Does the association between depression and smoking vary by body mass index (BMI) category?

Abstract

Objective

The purpose of this study was to explore how weight might influence the relationship between depression and smoking.

Methods

Data were obtained from a cross-sectional survey representative of women age 40-65 enrolled in Group Health Cooperative, a health plan serving members in Washington and northern Idaho (n = 4,640). We examined the relationships between depression and smoking in normal weight, overweight, and obese women using weighted multiple logistic regression with both minimal and full adjustment.

Results

Current depression was significantly associated with current smoking in obese women (adjusted odds ratio = 2.48, 95% confidence interval = 1.26−4.88) but not in underweight/normal or overweight women. Among ever smokers, obese women, but not other groups, were significantly less likely to have quit smoking in the past.

Conclusions

The association between smoking and depression in middle-aged women appears to be limited to the obese subset and may stem from a lesser likelihood of obese ever smokers to have quit. This population represents an important target for preventive medicine efforts.

Introduction

There is a strong relationship between depression and tobacco use initiation (Rohde et al. 2003), progression to heavy smoking (Breslau, Novak, & Kessler 2004;Rohde et al. 2004a) and nicotine dependence (Breslau, Kilbey, & Andreski 1991;Breslau, Kilbey, & Andreski 1993;Breslau, Novak, & Kessler 2004;Rohde, Kahler, Lewinsohn, & Brown 2004a), though the dependency criteria may be driving this last association (Breslau & Johnson 2000) with the Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition criteria more weakly predicting cessation compared to Fagerstrom Test for Nicotine Dependence. The relationship between depression and tobacco use cessation, however, is less clear. While some research has found that depression is a barrier to cessation (Rohde et al. 2004b), other studies have concluded that there are similar rates of cessation in depressed and non-depressed smokers (John et al. 2004a;John et al. 2004b). Furthermore, depressive symptoms have been shown to be unrelated to readiness to quit (Prochaska et al. 2004) and one longitudinal study demonstrated that depressed smokers were no more likely than non-depressed smokers to persist in their smoking (Johnson & Breslau 2006).

Perhaps the inconsistent relations between smoking status and depression are due to differences in the characteristics of the people in the various studies. Depression is associated with both obesity and unhealthy weight-related behaviors such as physical inactivity and binge drinking (Fan et al. 2009;Strine et al. 2008). Smokers who have weight-related behavioral risk-factors (low physical activity, high dietary fat intake) appear to have greater nicotine dependence and are somewhat less likely to successfully quit smoking (Sherwood et al. 2000). Given this, the purpose of this study was to explore how weight might influence the relationship between depression and smoking in a sample of middle aged women. We hypothesized that the relationship between depression and smoking cessation may be different for those who are normal weight, overweight, and obese.

Methods

Data for this study were obtained during November 2003 - February 2005 through a population-based survey of women age 40-65 enrolled in Group Health Cooperative, a health plan serving approximately 500,000 members in Washington and northern Idaho. This sample was selected for the larger study that aims to explore depression - obesity relationships because of evidence that the association between obesity and depression is stronger in women (Istvan, Zavela, & Weidner 1992) and because both this age group and women has a relatively high risk of depression compared to other age groups (Kessler et al. 2003). The membership of Group Health Cooperative includes those enrolled through employer-purchased contracts as well as risk-sharing contracts with Medicare and Medicaid. The Group Health enrollment population is demographically representative of the Greater Seattle area's population (Simon et al. 1996). Study participants were recruited from eight clinics that served higher proportions of ethnic minorities. All women over the age of 40 in the plan are periodically invited to complete breast cancer risk questionnaires as part of the Group Health Breast Cancer Screening Program (BCSP) (Taplin et al. 1990). The survey includes self-report of height and weight from which BMI (kg/m2) was calculated. The primary goal of the survey was to assess the relationship between obesity and depression. In order to have an adequate number of women with high BMIs, women in the target age group were first stratified by BMI category to oversample women with high BMIs. All women with a BMI > 30 and 12% of those with a BMI < 30 were invited to participate. A sample of women who either did not complete BCSP or whose BMI could not be determined from the questionnaire was also recruited. The overall survey response rate was 61.5%.

Surveys were conducted by telephone and included items on height, weight, race/ethnicity (American Indian or Alaskan Native, Asian, Black or African American, Native Hawaiian or other Pacific Islander, White), age, current and past tobacco use, income, marital status (never married; widowed, divorced, or separated; married), education (less than college graduate, college graduate) moderate physical activity (number of times per week), vigorous physical activity (number of times per week) and current and past depression. Current depression was assessed using the Patient Health Questionnaire (PHQ) which uses the nine American Psychiatric Association DSM–IV (American Psychiatric Association 1994) criteria for the diagnosis of major depression, lifetime depression was assessed using a modified PHQ where questions are modified to ask about the past. Current major depression was categorized as a positive response (“More than half the days” or “Nearly all days”) to at least one of two core symptoms (depressed mood or loss of interest) and a total of five positive symptoms. Weight status was classified as: underweight = BMI < 18.5, normal = BMI 18.5 − 24.9, overweight = BMI 25 −29.1, obese = BMI > 30. For regression analyses, “underweight” and “normal” were combined due to small numbers of underweight women. Current and ever smoking were each assessed by single items (“Do you smoke cigarettes now?”, “Have you smoked at least 100 cigarettes in your entire life?”). Those who did not report current smoking but did indicate that they had smoked in the past were categorized as former smokers. All measures were self-reported.

We used both minimally and fully adjusted multiple logistic regression models to estimate associations and 95% confidence intervals. Analyses were conducted using SAS, version 9.13 (SAS Inc., Cary, NC). To account for stratified sampling procedures and differential response rates across strata, analyses incorporated sampling weights (Cochran 1977) using SAS procedures PROC SURVEYFREQ, PROC SURVEYMEANS, and PROC SURVEYLOGISTIC.

Results

Table 1 shows descriptive statistics for all independent and dependent variables by BMI category. No BMI category was significantly more likely to include either current smokers or quitters; prevalence of lifetime and current depression increased with BMI category. When we examined the associations between current major depression and current smoking, there was a nearly significant association in the minimally adjusted model (adjusted for age and race; OR = 1.70, 95% CI = 0.99 − 2.86) and a significant association in the fully adjusted model (adjusted for age, race, education, income, marital status, moderate physical activity and vigorous physical activity; OR = 2.13, 95% CI = 1.13 − 4.03). However, Table 2 shows that in both the minimally and fully adjusted models, when stratified by BMI category, current depression was associated with current smoking in obese women only. BMI category was not associated with current smoking in any models.

Table 1.

Description of the sample by weight category. November 2003 - February 2005, Group Health Cooperative, USA.

		underweight			normal weight			overweight			obese			overall
Categorical variables:
		n	weighted n	%	n	weighted n	%	n	weighted n	%	n	weighted n	%	n	p-value*
Ever smoker
	no	11	30.0	60.2	370	1046.0	58.3	388	637.7	52.7	1735	791.3	51.5	2504	0.053
	yes	7	19.8	39.8	265	746.9	41.7	337	573.2	47.3	1483	745.8	48.5	2092
Current smoker
	no	15	41.2	82.7	576	1635.0	91.0	656	1104.0	90.9	2911	1369.0	89.0	4158	0.37
	yes	3	8.6	17.3	60	160.9	9.0	70	110.1	9.1	308	168.6	11.0	441
Quitter (among ever smokers only)	no	3	8.6	43.6	59	158.3	21.2	68	105.0	18.3	302	166.5	22.3	432	0.28
	yes	4	11.2	56.4	206	588.6	78.8	269	468.3	81.7	1181	579.2	77.7	1660
Current depression
	no	19	52.9	100	630	1779.0	98.2	697	1151.0	93.7	3015	1420.0	91.9	4361	<.001
	yes	0	0.0	0	12	32.3	1.8	35	77.8	6.3	229	125.6	8.1	276
Lifetime depression
	no	19	52.9	100	620	1753.0	96.9	682	1125.0	92.2	2933	1376.0	89.0	4254	<.001
	yes	0	0.0	0.0	21	55.5	3.1	47	95.6	7.8	310	169.3	11.0	378
Race
	American Indian or Alaskan Native	0	0.0	0.0	7	21.3	1.2	23	19.4	1.6	162	73.2	4.7	192	<.001
	Asian	5	13.2	23.9	66	182.6	10.1	48	97.1	7.9	99	51.6	3.3	218
	Black or African American	1	3.0	5.5	25	64.7	3.6	53	87.8	7.2	351	166.9	10.8	430
	Native Hawaiian or Other Pacific Islander	1	3.0	5.5	5	14.2	0.8	7	7.3	0.6	20	15.7	1.0	33
	White	13	36.1	65.1	541	1532.0	84.4	600	1017.0	82.8	2612	1239.0	80.1	3766
Marital status
	never	5	14.7	27.9	51	144.6	8.0	67	109.6	8.9	421	205.4	13.3	544	<.001
	widowed, divorced, or separated	3	8.1	15.4	136	375.5	20.7	172	292.6	23.8	887	433.4	28.0	1198
	married	11	30.0	56.7	457	1295.0	71.3	493	829.1	67.3	1937	907.6	58.7	2898
College graduate
	no	6	17.3	31.2	153	410.5	22.8	256	409.5	33.3	1519	713.5	46.2	1934	<.001
	yes	14	38.1	68.8	487	1394.0	77.2	476	819.1	66.7	1723	831.8	53.8	2700
Continuous variables:
		n	Mean	SE	n	Mean	SE	n	Mean	SE	n	Mean	SE	Mean	p-value^
Age		20	50.4	1.6	644	51.6	0.3	733	53.0	0.3	3245	52.0	0.2	52.1	0.15
Monthly income		10	3028.5	309.4	492	3737.7	110.8	531	3780.8	145.3	2444	3409.3	74.8	3631.3	0.024
Times/week moderate physical activity		18	6.3	1.1	639	6.6	0.1	729	5.5	0.2	3234	4.5	0.1	5.6	<.001
Times/week vigorous physical activity		18	3.8	0.9	640	5.0	0.2	728	3.9	0.2	3235	2.6	0.1	3.9	<.001

^*Rao-Scott Chi square test for differences between BMI categories

^{^}Weighted t-test for differences between BMI categories

Note: Due to small cell counts in the underweight category, underweight and normal weight categories were combined for statistical tests. Totals for each category range from 4,596 to 4,640 due to missing data on certain survey items.

Table 2.

Associations between current depression and current smoking and quitting smoking by weight status. November 2003 - February 2005, Group Health Cooperative, USA.

	Model 1*							Model 2^
	n	n weighted	Prevalence of current smoking or quitting smoking in depressed (weighted %)	Prevalence of current smoking or quitting smoking in non-depressed (weighted %)	OR	95% CI		n	n weighted	OR	95% CI
Associations between current depression and current smoking:
underweight/norma
l	650	1835.9	1.30	3.82	1.07	0.12	9.43	449	1415.2	1.52	0.15	15.63
overweight	724	1213.3	2.38	2.40	0.91	0.20	4.10	527	894.7	1.78	0.41	7.69
obese	3216	1536.4	11.26	3.26	2.46a	1.39	4.37	2439	1158.2	2.48b	1.26	4.88
Ever smokers: associations between current depression and having quit smoking in the past:
underweight/norma
l	272	761.6	2.57	8.33	0.75	0.07	7.81	212	601.8	0.7	0.07	6.67
overweight	337	573.2	4.71	5.05	1.17	0.25	5.49	241	421.8	0.75	0.17	3.23
obese	1482	745.4	21.99	7.14	0.40c	0.21	0.76	1114	558.5	0.43d	0.20	0.90

Note: The Breslow-Day test of homogeneity of the depression and current smoking relationship across weight category strata had p = 0.09. The Breslow-Day test of homogeneity of depression and quitting smoking relationship across weight category strata had p = 0.11.

^*Model 1 is adjusted for age and race

^{^}Model 2 is adjusted for age, race, education, income, marital status, moderate physical activity and vigorous physical activity

In this sample of middle-aged women (n=4,640), smoking and depression were associated only in obese women. It may stem from a lifetime lesser likelihood of obese smokers to have quit.

Among ever smokers, having quit smoking in the past was not significantly associated with either past depression or current BMI category in either minimally adjusted or fully adjusted models (results not shown). However, when examining quitting by BMI status, obese women, but not other groups, were less likely to have quit smoking in the past if they reported current depression (Table 2).

Discussion

To our knowledge, this study is the first to examine the association between depression and smoking by BMI category. In a large sample of middle-aged women, we found current depression to be associated with current smoking only among the subset of women who were obese. Among ever smokers, obese women were much less likely have quit.

The prevalence of smoking in a population depends on 1) the probability of individuals to initiate the behavior and 2) the probability of the behavior persisting. Our results suggest that smoking cessation may be especially challenging for female smokers who are both depressed and obese. It is notable that depression did not appear to be a substantial barrier to cessation for women who are not obese.

This study has several limitations. First we are unable to assess whether our results may have been due to selection bias because data on the smoking habits or depression of non-responders were not collected. Though we have no reason to believe this was the case, it is possible that obese smokers suffering from depression may have been more likely to participate compared to obese smokers not suffering from depression. A second limitation is that height and weight were based on self-report and women tend to under-report weight (Pirie et al. 1981). However, a previously published analysis of these data found that the under-reporting of weight in both depressed and non-depressed obese women was minimal and similar in magnitude to that seen in normal weight women (Jeffery et al. 2008). A third limitation is the use of self-reported smoking data. Past research where self reported smoking was validated with biochemical markers of nicotine uptake has shown self-reported smoking to be reasonably valid, but it may underestimate actual smoking (Gorber et al. 2009). A fourth limitation is the relatively small number of non-obese women, as evidenced by the wide confidence intervals around these estimates in Table 2. For this reason, these estimates should be viewed as imprecise and future research is needed to better categorize the associations between smoking, depression, and obesity in these weight groups.

Though we cannot know which way causality might flow from this exploratory analysis of cross-sectional data, it appears that the combination of both depression and obesity is associated with a lower likelihood that a woman who has been a smoker will be able to quit. Future research should examine depression and tobacco use by weight category longitudinally. Additionally, research that includes measures of smoking intensity and menopausal status could provide greater insight. Tobacco cessation may be especially difficult for those who are both depressed and obese due to compounding burdens and/or a greater nicotine dependency in this population. Since health risk behaviors do tend to cluster (Klesges et al. 1990;Shah et al. 1993) and because tobacco use and obesity are leading risk factors for chronic disease, this population represents an important target for preventive medicine efforts.