Anxiety disorders are the most prevalent mental disorders in developed countries. Obesity is hypothesized to be a risk factor for anxiety disorders but evidence supporting an association between these two conditions is not clear. The objectives of this paper were to systematically review the literature for a link between obesity and anxiety disorders in the general population, and to present a pooled estimate of association. We performed a systematic search for epidemiological articles reporting on obesity (explanatory variable) and anxiety disorders (outcome variable) in seven bibliographical databases. Two independent reviewers abstracted the data and assessed study quality. We found 16 studies (2 prospective and 14 cross-sectional) that met the selection criteria. Measures of effect from prospective data were mixed but cross-sectional evidence suggested a positive association between obesity and anxiety. The pooled odds ratio from cross-sectional studies was 1.4 (confidence interval: 1.2–1.6). Subgroup analyses revealed a positive association in men and women. Overall, a moderate level of evidence exists for a positive association between obesity and anxiety disorders. Questions remain regarding the role of obesity severity and subtypes of anxiety disorders. The causal relationship from obesity to anxiety disorders could not be inferred from current data; future etiologic studies are recommended.
Anxiety disorders are the most common psychiatric disorders in the developed world, affecting a quarter of the population during lifetime.1 These disorders are marked by psychological symptoms such as excessive worry, fear, apprehension, and physical symptoms such as fatigue, heart palpitations and tension. They present in the form of generalized anxiety disorder, panic disorder, post-traumatic stress disorders and specific phobias, among others.2 Anxiety disorders are highly persistent, typically chronic and frequently coexist with each other and with other psychiatric conditions. Further, they are known to increase the risk of a number of poor health outcomes such as chronic conditions, poor quality of life and mortality.3, 4, 5, 6
Obesity may be a risk factor for anxiety disorders. Obesity is an increasingly prevalent disease,7 characterized by excess bodyweight, and estimated to be one of the most important contributors to the burden of disease worldwide.8 Obesity may lead to anxiety disorders through various pathways. For instance, weight-related discrimination and stigma can be deeply distressing for obese individuals;9, 10 in addition, the negative effect of obesity on health and quality of life might be particularly stressful.6, 11, 12 Both pathways may subsequently result in anxiety disorders.
The current evidence linking obesity and anxiety disorders is not clear: some studies have identified an association between these two conditions while others have failed to observe a significant relationship. Although it is possible that no association exists, mixed or weak results might be caused by the heterogeneous nature of obesity or of anxiety disorders. The relation between obesity and anxiety disorders might differ among subgroups of the population with varying sociodemographic, behavioral and biological characteristics. For example, because of the greater social discrimination toward obese women,13 obesity might be more strongly related to anxiety disorders in women than in men.
A scoping search retrieved seven reviews relevant to obesity and psychological factors that included anxiety.14, 15, 16, 17, 18, 19, 20 A majority of the reviews identified some links; however, none specifically addressed anxiety disorders. Many of the papers were narrative reviews and differed methodologically from a systematic review. Some of the reviews also targeted clinical populations from which results may not apply to the general population. The purpose of this paper was therefore to systematically review the literature for an association between obesity and anxiety disorders in the community, and to present a pooled estimate of association. A comprehensive systematic review provides unbiased study selection and objective evaluation of study quality. Identifying obesity as a potential risk factor for anxiety disorders is crucial not only to further our knowledge regarding this prevalent mental disorder, but also to alleviate the resulting burden of obesity in the population.
Literature search strategy
We performed a systematic literature search in seven major databases: MEDLINE via PubMed (United States National Library of Medicine, Bethesda, MD, USA), EMBASE (Elsevier, Amsterdam, Netherlands), Scopus (Elsevier, Amsterdam, Netherlands), ProQuest Dissertations and Theses Online (ProQuest, Ann Arbor, MI, USA), ISI Web of Knowledge (Thomson Reuters, New York, NY, USA), Cumulative Index to Nursing and Allied Health Literature (Elsevier, Amsterdam, Netherlands) and PsycINFO (American Psychological Association, Washington, DC, USA) were searched from inception to May 2009. Database-specific search terms included the key words ‘obesity’, ‘anxiety disorder’, names of anxiety disorder subtypes and relevant synonyms. Examples of search strategies are provided in the annex. We conducted the search with no language restrictions, but reviewed only English or French papers. We identified additional studies through hand-searches of bibliography from primary studies, review articles and key journals, and through contacts with experts in the field.
Epidemiological studies that provided a quantitative measure of association between obesity (explanatory variable) and anxiety disorders or any specific type of anxiety disorder (outcome variable) were eligible. Studies based on samples of children or on pregnant women were excluded because body weight can fluctuate considerably during these life stages. Papers using non-representative sampling procedures were also excluded. For cohorts with multiple reanalyses, only one article with the most relevant stratification was chosen. Two investigators (GG and DN) independently screened all articles for those that met broad inclusion criteria.
Data extraction and quality assessment
Quality assessment of studies was performed together with data extraction by independent reviewers (GG and DN). Inter-rater reliability was estimated from kappa scores. Any disagreements were resolved by discussion or by a third party (NS). We contacted authors for additional information when necessary.
From each paper, the reviewers abstracted basic study information, sample characteristics, details on anxiety and obesity measurements, confounders and association estimates. The fully adjusted measure of association was reported if available. Critical appraisal checklists were used to evaluate the quality of the studies. Checklists were adapted from the Newcastle–Ottawa Quality Assessment Scale21 and critical appraisal articles.22, 23, 24 The Newcastle–Ottawa Quality Assessment Scale has established content validity and inter-rater reliability.21 Items reviewed included representativeness of sampling procedure, response rate, validity of measurement methods and control of at least three important confounders. A response rate of 60% or above was considered adequate. This cut-off has been used in previous systematic reviews of observational studies.25, 26 We identified pertinent confounders through the literature and included demographic variables (age, gender, ethnicity, marital status and socioeconomic status (SES)), health variables (chronic conditions and self-reported health) and lifestyle variables (physical activity, smoking status). Reporting on the methodological aspects of the studies rather than relying on a numerical score for quality is considered more appropriate for systematic reviews and meta-analyses.27 Therefore, we rated individual components of the checklist (criteria met; criteria not met; not reported) and provided an overall rating for the quality of the study (poor; moderate; high).
We analyzed prospective and cross-sectional studies separately. Odds ratio estimates (OR) and their 95% confidence intervals (CI) were graphically represented on Forest plots. Some heterogeneity between studies was expected and evaluated descriptively and statistically using the inconsistency index.28 For studies with available or calculable ORs, we conducted a meta-analysis from a random effects model.29 We carried out subgroup analyses by gender, type of anxiety assessment and confounder control. We also performed additional sensitivity analyses by removing larger studies and extreme findings. Publication bias was assessed on a Funnel plot and with Egger's test. Analyses were conducted using the statistical software Comprehensive Meta-Analysis (version 2.2.048, Biostat, Englewood, NJ, USA).
Figure 1 illustrates the flowchart of study selection. The search identified 4002 abstracts for perusal. Reviewers rejected 3925 studies that did not meet broad inclusion criteria. Of the 77 remaining studies, 64 papers were further excluded on review of their full text. The main reasons for exclusion were non-representative sampling and assessment of mental outcomes other than anxiety disorders. We found three additional studies through bibliography hand-searches, yielding 16 relevant publications for this systematic review.30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45 Inter-rater agreement was good for study screening (kappa statistic 0.72, 95% CI: 0.57–0.87) and for quality assessment (kappa 0.75, 95% CI: 0.62–0.88).
Tables 1 and 2 summarize the characteristics and main findings of the prospective and cross-sectional studies, respectively. Tables 3 and 4 present their quality assessment. Papers included in this review were published between 1962 and 2009; 13 of the 16 studies were published within the last 3 years. All studies were from the Western world: nine from Northwest Europe, eight from North America and two from Australia. Sample size ranged from 302 to 177 047, with a total of 346 289 individuals included in this review.
Only two longitudinal studies analyzed the effect of obesity on subsequent anxiety disorders. Findings were contradictory. In one study,33 obesity had a significant positive effect on anxiety disorder in men but not in women. In the other women-only study,38 obesity had a strong positive effect in women. Bjerkeset et al.,33 followed a large cohort (n=33 777) of Norwegian men and women for approximately 10 years (1984–1986 and 1995–1997). Psychiatric cases of anxiety were identified using the Hospital Anxiety and Depression Scale. Baseline measures of common mental symptoms served as a crude proxy to control for baseline anxiety disorders. Adjusted ORs supported a positive effect of obesity on anxiety symptoms. When stratified by gender, the association strengthened for men but became nonsignificant for women.
Kasen et al.,38 used a smaller sample (n=544) of American mothers interviewed on three occasions (1975, 1983 and 2002–2005). Past 6-month generalized anxiety disorder was ascertained from Diagnostic and Statistical Manual of Mental Disorder-IV diagnostic criteria. The study presented some limitations with respect to obesity assessment (participants were asked to recall pre-pregnancy weight, 5 years earlier on average) and generalizability (study limited to mothers only). Levels of depressive symptoms assessed at the second wave (1983) were used as a surrogate for baseline psychopathology. Obese women were found to be more at risk of developing generalized anxiety disorder compared with normal-weight women, even after controlling for important confounders.
In sum, prospective evidence for an effect of obesity on anxiety disorder is scarce and mixed. Further, only one study provided prospective evidence in men. Study quality was not consistent between the two studies. Both were undermined by differential loss to follow-up and both could not fully control for baseline anxiety disorders. As a result, it became difficult to assert whether it was past anxiety disorder or past weight status predicting future anxiety.
In all, 14 of the 16 studies included in this review were cross-sectional. Cross-sectional findings generally pointed to a weak but positive association between obesity and anxiety disorders. Seven studies showed a significant positive association30, 31, 32, 39, 41, 42, 43 while five showed a positive trend that was nonsignificant for at least one gender stratum.34, 35, 37, 44, 45 In the largest study on obesity and anxiety disorder to date, Zhao et al.45 analyzed data from 177 047 American adults. Participants self-reported any lifetime medically diagnosed anxiety disorder. A body mass index (BMI) was calculated from self-reported weight and height. A significant positive association was present in obese (BMI30) women but only in severely obese (BMI40) men, after adjustment.
No study reported a significant negative association but two studies described a negative trend for one of their stratum (gender stratum40 and anxiety subtype36). Using data from the Alberta Mental Health Survey (n=3 882), McLaren et al.40 found a positive trend between obesity and anxiety disorder in women but a negative trend in men, after controlling for important confounders including mood and eating disorders. Anxiety disorder was assessed with the Mini Neuropsychiatric Interview and obesity was determined from self-reported weight and height (BMI30). Hallstrom et al.36 analyzed the association of obesity with anxiety attacks or phobia in 800 women. Anxiety disorders were assessed during a psychiatric interview. Although weight and height were measured, obesity was determined from a non-standard BMI. Findings revealed a positive trend for phobia but a negative trend for anxiety attacks. Analyses were not adjusted for confounders.
A meta-analysis was performed on 13 of the cross-sectional studies30, 31, 32, 34, 35, 36, 37, 39, 40, 42, 43, 44, 45 for which an OR and its precision estimates were available or calculable (Figure 2). For studies that reported both lifetime and past-year anxiety prevalence, we favored past-year prevalence. The pooled OR of an association between obesity and anxiety was 1.40 (CI: 1.23–1.57). The inconsistency index was 84.3% (P-value < 0.001), suggesting high level of heterogeneity. Removing the two largest studies30, 45 independently or concurrently did not significantly alter our findings (pooled OR with both studies removed: 1.40, CI: 1.24–1.58). Removing the study with the highest OR32 had a minimal effect on the pooled OR (1.38, CI: 1.23–1.55).
Subgroup analyses were conducted to account for variability based on predefined hypotheses (Table 5). Pooled ORs by gender were positive and significant for men and women. Pooled ORs of anxiety assessed over lifetime and past-year were nearly identical. The pooled OR was lower for studies with adequate confounding control than for studies without. All subgroup analyses failed to improve study heterogeneity.
The funnel plot (Figure 3) was overall symmetric except for one small study to the extreme right of the mean. Egger's test produced a nonsignificant P-value (0.65) suggesting no publication bias.
We identified 16 observational studies (2 longitudinal and 14 cross-sectional studies) that examined the association between obesity and anxiety in adults from community samples. Overall evidence suggests that obesity is positively associated with anxiety disorders in adults. More than half of the studies and our pooled analysis showed significant higher odds of anxiety disorders in obese compared with non-obese individuals. In spite of this, the strength of evidence is moderate. A majority of the studies were undermined by methodological limitations and the observed associations were weak and often nonsignificant. Moreover, evidence was mainly cross-sectional and other plausible explanations could not be ruled out with current findings.
In our systematic review, we found similar association between obesity and anxiety by gender. Of the studies that stratified or restricted by gender, more than half supported a significant positive association for women and the remainder found a positive trend, except for one study that reported conflicting results by anxiety subtypes.36 The same positive association or trend was observed for men, except for one paper reporting a negative trend.40 Furthermore, the heterogeneity we found between studies was not explained by gender. However, other variables may moderate the association between obesity and anxiety, including the degree of obesity and the types of anxiety under investigation. Reviewed evidence suggests a stronger association between severe obesity (defined as a BMI35) and anxiety disorders compared with moderate obesity (BMI 30–35).32, 43, 45 Petry et al.46 stratified by obesity severity the same data used by Barry et al.30 They found increasing adjusted ORs of past-year and lifetime anxiety disorders across obesity severity.
Obesity might also associate differently with different subtypes of anxiety disorder (see Table 6). Seven studies in this review performed subgroup analysis by anxiety disorder subtypes.30, 36, 37, 38, 39, 43, 44 Evidence was largely mixed with the exception of specific phobia, which was consistently associated with obesity among studies that analyzed this subtype. In the only study that analyzed post-traumatic stress disorder (PTSD),43 the authors found post-traumatic stress disorder was most strongly associated with obesity compared with other anxiety subtypes. Obesity was also linked to social phobia, particularly in women.
Obesity may be linked to anxiety disorders through a number of ways. Studies have shown that obese people experience weight discrimination in both their public and private lives.13, 47 Obese individuals have poorer social support and social networks47 as well as lower SES48, 49, 50 compared with normal-weight individuals. These stifled life opportunities and limited psychosocial resources can put obese persons at greater risk of developing psychological distress9, 10 and subsequent anxiety disorders. Obese individuals might also harbor anti-fat biases toward themselves.49 They may feel pressure to gain control over their weight, which can be distressing, particularly when repeatedly failed attempts to lose weight are the norm. Dieting and eating preoccupation were found to be closely related to anxiety.51
In addition, obesity is a strong predictor of a number of chronic medical conditions.52, 53, 54 The adverse symptoms and functional impairments associated with obesity55, 56 and obesity-related chronic conditions could increase the risk of anxiety.6, 11, 12, 57 Obesogenic lifestyle factors such as caloric dense diets and lack of physical activity could contribute to the development or maintenance of anxiety disorders.58
Conversely, it is possible for anxiety disorders to lead to weight gain. The hypothalamic–pituitary–adrenal axis dysregulation is thought to contribute to appetite dysregulation and subsequent weight gain in stressed individuals.59, 60 Symptoms of anxiety disorders may not only increase appetite,61 but stimulate a craving for high-sugar and high-fat foods.59, 62, 63, 64 In addition, anxiety-associated chronic conditions (for example, asthma) can have a profound effect on functional health, which may lead to physical inactivity resulting in excess weight.
Obesity and anxiety disorders may also relate through a third variable including biological, environmental or individual factors. For example, obesity and anxiety disorders are both partly heritable diseases65, 66 and may share a common genetic basis. Alternatively, environmental factors such as environmental endocrine-disrupting chemicals could partly explain the obesity–anxiety link.67 These compounds can affect hormonal homeostasis involved in weight and emotional regulation.59, 60, 67 Individual factors might include childhood negative events (for example, child abuse), which have been found to predict excess weight68, 69 and anxiety disorders.70 Premorbid personality traits, such as neurocriticism, hypersensitivity to criticism and avoidant coping styles, can also predispose individuals to anxiety and obesity.12, 71, 72 It is noteworthy that an avoidant coping style has been associated with impaired eating behavior.73, 74 In addition, psychiatric conditions are often comorbid and some psychiatric illnesses have been found to lead to weight gain, including eating disorders,75, 76 mood disorders26, 46 and personality disorders.46
When considering the link from obesity to anxiety, the definition of anxiety should be kept in mind. Anxiety symptoms typically occur when an individual is exposed to specific stressors, while anxiety qualifies as a clinical disorder when symptoms are disruptive enough to interfere with functioning. As a result, anxiety disorder is contextual. Obese individuals could structure their life to avoid anxiety-inducing situations and their life decisions may therefore differ substantially from the life selected by normal-weight individuals. For instance, obese persons of lower SES report less obesity-related stigma in their lives than those of higher SES.77 The causal association between obesity and anxiety might therefore not be straightforward and may be compounded by feedback loops including individuals' aggregated psychosocial decisions.
Limitations of studies
The overall quality of studies was moderate but inconsistent: poor confounder control and information biases were the main limitations. The studies in our systematic review also presented some important differences, especially in their sample characteristics and in their measurements of anxiety and obesity, which may confound potential interpretations.
Nearly half of cross-sectional studies controlled for less than three important confounders.31, 35, 36, 37, 41, 42 Subgroup analysis showed that the pooled estimate for poorly adjusted studies was higher and presented greater variability than for adequately controlled studies. Poor confounder control could have an important effect on results. For instance, Hach et al.35 found obesity to be significantly associated with anxiety disorders in men but reported no significant association after adjustment for somatic conditions. Confounder selection was also inconsistent across studies. The most common covariates were age, gender and a proxy for SES, while health characteristics were rarely included. Only three studies controlled for physical comorbidities35, 39, 45 and three for psychiatric comorbidities.40, 43, 44 Adverse health conditions are predictors of both obesity and anxiety disorders12 and may have accounted for some of the observed association between obesity and anxiety.
Researchers used an array of instruments to measure anxiety, which encompassed varying levels of anxiety disorders. For instance, timeframe of reference for anxiety outcomes spanned from current to lifetime. Lifetime anxiety outcomes unanimously showed positive point estimates, whereas timeframes of a year or less found weaker and sometimes negative associations. However, subgroup analysis did not reveal any difference in pooled estimates between lifetime and past-year anxiety outcomes. Interestingly, publications that analyzed both past-year and lifetime anxiety outcomes30, 31, 39, 44 found slightly weaker associations for lifetime anxiety disorders. Longer compared with shorter timeframes of anxiety assessment may be more prone to recall bias. The number and types of anxiety disorders also varied across studies. The decision as to how many and which subtypes of anxiety to include in the analysis might have affected the strength and direction of association.
Although obesity was usually determined from BMI (30 kg m–2), as recommended by current guidelines,78 it was often based on self-reported data, which has been shown to underestimate the prevalence of obesity.79 Nevertheless, outcome results were not markedly different between self-reported and measured obesity. The obesity referent group in some studies might have also affected the estimate of association. Six studies included overweight (BMI 25–29.9) and/or underweight persons (BMI <18.5)31, 35, 39, 42, 43, 44 in the referent group. These weight categories have been associated with anxiety40, 45 and including them as reference could have biased results toward a negative effect.
Finally, most publications used cross-sectional data and causality could therefore not be inferred. In addition, all studies were from the Western world. Generalizability of the findings may not hold in different, more weight-accepting cultures.
Limitations of the systematic review and meta-analysis
Qualitative and quantitative results from this review should be interpreted with some caution. Although statistical analysis did not observe a publication bias, such a bias may still exist. Publishers and authors often favor publishing positive findings over negative ones.80 Language bias may also be of relevance. Although we did not restrict our search by language, we could only review the English and French literature. Finally, the studies included in our meta-analysis were heterogeneous. Although we used a random effects model to pool results, statistical methods may not be enough to overcome heterogeneity. Some authors have argued that heterogeneous studies should simply not be meta-analyzed, while others maintain that meta-analysis is still worthwhile as long as interpretation remains cautious.80
Our systematic review had several strengths. We undertook a comprehensive search from several bibliographical databases. Two researchers independently reviewed and assessed the quality of the studies. A thorough systematic review was performed instead of a solely quantitative analysis and study heterogeneity was carefully considered and examined. Throughout, we considered the evidence supporting a relationship between anxiety disorders and obesity, a topic that to our knowledge has not been specifically reviewed before.
Many research opportunities exist to clarify the causal relationship between obesity and anxiety. To begin, it is important to acknowledge the heterogeneous nature of obesity and anxiety. Obese people differ on the history, etiology and severity of their weight problem.19 The association between anxiety and obesity might therefore vary across determinants of obesity. For example, some studies found higher odds of anxiety with worse obesity severity. Subtypes of anxiety may affect the relationship with obesity but evidence for this phenomenon has so far been sparse and mixed. Population-based studies with large sample size or case–control designs could be helpful in adequately examining subgroups of obesity and anxiety disorders.
In addition, some methodological issues could be considered in future research. For example, assessing obesity from measured rather than self-reported data would reduce obesity misclassification and increase the validity of the data. Shorter time reference for assessing anxiety disorders (for example, past 12-month) should be favored over lifetime assessment because the latter is prone to poor recall. Confounding factors such as health characteristics and obesity-related comorbidities should be included in models in addition to the confounders already identified in previous studies. Finally, prospective data were rare in this systematic review and their results were mixed. High-quality longitudinal studies are recommended to elucidate the temporal relationship between obesity and anxiety.
After careful systematic review and meta-analysis of the literature, we found moderate evidence supporting a weak but positive association between anxiety and obesity. The causal effect of obesity on anxiety disorders could not be inferred from current data. Future high-quality longitudinal studies are recommended.
Conflict of interest
The authors declare no conflict of interest.
Funding for this project was provided by the Canadian Institutes for Health Research (CIHR Grant MOP-79464).
About this article
Current Obesity Reports (2018)