Stress and Adiposity: A Meta-Analysis of Longitudinal Studies




Psychosocial stress has been strongly implicated in the biology of adiposity but epidemiological studies have produced inconsistent results. The aim of this analysis was to bring together results from published, longitudinal, prospective studies examining associations between psychosocial stress and objectively measured adiposity in a meta-analysis. Searches were conducted on Medline, PsycINFO, Web of Science, and PubMed (to January 2009) and reference lists from relevant articles were examined. Prospective studies relating psychosocial stress (general life stress (including caregiver stress), work stress) to BMI, body fat, body weight, waist circumference, or waist-to-hip ratio were included. Analyses from 14 cohorts were collated and evaluated. There was no significant heterogeneity, no evidence of publication bias, and no association between study quality and outcomes. The majority of analyses found no significant relationship between stress and adiposity (69%), but among those with significant effects, more found positive than negative associations (25 vs. 6%). Combining results in a meta-analysis showed that stress was associated with increasing adiposity (r = 0.014; confidence interval (CI) = 0.002–0.025, P < 0.05). Effects were stronger for men than women, in analyses with longer rather than shorter follow-ups, and in better quality studies. We conclude that psychosocial stress is a risk factor for weight gain but effects are very small. Variability across studies indicates there are moderating variables to be elucidated.


Obesity prevalence shows accelerating trends worldwide (1,2) and is known to be associated with a range of risks to health and survival (3). Understanding the determinants of obesity is important for global public health and the economy (4,5). The etiology of obesity is multifaceted, involving interactions between individuals and their social, cultural, and physical environments, but psychosocial stress is one of the potential determinants that has attracted considerable attention (6,7).

The stress response is the generalized response to any factor that has the potential to overwhelm the body's compensatory ability to maintain homeostasis (8). Part of this response may involve metabolic changes that could directly increase abdominal adiposity (9). Stress may also affect food choice, both through lack of time for food preparation (10) and by increasing preferences for higher-fat, energy-dense foods (11,12); thereby promoting positive energy balance. On the deficit side, stress has been shown to reduce participation in leisure time physical activity (13); again potentially favoring positive energy balance.

However, the epidemiological literature linking stress to weight gain has produced inconsistent results. In a recent narrative review bringing together animal and human studies, there was considerable variability in the results (6). In the human literature, psychosocial stress was related to weight gain in schoolchildren in one study (14), and perceived stress mediated the association of neighborhood disorder with obesity in a community sample in Texas (15); a mechanism that could contribute to the association between low socioeconomic status (SES) and obesity (16). Acute stressful periods have also been associated with weight gain in several studies (17,18,19). However, many other studies have not identified any effect: in low-income young mothers, obesity was predicted by lack of parental support but not perceived stress (20), and in a recent large study in the United Kingdom there was no association between perceived stress and increases in weight or waist circumference over 5 years of adolescence (21). Work stress has been found to be associated with adiposity (22), but in the Whitehall sample (23), the association was only observed in men. Other studies have found significant effects in univariate analyses, which have not survived adjustment for confounding factors (24).

Despite the results being varied and inconclusive, it is possible that the totality of the findings could provide a more robust conclusion. We therefore conducted a systematic review and meta-analysis to determine whether the combined evidence from prospective studies supports an association between psychosocial stress and obesity in adults. We excluded studies on children from the review because indexes of adiposity are different in children and adults, and sources of stress are also different. There are many types of stressor (e.g., physical, physiological, and emotional), but this review focused on external stressors such as life events, caregiver stress, or work stress which have been the main focus of longitudinal studies examining the relationship with adiposity. Work stress was examined within the overall analysis and separately because of its importance for other health outcomes. The analysis was restricted to studies with an objectively measured weight outcome in order to avoid any bias associated with self-report, and we analyzed separately by studies that controlled for the baseline level of the outcome variable.

Methods and Procedures

Data sources and searches

A protocol was developed based on widely recommended methods for systematic reviews of observational studies (25,26). The following general bibliographic databases were searched: Medline (1966–January 2009); PsycINFO (1872–January 2009); Web of Science (1900–January 2009); PubMed (1950–January 2009) and reference lists from relevant articles were scrutinized. The search terms were: (body weight OR BMI OR obesity OR waist OR abdominal fat OR body fat) AND (psychological stress OR psychosocial stress OR work stress OR life event OR life stress OR chronic stress OR social support) AND (prospective OR longitudinal).

The systematic review and meta-analysis was limited to prospective studies. Criteria for inclusion/exclusion were as follows: (i) English language full-length publication in a peer-reviewed journal; (ii) prospective cohort design with participants ≥16 years old; (iii) investigating a longitudinal association of life stress-related factors (work stress, general life stress, caregiver stress) and body weight, BMI, central fat or obesity, skin fat, or waist circumference or the ratio of waist to hip; (iv) if more than one kind of stress-related psychosocial factor and adiposity indicator was assessed in one paper, the samples were included separately; (v) If the associations of stress-related psychosocial factors with obesity were separately assessed in men and women in one paper, the samples were included separately; (vi) if analyses of body weight were not adjusted for height, the studies were excluded; (vii) studies enrolling the populations suffering from severe illness (e.g., cancer, acquired immune deficiency syndrome), eating disorders or receiving some weight-affecting interventions (e.g., gastric banding, dieting trial, gastric bypass surgery) were excluded; (viii) studies using low SES as a stress indicator were excluded, because almost all other studies included SES as a covariate; (ix) studies with <1 year follow-up were excluded because the present review addressed longer-term effect of stress on adiposity; (x) if there were several papers in one cohort, the paper(s) with shorter follow-up, smaller sample sizes or poorer study quality indexes were excluded; (xi) analyses based on self-reported outcome variables were excluded.

Data extraction and quality assessment

Studies were coded according to predetermined criteria, which were later updated to incorporate additional aspects of the located studies. The final list of variables included the following: first author, publication year; cohort size with subject characteristics (country); subject age (years); follow-up duration (years); type of stress-related psychosocial factors (measurement method); controlled covariates; outcome (measurement method); quality score; and brief results and effect size (correlation coefficient, r).

Study quality was evaluated using established protocols (25) because quality can contribute to any potential bias associated with effect estimation. Good quality studies were deemed to have the following features: (i) consecutive or random recruitment of participants or representative populations, (ii) ascertainment of stress-related variables by validated instruments or clinical examination, (iii) ascertainment of outcome variables by clinical examination, and (iv) control for possible covariates including age, sex, smoking, and SES. Studies were categorized as higher or lower quality according to whether or not they fulfilled three or more of these criteria. Study inclusion and data extractions were conducted by one author (Y.C.) and verified by another (A.S.). Assessment of quality and validity were made independently by at least two reviewers and disputes were settled by consensus.

Data synthesis and analysis

The meta-analysis was conducted according to procedures described elsewhere (27,28). Where possible, effect sizes were calculated from the difference in adiposity indicators between the control and exposed groups and were transformed into r. When raw data were omitted from the publication, an F ratio was used for conversion to r. If no relevant convertible statistics were presented other than a P value, we calculated the t statistic from the P value and an r-sub (equivalent) (27,28,29). When a paper reported P < 0.05, P < 0.10, or ns, we computed r-sub (equivalent) with P values of 0.025, 0.05, 0.50 (one-tailed), respectively, which likely yielded a conservative estimate of effect size. In the event that insufficient data were reported to calculate an r value, the authors were contacted for further information. An inferential statistic or effect size was used to calculate a z-score for each study, which was weighted by sample size to calculate an overall z-score and probability.

In the present meta-analysis, each psychosocial stress category measure was viewed as a separate construct, and therefore several effect sizes could be derived from one study. Where more than one measure was reported for a construct (e.g., several measures of work stress), the weighted mean of the set of effect sizes (e.g., mean r for all work stress measures) was used to calculate the effect size. In addition to computing separate effect sizes for each construct, an aggregate effect size was obtained for each study by calculating the weighted mean of all effect sizes in that category. This procedure was carried out because some meta-analyses conservatively include only one effect size per study population. Random effects modeling was used (30) to account for the amount of variance caused by differences between studies as well as differences among participants within studies.

Provided there was sufficient information (≥4 studies), we performed sensitivity analyses according to the characteristics of study design (sample size and follow-up period), study population (age, gender), study quality, outcome variables (BMI, waist to hip, etc.), and whether the baseline value of the outcome variable had been included. We employed the Q-test for homogeneity between studies to test whether between-study variability in effect sizes exceeded that expected from the corresponding within-study variability. To detect publication biases, we estimated the degree of asymmetry using an unweighted regression asymmetry test (31). All analyses were performed using a Meta-Analysis Program developed in Japan (32).


The flow diagram for the review is presented in Figure 1. Table 1 and the Supplementary Table S1 online included 14 papers and excluded 42 papers, respectively. Table 2 summarizes the detailed characteristics of 32 separate analyses extracted from these enrolled papers.

Figure 1.

Flow diagram of systematic review (QUOROM statement flow diagram).

Table 1.  Prospective studies investigating the association of stress-related psychosocial factors and adiposity
inline image
Table 2.  Characteristics of the studies included in the meta-analyses
inline image

Study characteristics and quality

Results from 14 papers published between 1988 and 2009, involving participants from Europe and the United States, were included. General life stress (caregiver stress, major life events, etc.) was employed as a predictor in 56% (18/32) of analyses and work stress was evaluated in 44% (14/32) of analyses. The most widely used adiposity indicator was BMI (59%; 19/32), followed by waist to hip (16%; 5/32), waist circumference (16%; 5/32) and weight (9%; 3/32). Study quality scores (0–4) averaged 2.8, and the proportion of high-quality analyses (quality score ≥3) was 56% (18/32) (Table 3).

Table 3.  Results of meta-analyses, subgrouping, and sensitivity analyses
inline image

Study results and meta-analysis

Across all analyses, more demonstrated a significant positive effect (25%) than a negative effect (6%) of stress on adiposity, but 69% had a null effect (Table 2). As shown in Table 3, the overall combined r was 0.014 (confidence interval (CI) 0.002–0.025, P = 0.023), which was accompanied by no significant heterogeneity between analyses and no evidence for publication bias exception for analyses using weight as the outcome. The aggregate effect over the 14 papers was not significant, r = 0.011 (−0.007 to 0.029), P = 0.22.

Subgroup meta-analyses by follow-up period showed that analyses with longer follow-ups had a higher combined r (r for studies with ≥5 years follow-up period =0.016; CI 0.004–0.028, P = 0.009) compared with the overall effect size. Effects were also stronger for studies that had a higher methodological quality (r = 0.015; CI 0.002–0.027, P = 0.02) than those with a lower methodological quality (r = 0.007; CI −0.032 to 0.046, P = 0.72). Interestingly, a combined r of male population analyses (0.024; CI 0.006–0.042, P = 0.01) was higher than the r of female populations (0.017; CI −0.008–0.042, P = 0.17) or the overall effect size. Combined r in analyses that controlled for age, sex, smoking, and SES (0.013; CI −0.000–0.026, P = 0.06) was similar to the overall effect size. Dividing the studies into those focusing on general life stress and work stress did not show a significant association for either individually. Sensitivity analyses across adiposity outcomes indicated a tendency for weight to have a higher combined effect than other indicators, but there was no evidence that effects were stronger for indicators of central adiposity.


This systematic, meta-analytic review is the first to confirm that psychosocial stress is positively related to the development of adiposity in prospective studies, although the effects were modest and smaller than assumed in the lay literature. Confidence in the results is increased by the fact that the effect was stronger in studies that were methodologically stronger, controlled for the baseline level of the outcome variable or had longer follow-up.

There was no evidence that central adiposity was more sensitive to stress than general adiposity, although with a relatively small number of studies, a difference might be difficult to detect. In line with results from cross-sectional studies (6), a stronger effect was seen in men than women.

One of the strengths of this review is that the included studies assessed adiposity by physical examination rather than relying on self-reports which are prone to error (13). The majority of studies included SES as a covariate, an important methodological consideration given that lower SES is associated with both BMI and stress, particularly in women (33). However, SES could also moderate the strength of the relationship between stress and obesity, and therefore the effect sizes observed in this meta-analysis may underestimate the true relationship between stress and obesity. Most studies also controlled for the baseline level of the outcome variable and effects were stronger in this subset.

Major life-events and periods of acute stress appeared to play a greater part in the onset of obesity for men than women as has been reported before (6), although the mechanism is unclear. Men consistently show stronger physiological responses to stress than women, including greater cardiovascular and neuroendocrine activation (34), and higher cortisol levels after exposure to acute real-life psychological stress (e.g., examinations) or controlled laboratory experiments (35,36). Discrepancies between reported stress and untapped, stress-related, psychobiological events may explain the weak association between some measures of life stress and adiposity in this analysis. Illustrating this point, we have previously found that impaired systolic BP recovery but not subjective stress, predicted increases in waist to hip in men (37).

Dividing the studies into those focusing on general stress and work stress did not show significant results in either group of studies alone, but power for either subset alone was reduced. It should be noted that three out of the four studies finding significant inverse associations with BMI were in the work stress field, which contributed to the null effect. Measures of work stress may also have been limited or insensitive. For example, single-item assessments were employed in three studies to index work dissatisfaction (38), insecurity (39) and work influence/control (24) respectively, and only one study employed a general validated measure of work stress (40). A further limitation is that the outcome may be influenced by missing data from people suffering stress-induced work leave, an influential factor because unemployment is related to BMI increases (41).

The association of stress with eating behavior is likely to be complex, with multiple factors such as perception of the stressor (11,42), current weight status, and psychological eating traits (18,43) modifying the effect. For example, individuals have reported eating less as well as more, under conditions of stress, which may explain the weak findings in this area (42). Unfortunately, the studies included in the present review rarely included data on food intake, and where activity was included, it was often only as a covariate. There were also few studies that measured psychological eating traits to examine effect-modification.

A key limitation is that multiple stressors often coexist, but we were unable to discriminate between sources, severity and types of stress because of the paucity of prospective studies in this field. The role of co-occurring psychological disorders in the relationship between stress and adiposity could also not be determined in this review, and there may be further dissociations or inter-relationships between the various measures of psychosocial stress. The literature on psychological distress and depression in weight gain was not included. This topic has been recently reviewed elsewhere (44) and interpretation is complicated by the influence of psychotropic medication on weight change. We focused on exposure to external stressors rather than emotional states that may have multiple determinants. Additionally, most of the studies included in this review involved single assessments of stress, which cannot fully reflect people's experience over long follow-up periods.

This meta-analysis indicates that psychosocial stress promotes weight gain, albeit with a modest effect size. The results highlight the importance of further prospective research. They also emphasize the need to understand the behavioral and psychobiological mechanisms in order to identify the factors that mediate and moderate obesogenic effects of stress.


Supplementary material is linked to the online version of the paper at http:www.nature.comoby


Y.C. is supported by the Kanae Foundation for the Promotion of Medical Science and the National Prevention Research Initiative, UK. A.S. is supported by the British Heart Foundation, J.W. and K.L.W. are supported by Cancer Research United Kingdom and E.L.G. by the University of Roehampton, UK.


The authors declared no conflict of interest.