Journal of Behavioral Medicine

, Volume 37, Issue 4, pp 577–586

Associations between depressive symptoms, self-efficacy, eating styles, exercise and body mass index in women

Authors

    • Department of Global Community Health and Behavioral SciencesTulane University School of Public Health and Tropical Medicine
  • Janet C. Rice
    • Department of Biostatistics and BioinformaticsTulane University School of Public Health and Tropical Medicine
  • Marsha Broussard
    • Louisiana Public Health Institute
  • Carolyn C. Johnson
    • Department of Global Community Health and Behavioral SciencesTulane University School of Public Health and Tropical Medicine
  • Larry S. Webber
    • Department of Biostatistics and BioinformaticsTulane University School of Public Health and Tropical Medicine
Article

DOI: 10.1007/s10865-013-9526-5

Cite this article as:
Clum, G.A., Rice, J.C., Broussard, M. et al. J Behav Med (2014) 37: 577. doi:10.1007/s10865-013-9526-5

Abstract

This article explores cross-sectional associations between depressive symptoms and body mass index (BMI) in women working in schools in the Greater New Orleans area. Self-efficacy for eating and exercise, eating styles, and exercise are examined as potential pathways. This is a secondary data analysis of 743 women who were participating in a workplace wellness randomized controlled trial to address environmental factors influencing eating and exercise behaviors using baseline data prior to the intervention. BMI was the primary outcome examined. Path analysis suggested that increased depressive symptoms were associated with increased BMI in women. Indirect effects of depressive symptoms on BMI were found for increased healthy eating self-efficacy, increased emotional eating, and decreased exercise self-efficacy. The association between greater healthy eating self efficacy and BMI was unexpected, and may indicate a suppressor effect of eating self-efficacy in the relationship between depressive symptoms and BMI in women. The findings suggest the importance of depressive symptoms to BMI in women. Targets for interventions to reduce BMI include targeting depressive symptoms and related sequelae including self-efficacy for exercise, and emotional eating. Further investigation of eating self-efficacy and BMI are recommended with particular attention to both efficacy for health eating and avoidance of unhealthy foods.

Keywords

Depressive symptomsBMISelf-efficacyEmotional eating

The health and economic burden associated with both depression and obesity is an important public health issue (Simon et al., 2011). The National Comorbidity Survey suggests that lifetime episodes of major depressive disorder are experienced by 16.2 % of adults, with women at increased risk for lifetime major depressive disorder relative to men (Kessler et al., 2003). The prevalence of obesity among women in the United States is 35.8 %, a figure that has remained relatively stable in the past decade (Flegal et al., 2012). Both depression and obesity are associated with functional impairment and poor long term health outcomes, as well as high health care costs. Meta-analyses examining findings from longitudinal studies suggest that there is a significant association between depression and obesity (Scott et al., 2008), with depression leading to a 58 % increased risk for obesity (Blaine, 2008; Luppino et al., 2010). The impact of depression on later obesity is particularly strong for women (Blaine, 2008; de Wit et al., 2010). Both obesity and depression are associated with higher health care costs across inpatient and outpatient services (Simon et al., 2011). These data point to the importance of understanding pathways between depression and overweight /obesity, as intervening in these pathways can have a significant impact on both health outcomes.

Behavioral factors such as eating and exercise exert strong influence on body mass index (BMI), one of the most common measures of overweight/obesity. Psychological factors have also been hypothesized as important influences on eating and exercise behavior and hence weight outcomes. Symptoms associated with depression can have direct links to these behaviors via the symptoms themselves. For example, symptoms of depression include disturbances in appetite, and may result in both increased and decreased appetite. Depression has been associated with increased intake of energy-dense sweet foods (Jeffery et al., 2009) and fast food (Crawford et al., 2011), as well as decreased consumption of fruits and vegetables and non-sweet food consumption (Agurs-Collins & Fuemmeler, 2011; Jeffery et al., 2009; Konttinen et al., 2010). The increased consumption of unhealthy, energy dense foods has been hypothesized to be an outcome of emotional eating associated with depression (Konttinen et al., 2010). Emotional eating is eating in response to negative emotion. It is possible that greater intake of energy dense, sweet foods may be the result of attempts to regulate negative affect associated with depression (Jeffery et al., 2009). In support of this, emotional eating has been shown to mediate between depressive symptoms and body mass index, waist circumference, and percent body fat (Konttinen et al., 2010), indicating the importance of emotional eating in pathways between depression and overweight/obesity.

In addition to emotional eating, some research has focused on the concept of dietary restraint as an important psychological factor in overweight/obesity. Cognitive restraint is deliberately curtailing food consumption as a weight control strategy. Research suggests that cognitive restraint has complex associations with eating behavior depending on a broad range of factors such as measurement of the construct of dietary restraint and whether one is overweight or of normal weight (Johnson et al., 2012). Self regulation research has shown that impulsivity, in particular, negative urgency or the likelihood of acting rashly in distress, is associated with a host of health risk behaviors including overeating, binging, and alcohol use (Anestis et al., 2007, 2009; Churchill & Jessop, 2011; Fischer et al., 2004). Cognitive restraint has been shown to be associated with successful weight loss and maintenance, and depressive symptoms may impact cognitive restraint (Teixeira et al., 2010; Tice et al., 2001), suggesting its relevance for investigation in examinations of depression and overweight/obesity.

Physical activity is also an important component of weight management and overweight/obesity, and physical activity is diminished in depressed persons. Depression is associated with increased sedentary behavior and decreased physical activity, particularly in women and those above age 40 (Martinsen, 2008). Conversely, a burgeoning literature has demonstrated that exercise is an efficacious intervention for reducing and preventing depressive symptoms (Krogh et al., 2011; Martinsen, 2008). Failure to engage in regular exercise may be an important influence on overweight and obesity identified in depressed persons (Markowitz et al., 2008).

Self-efficacy for physical activity and eating behavior has been supported as an important link between depression and overweight/obesity (Teixeira et al., 2010), particularly in women. Self-efficacy, or the belief that one can attain a desired outcome, is thought to be associated with engagement in behavior to attain that outcome (Bandura, 1997). Self-efficacy is diminished in depressed individuals across a range of behaviors, and may be influencing the reduced engagement in exercise and poor eating behavior (Markowitz et al., 2008; Teixeira et al., 2002). In women, depression has been associated with lower weight control self-efficacy and higher body weight and less likelihood of successful weight loss after intervention (Linde et al., 2004, 2011). Physical exercise self-efficacy has been deemed a critical component in successful weight loss and weight loss maintenance (Teixeira et al., 2010). The influence of depression on physical exercise self-efficacy could be an important link to overweight and obesity in depressed individuals. Research suggests that depression is also associated with decreased physical activity self-efficacy (Clark et al., 1991; Craft et al., 2008). In support of this, a recent investigation of the role of self-efficacy for physical exercise on adiposity indicators was conducted with a large, population-based sample of Finnish men and women. The results supported a mediational role for physical activity self-efficacy in the association between depressive symptoms and adiposity (Konttinen et al., 2010).

Given the important public health implications of both depression and obesity, as well as the association between depressive symptoms and obesity, identification of associations between the two can help to inform intervention development. To date, few studies have targeted depressive symptoms in weight loss interventions, despite the evidence that depression is associated with key components of successful weight loss and maintenance, specifically self-efficacy for physical exercise and eating behavior. Both emotional eating and physical exercise self-efficacy have been identified as pathways between depressive symptoms and indicators of overweight and obesity (Konttinen et al., 2010). In this study, we propose to expand this model and test it in women living in South Louisiana, a population and region at great risk for obesity (CDC, 2009). We conducted a secondary data analysis using cross-sectional baseline data from a completed randomized controlled trial (RCT) of an environmental intervention to reduce overweight and obesity in faculty and staff in a school setting. We hypothesized that depressive symptoms would be associated with body mass index through two pathways. First, depressive symptoms will be negatively associated with self-efficacy for healthy eating behavior, cognitive restraint, and positively associated with increased emotional eating, and BMI. Additionally, we proposed that depressive symptoms would be positively associated with reduced physical activity self-efficacy, reduced engagement in physical exercise, and increased body mass index.

Method

Participants

Secondary analysis of baseline data from an RCT was conducted. Twenty-two schools in the Greater New Orleans area participated in the RCT called ACTION, which was a wellness program designed to address eating and physical activity barriers among adult school personnel (Webber et al., 2007, 2012). The primary focus of the intervention was to address school environmental factors influencing eating and physical activity behaviors. Examples of the intervention included increasing opportunities for exercise, increasing access to healthy snack foods, and a wellness area located within the school that had opportunities for exercise and access to educational materials. Additional educational content varied across the two and a half year intervention and included goal setting, pedometer use, and addressing healthy eating strategies. A process evaluation of the RCT and more detail on the intervention content is available for review (Johnson et al., 2010; Webber et al., 2007). All analyses here in this study were conducted with 743 women who provided baseline data collected in fall of 2006, prior to the start of the intervention, and hence data were cross-sectional and not affected by the intervention. All procedures and materials were approved by the Tulane IRB board.

Measures

Body mass index (kg/m2) was calculated at baseline with the formula weight (kg)/height (m)2. Weight was measured with a calibrated metric digital scale, and height was measured with a portable stadiometer. Height and weight were measured to the nearest 0.1 and 0.5 kg, and taken in duplicate.

The Three Factor Eating Questionnaire (TFEQ-R21(Tholin et al., 2005) assessed three components of eating behavior, uncontrolled eating, emotional eating, and cognitive restraint. The TFEQ-R21 is adapted from the TFEQ-R18 (Anglé et al., 2009; Karlsson et al., 2000; Stunkard & Messick, 1985). The TFEQ-18 is widely utilized and the 18 item version has demonstrated good internal reliability with scores above 0.70. Cognitive restraint alpha for adults was reported to be 0.84 and emotional eating 0.87 (de Lauzon-Guillain et al., 2004, 2009). Scores are anchored from 1 to 4 and summed for each subscale. In this study, we restricted our analyses to emotional eating and cognitive restraint due to the theoretical relevance of these factors to depression between uncontrolled and emotional eating. In this study, alpha for these scales was 0.94 for emotional eating, and 0.77 for cognitive restraint.

Depressive symptoms were assessed with the Center for Epidemiologic Studies Depression Scale. This is a 20 item scale assessing depressive symptom frequency in the past week. Scores are anchored on a 4 point scale ranging from 0 (rarely or none of the time) to 3 (most or all of the time). Total scores range from 0 to 60, reliability and validity of the scale are excellent. The CES-D has been used in numerous populations and has good internal consistency, high test–retest stability, and strong concurrent validity as compared to clinical and self report criteria (Bromberger et al., 2007; Radloff, 1977). Scores of 16 or higher are used to indicate the clinical relevance of depressive symptoms. In this study, items were summed to create a continuous score of depressive symptoms. Alpha for this study was 0.90.

Self-efficacy for healthy eating behaviors consisted of items assessing participant’s confidence in eating healthier foods and maintaining a weight loss program. Items were utilized in prior studies conducted by study authors such as CATCH-ON (Johnson et al., 2003) and guided by instruments with sound psychometric properties (e.g. (Sallis et al., 1988). A panel of experts reviewed items used for this study. A Likert scale ranging from “extremely confident (1)” to “extremely not confident (5)” asked participants to rate their confidence, for example, to “eat at least a total of 5 servings of fruits and vegetables every day”, “select packaged food items based on reading the nutrition label”, “eat small portion sizes”, “substitute seafood or poultry for red meat.” Items were reverse scored and summed so that higher scores indicated greater self-efficacy for the behavior. Alpha for this study was 0.78.

Self-efficacy for physical activity behaviors was assessed with five items that asked about a participant’s confidence in engaging in regular physical activity. Items were similarly drawn from prior studies and reviewed by a panel of experts in nutrition and exercise. Examples included confidence in being “physically active at least 5 times a week”, “planning a physical activity program that you like and sticking with it”, and “exercising even when you don’t feel like it.” Items were rated on a 5 point scale from 1 (extremely confident) to 5 (extremely not confident). Items were reverse scored and summed so that higher scores indicated greater physical activity self-efficacy. Alpha for this study was 0.90.

Moderate to vigorous physical activity (MVPA) was measured with an Actigraph Accelerometer. Data were stored in 30 s counts/minute. Participants wore the accelerometer for a six day period. Sedentary is <1,951 counts per minute, light as 1,952–5,724 counts per minute, and moderate to vigorous is 5,725 or more counts per minute (Freedson et al., 1998). MVPA was the variable of interest in this study; therefore, the two categories moderate and vigorous activity were collapsed and analyzed together. Physical activity data was log transformed.

Statistical analyses

Pearson Product-Moment Correlation Coefficients were calculated to examine bivariate associations. SAS (SAS Institute, Cary, NC, USA) version 9.1 was used to assess pathways between depressive symptoms and BMI in our sample of female school personnel. We included age and race as exogenous variables in the model.

We tested the overall model fit using a variety of indices, including cut off values close to 0.95 for comparative fit index (CFI) and normal fit index (NFI), and cutoff value close to .06 for the root mean square error of approximation (RMSEA) as recommended by Hu & Bentler, (1999). We tested indirect effects using Sobel’s test of significance (MacKinnon et al., 2002; Sobel, 1982). Participants with missing data were excluded from analyses.

Results

Descriptive statistics

Details regarding baseline variables of school and study participants are available in Webber and colleagues (Webber et al., 2007, 2012). Of the 852 participants in the baseline assessment, a small number of men, pregnant and lactating women, and those without viable accelerometer data were removed from the dataset, leaving a total of 743 women. Overall, less than 5 % of data was missing in the sample, and there were no differences on study variables for those missing data versus those not missing data. The average age of participants was 46.7 years (SD = 10.8) with ages ranging from 19 to 72. Descriptive data for the study sample are shown in Table 1. Of the 743 women, 563 were White, 135 were African American, and 96 identified as other. We collapsed African Americans and other race/ethnicity into one group comprising 24.6 % of the sample with Whites comprising the remaining 76.4 %. The average level of past week depressive symptoms reported on the CES-D was 13.2. Sixty-five percent of women had a score <16, thus 35 % had a score of 16 or higher on the CES-D, which has been used to indicate clinically meaningful symptoms of depression. The mean BMI for women was 30.08 (SD = 7.9) and ranged from 15.8 to 74.8. In our sample, 29.2 % of women were normal weight (BMI < 25), 28.8 % were overweight (BMI 25–29.9), and 43.0 % were obese (BMI ≥ 30). Relative to adults in New Orleans, rates in this female only sample are slightly lower for overweight (35.6 %) and higher for obesity (28.8 %) (CDC, 2009).
Table 1

Description of demographics and study measures at baseline (N = 743)

Demographics

Percent

M

SD

Range

Age

 

46.69

10.75

(19–72)

Race-White

76.4

   

Race—African American/Other

24.6

   

Measures

 CES-D Depression

 

13.25

9.3

(0–57.89)

 CES-D <16

65.0

   

 CES-D ≥16

35.0

   

Eating self efficacy

 

3.87

0.64

(1–5)

Exercise self efficacy

 

3.31

0.94

(1–5)

Emotional eating

 

40.68

29.63

(0–100)

Cognitive restraint

 

48.94

19.99

(0–100)

Moderate/vigorous activity (log)

 

0.12

0.74

(−0.69 to 3.55)

BMI

 

30.08

7.93

(15.76–74.81)

 Normal

29.2

   

 Overweight

28.8

   

 Obese

43.0

   
Bivariate associations between study variables can be viewed in Table 2. BMI was significantly and positively associated with African American race, depressive symptoms, and emotional eating. BMI was significantly and negatively associated with exercise self-efficacy, cognitive restraint, and physical activity. Increased depressive symptoms were associated with lower eating self-efficacy, lower exercise self-efficacy, lower cognitive restraint, and greater reported emotional eating style at the bivariate level.
Table 2

Correlation coefficients among study variables (N = 743)

Variables

1

2

3

4

5

6

7

8

9

1. Age

        

2. Race

−.05

       

3. CES-D depression

−.06

.02

      

4. Eating self efficacy

.11**

−.07*

−.17***

     

5. Exercise self efficacy

.05

.12**

−.11**

.39***

    

6. Emotional eating

.03

−.18***

.26***

−.03

−.18***

   

7. Cognitive restraint

.02

−.05

−.10**

.35***

.18***

−.04

  

8.Moderate/vigorous activity (Log)

−.20***

−.03

−.033

.06

.15***

−.01

.17***

 

9. BMI

.03

.16***

.14***

.00

−.15***

.30***

−.13**

−.16***

p < .05; ** p < .01; *** p < .001

Testing the model

Our hypothesized model is depicted in Fig. 1. Results suggested that the fit of the model was adequate according to recommended cut off points at, CFI = 0.94, NFI = 0.94, RMSEA = 0.09. The Chi Square test was 37.38 with 5 degrees of freedom, significant at <.001. This Chi Square test is considered extremely sensitive to sample size and is likely to be significant at N > 400 as is the case here, and therefore is not typically suggested as an appropriate fit indice. The model explained 18 % of the variance in BMI. See Table 3 for direct effects of study variables and standard errors. Table 4 displays the direct and indirect effects of depressive symptoms on BMI through study variables. The path coefficient for the total effect of depressive symptoms on BMI was significant at 0.121 (p < 0.001), however, with all variables entered in the model, the direct effect of depressive symptoms on BMI was reduced to 0.047 (p > 0.10), therefore, we tested indirect relationships between depressive symptoms and BMI. The path coefficient between depressive symptoms and eating self-efficacy was significant, b = −0.011, p < 0.001 and in the expected direction, depressive symptoms were associated with decreased eating self-efficacy. The association between eating self-efficacy and BMI was also significant, b = 1.595. p < 0.001, however, it was not in the expected direction, greater eating self-efficacy was associated with greater BMI. The indirect pathway between depressive symptoms and BMI through eating self-efficacy was also significant (b = −0.018, p < .01).
https://static-content.springer.com/image/art%3A10.1007%2Fs10865-013-9526-5/MediaObjects/10865_2013_9526_Fig1_HTML.gif
Fig. 1

Structural equation model examining pathways between depressive symptoms and BMI

Table 3

Direct effects and standard errors of study variables

Study variable

Dependent variable

Direct effect (Standard Error)

Depressive symptoms

Eating self-efficacy

−.011 (.003)***

Exercise self-efficacy

−.011 (.004)**

Emotional eating

.831 (.112)***

Cognitive restraint

−.091 (.074)

Moderate/vigorous physical activity

−.002 (.003)

Eating self-efficacy

BMI

1.595 (.476)***

Exercise self-efficacy

BMI

−1.119 (.311)***

Emotional eating

BMI

.080 (.009)***

Cognitive restraint

BMI

−.039 (.014)**

Moderate/vigorous physical activity

BMI

−1.304 (.372)***

p < .05; ** p < .01; *** p < .001

Table 4

Direct and indirect effects: Associations among CES-D and BMI

Dependent variables

Predictor CES-D symptoms

Eating self efficacy

 Direct

−0.011**

 Indirect

−0.018**

Exercise self efficacy

 Direct

−0.011**

 Indirect

.012*

Cognitive restraint

 Direct

−0.091

 Indirect

0.004

Emotional eating

 Direct

0.831**

 Indirect

0.067**

Mod/vigorous physical activity

 Direct

−0.002

 Indirect

0.003

BMI

 Direct

0.047

 Indirect

p< .05; ** p< .01; *** p < .001

We also explored the indirect pathway from depressive symptoms to emotional eating, cognitive restraint and BMI. Depressive symptoms were positively associated with increased emotional eating, b = .831 (p < 0.001), and emotional eating was positively associated with greater BMI, b = 0.080, (p < 0.001). The indirect effect of depressive symptoms on BMI through emotional eating was significant at b = 0.067 (p < 0.001). Depressive symptoms were not significantly associated with cognitive restraint, b = −0.091, however, greater cognitive restraint was associated with decreased BMI, b = −0.039 (p < 0.01). The lack of association between depressive symptoms and cognitive restraint indicated that cognitive restraint did not show an indirect effect between depressive symptoms and BMI.

Depressive symptoms also had a significant direct effect on exercise self-efficacy, such that greater depressive symptoms were associated with lower exercise self-efficacy, b = −0.011, (p < 0.01). Lower exercise self-efficacy was associated with increased BMI, b = −1.119 (p < 0.001), and the indirect effect of depressive symptoms on BMI through exercise self-efficacy was significant at b = 0.012, (p < 0.05). We also explored the association of the pathway between depressive symptoms, MVPA, and BMI. There was a non-significant association between depressive symptoms and MVPA, b = −0.002. Greater MVPA was associated with decreased BMI, b = −1.304 (p < 0.001), but because depressive symptoms were not associated with MVPA, there was no evidence of an indirect effect.

Because there is some evidence to suggest racial differences in BMI, we examined interactions between race and study variables and tested our model stratifying by race. There were no statistically significant interactions between race and study variables. As a follow up, we stratified the sample by race and ran the full model exclusively on the African American and those who identified as other participants. Model fit was good for the subset of African-Americans and others. Given these results, we combined Caucasians, African Americans, and those who identified as other in the model.

Discussion

This study builds on the body of literature demonstrating associations between depressive symptoms and BMI, extending it to a sample of employed, largely middle aged women from south Louisiana, a group at high risk for obesity (CDC, 2009). Further, we examined theoretically supported factors through which depressive symptoms might be associated with BMI in women, including psychological components such as self-efficacy and eating styles, and objective measurement of exercise. Women in this study were obese with BMI scores above 30; approximately 71 % of the sample was either overweight or obese. This finding is consistent with research that shows high rates of obesity in the United States and particularly in the South (CDC, 2009). Using a traditional cutoff score of 16 on the CES-D, 35 % of women in this study reported clinically significant depressive symptoms, which is elevated relative to other samples of women. In mid-life women, 26 % in the Seattle Mid-life Women’s Health cohort (Woods & Mitchell, 1997), and 23 % in the Study of Women’s Health Across the Nation (SWAN) had clinically significant symptoms (Bromberger et al., 2007). Baseline data in this study were collected in the Fall of 2006, approximately one year after Katrina. It is possible that higher rates of depressive symptoms may be accounted for by exposure to that event (Wadsworth et al., 2009). The high rates of depression and overweight/obesity in our study and the significant association between depression and overweight/obesity suggests that women in this geographical area and age range may be an important target for intervention. Longitudinal studies suggest that early life depression may be a risk factor for obesity in adulthood, particularly for women (Blaine, 2008). Our study highlights psychological factors that may be targeted through interventions to reduce overweight/obesity in those with depressive symptoms.

In general, our study supports the concept that depressive symptoms are associated with decreased self-efficacy for behaviors relevant to overweight and obesity. At the bivariate level, depressive symptoms were associated with decreased self efficacy for exercise and healthy eating. This is consistent with literature demonstrating lower self-efficacy in individuals with depression (Kanfer & Zeiss, 1983; Lewinsohn, 1974) and associations with increased adiposity in women and men (Konttinen et al., 2010). Unexpectedly, greater healthy eating self-efficacy was associated with greater BMI, and demonstrated a significant indirect effect from depressive symptoms to BMI. We explored several potential explanations for this unexpected association, and statistically, the results remained the same. It is possible that high correlations among study variables such as eating and physical exercise self-efficacy may have contributed to this finding. Given the lack of association between eating self-efficacy and BMI at the bivariate level, and the unexpected and positive association in the model, eating self-efficacy may be acting as a suppressor variable. There is also the possibility of measurement bias because our measure assessed efficacy for performing positive strategies for healthy eating and not efficacy in avoiding unhealthy foods. Avoidance of unhealthy foods may be more germane to BMI than intake of healthy foods. A recent study that included Southern Louisiana and Los Angeles County in its sample estimated caloric intake in the last 24 h (Cohen et al., 2010). Overconsumption of discretionary calories, which include foods high in sugar and fat and low in essential nutrients, was in evidence, exceeding guidelines by more than 120 % in Southern Louisiana, while consumption of fruits and vegetables fell 20 % short of guidelines in Southern Louisiana. The authors concluded that decreasing discretionary caloric intake such as salty snacks, cookies, candy and sugar sweetened beverages may be a more relevant target than increase of fruits and vegetables (Cohen et al., 2010). Likewise, self-efficacy for these different consumption patterns may be important to assess. Thus, future assessments of eating self-efficacy should include avoidance of discretionary foods and enactment of healthy eating strategies.

As noted, self-efficacy for exercise was lower in women who reported more depressive symptoms and associated with greater BMI. The indirect effect was significant, indicating that the association between depressive symptoms and BMI was partially accounted for by decreased exercise self-efficacy. Longitudinal studies demonstrate declines in physical activity associated with depressive symptoms (Roshanaei-Moghaddam et al., 2009), and decreased self-efficacy may be an important link (Craft et al., 2008; Trost et al., 2002). Exercise self-efficacy has also been shown to be an important component of weight loss and weight loss maintenance in overweight and obese women participating in weight loss intervention (Teixeira et al., 2010). Although depressive symptoms were associated with decreased exercise self-efficacy, they were not associated with objective measurement of engagement in moderate to vigorous physical activity. A similar lack of association between these variables was found in a study of depressive symptoms and physical activity in adolescent girls (Johnson et al., 2008). However, other research has demonstrated a link between depression and exercise (Roshanaei-Moghaddam et al., 2009). Individuals with major depression in a community based, longitudinal study were at greater risk for transitioning from an active to inactive lifestyle (Patten et al., 2009) and latent changes in depressive symptoms have been associated with latent changes in physical activity in older adults (Lindwall et al., 2011). Interventions that include exercise have also been shown to reduce depressive symptoms (Martinsen, 2008; Rethorst et al., 2009), thus suggesting that exercise is an important strategy for weight control in those with depression. Targeting self-efficacy for physical exercise may be a strategic focus to increase engagement in physical activity and influence overweight and obesity in women with depressive symptoms.

Women reporting greater depressive symptoms also reported engaging in more emotional eating and emotional eating was associated with increased BMI. The indirect effect of depressive symptoms on BMI through emotional eating was significant, indicating that emotional eating is an eating style that may be important in the association between depressive symptoms and BMI. This finding is consistent with the Finnish study demonstrating that emotional eating was important in the association between depressive symptoms and adiposity indicators in men and women (Konttinen et al., 2010). Depressive symptoms have been linked to emotional eating but not other forms of eating styles, such as external eating, or overeating in response to food cues (Ouwens et al., 2009). Other research supports indirect pathways between depressive symptoms and emotional eating, including difficulties identifying feelings (alexithymia), and impulsivity (Ouwens et al., 2009). Further parsing of the association between depressive symptoms and emotional eating may yield targets for assisting those with depressive symptoms in finding alternate strategies for modulating negative emotion and reducing the likelihood of overweight and obesity as a consequence of depressive symptoms and emotional eating. Emotion education and regulation strategies may prove to be an important component to weight loss for those with depression.

Cognitive restraint did not demonstrate an indirect effect between depressive symptoms and BMI in this sample of women. However, cognitive restraint has proven to be an important component in weight loss interventions (Teixeira et al., 2010) and associated with decreased food intake in higher BMI samples (Johnson et al., 2012). How cognitive restraint influences food intake in the context of depressive symptoms, and its association with impulsivity, urgency, and negative affect bears further exploration. Other potential avenues for future research related to eating styles and BMI include biological influences. Recent research investigating genetic influences on depressive symptoms and emotional eating suggest a possible role for decreased serotonin (5-HTTLPR polymorphism in serotonin transporter gene) in increasing emotional eating in response to depressive symptoms particularly for girls carrying this 5-HTTLPR genotype (van Strien et al., 2010). Reduced serotonin activity has been associated with increased BMI in adolescents (Fuemmeler et al., 2008). Dopamine receptors have also been linked to increased sweet food and calorie dense foods in young women with depressive symptoms (Agurs-Collins & Fuemmeler, 2011). Genetic influences may play a role in understanding the association between depressive symptoms, eating style, and food consumption in overweight and obesity (Stice et al., 2010) and could serve as markers for those most at risk in planning prevention activities.

This study has several limitations, most notably the reliance on cross-sectional data. Although we have constructed a model using relevant theory and research, the use of cross-sectional data limits our interpretation of the findings. For example, numerous studies suggest a bidirectional association between depression and obesity such that depression is linked to later obesity/overweight and obesity/overweight is associated with ensuing depression (Luppino et al., 2010), our results are unable to address this bidirectional association. Recent methodological work suggests the importance of longitudinal models for understanding pathways between variables to avoid model bias (Maxwell et al., 2011; Shrout, 2011). Longitudinal, prospective designs are necessary to confirm the associations we have identified. Intervention research has shown that decreasing BMI has can improve depressive symptoms, and exercise has also shown short term impact on depressive symptoms (Krogh et al., 2011; Rethorst et al., 2009). Because of the cross sectional nature of our data, we cannot rule out the possibility that women with higher BMI are suffering from depressive symptoms as a result of their BMI, or that having a higher BMI reduces efficacy for eating and exercise. However, intervention research suggests that increasing self-efficacy for eating and exercise and decreasing emotional eating are necessary for weight loss and weight loss maintenance, thus providing indirect support for targeting depressive symptoms in weight loss interventions (Annesi, 2011; Ouwens et al., 2009; Teixeira et al., 2010). Our study is also limited by reliance on self report measures for certain constructs, this could be addressed through more proximal assessment of symptoms, eating styles and self-efficacy, such as through daily diaries or ecological momentary assessment. These assessment methods may provide new insights into pathways between depressive symptoms and BMI. Where possible, we included objective indicators such as BMI and physical activity and these should continue to be incorporated in future studies. In spite of these limitations, our study presents novel findings showing associations between depressive symptoms, self efficacy for eating and exercise behaviors, and eating styles and BMI in women. Interventions that target self-efficacy for exercise, and that directly address emotional eating are warranted to assist women with depressive symptoms in addressing overweight and obesity. Given the high rates of both depression and obesity in women, there is an urgent need for new strategies to address these comorbidities.

Copyright information

© Springer Science+Business Media New York 2013