A longitudinal study evaluating adverse childhood experiences and obesity in adulthood using the Canadian Longitudinal Study on Aging (CLSA)

Authors: Vanessa De Rubeis, Andrea Gonzalez, Jean-Éric Tarride, Lauren E Griffith, Laura N Anderson

Abstract

Background

Adverse childhood experiences (ACEs) are a risk factor for obesity; however, the causal mechanisms are not well understood. Objectives were to measure the impact of ACEs on adulthood obesity and to investigate whether the association was mediated by nutrition and stress.

Methods

A longitudinal study was conducted using adults aged 46–90 years (n = 26 615) from the Canadian Longitudinal Study on Aging. Participants were asked to recall ACEs from <18 years of age. Body mass index (BMI), waist circumference and per cent body fat were measured (2015–18) and obesity was defined using standard cut points. Nutrition was measured using data from the Short Diet Questionnaire and stress was measured using allostatic load. Multinomial logistic regression was used to estimate odds ratios (ORs) and 95% CIs for each obesity measure. Causal mediation methods were used to determine whether nutrition and stress were mediators.

Results

There were 66% of adults who had experienced one or more ACE. The odds of obesity defined by BMI and waist circumference increased in a dose–response manner with increasing number of ACEs (P trend <0.001). For instance, adults with four to eight ACEs, compared with none, had greater odds of obesity, defined by BMI (adjusted OR: 1.54; 95% CI: 1.28–1.75) and waist circumference (adjusted OR: 1.30; 95% CI: 1.15–1.47). There was no evidence of mediation by stress or nutrition.

Conclusions

Adversity experienced in early life is strongly associated with obesity among Canadian adults. Further research is needed to identify other mechanisms for this association to inform obesity prevention strategies.

Background

National data suggest that 27% of Canadians aged ≥18 years—about 7.3 million adults—have obesity.^1^ Although body mass index (BMI) is typically used in epidemiologic research to measure obesity, other measures of adiposity, such as waist circumference and per cent body fat, may provide further information in terms of disease risk.^2^ Waist circumference provides an indication of excess fat that is located in the abdominal region, which may put an individual at greater risk of disease compared with fat that is located in other regions of the body.^3^ Similar to many chronic diseases, obesity has a long latency period. Therefore it is important to understand the early-life determinants for obesity and mechanisms across the life course to inform obesity prevention.^4^ Although numerous genes and epigenetic variations in multiple biologic pathways have been associated with obesity, environmental factors during early life are also critical, signifying the importance of exploring early origins of obesity.^5–7^

Adverse childhood experiences (ACEs) have a profound impact on disease risk across the life course.^8^^,^^9^ There are several possible frameworks that could explain how experiences in early life alter the body’s functioning leading to the development of diseases, including obesity.^10–12^ These include sensitive or critical periods of development, chain of risk or accumulation of risk.^12^ Systematic reviews have consistently reported that people with a history of adversity in childhood had greater odds of developing obesity across the life course.^13–16^ It was noted that regardless of the method of assessment and with either continuous or categorical assessment of body weight, the association between early-life adversity and increased risk of obesity persisted.^14^ There has been a call for future research to use large, population-based samples to further explore the association between ACEs and obesity, and to explore sex differences, as it has been found that females typically report more ACEs, but it is not clear how this impacts obesity development.^13^^,^^17^

A recent systematic review exploring plausible mechanisms following exposure to ACEs to obesity development noted that the availability of data often limits the exploration of potential pathways to disease development.^13^ Expanding on the Developmental Origins of Health and Disease framework may inform pathways to obesity development, as it hypothesizes two main pathways to disease development through nutrition and stress.^5^^,^^18^ ACEs have been linked to changes in diet^19^ and increased stress,^20^ which are also linked to obesity,^21^^,^^22^ therefore making them potential mediators. Thus, the objectives of this study were to evaluate the association between ACEs and obesity in adults aged 46–90 years in Canada and to investigate whether the association between ACEs and obesity was mediated by nutrition and stress using causal mediation methods. Sex differences were also examined.

Methods

Study design and participants

A longitudinal was conducted study using data from the Canadian Longitudinal Study on Aging (CLSA). The CLSA has collected data on >50 000 community dwelling adults from Canada. The complete description of the methodology of the CLSA can be found elsewhere.^23^ Briefly, participants were recruited using a population-based sampling strategy. People who resided in the 10 Canadian provinces could complete interviews in either English or French, did not reside in an institution or on a Federal First Nations reserve, were not a full-time member of the Canadian Armed forces and were cognitively able to participate on their own were eligible for inclusion. This study uses data collected at baseline (2011–15) and Follow-up 1 (2015–18). The CLSA has two the Tracking Cohort (n = 21 241) and the Comprehensive Cohort (n = 30 097), which vary in how data were collected. Only participants in the Comprehensive Cohort were eligible for inclusion in this study, as they provided anthropometric measures to define obesity, as well as biomarkers that were used to generate the measure of stress. Ethics approval for the current study was obtained from the Hamilton Integrated Research Ethics Board on 24 November 2020.

Measurement of ACEs

At Follow-up 1 (2015–18), participants were asked to recall events that occurred before the age of 16 years within their family related to physical abuse, sexual abuse, emotional abuse, neglect and exposure to intimate partner violence. Participants were also asked whether they had experienced the death of a parent, parental divorce/separation or living with a family member with mental health problems before the age of 18 years. Questions asked were adapted from the Childhood Experience of Violence Questionnaire and the National Longitudinal Study of Adolescent to Adult Health Wave III questionnaire.^24^^,^^25^ The number of reported ACEs were summed to create a total ACEs variable, ranging from 0 to 8. This method has previously been used in the CLSA.^26^ Since relatively few people reported more than five ACEs, four or more ACEs were collapsed into one group.

Measurement of obesity

All obesity measures were taken by trained research assistants at Follow-up 1 (2011–15), who followed standardized protocols to ensure valid and reliable measurement.^27^^,^^28^ Participants’ measured height and weight were used to calculate the BMI (kg/m^2^). The BMI was categorized using the World Health Organization (WHO) standard cut-offs for defining obesity^29^^,^^30^: normal weight (≤24.9 kg/m^2^), overweight (25–29.9 kg/m^2^) and obesity (≥30 kg/m^2^). A cut-off of ≥88 cm for females and ≥102 cm for males was used to define obesity by waist circumference. Participants’ body fat (%BF) was measured using Hologic Discovery A Dual-Energy X-Ray Absorptiometry (DXA) machines.^31^ Although there are no well-established cut points to define obesity using %BF, the WHO suggests using >35% for females and >25% for males.^32^

Measurement of mediators

Measurement of nutrition

To measure nutrition, the unhealthy diet score was derived using data from the CLSA Short Diet Questionnaire (SDQ)^33^ at CLSA baseline (2011–15). The unhealthy diet score was created based on a methodology from the Prospective Urban Rural Epidemiological healthy diet score and has previously been applied to the CLSA data.^34^ Seven food groups (fruits, vegetables, legumes, nuts, fish, dairy and meat), measured in the number of servings per day, were divided into quintiles based on the sample distribution. A cumulative score was created by adding the quintile of consumption for each food group, creating a score ranging from 0 to 28 (where 0 is the healthiest diet and 28 is the unhealthiest diet).

Measurement of stress

Accumulated life stress was measured using an index of allostatic load, which is a measure of physiological dysregulation, as the result of exposure to chronic stress.^26^ This score was derived using several haematological, cardiometabolic and clinical biomarkers including white blood cells, HbA1c, albumin, alanine aminotransferase, creatinine, haemoglobin, ferritin, C-reactive protein, total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, triglycerides, systolic and diastolic blood pressure and heart rate taken at baseline (2011–15). Based on the sample distribution of each biomarker and following recommendations,^35^ high risk was defined using the upper or lower 25th percentile as appropriate for each biomarker. The number of biomarkers that fell within the high-risk category were summed to create the allostatic load index. This method has been previously used in the CLSA^26^ and has been recommended as a measure of accumulated stress.^36^^,^^37^ Allostatic load ranged from 0 to 13, with a score of 0 indicating less accumulated stress and a score of 13 indicating more accumulated stress.

Measurement of other variables

Potential confounding variables included participants’ age, sex, education level, racial background, household income, smoking status and alcohol intake. Participants’ sex was categorized as male or female and age at Follow-up 1 (2015–18) was grouped into categories of 46–54, 55–64, 65–74 and 75–90 years. At baseline, participants were asked to report their racial background, categorized into White and other.

Statistical analysis

All statistical analyses were conducted using SAS 9.4. Multinomial logistic regression was used to estimate odds ratios (ORs) and 95% CIs. Unadjusted and adjusted models (adjusted for age, sex and racial background) were run separately for all three obesity outcome measures. Models were also stratified by sex and a Wald test for interaction was conducted. Inflation weights were applied to descriptive analyses to account for sample misrepresentation, coverage error and non-response, which improved the overall precision of estimates.^38^ Analytic weights were used for regression analyses to ensure results represent the association among variables at the population level, rather than the association between variables within the selected sample.^38^ Complete case analysis was conducted as there were minimal missing data (no variable had >10% missing).

Causal mediation analysis was conducted following the principles outlined by Vander Weele.^39^^,^^40^ Exposure–mediator, mediator–outcome and exposure–outcome confounders were identified to control for potential biases and to meet the assumptions of causal mediation analysis (Figure 1). Given the time that had elapsed between ACEs (exposure) and obesity (outcome), there is a greater risk of violation of the fourth assumption. We outline other possible pathways that may explain how ACEs may lead to obesity development but were beyond the objective of the current project. We allowed for exposure–mediator interaction in all models. PROC CAUSALMED was used to estimate the total effect, direct effect and indirect effects. This procedure cannot handle multi-category exposure or outcome variables.^40^ ACEs were dichotomized as (0–3 ACEs) vs (4–8 ACEs); this categorization was chosen as 4–8 ACEs has consistently been found to be associated with increased outcomes.^8^ Obesity was also dichotomized as no obesity (≤29.9 kg/m^2^) vs obesity (≥30 kg/m^2^). The analysis for each mediator (unhealthy diet score and allostatic load) were conducted separately as it was unlikely that either variable influenced the other, given they were both measured at the same time point (CLSA baseline 2011–15). Separate mediation analyses were run for all three obesity outcomes (BMI, waist circumference and %BF).

Sensitivity analysis

As a sensitivity analysis, we ran the adjusted associations between the exposure (ACEs) and mediators (nutrition and stress), and the mediators (nutrition and stress) and outcomes (obesity defined by BMI, waist circumference and %BF) to explore the individual associations. We also conducted a sensitivity analysis to explore how the severity of ACEs influenced the association between ACEs and obesity, and the subsequent mediation analyses by creating a new ACEs score and rerunning the mediation analysis without parental separation/divorce. It is possible that the effect of obesity on stress may differ in older adults,^41^ thus we conducted sensitivity analysis removing participants in the oldest age group (75–90 years).

Results

This study included 26 615 participants from the CLSA at Follow-up 1 (2011–15) (Figure 2). Over 63% of participants reported one or more ACE (n = 16 745) (Figure 3). A high proportion of people had obesity defined by BMI (31%), waist circumference (43%) and %BF (74%). A more detailed description of the included participants can be found in Table 1.

As the number of reported ACEs increased, the odds of obesity also increased for obesity defined by BMI and waist circumference, although this pattern was not consistent for obesity defined by %BF. The P trend for obesity defined by BMI and waist circumference was *P = *0.001, indicating a dose–response association (Table 2). For all three obesity measures, people who reported 4–8 ACEs compared with those who reported no ACEs had the highest odds of obesity in adulthood defined by BMI (adjusted OR: 1.54; 95% CI: 1.28–1.75), waist circumference (adjusted OR: 1.30; 95% CI: 1.15–1.47) and %BF (adjusted OR: 1.27; 95% CI: 1.10–1.48) (Table 2). Adjustment for confounders only slightly changed ORs across outcomes (unadjusted results are shown in Table 2).

When the results were stratified by sex, similar associations remained. For instance, males who reported 4–8 ACEs compared with none were 64% more likely to have obesity defined by BMI (adjusted OR: 1.64; 95% CI: 1.26–2.14), whereas females who reported 4–8 ACEs compared with none were 44% more likely to have obesity (adjusted OR: 1.44; 95% CI: 1.18–1.75) (Table 3). For obesity defined by waist circumference, associations were similar. For example, among those who reported 4–8 ACEs, males had a slightly higher OR (adjusted OR: 1.30; 95% CI: 1.06–1.59) compared with females (adjusted OR: 1.29; 95% CI: 1.10–1.51) (Table 3), whereas for %BF, females with 4–8 ACEs compared with none had higher odds of obesity (adjusted OR: 1.35; 95% CI: 1.10–1.65) compared with males with 4–8 ACEs compared with 0 (adjusted OR: 1.19; 95% CI: 0.94–1.49). Across almost all stratified associations, there was no evidence of a difference by sex and nearly all confidence intervals overlapped.

Mediation by nutrition and stress

We did not find exposure–mediator interactions for any models; therefore, it can be assumed that the controlled direct effect and natural direct effect were not different, and therefore we only present the natural direct effect (Table 4). There was limited evidence of mediation by both nutrition and stress for obesity defined by BMI, waist circumference and %BF as all indirect effects were close to null and, for the nutrition mediation models, all indirect effect confidence intervals included one (Table 4). The percentage mediated also suggested limited mediation by nutrition and stress as the CIs for all percentages were either very wide or included zero. Mediation analyses were stratified by sex; however, results did not differ between males and females (results not shown).

Sensitivity analysis

Results for all sensitivity analyses can be found in the Supplementary Information (available as Supplementary data at IJE online). We did not find an association between ACEs and nutrition (Supplementary Table S2, available as Supplementary data at IJE online) and all associations between nutrition and obesity outcomes were close to null (Supplementary Table S3, available as Supplementary data at IJE online). Conversely, the adjusted beta for the association between ACEs and stress was 0.18 (95% CI: 0.05–0.31) (Supplementary Table S2, available as Supplementary data at IJE online) and an association was found between stress and all obesity outcomes (Supplementary Table S3, available as Supplementary data at IJE online). This may warrant further exploration since our measure of stress was found to be slightly associated with both our exposure and outcome, which is consistent with our mediation findings whereby the indirect effects were slightly above 1, although the per cent mediated CIs were very wide. Results of our sensitivity analysis evaluating the type of ACEs suggest that both maltreatment ACEs and family dysfunction ACEs were associated with increased odds of obesity defined by each measure (Supplementary Table S4, available as Supplementary data at IJE online). Removal of parental separation or divorce from the ACE score did not appear to change results (Supplementary Table S5, available as Supplementary data at IJE online). Restricting the age of the sample to a younger group did not appear to change results (Supplementary Table S6, available as Supplementary data at IJE online).

Discussion

Consistent with previous research, the results of our study demonstrate a strong, dose–response association between ACEs and obesity in adults aged 46–90 years, however, our findings are novel given the use of multiple obesity measures taken by trained research assistants, rather than self-report. We found those who reported 4–8 ACEs had the greatest odds of obesity, which was consistent across all measures of obesity. We did not identify differences in associations by sex. When exploring factors that explained the association between ACEs and obesity, our study did not find evidence that either nutrition or stress mediated this association.

In our large population-based sample of 26 615 Canadian adults, two-thirds reported experiencing at least one ACE. This is similar to other estimates of population-based studies,^42^ which is concerning since ACEs are a known risk factors for several conditions.^20^ Given the public health impact of obesity, understanding risk factors in early life is important, leading to new strategies for obesity prevention or treatment.^1^ Whereas ACEs have consistently been shown to increase the risk of obesity, the pathways to disease development are understudied and not well understood.^13^^,^^14^^,^^43^ A recent systematic review evaluating associations between ACEs and adulthood obesity identified few studies that specifically studied obesity as the outcome, none of which was conducted in the Canadian population.^13^ The most commonly cited explanations for this association included social disruption (e.g. income), health behaviours, chronic stress response and mental health.^13^ Another review that focused more broadly on adversity throughout childhood, obesity and binge-eating disorder found the most common pathways were depression, post-traumatic stress disorder and interpersonal and neurobiological factors.^11^ There are several pathways that explain how adversity during childhood leads to obesity development and it is possible that these pathways are intertwined, meaning that they work together after exposure in the development of obesity. As outlined in our conceptual model, there are several other pathways that explain how ACEs lead to obesity, although they were beyond the scope of the current study. For instance, there is some evidence on mediation through mental health.^44^^,^^45^ It has been suggested that intervention strategies should be aimed at different pathways of disease development following exposure to ACEs, which should be considered in the context of obesity development.^26^^,^^46^ It has been found that that nutrition and measures of accumulated stress explain the pathway between ACEs and the development of other diseases.^26^^,^^46^ Although we did not find these to be mediators, it is still possible that these could explain the development of obesity following ACEs, and null findings may be related to the timing of assessment or that these factors work in tandem with other pathways (e.g. mental health, lifestyle or health behaviour factors, interpersonal factors or neurobiological factors).^11^^,^^13^ Future researchers who have access to different measures of nutrition and stress may consider further exploring the role of these variables in this association, as well as studying multiple mediators along this pathway. The findings from this study still remain as an important contribution to the literature, as it has been noted that there is limited availability of data to explore potential mechanistic pathways.^13^

Similarly to previous research, our study did find that females reported experiencing more adverse experiences compared with males.17 Other studies evaluating the impact that ACEs have on outcomes later in life reported stronger associations with outcomes including PTSD^17^ and multimorbidity^26^ among females has been found. However, we did not find any evidence of sex differences for the associations with obesity defined by BMI, waist circumference and %BF. Future research should explore both sex and gender differences, as it is possible that there are differences in how ACEs are biologically or socially embedded, and that the physiological or societal changes that occur following adversity may differentially predispose someone to have an increased risk of disease development, including obesity.^47^

There are several strengths of our study including the large population-based sample and the use of longitudinal data. This study also is one of the first to use objective measures of obesity rather than self-report on a large sample, including the assessment of body fatness using DXA scans. The use of mediation methodologies is also a strength of this study as it has been noted that there are biases associated with more traditional approaches, such as the inability to control for exposure–outcome and mediator–outcome confounders.^48^ Limitations of this study include the lack of ethnic diversity, which limits the ability to generalize results. Temporality is a requirement for causal mediation. In our study ACEs were specific to experiences that occurred before the age of 18 years and measurements of nutrition, stress and obesity were prospectively collected later in adulthood; however, ACEs were based on self-reported recall and we cannot exclude the possibility that the temporality assumption could have been violated by recall bias. Yet, previous studies have reported that ACEs can be recalled in later adulthood accurately.^49^ In addition, people with obesity and ACEs may have differentially participated in our study, therefore potentially introducing selection bias in this study. Allostatic load is commonly used to measure accumulated stress, although measures of obesity, including BMI, are typically included in the allostatic load index.^36^ Given that these were outcomes of interest, this was removed from the allostatic load index. Future research may consider using other objective or subjective tools to measure indicators of stress, as this may be alleviate measurement error or biases associated with the allostatic load measure used in our study.^50^ Assessment of nutrition and allostatic load were taken relatively close to when the obesity measures were taken, potentially limiting the ability to identify them as mediators. Future research may consider assessment of nutrition or stress earlier in life or assessment at different time points, allowing a multiple mediation analysis to be conducted since it is possible that the development of obesity following exposure to ACEs occurs through multiple pathways. We defined obesity using standard cut-offs for BMI, waist circumference and %BF, although these cut-offs have been noted to vary across different samples or groups of people.^51^ For instance, there is limited consensus on a standard cut-off to define obesity using %BF.^32^ For this current study, we identified a cut-off that is often used in research^32^^,^^52^^,^^53^ since there is no established cut-off from the WHO; however, this may be a potentially liberal cut-off, which would explain the high prevalence of obesity defined by %BF found in our study. Continued research on cut-offs to define obesity is needed, as this may improve obesity prevention and interventions. There may also be the potential for measurement error associated with the dichotomization of ACEs for the mediation analysis, which may potentially bias the decomposition of effects in the mediation analysis. Although we controlled for exposure–mediator, mediator–outcome and exposure–mediator confounders, it is possible there is residual confounding, which could potentially bias findings as the assumptions of causal mediation may be violated.^54^ Interpretation of findings should be done with caution given the potential violation of the fourth assumption, which may be related to the time between exposure and outcome, and the use of observational data. Future studies might consider the use of the G-computation formula to adjust for time-varying confounding, which may overcome issues associated with the fourth assumption of causal mediation methodologies.

ACEs dramatically influence obesity later in life, whereby the more ACEs a person experiences, the greater the risk of obesity in later life. Although we did not find evidence to support the role of nutrition and stress as mediators along the pathway to disease development, these findings can be used to inform future research questions or hypotheses about possible plausible mechanisms. Understanding the pathway to obesity development following ACEs is critical to inform strategies to identify people at a greater risk of disease later in life, allowing potential interventions. Although adversity may occur very early on in life, it is apparent the detrimental effects last beyond childhood, contributing to negative outcomes later in life.

Ethics approval

Ethics approval was received from the Hamilton Integrated Research Ethics Board (HiREB) (project ID: 12886). This study conforms to the Declaration of Helsinki.

Supplementary Material