Associations between weight-related eating behaviors and adiposity in postmenopausal Japanese American and white women

Abstract

The purpose of this study was to test the associations between cognitive and psychological eating behavior traits and detailed measures of adiposity and body fat distribution using imaging-based methods in a cross-sectional study. Eating behavior traits (compensatory and routine restraint, external eating, and emotional eating) were assessed using the validated Weight-Related Eating Questionnaire, and measures of adiposity using anthropometry, dual energy X-ray absorptiometry (DXA), and magnetic resonance imaging (MRI). Each adiposity outcome of interest (total fat, ratio of trunk fat to periphery fat, visceral and subcutaneous fat as % of abdominal area, and % liver fat) was regressed on the four eating behaviors while adjusting for age and race/ethnicity. This study included a total of 60 postmenopausal Japanese American (n=30) and white (n=30) women (age: 60-65y, BMI: 18.8-39.6 kg/m²). Weight-related eating behavior traits did not differ by ethnicity. Higher external eating scores were associated with measures of total adiposity, including higher BMI (β = 0.36, p = 0.02) and DXA total fat mass (β = 0.41, p = 0.001), and with MRI abdominal subcutaneous fat (β = 0.55, p = 0.001). Higher routine restraint scores were associated with visceral adiposity (β = 0.42, p = 0.04). Our findings suggest that different weight-related eating behavior traits might increase not only total adiposity but also abdominal and visceral fat deposition associated with higher metabolic risks. Future research, preferably in a prospective study of men and women and including biomarkers related to psychological stress, will be needed to explore potential underlying biological mechanisms.

1.0 Introduction

Body fat deposition in the abdominal region is a known predictor of cardiovascular disease and diabetes [1], as well as some obesity-related cancers [2, 3]. Compared to peripheral or subcutaneous abdominal adiposity, intra-abdominal visceral adiposity has a greater negative impact on health outcomes, with recent evidence suggesting several biological differences between subcutaneous and visceral abdominal adiposity [4]. While the tendency for some individuals to disproportionately store body fat in the abdomen might be influenced by race/ethnicity or genetic traits [5, 6], there is a need to explore other, potentially modifiable, behavioral characteristics that contribute to elevated chronic disease and cancer risk related to increases in abdominal adiposity [4].

Much of the existing eating behavior research related to obesity has focused on identifying specific components of the diet (e.g., sugar sweetened beverages, dietary fat) and/or dietary patterns (e.g., breakfast skipping, fast food consumption) that might promote excessive weight gain [7-9]. Fewer studies have explored the role in obesity or central adiposity of cognitive and psychological aspects of eating behavior traits, which have been demonstrated to develop in childhood and persist through adulthood [10-14]. There are three theory-based constructs that describe these aspects of eating behavior. Dietary restraint [15], which reflects a rigid approach to weight control (routine restraint) as in chronic dieting, and a flexible approach to weight control (compensatory restraint), such as consciously eating more or less around times of overeating [16-18]. External eating [19] refers to the susceptibility to eat in response to the hedonic properties of food or other social/environmental influences. Emotional eating [20] reflects eating that occurs in response to negative affect. These eating behaviors, particularly external and emotional eating, have been associated with high-fat, high-sugar patterns of dietary intake [21-23] related to higher levels of adiposity, and might contribute to explaining differences in body fat distribution [24].

To date, studies on eating behavior traits and obesity have been almost exclusively based on crude measures of adiposity (e.g., body mass index (BMI) and/or waist circumference) and, therefore, have limitations of potentially misclassifying total or abdominal adiposity. Furthermore, eating behaviors could influence body fat accumulation in the abdominal versus peripheral region or in the visceral versus subcutaneous compartment of the abdomen. For example, chronic dietary restraint, characterized by dieting and food avoidance has been associated with higher perceived stress, measured by higher levels of cortisol in saliva [25] or urine [26, 27]. Also, increases in cortisol levels have also been positively associated with subsequent consumption of energy-dense foods [28-30] and fat deposition in the visceral compartment [31]. To our knowledge, only one prior study has examined the associations between eating behavior traits and image-based direct measures of abdominal and intra-abdominal fat distribution [24]. Provencher et al. [24], using the Three Factor Eating Questionnaire, demonstrated that flexible restraint had a negative correlation, and, rigid restraint and disinhibition, which encompasses external and emotional eating, had a positive correlation with greater total, abdominal, and visceral and subcutaneous adiposity in the abdomen. These findings provided support that theory-based eating behaviors might contribute to explaining the distribution of total and abdominal adiposity.

The purpose of our study was to test the association of eating behaviors not only with total body fat but also with regional body fat distribution adjusted for total adiposity, using imaging-based methods of dual energy X-ray absorptiometry (DXA) and magnetic resonance imaging (MRI) scans. The analyses utilize data collected for a pilot study exploring the differences in adiposity and abdominal fat distribution between Asian American and white women. The study was conducted to investigate prior observations that Japanese American women showed higher risks for breast cancer [32] and diabetes [33] compared to whites at any given BMI level even after accounting for other risk factors. It was hypothesized that differences in body fat distribution might explain the heterogeneity in disease risk. Findings from the pilot study confirmed that Asian women have greater abdominal and visceral adiposity than white women with similar body mass indexes [34]. The sample size of 30 per group was selected as the minimum required to detect the large differences in fat distribution between Japanese and white women expected based on preliminary data.

Based on previous literature, various eating behaviors are associated with both increased adiposity and biomarkers of acute and chronic psychological stress. It was, therefore, hypothesized that routine restraint, external eating and emotional eating would be associated with greater total adiposity, whereas compensatory restraint would be associated with lower total adiposity. We also hypothesized that routine restraint and emotional eating behaviors may be related to greater visceral as opposed to subcutaneous adiposity.

2.0 Participants and methods

2.1 Participants

White (n=30) and Japanese American (n=30) postmenopausal women living in Hawaii, age 60-65 years were recruited from the ongoing Multiethnic Cohort Study. Details of recruitment and data collection methods are fully described elsewhere [34]. Briefly, recruitment was stratified by race/ethnicity and BMI categories based on self-reported weight and height. Exclusion criteria included: current or recent smoking (<2 years); use of medications that may interfere with metabolism and body composition (chemotherapy, insulin or weight-loss drugs); substantial weight change (loss or gain of 20 pounds or more) in the past 6 months; pre- or perimenopausal status (had menses in the past 12 months); a measured BMI outside the target range (18.5–40 kg/m²); and a report of any soft or metal implants/objects in the body that could bias the body composition estimates or put subjects at risk in the MRI magnetic field. The study was approved by the Institutional Review Boards of the University of Hawaii and the Queen’s Medical Center, Honolulu, and all participants signed an informed consent.

2.2 Body composition measures

As described in detail in the original publication [34], subjects fasted overnight for an early-morning examination that included anthropometry measurements, DXA/MRI scans and fasting blood collection. All the participants wore a hospital gown for the anthropometry and DXA examinations, and comfortable shirts and pants without any metal parts for the MRI scan.

Anthropometric measures including height, weight, and waist circumference taken at the navel were collected using standardized protocols. Each measurement was taken twice, and a third time if the two measures differed by >0.1 kg for weight or by >0.5 cm for height or waist, after which the average of the two closest values was used in the analysis [34].

DXA was used to measure total and regional fat mass in the trunk, arms and legs. DXA measurements were completed by a certified radiographic technician using a whole-body DXA scan (GE Lunar Prodigy, Madison, WI, USA) [35]. Calibration using a manufacturer’s phantom was performed daily. Measures of total and regional body fat mass in the trunk, arms, and legs were used to calculate trunk-to-periphery fat mass ratio (fat mass in the trunk divided by the sum of fat mass in the arms and legs).

Abdominal MRI scans were additionally conducted to determine the trunk fat distribution in visceral, subcutaneous and hepatic compartments, which cannot be attained directly using DXA. MRI scans were obtained at a separate appointment from 20 Japanese American and 28 white women among the 60 participants. Due to scheduling limitations, the time between DXA and MRI measurements were within one to thirteen weeks (median 7.6 weeks). However, the body fat distributions based on DXA were found to be similar for women whose visits were close in time and those with visits spaced further apart. MRI was performed in a single session on a research-dedicated 3 Tesla TIM Trio scanner (Siemens Medical Systems, Erlangen, Germany; software version VB13). Details of the MRI protocol have been reported [34]. The abdominal and peritoneal walls were manually traced to measure the total fat area (mm²) and the visceral fat area (intraperitoneal, including fat inside internal organs), respectively, at the L4-L5 lumbar spine position [34]. The visceral fat area was then subtracted from total fat area to obtain the subcutaneous fat area (extraperitoneal). Percent liver fat was measured in a circular region of interest (15-25 cm²) in the lateral portion of the right lobe of the liver [36].

2.3 Eating behaviors

Four theory-based, cognitive/psychological aspects of eating behavior labeled compensatory restraint, routine restraint, susceptibility to external cues, and emotional eating were assessed with the 16-item Weight-Related Eating Questionnaire (WREQ) [17], which has been validated as an online survey in a diverse sample of males and females confirming its unbiased generalizability across gender, age (young adults, 18-24 years and adults, 25-81 years), race (Japanese Americans and whites), and BMI subgroups (under/normal weight and overweight/obese) [37]. Routine restraint (3 items) and compensatory restraint (3 items) reflect two subscales of cognitive dietary restraint that represent rigid and flexible approaches to weight control, respectively. Routine restraint is defined as the perceived routine restriction of energy intake to control weight by means of calorie counting, food avoidance/refusal, and purposefully intending to eat less than normal, consistent with dieting. Alternatively, compensatory restraint refers to an intentional restriction of energy intake that predominantly occurs before or following an episode of overeating. External eating (5 items) refers to eating that occurs in response to external oro-sensory cues (e.g., social and environmental cues or hedonic properties of food). Emotional eating (5 items) is defined as eating in response to negative emotions (e.g., anxiety, stress, and disappointment). Subscale scores are the average of responses on a five-point scale: 1 to 5 assigned to responses (“not at all, slightly, more or less, pretty well, completely”) to how each item “describes” the participant’s eating behavior. In a previously published validation study conducted in a diverse population [37], the subscales of the WREQ have shown good internal consistency (Cronbach’s alpha = 0.67 to 0.91; four-week test-retest reliability coefficients r = 0.74 to 0.86). In the current study, Cronbach’s alpha ranged from 0.72 (routine restraint) to 0.90 (emotional eating). Each participant completed the WREQ following their body composition testing in paper and pen format while at the clinic. The study coordinator was available to aid with the questionnaires; however, none of the participants reported having difficulty completing the WREQ.

2.4 Statistical analyses

All analyses were conducted using the SAS Statistical Software, version 9.2 (SAS Institute, Inc., Cary, NC, USA). Demographics, body composition, and eating behavior characteristics were compared between Japanese American and white women using analysis of covariance. To account for correlations among the WREQ subscales (r = 0.17 to 0.69) [17, 37], each body composition measure was regressed on the four continuous eating behavior variables. All models were adjusted for age and race/ethnicity. The addition of height had little influence on the observed associations, and this variable was not included in the final regression models. In separate models, interaction terms (race/ethnicity-by-eating behaviors) were included to determine whether the associations with body composition variables varied between Japanese American women and white women. Regression diagnostics were performed to confirm that model assumptions were met. Only % liver fat required log transformation and is presented as geometric means in tables. Participants missing MRI-specific measures (n=12) were excluded from analyses with these outcomes. Significance was defined as p<0.05.

3.0 Results

The women in this study (50% white, 50% Japanese American) had, on average, an age of 63.4±1.4 years, a weight of 66.3±5.3 kg, a height of 157.5±5.3 cm, and a BMI of 26.7±4.9 kg/m². By study design, there were no race/ethnicity differences on measures of total adiposity as measured by DXA. However, after adjusting for age, height, and total body fat, Japanese American women had significantly greater measures of abdominal adiposity (waist circumference, trunk fat mass, and trunk-to-periphery fat ratio), and of visceral fat (percent of abdominal area), but no statistically significant differences were found in subcutaneous fat or percent liver fat (Table 1), as reported earlier in detail [34]. There were no race/ethnic differences in mean eating behavior scores, and scores for each subscale were similar to our previous findings in these populations (Table 2) [37].

Table 1.

Comparison of adjusted means^* of body composition measures between white and Japanese American women

Body composition measures	White	Japanese American	P-value	ω2
Total adiposity
Body mass index (kg/m2)	26.9 (25.2,28.7)	26.5 (24.7,28.3)	0.718	--
DXA total fat (kg)	28.8 (25.0,32.5)	25.5 (21.8,29.3)	0.280	--
Abdominal adiposity
Waist circumference at navel (cm)	92.8 (90.9,94.7)	97.4 (95.5,99.3)	0.002	0.02
DXA trunk fat (kg)	13.4 (13.4,14.6)	15.3 (14.7,15.9)	0.008	0.10
DXA trunk-to-periphery fat ratio	1.1 (1.0,1.2)	1.4 (1.3,1.5)	0.008	0.10
Abdominal adiposity distribution
MRI subcutaneous fat area (cm2)	190.1 (169.5,210.7)	200.8 (176.4,225.2)	0.520	--
MRI % subcutaneous fat (% abdominal area)	30.3(27.1,33.4)	33.4 (29.7,37.2)	0.207	--
MRI visceral fat area (cm2)	127.5 (98.6,156.4)	153.0 (118.7,187.3)	0.275	--
MRI % visceral fat (% abdominal area)	18.5 (16.0,21.0)	23.9 (20.9,26.9)	0.011	0.09
MRI % liver fat (geometric mean)	3.8(3.5,4.1)	5.9 (5.6,6.3)	0.100	--

DXA – dual energy x-ray absorptiometry; MRI – magnetic resonance imaging

^*All measures of adiposity were adjusted for age. DXA total fat, trunk fat, trunk-to-periphery fat ratio, and % liver fat were further adjusted for height and total fat (where appropriate). Measures of adiposity distribution were adjusted for age and total fat only.

P-values and omega-squared (ω²) represent the effect of group designation on each outcome variable.

Sample included n=30 White women and n=30 Japanese American women for anthropometry and DXA variables and n=28 White women and n=20 Japanese American women for MRI variables.

Table 2.

Comparison of adjusted means^* in weight-related eating questionnaire scores between White and Japanese American women

Weight-related eating behaviors	White	Japanese American	P-value
Compensatory restraint	2.98 (2.61,3.35)	2.89 (2.52,3.26)	0.743
Routine Restraint	1.96 (1.68,2.25)	1.83 (1.55,2.12)	0.512
External Eating	2.39 (2.11,2.66)	2.43 (2.16,2.71)	0.813
Emotional Eating	1.96 (1.63,2.29)	1.96 (1.63,2.30)	0.996

Weight-related eating questionnaire subscale scores range from 1 (does not describe me at all) – 5 (describes me completely).

^*All means (95% CI) were adjusted for age.

Sample included n=30 White women and n=30 Japanese American women.

Multiple regression analyses (Table 3) demonstrated that after adjusting for age and race/ethnicity, higher external eating scores were associated with significantly greater BMI, waist circumference, total body fat, and trunk fat, as well as greater abdominal subcutaneous fat. Also, higher routine restraint scores were positively associated with greater visceral fat. There was the suggestion of an inverse association between compensatory restraint scores and trunk-to-periphery fat ratio (p=0.08) and % visceral fat (p=0.07); however, there was no association between emotional eating scores and body fat distribution. All associations were consistent between Japanese American women and white women (data not shown, p’s for interaction > 0.05).

Table 3.

Multivariate regression models of Weight-Related Eating Questionnaire scores and body composition measures (n=60)

Body composition measures	R2 eating behaviors only	R2 eating behaviors, age, race	Compensatory restraint		Routine restraint		External eating		Emotional eating
			β (SE)	p- value	β (SE)	p- value	β (SE)	p- value	β (SE)	p- value
Total adiposity
Body mass index	0.154†	0.161	−0.123 (0.177)	0.490	−0.048 (0.180)	0.790	0.364 (0.155)	0.023	0.055 (0.169)	0.743
DXA total fat	0.142†	0.193†	−0.192 (0.173)	0.272	0.027 (0.176)	0.877	0.411 (0.153)	0.001	−0.053 (0.166)	0.750
Abdominal adiposity
Waist circumference at navel	0.141†	0.148	−0.131 (0.178)	0.462	−0.050 (0.181)	0.781	0.378 (0.157)	0.019	−0.004 (0.170)	0.981
DXA trunk fat	0.112	0.131	−0.243 (0.180)	0.183	0.068 (0.183)	0.710	0.336 (0.159)	0.039	−0.036 (0.173)	0.836
DXA trunk-to- periphery fat	0.050	0.192†	−0.313 (0.173)	0.077	0.170 (0.176)	0.340	−0.043 (0.155)	0.783	0.038 (0.169)	0.821
Abdominal adiposity distribution
MRI subcutaneous fat area	0.238*	0.241†	−0.083 (0.189)	0.663	0.199 (0.199)	0.325	0.484 (0.158)	0.004	−0.088 (0.183)	0.633
MRI % subcutaneous fat	0.208*	0.238†	−0.046 (0.189)	0.811	0.057 (0.200)	0.778	0.545 (0.161)	0.001	−0.242 (0.190)	0.208
MRI visceral fat area	0.128	0.140	−0.305 (0.201)	0.136	0.429 (0.212)	0.050	−0.042 (0.175)	0.810	0.263 (0.202)	0.201
MRI % visceral fat	0.131	0.210	−0.359 (0.192)	0.069	0.423 (0.203)	0.044	−0.008 (0.164)	0.963	0.163 (0.193)	0.403
MRI % liver fat	0.076	0.171	−0.278 (0.197)	0.167	0.184 (0.208)	0.382	−0.186 (0.167)	0.270	0.160 (0.193)	0.411

WREQ = Weight-Related Eating Questionnaire; β – standardized parameter estimate in the models including age, ethnicity and all WREQ variables; SE – standard error; R² = percent of variance accounted for by variables in the models; DXA – dual energy X-ray absorptiometry; MRI – magnetic resonance imaging

Sample included n=30 white women and n=30 Japanese American women for anthropometry and DXA variables and n=28 white women and n=20 Japanese American women for MRI variables.

^*P<0.05 for the model F test

^†P<0.10 for the model F test

4.0 Discussion

The purpose of this study was to test the associations of the four weight-related eating behavior traits, assessed by the validated WREQ, to detailed measures of total, abdominal and intra-abdominal adiposity. Higher external eating scores were associated with higher measures of total adiposity (greater BMI and total fat) and abdominal adiposity (waist circumference and trunk fat), particularly subcutaneous abdominal adiposity, after adjusting for total adiposity. Higher routine restraint scores were positively associated with greater visceral adiposity. Additionally, there were non-significant suggestions that compensatory restraint scores were negatively associated with trunk-to-periphery fat and visceral adiposity. Emotional eating scores were not significantly associated with adiposity or abdominal adiposity distribution. Although further research is needed to replicate these findings, preferably using a prospective study design, they suggest that different weight-related eating behavior traits influence the deposition of adipose tissue in a manner that could negatively impact disease risk in some populations.

The current study is similar to one conducted by Provencher et al. [24], which used the Three Factor Eating Questionnaire (TFEQ) [38, 39] for the associations of eating behaviors with direct measures of visceral and subcutaneous adiposity. Among 352 middle-aged women (age 42.0±14.6 years; BMI 28.8±8.2 kg/m²), they observed significant and positive bivariate correlations of “rigid restraint” (equivalent to our “routine restraint”) or disinhibition (combining our “external eating” and “emotional eating”) with both visceral and subcutaneous fat areas [39] (r’s = 0.12 to 0.49, p<0.05). Differences between the Provencher study and ours is likely due to different categorizations in the instruments and also different statistical modeling. The WREQ separately assesses two, moderately correlated constructs of external and emotional eating, whereas the TFEQ assessed disinhibition that combined situational and emotional eating subscales [17, 18, 37]. Also, we assessed multivariable-adjusted independent associations for each eating behavior, whereas Provencher et al. examined bivariate Spearman correlations. As a result, final categorization we were able to show an independent association of routine restraint with visceral adiposity, and of external eating with subcutaneous adiposity. Also, we observed a stronger association for the relative compared to the absolute visceral or subcutaneous fat area, which supports independent associations for either adipose compartment. As such, this study makes a unique contribution by providing more detailed infromation for the associations between eating behaviors and measures of adiposity.

As discussed above, we found that external eating was positively associated with measures of adiposity, including greater BMI, total body fat, waist circumference, and trunk fat, but not visceral adiposity. These associations were hypothesized based on Schachter’s theory of externality [40], which states that some individuals are at risk of obesity as a result of the susceptibility to eat in response to cues in the external environment (e.g., food-related cues) more so than internal cues (e.g., physiological states of hunger and fullness). However, empirical support for the associations between external eating and measures of total adiposity, particularly BMI, has not been consistent [41, 42]. In two validation studies of the WREQ, one with 721 university students (70% female, age 18.9±1.1 years; BMI 23.1± 3.5 kg/m²) and another with 621 adults (61% female; age 34.8± 16.6 years; BMI 25.7±6.2 kg/m²), the relationships between external eating and current BMI as well as weight change rate were not significant [17, 37]. These findings have not been limited to the WREQ as similarly null associations have been observed between external or uncontrolled eating and BMI using the Dutch Eating Behavior Questionnaire and the revised 18-item Three Factor Eating Questionnaire, respectively [17, 43, 44]. The reasons for observing significant associations between external eating and measures of adiposity in this study and not others are unclear. Future research will be needed to explore possible moderators and mediators of this relationship.

Our observations are consistent with the hypothesis that restrained eating (i.e., routine restraint) may stimulate, or coincide with, cortisol secretion and the sympathetic nervous system, and thus may be conducive to visceral adipogenesis [45], whereas external eating may occur without stress responses and thus, increase overall, mostly subcutaneous, adiposity. Routine restraint reflects a more rigid approach to weight control, and the chronic aspects of dieting have been associated with depression, stress, and issues of self-esteem [46]. Measures of dietary restraint are consistently demonstrated to be positively associated with higher urinary cortisol excretion in pre- and post-menopausal women [25-27, 47], as well as higher BMI and weight gain [48]. On the other hand, rigid eating or routine restraint may simply be a confounding trait among people who are susceptible to stress and cortisol responses, and who, thus, develop more visceral adiposity in response to the higher concentration of vagus nerve ending stimuli in the visceral than subcutaneous adipose tissue [49]. However, even in such a case, modifying routine restraint eating may be beneficial to mitigate the cycle of overeating and stress. We also observed inverse associations between compensatory restraint and measures of adiposity, though the associations for trunk-to-peripheral fat and visceral adiposity were only of borderline significance. These findings are consistent with the literature on dietary restraint that suggests that using a flexible approach towards weight control protects against weight gain [24, 50]. Published findings have also shown that compensatory restraint is associated with greater fruit and vegetable consumption in young adults [17] and is consistent with trying to prevent weight gain, as well as trying to lose weight [37]. Although these findings may be confounded by a personality trait that affects both the eating behaviors and adipogenicity, they suggest that future interventions may be targeted to curb the detrimental effects of rigid and chronic dieting, as well as to encourage more flexible approaches to eating and weight control.

There was little evidence that emotional eating was associated with abdominal adiposity: although our study’s power was suboptimal for detection of modest associations, the slopes for this relationship in the regression models are close to null. To our knowledge, there are no other studies that have explored the relationship between emotional eating scores with image-based direct measures of abdominal adiposity. However, associations between emotional eating and measures of BMI, waist circumference, body fat, and/or weight change are generally positive in adult populations [14, 17, 37, 51-53]. Reasons for this attenuated association in our sample are unknown, but it may be related to the specificity of the sampled population of postmenopausal women. The mean emotional eating scores were low, suggesting that few women in this sample would describe themselves as susceptible to eating in response to negative effect. However, the mean emotional eating score among the mature women in the current study (1.96) were only slightly lower than that observed among younger women (2.13) [37]. Replication of the findings will be important in clarifying this association. On the other hand, episodes of stress, which could initiate an episode of emotional eating, are known to induce acute increases in cortisol [28, 29], which might have only a small and transient influence on visceral adiposity [29].

This study adds to the very limited research exploring the association between the cognitive/psychological aspects of eating behavior with detailed body composition measures. One of the strengths of our study was the stratified recruitment that ensured equal distribution of BMI within race/ethnic groups. With the balanced distribution for total adiposity between whites and Japanese Americans, we were able to show that the associations between eating behavior and adiposity were similar, although the latter group has proportionately greater abdominal and visceral fat. Also, the use of the WREQ and multivariable-adjusted modeling allowed us to distinguish the effects of two different weight control strategies (compensatory and routine restraint) as well as of external and emotional disinhibition. To our knowledge, the current study is among the first to demonstrate that the eating style can account for differences in the deposition of abdominal adipose tissue.

Despite the strengths of the study, generalizability of our findings is limited to postmenopausal women of ages 60 to 65 and will need to be replicated in other populations. Like the previous study by Provencher et al. [24], we used a cross-sectional design and, thus, lack evidence for the temporal sequence between eating behaviors and adiposity development. However, considering the cumulated research that demonstrated the early development and stability of eating behavior traits from childhood [10, 13], as evidenced by an association of early eating behaviors with long-term weight gain [14], our results add to the supportive evidence that certain eating behaviors may be modified to prevent total, abdominal or visceral adiposity. Our sample size was small, which may have limited out ability to detect modest associations as statistically significant (e.g., emotional eating and adiposity) and our ability to explore any moderation effects of race on the association between eating behaviors and measures of adiposity, though no modification by race was hypothesized. Yet, we were able to observe several significant and potentially important associations in the current and the previously published original study analyses [34]. In addition, the WREQ is a self-reported, subjective assessment of the cognitive/psychological aspects of eating behavior. Some individuals may be unaware of eating episodes that are motivated by external (social/environmental cues or hedonic properties of palatable foods) or emotional cues [37]; however, this random misclassification would most likely have attenuated rather than strengthened the observed associations. Lastly, other factors could have influenced both adiposity and eating behaviors, and thus could have confounded the association, such as the independent effect of stress. Although other factors, such as physical activity and dietary composition, might be correlated with both eating behaviors and adiposity distribution, we purposefully did not control for them as it is believed that they would be intermediaries in the pathway from eating behavior to adiposity and adjustment would unduly attenuate the relationships of interest.

5.0 Conclusions

Our findings suggest that several body weight-related eating behavior traits were associated with measures of body fat amount and distribution. Future research will be needed to replicate these findings prospectively in a larger and more diverse sample of people, including both sexes and across broader age and race/ethnicity groups. In addition, the assessment of biomarkers related to stress (e.g., cortisol) and inflammation would help to understand the underlying physiological mechanisms linking eating behavior traits to the distribution of abdominal adipose tissue.

Highlights.

We test associations between weight-related eating behaviors and objective measures of total adiposity and abdominal adiposity. > Visceral and subcutaneous abdominal adipose was measured by DXA and MRI. > External eating is associated with subcutaneous abdominal adiposity. > Routine dietary restraint is associated with visceral abdominal adiposity. > Different weight-related eating behaviors are associated with depot-specific adiposity.