Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 May 1.
Published in final edited form as: J Abnorm Psychol. 2021 May;130(4):388–398. doi: 10.1037/abn0000669

How Changing Life Roles Predict Eating Disorder Pathology over 30-Year Follow-Up

Madeline R Wick 1, Tiffany A Brown 2, Elizabeth H Fitzgerald 1, Pamela K Keel 1
PMCID: PMC8244170  NIHMSID: NIHMS1697031  PMID: 34180703

Abstract

Epidemiological data support higher prevalence of eating disorders in midlife than previously believed. Yet, few studies have examined risk factors unique to adult development. The present study examined how changes in life roles (educational, marital, and parental status) predicted disordered eating as participants transitioned from their twenties to their fifties. Participants (N=624 women and N=276 men) completed baseline assessments in college and at 10-, 20-, and 30-year follow-up, with 72% of women and 67% of men completing 30-year follow-up. Multilevel models examined how changes in life roles predicted changes in disordered eating. For women, obtaining a graduate degree predicted decreased eating pathology initially, but over time predicted subsequent increases in Drive for Thinness. Men’s eating pathology was not affected by obtaining a graduate degree. Changes in marital status demonstrated no significant association with disordered eating for either gender. Becoming a parent predicted a significant decrease in Drive for Thinness at the subsequent assessment but no further declines with age; whereas those who never had children showed significant decreases in both Drive for Thinness and Bulimia with age. For both women and men, becoming a parent may decrease the importance of shape and weight as sources of self-evaluation. However, women obtaining advanced degrees and parents may experience shifts in eating pathology related to the “Career-and-Care-Crunch” according to Mehta and colleagues’ (2020) recent conceptualization of adult developmental stages. Pending independent replication, future research might design interventions for those whose role transitions put them at greater risk for disordered eating during midlife.

Keywords: longitudinal, adult development, gender, disordered eating

General Scientific Summary:

Over the course of adult development, changing life roles predict changes in disordered eating, and some effects differ between women and men. Becoming a parent is linked to changes in eating pathology in both women and men; whereas obtaining a graduate degree predicted changes in disordered eating for women but not men.

Once thought to be rare in older age groups, recent epidemiological data estimate 3.5% of midlife women and between 1 and 2% of midlife men meet full diagnostic criteria for an eating disorder (ED) (Mangweth-Matzek & Hoek, 2017). These findings support the need to employ longitudinal designs to identify factors that contribute to ED risk beyond young adulthood. Yet, most longitudinal studies have focused on cohorts followed from adolescence through young adulthood. Recently, Brown and colleagues (2020) found that risk factors established in younger cohorts, such as dieting frequency, prospectively predicted higher drive for thinness and bulimic symptoms in men and women as they transitioned from their twenties to their fifties. In addition, women and men differed in the potency of dieting frequency as a predictor of eating pathology over time. Specifically, as men aged, dieting frequency became a less potent predictor of drive for thinness whereas its potency did not change for women. These findings support the need to examine risk factors unique to adult development and potential gender differences in effects across the lifespan. The current study extends this previous work by evaluating how changes in educational, marital, and parental status predict changes in eating pathology for women and men as they enter middle adulthood and whether gender moderates these associations. Furthermore, the current study includes BMI as a covariate to determine whether the relationship between life roles and disordered eating is independent of changes in BMI over time.

Developmental Transitions Unique to Adulthood

The vast majority of longitudinal risk factor studies for EDs focus on adolescents and young adults (e.g., Herpertz-Dahlmann et al., 2015; Levinson et al., 2020), among whom shifts in life roles are limited to transitioning from middle school to high school to college. Changes in educational status are relatively uniform across participants and confounded by age, limiting any value in examining educational status as a predictor of changes in eating pathology. Most individuals in teenage years and early twenties are neither married (U.S. Census Bureau, 2017) nor have children (Martin et al., 2018). According to Mehta, Arnett, Palmer, and Nelson (2020), emerging adulthood (18–29 years) is characterized by pursuing tertiary education and exploration of careers and relationships, potentially maintaining the relevance of risk factors examined in younger samples.

During established adulthood (30–45 years) (Mehta et al., 2020), individuals are more likely to have completed school, to get married (U.S. Census Bureau, 2017) and to begin having children (Martin et al., 2018). This period is characterized by the “Career-and-Care-Crunch”, in which demands for professional advancement and childrearing collide (Mehta et al., 2020). This ‘Crunch’ is particularly salient for women who continue to provide the majority of childcare, even when both parents work outside the home. Compared to their twenties, women in their thirties and forties spend more time with men, children, and older adults – that is, they affiliate more with groups at lower risk for eating pathology. In addition, becoming a spouse and parent introduces new sources for evaluation of self-worth which may displace the importance of weight and shape. Combined, these changing life roles may contribute to lower ED risk for women as they age (Keel et al., 2007). However, as men age, they may be exposed to more dieting among their same-gender peers (Brown et al., 2020; Keel et al., 2007), thus increasing their ED risk (Gravener et al., 2008). Both increasing body weight (Flegal et al., 2016) and secular shifts in body ideals for men (Thompson & Cafri, 2007) may further contribute to greater emphasis on weight and shape. However, like women, men may benefit from basing self-evaluation on their roles as husbands and fathers.

During middle adulthood (45–64 years) (Mehta et al., 2020), women and men experience another set of unique shifts in life roles. Compared to established adulthood, adults in their fifties have advanced in their careers and have older children who demand less physical care (Mehta et al., 2020). At midlife, many adults explore new educational opportunities and may begin dating after a first marriage ends in divorce (Montenegro, 2003). Further, those with partners may seek to reconnect romantically to protect against separation (Copeland, 2014). This might include putting more effort into being physically attractive to one’s partner, an established risk factor for disordered eating in younger samples (Eisenberg et al., 2013). Despite the potential impact of these transitions on changes in disordered eating over adult development, very limited information exists regarding associations between life roles and disordered eating.

Association between Life Roles and Disordered Eating

Cross-sectional designs have produced largely mixed results regarding the association between life roles and disordered eating in adult samples. For example, being married predicted decreased eating pathology in one study of women (Borjorquez et al., 2018) but not another (Mangweth-Matzek et al., 2014), and neither study supported an association between being a parent and disordered eating (e.g., Borjorquez et al., 2018; Mangweth-Matzek et al., 2014). A review (Mitchison & Hay, 2014) found an equal number of studies supporting a positive and a negative association between higher educational attainment and disordered eating and concluded there was no significant association between the two variables.

Although these studies’ positive and null findings could be interpreted as reflecting a combination of Type I error and true null results, it is important to note that participants’ ages ranged by 20 to 48 years within cohorts (Borjorquez et al., 2018; Mangweth-Matzek et al., 2014). Thus, cross-sectional designs have confounded cohort effects with developmental effects. Further, where significant associations have been found, it is not possible to distinguish among risk factors, correlates, and consequences of disordered eating. These limitations underscore the need to move beyond cross-sectional designs to determine whether changes in life roles predict changes in disordered eating.

To our knowledge, only two longitudinal studies have examined the influence of life roles on disordered eating by following participants into adulthood (Keel et al., 2007; Vogeltanz-Holm et al., 2000). Vogeltanz-Holm and colleagues (2000) found that being married predicted less intense dieting at 5-year follow-up in women who completed baseline surveys when they were between 21 and 49 years of age. However, this study only included women, leaving it unclear whether patterns would generalize to men, and the wide age range introduces cohort effects. Keel et al. (2007) followed the cohort included in the current investigation from a mean age of 20 to 40 years. Replicating and extending findings from Vogeltanz-Holm et al. (2000), marriage was associated with decreased disordered eating over time in women, but this effect was not found in men. In addition, becoming a parent was associated with decreased drive for thinness in both women and men and decreased bulimic symptoms in women. Finally, changes in educational attainment were not significantly associated with changes in disordered eating over time. Importantly, models were run in women and men separately. This precluded formal statistical testing of gender as a moderator and limited conclusions about whether women and men differ in associations between life roles and eating pathology. In addition, BMI was not included as a covariate in life role models presented by Keel et al. (2007). Finally, results followed participants into established adulthood but not did extend to middle adulthood (Mehta et al., 2020).

The Present Study

The present study examined shifting life roles, including changes in educational attainment, marital status, and parental status as predictors of disordered eating in a cohort of women and men assessed when they were in college and again at 10-, 20-, and 30-year follow-up. This study contributes to the existing literature in three important ways. First, it includes participants at a unique developmental stage (mean age (SD) = 50 (2) years). This contributes to the literature significantly as there are only a small number of studies, and almost no longitudinal studies, that have examined ED risk during middle adulthood. Second, while life roles as predictors for eating pathology were examined separately in men and women in Keel et al. (2007), the current paper examined gender as a moderator to determine if these effects differ significantly between men and women. Finally, we included BMI as a time-varying predictor to determine whether the relationship between life roles and disordered eating is independent of changes in BMI over time. Thus, to our knowledge, the current paper is the first to examine changing life roles as predictors of disordered eating into middle adulthood, whether gender moderates the association between life roles and changes in eating pathology, and whether changes in BMI account for these associations.

Based on findings from our prior assessment wave (Keel et al., 2007), we hypothesized that being married would predict decreased disordered eating and that gender would significantly moderate this association. We also hypothesized that parenthood would predict decreased disordered eating in both women and men. Given mixed findings regarding associations between educational attainment and disordered eating (Mitchison & Hay, 2014), no a priori hypothesis was made for this variable.

Methods

Participants

College students (N = 900; 624 women and 276 men) from a prestigious northeastern university in 1982 were followed prospectively at 10-year intervals. At 30-year follow-up in 2012, participants’ mean (SD) age was 50.53 (±2.06) years. At follow-up, the racial and ethnic breakdown was: 84% White, non-Hispanic, 8% Asian, 4% African American, 4% Hispanic, and 0.2% Other.

Procedures

In 1982, participants were randomly sampled from their university to participate in a survey of health and eating behaviors. Baseline participation rates were high (women: n = 624 [78%], men: n = 276 [69%]). Participants were followed at 10-year intervals, resulting in 10-, 20-, and 30-year follow-up data (Heatherton et al., 1997; Keel et al., 2007; Brown et al., 2020). Participation rates were strong at 10- (women: 82%; men: 76%; Heatherton et al., 1997) and 20-year follow-up (women: 75%; men: 70%; Keel et al., 2007). In 2012, all living participants were contacted for follow-up. At 30-year follow-up, 15 participants (9 women, 6 men) were deceased, and, of those still living, 72% of women (N = 440) and 67% of men (N = 181) participated (see Brown et al., 2020 for additional details). All study procedures were approved by the Institutional Review Board, and all participants provided informed consent prior to participation at each assessment wave.

Measures

Life Roles.

Life roles were assessed using single items for educational attainment (1 = high school degree, 2 = college degree, 3 = graduate degree), marital status (0 = not married, 1 = married), and parental status (0 = no children, 1 = have children) at 10-, 20-, and 30- year follow-up. Life roles were not asked at baseline when participants were in college and were coded to reflect participants’ status as full-time college students, which required all participants to have high school degrees and none to have completed college. In addition, participants were coded as unmarried and having no children at baseline, consistent with prior work (Keel et al., 2007). We originally hoped to examine occupational status as well. However, review of the survey item first used in the 1992 survey (Heatherton et al., 1997) revealed that responses did not distinguish between current and past employment, making the variable incapable of capturing how changes in occupational status might be related to changes in eating pathology.

Eating Pathology.

Eating pathology was assessed using the Eating Disorders Inventory (EDI; Garner et al., 1983) Drive for Thinness and Bulimia subscale versions available in 1982. Reflecting how bulimia was defined in the Diagnostic and Statistical Manual, 3rd Edition (American Psychiatric Association, 1980), the EDI Bulimia scale focuses heavily on binge eating. Measurement invariance for the EDI across genders and through age 50 has been established in this cohort (Brown et al., 2020). The internal consistencies of the Drive for Thinness (α=.83–.90) and Bulimia subscales (α=.72–.86) were good in the present study.

Data Analyses

Data distributions were examined for normality, and both Drive for Thinness and Bulimia scores were log transformed to correct for positive skew. Little’s MCAR test was significant, χ2(6) = 111.48, p = .03, indicating that data were not missing completely at random. As reported previously (Brown et al., 2020), participation at 30-year follow-up was not significantly associated with gender, baseline BMI, or EDI Drive for Thinness or Bulimia scale scores (all p-values >.10). Racial/ethnic minority participants were less likely to complete 30-year follow-up compared to non-Hispanic White participants (non-Hispanic White: 82.01%; Other Race/Ethnicity: 68.33%, χ2[1, n=648] = 11.21, p = .001). Furthermore, those who remained single at 10-year follow-up were significantly less likely to participate at 30-year follow-up (single: 66.29%; married: 78.06%, χ2[1, n=713] = 12.30, p < .001). No other significant differences were found. Missing data for Drive for Thinness and Bulimia scale scores were imputed using expectation-maximization (EM) (Schafer & Graham, 2002), consistent with our prior research (Brown et al., 2020; Keel et al., 2007). Missing data for dichotomous variables (e.g., life roles) were not imputed.

Multilevel models were used to test hypotheses regarding whether changing life roles predicted changes in eating pathology (EDI Drive for Thinness and Bulimia scores) over time, with gender as a moderator, first without and then with BMI as a time-varying predictor. Within-person changes in eating pathology were modeled at Level 1, and life roles, gender, and the interaction between life roles, gender, and age were modeled at Level 2. A random intercept, random slopes model was fit to data, such that participants’ intercepts and rate of change over time could vary, representing the least restrictive approach. Other covariance structures were examined; however, these alternatives were rejected because they did not converge, did not produce a positive definite Hessian matrix, or provided an inferior fit to the data according to AIC and BIC. Full information maximum likelihood estimation (FIML) was used. Predictor variables were grand mean centered across waves, an unstructured covariance matrix was used, and a linear effect of time was specified for all models, consistent with observed changes in Drive for Thinness and Bulimia Scores across assessment waves (Brown et al., 2020).

Two sets of models were run. In the first set of models, we examined how a change in life roles predicted changes in eating pathology when the change in life role occurred between the prior and subsequent assessment (“intervening change” models). For example, these models tested how getting married between college and 10-year follow-up predicted Drive for Thinness at 10-year follow-up compared to staying single over that time period, controlling for Drive for Thinness in college. In the second set of models, we examined how an established change in life roles predicted subsequent changes in eating pathology (“time-lagged change” models). For example, these models tested how getting married between college and 10-year follow-up predicted changes in Drive for Thinness from 10- to 20-year follow-up, offering important information about the sustained impact of changes in life roles. All analyses were run using Statistical Package for the Social Sciences (IBM SPSS, Version 25).

Results

Table 1 displays the proportion of women and men who obtained college and graduate degrees, were married, and became parents at each follow-up wave. There were no significant gender differences in educational attainment. There were no significant differences between genders in likelihood of being married in their thirties or fifties; however, men were more likely than women to be married in their forties. There were no significant gender differences in likelihood of being a parent at 10- or 20-year follow-up; however, by their fifties, men were more likely than women to be a parent. BMI was higher for men compared to women across follow-up.

Table 1.

Life Roles at 10-, 20-, and 30-Year Follow-up in Women and Men

10-Yr Follow-up Mean (SE) Age = 29.94 (0.07) 20-Yr Follow-up Mean (SE) Age = 40.11 (0.08) 30-Yr Follow-up Mean (SE) Age = 50.53 (0.08)
Women n=509 Men n=206 Women n=465 Men n=189 Women n=440 Men n=181
n (%) n (%) X2 (df) n (%) n (%) X2 (df) n (%) n (%) X2 (df)
Educational Attainment 0.42 (2) 1.42 (2) 0.48 (2)
 High School 2 (0.4) 1 (0.5) 2 (0.4) 1 (0.5) 1 (0.2) 1 (0.6)
 College Degree 133 (26.1) 49 (23.8) 110 (23.7) 37 (19.6) 93 (21.1) 40 (22.1)
 Graduate Degree 373 (73.3) 155 (75.2) 349 (75.1) 151 (79.9) 339 (77.0) 138 (76.2)
Married 253 (49.7) 107 (51.9) 0.24 (1) 320 (68.8) 148 (78.3) 5.95 (1)* 337 (76.6) 150 (82.9) 2.54 (1)
Have children 120 (23.6) 49 (23.8) 0.05 (1) 314 (67.5) 133 (70.4) 0.19 (1) 340 (77.3) 157 (86.7) 7.75 (1)**
mean (SD) mean (SD) t (df) mean (SD) mean (SD) t (df) mean (SD) mean (SD) t (df)
Body Mass Index 22.03 (3.30) 24.13 (2.70) 9.26 (898)*** 23.38 (4.09) 25.45 (2.90) 7.61 (724.01)*** 24.05 (4.50) 26.12 (3.23) 6.90 (716.53)***
Open in a new tab

Note. The number/percentage reflects those who completed surveys at each wave. The percentage of missing data at 10-year follow-up was as follows: Educational Attainment 20.8%; Married 20.8%; Have Children 14.2%; Body Mass Index 21.0%. Percentage of missing data at 20-year follow-up: Educational Attainment 27.8%; Married 27.3%; Have Children 30.1%; Body Mass Index 27.8%. Percentage of missing data at 30-year follow-up: Educational Attainment 32.0%; Married 31.3%; Have Children 31.7%; Body Mass Index 31.8%. Degrees of freedom reflect adjustments when heterogeneity of variance was found.

*

p<.05,

**

p<.01,

***

p<.001

Associations Between Life Roles and Eating Pathology

Table 2 displays associations between life roles and eating pathology at each follow-up wave. Those who married were more likely to have children at each follow-up and had higher BMIs at 10-year follow-up compared to those who were single. Parents had higher BMIs compared to non-parents at 10- year follow-up. At 20- and 30-year follow-up, being married was associated with lower Drive for Thinness and lower Bulimia scores. In addition, being a parent was associated with lower Drive for Thinness at 20-year follow-up. Higher BMI was associated with higher Drive for Thinness and Bulimia scores across all follow-up waves. No other associations were significant.

Table 2.

Bivariate Associations Between Predictors and Outcome at Each Follow-Up Wave

10-Year Follow-Up 1
χ2 / t (df)		 2
χ2 / t (df)		 3
χ2 / t (df)		 4
r (n)
1. Educational Attainment -
2. Marital Status 0.23 (1) -
3.Parental Status 0.01 (1) 150.35 (1)*** -
4. Body Mass Index 0.37 (263.14) −2.47 (711)* −2.04 (770)* -
5. EDI Drive for Thinness −0.30 (708) 0.07 (711) 0.99 (770) .22 (900)***
6. EDI Bulimia 0.02 (708) 0.43 (711) 0.80 (341.12) .29 (900)***
20-Year Follow-Up 1
χ2 / t (df)		 2
χ2 / t (df)		 3
χ2 / t (df)		 4
r (n)
1. Educational Attainment -
2. Marital Status 0.02 (1) -
3. Parental Status 1.90(1) 245.43 (1)*** -
4. Body Mass Index −0.25 (216.14) 0.54 (296.28) 1.63 (290.43) -
5. EDI Drive for Thinness 1.00 (645) 2.16 (289.89)* 2.67 (280.90)** .20 (900)***
6. EDI Bulimia 1.48 (222.10) 2.10 (292.13)* 1.00 (627) .34 (900)***
30-Year Follow-Up 1
χ2 / t (df)		 2
χ2 / t (df)		 3
χ2 / t (df)		 4
r (n)
1. Educational Attainment -
2. Marital Status 4.44 (1)* -
3. Parental Status 0.08 (1) 154.91 (1)*** -
4. Body Mass Index 0.64 (608) 0.92 (175.07) 0.65 (152.37) -
5. EDI Drive for Thinness −1.41 (608) 2.08 (184.20)* 0.98 (613) .19 (900)***
6. EDI Bulimia 0.82 (608) 2.03 (182.05)* 0.65 (613) .38 (900)***
Open in a new tab

Note. Fewer than 5 participants had not obtained a college degree by 10-year follow-up and were omitted from analyses to prevent violating minimum cell size assumptions. Degrees of freedom reflect adjustments when heterogeneity of variances was found. Educational attainment: 1 = high school degree, 2 = college degree, 3 = graduate degree; marital status: 0 = not married, 1 = married; parental status: 0 = no children, 1 = have children.

*

p<.05,

**

p<.01,

***

p<.001.

Changing Life Roles as Predictors of Eating Pathology in Women and Men across Adulthood

Findings from multilevel model analyses examining associations between changes in life roles and changes in disordered eating, with and without BMI as a covariate, are presented in Table 3 for intervening models and Table 4 for time-lagged models. There were no significant effects of marital status in any model.

Table 3.

Intervening Models Predicting Changes in Eating Pathology across Adulthood

Drive for Thinness Bulimia
Without BMI Partial R2 With BMI Partial R2 Without BMI Partial R2 With BMI Partial R2
Fixed Effects γ (SE) γ (SE) γ (SE) γ (SE)
 Intercept (γ00) 11.47 (0.19)*** .70 2.79 (0.73)*** .01 10.64 (0.14)*** .79 1.65 (0.51)** .01
 Age (γ10) −0.02 (0.01) <.01 −0.06 (0.01)*** .01 −0.06 (0.01)*** .02 −0.09 (0.01)*** .06
 BMI -- -- 0.37 (0.03)*** .08 -- -- 0.39 (0.02)*** .16
 Gender 3.31 (0.42)*** .04 4.12 (0.41)*** .06 1.42 (0.31)*** .01 2.28 (0.29)*** .04
 EA −0.25 (0.18) <.01 −0.23 (0.17) <.01 −0.49 (0.13)*** .01 −0.49 (0.12)*** .01
 MS 0.18 (0.26) <.01 0.11 (0.25) <.01 −0.06 (0.18) <.01 −0.13 (0.18) <.01
 PS −0.88 (0.28)** .01 −0.92 (0.28)** .01 −0.29 (0.20) <.01 −0.34 (0.19) <.01
 Gender × Age −0.06 (0.03)* <.01 −0.05 (0.03)† <.01 −0.06 (0.02)** <.01 −0.05 (0.02)* <.01
 EA × Age 0.07 (0.01)*** .02 0.07 (0.01)*** .02 0.04 (0.01)*** .01 0.04 (0.01)*** .01
 MS × Age −0.02 (0.03) <.01 −0.02 (0.03) <.01 −0.02 (0.02) <.01 −0.02 (0.02) <.01
 PS × Age 0.07 (0.03)* <.01 0.09 (0.03)** <.01 0.06 (0.02)** <.01 0.07 (0.02)*** .01
 EA × Gender −0.80 (0.39)* <.01 −0.82 (0.38)* <.01 −0.17 (0.29) <.01 −0.22 (0.27) <.01
 MS × Gender 0.28 (0.56) <.01 0.38 (0.55) <.01 −0.23 (0.40) <.01 −0.11 (0.39) <.01
 PS × Gender −0.07 (0.62) <.01 −0.18 (0.61) <.01 0.06 (0.44) <.01 −0.08 (0.42) <.01
 EA × Age × Gender 0.12 (0.03)*** .01 0.10 (0.03)** .01 0.06 (0.02)** .01 0.04 (0.02)† <.01
 MS × Age × Gender −0.04 (0.06) <.01 −0.03 (0.06) <.01 −0.02 (0.04) <.01 −0.02 (0.04) <.01
 PS × Age × Gender −0.03 (0.07) <.01 −0.01 (0.06) <.01 0.05 (0.05) <.01 0.07 (0.05) <.01
Variance
 Within-Person 10.54 (0.44)*** 10.34 (0.43)*** 5.23 (0.22)*** 4.99 (0.20)***
 Intercept 15.90 (0.95)*** 14.32 (0.87)*** 8.96 (0.52)*** 7.45 (0.44)***
 Slope 0.01 (<0.01)*** 0.01 (<0.01)*** 0.01 (<0.01)*** 0.01 (<0.01)***
Fit Indices
 AIC 16614.5 16471.0 14920.7 14613.6
 BIC 16733.5 16595.9 15039.7 14738.5
Open in a new tab

Note. EA=Educational Attainment; MS=Marital Status; PS=Parental Status;

†

p = .05,

*

p < .05,

**

p < .01,

***

p < .001

Table 4.

Time-Lagged Models Predicting Changes in Eating Pathology across Adulthood

Drive for Thinness Bulimia
Without BMI Partial R2 With BMI Partial R2 Without BMI Partial R2 With BMI Partial R2
Fixed Effects γ (SE) γ (SE) γ (SE) γ (SE)
 Intercept (γ00) 11.28 (0.20)*** .67 5.89 (0.83)*** .03 10.23 (0.14)*** .77 5.17 (0.58)*** .05
 Age (γ10) 0.04 (0.02)* <.01 0.01 (0.02) <.01 <0.01 (0.01) <.01 −0.03 (0.01) <.01
 BMI -- -- 0.24 (0.04)*** .03 -- -- 0.23 (0.03)*** .05
 Gender 2.70 (0.52)*** .02 3.16 (0.45)*** .03 1.10 (0.32)** .01 1.53 (0.31)*** .01
 EA −0.09 (0.18) <.01 −0.07 (0.18) <.01 −0.26 (0.12)* <.01 −0.24 (0.12)† <.01
 MS −0.41 (0.32) <.01 −0.43 (0.33) <.01 −0.10 (0.22) <.01 −0.14 (0.23) <.01
 PS 0.30 (0.39) <.01 0.25 (0.40) <.01 0.16 (0.27) <.01 0.13 (0.28) <.01
 Gender × Age −0.01 (0.04) <.01 −0.01 (0.04) <.01 <0.01 (0.03) <.01 <0.01 (0.03)
 EA × Age 0.03 (0.02) <.01 0.03 (0.02) <.01 <0.01 (0.01) <.01 <0.01 (0.01) <.01
 MS × Age 0.04 (0.05) <.01 0.03 (0.05) <.01 0.01 (0.03) <.01 <0.01 (0.03) <.01
 PS × Age −0.03 (0.05) <.01 −0.03 (0.05) <.01 −0.02 (0.04) <.01 −0.01 (0.04) <.01
 EA × Gender 0.26 (0.39) <.01 0.37 (0.39) <.01 −0.04 (0.27) <.01 0.06 (0.27) <.01
 MS × Gender −0.80 (0.72) <.01 −0.72 (0.73) <.01 0.13 (0.50) <.01 0.22 (0.52) <.01
 PS × Gender 0.20 (0.89) <.01 −0.06 (0.91) <.01 −0.23 (0.62) <.01 −0.52 (0.64) <.01
 EA × Age × Gender −0.02 (0.05) <.01 −0.03 (0.05) <.01 0.03 (0.03) <.01 0.02 (0.03) <.01
 MS × Age × Gender 0.09 (0.11) <.01 0.08 (0.11) <.01 −0.06 (0.07) <.01 −0.06 (0.07) <.01
 PS × Age × Gender −0.02 (0.11) <.01 0.02 (0.11) <.01 0.06 (0.08) <.01 0.10 (0.08) <.01
Variance
 Within-Person 6.42 (0.41)*** 6.68 (0.43)*** 3.05 (0.20)*** 3.34 (0.22)***
 Intercept 16.40 (0.96)*** 14.90 (0.91)*** 8.93 (0.52)*** 7.57 (0.47)***
 Slope 0.01 (<0.01)*** 0.01 (<0.01)** 0.01 (<0.01)*** 0.01 (<0.01)**
Fit Indices
 AIC 11032.3 10992.9 9739.5 9671.0
 BIC 11143.9 11110.0 9851.0 9788.1
Open in a new tab

Note. EA=Educational Attainment; MS=Marital Status; PS=Parental Status;

†

p = .05,

*

p < .05,

**

p < .01,

***

p < .001

Drive for Thinness Intervening Model.

We found a significant main effect of gender; women endorsed higher Drive for Thinness than men at all ages (t(1694.72) = 7.96, p <.001). There was also a significant gender by age interaction (t(1603.94) = −2.05, p =.04); women experienced declines in Drive for Thinness over time (b = −0.04, p = .01), while men did not (b = 0.02, p = .37).

There was a significant interaction between age and educational attainment (t(1441.83) = 4.83, p <.001). Those who obtained a graduate degree experienced increasing Drive for Thinness with age (b = 0.04, p = .007), and those who did not obtain a graduate degree experienced decreasing Drive for Thinness (b = −0.08, p < .001). Further, there was a significant interaction between gender and educational attainment (t(1595.69) = −2.02, p = .04). Women who obtained a graduate degree reported a decrease in their Drive for Thinness at the subsequent assessment (b = −0.49, p = .02) not observed in men (b = 0.30, p = .36). There was also a significant three-way interaction for age, gender, and educational attainment (t(1352.81) = 3.78, p < .001; see Figure 1). While women who obtained a graduate degree experienced an initial drop in Drive for Thinness, they experienced subsequent increases in Drive for Thinness over time (b = 0.05, p = .02). In contrast, women who did not obtain a graduate degree experienced decreasing Drive for Thinness across adulthood (b = −0.14, p < .001). For men, obtaining a graduate degree did not significantly impact the trajectory of Drive for Thinness (t(1421.66) = −0.01, p = .60).

Figure 1.

Figure 1.

Open in a new tab

Depiction of the three-way interaction between educational attainment, gender and age in the intervening model predicting Drive for Thinness. For women (left), those who obtained a graduate degree experienced an initial drop in Drive for Thinness but over time experienced subsequent increases in Drive for Thinness. Women who did not obtain a graduate degree generally experienced decreasing Drive for Thinness across adulthood. For men (right), obtaining a graduate degree did not significantly impact the trajectory of Drive for Thinness.

Parental status produced a significant main effect; when a participant became a parent, their Drive for Thinness dropped at the subsequent assessment (t(1626.01) = −3.15, p = .002). Furthermore, there was a significant interaction between age and parental status (t(1978.30) = 2.51, p =.01). Among those who never became parents, Drive for Thinness decreased significantly over time (b = −0.05, p = .01). In contrast, those who became parents showed no subsequent declines with age (b = 0.03, p = .19).

With BMI as a covariate, there was a significant effect of age (t(1635.27 = −4.39, p <.001) indicating that Drive for Thinness decreased over time. BMI also demonstrated a main effect (t(1847.67) = 10.15, p <.001). Higher BMIs were associated with higher Drive for Thinness over follow-up. All other results were comparable between models with and without BMI, including the gender by age interaction at the threshold for significance (t(1606.04) = −1.95, p = .05).

Bulimia Intervening Model.

There was a significant main effect of gender; women had higher Bulimia scores than men across adulthood (t(1664.47) = 4.62, p <.001). We also found a significant main effect of age; Bulimia scores significantly declined over time (t(1657.31) = −5.88, p <.001). There was also a significant gender by age interaction (t(1673.55) = −2.65, p = .008). Women experienced significant declines in Bulimia scores over time (b = −0.07, p < .001), not observed in men (b = −0.02, p = .37).

Regarding life roles, there was a significant main effect of educational attainment; participants who obtained graduate degrees experienced a drop in their Bulimia scores at the subsequent assessment (t(1669.96) = −3.76, p <.001). There was also a significant interaction between age and educational attainment (t(1487.15) = 4.33, p < .001). Among those who never obtained a graduate degree, Bulimia scores decreased significantly over time (b = −0.10, p < .001), whereas no further declines were observed among those with graduate degrees (b = −0.02, p = .16). Finally, there was a significant three-way interaction for gender × degree × age for Bulimia (t(1388.93) = 2.84, p = .005; see Figure 2). Women without graduate degrees experienced significant declines in Bulimia across adulthood (b = −0.13, p < .001) not observed in women with graduate degrees (b = −0.02, p < .222). In contrast, for men, obtaining a graduate degree did not impact Bulimia across adulthood (t(1318.38) = −0.02, p = .99).

Figure 2.

Figure 2.

Open in a new tab

Depiction of the three-way interaction between educational attainment, gender, and age in the intervening model predicting Bulimia scores. For women (left), those who did not obtain a graduate degree experienced significant declines in Bulimia across adulthood not observed in women with graduate degrees. For men (right), obtaining a graduate degree did not significantly impact the trajectory of Bulimia scores.

There was a significant interaction between age and parental status (t(1997.80) = 2.97, p = .007). Participants who never became parents experienced declines in Bulimia scores across adulthood (b = −0.08, p < .001), not observed in those who became parents (b = −0.02, p = .15).

In the model covarying for BMI, higher BMI was associated with higher Bulimia scores (t(1696.51) = 18.30, p <.001). Otherwise, the pattern of results was comparable. When covarying for BMI, the three-way interaction for gender, degree, and age was at the level of significance (t(1370.67) = 1.96, p = .05).

Drive for Thinness Time-Lagged Model.

There were significant main effects of age (t(1223.51) = 2.08, p = .04) and gender (t(1576.29) = 6.03, p < .001), such that Drive for Thinness increased with age, and women had higher Drive for Thinness than men across adulthood. This change in the effect of age reflects eliminating baseline Drive for Thinness scores from the time-lagged model and suggests that the drop in scores after college largely accounts for the effect of age in intervening models.

There was a marginally significant age by educational status interaction (t(1107.67) = 1.76, p = .079), such that those who obtained graduate degrees subsequently experienced increases in Drive for Thinness (b = 0.09, p = .02), not observed in those without graduate degrees (b = 0.01, p = .76), consistent with patterns observed from intervening models.

In the model covarying for BMI, results were comparable except that there was no longer a significant effect of age (t(1245.70) = 0.70, p = .49), but there was a significant effect of BMI (t(1554.93) = 6.65, p <.001), such that higher BMI prospectively predicted higher Drive for Thinness.

Bulimia Time-Lagged Model.

There was a significant main effect of gender, such that women had higher Bulimia scores than men across adulthood (t(1540.76) = 3.44, p = .001).

Additionally, there was a significant main effect of educational attainment, such that obtaining a graduate degree prospectively predicted subsequent decreases in Bulimia scores (t(1165.78) = −2.10, p = .04).

In the model covarying for BMI, higher BMI prospectively predicted higher Bulimia scores (t(1445.81) = 8.97, p <.001). All other results were comparable, including the effect of educational attainment at the level of significance (t(1156.43) = −1.96, p = .05).

Discussion

The present study examined how changing life roles predicted changes in eating pathology across adulthood and whether the effects differed between women and men. Educational attainment did not differ between women and men, consistent with recent data on educational attainment in the United States (Statista, 2019). However, the effects of higher education on eating pathology differed between genders. Earning a graduate degree predicted an initial drop in Drive for Thinness only for women. Women with graduate degrees demonstrated age-related increases in Drive for Thinness whereas women who never obtained a graduate degree experienced steady declines with age. Men were more likely than women to be married by age 40 and more likely than women to have children by age 50. The hypothesis that marriage would predict decreased eating pathology only in women was not supported because marital status did not predict changes in eating pathology for women or men. Becoming a parent predicted a subsequent decrease in Drive for Thinness for both women and men, partially supporting hypotheses. However, those who never had children experienced decreases in Drive for Thinness and Bulimia with age, not observed in parents. Patterns reveal novel associations between changes unique to adult development and eating pathology that merit attention in theoretical models of risk and intervention.

Consistent with prior literature (e.g., Heatherton et al., 1997; Keel et al., 2007), women had higher Drive for Thinness and Bulimia than men. Additionally, Drive for Thinness and Bulimia decreased in women but not men in intervening models. This may reflect a statistical artifact, given that women’s disordered eating had more room to decline after college than men’s. Alternatively, findings could reflect greater maintenance of eating pathology among men who endorsed problems in college (Brown et al., 2020), highlighting the importance of examining gender differences across adult developmental stages as suggested by Mehta et al. (2020). Shifting from predominantly same-gender peer groups to more mixed gender peer groups may differentially affect women and men’s disordered eating. Consistent with findings from Gravener et al. (2008), as women age, they may benefit from leaving peer groups dominated by other women. Conversely, as men join peer groups with more females, their increased exposure to peers with higher eating pathology may increase their risk. Of interest, in time-lagged models, Drive for Thinness increased with age and this effect was due to age-related increases in BMI. This suggests that both women and men experience increasing concerns about weight and restricting food intake as they age, especially as they experience increased weight gain (Stenholm et al., 2015).

Gender differences in the impact of educational attainment may reflect the unique effects of the “Career-and-Care-Crunch” for women (Mehta et al., 2020). Intervening models indicated that women experienced initial decreases in Drive for Thinness after obtaining their graduate degrees (with a nonsignificant decrease in Bulimia as well). Over time, however, women who obtained graduate degrees experienced increased Drive for Thinness and Bulimia as they aged. Women obtaining graduate training are more likely to delay getting married and having children which may temporarily delay stresses linked to balancing childcare with professional advancement (Mehta et al., 2020). Our models did not explore interactions among life roles, such as the interaction between gender, educational attainment, parental status, and age due to concerns about power. However, future studies may provide greater insight into how life roles combine to influence risk for eating pathology differentially for women and men.

Although men were more likely than women to be married by age 40, marital status constituted neither a risk nor protective factor against Drive for Thinness or Bulimia over the 30-year follow-up duration. Keel et al. (2007) found that, for women, marriage was associated with decreased Drive for Thinness and Bulimia by age 40. Prior research has also suggested that marital satisfaction contributes to weight gain in couples, such that dissatisfied partners are more likely to contemplate divorce and gain less weight (Meltzer et al., 2013). Our study only addressed marital status, not marital satisfaction. It is possible that the beneficial effects of marriage observed at 20-year follow-up (Keel et al., 2007) wane depending on the quality of partnerships. In addition, marital status was robustly associated with parental status, which may have absorbed some of the variance in disordered eating explained by a stable partnership in our multivariable model. This seems particularly likely because marital status demonstrated significant concurrent associations with lower Drive for Thinness and Bulimia scores in bivariate analyses at 20- and 30-year follow-up, consistent with prior findings.

For both men and women, becoming a parent predicted a subsequent decline in Drive for Thinness, consistent with findings for women from Keel et al. (2007). Societal expectations and biological drives both act as powerful motivators for parents to care for a child (Bjorklund & Yunger, 2001). For example, previous research has shown that parents work to eat healthier to be good role models to their children (Palfreyman et al., 2013). Parents also gain another domain in which to evaluate themselves – their parenting abilities (Coleman & Karraker, 2008). Thus, overvaluation of weight and shape may be displaced by concerns about parenting ability and children’s well-being, especially during the “Career-and-Care-Crunch” of established adulthood (Mehta et al., 2020). However, participants who never had children demonstrated significant declines in both Drive for Thinness and Bulimia with age not observed in parents. This could reflect “catching up” over time or the complex association between parenthood and well-being (Nelson, Kushlev, & Lyubomirsky, 2014).

Although intervening models generated several significant findings, fewer significant associations emerged from time-lagged models. The most notable finding reflected the sustained effect of educational attainment in predicting a subsequent decrease in Bulimia scores. This could reflect the same processes discussed for intervening models or could reflect other variables not included in models, such as likelihood of working outside of the home or income levels, both of which are linked to having advanced degrees (Carnevale et al., 2011).

Of note, effect sizes were very small (Cohen, 1992), suggesting that although changing life roles predicted changes in eating pathology, they account for a small portion of variance in these changes. As such, our results do not suggest that specific life roles such as obtaining a graduate degree or having children should be avoided or sought out, but rather individuals and their treatment providers should consider how life roles are associated with eating pathology. Importantly, the cumulative effect of life role changes was similar to the effect of gender in intervening models, reinforcing that small effect sizes may contribute to clinically significant outcomes (West, 2007).

The present study had several notable strengths. First, we followed participants from emerging adulthood into middle adulthood, representing the first longitudinal examination of this adult developmental stage without the confound of cohort effects. Additionally, including women and men in each model permitted the first study of gender differences on this topic. Further, we experienced low attrition in our sample, retaining 72% of women and 67% of men at 30-year follow-up. This increases confidence that findings are representative of the original sample. Our dependent variables demonstrated strong psychometric properties and are widely used in the literature, increasing the ability to compare findings across studies.

Despite these strengths, our study had limitations. Men represented approximately 30% of our sample, reflecting the sampling design of the parent study initiated in 1982 and gender differences in baseline participation. Future longitudinal studies should over-recruit men to offset their lower likelihood of participation. Although repeated assessments over 10-year intervals captured meaningful changes across adulthood, intervening models may not reflect how changes in life roles predicted changes in disordered eating. For example, apparent benefits of becoming a parent may represent improvements that came before having a child and increased the likelihood of becoming a parent rather than consequences of parenthood. Our time-lagged models address this limitation by establishing temporal precedence. However, time-lagged models do not capture recent changes in life roles. For example, if an individual had no children between college and 10-year follow-up and then became a parent between 10- and 20-year follow-up, the time-lagged models would not capture how this change in life role or others (e.g., divorce) predict eating pathology at 20-year follow-up. Thus, we conducted both intervening and time-lagged analyses because the strengths of one offset the limitations of the other. Relatedly, like any longitudinal study, we cannot infer causation from any of our findings because we cannot rule out the potential influence of underlying third variables. Of note, while we included educational attainment in our models as a predictor of eating pathology, unmeasured third variables such as socioeconomic status (Titus, 2006) or treatment access (Fletcher & Frisvold, 2012) may explain this relationship. Similarly, many of our female participants may have experienced unmeasured hormonal changes due to menopause at 30-year follow-up. Hormonal factors or their consequences may account for some of our findings (Baker & Runfola, 2016; Thompson & Bardone-Cone, 2019). Moreover, the majority of our sample obtained at least an undergraduate degree and results may not generalize to those of differing educational backgrounds. Additionally, generalizability to racial and ethnic minority groups is limited by the predominantly White, non-Hispanic make-up of our sample and the reduced likelihood of participation at 30-year follow-up among racial/ethnic minority participants. Those who remained single at 10-year follow-up were less likely to participate at 30-year follow-up, which may bias results for marital status. To minimize method variance in our longitudinal design, survey items regarding eating pathology and life roles were maintained from prior surveys. Previous research has shown that men’s eating pathology may present as drive for muscularity (Lavender et al., 2017), and our models may not fully capture male participants’ disordered eating given the reliance on Drive for Thinness and Bulimia scores. Unfortunately, the original question concerning marital status (Heatherton et al., 1997) did not probe for same-sex partnerships, and legal changes in marriage laws may have changed the meaning of this item over the course of follow-up for participants. Furthermore, we did not include occupational status as a life role in the present study because the original item did not distinguish between current and past employment. Greater information on who is working outside of the home and when those changes occur could shed additional light on the course of eating pathology across adult development in women and men related to the “Career-and-Care-Crunch” (Mehta et al., 2020).

Conclusions

Disordered eating across adult development represents an important focus for future research. Our results have implications for theories of ED risk and maintenance. Specifically, present findings implicate identity and social environment as potential risk and protective factors for eating pathology. Each of the life roles measured reflects changes in how women and men may define themselves in the context of new social environments and responsibilities to the self and others. In addition, each role introduces potential sources of stress. Future work may capitalize on current findings to explore what aspects of these changes are helpful and which aspects may be harmful. Greater understanding could advance interventions to reduce ED risk across adult development.

Funding:

This work was supported by grants from the Milton Fund and National Institute of Mental Health (R01MH63758, PI: Keel). We thank the Henry Murray Center of the Radcliffe Institute for providing access to data from Todd Heatherton’s Follow-up and Replication of Prevalence among College Students and the Alumni Office of Harvard University for supplying addresses for participants at follow-up.

Footnotes

Research ethics committee approval: This study, “Bulimic Syndromes: Secular & Longitudinal Trends II” was approved by the Florida State University Institutional Review Board.

References

  1. American Psychiatric Association. (1980). Diagnostic and statistical manual of mental disorders (3rd ed.). Washington, DC: Author. [Google Scholar]
  2. Baker JH & Runfola CD (2016). Eating disorders in midlife women: A perimenopausal eating disorder? Maturitas, 85, 112–116. 10.1016/j.maturitas.2015.12.017. [DOI] [PubMed] [Google Scholar]
  3. Borjorquez I, Bustos J, Valdez V, & Unikel C (2018). Life course, sociocultural factors and disordered eating in adult Mexican women. Appetite, 121, 207–214. 10.1016/j.appet.2017.11.085. [DOI] [PubMed] [Google Scholar]
  4. Brown TA, Forney KJ, Klein KM, Grillot C, & Keel PK (2020). A 30-year Longitudinal Study of Body Weight, Dieting, and Eating Pathology Across Women and Men From Late Adolescence to Later Midlife. Journal of Abnormal Psychology, 129, 376–386. 10.1037/abn0000519. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Carnevale A, Strohl J & Melton M (2011). What’s it worth? The economic value of college majors. Georgetown University, Center on Education and the Workforce. [Google Scholar]
  6. Cohen J (1992). A power primer. Psychological Bulletin, 112, 155–159. [DOI] [PubMed] [Google Scholar]
  7. Coleman PK & Karraker KH (2008). Parenting self-efficacy among mothers of school-age children: Conceptualization, measurement, and correlates. Family Relations, 49, 13–24. 10.1111/j.1741-3729.2000.00013.x. [DOI] [Google Scholar]
  8. Copeland L (2014). The 4 ways men and women connect in middle age. Retrieved from https://www.huffpost.com/entry/dating-after-50_b_5647768.
  9. Eisenberg ME, Berge JM, & Neumark-Sztainer D (2013). Dieting and encouragement to diet by significant others: Associations with disordered eating in young adults. American Journal of Health Promotion, 27, 370–377. 10.4278/ajhp.120120-QUAN-57. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Flegal KM, Kruszon-Moran D, Carroll MD, Fryar CD, & Ogden CL (2016). Trends in obesity among adults in the United States, 2005 to 2014. The Journal of the American Medical Association, 315, 2284–2291. 10.1001/jama.2016.6458. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Fletcher JM & Frisvold DE (2012). Higher education and health investments: Does more schooling affect preventative health care use? Journal of Human Capital, 3, 144–176. 10.1086/645090. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Garner DM, Olmstead MP, & Polivy J (1983). Development and validation of a multidimensional eating disorder inventory for anorexia nervosa and bulimia. International Journal of Eating Disorders, 2, 15–34. . [DOI] [Google Scholar]
  13. Gravener JA, Haedt AA, Heatherton TF, & Keel PK (2008). Gender and age differences in associations between peer dieting and drive for thinness. International Journal of Eating Disorders, 41, 57–63. 10.1002/eat.20438. [DOI] [PubMed] [Google Scholar]
  14. Heatherton TF, Mahamedi F, Streipe M, Field AE, & Keel PK (1997). A 10-year longitudinal study of body weight, dieting, and eating disorder symptoms. Journal of Abnormal Psychology, 106, 117–125. 10.1037/0021-843X.1061.117. [DOI] [PubMed] [Google Scholar]
  15. Herpertz-Dahlmann B, Dempfle A, Konrad K, Klasen F, & Ravens-Sieberer U (2015). Eating disorder symptoms do not just disappear: The implications of adolescent eating-disordered behaviour for body weight and mental health in young adulthood. European Child & Adolescent Psychiatry, 24, 675–684. 10.1007/s00787-014-0610-3. [DOI] [PubMed] [Google Scholar]
  16. Keel PK, Baxter MG, Heatherton TF, Joiner TE Jr. (2007). A 20-year longitudinal study of body weight, dieting, and eating disorder symptoms. Journal of Abnormal Psychology, 116, 422–432. 10.1037/0021-843X.116.2.422. [DOI] [PubMed] [Google Scholar]
  17. Lavender JM, Brown TA, & Murray SB (2017). Men, muscles, and eating disorders: An overview of traditional and muscularity-oriented disordered eating. Current Psychiatry Reports, 19, 32. 10.1007/s11920-017-0787-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Levinson JA, Sarda V, Sonneville K, Calzo JP, Ambwani S, & Austin SB (2020). Diet pill and laxative use for weight control and subsequent incident eating disorder in US young women: 2001–2016. American Journal of Public Health, 110, 109–111. 10.2105/AJPH.2019.305390. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Mangweth-Matzek B & Hoek HW (2017). Epidemiology and treatment of eating disorders in men and women of middle and older age. Current Opinions in Psychiatry, 30, 446–451. 10.1097/YCO.0000000000000356. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Mangweth-Matzek B, Hoek HW, Rupp CI, Lackner-Seifert K, Frey N, Whitworth AB, et al. (2014). Prevalence of eating disorders in middle-aged women. International Journal of Eating Disorders, 47, 320–325. 10.1002/eat.22232. [DOI] [PubMed] [Google Scholar]
  21. Martin JA, Hamilton BE, Osterman MJK, Driscoll AK, & Drake P (2018). Births: Final data for 2016. National Vital Statistics Reports, 67, 1–54. [PubMed] [Google Scholar]
  22. Mehta CM, Arnett JJ, Palmer CG, & Nelson LJ (2020). Established adulthood: A new conception of ages 30 to 45. American Psychology, 75, 431–444. 10.1037/amp0000600. [DOI] [PubMed] [Google Scholar]
  23. Meltzer AL, Novak SA, McNulty JK, Butler EA, & Karney BR (2013). Marital satisfaction predicts weight gain in early marriage. Health Psychology, 32, 824–827. 10.1037/a0031593. [DOI] [PubMed] [Google Scholar]
  24. Mitchison D & Hay PJ (2014). The epidemiology of eating disorders: Genetic, environmental, and societal factors. Clinical Epidemiology, 6, 89–97. 10.2147/CLEP.S40841. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Montenegro XP (2003). Lifestyles, dating, and romance: A study of midlife singles. Retrieved from https://assets.aarp.org/rgcenter/general/singles.pdf.
  26. Nelson SK, Kushlev K, & Lyubomirsky S (2014). The pains and pleasures of parenting: when, why, and how is parenthood associated with more or less well-being? Psychological Bulletin, 140(3), 846–895. 10.1037/a0035444 [DOI] [PubMed] [Google Scholar]
  27. Palfreyman Z, Haycraft E, & Meyer C (2013). Unintentional role models: Links between maternal eating psychopathology and the modelling of eating behaviours. European Eating Disorders Review, 21, 195–201. 10.1002/erv.2219. [DOI] [PubMed] [Google Scholar]
  28. Pearson CM, Wonderlich SA, & Smith GT (2015). A risk and maintenance model for bulimia nervosa: From impulsive action to compulsive behavior. Psychological Review, 122, 516–535. 10.1037/a0039268. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Schafer JL, & Graham JW (2002). Missing data: Our view of the state of the art. Psychological Methods, 7, 147–177. 10.1037/1082-989X.7.2.147. [DOI] [PubMed] [Google Scholar]
  30. Statista. (2019). Percentage of the U.S. population with a college degree 1940–2018, by gender. Retrieved from https://www.statista.com/statistics/184272/educational-attainment-of-college-diploma-or-higher-by-gender/.
  31. Stenholm S, Vahtera J, Kawachi I, Pentti J, Halonen JI, Westerlund H, … Kivimäki M (2015). Patterns of weight gain in middle-aged and older US adults, 1992–2010. Epidemiology, 26(2), 165–168. 10.1097/ede.0000000000000228 [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Thompson KA & Bardone-Cone AM (2019). Menopausal status and disordered eating and body image concerns among middle-aged women. International Journal of Eating Disorders, 52, 314–318. 10.1002/eat.23030. [DOI] [PubMed] [Google Scholar]
  33. Thompson KJ & Cafri G (Eds.). (2007). The muscular ideal: Psychological, social, and medical perspectives. American Psychological Association. [Google Scholar]
  34. Titus MA (2006). Understanding college degree completion of students with low socioeconomic status: The influence of the institutional financial context. Research in Higher Education, 47, 371–398. 10.1007/s11162-005-9000-5. [DOI] [Google Scholar]
  35. U.S. Census Bureau (2017). Estimated median age at first marriage: 1890 to present. Retrieved from https://www.census.gov/data/tables/time-series/demo/families/marital.html.
  36. Vogeltanz-Holm ND, Wonderlich SA, Lewis BA, Wilsnack SC, Harris TR, Wilsnack RW, & Kristjanson AF (2000). Longitudinal predictors of binge eating, intense dieting, and weight concerns in a national sample of women. Behavior Therapy, 31, 221–235. 10.1016/S0005-7894(00)80013-1. [DOI] [Google Scholar]
  37. West R (2007). The clinical significance of ‘small’ effects of smoking cessation treatments. Addiction, 102, 506–509. 10.1111/j.1360-0443.2007.01750.x [DOI] [PubMed] [Google Scholar]

RESOURCES