Introduction

Stress and obesity share a complex relationship, with cyclical and bidirectional associations across the life course [1, 2]. As described in a conceptual model by van der Valk et al., the bidirectional interplay between obesity, chronic stress, and glucocorticoid action is impacted by numerous individual level characteristics, including genetics, lifestyle, medications and mental distress [1]. “It is well known that stress and obesity are associated”, and many mechanistic pathways that lead to disease development exist, including health behaviors, glucocorticoid activation, and mental health [1, 3, 4]. However, having obesity has also been found to increase stress due to several reasons, including comorbidities that limit daily activities, and weight stigma or bias, which may induce a prolonged stress response [3]. Adverse childhood experiences (ACEs), defined as a wide range of negative events, including abuse, neglect, witnessing violence, parental mental illness or incarceration of a family member [5], are one example of an individual level factor that may impact both obesity and stress [4,5,6,7,8]. Although ACEs take place early in life, the effects have been found to extend beyond childhood or adolescence into older adulthood [4]. Following exposure to ACEs there may be a prolonged stress response, which is also known as toxic stress, which may make people with ACEs more susceptible to stress or worse experiences during a stressful event [5,6,7, 9].

In Canada, throughout the first year of the pandemic (March 2020-March 2021), strict public health preventive measures were in place including work and school closures, and travel bans [10]. In turn, this created wide-reaching implications on population health including an increase in stress [11,12,13]. Obesity was identified as a risk factor for increased COVID-19 morbidity and mortality early in the pandemic and this may have contributed to increased weight bias and stress among people living with obesity [14,15,16]. There is limited research on the impact of living with obesity on stress during the COVID-19 pandemic. A systematic review has found that exposure to disasters increases cardiometabolic risk, including obesity, across the life course, however, research has not yet explored how obesity may influence stress experienced during a disaster [17]. Stress during a disaster can be measured objectively, using reports of stressors, or subjectively, measuring perceptions of a disaster [18]. ACEs are an established risk factor for both stress [5,6,7] and later life obesity [4, 8]. ACEs have been associated with higher psychological symptoms following a natural disaster [19] and it is possible people with obesity who experienced ACEs may have experienced an added burden of stress during the pandemic [20,21,22]. Differences also exist by sex, whereby females have higher psychological symptoms following a disaster [19], and the prevalence of ACEs and obesity vary among males and females [23, 24].

It is possible people with obesity may have experienced greater stress during the COVID-19 pandemic, and life course epidemiology frameworks help to identify potential distal risk factors, such as ACEs, that may have led to differential experiences [25]. This is of importance in the context of the pandemic, and beyond, as the findings from this research will contribute to understanding the complex relationship stress and obesity share. The objectives of this study were to evaluate the association between both obesity and ACEs and stress during the pandemic (stressors and perceived consequences) and to determine if the association between obesity and stress was modified by ACEs. Differences in the association between obesity and ACEs by sex were also evaluated.

Methods

Study design and participants

We conducted an analysis using longitudinal data collected as part of the Canadian Longitudinal Study on Aging (CLSA). The methodology of the CLSA has been published elsewhere [26]. Briefly, the CLSA is a national longitudinal study of adults aged 45–85 at the time of recruitment (2011–2015). At the time of recruitment, participants provided informed consent. Adults from the 10 Canadian provinces were recruited using population-based sampling strategies [26]. Participants were eligible for inclusion into the CLSA if they could complete interviews in English or French, were cognitively able participate on their own, were not in an institution, did not reside in a Canadian territory (The Northwest Territories, the Yukon and Nunavut) or on a Federal First Nations reserve, and were not a full-time member of the Canadian Armed forces. Ethics approval for this study was received from the Hamilton Integrated Research Ethics Board (HiREB).

The CLSA is comprised of the Tracking cohort and the Comprehensive cohort. Data for the Tracking cohort were collected using telephone interviews, whereas data for the Comprehensive cohort were collected via in-home interviews and clinical data collection site visits. Data are collected every three years and all participants will be followed for 20 years, or until death or loss-to-follow-up. Data for this analysis were collected at baseline in 2011–2015 and at follow-up 1 in 2015–2018. At the start of the COVID-19 pandemic, the CLSA COVID-19 Questionnaire Study was implemented, which collected longitudinal data from April 2020 to December 2020. Specific details about when data for this study were collected can be found in Table 1.

Primary exposures

Obesity

Obesity was measured at CLSA follow-up 1 (2015–2018). For individuals in the Comprehensive cohort (n = 15,582), height and weight were measured by trained research assistants. These measurements were used to calculate body mass index (BMI) (kg/m2). For individuals in the Tracking cohort (n = 8390), height and weight were assessed using self-report, which were then used to calculate BMI. A correction factor developed by Statistics Canada was applied to the self-reported BMI to account for bias associated with self-report [27]. These correction equations were generated using the 2005 Canadian Community Health Survey with consideration of several sociodemographic variables separately for males and females [28]. Self-reported BMI was slightly underestimated compared to the corrected BMI, which is consistent with the literature [27]. BMI was categorized following World Health Organization standard cut-offs [29]: normal weight (≤24.9 kg/m2), overweight (25–29.9 kg/m2), obesity class I (30–34.9 kg/m2), obesity class II (35–39.9 kg/m2) and obesity class III (≥40 kg/m2). Obesity was further classified into 3 subgroups, as research has found variation in risk of health outcomes across the subtypes [30]. Underweight individuals were included in the normal weight category given the small sample size.

To measure ACEs, at CLSA follow-up 1, participants were asked about 11 experiences before the age of 16 related to physical abuse, sexual abuse, emotional abuse, neglect, and exposure to intimate partner violence. Participants were also asked about three experiences before the age of 18 related to death of a parent, parental divorce/separation and living with a family member with mental health problems. These questions were adapted from the Childhood Experience of Violence Questionnaire and the National Longitudinal Study of Adolescent to Adult Health Wave III questionnaire [31, 32]. Based on responses to dichotomized yes/no questions, a cumulative score was created by summing the total number of ACEs participants reported [31]. Since only a small proportion of people reported 5 to 8 ACEs (4%), those reporting four or more were collapsed into one group. A cumulative ACEs score was used rather than subgroups by severity, as research has found this to be a better assessment of cumulative exposure, and has been found to be associated with health outcomes [33].

Measurement of outcomes (stress)

Stress was measured in two ways: (1) stressors and (2) the perceived consequences of the pandemic. These questions have previously been used in disaster research [18, 34,35,36] to study objective and subjective stress following a disaster such as the COVID-19 pandemic. The development of these questions were modified from gold-standard measurement tools [18].

Stressors

Stressors were measured at CLSA COVID-19 Questionnaire Study Exit Survey (September 2020-December 2020). Participants were asked, “Which of the following have you experienced during the COVID-19 pandemic?” where participants could select one or more of the following options: participant was ill, someone close to the participant was ill, someone close to the participant died, loss of income, unable to access necessary food and supplies, unable to access healthcare, unable to access usual prescriptions, increased conflict, separation from family, increased caregiving, unable to care for those who require assistance due to limitations, and breakdown in family relationships. The 12 stressors were classified into four domains for this analysis; (1) health (participant was ill, someone close to the participant was ill, someone close to the participant died), (2) resources (loss of income, unable to access necessary food and supplies, unable to access healthcare, unable to access usual prescriptions), (3) relationships (increased conflict, separation from family, breakdown in family relationships), and (4) caregiving (increased caregiving, unable to care for those who require assistance due to limitations). To create each domain, the total number of stressors within each category was summed. The range of values for each domain varied depending on how many stressors fell within the category. For instance, the health domain ranged from 0 to 3, whereas the resources domain ranged from 0 to 4. In addition, a cumulative stressor score was created by summing the total number of stressors participants experienced across all domains [37]. The cumulative stressor score ranged from 0 to 12.

Perceived consequences of the pandemic

As a subjective measure of perceived stress, participants were asked “Taking everything about COVID-19 into account, how would you describe the consequences of COVID-19 on you and your household?” during the CLSA COVID-19 Questionnaire Study Exit Survey (September 2020-December 2020) [18, 34, 35]. Response options were very negative, negative, neutral, positive, and very positive. Very few participants reported the consequences of the pandemic as very negative or very positive, so these categories were combined with negative and positive response options, respectively. The neutral category was further combined with the positive and very positive category to create a binary variable, since the objective of the analysis was to explore negative/very negative perceived consequences of the pandemic compared to all other perceptions.

Measurement of potential confounding variables

All remaining variables were measured at CLSA baseline (2011–2015), CLSA follow-up 1 (2015–2018), CLSA COVID-19 Baseline Survey (April 2020-June 2020) or the CLSA COVID-19 Exit Survey (September 2020-December 2020). These variables were chosen based on the framework proposed by van der Valk et al., identifying characteristics that are related to the association of stress and obesity [1]. Participant sex (male or female) and racial background (white or non-white) were collected at CLSA baseline. Participant age at CLSA COVID-19 Baseline Survey was categorized as 50–64 years, 65–74 years, and 75–96 years. Physical activity, household income, alcohol consumption and depression were measured at CLSA follow-up 1. The Physical Activity Scale for the Elderly (PASE) was used to assess level of physical activity for the previous seven days [38]. Based on the World Health Organization (WHO) guidelines [39], physical activity was dichotomized into ≤150 min/week of moderate-intensity or ≤75 min/week of vigorous-intensity versus >150 min/week of moderate-intensity or >75 min/week of vigorous-intensity. Household income was categorized into less than $50,000,$50,000 to less than $100,000,$100,000 to less than $150,000, and$150,000 or more, and alcohol consumption over the past 12 months was categorized as did not drink in the last 12 months, occasional drinker, and regular drinker (at least once a month). Depression was assessed using the Center for Epidemiologic Studies Short Depression (CESD) scale [40], where a score of ≥10 indicates risk for clinical depression.

Statistical analysis

All statistical analyses were completed using SAS 9.4. Statistical code is available upon request. The associations between both obesity and ACEs were independently evaluated in relation to the three primary outcomes, (1) the stressor domains, (2) total stressor score, and (3) the perceived consequences of the pandemic. PROC GENMOD was used to estimate relative risks (RRs) and 95% confidence intervals (CIs). For all outcomes, a log link function was used, however the distribution used varied for different outcomes. For the stressor domains, a Poisson distribution was assumed as this was a count variable. Although the total stressor variable was also a count variable, a negative binomial distribution was assumed given the overdispersion. Finally, a binomial distribution was assumed for the binary perceived consequences of the pandemic variable. All models were adjusted for potential confounders that were hypothesized a priori to be predictors of both the exposures and outcome variables. These included sex, age group, racial background, physical activity, household income, alcohol consumption and depression [1]. For the association between ACEs and stress, an additional model was run adding obesity to the fully adjusted model, given the potential mediating role of obesity. All variables had less than 5% of participants missing, and a complete case analysis was conducted. A sensitivity analysis was conducted to explore differences in associations by severity of ACEs. We explored the association between maltreatment ACEs and measures of stress, and family dysfunction ACEs and measures of stress.

For the association between obesity and measures of stress, interaction by both ACEs and sex were assessed separately on both the additive and multiplicative scales. In epidemiologic research, interaction is often only explored on the multiplicative scale, however, the assessment of interaction on the additive scale has significant public health importance as it can contribute to better allocation of resources and identification of high-risk subgroups [41]. STROBE guidelines recommend presenting the separate effects of exposures and modifiers, as well as joint effects to ensure readers can assess interaction on either scale [41]. To determine if the associations between obesity and measures of stress were modified by ACEs, a dichotomous ACEs variable was created. Individuals who reported no ACEs were categorized as none, and those who reported one or more ACES, were categorized as yes. Using the framework proposed by Knol and VanderWeele [41], interaction was tested on the additive scale using the relative excess risk due to interaction (RERI) and on the multiplicative scale using the ratio of relative risk (RRR). The 95% CI for the RERI were calculated using the delta method [41,42,43].

Results

A total of 23,972 participants were included in this analysis. A detailed flowchart of the analytic sample can be found in Fig. 1 and characteristics of the study population are presented in Table 1. The CLSA COVID-19 Questionnaire Study participants are generally comparable to the full CLSA sample, however, this subgroup had a slightly higher mean household income and higher education than the full sample [44]. Over three quarters (76%) of the participants reported at least one stressor and 63% reported perceiving the consequences of the pandemic as negative or very negative (Table 2).

Obesity and stress

A consistent dose-response association was observed for the association between obesity and total stressors, and the health and resources domains, whereby as obesity level increased, the risk of reporting an additional stressor also increased. For instance, adults with class III severe obesity (≥40 kg/m2), compared to normal weight (≤24.9 kg/m2), had a greater risk of reporting an additional stressor for the total number of stressors (adjusted RR: 1.19; 95% CI: 1.12, 1.27), within the health domain (adjusted RR: 1.25; 95% CI: 1.12, 1.39), and within the resources domain (adjusted RR: 1.38; 95% CI: 1.25, 1.53) (Table 3). Obesity was slightly associated with an increased risk of perceiving the consequences of the pandemic as negative/very negative (obesity class III adjusted RR: 1.05; 95% CI: 1.00, 1.11) (Table 3). When ACEs were added to the model of obesity and stress the results did not change suggesting that ACEs was not a confounder of this association. Results only slighted changed after adjustment for confounding variables (Supplemental Table A1).

There were 14,607 (61%) adults who experienced one or more ACE (Table 1). Across all outcomes, there was a strong dose-response association, whereby as the number of ACEs increased, the risk of reporting an additional stressor and within each stressor domain also increased (Table 3). For instance, among those with the highest number of ACEs, the risk of an additional stressor within the resources domain was 53% greater than for those with no ACEs (adjusted RR: 1.53; 95% CI: 1.43, 1.63). Similarly, the adjusted RR estimates for the risk of perceiving the consequences of the pandemic as negative/very negative also increased as the number of ACEs increased (Table 3). Those with 4 to 8 ACEs, compared to none, were 32% more likely to perceive the pandemic as negative/very negative compared to neutral/positive/very positive (adjusted RR: 1.32; 95% CI: 1.19, 1.47). The adjusted results were attenuated slightly but even after adjustment for potential confounders identified a priori, the results remained similar and statistically significant (Supplementary Table A1). To evaluate whether the association between ACEs and stress was explained fully or in part by obesity we ran a model adjusting for obesity. When adding obesity to the models for ACEs and all outcomes, associations were attenuated only slightly suggesting that obesity may not mediate the association between ACEs and stress (results not shown).

The association between obesity and stress by ACEs

The tests for interaction by ACEs on both the additive and multiplicative scales for the association between obesity and stress are provided in Table 4. There was no consistent evidence of interaction on either the additive or multiplicative scale by ACEs on any of the outcomes. Only the multiplicative interaction between class III obesity and ACEs for stressors within the health domain was statistically significant (RERI = –0.34; 95% CI: –0.71, 0.02; RRR = 0.74; 95% CI: 0.58, 0.96), meaning among people with class III obesity, those with ACEs, compared to those with no ACEs, were less likely to report an additional stressor within the health domain.

Sex differences

The tests for interaction of obesity by sex on the additive and multiplicative scale are shown in Supplemental Table A3. There was consistent evidence of negative interaction on the multiplicative scale for class III obesity, such that females with class III obesity were less likely to report stress outcomes during the pandemic than males. For example, the joint exposure of having obesity class III and being female was associated with lower reports of stressors in the health domain (RERI = –0.27; 95% CI: –0.59, 0.05; RRR = 0.79; 95% CI: 0.63, 1.00), resources domain (RERI = –0.31; 95% CI: –0.61, –0.01; RRR = 0.80; 95% CI: 0.65, 0.98), caregiving domain (RERI = –0.60; 95% CI: –1.10, –0.09; RRR = 0.62; 95% CI: 0.44, 0.88), total stressors (RERI = –0.25; 95% CI: –0.43, –0.07; RRR = 0.80; 95% CI: 0.70, 0.91) and perceiving the pandemic as negative or very negative (RERI = 0.01; 95% CI: –0.07, 0.09; RRR = 0.89; 95% CI: 0.80, 0.99).The tests for interaction of ACEs by sex are shown in Supplemental Table A2. There was limited evidence of interaction on either scale, and all of the tests except one were not statistically significant. The only significant test for interaction was on the multiplicative scale for the perceived consequences of the pandemic such that, females with 4–8 ACEs were less likely to perceive the pandemic as negative or very negative compared to males (RRR = 0.91; 95% CI: 0.84, 0.99, p = 0.02).

Sensitivity analysis

Sensitivity analyses to explore differences in associations between child maltreatment ACEs and measures of stress, and family dysfunction ACEs and measures of stress can be found in Table A4. For both maltreatment ACEs and family dysfunction ACEs, the greatest risk of reporting an additional stressor, or perceiving the consequences of the pandemic as negative/very negative were among those who reported the most ACEs. Associations were slightly larger for those who reported the most maltreatment ACEs.

Discussion

The findings from this study contribute to our understanding of the experiences of stress among people with obesity during the COVID-19 pandemic and the cyclical relationship between obesity and stress. A dose-response association was found between obesity some measures of stress (total stressors, resources domain, and health domain). Although ACEs did not modify this association, it was found to be independently associated with stress experienced during the pandemic, as we identified a strong dose-response association between ACEs and all measures of stress. It was hypothesized that obesity may mediate the association between ACEs and stress experienced during the pandemic given the cyclical association stress and obesity share, however, the preliminary mediation analysis did not find obesity to be a mediator. This may be related to the timing of the measurement or that there are multiple pathways whereby ACEs influenced pandemic-related stress. Future research exploring the mechanisms behind these associations is needed, as it can be used to inform the response to future disasters or stressful events.

The joint exposures of obesity and sex were significant, meaning sex modifies these associations. Females with ACEs and with class III obesity were less likely to report an additional stressor compared to males. Our study appears to be the first to evaluate the joint effect of sex with other distal and proximal factors to stress caused by population-level adversity, such as the COVID-19 pandemic. However, studies that evaluated sex independently found females compared to males typically have higher reports of psychological related outcomes following disasters [19, 45].