Longitudinal Analysis of Psychosocial Stressors and Body Mass Index in Middle-Aged and Older Adults in the United States

Abstract

Objectives

Psychosocial stress may be a risk factor for obesity and overweight in middle-aged and older adults. However, research on psychosocial stress and excess body weight has typically been cross-sectional and focused on single stressors.

Methods

Using 3 waves of data from the Health and Retirement Study, we conducted longitudinal analyses to assess associations between 5 psychosocial stressors—individually and in combination—and body mass index (BMI), adjusting for sociodemographic factors, alcohol use, and smoking history. We tested interaction effects between race and gender with stressors on BMI.

Results

A total of 3,956 participants were included in the main analyses. Most participants were White (88.04%) and more than half were female (60.39%). Perceived discrimination, financial stress, and relationship stress were positively associated with BMI. A greater cumulative stress burden was associated with higher BMI. In stratified analyses, greater financial stress was associated with higher BMI among White participants, whereas greater neighborhood stress was associated with lower BMI among Black participants. Greater relationship stress, financial stress, cumulative high stress, and overall cumulative stress burden were associated with higher BMI for women, but not men.

Discussion

Different sources of stress may be risk factors to weight gain and affect BMI in adults. White and female adults may be more susceptible to the obesogenic effects of stressors. Reducing exposure to stress may help reduce the burden of high BMI among middle-aged and older adults.

The obesity epidemic is a major public health issue in the United States. Individuals with overweight or obesity (i.e., body mass index [BMI] between 25.0 and 29.9 kg/m² and a BMI of 30 kg/m² or more, respectively) are at increased risk for many of the leading causes of morbidity and mortality in the United States relative to individuals with a healthy weight (Vucenik & Stains, 2012; Xu et al., 2018). Since 2005, rates of obesity have continued to increase across the United States among all racial/ethnic groups, but Black and Latino adults continue to have higher rates of obesity than White adults (Hales et al., 2018). The prevalence of high BMI is most pronounced among middle-aged and older adults, with 43% of middle-aged adults and 41% of older adults having obesity, thus placing them at a higher risk for metabolic and cardiovascular disease and cancer.

There is compelling evidence that psychological and social stress (psychosocial stress) are risk factors for obesity (Cuevas et al., 2019; Wardle et al., 2011). For example, psychological distress due to stress exposure can increase cravings for and excess eating of foods high in fat and sugar (Scott et al., 2012). Psychosocial stress can also have a negative effect on physical activity outcomes (Stults-Kolehmainen & Sinha, 2014). A meta-analysis of 14 longitudinal studies concluded that indicators of psychosocial stress including acute stressful periods and job stress are risk factors for weight gain (Wardle et al., 2011). While most studies included in the meta-analysis were limited to exploring a small number of individual stressors such as childhood adversity, job stress (or job demands), and caregiver stress, emerging research suggests that other sources of stress may be risk factors for increase in BMI. Financial hardship (e.g., difficulty paying bills), relationship problems, and neighborhood stress have been found to be associated with subsequent weight gain among adults (Conklin et al., 2014; Kouvonen et al., 2011; Siahpush et al., 2014). Yet, despite the possible obesogenic influence of a broad range of stressors, existing research examining the contribution of psychosocial stressors to increased obesity risk is limited. Stress is multidimensional and multiple psychosocial stressors often coexist in individuals’ lives (Wardle et al., 2011).

Examining multiple stressors simultaneously provides insight into the different ways in which stress can increase BMI and potential mechanisms by which stress can increase BMI. For instance, poor marital quality is associated with health-compromising behaviors, such as poor diet and sedentary behaviors (Robles et al., 2014). The co-occurrence of stressors can have additive effects on BMI due to the multiple pathways by which stressors can increase BMI (e.g., poor sleep, physical inactivity, poor mental health, etc.). The cumulative stress hypothesis posits that the more stress exposure a person experiences, the greater the likelihood that they will succumb to a disease (Nederhof & Schmidt, 2012). Previous studies have examined the accumulation of stressor exposure by totaling the number of stressor scores in the top quintile of the sample distribution. Compared to individuals who were not in the top quintile for any of the stressor domains, individuals with two or more high stressors typically had significantly shorter sleep durations, higher depressive symptoms, poorer self-rated health, and higher odds of having obesity (Cuevas et al., 2019; Slopen & Williams, 2014; Sternthal et al., 2011).

Differences in levels of exposure to a variety of stressors by race and gender may play a role in the relationship between stress and obesity. For instance, Black adults report greater exposure to discrimination, financial stress, acute life events, neighborhood stress, and relationship problems compared to their White counterparts (Sternthal et al., 2011). Women also experience greater exposure to stressors (e.g., employment discrimination, financial stress) compared to men (Mayor, 2015; Milner et al., 2019). The disproportionate burden of stress exposure can lead to a greater risk of obesity for Black individuals and women. Prior research also suggests that women and men, as well as Black and White adults, differ in their responses to stressors (Jackson et al., 2010; Mayor, 2015). Jackson and colleagues (2010) revealed that Black adults engage in greater obesity-related behaviors (e.g., overeating) as a means to cope with stressors. While these behaviors led to fewer depressive symptoms, they also led to a higher burden of chronic health conditions (Jackson et al., 2010). Laboratory studies illustrated that female participants are more likely to engage in stress-induced emotional eating than male participants (Devonport et al., 2019; van Strien, 2018). A recent cross-sectional study revealed that the association between cumulative stress and obesity is stronger for women than men (Cuevas et al., 2019). To date, a limited number of studies have examined the association between stressors and BMI. As much of this scholarship has been cross-sectional, a longitudinal examination between a variety of stressors and BMI is warranted.

The Health and Retirement Study (HRS) provides a unique opportunity to longitudinally assess the association between psychosocial stressors and BMI among a large and racially diverse sample of middle-aged and older adults. Using three waves of data collected 4 years apart (i.e., in total, over 8 years), the present study aims to quantify the longitudinal association between a wide range of psychosocial stressors and BMI among non-Hispanic Black and non-Hispanic White middle-aged and older adults. We also assessed the association between cumulative exposure to high stress and BMI. We hypothesized that higher levels of psychosocial stressors will be associated with higher BMI. These associations will be stronger for non-Hispanic Black participants as compared to non-Hispanic White participants and women as compared to men.

Method

Sample

The HRS is a longitudinal epidemiological cohort study of a nationally representative sample of individuals 50 years and older living in the United States. The original HRS core sample design was a multistage area probability sample of households which began in 1992 with oversamples of Black and Latino adults, and residents of Florida. Subsequently, participants took part in a biennial interview that covers a range of topics including socioeconomic status, work, retirement, health, and health care utilization. HRS employs a steady-state design, replenishing the sample every 6 years with younger cohorts.

In 2006, HRS implemented an enhanced face-to-face interview that included a psychosocial questionnaire (Clarke et al., 2008) and biomarker assessment, including measured height and weight (Crimmins et al., 2008). Half of the HRS participants completed this interview in 2006 and were given follow-up interviews 4 and 8 years later. We used the 2006 assessments as baseline measures in our analysis, and the 2010 and 2014 assessments as follow-up measures. The study was reviewed and approved by the University of Michigan’s Health Sciences Institutional Review Board as well as Tufts University’s Institutional Review Board.

Analytic Sample

A total of 278 respondents (5% of baseline responses) did not complete the psychosocial questionnaire in 2006 and therefore, were excluded from the sample. There were no differences in race, gender, education, obesity, or stress exposure between included and excluded individuals at p-value < .05. In the 2006 assessments, a total of 5,620 participants completed all of the measures of interest; 4,238 of these initial respondents completed the 2010 and 2014 assessments. Those who did not identify as non-Hispanic White or non-Hispanic Black were excluded from the proposed study due to small sample sizes (n = 137 and 145, respectively), resulting in an analytic sample of 3,956 participants across all three waves.

Dependent Variables

Study interviewers measured respondents’ height and weight. BMI is calculated as weight (kg) divided by height (m) squared (Shah & Braverman, 2012). BMI data were collected in the three waves of this study (i.e., 2006, 2010, and 2014).

Psychosocial Stressors

HRS collected data on a wide range of stressors. We examined five domains of psychosocial stressors (acute life events, financial stress, neighborhood stress, relationship stress, and life discrimination) and constructed a measure of cumulative exposure across these stress domains. While other stressors were measured in the initial 2006 assessment (work discrimination and job stress), subsequent waves did not include data on these stressors. The individual domains and cumulative measure were constructed for all three points of assessment. Each domain included one or more measures, as described below (see Supplementary Appendix A for further detail). For domains that included multiple measures, we transformed each measure into a z-score and summed them together. We then restandardized the sum score into a z-score to allow for comparisons across domains (Slopen & Williams, 2014; Sternthal et al., 2011).

Acute life events included two major life event inventories: (a) acute life events over the life span (seven items), such as death of a loved one and experience of life-threatening illness; and (b) acute life events in the past 5 years (six items), such as being a victim of fraud and experiencing involuntary job loss. Financial stress included two measures: (a) financial distress (two items), and (b) lack of financial autonomy (two items). Neighborhood stress included four items about neighborhood safety, vandalism, and community disorder (i.e., cleanliness and deserted houses). Relationship stress included four measures related to demands and criticism from others: (a) spouses (four items); (b) children (four items); (c) other family members (four items); and (d) friends (four items). Lifetime discrimination measured racial and nonracial discrimination using questions from an inventory of major discriminatory events (six items) and a shortened version of the Everyday Discrimination Scale (five items; Williams et al., 1997).

For each point of assessment, a cumulative high-stress score was created to identify individuals experiencing high levels of stress across multiple domains. Consistent with previous research (Boen, 2019; Slopen & Williams, 2014; Sternthal et al., 2011), each stressor domain was dichotomized to contrast the top quintile versus others. Focusing on the top quintile allows us to capture both severity and accumulation of stressors (Williams & Mohammed, 2009). The cumulative high-stress score reflects the number of domains in which the individual was in the top quintile of stress exposure. The score ranged from 0 to 5, with a higher score reflecting higher cumulative high-stress exposure. We also created a composite score of cumulative stress burden using the individual stress domains. We summed the standardized scores without using any cutoffs, and then standardized the summed score. The cumulative stress burden score reflects the total stress exposure across all the stress domains in this study.

Covariates: Sociodemographics and Health Behaviors

We included the following sociodemographic covariates in the adjusted regression models: gender (male or female), race (Black or White), education (e.g., less than high school, General Educational Development, high school graduate, or some college), and having any children (yes or no). We also included four time-varying sociodemographic covariates, age (in years), marital/partner status (e.g., married, partnered, or widowed), employment status (e.g., works full-time, works part-time, partly retired, or disabled), and household income. We included two time-varying health behaviors—smoking (current, former, or never) and alcohol consumption (does or does not drink alcohol)—as covariates as they are potential confounders of the relationship between stress and obesity. Individuals who experience stress are more likely to smoke and consume alcohol; these behaviors are associated with lower BMI (Kassel et al., 2003; Kaufman et al., 2012; Traversy & Chaput, 2015).

Last, given that childhood adversity is associated with both greater stress exposure in adulthood and higher BMI (Albott et al., 2018; Danese & Tan, 2014), we adjusted for childhood adversity. Childhood adversity was assessed at baseline retrospectively using a count of six events before the age of 18 (i.e., experiencing a difficult living arrangement, exposure to frequent parental substance use, being in trouble with the police, being physically abused by a parent, whether their parents divorced before they reached the age of 16, and whether one or both parents died before they reached the age of 16; Clarke et al., 2008).

Statistical Analysis

We examined associations between cumulative stress and individual stressor variables from the preceding assessments (2006 or 2010) in relation to future (2010 or 2014) BMI. To avoid potential multicollinearity between the stressors, we quantified the association between each stressor variable and BMI using separate models. Mixed effect models with unstructured covariance were used to account for the repeated measurements from the same individual in multiple years. The formula for testing the mixed effect models was $\mathrm{B}\mathrm{M}{\mathrm{I}}_{j,t}={\mathrm{\beta }}_0+{\mathrm{\beta }}_1{S}_{j,t-1}+{\mathrm{\beta }}_2{V}_{j,t}+{\mathrm{\mu }}_{j,t}+{\mathrm{\mu }}_t$, where S_j,t−1 indicated the level of stress from the prior wave, V_j,t indicated time-varying and time constant covariates, and μ _t represented the overall error term. We specified a random intercept by including μ _j,t for random error at person level. All models were adjusted for covariates, including age, gender, race, education, marital/partner status, having any children, and childhood stressors. We did not adjust for prior BMI as a lagged dependent variable in a mixed model would violate the assumption of independence from μ _t and lead to biased estimates for all variables in the model (Allison, 2015). Analyses were conducted using Stata software (StataCorp, 2013).

To test our hypothesis that psychosocial stress may display a stronger association with weight gain among Black participants in comparison to White participants, we tested for effect modification by race/ethnicity using a race/ethnicity by stress interaction term for all stressors. We also tested gender as an effect modifier using the same approach. We performed stratified analyses if interactions at p < .05 were detected.

Results

A total of 3,956 participants were included in the main analyses, of whom 3,482 (88.04%) were White, and 473 (11.96%) were Black. More than half of participants were female (60.39%). The average age at baseline was 65.21 (SE = 9.20) years old. Approximately half of the participants had some college experience (50.35%). Table 1 presents the distribution of sociodemographic characteristics by year of interview (2006, 2010, and 2014). Mean stressor z-scores across the 3 years are presented in Table 2.

Table 1.

Sample Characteristics (N = 3,956) Across the Three Waves

	2006		2010		2014
	Mean	SE	Mean	SE	Mean	SE
Early life adversity	−0.0043	1.00	—	—	—	—
Age	65.21	9.20	69.42	9.26	73.15	9.21
Household income	10.76	1.10	10.66	1.17	10.69	1.07
Body mass index	29.10	5.61	29.46	5.89	29.33	6.05
	n	%	n	%	n	%
Race
White	3,482	88.04	—	—	—	—
Black	473	11.96	—	—	—	—
Gender
Male	1,567	39.61	—	—	—	—
Female	2,389	60.39	—	—	—	—
Education
<HS, no degree	453	11.45	—	—	—	—
GED	182	4.60	—	—	—	—
HS graduate	1,329	33.61	—	—	—	—
Some college, HS/GED	762	19.26	—	—	—	—
Some college, AA/<BA	193	4.88	—	—	—	—
College BA	600	15.17	—	—	—	—
MA/MBA	343	8.67	—	—	—	—
JD/MD/PhD	93	2.35	—	—	—	—
Parental status
Has kids	3,516	88.98	—	—	—	—
No kids	435	11.02	—	—	—	—
Marriage status
Married	2,768	69.97	2,567	64.91	2,355	59.53
Married, spouse absent	6	0.15	29	0.73	24	0.61
Partnered	139	3.51	118	2.98	108	2.73
Separated/divorced	400	10.11	408	10.32	397	10.04
Widowed	546	13.80	726	18.36	968	24.47
Never married	97	2.45	107	2.71	104	2.63
Employment status
Works full-time	1,119	28.29	767	19.39	488	12.34
Works part-time	230	5.81	173	4.37	98	2.48
Unemployed	60	1.52	80	2.02	32	0.81
Partly retired	444	11.22	419	10.59	393	9.93
Retired	1,820	46.01	2,348	59.37	2,830	71.54
Disabled	47	1.19	28	0.71	18	0.46
Not in labor force	236	5.97	140	3.54	97	2.45
Smoke
Never smoker	1,819	46.28	1,818	46.27	1,819	46.30
Former smoker	1,638	41.68	1,705	43.40	1,755	44.67
Current smoker	473	12.04	406	10.33	355	9.04
Drink
Does not drink alcohol	1,688	42.68	1,733	43.82	1,867	47.19
Drink alcohol	2,267	57.32	2,222	56.18	2,089	52.81
Body mass index categories
Underweight	29	0.74	44	1.12	62	1.58
Normal weight	1,041	26.62	1,048	26.72	1,120	28.51
Overweight	1,534	39.22	1,468	37.43	1,449	36.89
Obesity	1,307	33.43	1,362	34.73	1,297	33.02

Notes: Income variable reflects the log of total household income in dollars. GED = General Educational Development; HS = high school.

Table 2.

Distribution of Each Stress Domain for the Full Sample Across Three Waves (N = 3,956)

	2006		2010		2014
Stressors	Mean	SE	Mean	SE	Mean	SE
Acute stress	−0.01	0.99	−0.01	0.99	0.001	1.00
Discrimination stress	−0.01	0.98	−0.002	1.00	−0.01	0.97
Financial stress	−0.03	0.98	−0.02	1.00	−0.02	1.00
Neighborhood stress	−0.04	0.98	−0.04	0.97	−0.03	0.98
Relationship stress	−0.03	0.98	−0.02	0.98	−0.02	0.98
Cumulative high stress	0.87	1.01	0.88	1.03	0.89	1.03
Cumulative stress burden	−0.01	0.98	0.002	1.02	0.01	1.01

Discrimination, financial stress, relationship stress, cumulative high stress, and cumulative stress burden increased across the three waves (see Table 3). There was no evidence that neighborhood stress changed over time, whereas acute stress decreased between 2006 and 2014.

Table 3.

Change in Stressors Across the Three Waves (N = 3,956)

	Acute stress	Discrimination stress	Financial stress	Neighborhood stress	Relationship stress	Cumulative high stress	Cumulative stress burden
2006 (Ref)
2010	−0.03 (−0.07, 0.02)	0.07 (0.02, 0.11)	0.06 (0.02, 0.10)	0.002 (−0.04, 0.04)	0.05 (0.01, 0.09)	0.06 (0.02, 0.10)	0.05 (0.01, 0.09)
2014	−0.05 (−0.09, −0.001)	0.11 (0.06, 0.15)	0.13 (0.08, 0.17)	0.02 (−0.03, 0.06)	0.10 (0.05, 0.14)	0.12 (0.08, 0.17)	0.10 (0.06, 0.14)

Notes: All models controlled for age, education, gender, income, race, smoking, and drinking. Values in parentheses correspond to 95% CIs for each estimate.

Main Analyses

Greater discrimination (B = 0.12, 95% CI [0.02, 0.21], p = .02), financial stress (B = 0.16, 95% CI [0.06, 0.26], p < .01), and relationship stress (B = 0.20, 95% CI [0.10, 0.31], p < .01) were positively associated with BMI (Table 4). Greater cumulative stress burden was associated with higher BMI (B = 0.21, 95% CI [0.10, 0.32, p < .01). However, acute stress (B = 0.05, 95% CI [−0.03, 0.13], p = .24), neighborhood stress (B = −0.02, 95% CI [−0.11, 0.07], p = .66), and cumulative high stress were not associated with BMI (B = 0.08, 95% CI [−0.01, 0.17], p = .08). See Supplementary Materials for estimates of all variables, including sociodemographic factors and health behaviors, for each model.

Table 4.

Longitudinal Association Between Stressors and BMI (N = 3,956)

Stressors	B	CI (95%)
Acute stress	0.05	−0.03, 0.13
Discrimination stress	0.12	0.02, 0.21
Financial stress	0.16	0.06, 0.26
Neighborhood stress	−0.02	−0.11, 0.07
Relationship stress	0.20	0.10, 0.31
Cumulative high stress	0.08	−0.10, 0.17
Cumulative stress burden	0.21	0.10, 0.32

Notes: Separate models were tested to quantify the association between each stress and body mass index (BMI). All models controlled for age, education, gender, income, race, smoking, and drinking.

We tested two-way interactions between the stressors and race on BMI. Interaction terms between race and financial stress and neighborhood stress were significant. We conducted race-stratified analyses for the two individual stressors; greater financial stress was associated with higher BMI among White participants (B = 0.22, 95% CI [0.12, 0.32], p < .01), whereas greater neighborhood stress was associated with lower BMI among Black participants (B = −0.34, 95% CI [−0.61, −0.07], p = .02; see Supplementary Materials).

We also tested two-way interactions between the stressors and gender on BMI. Interaction terms between gender and relationship stress, financial stress, cumulative high stress, and cumulative stress burden were significant. Stratifying the analyses by gender revealed that greater relationship stress (B = 0.26, 95% CI [0.12, 0.40], p < .001), financial stress (B = 0.12, 95% CI [0.09, 0.36], p = .001), cumulative high stress (B = 0.12, 95% CI [0.002, 0.25], p = .047), and overall cumulative stress burden (B = 0.25, 95% CI [0.11, 0.39], p = .001) were associated with higher BMI among women, but not men.

Discussion

We examined the association between a range of stressors and BMI in middle-aged and older American adults. In the sample overall, greater exposure to discrimination, financial stress, and relationship stress were associated with higher BMI, and neighborhood stress and acute life events were not associated with BMI. We examined the association between high stress and BMI by dichotomizing each stressor domain to contrast the top quintile versus others. A cumulative high-stress score reflected the number of domains in which the individual was in the top quintile of stress exposure. We did not find an association between cumulative high stress across the five stressors and BMI. However, when we removed the cutoffs from the summation, we did find that greater cumulative stress burden was associated with higher BMI.

Our results are consistent with prior studies that showed greater exposure to discriminatory experiences to be associated with higher adiposity across various racial/ethnic and gender groups (Hunte, 2011; Hunte & Williams, 2009; Sutin & Terracciano, 2013; Thorpe et al., 2017), with discrimination being a known risk factor for many common diseases that disproportionately affect racial/ethnic minorities (Paradies et al., 2015). For instance, Hunte (2011) found in a predominantly White sample of men and women that those who reported high levels of interpersonal everyday discrimination had an increased waist circumference over a 9-year period. How discrimination increases adiposity remains unknown, but research suggests that discrimination may increase susceptibility to adopting obesity-related behaviors as a way to cope with the stressor (e.g., physical inactivity; Jackson & Steptoe, 2017). Greater attention to the potential pathways can help us better understand how to reduce the impact of discrimination on BMI.

Our findings also corroborate evidence of an association between greater financial stress and other measures of adiposity. Previous prospective studies, using European and Australian samples, have found associations between persistent financial hardship (e.g., difficulty paying bills) and weight gain, as well as an increase in BMI (Conklin et al., 2014). Financial stress can influence a person’s health behavior especially if they live in environments with limited access to healthy food. According to the scarcity theory (Mullainathan & Shafir, 2013), financial stress can decrease a person’s cognitive capacity to make healthy choices and trigger health-compromising behaviors (Beenackers et al., 2018). Living near fast food outlets or in environments with poor access to health-promoting resources can exacerbate the obesogenic effects of financial stress. Along with assessing the behavioral mechanisms, future research should examine how the food environment may play a role in the relationship between financial stress and obesity.

While findings on social relationships and weight change have been mixed, our study aligns with prior studies that suggest negative interpersonal relationships (e.g., adverse exchanges and conflict) are associated with greater adiposity (Kershaw et al., 2014). Negative relationships, such as poor marital quality, are predictive of short- and long-term poor health outcomes and can exacerbate the effects of other stressors on health (Robles et al., 2014). Persistent negative relationships can increase weight gain through obesity-related behaviors such as poor sleep and diet quality (Robles et al., 2014). Further research is needed to better understand how weight-related health behaviors can mediate and moderate the link between persistent negative social relationships and BMI gain.

It is important to note, however, when we did not impose cutoffs to the scoring, we did find that greater accumulation of stressors was associated with higher BMI. Imposing cutoffs may mischaracterize the association between cumulative stress and BMI due to the restricting of the variance and the considerable loss of power when dichotomizing variables (Deyi et al., 1998). Without the cutoffs, we found support for the cumulative stress hypothesis, which posits that the more stress exposure a person experiences over their lifetime, the greater risk they have for developing a disease (Nederhof & Schmidt, 2012; Sternthal et al., 2011). Thus, our findings suggest that cumulative exposure to stressors may still place middle-aged and older adults at increased risk for overweight and obesity. Future research should also examine the potential biobehavioral and psychological pathways that can explain the link between cumulative stress and BMI.

We found that greater neighborhood stress was associated with higher BMI for White participants, but lower BMI for Black participants. Survival bias may account for these findings as surviving Black participants may be able to cope more effectively with neighborhood stress than White participants. However, given the age of the study participants, lower BMI may also be a sign of frailty characterized by loss of bone density and muscle mass. Older adults living in high-stress neighborhoods (e.g., residential instability, presence of vandalism/graffiti, deserted houses) have higher frailty risks, such as muscle weakness and weight loss (Caldwell et al., 2019; Duchowny et al., 2020). Therefore, it is possible that greater neighborhood stress can lead to greater frailty and, in turn, lower BMI in Black participants. Nevertheless, given that we did not find similar patterns across the other stressors for Black participants, another possible explanation for our findings is that neighborhood stress has a distinct relationship with BMI due to its connection with context/place. Individuals who report higher neighborhood disorder tend to live in more urban, densely populated neighborhoods (Walton et al., 2008; York Cornwell & Hall, 2017). Yet, urban residence is associated with lower BMI compared to those living in rural areas (Lundeen et al., 2018). The inverse relationship between neighborhood stress and BMI may be a function of the Black participants living in urban neighborhoods. Further research is needed to better understand the interrelationship between race, neighborhood perceptions and characteristics, urbanicity, and BMI across different age cohorts.

We also explored whether the relationship between the stressors and BMI differed based on gender. We found that financial stress, relationship stress, and cumulative stress were positively associated with higher BMI among women, but not men. These findings are consistent with the broader literature. Conklin et al. (2014) found persistent financial hardships are associated with weight gain for women over 11 years, but not men. A cross-sectional study found that financial stress was associated with greater odds of obesity for women, but not men (Cuevas et al., 2019). Conklin et al. (2014) suggested that women are often in gendered roles that place them as main providers of food and clothing in the household. Persistent financial downturns may disproportionately affect women as the loss of financial support threatens their ability to fulfill their role obligations (Conklin et al., 2014). In addition, Conklin and colleagues suggested that women are more likely than men to fulfill additional social roles related to caregiving. Due to the greater number of roles and demands, financial stressors can worsen the effects of these demands (Conklin et al., 2014). The effects of relationship stress can be an added burden for women. Studies suggest that poor marital quality may have stronger physiological effects on women compared to men given that women may be more responsive to the emotional quality of the relationships than men (Robles et al., 2014).

While we found positive associations between some stressors and BMI, it is important to note that each stressor may operate differently to increase BMI (Tomiyama, 2018). Stress can increase BMI through different pathways, including behavioral, psychological, and physiological means. While this study advances our understanding of the role of stressors on possible BMI gain, this study did not examine the potential pathways by which stressors increase weight gain. Illuminating the pathways could provide insight for weight management interventions.

Limitations

We used a single measure of adiposity (i.e., BMI). BMI is a reliable and valid measure for adiposity and has utility in helping to identify adults at risk for cardiovascular disease and other illnesses (Rothman, 2008). However, other adiposity measures may be more efficient for identifying clinical risks (Janssen et al., 2004). Also, research suggests that body shape and composition differ across BMI categories in non-Hispanic White Americans, non-Hispanic Black Americans, and Mexican Americans, which can distort the relationship between race/ethnicity, BMI, and disease (Heymsfield et al., 2016). Therefore, using other adiposity measures would allow for a better understanding of the effects of stressors on weight gain. We were not able to test a three-way interaction between race, gender, and the stressors due to unequal sample sizes for Black and White respondents. Assessing the effects of stressors on BMI across subgroups can help us identify potential vulnerable groups. Despite examining a wide range of stressors, there were still other stressors that were not investigated in this study. Stressors in this study were mainly chosen due to the availability of data in all three time points of assessment. However, there are other stressors that have obesogenic influence. For instance, workplace stress and caregiving stress are found to be risk factors for weight gain (Wardle et al., 2011), and will be important to consider in future prospective research. This study included only White and Black respondents. Psychosocial stressors may be obesogenic risk factors for other groups, such as Hispanics/Latinos, who have higher prevalence of obesity (Ogden et al., 2014) and report greater exposure to psychosocial stressors than White adults (Cuevas et al., 2019; Sternthal et al., 2011).

Conclusion

There is growing evidence that stress is a risk factor for increased BMI. We found that perceived discrimination, financial stress, and relationship stress are positively associated with BMI. Greater cumulative stress burden across the five stress domains is associated with higher BMI. Prevention and intervention measures to prevent weight gain among middle-aged and older adults will benefit from approaches to reduce exposure and/or responses to psychosocial stressors. The associations between certain psychosocial stressors and BMI may differ based on race and gender, which suggests that the very same interventions that reduce one specific type of stress may not have identical effects across subgroups of society.

Funding

The development of the manuscript was partially supported by Cancer Disparities Research Network/Geographic Management Program (GMaP) Region 4 funded by 3 P30 CA006927-52S2 and by Clinical and Translational Science Institute Mentored Career Development Award (KL2 TR002545).

Conflict of Interest

None declared.

Acknowledgments

The reported study was not preregistered.

Author Contributions

A. G. Cuevas proposed the project, analyzed the data, and drafted the manuscript. S. Greatorex-Voith drafted the methods and results and provided critical input to analytic approach. S. Assari provided critical input to analytic approach, assisted with the interpretation of results and critically reviewed and revised the manuscript. N. Slopen provided critical input to analytic approach, assisted with the interpretation of results and critically reviewed and revised the manuscript. C. D. Economos provided input to analytic procedures and contributed critical input to project framing. All authors provided revisions and approved a final version of manuscript.