Objective. Stress in childhood relates to both childhood depression and elevated adult body mass index (BMI), a measure of relative overweight. However, there are limited data on the association between major depression in childhood and BMI in adulthood. The current study examines this association.
Method. Children 6 to 17 years old with major depression (n = 90) or no psychiatric disorder (n = 87) were identified at Columbia Presbyterian Medical Center and followed up 10 to 15 years later. Psychiatric status at intake and follow-up was assessed via standardized psychiatric interviews. BMI during adulthood was recorded so that the association between depression and BMI could be considered over time.
Results. Participants with childhood major depression had a BMI of 26.1 ± 5.2 as adults, compared with a BMI of 24.2 ± 4.1 in healthy comparisons (t175 = 2.7). This association could not be explained by a number of potentially confounding factors, including age, gender, cigarette or alcohol use, social class, and pregnancy or medication history. Although poverty during adulthood also predicted adult BMI, both the association between poverty and adult BMI (t152 = 2.9), as well as between childhood depression and adult BMI (t152 = 2.2) were significant in a multivariate model. Finally, duration of depression between childhood and adulthood also emerged as a predictor of adult BMI.
Conclusions. Depression during childhood is positively associated with BMI during adulthood. This association cannot be explained by various potential confounding variables and may develop over time as children pass into their adult years. body mass index, depression, children, adolescents.
Current public health initiatives recognize the importance of identifying early life influences on adult body mass index (BMI), a measure related to obesity.1–7 This emphasis derives from the limited availability of effective treatments for obesity, coupled with the current view of obesity as a developmental condition.5,8–15 Preventive interventions might reduce the risk for obesity by targeting childhood factors that predict increases in BMI. Such prevention efforts require sound understanding of childhood correlates of adult BMI, a focus of recent epidemiologic studies.1–7
Recent findings note that adult BMI can be predicted by social factors measured during childhood, including poverty, social stress, and harsh parenting.8–11,14 Similarly, retrospective studies in adults suggest that various childhood social stressors may impact on the risk for obesity and other adverse medical consequences.12,13 Each of these social factors, in turn, exhibits strong and consistent associations with another potential correlate of adult BMI, childhood psychopathology.16–19Epidemiologic studies, however, provide weak and inconsistent evidence of cross-sectional associations between psychopathology and BMI in either childhood or adulthood.20–30 Nevertheless, as in studies of social factors, recent studies suggest that there may be more powerful longitudinal than cross-sectional associations between childhood psychopathology and adult BMI.31
A recent study from our group examined relationships between psychopathology in childhood and BMI in adulthood among individuals randomly selected from the community.31 Childhood symptoms of conduct disorder predicted higher adult BMI. Although childhood symptoms of major depression also predicted adult BMI in the sample while not controlling for conduct disorder symptoms, the association between BMI and depression was nonsignificant when covarying for level of childhood conduct disorder symptoms. Regardless, given the tie between social factors and conduct disorder,17,19 these findings are consistent with the above-noted reports documenting the impact of childhood social factors on both childhood psychiatric risk and adult BMI.
The current study is designed to address limitations in this previous study.31 Very few participants in the epidemiologic sample had any history of psychiatric contact, and relatively few participants met criteria for clinical diagnoses, particularly for depression. As a result, we did not examine the association between any childhood psychiatric diagnosis and adult BMI. Hence, the extent to which the previous findings applied to cases typically encountered in clinical settings remains unclear. Moreover, this limitation was compounded by the use of lay interviewers as opposed to clinicians to obtain psychiatric symptom ratings. Given recent efficacy data for pharmacological and psychotherapeutic treatments in clinically defined cases of childhood major depression,32,33 it is particularly important to examine associations between childhood depression and adult BMI.
A second limitation in the previous study31 related to the fact that only current depressive symptoms at 2 time points were assessed. As a result, the degree to which chronic depression during childhood might exert a cumulative effect on BMI across development could not be examined. The current report addresses these limitations by examining the relationship between a clinical diagnosis of major depression in childhood and BMI in adulthood. Cases were identified from a clinical setting, diagnosed independently by 2 psychiatrists, and reassessed blind to all other data as adults. Both current depressive symptoms as well as symptoms between childhood and adulthood were rated.
The overall design of the current study was a prospective case–control study that involved a follow-up assessment conducted 10 to 15 years after an initial assessment. The eligible sample of probands for the initial clinical follow-up comprised a total of 199 children and adolescents, ages 6 to 17 years, who presented with a diagnosis of major depression.
The eligible sample of comparisons comprised a total of 176 children of similar ages with no lifetime history of a psychiatric disorder. From among a total of 375 participants, data on psychiatric outcome were available for a total of 284 participants, 156 probands and 128 comparisons. The current report sought to relocate participants with baseline and follow-up psychiatric data to acquire data on BMI. BMI data ultimately were available for a total of 177 participants, comprising 90 probands and 87 comparisons. Previous data on initial and follow-up psychiatric diagnosis have been previously described.32–44 This study was approved by the institutional review board, and all families provided informed consent for participation in all phases of the study.
Selection of Probands
Probands with a diagnosis of major depression were selected from a psychiatric clinic at Columbia Presbyterian Medical Center, when they presented for treatment of feeling sad or depressed. Potential cases were screened and then reevaluated in 2 weeks by 2 independent psychiatrists, using the Schedule for Affective Disorders and Schizophrenia for School-age Children.40 Probands were included in the current study only if they met criteria for major depression at both assessments. Interrater and test–retest reliabilities of this assessment were both satisfactory, as has been reported previously.40
Participants were excluded from the study if: 1) depressive symptoms were temporally related to medical illness or the use of prescribed or illicit substances; 2) participants were above the 95th percentile on weight/height ratios (this variable served as an exclusion criteria when the cohort was initially assembled because of the research team's interest in neuroendocrinological aspects of pediatric major depression); 3) IQ was lower than 70; and 4) participants met criteria for anorexia nervosa, autism, or schizophrenia.
Selection of Comparisons
Healthy participants were recruited at the same time through local advertisements and rosters of local school classrooms, as described in previous reports.34–39 Comparisons had been actively recruited from the same community and at the same time as the probands using various methods. An effort was made to randomly sample demographically matched groups of psychiatrically healthy children and adolescents from community schools, but this procedure resulted in relatively low participation rates. As a result, school-based samples of comparison participants were supplemented with samples of volunteers recruited through advertisements. As shown in Table 1, probands and comparisons were matched on age and gender. Only children and adolescents with no current or past psychiatric history and no use of prescribed or illicit substances were eligible for participation. These determinations were made by structured diagnostic assessments by a psychiatrist. Comparisons were also required to meet the weight/height and IQ exclusion criteria listed above.
Psychiatric Follow-Up Assessment
As with the initial psychiatric assessment, diagnostic information was collected from both the participant and an adult informant, who was typically a parent. This assessment used the Schedule for Affective Disorders and Schizophrenia-Lifetime Disorders to assess mood, anxiety, behavioral, and psychotic disorders.46,47 Interviewers were each clinically trained and experienced in the standardized assessment of psychopathology. Information on demographic variables and other potential correlates of BMI were also obtained. In addition, the duration of depression across the follow-up was estimated from this assessment. This estimation was based on dates/duration of depressive episodes between the initial and follow-up assessment. Changes in eating behavior associated with each depressive episode were also assessed. Specifically, questions were asked about changes in appetite, the urge to overeat, increases or decreases in food intake, as well as increases or decreases in body weight.
Assessment of BMI
BMI is frequently used to quantify overweight in epidemiologic research, partially because of the ease with which it is measured.1–7,48 Despite some disadvantages as a measure to assess relative overweight, BMI is consistently related to morbidity from illness associated with obesity.1–7
At study intake, height and weight were measured in a subset of participants as part of a comprehensive physical examination conducted by a pediatrician using a standard scale. These participants were participants in studies on the biology of childhood depression.36,41 Given that childhood BMI depends on age, childhood BMI is treated both as raw data and as percentile data, using age- and gender-specific norms from the First National Health and Nutrition Examination Survey.49 In adulthood, height and weight were assessed by self-report. BMI was calculated (BMI = weight [in kg] divided by height squared [in m2]). Although self-reported BMI is measured with some bias, validation studies suggest that this bias is unlikely to affect conclusions about associations between BMI and psychopathology, particularly in longitudinal studies.50–52
Assessment of Potential Confounding Variables
The relationship between BMI and various potential confounding variables was examined. These variables included: age, gender, and ethnicity. The relationship between BMI and social class was also examined using various indices assessed in adulthood, including Hollingshead ratings, household income, and educational attainment. A series of items on physical health status were assessed in adulthood. As described elsewhere,42,44 these items were summed to create continuous scales that indexed physical health from childhood through early adulthood. Number of pregnancies was also assessed in adulthood by self-report, given evidence of an association between number of pregnancies and obesity.31 Finally, lifetime use of cigarettes, alcohol, and medications was assessed by self-report, as described in detail elsewhere, and served as potential confounders.42,44 For cigarette use, participants were classified based on assessments in adulthood as having ever versus never smoked daily for at least 1 month or longer. For alcohol use, participants were classified as having ever versus never used alcohol regularly if they drank at least 3 days per week for 1 month or longer. Psychotropic medication use was quantified as the number of 6-month periods of exposure.
All results rely on 2-tailed statistical tests with α = 0.05. Bivariate associations among BMI, psychiatric status, and various potential confounding variables were examined using unpairedt tests, Pearson correlation coefficients, or odds ratios. Analysis of covariance or logistic regression models were fit to consider associations between childhood depression and BMI in either childhood or adulthood, while controlling for the effects of potential confounding variables. These analyses treated BMI both as a continuous and as a dichotomous variable, using standard BMI-derived cutpoints to categorize participants. For continuous analyses, both raw BMI values and log-transformed values were used, generating similar results. To facilitate comparisons with previous studies, results for raw values are presented. Based on National Institutes of Health guidelines,53 a BMI of 25.0 was used to define overweight status. Only 20 individuals in the sample exhibited a BMI >30.0, a standard definition of obesity.53 This included 13 of 156 probands (8.3%) and 7 of 128 comparisons (5.5%). Analyses did not consider predictors of obesity, because results from these analyses are statistically unstable because of the small numbers of obese individuals. Finally, because previous studies have noted gender specificity in risk factors for obesity, we also fit gender interactions for all variables.
As noted above, BMI data were acquired from 90 probands and 87 comparisons. Table 1 summarizes data in these 177 participants on each of the variables examined in the current study. Available data in participants who did not provide adult BMI data are also presented inTable 1. On all 3 measures of social class, there were tendencies for participants without BMI data to come from lower social strata than subsamples with BMI data. This appeared somewhat more likely among comparisons than among probands, although there were no statistically significant differences in any social class measure between any 2 groups with and without BMI data. Although probands were significantly older than were controls, age differences were generally small in magnitude. Probands also had significantly more pregnancies than did comparisons.
Association Between Childhood Depression and Adult BMI
Table 2 compares BMI data in adulthood between participants who had been either depressed or psychiatrically healthy in childhood. As shown in Table 2, as adults, participants with childhood depression were on average 1.9 ± 4.7 kg/m2 larger in BMI than were participants who had been healthy in childhood (t175 = 2.7; P < .01).
The association between childhood depression and adult BMI was also examined after dichotomizing participants based on their BMI in adulthood to provide an index of overweight. In a bivariate logistic model, childhood depression predicted a twofold increased risk for adult overweight status (odds ratio = 1.9; 95% confidence interval = 1.02–3.4; Wald X2 = 3.8;P < .05).
Potential Confounding Variables
Table 2 also presents the relationship between various potential confounding variables and childhood depression status. As shown inTable 2, relative to healthy comparisons, formerly depressed children were older when they provided BMI data, exhibited lower levels of physical health, lower adult income, and higher rates of cigarette, medication, and alcohol use. Although there were no differences between groups in gender, models were fit including gender and gender-by-depression status interactions as predictors of BMI, in light of previous findings on gender-specific associations between BMI and psychopathology.31 No associations or interactions with gender emerged.
Table 3 examines the relationship between each potential confounding variable and BMI in adulthood. As shown inTable 3, income at follow-up was the only potentially confounding variable that showed a relationship with adult BMI. Other measures of social class, including educational attainment and Hollingshead ratings, correlated with adult income, but only adult income exhibited an association with adult BMI. When the association between childhood depression and adult BMI was modeled while covarying for social class, both the association between depression and BMI (t152 = 2.2; P < .05), as well as the association between social class and BMI (t152 = 2.9; P < .05), were significant.
In bivariate analyses, household income in adulthood was the only potential confounder exhibiting even a trend for an association with overweight. Hence, a logistic model was also fit predicting overweight from childhood depression and household income. In this model, both childhood depression (odds ratio = 1.7; 95% confidence interval = 0.9–3.2; Wald X2 = 2.7;P < .10) and poverty (income below $9999; odds ratio = 2.6; 95% confidence interval = 0.8–7.6; Wald X2 = 2.6; P < .10) predicted increased risk for overweight status at trend levels.
Total lifetime physical health problems also showed a trend for an association with adult BMI, as shown in Table 3. When both childhood depression and physical health problems were included in a model, only the association with childhood depression (t174 = 2.6; P < .01) but not with health problems (t174 = 1.2; P = .15) remained. BMI did not differ among participants who were (n = 11; BMI = 25.3 ± 4.8) or were not (n = 166; BMI = 25.2 ± 4.8) currently depressed at the time of the adult assessment (t175 = 0.1). Moreover, BMI did not relate to changes in eating patters associated with depressive episodes. However, there was evidence of an association with duration of depression from childhood into adulthood. In an analysis of covariance examining predictors of adult BMI, both low household income (F = 4.6; P < .05) and duration of depression (F = 4.4; P < .05) predicted higher adult BMI. In this analysis, the association between childhood depression and adult BMI was reduced to a trend (F = 3.6; P = .058).
Finally, the association between childhood depression and adult BMI persisted in analyses that controlled for any other potential confounding variables in Table 3 showing weaker associations with BMI. Beyond these potential confounders in Table 3, we also considered the potential impact of childhood conduct disorder on adult BMI, based on previous findings.31 Only 8 of the 90 depressed probands had a history of conduct disorder. Although there was no statistical difference in BMI between probands with and without conduct disorder, this statistical contrast possesses limited statistical power.
Childhood BMI, Childhood Depression, and Adult BMI
BMI data in both childhood and adulthood were only available for 65 participants, including 46 depressed participants and 19 psychiatrically healthy participants. Although BMI in adulthood was larger in formerly depressed children, as shown in Table 2, the raw BMI in childhood was comparable between depressed (19.4 ± 3.6) and healthy participants (19.7 ± 3.5;t63 = 0.25; P = .80). Similarly, using age/gender-specific normative data from First National Health and Nutrition Examination Survey, formerly depressed children (57 ± 29 percentile) did not differ from healthy children (52 ± 63 percentile) in BMI percentiles (t63 = 0.25; P = .80). Moreover, childhood depression predicted adult BMI when BMI in childhood as well as adulthood household income were included in a statistical model as covariates (F[1,60] = 4.5;P < .05). In this model, adult BMI was predicted by both childhood depression (t60 = 2.9;P < .01) and childhood BMI (t60 = 2.5; P < .05) but not by adulthood household income (t60 = 1.1; P > .50). The same associations emerged when percentile BMI from childhood as opposed to absolute BMI was used. Similarly, when duration of depression was included as a covariate, adult BMI was still predicted by both childhood depression (F[1,52] = 5.1;P < .05) and childhood BMI (F[1,52] = 6.5; P < .01) but by neither adulthood household income (F[1,52] = 0.4.; P = .50) nor duration of depression (F[1,52] = 1.5; P = .22). Finally, in logistic models predicting overweight, there was no relationship between childhood BMI and adult overweight status (Wald X2 = 0.2; P > .50).
As reviewed elsewhere,20 most studies find weak or inconsistent cross-sectional relationships between psychopathology and either categorical measures of obesity or continuous measures, such as BMI. There also were no cross-sectional associations in the current study between depression and BMI in either childhood or adulthood, although the exclusion of obese children from the study may influence this finding. Few studies have examined prospective relationships between psychopathology in childhood and BMI in adulthood. None have considered the impact of major depression in childhood or the duration of depressive symptoms across development on BMI in adulthood. In the current study, a clinical diagnosis of major depression in childhood was associated with a higher BMI during adulthood. Moreover, the overall duration of depressive symptoms from childhood into adulthood emerged as a particularly strong predictor of elevated BMI in adulthood.
If the current findings are replicated, they would suggest that childhood depression, much like other indicators of childhood stress, predicts adult BMI. At least 3 plausible mechanisms could account for such an association between adult BMI and childhood depression. First, elevated BMI during childhood could directly contribute to the risk for depression in childhood, while also influencing adult BMI. Such an association between childhood BMI and childhood depression could arise because of the social stigma associated with obesity. Childhood BMI did relate to adult BMI in the current study, as in previous studies.5,7 Nevertheless, this mechanism is unlikely to account for the association between childhood depression and adult BMI found in the current report, because children with obesity were explicitly excluded from the study. Moreover, in the remaining nonobese children, there was no evidence of any cross-sectional association between BMI and depression in childhood, despite the presence of an association between childhood depression and adult BMI.
Second, childhood depression could contribute to elevated BMI by affecting factors associated with elevated BMI during adulthood. For example, depression could affect diet or activity levels that could, in turn, lead to elevations in BMI. Similarly, medication or illicit substance use in depressed individuals could affect adult BMI. Although there was no evidence that either diet or medication use related to adult BMI in the current study, the current study only cursorily assessed alterations in diet. Effects of depression on other unmeasured dietary factors as well as on activity levels could play a role. Third, both childhood depression and adult BMI could be influenced by third factors. These third factors could include both social and biological risk factors for overweight status that previously have been linked to depression, as outlined below.
Longitudinal data may provide key insights on the potential role played by each of the above-noted mechanisms in linking childhood psychopathology and adult BMI. We possessed insufficient data on BMI during childhood in the current as well as the previous study to support any definitive conclusions on cross-sectional associations between childhood BMI and psychopathology.31 However, in the cases in which data were available at 2 time points, childhood depression predicted an increase in BMI across development. Nevertheless, given the dearth of data on childhood BMI, the precise age remains to be established during which an association might emerge between childhood psychopathology and adult BMI.
In the current study, the degree to which participants remained depressed across development represented a particularly salient predictor of adult BMI. This is consistent with the lack of a cross-sectional association between BMI and depression. This finding may suggest a particular role for cumulative influences on BMI over time, either from a direct effect of depression on BMI or through the effects of common third factors on both depression and BMI. Studies examining this issue might further consider the degree to which cross-sectional relationships between psychopathology and BMI vary as a function of developmental stage. Finally, studies might further consider the degree to which childhood psychopathology predicts an increase in BMI between childhood and adulthood.
There are limitations to our findings that might be addressed in future studies. First, as noted above, it is crucial to obtain more data on childhood BMI and various forms of psychopathology in samples followed longitudinally. Such data will stimulate insights on the mechanism behind a potential association between childhood depression or other psychopathologies and adult BMI. Second, we had no data on family history of obesity. Because both psychopathology and obesity are familial conditions,7,19,35,37 family history of obesity could play a role in the association between childhood psychopathology and adult BMI. Although minimal research examines the association between family history of depression and obesity, the current findings linking childhood depression to adult BMI emphasize the need for studies in this area. If such research did in fact reveal an association between family history of depression and obesity, this would raise questions as to whether the association between childhood depression and adult BMI might mediate the association between family history and obesity. Moreover, given data on therapeutics,32,33 treatment of childhood depression could provide an avenue for reducing the risk for obesity in children born to obese adults. Given the paucity of data coupled with the potential implications for treatment, research that simultaneously examines associations among childhood depression, adult BMI, and family history of depression may provide insights on possible mechanisms for the associations found in the current report.
Third, the overall difference between early adult BMI in formerly depressed as opposed to healthy participants was moderate. As a result, childhood depression should perhaps best be viewed as only one of many potentially modifiable childhood predictors of adult BMI. Nevertheless, the association between childhood depression and adult BMI in the current study was comparable in magnitude to those between childhood and adult BMI, as well as between poverty and adult BMI, associations thought to carry considerable clinical significance.
Fourth, there was considerable attrition in the sample with complete BMI data. As considerable data were available for >80% of the sample, the study could assess the possible impact of attrition. On most measures, there were small differences between participants with and without adult BMI data. There was some evidence of particularly high degrees of attrition in participants from the lower social strata, which could bias our results in various ways. Most importantly, participants in epidemiologic studies with adverse outcomes, including poverty, are typically more difficult to locate. As a result, the current findings, as with previous findings,42–44 should be interpreted against the losses to follow-up. Specifically, because of these types of attrition-related biases, the current findings may not accurately capture the true magnitude of any association between childhood depression and adult BMI.
Finally, the use of self-report BMI represents another potential limitation. For example, because Taylor and Brown54 note that depression relates to individual's self-perceptions, depression could affect self-report of BMI. Nevertheless, previous studies suggest that this is unlikely to account for the current results.50–52 Moreover, to the extent that such a factor were to be involved, one would expect this to produce association with current depression, given that depression's effect on perception is state-dependent. Associations with current depression were not found in the current study.
Despite these limitations, there are reasons to believe that our findings may be worthy of additional study. The direction and the magnitude of associations between BMI and psychopathology in the current study resemble those in our previous epidemiologic study, although associations were stronger for conduct disorder as opposed to depression in that study.31 Moreover, consistent with other studies, the oft-noted association between BMI and poverty emerged in the current study, although it did not account for the relationship between BMI and depression. Various other forms of social adversity in childhood also predict adult BMI, independent of any effect on childhood BMI.8–15 Given the strong tie between childhood adversity and depression,16,19 the longitudinal association with BMI in the current study is consistent with studies on associations between childhood adversity and BMI over the lifespan.8–15
A link between childhood depression and adult BMI also supports findings from recent neurobiological investigations. Emerging evidence suggests that brain systems that are either affected by stress or which moderate an organism's response to stress play a role in disorders of both mood and weight regulation. Specifically, potential common neurobiological abnormalities involved in disorders of mood and weight regulation include deficits in the serotonin system and the hypothalamic-pituitary-adrenal axis, as well as deficits in the brain circuits that underlie hedonic regulation.15,55–59
The current study found that depressed children exhibit a larger BMI as adults than do nondepressed comparisons. This relationship was found in a subsample of participants who did not differ on childhood BMI. This relationship could not be attributed to a number of potential confounding variables. Given the treatable nature of childhood depression,32,33 replications of this finding may inform current efforts to address growing rates of obesity through developmentally oriented interventions.5,10,13
This work was supported by Scientist Development Award for Clinicians K20-MH01391 from the National Institute of Mental Health (to D.S.P.), National Institute of Mental Health Grant MH50666 and NARSAD Established Investigator Award (to M.M.W.), and an Aaron Diamond Post-Doctoral Research Fellowship Award (to R.B.G.).
- Received March 22, 2000.
- Accepted August 8, 2000.
Reprint requests to (D.S.P.) New York State Psychiatric Institute, 1051 Riverside Dr, Unit 74, New York, NY 10032. E-mail:
- BMI =
- body mass index
- Yao CH,
- Slattery ML,
- Jacobs DR,
- Folsom AR,
- Nelson ET
- Crisp AH, Douglas JWB, Ross JM, Stonehill E. Some developmental aspects of disorders of weight. J Psychosom Res. 1970;313–320
- Rutter M, Giller H. Juvenile Delinquency: Trends and Perspectives. New York, NY: Guilford Press; 1984
- Moore ME,
- Stunkard A,
- Srole L
- Crisp AH,
- McGuiness B
- Stunkard AJ,
- Albaum JM
- Stevens J,
- Keil JE,
- Waid LR,
- Gazes PC
- National Heart, Lung, and Blood Institute. Clinical Guidelines for Identification, Evaluation, and Treatment of Overweight and Obesity in Adults: Evidence Report. Bethesda, MD: National Heart, Lung, and Blood Institute; 1998
- Hoebel BG, Liebowitz SF. Brain monoamines in the modulation of self-stimulation, feeding and body weight. In: Weinger HA, Hofer MA, Stunkard AJ, eds. Brain Behavior and Bodily Diseases. New York, NY: Raven Press; 1981:103–142
- Rolls ET. The Brain and Emotion. New York, NY: Oxford Press; 1999
- Copyright © 2001 American Academy of Pediatrics