Obesity is a prevalent global-health problem associated with substantial morbidity, impairment and economic burden. Because most readily available forms of treatment are ineffective in the long term, it is essential to advance knowledge of obesity prevention by identifying potentially modifiable risk factors. Findings from experimental studies in non-human primates suggest that adverse childhood experiences may influence obesity risk. However, observations from human studies showed heterogeneous results. To address these inconsistencies, we performed Medline, PsycInfo and Embase searches till 1 August 2012 for articles examining the association between childhood maltreatment and obesity. We then conducted a meta-analysis of the identified studies and explored the effects of various possible sources of bias. A meta-analysis of 41 studies (190 285 participants) revealed that childhood maltreatment was associated with elevated risk of developing obesity over the life-course (odds ratio=1.36; 95% confidence interval=1.26–1.47). Results were not explained by publication bias or undue influence of individual studies. Overall, results were not significantly affected by the measures or definitions used for maltreatment or obesity, nor by confounding by childhood or adult socioeconomic status, current smoking, alcohol intake or physical activity. However, the association was not statistically significant in studies of children and adolescents, focusing on emotional neglect, or adjusting for current depression. Furthermore, the association was stronger in samples including more women and whites, but was not influenced by study quality. Child maltreatment is a potentially modifiable risk factor for obesity. Future research should clarify the mechanisms through which child maltreatment affects obesity risk and explore methods to remediate this effect.
Obesity is a prevalent global-health problem associated with substantial morbidity, impairment and economic burden. Obesity affects 17% of children and adolescents and 36% of adults in the United States,1, 2 but is also rapidly becoming a major public health concern in developing countries.3 Obesity is a risk factor associated with complex physiological abnormalities4 and with high risk of developing Type 2 diabetes, cardiovascular disease and functional disability,5, 6 thus influencing both direct and indirect health costs. Because most readily available forms of treatment are ineffective in the long term,7 it is essential to advance knowledge of obesity prevention by identifying potentially modifiable risk factors.
Childhood experiences may contribute setting life-long trajectories for obesity. Experimental research in animal models starting in the womb showed that brain regions regulating energy balance continue to develop till early postnatal life in response to environmental stimuli and endocrine signals.8, 9, 10 Similarly, the development of energy balance regulation in humans may be significantly influenced by environmental influences starting in the prenatal period and continuing throughout childhood.11, 12 Diet in childhood has long been investigated as an important determinant of later obesity.13, 14 More recently, studies in non-human primates suggested that stressful psychosocial experiences in childhood could also be associated with the development of obesity,15, 16 but evidence in humans remains heterogeneous.
The aim of this study was to examine comprehensively whether a prevalent, severe and often chronic childhood stressor, namely maltreatment, is associated with obesity risk. Our work builds upon a previous systematic review showing that individuals reporting interpersonal violence in childhood were at elevated obesity risk in most published studies,17 and upon meta-analytical research showing that adults who experienced psychosocial stress had small, longitudinal increase in adiposity.18 We performed a meta-analysis of studies investigating the association between childhood maltreatment and obesity over the life-course. We also explored the effects of various possible sources of artifact or bias on the results of the meta-analysis.
Materials and methods
We have attached a copy of our protocol for the systematic review and the PRISMA and MOOSE checklists as supporting information. An ethics statement was not required for this research.
In this meta-analysis, we included original, peer-reviewed full papers satisfying the following criteria: (i) definition of childhood adversities consistent with maltreatment (that is, physical abuse, sexual abuse, emotional abuse, physical neglect, emotional neglect or family conflict/violence) and occurring before age 18 years; (ii) definition of obesity based on weight and height (that is, categorical measures of body mass index (BMI) above a preset value, continuous measure of BMI or corresponding z-scores in studies of children) or waist circumference; (iii) inclusion of a control group with no maltreatment history and no obesity.
We searched Medline, PsycInfo and Embase databases for articles describing the relationship between childhood maltreatment (search terms: child* maltreatment, child* abuse, child* neglect, family conflict, early experience) and obesity (search terms: obesity, overweight, body-mass index, BMI, body size, adiposity, waist circumference) in human subjects, written in English and published by 1 August 2012.
Two authors independently extracted data from eligible articles. Inconsistencies were resolved in consensus meetings and checked with the authors of the primary studies when necessary.
We extracted adjusted effect sizes whenever possible, to measure the effects of maltreatment independent of the influence of potential intervening variables. Extracted data was converted to odds ratios effect sizes19 reflecting the probability of unfavorable outcomes, with odds ratios above 1 reflecting increased likelihood of obesity in individuals with a history of childhood maltreatment compared to non-maltreated individuals. Where only continuous outcomes were reported, risk of unfavorable outcomes was derived using validated methods.20 In addition to identifying and calculating effect sizes, we also collected and coded information about mean age of the sample, gender distribution, ethnicity, maltreatment assessment (retrospective, prospective), maltreatment measure (questionnaire, interview, records), maltreatment definition (physical abuse, sexual abuse, emotional abuse, physical neglect, emotional neglect), obesity measure (self-report, examination), obesity definition (obesity cutoff, BMI scores, waist circumference) and covariates included in the adjusted effect size (child socioeconomic status, adult socioeconomic status, current smoking, current alcohol intake, current physical activity, current depression). Heterogeneity between studies was tested with Cochran’s Q-test.21 We carried out meta-analyses using random-effects models, which include both sampling and study-level errors. Meta-analyses were performed using the metan programs applied in STATA 12 (StataCorp, College Station, TX, USA). To measure the impact of maltreatment on obesity, we estimated the number needed to treat,22 defined as NNT=1/((1−OR) × Po), where Po is the prevalence of obesity in the population, following recommendations by the Cochrane Collaboration.23 The NNT indicates the number of maltreatment cases that need to be treated (or prevented) to avoid one case of obesity. Additional analyses explored the effects of various possible sources of artifact or bias on the results.
We assessed the presence of publication bias visually by funnel plot24 and formally by its direct statistical analog, Begg’s adjusted rank correlation test,25 using the metabias program applied in STATA. In the presence of significant rank correlation tests, we adopted a nonparametric ‘trim-and-fill’ method,26 using the metatrim program applied in STATA to examine the extent to which publication bias may have contributed to the meta-analytical results.
We assessed the undue influence of individual studies on the overall meta-analysis results by testing changes in the estimate across permutations where each study was omitted in turn using the metaninf program applied in STATA.
We assessed the sensitivity of meta-analytical results to different definitions and measures of childhood maltreatment and obesity through subgroup analyses.
We tested for meta-analytical evidence of confounding or mediation by childhood/parental socio-economic status, adulthood socioeconomic status, current smoking, current alcohol intake, current physical activity, and current depression through subgroup analyses contrasting effect sizes that were adjusted or unadjusted for a specific potential intervening variable.27
Finally, we assessed the moderation of the meta-analytical results by gender distribution (% female), ethnicity distribution (% white), mean age of the sample and study quality through meta-regression using the metareg program applied in STATA. The quality of epidemiological studies was assessed with the Newcastle–Ottawa Scale,28 which has been recommended by the Cochrane collaboration23 and has been used in previous publications.29
The association between childhood maltreatment and obesity was tested in 44 data sets from 41 studies with a total of 190 285 participants (Table 1).30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70 We identified significant heterogeneity across studies (I2=74.8%, Q=170.46, d.f.=423, P<0.001) and performed analysis using a random-effects model. A meta-analysis of these studies showed that, compared to those without a history of childhood maltreatment, maltreated individuals were more likely to be obese (odds ratio (OR)=1.36; 95% confidence interval (CI)=1.26–1.47) (Figure 2). We performed additional analyses to explore the effect of various possible sources of artifact or bias on these results.
We examined the possibility of publication bias. The funnel plot showed asymmetrical distribution of the studies and the Begg’s adjusted rank correlation test was significant (z=2.16, P=0.030; see Supplementary Figure S1), suggesting the possibility of publication bias. However, trim-and-fill procedure with 12 imputed studies achieved a combined effect size that was smaller but still significant (OR=1.21; 95% CI=1.12–1.32), suggesting that potential publication bias did not significantly affect the meta-analytical results.
We examined the possibility of undue influence of individual studies on the overall meta-analysis results. We carried out 44 permutations, omitting each study in turn (see Supplementary Figure S2). There was no significant heterogeneity across permutations (I2=0.0%, heterogeneity χ2=1.73, d.f.=43, P=1.000). This suggested that no individual study had undue influence on the overall meta-analysis results.
We examined the sensitivity of meta-analysis results to different definitions and measures of childhood maltreatment and obesity (see Figure 3a). Childhood maltreatment was associated with obesity, regardless of whether maltreatment was assessed through retrospective report (OR=1.34; 95% CI=1.24–1.45) or prospective observation (OR=1.60; 95% CI=1.18–2.17). Childhood maltreatment was associated with obesity, regardless of whether maltreatment was measured through questionnaire (OR=1.35; 95% CI=1.25–1.47), interview (OR=1.34; 95% CI=1.06–1.69) or records (OR=1.62; 95% CI=1.08–2.44). Childhood maltreatment was associated with obesity when maltreatment was defined as sexual abuse (OR=1.43; 95% CI=1.27–1.62), physical abuse (OR=1.29; 95% CI=1.15–1.43), physical neglect (OR=1.29; 95% CI=1.02–1.62) or emotional abuse (OR=1.24; 95% CI=1.10–1.40), but not when it was defined as emotional neglect (OR=1.21; 95% CI=0.92–1.60). Childhood maltreatment was associated with obesity, regardless of whether obesity was measured through self-report (OR=1.43; 95% CI=1.26–1.61) or examination (OR=1.33; 95% CI=1.20–1.48). Finally, childhood maltreatment was associated with obesity, regardless of whether obesity was defined through categorical measures of body mass (OR=1.47; 95% CI=1.31–1.65), continuous measure of body mass (OR=1.30; 95% CI=1.24–1.36) or continuous measure of waist circumference (OR=1.28; 95% CI=1.11–1.46).
We explored the potential confounding or mediation of meta-analysis results by key variables (see Figure 3b). Childhood maltreatment was associated with obesity, regardless of whether the study estimates were or were not adjusted for childhood socioeconomic status (unadjusted: OR=1.41, 95% CI=1.30–1.53; adjusted: OR=1.30, 95% CI=1.07–1.57), adult socioeconomic status (unadjusted: OR=1.46, 95% CI=1.26–1.68; adjusted: OR=1.38, 95% CI=1.24–1.54), current smoking (unadjusted: OR=1.42, 95% CI=1.29–1.55; adjusted: OR=1.26, 95% CI=1.08–1.47), current alcohol intake (unadjusted: OR=1.41, 95% CI=1.29–1.54; adjusted: OR=1.25, 95% CI=1.04–1.49) and current physical activity (unadjusted: OR=1.40, 95% CI=1.28–1.54; adjusted: OR=1.24, 95% CI=1.09–1.41). In contrast, the association between childhood maltreatment and obesity was nonsignificant when the estimate was adjusted for current depression (unadjusted: OR=1.43, 95% CI=1.31–1.55; adjusted: OR=1.10, 95% CI=0.95–1.27).
Finally, we examined the potential moderation of meta-analysis results by key variables. Meta-regression analyses showed that the association between childhood maltreatment and obesity was stronger in studies including more women (see Figure 4a; k=43, b=0.003, P=0.052) and more whites (see Figure 4b; k=28, b=0.004, P=0.047). Although meta-regression analysis showed that mean age of the sample did not moderate the association between childhood maltreatment and obesity (see Figure 4c; k=43, b=0.003, P=0.403), this association was significant in studies of adults (OR=1.38, 95% CI=1.28–1.50) but not in studies of children and adolescents (OR=1.13, 95% CI=0.92–1.39; see Figure 3b). On the basis of recent US prevalence estimates of obesity,2 NNT was 7 (95% CI=6–10) for adult obesity. Finally, study quality did not moderate the association between childhood maltreatment and obesity (see Supplementary Table S1 and Figure 4d; k=44, b=0.022, P=0.649).
This meta-analysis addressed the possible developmental origins of heterogeneity in risk for obesity. Results suggest that childhood maltreatment predicts obesity, largely regardless of the measures and the definitions used and independent of several potential intervening variables. Stressful psychosocial experiences in childhood might thus be conceptualized as potentially modifiable risk factors for obesity. Assuming causality, results suggest that prevention or effective treatment of seven cases of childhood maltreatment could avoid one case of obesity in adulthood. Yet, it is unclear if and how the effect of maltreatment on obesity could be modified through intervention.
These results should be evaluated in the context of several potential limitations.
First, the results may be influenced by selective publication of positive studies. Although we identified evidence of publication bias, the trim-and-fill procedure suggested that this bias was unlikely to significantly affect the meta-analysis results.
Second, the results may be influenced by undue effect of individual studies. However, permutations that serially excluded single studies in turn suggested that this was not the case.
Third, the results may be sensitive to how exposure and outcome were ascertained. However, subgroup analyses showed that childhood maltreatment was associated with obesity, regardless of the measures and definitions used. Of note, results based on retrospective self-reports of maltreatment were similar to results based on prospectively collected and objectively assessed data, minimizing the chances that results were simply due to recall bias.
Fourth, the results may be influenced by reverse causality, that is, they might have emerged simply because children who were obese in the first place were more likely to experience subsequently maltreatment or to be classified as maltreated. However, we noted that the association between maltreatment and obesity was nonsignificant in studies of children and adolescents (Figure 3b). Furthermore, studies that have examined within-individual changes in body mass showed that childhood maltreatment was associated with progressive increase of body mass over time, after maltreatment had occurred.58, 60, 71 Therefore, it is unlikely that the meta-analysis results are simply due to reverse causality.
Fifth, the results may be influenced by confounding bias. However, studies yielded similar results, regardless of whether they statistically adjusted for key potential intervening variables, such as childhood/parental socioeconomic status, adult socioeconomic status, smoking, alcohol intake or physical exercise. Therefore, these variables were unlikely to explain the association between childhood maltreatment and obesity. In contrast, this association was nonsignificant among studies adjusting for depression. Because childhood maltreatment is an important predictor of depression over the life-course29 and depression is prospectively associated with obesity,72 this finding suggests that depression may be a key factor mediating the effect of childhood maltreatment on obesity.
Sixth, the results may be influenced by characteristics of the samples examined. Consistent with the experimental findings in non-human primates,16 we noted that women might be more vulnerable to the effects of maltreatment on obesity. Furthermore, white individuals may be more vulnerable to the effect of maltreatment on obesity. However, because of the variability in the strength of association between BMI and body composition across different ethnic groups,73 these findings might have alternatively arisen because of the inadequacy of BMI to capture the effect of maltreatment on body composition in non-white ethnic groups. In contrast, we did not observe a linear association between mean age of the sample and effect sizes.
Seventh, the results may be influenced by study quality. However, differences in quality ratings were not associated with differences in effect sizes across studies.
Finally, the results may be influenced by residual biases (for example, measurement error, confounding by unmeasured genetic or prenatal factors), which cannot be excluded in meta-analyses of observational studies. Therefore, it is comforting that these results sit within a nomological network that includes coherent predictions from childhood maltreatment to obesity-related outcomes, such as Type 2 diabetes and cardiovascular disease,74 as well as experimental evidence from animal models linking early life stress to obesity.15, 16
Despite these potential limitations, the study results have important implications in several areas.
With regard to research implications, our results suggest that, like nutritional stressors in early life,11, 12 childhood maltreatment may be associated with a ‘thrifty’ phenotype characterized by increased energy intake and storage, and/or reduced energy expenditure. Because they act on developing, plastic systems,8, 9, 10 childhood stressors like maltreatment may exert enduring effects, presumably through epigenetic mechanisms.75, 76 Future research will need to clarify the biological mechanisms underlying these clinical findings, for example, by directly testing the adipogenic effects of neurobiological, endocrine and immune abnormalities previously described in maltreated individuals.77 On the one hand, these abnormalities might increase energy intake. For example, smaller volume of the prefrontal cortex in maltreated vs non-maltreated individuals might be associated with functional impairment in inhibitory control of feeding behavior,78 and maltreated individuals have shown high risk of disordered eating in some clinical studies.79 Furthermore, chronically elevated cortisol levels may increase the salience of pleasurable/compulsive activities, and lead to overeating in the face of stressful experiences to reduce further hypothalamic-pituitary-adrenal axis activation.80, 81 Stress-related food-reward dependence might emerge because of the effect of cortisol,80, 81 leptin,82, 83 endocannabinoids84, 85 and opioids on dopamine transmission. This ‘food addiction’ mechanism86, 87 may be supported by the broader clinical evidence that maltreated individuals are at high risk of developing addiction problems.88 Finally, chronically elevated cortisol levels could favor fat accumulation by inhibiting lipolysis.80, 81 On the other hand, biological abnormalities linked to maltreatment might reduce energy expenditure. For example, elevated inflammation levels described in maltreated individuals89 might induce fatigue and reduced activity,90, 91 consistent with the clinical evidence that maltreated individuals are at high risk of depression.29 A better understanding of the mechanisms of biological embedding of childhood maltreatment may be particularly useful with regard to obesity, because we observed that maltreatment predicted obesity in studies of adults but not in studies of children and adolescents. It should be noted that large differences in effect sizes across studies of children and adolescents might be at least partly explained by differences in pubertal (for example, Tanner) stages,92 and most studies did not provide data to test this hypothesis. Furthermore, longitudinal inferences from cross-sectional observations at different ages may have limited validity.93 However, results so far suggest the presence of an ‘incubation period’94 between exposure and outcome, during which it might be possible to remediate the biological changes linked to maltreatment before the onset of clinical end points.
With regard to clinical implications, our results suggest that a comprehensive assessment of obese patients should include evaluation of childhood psychosocial environment and experiences. Studies that have examined within-individual changes in body mass showed that childhood maltreatment may be associated not only with obesity occurrence but also with its persistence and exacerbation over time,58, 60, 71 similarly to what has been described for depression.29 This evidence suggests that interventions targeting childhood maltreatment might improve outcomes of obesity in maltreated individuals. However, it is currently unknown if clinical interventions in maltreated individuals could remediate these adipogenic effects. Future clinical trials will be needed to test whether currently available interventions targeting childhood maltreatment and its psychological sequelae95 could modify obesity risk. For example, it is unknown if interventions effective in preventing the recurrence of child maltreatment, such as the parent–child interaction therapy,96, 97 have any impact on body mass, eating behaviors and associated biological abnormalities. Similarly, it is unknown if interventions effective in treating maltreatment-related post-traumatic stress disorder, such as trauma-focused cognitive behavioral therapy,98, 99 could help prevent or reduce obesity, disordered eating and associated biological abnormalities in maltreated individuals.
With regard to public health implications, our results support and expand previous findings suggesting that—in addition to the significant genetic and prenatal influences—body mass is significantly shaped by experiences in childhood.8, 9, 10, 11, 12, 15, 16 Because childhood psychosocial experiences influence obesity risk, obesity should be seen not only as a clinical problem but also as a societal problem. Public health solutions to this problem might be important for at least two reasons. On the one hand, child maltreatment is prevalent in the population. Nearly one in five children experience maltreatment,100 with cumulative prevalence of 5–30% for any sexual abuse (1–10% for penetrative sexual abuse), 5–35% for physical abuse, 4–9% for emotional abuse, 6–12% for neglect and frequent co-occurrence of different forms of maltreatment.101 On the other hand, although not all maltreated children will develop obesity, it is unclear how to identify those who will and if clinical interventions could remediate the developmental liability to obesity described in this subgroup. Future studies will be needed to test whether universal or targeted preventive interventions that are effective in reducing the occurrence of child maltreatment95 are also effective in reducing the incidence of obesity in the population. Improving the psychosocial environment children live in might contribute to the primordial prevention of obesity and its consequence.102
Dr AD was supported by a Clinical Lectureship from the London Deanery, by the National Heart, Lung and Blood Institute of the US National Institutes of Health under Award Number R21HL109396, and by the UK Medical Research Council's grants G1002190 and G9806489. The content is solely the responsibility of the authors and does not necessarily represent the official views of the US National Institutes of Health or the UK Medical Research Council. No funding bodies had any role in study design, data collection and analysis, decision to publish or preparation of the manuscript. Dr AD had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
About this article
Supplementary Information accompanies the paper on the Molecular Psychiatry website (http://www.nature.com/mp)