [2]. Of particular concern are findings that obesity results in a 1.8 times greater disease burden (in health-related quality of life and mortality) for women than for men [3] and that women who smoke lose the cardiovascular health advantages generally enjoyed by women who do not smoke [4]. Further,
1054-139X/$ - see front matter 䉷 2011 Society for Adolescent Health and Medicine. All rights reserved. doi:10.1016/j.jadohealth.2011.04.001
512
A.E. Hussaini et al. / Journal of Adolescent Health 49 (2011) 511–517
recent reports suggest that gender may play a moderating role in the relationship between body weight and smoking, with overweight or obese adult women more likely to be current smokers than overweight or obese men [5]. Both adult obesity and smoking are health risks often established during childhood and adolescence. Currently, 18% of U.S. adolescents are obese [6]; 50%– 80% of obese adolescents will become obese adults [7]. Among adult smokers, 80% report beginning to smoke before the age of 18 years [8]. Interventions aimed at decreasing the concurrent health risks associated with adult obesity and cigarette smoking will need to include adolescents. Distinct vulnerabilities and protections related to obesity and risk behavior initiation have been described for males and females [9,10]. Adolescent females have the additional stress of conforming to a lean body image ideal, which has been noted in obese girls who are at least as young as 5 years old [11]. Understanding the unique factors that promote smoking cessation among adolescent females will be critical to our ability to intervene to prevent the concurrent health risks of obesity and smoking. To assess those females experiencing the greatest comorbidities of obesity and smoking, the goals of this study are to describe the relationship between adolescent obesity at wave 1 and level of nicotine addiction at wave 3 (6 years later), and further, to determine which psychosocial factors, measured at wave 2, mediate any existing relationship. Methods Sample The data for this study were drawn from waves 1, 2, and 3 in-home surveys of the National Longitudinal Adolescent Health Survey (Add Health), a school-based sample of students in grades 7 through 12. The schools selected for the study comprise a stratified sample of all U.S. high schools systematically selected proportional to their size and nationally representative with respect to region, urbanicity, school size, school type, and ethnicity. A total of 134 schools, including high schools and feeder middle schools, participated in the study. Between April and December 1995, 20,745 students were selected to participate in an in-home survey. Of these, 14,738 students were re-interviewed at wave 2, between April and August, 1996. Six years after the first in-home survey, 15,197 students were interviewed at wave 3. Figure 1 shows exclusion criteria resulting in a sample of 5,680 female adolescents, 4,102 (72% of the sample) of whom had valid data required for the primary analysis [12]. Because some students had timed out of school at wave 2, there was a population of the primary analysis subjects (about 1,500) who were missing data on grade point average (GPA), which was required for the mediator analysis. Rather than excluding these students with listwise deletion, we chose to use the missing data analysis procedure, mutivariate imputation by chained equations (ICE), available in Stata (StataCorp, College Station, TX) [13]. Measures Independent variables (wave 1). BMI based on self-reported data has previously been validated in Add Health [14]. Obese females were those whose calculated BMI at wave 1 was ⱖ95th percentile based on the CDC growth charts. For participants aged ⱖ18
Figure 1. Flowchart depicting determination of study sample.
years at wave 1, obesity was defined as calculated BMI of ⱖ30. A BMI of ⱖ30 is consistent with the CDC definition for obesity starting at the age of 20 years. We applied the adult definition of obesity (BMI, ⱖ30) to the 18- and 19-year-olds in the sample for consistency and in consideration of evolving views related to categorization of this transitional age group [6,15]. Height, weight, race, age, and menarchal age were self-reported. For the 139 females who failed to report menarchal age, 12.4 years was assigned based on the national average of girls of the same birth cohort [16]. Socioeconomic status (SES) was based on self-report of parental education and occupation using previously described scoring methods, with higher score indicating higher SES [17]. Parental smoking was self-reported as yes if one or both resident parents smoked. Number of friends who smoked was selfreported and ranged from one to three. Given that the outcome variable was level of nicotine addiction, wave 1 smoking was not an exclusion criterion, but rather served as an indicator of baseline smoking, and was defined continuously as the number of cigarettes smoked per month. Potential mediators (wave 2) of the relationship between obesity (wave 1) and level of nicotine addiction (wave 3). Depressive symptoms were measured using the previously validated Center for Epidemiologic Studies Depression scale of 18 items in Add Health (Cronbach’s ␣ ⫽ .88), resulting in a continuous scale with higher scores indicative of increased depressive symptoms [18,19]. Self-esteem was measured using the Rosenberg SelfEsteem scale, consisting of six items in Add Health (Cronbach’s ␣ ⫽ .86) [19,20], resulting in a continuous scale that was reverse coded such that higher scores were indicative of lower selfesteem. Family (Cronbach’s ␣ ⫽ .83) and school (Cronbach’s ␣ ⫽ .82) connectedness were scales previously identified in Add Health, with higher scores indicative of increased connectedness [12,21,22]. Student GPA was calculated by averaging numerical equivalents of self-reported letter grades in English, Mathematics, History, and Science. Outcome variable (wave 3). The use of self-report smoking data among adolescents has been previously validated [23]. Subjects reporting never having tried cigarette smoking, not even one or
A.E. Hussaini et al. / Journal of Adolescent Health 49 (2011) 511–517
two puffs, were classified as never smokers. Those reporting that they had tried cigarette smoking were classified as ever smokers. Among the ever smokers, those who reported ever having smoked regularly (at least one cigarette per day for at least 30 days) and smoking currently (having smoked a cigarette in the last 30 days) were eligible to complete the Fagerstrom Test for Nicotine Dependence (FTND) to determine level of nicotine addiction using previously determined scoring cutoffs [24]. Those scoring 0 –2 were categorized as not addicted; 3– 4, low addiction; 5, moderate addiction; and ⱖ6, high addiction. Never smokers and ever smokers (those not regularly and currently smoking) were not eligible to complete the FTND and were simply assigned to the 0 –2 category, also referred to as the “not addicted” nicotine category. The FTND is a modified version of the Fagerstrom Tolerance Questionnaire which has been validated for use in adolescents [25]. Statistical analysis All analyses were conducted using Stata 11E and surveycorrected statistical procedures to correct for the clustered and stratified nature of the Add Health sample [26]. To impute the data required for the modeling analysis among subjects missing data, the missing data analysis procedure ICE, a multivariate imputation procedure by chained equations available in Stata [13], was used. Five copies of the data, each with missing values imputed, were independently assessed in the multivariate multinomial logistic regression analyses. Estimates of parameters of interest were averaged across the five copies to give a single mean estimate, and standard errors were adjusted according to Rubin’s rules [27]. The imputation procedure included all of the variables in models 1 and 2, listed in Table 2. Selection of potential mediators was based on previously reported factors known to be associated with obesity and smoking among females and available in the Add Health data set. To complete the mediation analysis as described by Baron and Kenny, demonstration of three relationships was required [28]. In the first step, multinomial logistic regression was used to determine the association between wave 1 obesity and level of wave 3 nicotine addiction (Figure 2, relationship A). Covariates in this model included demographic variables of race, age, menarchal age, and SES as well as parent and friend smoking and wave 1 smoking. In the second step, linear regression models were used to assess the relationship between obesity and each potential mediator of obesity individually, including depressive symptoms, low self-esteem, family connectedness, school connectedness, GPA, trying to lose weight, and physical activity (Figure 2, relationship B). All aforementioned covariates were included in this second set of models. The third step used multiple logistic regression models to individually assess the relationship between each potential mediator and level of wave 3 nicotine addiction (Figure 2, relationship C). All aforementioned covariates were included in this third set of models. To avoid false negatives by prematurely excluding potential mediators, the inclusion criteria of potential mediators was widened (p ⬍ .10, two-tailed) in our testing of relationships B and C. Those potential mediators that were determined to have significant relationships with obesity or level of nicotine addiction were further evaluated using the coefficient product test. A significant test (p ⬍ .05, two-tailed test) is evidence for partial mediation. Mediators that were significant based on the coefficient product test were included individually in a final model
513
used to determine whether the relationship between obesity and level of nicotine addiction (represented by the  coefficient of obesity in the model) was reduced by the mediator being tested. The degree to which  is reduced when the potential mediators are added to the equation is indicative of the strength of the mediation. Confirming full mediation requires that the coefficient corresponding to the mediated effect is near zero when the mediator term is included in the model. Results Descriptive characteristics of sample Of the sample, 439 females (11%) had a BMI of ⱖ95th percentile for age and gender at wave 1. The decision to use the adult definition of obesity (BMI, ⱖ30) for 18- and 19-year-olds affected a single 19-year-old female in our sample whose BMI of 30.23 resulted in her being classified as obese based on adult criteria and nonobese based on the 95th percentile BMI for her age. In all, 57% of the sample was White, 18% Black, 16% Hispanic, 7% Asian, 2% Native American, and ⬍1% other. The average age at wave 1 was 15 (⫾1.6) years and the average age at wave 3 was 21 (⫾1.6) years. Table 1 shows weighted differences between obese and nonobese female adolescents in level of wave 3 nicotine addiction,
Table 1 Survey-weighted characteristics of sample participants who completed waves 1, 2, and 3 of Add Health by obesity status Characteristics
Self-reported nicotine addiction (wave 3) High addiction: % (N) Moderate addiction Low addiction Not addicted Characteristics (wave 1) Race: % (N) White Black Hispanic Asian Native American Other race Age: mean years (SE) Menarchal age: mean years (SE) Family SES: mean (SE) Resident parent smokes: % families (N) Friends smoke: Number of friends (SE) Cigarettes per month (W1): Number of cigarettes (SE) Mediators (wave 2) means from imputed data Depressive symptoms: mean (SE) Low self-esteem: mean (SE) Family connectedness: mean (SE) School connectedness: mean (SE) GPA: mean (SE)
Obese N ⫽ 439
Nonobese N ⫽ 3,663
8.43 (37) 2.28 (10) 6.38 (28) 82.92 (364)
3.90 (143) 3.17 (116) 5.68 (208) 87.25 (3,196)
50.11 (220) 23.46 (103) 18.22 (80) 3.64 (16) 3.87 (17) .68 (3) 14.98 (.08) 11.82 (1.26)
57.52 (2,107) 16.84 (617) 15.21 (557) 7.40 (271) 2.16 (79) .87 (32) 15.07 (.03) 12.15 (1.27)
5.46 (.12) 59 (259)
6.20 (.04) 50 (1,832)
p
.04* .40 .19 .91 ⬍.001**
.92 (.05)
.81 (.02)
.12 ⬍.001 ⬍.001 ⬍.001* .03
35.67 (119.55)
28.38 (105.78)
12.59 (.48)
11.06 (.19)
⬍.001
12.01 (.23) 22.20 (.17)
11.11 (.09) 22.32 (.06)
⬍.001 .03
17.23 (.36)
17.73 (.19)
⬍.001
2.68 (.05)
2.99 (.03)
⬍.001
Total of ⬍100% resulting from rounding. * p value for this addiction category versus all other others. ** RxC 2 test.
.26
514
A.E. Hussaini et al. / Journal of Adolescent Health 49 (2011) 511–517
Table 2 Survey-weighted multinomial logistic regression results (odds ratios and confidence intervals) predicting self-reported level of nicotine addiction (N ⫽ 4,102) Characteristics
Model 1 Low addiction
Independent variable Obesity Demographics and covariates Race (reference race: White) Black Hispanic Asian Native American Other race Higher age Higher menarchal age Higher family SES Pack (20 cigarettes) per month (W1)a Family smokes More friends smoke Mediators Higher depressive symptoms Lower self-esteem Higher family connectedness Higher school connectedness Higher GPA
Model 2 (mediators imputed) Moderate addiction
High addiction
Low addiction
Moderate addiction
High addiction
.95 (.56–1.59)
.80 (.35–1.81)
2.12 (1.22–3.68)
.87 (.52–1.46)
.73 (.32–1.65)
1.79 (.96–3.36)
1.00 .73 (.38–1.40) .37 (.18–.76) .55 (.19–1.56) 1.70 (.65–4.42) 3.91 (1.29–11.86) .77 (.69–.87) .93 (.81–1.08) .93 (.86–.99) 1.05 (1.02–1.08)
1.00 .45 (.21–.96) .19 (.06–.65) .22 (.03–1.39) 2.32 (.81–6.69) 3.28 (.39–27.36) .88 (.74–1.06) 1.01 (.83–1.22) .86 (.79–.94) 1.07 (1.04–1.09)
1.00 .29 (.13–.63) .25 (.12–.55) .65 (.20–2.10) 1.74 (.66–4.56) 2.38 (.45–12.49) .83 (.72–.95) .91 (.79–1.05) .92 (.84–.98) 1.10 (1.07–1.13)
1.00 .69 (.35–1.37) .30 (.14–.63) .51 (.18–1.46) 1.64 (.69–3.90) 4.08 (1.29–12.94) .77 (.68–.87) .94 (.81–1.10) .94 (.87–1.02) 1.05 (1.02–1.08)
1.00 .30 (.14–.66) .13 (.04–.44) .19 (.03–1.30) 2.06 (.71–5.92) 3.64 (.48–27.89) .84 (.69–1.02) 1.04 (.85–1.28) .93 (.84–1.02) 1.05 (1.02–1.08)
1.00 .22 (.09–.47) .15 (.06–.36) .50 (.14–1.85) 1.59 (.58–4.41) 2.76 (.55–13.82) .80 (.69–.91) .93 (.80–1.08) .99 (.89–1.09) 1.09 (1.05–1.12)
1.55 (1.11–2.14) 1.72 (1.45–2.03)
2.13 (1.26–3.60) 1.74 (1.41–2.15)
4.72 (2.89–7.72) 1.57 (1.27–1.94)
1.49 (1.07–2.08) 1.63 (1.38–1.92)
1.96 (1.14–3.34) 1.51 (1.21–1.89)
4.40 (2.68–7.21) 1.36 (1.09–1.70)
1.01 (.99–1.03) 1.06 (.99–1.13) 1.01 (.94–1.08) 1.01 (.97–1.04) .75 (.60–.94)
1.04 (1.01–1.08) .95 (⫺.88–1.03) .98 (.91–1.05) .97 (.93–1.00) .52 (.32–.83)
1.04 (1.01–1.07) 1.01 (.95–1.06) 1.01 (.95–1.08) .98 (.95–1.02) .48 (.32–.74)
Odds ratios are versus base outcome category of Fagerstrom score 0-2 (nonaddicted). Bolded values are significant at the p ⬍ .05 level. a Coefficients and confidence intervals reported as packs per month rather than as cigarettes per month.
demographic characteristics, parent and friend smoking, potential mediating variables, and wave 1 smoking. More than twice as many obese females experienced the highest level of nicotine addiction as nonobese females (8.43% vs. 3.90%, p ⫽ .04). Obese females were more likely to be Black, Hispanic, or Native American and were more likely to report younger menarchal age and to live in families with lower SES (p ⬍ .001). They were also more likely to live with parents who smoked, and more likely to have more friends who smoked. At wave 1, obese females reported a higher number of cigarettes smoked per month, although this difference was not statistically significant. With regard to potential mediators, obese females were significantly more likely to report depressive symptoms, lower self-esteem, lower school and family connectedness, and lower GPAs.
Model 1 results in Table 2 show that obesity more than doubled the risk of the highest level of nicotine addiction after controlling for all demographic variables, parent and friend smoking, and wave 1 cigarette smoking (OR, 2.12; 95% CI, 1.22– 3.68). Of all variables, parental smoking was the strongest predictor of the highest level of nicotine addiction (OR, 4.72; 95% CI, 2.89 –7.72). The second set of linear regression models demonstrated that wave 1 obesity was significantly associated with depressive symptoms, low self-esteem, and GPA. The third set of multiple logistic regression models showed that level of wave 3 nicotine addiction was significantly associated with depressive symptoms, low self-esteem, family connectedness, school connectedness, and GPA. Although all of these potential mediators were then evaluated using the coefficient product test, only GPA was found to be significant (p ⬍ .05, two-tailed). Model 2 results in Table 2 show that when GPA was included as a mediator, the relationship between obesity and high level nicotine addiction was reduced to statistical insignificance (p ⫽ .07), and higher GPA was strongly protective against high addiction (OR, .48; 95% CI, .32–.74). Discussion
Figure 2. Conceptual model of the mediated association between obesity and level of nicotine addiction in females.
Our findings suggest that obese, female adolescents are more than twice as likely to become highly nicotine-addicted smokers as their normal weight peers, even after accounting for demographic covariates, parent and friend smoking, and baseline smoking. However, in our sample, obese females were not significantly more likely to experience low or moderate levels of addiction. Further, the relationship between obesity and nicotine addiction is mediated by GPA. Previous research has explored the associations between obesity and low GPA, and separately the associations between low GPA and cigarette smoking. Ding reported that, on average, obese
A.E. Hussaini et al. / Journal of Adolescent Health 49 (2011) 511–517
515
Table 3 Obese females’ efforts to change weight by smoking statusa Response to Add Health Question ⬙What are you currently doing about your weight?⬙
Not addicted
Low addiction
Moderate addiction
High addiction
Total
Trying to lose weight: n (%) Trying to gain weight: n (%) Trying to stay the same weight: n (%) Not trying to do anything about weight: n (%) Totala
249 (68.4) 9 (2.5) 23 (6.3) 83 (22.8) 364 (100)
18 (64.3) 0 (0) 0 (0) 10 (35.7) 28 (100)
8 (80) 0 (0) 0 (0) 2 (20) 10 (100)
26 (70.3) 0 (0) 0 (0) 11 (29.7) 37 (100)
301 (68.6) 9 (2.1) 23 (5.2) 37 (100) 439 (100)
Pearson 2(9) ⫽ 9.2490, Pr ⫽ .415. a Total of ⬎100% resulting from rounding.
adolescents had GPAs that were .43 lower than their nonobese peers (p ⬍ .01) [29], and Falkner reported that obese females were 1.5 times more likely to be held back a grade. Perception of academic achievement is also lower among obese females, who are 2.1 times more likely to consider themselves as poor students as compared with their average-weight peers [30]. Interestingly, Huang found that females with higher BMIs underestimated their academic performance as compared with their thinner peers (p ⬍ .01), even when objective data did not support this belief [31]. Such perceptions may stem from the social and psychological impact of being an obese student. Obese adolescents are more likely to report being marginalized by peers, teased about weight issues, and bullied [9]. Even teachers report believing that overweight students are less academically capable [32]. Physical health issues may also affect school performance. This combination of factors may contribute to the finding that obese school children and adolescents are 4 times more likely than healthy peers to report impaired school function, and that obese students report increased rates of school absenteeism as compared with their nonobese peers [33]. Academic success reflects not only comprehension and mastery of material, but also consistent school attendance and school participation. We hypothesize that obese females may have less confidence academically and experience increased social marginalization and absenteeism, resulting in the lower GPA variable. Although the specific relationship between lower GPA and nicotine addiction has not been fully elucidated, research has demonstrated strong associations between lower academic achievement and cigarette smoking. Adolescents with a lower GPA are more likely to adopt smoking earlier and become regular smokers as compared with peers with higher GPAs [34]. This association may be explained in part by having less parental monitoring and a more relaxed parental attitude toward youth smoking [35]. Lower GPA and increased likelihood of smoking may also be associated with exposure to parental second-hand smoke which is implicated in children’s lower academic success and an increased likelihood of nicotine addiction during adolescence [36]. Students with lower GPAs may also spend more time with marginalized peers who smoke cigarettes [9,37]. Independently, the links between obesity and low GPA and low GPA and cigarette smoking are remarkable. A full understanding of the reasons why GPA predicts nicotine addiction and not just cigarette smoking in our sample will require further focused investigation. Obese adolescents may turn to smoking to make social connections with other marginalized peers or cope with the stress resulting from the academic and social consequences of obesity. The end result of this strategy may be nicotine addiction. Further understanding these associations and identification of potential intervention points may lead to effec-
tive interventions that decrease the likelihood of developing the dual health risks of obesity and nicotine addiction. In our sample, living with parents who smoked was the strongest predictor of nicotine addiction in young adults. This is consistent with previously published data that demonstrate family smoking as a major risk factor for smoking initiation among adolescents [38]. Although many researchers and clinicians assume that the strong effect of parental smoking on adolescent smoking behavior results from role-modeling smoking behavior or increased access to cigarettes, children and adolescents who are exposed to their parents’ (and peers’) second-hand tobacco smoke are also at risk for developing nicotine addiction [36]. Exploration of the link between parental and adolescent nicotine addiction is another important focal point for research and intervention. As illustrated in Table 3, our findings did not support the broadly accepted perception that obese teenagers smoke to lose weight. In previous studies, adolescents, especially girls, have reported beginning and continuing to smoke for weight control and weight loss. In fact, Strauss and Mir found a twofold increased risk of smoking among normal-weight adolescent girls who had been dieting. However, they found no increased risk of smoking among overweight adolescents (girls or boys) trying to lose weight [39]. This may be because overweight adolescents derive relatively less pleasure from nicotine as compared with food [40]. Thus, some obese adolescents may experiment with cigarettes but not progress to a regular habit because they find it less rewarding (e.g., fewer pleasurable effects, more unpleasurable effects), whereas other obese females, as suggested by our sample, may be at increased risk for nicotine addiction. There are several important limitations to these analyses. First, our sample reflects risks for those adolescents who completed all three waves of Add Health. The females who completed the three waves were more likely to be younger at wave 1, White, and have a higher SES. Older, minority, and lower SES females were not as likely to complete all three waves and be eligible for this longitudinal sample. The importance of GPA as a mediator may be different than in a more diverse sample. Second, our exploration of mediating covariates is limited by the available data in Add Health; inclusion of other covariates may deepen the understanding of this research question. Third, there is an uneven time frame between waves. Obesity and GPA are measured with a 1-year time lag in this longitudinal study and related to level of nicotine addiction 6 years after the exposure is measured. Given the 1-year time lag between waves 1 and 2, we explored the possibility that the potential mediators could be markers of the same constructs occurring even before the exposure variable. To test this hypothesis, we conducted supplemental analyses to examine potential reverse causality between wave 2 mediators
516
A.E. Hussaini et al. / Journal of Adolescent Health 49 (2011) 511–517
and the wave 1 obesity measure. Specifically, we ran regression models examining whether wave 2 mediators were significant predictors of obesity at wave 1. We found no significant associations. Although the short time lag between waves 1 and 2 is a significant limitation in Add Health, we believe that it is still one of the premier data sets for understanding longitudinal behavior change among U.S. adolescents. Further, we believe that the exploration of the relationship between obesity and nicotine addiction will be a helpful contribution to this critical health issue. Fourth, because all questions comprising the FTND were not included at wave 1, a true time lagged control for nicotine addiction was not possible. Although the difference in the average number of cigarettes smoked by obese and nonobese females at wave 1 was not statistically significant, number of cigarettes smoked at wave 1 was included in all modeling as a proxy control for the outcome variable. Fifth, family history of obesity was not controlled for. The only measure available was self-reported history of parental obesity from the in-home, but not necessarily the biological parent. This measurement was not thought to be accurate enough to be included in the analysis. Finally, given that Add Health data were collected over a period of 6 years, there were losses to follow-up from the baseline sample. However, the proportion of obese females and females who smoke was the same in wave 1 as our longitudinal study sample [33]. Therefore, the differential attrition that was noted in the demographic variables in the longitudinal sample did not result in disproportionate loss of obese females or female smokers. Acknowledgments At the time of manuscript preparation, Dr Aliya Hussaini was supported by the Robert Wood Johnson Foundation through the Robert Wood Johnson Clinical Scholars Program at the University of Pennsylvania. This research uses data from Add Health, a program project designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris, and funded by a grant P01-HD31921 from the National Institute of Child Health and Human Development, with cooperative funding from 17 other agencies. Special acknowledgment is due Ronald R. Rindfuss and Barbara Entwisle for assistance in the original design. Persons interested in obtaining data files from Add Health should contact Add Health, Carolina Population Center, 123 W. Franklin Street, Chapel Hill, NC 27516-2524 ([email protected]). The authors thank Janet Audrain-McGovern, Ph.D., Associate Professor of Psychology in the Department of Psychiatry and member of the Cancer Center at the University of Pennsylvania and Donald F. Schwarz, M.D., M.P.H., Deputy Mayor of Health and Opportunity and Health Commissioner for the City of Philadelphia for their thoughtful contributions and review of this manuscript. References [1] Healton CG, Vallone D, McCausland KL, et al. Smoking, obesity, and their co-occurrence in the United States: Cross sectional analysis. Br Med J 2006; 333:25– 6. [2] Freedman DM, Sigurdson AJ, Rajaraman P, et al. The mortality risk of smoking and obesity combined. Am J Prev Med 2006;31:355– 62. [3] Muennig P, Lubetkin E, Jia H, Franks P. Gender and the burden of disease attributable to obesity. Am J Public Health 2006;96:1662– 8. [4] Grundtvig M. Impact on nurse-based heart failure clinics on drug management and hospital admissions by self monitoring through a common database. Presented at: European Society of Cardiology Congress; September 3, 2008; Munich, Germany.
[5] Park E. Gender as a moderator in the association of body weight to smoking and mental health. Am J Public Health 2009;99:146 –51. [6] Ogden CL, Carroll MD, Curtin LR, et al. Prevalence of high body mass index in US children and adolescents 2007–2008. JAMA 2010;303:242–9. [7] Whitaker RC, Wright JA, Pepe MS, et al. Predicting obesity in young adulthood from childhood and parental obesity. N Engl J Med 1997;337: 869 –73. [8] Preventing tobacco use among young people: A report of the surgeon general. Atlanta, GA: Centers for Disease Control and Prevention. U.S. Department of Health and Human Services, Public Health Service, 1994. [9] Janssen I, Craig WM, Boyce WF, Pickett W. Associations between overweight and obesity with bullying behaviors in school-aged children. Pediatrics 2004;113:1187–94. [10] Tiggeman R, Rothblum ED. Gender differences in social consequences of perceived overweight in the United States and Australia. Sex Roles 1988; 18:75– 86. [11] Davison KK, Birch LL. Weight status, parent reaction, and self-concept in five-year-old girls. Pediatrics 2001;107:46 –53. [12] Resnick MD, Bearman PS, Blum RW, et al. Protecting adolescents from harm. Findings from the National Longitudinal Study on adolescent health. JAMA 1997;278:823–32. [13] Royston P. Multiple imputation of missing values: Update. Stata J 2005;5: 1–14. [14] Goodman E, Hinden BR, Khandelwal S. Accuracy of teen and parental reports of obesity and body mass index. Pediatrics 2000;106:52– 8. [15] Gordon-Larsen P, Adair LS, Nelson MC, Popkin BM. Five-year obesity incidence in the transition period between adolescence and adulthood: The National Longitudinal Study of adolescent health. Am J Clin Nutr 2004;80: 569 –75. [16] McDowell MA, Brody DJ, Hughes JP. Has age at menarche changed? Results from the National Health and Nutrition Examination Survey (NHANES) 1999 –2004. J Adolesc Health 2007;40:227–31. [17] Bearman PS, Moody J. Suicide and friendships among American adolescents. Am J Public Health 2004;94:89 –95. [18] Radloff LS. The use of the center for epidemiological studies depressions scale in adolescents and young adults. J Youth Adolesc 1991;20:149 – 66. [19] Warren JT, Harvey SM, Henderson JT. Do depression and low self-esteem follow abortion among adolescents? Evidence from a National Study. Perspect Sex Reprod Health 2010;42:230 –5. [20] Rosenberg M. Society and the adolescent self-image. Princeton, NJ: Princeton University Press, 1965. [21] Wainwright JL, Russell ST, Patterson CJ. Psychosocial adjustment, school outcomes, and romantic relationships of adolescents with same-sex parents. Child Dev 2004;75:1886 –98. [22] Pesa JA, Syre TR, Jones E. Psychosocial differences associated with body weight among female adolescents: The importance of body image. J Adolesc Health 2000;26:330 –7. [23] Patrick DL, Cheadle A, Thompson DC, et al. The validity of self-reported smoking: A review and meta-analysis. Am J Public Health 1994;84:1086 – 93. [24] Fagerstrom KO, Heatherton TF, Kozlowski LT. Nicotine addiction and its assessment. Ear Nose Throat J 1992;69:763–7. [25] Prokhorov AV, De Moor C, Pallonen UE, et al. Validation of the modified Fagerstrom tolerance questionnaire with salivary cotinine among adolescents. Addict Behav 2000;25:429 –33. [26] Chantalla K. Guidelines for analyzing add health data. Chapel Hill, NC: Carolina Population Center, University of North Carolina Chapel Hill, 2006. [27] Rubin DB. Inference and missing data (with discussion). Biometrika 1976; 63:581–92. [28] Baron RM, Kenny DA. The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. J Pers Soc Psychol 1986;51:1173– 82. [29] Ding W, Lehrer SF, Rosenquist JN, et al. The impact of poor health on education: New evidence using genetic markers, 2006 [NBER Working Paper No. W12304]. Available at: http://ssrn.com/abstract⫽910836. Accessed March 21, 2011. [30] Falkner NH, Neumark-Sztainer D, Story M, et al. Social, educational, and psychological correlates of weight status in adolescents. Obes Res 2001;9: 32– 42. [31] Huang TT, Goran MI, Spruijt-Metz D. Associations of adiposity with measured and self-reported academic performance in early adolescence. Obesity 2006;14:1839 – 45. [32] Puhl R, Brownell KD. Bias, discrimination and obesity. Obes Res 2001;9: 788 – 805. [33] Schwimmer J, Burwinkle T, Varni J. Health Related quality of life of severely obese children and adolescents. JAMA 2003;289:1813–9. [34] Audrain-McGovern J, Rodriguez D, Tercyak KP, et al. Identifying and characterizing adolescent smoking trajectories. Cancer Epidemiol Biomarkers Prev 2004;13:2023–34.
A.E. Hussaini et al. / Journal of Adolescent Health 49 (2011) 511–517
[35] Forrester K, Biglan A, Severson HH, Smolkowski K. Predictors of smoking onset over two years. Nicotine Tob Res 2007;9:1259 – 67. [36] Collins NB, Wileyto EP, Murphy MF, MunafÓ MR. Adolescent environmental tobacco smoke exposure predicts academic achievement test failure. J Adolesc Health 2007;41:363–70. [37] Jessor R. Risk behavior in adolescence: A psychosocial framework for understanding and action. In: Rogers D, Ginzburg E, eds. Adolescents at Risk: Medical and Social Perspectives. Boulder, CO: Westview Press, 1992:19 –34.
517
[38] Barman SK, Pulkkinen L, Kaprio J, Rose RJ. Inattentiveness, parental smoking and adolescent smoking initiation. Addiction 2004;99: 1049 – 61. [39] Strauss RS, Mir HM. Smoking and weight loss attempts in overweight and normal weight adolescents. Int J Obes 2001;25:1381–5. [40] Blendy JA, Strasser A, Walters CL, et al. Reduced nicotine reward in obesity: Cross comparison in human and mouse. Psychopharmacology 2005;180: 306 –15.
www.jahonline.org Original article
Adolescent Obesity as a Risk Factor for High-Level Nicotine Addiction in Young Women Aliya Esmail Hussaini, M.D., M.Sc.a,b,*, Lisa Marie Nicholson, Ph.D.c, David Shera, Sc.D.d, Nicolas Stettler, M.D., M.S. C.E.e, and Sara Kinsman, M.D., Ph.D.f a
Robert Wood Johnson Clinical Scholars Program, University of Pennsylvania, Philadelphia, Pennsylvania Michael & Susan Dell Foundation, Austin, Texas Institute for Health Research and Policy, Department of Public Health, University of Illinois at Chicago, Chicago, Illinois d Merck Research Laboratories, WestPoint, Pennsylvania e Division of Gastroenterology, Hepatology, and Nutrition, Children’s Hospital of Philadelphia, Philadelphia, Pennsylvania f Division of Adolescent Medicine, Children’s Hospital of Philadelphia, Philadelphia, Pennsylvania b c
Article history: Received September 28, 2010; Accepted April 1, 2011 Keywords: Adolescent obesity; Smoking; Mediation analysis
A B S T R A C T
Purpose: Obesity and cigarette smoking are two of the most frequent and preventable causes of disease and death in the United States; both are often established during youth. We hypothesized that obese, adolescent girls would be at higher risk for nicotine addiction in young adulthood, and that particular individual and social factors would mediate this association. Methods: Students surveyed in the National Longitudinal Study of Adolescent Health (Add Health), a nationally representative school-based and in-home survey conducted in three waves, comprised the sample. More than 4,000 respondents were used for the multivariate linear and logistic regression analyses used to determine the association between obesity and level of nicotine addiction. Potential mediation effects of the association were also examined. Results: Obesity doubled the risk of the highest level of nicotine addiction after controlling for demographic factors, parent and friend smoking, and baseline smoking (OR, 2.12; 95% CI, 1.22–3.68). Family smoking was the strongest predictor of nicotine addiction (OR, 4.72; 95% CI, 2.89 –7.72). Grade point average was a partial mediator of this relationship (OR, .48; 95% CI, .32–.74). Conclusions: Obese, adolescent females are at increased risk for high-level nicotine addiction in young adulthood as compared with their nonobese peers. Grade point average partially mediates the association, and may represent a confluence of factors including increased absenteeism, social marginalization, biases, and lack of confidence in academic ability. Obese, adolescent females may require targeted interventions to address their risk of subsequent high-level nicotine addiction, especially if risk factors such as parental smoking and poor school performance are present. 䉷 2011 Society for Adolescent Health and Medicine. All rights reserved.
Obesity and cigarette smoking are two of the most frequent and preventable causes of disease and death in the United States. Approximately nine million adults, 4.7% of the adult U.S. population, report having both a body mass index (BMI) of ⱖ30 and
* Address correspondence to: Aliya Esmail Hussaini, M.D., M.Sc., Michael & Susan Dell Foundation, PO Box 163867, Austin, TX 78716. E-mail address: [email protected] (A.E. Hussaini).
smoking cigarettes [1]. These two health risks are individually associated with an increased likelihood of mortality and morbidity. The combination of obesity and smoking results in a 2.7- to 3.8-fold increased risk for mortality, as compared with nonobese, nonsmokers [2]. Of particular concern are findings that obesity results in a 1.8 times greater disease burden (in health-related quality of life and mortality) for women than for men [3] and that women who smoke lose the cardiovascular health advantages generally enjoyed by women who do not smoke [4]. Further,
1054-139X/$ - see front matter 䉷 2011 Society for Adolescent Health and Medicine. All rights reserved. doi:10.1016/j.jadohealth.2011.04.001
512
A.E. Hussaini et al. / Journal of Adolescent Health 49 (2011) 511–517
recent reports suggest that gender may play a moderating role in the relationship between body weight and smoking, with overweight or obese adult women more likely to be current smokers than overweight or obese men [5]. Both adult obesity and smoking are health risks often established during childhood and adolescence. Currently, 18% of U.S. adolescents are obese [6]; 50%– 80% of obese adolescents will become obese adults [7]. Among adult smokers, 80% report beginning to smoke before the age of 18 years [8]. Interventions aimed at decreasing the concurrent health risks associated with adult obesity and cigarette smoking will need to include adolescents. Distinct vulnerabilities and protections related to obesity and risk behavior initiation have been described for males and females [9,10]. Adolescent females have the additional stress of conforming to a lean body image ideal, which has been noted in obese girls who are at least as young as 5 years old [11]. Understanding the unique factors that promote smoking cessation among adolescent females will be critical to our ability to intervene to prevent the concurrent health risks of obesity and smoking. To assess those females experiencing the greatest comorbidities of obesity and smoking, the goals of this study are to describe the relationship between adolescent obesity at wave 1 and level of nicotine addiction at wave 3 (6 years later), and further, to determine which psychosocial factors, measured at wave 2, mediate any existing relationship. Methods Sample The data for this study were drawn from waves 1, 2, and 3 in-home surveys of the National Longitudinal Adolescent Health Survey (Add Health), a school-based sample of students in grades 7 through 12. The schools selected for the study comprise a stratified sample of all U.S. high schools systematically selected proportional to their size and nationally representative with respect to region, urbanicity, school size, school type, and ethnicity. A total of 134 schools, including high schools and feeder middle schools, participated in the study. Between April and December 1995, 20,745 students were selected to participate in an in-home survey. Of these, 14,738 students were re-interviewed at wave 2, between April and August, 1996. Six years after the first in-home survey, 15,197 students were interviewed at wave 3. Figure 1 shows exclusion criteria resulting in a sample of 5,680 female adolescents, 4,102 (72% of the sample) of whom had valid data required for the primary analysis [12]. Because some students had timed out of school at wave 2, there was a population of the primary analysis subjects (about 1,500) who were missing data on grade point average (GPA), which was required for the mediator analysis. Rather than excluding these students with listwise deletion, we chose to use the missing data analysis procedure, mutivariate imputation by chained equations (ICE), available in Stata (StataCorp, College Station, TX) [13]. Measures Independent variables (wave 1). BMI based on self-reported data has previously been validated in Add Health [14]. Obese females were those whose calculated BMI at wave 1 was ⱖ95th percentile based on the CDC growth charts. For participants aged ⱖ18
Figure 1. Flowchart depicting determination of study sample.
years at wave 1, obesity was defined as calculated BMI of ⱖ30. A BMI of ⱖ30 is consistent with the CDC definition for obesity starting at the age of 20 years. We applied the adult definition of obesity (BMI, ⱖ30) to the 18- and 19-year-olds in the sample for consistency and in consideration of evolving views related to categorization of this transitional age group [6,15]. Height, weight, race, age, and menarchal age were self-reported. For the 139 females who failed to report menarchal age, 12.4 years was assigned based on the national average of girls of the same birth cohort [16]. Socioeconomic status (SES) was based on self-report of parental education and occupation using previously described scoring methods, with higher score indicating higher SES [17]. Parental smoking was self-reported as yes if one or both resident parents smoked. Number of friends who smoked was selfreported and ranged from one to three. Given that the outcome variable was level of nicotine addiction, wave 1 smoking was not an exclusion criterion, but rather served as an indicator of baseline smoking, and was defined continuously as the number of cigarettes smoked per month. Potential mediators (wave 2) of the relationship between obesity (wave 1) and level of nicotine addiction (wave 3). Depressive symptoms were measured using the previously validated Center for Epidemiologic Studies Depression scale of 18 items in Add Health (Cronbach’s ␣ ⫽ .88), resulting in a continuous scale with higher scores indicative of increased depressive symptoms [18,19]. Self-esteem was measured using the Rosenberg SelfEsteem scale, consisting of six items in Add Health (Cronbach’s ␣ ⫽ .86) [19,20], resulting in a continuous scale that was reverse coded such that higher scores were indicative of lower selfesteem. Family (Cronbach’s ␣ ⫽ .83) and school (Cronbach’s ␣ ⫽ .82) connectedness were scales previously identified in Add Health, with higher scores indicative of increased connectedness [12,21,22]. Student GPA was calculated by averaging numerical equivalents of self-reported letter grades in English, Mathematics, History, and Science. Outcome variable (wave 3). The use of self-report smoking data among adolescents has been previously validated [23]. Subjects reporting never having tried cigarette smoking, not even one or
A.E. Hussaini et al. / Journal of Adolescent Health 49 (2011) 511–517
two puffs, were classified as never smokers. Those reporting that they had tried cigarette smoking were classified as ever smokers. Among the ever smokers, those who reported ever having smoked regularly (at least one cigarette per day for at least 30 days) and smoking currently (having smoked a cigarette in the last 30 days) were eligible to complete the Fagerstrom Test for Nicotine Dependence (FTND) to determine level of nicotine addiction using previously determined scoring cutoffs [24]. Those scoring 0 –2 were categorized as not addicted; 3– 4, low addiction; 5, moderate addiction; and ⱖ6, high addiction. Never smokers and ever smokers (those not regularly and currently smoking) were not eligible to complete the FTND and were simply assigned to the 0 –2 category, also referred to as the “not addicted” nicotine category. The FTND is a modified version of the Fagerstrom Tolerance Questionnaire which has been validated for use in adolescents [25]. Statistical analysis All analyses were conducted using Stata 11E and surveycorrected statistical procedures to correct for the clustered and stratified nature of the Add Health sample [26]. To impute the data required for the modeling analysis among subjects missing data, the missing data analysis procedure ICE, a multivariate imputation procedure by chained equations available in Stata [13], was used. Five copies of the data, each with missing values imputed, were independently assessed in the multivariate multinomial logistic regression analyses. Estimates of parameters of interest were averaged across the five copies to give a single mean estimate, and standard errors were adjusted according to Rubin’s rules [27]. The imputation procedure included all of the variables in models 1 and 2, listed in Table 2. Selection of potential mediators was based on previously reported factors known to be associated with obesity and smoking among females and available in the Add Health data set. To complete the mediation analysis as described by Baron and Kenny, demonstration of three relationships was required [28]. In the first step, multinomial logistic regression was used to determine the association between wave 1 obesity and level of wave 3 nicotine addiction (Figure 2, relationship A). Covariates in this model included demographic variables of race, age, menarchal age, and SES as well as parent and friend smoking and wave 1 smoking. In the second step, linear regression models were used to assess the relationship between obesity and each potential mediator of obesity individually, including depressive symptoms, low self-esteem, family connectedness, school connectedness, GPA, trying to lose weight, and physical activity (Figure 2, relationship B). All aforementioned covariates were included in this second set of models. The third step used multiple logistic regression models to individually assess the relationship between each potential mediator and level of wave 3 nicotine addiction (Figure 2, relationship C). All aforementioned covariates were included in this third set of models. To avoid false negatives by prematurely excluding potential mediators, the inclusion criteria of potential mediators was widened (p ⬍ .10, two-tailed) in our testing of relationships B and C. Those potential mediators that were determined to have significant relationships with obesity or level of nicotine addiction were further evaluated using the coefficient product test. A significant test (p ⬍ .05, two-tailed test) is evidence for partial mediation. Mediators that were significant based on the coefficient product test were included individually in a final model
513
used to determine whether the relationship between obesity and level of nicotine addiction (represented by the  coefficient of obesity in the model) was reduced by the mediator being tested. The degree to which  is reduced when the potential mediators are added to the equation is indicative of the strength of the mediation. Confirming full mediation requires that the coefficient corresponding to the mediated effect is near zero when the mediator term is included in the model. Results Descriptive characteristics of sample Of the sample, 439 females (11%) had a BMI of ⱖ95th percentile for age and gender at wave 1. The decision to use the adult definition of obesity (BMI, ⱖ30) for 18- and 19-year-olds affected a single 19-year-old female in our sample whose BMI of 30.23 resulted in her being classified as obese based on adult criteria and nonobese based on the 95th percentile BMI for her age. In all, 57% of the sample was White, 18% Black, 16% Hispanic, 7% Asian, 2% Native American, and ⬍1% other. The average age at wave 1 was 15 (⫾1.6) years and the average age at wave 3 was 21 (⫾1.6) years. Table 1 shows weighted differences between obese and nonobese female adolescents in level of wave 3 nicotine addiction,
Table 1 Survey-weighted characteristics of sample participants who completed waves 1, 2, and 3 of Add Health by obesity status Characteristics
Self-reported nicotine addiction (wave 3) High addiction: % (N) Moderate addiction Low addiction Not addicted Characteristics (wave 1) Race: % (N) White Black Hispanic Asian Native American Other race Age: mean years (SE) Menarchal age: mean years (SE) Family SES: mean (SE) Resident parent smokes: % families (N) Friends smoke: Number of friends (SE) Cigarettes per month (W1): Number of cigarettes (SE) Mediators (wave 2) means from imputed data Depressive symptoms: mean (SE) Low self-esteem: mean (SE) Family connectedness: mean (SE) School connectedness: mean (SE) GPA: mean (SE)
Obese N ⫽ 439
Nonobese N ⫽ 3,663
8.43 (37) 2.28 (10) 6.38 (28) 82.92 (364)
3.90 (143) 3.17 (116) 5.68 (208) 87.25 (3,196)
50.11 (220) 23.46 (103) 18.22 (80) 3.64 (16) 3.87 (17) .68 (3) 14.98 (.08) 11.82 (1.26)
57.52 (2,107) 16.84 (617) 15.21 (557) 7.40 (271) 2.16 (79) .87 (32) 15.07 (.03) 12.15 (1.27)
5.46 (.12) 59 (259)
6.20 (.04) 50 (1,832)
p
.04* .40 .19 .91 ⬍.001**
.92 (.05)
.81 (.02)
.12 ⬍.001 ⬍.001 ⬍.001* .03
35.67 (119.55)
28.38 (105.78)
12.59 (.48)
11.06 (.19)
⬍.001
12.01 (.23) 22.20 (.17)
11.11 (.09) 22.32 (.06)
⬍.001 .03
17.23 (.36)
17.73 (.19)
⬍.001
2.68 (.05)
2.99 (.03)
⬍.001
Total of ⬍100% resulting from rounding. * p value for this addiction category versus all other others. ** RxC 2 test.
.26
514
A.E. Hussaini et al. / Journal of Adolescent Health 49 (2011) 511–517
Table 2 Survey-weighted multinomial logistic regression results (odds ratios and confidence intervals) predicting self-reported level of nicotine addiction (N ⫽ 4,102) Characteristics
Model 1 Low addiction
Independent variable Obesity Demographics and covariates Race (reference race: White) Black Hispanic Asian Native American Other race Higher age Higher menarchal age Higher family SES Pack (20 cigarettes) per month (W1)a Family smokes More friends smoke Mediators Higher depressive symptoms Lower self-esteem Higher family connectedness Higher school connectedness Higher GPA
Model 2 (mediators imputed) Moderate addiction
High addiction
Low addiction
Moderate addiction
High addiction
.95 (.56–1.59)
.80 (.35–1.81)
2.12 (1.22–3.68)
.87 (.52–1.46)
.73 (.32–1.65)
1.79 (.96–3.36)
1.00 .73 (.38–1.40) .37 (.18–.76) .55 (.19–1.56) 1.70 (.65–4.42) 3.91 (1.29–11.86) .77 (.69–.87) .93 (.81–1.08) .93 (.86–.99) 1.05 (1.02–1.08)
1.00 .45 (.21–.96) .19 (.06–.65) .22 (.03–1.39) 2.32 (.81–6.69) 3.28 (.39–27.36) .88 (.74–1.06) 1.01 (.83–1.22) .86 (.79–.94) 1.07 (1.04–1.09)
1.00 .29 (.13–.63) .25 (.12–.55) .65 (.20–2.10) 1.74 (.66–4.56) 2.38 (.45–12.49) .83 (.72–.95) .91 (.79–1.05) .92 (.84–.98) 1.10 (1.07–1.13)
1.00 .69 (.35–1.37) .30 (.14–.63) .51 (.18–1.46) 1.64 (.69–3.90) 4.08 (1.29–12.94) .77 (.68–.87) .94 (.81–1.10) .94 (.87–1.02) 1.05 (1.02–1.08)
1.00 .30 (.14–.66) .13 (.04–.44) .19 (.03–1.30) 2.06 (.71–5.92) 3.64 (.48–27.89) .84 (.69–1.02) 1.04 (.85–1.28) .93 (.84–1.02) 1.05 (1.02–1.08)
1.00 .22 (.09–.47) .15 (.06–.36) .50 (.14–1.85) 1.59 (.58–4.41) 2.76 (.55–13.82) .80 (.69–.91) .93 (.80–1.08) .99 (.89–1.09) 1.09 (1.05–1.12)
1.55 (1.11–2.14) 1.72 (1.45–2.03)
2.13 (1.26–3.60) 1.74 (1.41–2.15)
4.72 (2.89–7.72) 1.57 (1.27–1.94)
1.49 (1.07–2.08) 1.63 (1.38–1.92)
1.96 (1.14–3.34) 1.51 (1.21–1.89)
4.40 (2.68–7.21) 1.36 (1.09–1.70)
1.01 (.99–1.03) 1.06 (.99–1.13) 1.01 (.94–1.08) 1.01 (.97–1.04) .75 (.60–.94)
1.04 (1.01–1.08) .95 (⫺.88–1.03) .98 (.91–1.05) .97 (.93–1.00) .52 (.32–.83)
1.04 (1.01–1.07) 1.01 (.95–1.06) 1.01 (.95–1.08) .98 (.95–1.02) .48 (.32–.74)
Odds ratios are versus base outcome category of Fagerstrom score 0-2 (nonaddicted). Bolded values are significant at the p ⬍ .05 level. a Coefficients and confidence intervals reported as packs per month rather than as cigarettes per month.
demographic characteristics, parent and friend smoking, potential mediating variables, and wave 1 smoking. More than twice as many obese females experienced the highest level of nicotine addiction as nonobese females (8.43% vs. 3.90%, p ⫽ .04). Obese females were more likely to be Black, Hispanic, or Native American and were more likely to report younger menarchal age and to live in families with lower SES (p ⬍ .001). They were also more likely to live with parents who smoked, and more likely to have more friends who smoked. At wave 1, obese females reported a higher number of cigarettes smoked per month, although this difference was not statistically significant. With regard to potential mediators, obese females were significantly more likely to report depressive symptoms, lower self-esteem, lower school and family connectedness, and lower GPAs.
Model 1 results in Table 2 show that obesity more than doubled the risk of the highest level of nicotine addiction after controlling for all demographic variables, parent and friend smoking, and wave 1 cigarette smoking (OR, 2.12; 95% CI, 1.22– 3.68). Of all variables, parental smoking was the strongest predictor of the highest level of nicotine addiction (OR, 4.72; 95% CI, 2.89 –7.72). The second set of linear regression models demonstrated that wave 1 obesity was significantly associated with depressive symptoms, low self-esteem, and GPA. The third set of multiple logistic regression models showed that level of wave 3 nicotine addiction was significantly associated with depressive symptoms, low self-esteem, family connectedness, school connectedness, and GPA. Although all of these potential mediators were then evaluated using the coefficient product test, only GPA was found to be significant (p ⬍ .05, two-tailed). Model 2 results in Table 2 show that when GPA was included as a mediator, the relationship between obesity and high level nicotine addiction was reduced to statistical insignificance (p ⫽ .07), and higher GPA was strongly protective against high addiction (OR, .48; 95% CI, .32–.74). Discussion
Figure 2. Conceptual model of the mediated association between obesity and level of nicotine addiction in females.
Our findings suggest that obese, female adolescents are more than twice as likely to become highly nicotine-addicted smokers as their normal weight peers, even after accounting for demographic covariates, parent and friend smoking, and baseline smoking. However, in our sample, obese females were not significantly more likely to experience low or moderate levels of addiction. Further, the relationship between obesity and nicotine addiction is mediated by GPA. Previous research has explored the associations between obesity and low GPA, and separately the associations between low GPA and cigarette smoking. Ding reported that, on average, obese
A.E. Hussaini et al. / Journal of Adolescent Health 49 (2011) 511–517
515
Table 3 Obese females’ efforts to change weight by smoking statusa Response to Add Health Question ⬙What are you currently doing about your weight?⬙
Not addicted
Low addiction
Moderate addiction
High addiction
Total
Trying to lose weight: n (%) Trying to gain weight: n (%) Trying to stay the same weight: n (%) Not trying to do anything about weight: n (%) Totala
249 (68.4) 9 (2.5) 23 (6.3) 83 (22.8) 364 (100)
18 (64.3) 0 (0) 0 (0) 10 (35.7) 28 (100)
8 (80) 0 (0) 0 (0) 2 (20) 10 (100)
26 (70.3) 0 (0) 0 (0) 11 (29.7) 37 (100)
301 (68.6) 9 (2.1) 23 (5.2) 37 (100) 439 (100)
Pearson 2(9) ⫽ 9.2490, Pr ⫽ .415. a Total of ⬎100% resulting from rounding.
adolescents had GPAs that were .43 lower than their nonobese peers (p ⬍ .01) [29], and Falkner reported that obese females were 1.5 times more likely to be held back a grade. Perception of academic achievement is also lower among obese females, who are 2.1 times more likely to consider themselves as poor students as compared with their average-weight peers [30]. Interestingly, Huang found that females with higher BMIs underestimated their academic performance as compared with their thinner peers (p ⬍ .01), even when objective data did not support this belief [31]. Such perceptions may stem from the social and psychological impact of being an obese student. Obese adolescents are more likely to report being marginalized by peers, teased about weight issues, and bullied [9]. Even teachers report believing that overweight students are less academically capable [32]. Physical health issues may also affect school performance. This combination of factors may contribute to the finding that obese school children and adolescents are 4 times more likely than healthy peers to report impaired school function, and that obese students report increased rates of school absenteeism as compared with their nonobese peers [33]. Academic success reflects not only comprehension and mastery of material, but also consistent school attendance and school participation. We hypothesize that obese females may have less confidence academically and experience increased social marginalization and absenteeism, resulting in the lower GPA variable. Although the specific relationship between lower GPA and nicotine addiction has not been fully elucidated, research has demonstrated strong associations between lower academic achievement and cigarette smoking. Adolescents with a lower GPA are more likely to adopt smoking earlier and become regular smokers as compared with peers with higher GPAs [34]. This association may be explained in part by having less parental monitoring and a more relaxed parental attitude toward youth smoking [35]. Lower GPA and increased likelihood of smoking may also be associated with exposure to parental second-hand smoke which is implicated in children’s lower academic success and an increased likelihood of nicotine addiction during adolescence [36]. Students with lower GPAs may also spend more time with marginalized peers who smoke cigarettes [9,37]. Independently, the links between obesity and low GPA and low GPA and cigarette smoking are remarkable. A full understanding of the reasons why GPA predicts nicotine addiction and not just cigarette smoking in our sample will require further focused investigation. Obese adolescents may turn to smoking to make social connections with other marginalized peers or cope with the stress resulting from the academic and social consequences of obesity. The end result of this strategy may be nicotine addiction. Further understanding these associations and identification of potential intervention points may lead to effec-
tive interventions that decrease the likelihood of developing the dual health risks of obesity and nicotine addiction. In our sample, living with parents who smoked was the strongest predictor of nicotine addiction in young adults. This is consistent with previously published data that demonstrate family smoking as a major risk factor for smoking initiation among adolescents [38]. Although many researchers and clinicians assume that the strong effect of parental smoking on adolescent smoking behavior results from role-modeling smoking behavior or increased access to cigarettes, children and adolescents who are exposed to their parents’ (and peers’) second-hand tobacco smoke are also at risk for developing nicotine addiction [36]. Exploration of the link between parental and adolescent nicotine addiction is another important focal point for research and intervention. As illustrated in Table 3, our findings did not support the broadly accepted perception that obese teenagers smoke to lose weight. In previous studies, adolescents, especially girls, have reported beginning and continuing to smoke for weight control and weight loss. In fact, Strauss and Mir found a twofold increased risk of smoking among normal-weight adolescent girls who had been dieting. However, they found no increased risk of smoking among overweight adolescents (girls or boys) trying to lose weight [39]. This may be because overweight adolescents derive relatively less pleasure from nicotine as compared with food [40]. Thus, some obese adolescents may experiment with cigarettes but not progress to a regular habit because they find it less rewarding (e.g., fewer pleasurable effects, more unpleasurable effects), whereas other obese females, as suggested by our sample, may be at increased risk for nicotine addiction. There are several important limitations to these analyses. First, our sample reflects risks for those adolescents who completed all three waves of Add Health. The females who completed the three waves were more likely to be younger at wave 1, White, and have a higher SES. Older, minority, and lower SES females were not as likely to complete all three waves and be eligible for this longitudinal sample. The importance of GPA as a mediator may be different than in a more diverse sample. Second, our exploration of mediating covariates is limited by the available data in Add Health; inclusion of other covariates may deepen the understanding of this research question. Third, there is an uneven time frame between waves. Obesity and GPA are measured with a 1-year time lag in this longitudinal study and related to level of nicotine addiction 6 years after the exposure is measured. Given the 1-year time lag between waves 1 and 2, we explored the possibility that the potential mediators could be markers of the same constructs occurring even before the exposure variable. To test this hypothesis, we conducted supplemental analyses to examine potential reverse causality between wave 2 mediators
516
A.E. Hussaini et al. / Journal of Adolescent Health 49 (2011) 511–517
and the wave 1 obesity measure. Specifically, we ran regression models examining whether wave 2 mediators were significant predictors of obesity at wave 1. We found no significant associations. Although the short time lag between waves 1 and 2 is a significant limitation in Add Health, we believe that it is still one of the premier data sets for understanding longitudinal behavior change among U.S. adolescents. Further, we believe that the exploration of the relationship between obesity and nicotine addiction will be a helpful contribution to this critical health issue. Fourth, because all questions comprising the FTND were not included at wave 1, a true time lagged control for nicotine addiction was not possible. Although the difference in the average number of cigarettes smoked by obese and nonobese females at wave 1 was not statistically significant, number of cigarettes smoked at wave 1 was included in all modeling as a proxy control for the outcome variable. Fifth, family history of obesity was not controlled for. The only measure available was self-reported history of parental obesity from the in-home, but not necessarily the biological parent. This measurement was not thought to be accurate enough to be included in the analysis. Finally, given that Add Health data were collected over a period of 6 years, there were losses to follow-up from the baseline sample. However, the proportion of obese females and females who smoke was the same in wave 1 as our longitudinal study sample [33]. Therefore, the differential attrition that was noted in the demographic variables in the longitudinal sample did not result in disproportionate loss of obese females or female smokers. Acknowledgments At the time of manuscript preparation, Dr Aliya Hussaini was supported by the Robert Wood Johnson Foundation through the Robert Wood Johnson Clinical Scholars Program at the University of Pennsylvania. This research uses data from Add Health, a program project designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris, and funded by a grant P01-HD31921 from the National Institute of Child Health and Human Development, with cooperative funding from 17 other agencies. Special acknowledgment is due Ronald R. Rindfuss and Barbara Entwisle for assistance in the original design. Persons interested in obtaining data files from Add Health should contact Add Health, Carolina Population Center, 123 W. Franklin Street, Chapel Hill, NC 27516-2524 ([email protected]). The authors thank Janet Audrain-McGovern, Ph.D., Associate Professor of Psychology in the Department of Psychiatry and member of the Cancer Center at the University of Pennsylvania and Donald F. Schwarz, M.D., M.P.H., Deputy Mayor of Health and Opportunity and Health Commissioner for the City of Philadelphia for their thoughtful contributions and review of this manuscript. References [1] Healton CG, Vallone D, McCausland KL, et al. Smoking, obesity, and their co-occurrence in the United States: Cross sectional analysis. Br Med J 2006; 333:25– 6. [2] Freedman DM, Sigurdson AJ, Rajaraman P, et al. The mortality risk of smoking and obesity combined. Am J Prev Med 2006;31:355– 62. [3] Muennig P, Lubetkin E, Jia H, Franks P. Gender and the burden of disease attributable to obesity. Am J Public Health 2006;96:1662– 8. [4] Grundtvig M. Impact on nurse-based heart failure clinics on drug management and hospital admissions by self monitoring through a common database. Presented at: European Society of Cardiology Congress; September 3, 2008; Munich, Germany.
[5] Park E. Gender as a moderator in the association of body weight to smoking and mental health. Am J Public Health 2009;99:146 –51. [6] Ogden CL, Carroll MD, Curtin LR, et al. Prevalence of high body mass index in US children and adolescents 2007–2008. JAMA 2010;303:242–9. [7] Whitaker RC, Wright JA, Pepe MS, et al. Predicting obesity in young adulthood from childhood and parental obesity. N Engl J Med 1997;337: 869 –73. [8] Preventing tobacco use among young people: A report of the surgeon general. Atlanta, GA: Centers for Disease Control and Prevention. U.S. Department of Health and Human Services, Public Health Service, 1994. [9] Janssen I, Craig WM, Boyce WF, Pickett W. Associations between overweight and obesity with bullying behaviors in school-aged children. Pediatrics 2004;113:1187–94. [10] Tiggeman R, Rothblum ED. Gender differences in social consequences of perceived overweight in the United States and Australia. Sex Roles 1988; 18:75– 86. [11] Davison KK, Birch LL. Weight status, parent reaction, and self-concept in five-year-old girls. Pediatrics 2001;107:46 –53. [12] Resnick MD, Bearman PS, Blum RW, et al. Protecting adolescents from harm. Findings from the National Longitudinal Study on adolescent health. JAMA 1997;278:823–32. [13] Royston P. Multiple imputation of missing values: Update. Stata J 2005;5: 1–14. [14] Goodman E, Hinden BR, Khandelwal S. Accuracy of teen and parental reports of obesity and body mass index. Pediatrics 2000;106:52– 8. [15] Gordon-Larsen P, Adair LS, Nelson MC, Popkin BM. Five-year obesity incidence in the transition period between adolescence and adulthood: The National Longitudinal Study of adolescent health. Am J Clin Nutr 2004;80: 569 –75. [16] McDowell MA, Brody DJ, Hughes JP. Has age at menarche changed? Results from the National Health and Nutrition Examination Survey (NHANES) 1999 –2004. J Adolesc Health 2007;40:227–31. [17] Bearman PS, Moody J. Suicide and friendships among American adolescents. Am J Public Health 2004;94:89 –95. [18] Radloff LS. The use of the center for epidemiological studies depressions scale in adolescents and young adults. J Youth Adolesc 1991;20:149 – 66. [19] Warren JT, Harvey SM, Henderson JT. Do depression and low self-esteem follow abortion among adolescents? Evidence from a National Study. Perspect Sex Reprod Health 2010;42:230 –5. [20] Rosenberg M. Society and the adolescent self-image. Princeton, NJ: Princeton University Press, 1965. [21] Wainwright JL, Russell ST, Patterson CJ. Psychosocial adjustment, school outcomes, and romantic relationships of adolescents with same-sex parents. Child Dev 2004;75:1886 –98. [22] Pesa JA, Syre TR, Jones E. Psychosocial differences associated with body weight among female adolescents: The importance of body image. J Adolesc Health 2000;26:330 –7. [23] Patrick DL, Cheadle A, Thompson DC, et al. The validity of self-reported smoking: A review and meta-analysis. Am J Public Health 1994;84:1086 – 93. [24] Fagerstrom KO, Heatherton TF, Kozlowski LT. Nicotine addiction and its assessment. Ear Nose Throat J 1992;69:763–7. [25] Prokhorov AV, De Moor C, Pallonen UE, et al. Validation of the modified Fagerstrom tolerance questionnaire with salivary cotinine among adolescents. Addict Behav 2000;25:429 –33. [26] Chantalla K. Guidelines for analyzing add health data. Chapel Hill, NC: Carolina Population Center, University of North Carolina Chapel Hill, 2006. [27] Rubin DB. Inference and missing data (with discussion). Biometrika 1976; 63:581–92. [28] Baron RM, Kenny DA. The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. J Pers Soc Psychol 1986;51:1173– 82. [29] Ding W, Lehrer SF, Rosenquist JN, et al. The impact of poor health on education: New evidence using genetic markers, 2006 [NBER Working Paper No. W12304]. Available at: http://ssrn.com/abstract⫽910836. Accessed March 21, 2011. [30] Falkner NH, Neumark-Sztainer D, Story M, et al. Social, educational, and psychological correlates of weight status in adolescents. Obes Res 2001;9: 32– 42. [31] Huang TT, Goran MI, Spruijt-Metz D. Associations of adiposity with measured and self-reported academic performance in early adolescence. Obesity 2006;14:1839 – 45. [32] Puhl R, Brownell KD. Bias, discrimination and obesity. Obes Res 2001;9: 788 – 805. [33] Schwimmer J, Burwinkle T, Varni J. Health Related quality of life of severely obese children and adolescents. JAMA 2003;289:1813–9. [34] Audrain-McGovern J, Rodriguez D, Tercyak KP, et al. Identifying and characterizing adolescent smoking trajectories. Cancer Epidemiol Biomarkers Prev 2004;13:2023–34.
A.E. Hussaini et al. / Journal of Adolescent Health 49 (2011) 511–517
[35] Forrester K, Biglan A, Severson HH, Smolkowski K. Predictors of smoking onset over two years. Nicotine Tob Res 2007;9:1259 – 67. [36] Collins NB, Wileyto EP, Murphy MF, MunafÓ MR. Adolescent environmental tobacco smoke exposure predicts academic achievement test failure. J Adolesc Health 2007;41:363–70. [37] Jessor R. Risk behavior in adolescence: A psychosocial framework for understanding and action. In: Rogers D, Ginzburg E, eds. Adolescents at Risk: Medical and Social Perspectives. Boulder, CO: Westview Press, 1992:19 –34.
517
[38] Barman SK, Pulkkinen L, Kaprio J, Rose RJ. Inattentiveness, parental smoking and adolescent smoking initiation. Addiction 2004;99: 1049 – 61. [39] Strauss RS, Mir HM. Smoking and weight loss attempts in overweight and normal weight adolescents. Int J Obes 2001;25:1381–5. [40] Blendy JA, Strasser A, Walters CL, et al. Reduced nicotine reward in obesity: Cross comparison in human and mouse. Psychopharmacology 2005;180: 306 –15.