1 INTRODUCTION

Obesity is a chronic, relapsing and progressive disease with excessive subcutaneous or visceral fat accumulation, leading to adverse health effects.1 The prevalence of obesity has been increasing globally, in part due to rapid modernization and industrialization.2, 3 In 2016, the World Health Organization (WHO) estimated that more than 1.9 billion adults aged 18 years and older are overweight, of whom 650 million have obesity. Obesity has a negative effect on overall health as it increases the risk for type 2 diabetes mellitus, coronary artery disease, musculoskeletal disorders and cancers.4

Obesity can be linked to psychosocial dysfunction and social disabilities related to stigmatization.5 Over the past few decades, there has been an increasing amount of evidence that shows a bidirectional association between mental illnesses and obesity.6-8 Depression and anxiety are the common mood disorders that frequently co-exist with obesity, although the association is stronger with depression (odds ratio [OR] 1.21–5.80) than anxiety (OR 1.27–1.40), as reported in 15 studies on obesity and depression and 4 studies on obesity and anxiety.6 A meta-analysis of 14 cross-sectional studies reported that obesity had an OR of 1.4 for anxiety.9 There are several hypotheses regarding the close link between obesity and depression. Leptin dysregulation in people with obesity has been implicated as a risk factor to develop depressive symptoms via its involvement in the amygdala and hippocampal neurogenesis and neuroplasticity.10 Systemic and neuro-inflammation, via an increased release of pro-inflammatory cytokines, have been implicated as potential mediators that link obesity and adiposity with depression.11 Other mechanisms include neuroendocrine dysregulation of the hypothalamic–pituitary–adrenal axis and alterations of peripheral cortisol metabolism, which also contribute to insulin resistance in people with obesity, leading to cardiometabolic diseases.12

The pathogenesis of obesity involves not just a sustained positive calorie balance, but also a change of a body weight set point, which explains the tendency of people with obesity to regain lost weight over time.13 Bariatric surgery is an effective therapeutic intervention for severe obesity to reduce substantial body weight, and improve or even induce remission of common co-existing medical conditions, including type 2 diabetes mellitus, hyperlipidaemia and hypertension.14-17 The most common type of bariatric surgery performed worldwide is Roux-en-Y gastric bypass, although in the last 2 years, the number of sleeve gastrectomy recorded is greater than Roux-en-Y gastric bypass.18 Structurally, bariatric surgery reduces caloric intake by modifying the anatomy of the gastrointestinal tract via either restrictive or malabsorptive effects. Functionally, it also reduces appetite and improves insulin sensitivity, mediated by incretin hormones such as glucagon-like peptide-1 (GLP-1) and pancreatic peptide YY (PYY).19

Compared to non-surgical methods, people with obesity and type 2 diabetes who underwent bariatric surgery achieved greater weight loss, remission rates in diabetes and dyslipidemia 5 years post-surgery, as reported in a population-based propensity score-matched cohort study.17 A randomized control trial at the University of Pittsburgh reported that bariatric surgeries were more effective than lifestyle interventions for people with obesity and type 2 diabetes mellitus, and were associated with a 19%–30% remission rate in diabetes mellitus.20 In a retrospective case–control study in Canada, bariatric surgery conferred an 89% relative risk reduction in mortality and a 53% relative risk reduction of psychiatric disorders after 5 years.21 Currently, bariatric surgery is recommended for patients with Class III obesity (defined as body mass index [BMI] ≥40 kg/m2) or Class II obesity (BMI ≥35–40 kg/m2) with co-morbidities.22, 23

We aimed to systematically review the literature and examine if bariatric surgery was associated with a reduction in the prevalence and severity of depressive and anxiety symptoms among people with obesity. We also examined if these changes were modified by the age of the subjects and their post-operative BMI.

2 MATERIALS AND METHODS 2.1 Data sources and extraction

We performed a systematic search of all English-language medical literature published from inception till March 2020 using PubMed, CINAHL and Ovid using the following MESH headings: ‘psycho$’, ‘depression’, ‘anxiety’, ‘mood disorders’, ‘mental health’, ‘bariatric surgery’ and ‘weight loss surgery’. We also looked into references of the selected papers. When the papers were not available or information of the study cohort was inadequate, we attempted to contact the respective authors via email to obtain the full paper and more detailed data. To be included in the analysis, studies had to fulfil the following criteria: (i) recruited subjects fulfilled criteria for bariatric surgery, that is, BMI ≥40 kg/m2 or BMI ≥35 kg/m2 with co-morbidities, (ii) reported the proportion of depressive or anxiety symptoms both pre- and post-operatively, and/or (iii) reported the scores of validated questionnaires for mood disorder pre- and post-operatively. Studies were excluded if subjects recruited had other concurrent medical illnesses or diseases which might affect the results, or only reported results on health-related quality of life (HRQOL). Review articles, posters, abstracts and thesis were also excluded. The titles and abstracts obtained through the electronic search were screened followed by the analysis of the full-text articles by two independent reviewers (HHL and HSL). All duplicates were removed. Data such as age, gender ratio, sample size, types of surgery performed, duration of follow-up, as well as pre- and post-operative data (BMI, the proportion of depressive and anxiety symptoms, and the scores of various assessment tools for mood disorders of interest) were documented using a standardized data extraction form. Wherever data were not provided numerically, they would be read off graphs. Data from eligible studies were extracted by HHL, and all extracted data were reviewed by AY.

2.2 Quality assessment

HHL and HSL independently assessed the quality of the methodology and reporting of the studies using the Newcastle–Ottawa Scoring (NOS) Scale. Any discrepancies were resolved by the third reviewer (LLL). The NOS scale was developed to assess the quality of non-randomized case–control studies for interpretation of meta-analysis results. It uses a ‘star system’ which judges the studies in three broad categories, namely the selection of study group, group comparability and ascertainment for the outcome of interest (exposure). In the original NOS Scale for case–control studies, each study can be awarded a maximum of one star for each numbered item (four in the Selection category and three in the Exposure category) and a maximum of two stars in the Comparability category. For the adapted NOS Scale for cross-sectional studies, each study can be awarded a maximum of five stars in the Selection category, two stars for the Comparability category and three stars for the Outcome category. Both scales have a maximum score of 10.

2.3 Statistical analysis 2.3.1 Qualitative

All abstracted information was tabulated. A qualitative meta-analysis was conducted to summarize, compare and contrast the abstracted data.

2.3.2 Quantitative

All data analyses were performed using Stats Direct (version 2.7.9). The presence of heterogeneity between the trials was tested using the I-squared (I2) statistic. I2 of more than 40% indicated a significant heterogeneity. If the I2 was significant, we pooled the data by using random effects model (DerSimonian–Laird); if not, we pooled the data by using fixed effects model (Hedges–Olkin). We also assessed publication bias with the Begg–Mazumdar and Egger test. For dichotomous data, namely the respective proportion of patients with anxiety and depressive symptoms, the pooled OR with 95% confidence intervals (CI) were estimated based on the random effects model. If the mean and standard deviation (SD) of continuous outcomes were presented in the original articles, we analysed them by using the effect size meta-analysis and presented the results as weighted mean difference with 95% CI. The risk of developing adverse effects after bariatric surgery was reported as a proportional rate with 95% CI.

2.4 Assessment tools for screening of mood disorders 2.4.1 Hospital Anxiety and Depression Score

This validated questionnaire comprises seven questions for anxiety symptoms and seven questions for depressive symptoms, rated on a 4-point severity scale. The maximum score for each subscale is 21. A total Hospital Anxiety and Depression Score (HADS) score of ≥11 indicates a definitive case.24, 25

2.4.2 Beck's Depression Inventory

This validated and widely used 21-item self-administered questionnaire evaluates the severity of depressive symptoms on a 4-point scale. The score ranges from 0 to 63. In non-clinical populations, a score of >20 indicates the presence of depressive symptoms, requiring further psychiatric evaluation.

2.4.3 Generalized Anxiety Disorder Assessment-7

The Generalized Anxiety Disorder Assessment (GAD-7) is a seven-item validated instrument that is used to assess the severity of generalized anxiety disorder. Each item asks the individual to rate the severity of his/her symptoms in the past 2 weeks. The total score ranges from 0 to 21. A score of 5, 10 and 15 represents the cut-off point for mild, moderate and severe anxiety symptoms, respectively. Further evaluation is recommended if the total score is 10 or more.

2.4.4 Patient Health Questionnaire-9

The Patient Health Questionnaire (PHQ) is a diagnostic tool for eight mental disorders. The PHQ-9 is a self-administered nine-item depression module from the full version of PHQ. Each item asks the individual to mark the presence of depressive symptoms in the past 2 weeks. The total score ranges from 0 to 27. A score of 5, 10, 15 and 20 represents mild, moderate, moderately severe and severe depressive symptoms, respectively.

2.5 Results

Our initial search identified a total of 3471 articles with 1701 articles from PubMed, 461 articles from CINAHL and 1309 articles from Ovid. After screening of titles, abstracts and full texts, followed by elimination of duplicate publications, 103 articles were identified. Among these, 61 articles were further excluded: 42 did not include comparative data pre- and post-operatively; 11 due to incomplete data; 1 due to overlap cohort.

2.6 Data synthesis

A total of 11,255 subjects from 49 articles were included in the present meta-analysis (Figure 1). All were prospective studies, except for one cross-sectional26 and three retrospective study design.27-29 Roux-en-Y gastric bypass was performed in 31 studies, biliopancreatic diversion in 4, gastric banding in 24, sleeve gastrectomy in 15 and vertical gastroplasty in 7. A total of 24 studies examined the proportion of anxiety symptoms and 28 studies compared the proportion of depressive symptoms pre- and post-surgery. Eight studies compared HADS (anxiety component), three examined GAD-7 scores and two reported SCL90R scores pre- and post-surgery. Twelve studies compared Beck’s Depression Inventory (BDI) scores, nine examined HADS (depression component), five reported PHQ-9 scores and two compared SCL90R scores pre- and post-surgery. The baseline characteristics of all studies are described in Table 1. The sample size ranged between 21 and 3045. For prospective studies, the duration of follow-up ranged between 3 and 120 months, with a mean of 34.0 months.

TABLE 1. Baseline demographic Reference Site Study type Sample size, n Age, mean ± SD, median (range) Female: male, n Type of surgery Duration of study, months NOS score Gertler 1986 Australia Prospective 20 37.2 (24–52) 0:20 RYGB, GP, GG, VG NA 7 Gertler 1986 Australia Prospective 133 34.9 (18–60) 133:0 RYGB, GP, GG, VG NA 7 Waters 1991 USA Prospective 157 36.3 (19–67) 132:25 RYGB 36 7 Maddi 2001 USA Prospective 52 38.4 (19–56) 46:6 12 8 Dixon 2003 Australia Prospective 487 41.2 ± 9.7 414:73 GB 48 7 Mamplekou 2005 Greece Prospective 59 37.7 ± 10.7 45:14 VG 24 7 Burgmer 2007 Germany Prospective 149 38.8 ± 10.3 102:47 GB, VG 24 8 Karlsson 2007 Sweden Prospective 655 47.0 ± 5.7 RYGB, GB, VG 120 8 Nickel 2007 Germany, Austria Prospective 21 38.0 ± 9.5 50:0 GB 72 7 Scholtz 2007 UK Prospective 29 39.0 ± 9.0 28:1 GB 60 8 Schowalter 2008 Germany Prospective 248 38.5 ± 10.1 200:48 GB 67 8 Andersen 2010 Norway Prospective 50 37.9 ± 7.9 28:22 BPD 24 8 Kruseman 2010 Switzerland Prospective 141 40.0 ± 10.0 80:0 RYGB 96 7 Thonney 2010 Switzerland Prospective 43 39.2 ± 1.4 43:0 RYGB 24 9 Assimakopoulos 2011 Greece Prospective 59 36.0 (18–56) 59:0 RYGB, BPD, SG 12 7 Lier 2011 Norway Prospective 127 41.3 ± 10.3 94:33 RYGB 12 8 Strain 2011 USA Retrospective 77 52.3 ± 12.7 54:23 SG 24.6 6 Zwaan 2011 Germany Prospective 107 37.5 ± 9.7 75:32 RYGB, GB 36 8 Jarvholm 2012 Sweden Prospective 37 16.6 ± 1.3 25:12 RYGB 3 8 Ortega 2012 Spain Prospective 60 44.1 ± 10.9 46:14 RYGB 12 7 Rutledge 2012 USA Prospective 55 52.5 ± 7.4 17:38 RYGB, GB 60 8 Sysko 2012 USA Prospective 101 15.8 ± 1.1 73:28 GB 12 7 Lier 2013 Norway Prospective 127 41.3 ± 10.3 125:44 RYGB 12 8 Burgmer 2014 Germany Prospective 148 38.8 ± 10.2 101:47 GB, VG 96 8 Castellini 2014 Italy Prospective 83 45.3 ± 10.1 75:8 RYGB, BPD, GB 12 7 Hayden 2014 Australia Prospective 204 45.2 ± 11.5 168:36 GB 24 8 Ivezaj 2014 USA Prospective 107 42.7 ± 10.5 94:13 GB 12 7 Matini 2014 Iran Prospective 67 36.8 ± 8.5 55:12 RYGB 6 9 Mitchell 2014 USA Prospective 2146 46.0 (18–76) 1685:461 RYGB, GB, SG, BPD 36 7 Strain 2014 USA Prospective 105 43.5 ± 10.8 76:29 RYGB, BPD, GB, SG 24 7 Tae 2014 Brazil Prospective 32 41.0 ± 11.6 32:0 10 8 Booth 2015 UK Prospective 3045 45.9 ± 10.2 2406:639 RYGB, GB, SG 96 7 Herpertz 2015 Germany Prospective 152 39.1 ± 10.3 102:50 GB, VG 108 8 Sockalingam 2015 USA Prospective 164 43.5 ± 9.7 133:31 RYGB, SG 12 7 White 2015 USA Prospective 357 43.7 ± 10.0 307:50 RYGB 24 7 Jarvholm 2016 Sweden Prospective 82 16.8 ± 1.2 57:25 RYGB 24 8 Kalarchian 2016 USA Prospective 165 46.0 (21-68) 135:30 RYGB, GB@ 36 8 Mack 2016 Germany Prospective 75 45.2 ± 11.6 48:27 SG 48 9 Efferdinger 2017 Austria Prospective 45 44.1 ± 13.3 34:11 RYGB, SG 6 7 Osterhues 2017 Germany Cross-sectional 192 40.7 ± 11.4 136:56 RYGB, SG, GB 19 8 Sockalingam 2017 Canada Prospective 156 45.2 ± 9.3 126:30 RYGB, SG 24 7 Alabi 2018 Mexico Retrospective 73 38.1 ± 9.1 56:17 GB 12 8 Ho 2018 Canada Retrospective 365 44.7 ± 10.0 292:73 RYGB, SG 12 8 Hunsaker 2018 USA Prospective 139 19.2 ± 1.3 162:40 RYGB, SG, GB 24 8 Monte 2018 USA Prospective 59 51.0 ± 13.0 51:8 SG 6 8 Pinto-Bastos 2018 Portugal Prospective 122 42.8 ± 11.4 108:14 RYGB, SG 6 7 Pinto-Bastos 2018 Portugal Prospective 116 47.9 ± 9.2 107:9 6 7 Ribeiro 2018 Brazil Prospective 281 40.7 ± 9.8 225:56 RYGB 60 6 Kalarchian 2019 USA Prospective 104 45.0 (21–68) 84:20 RYGB, GB 60 8 Kalarchian 2019 USA Prospective 69 47.0 (40–54) 58:11 60 8 Hawkins 2020 Canada Prospective 190 44.5 ± 10.6 156:34 RYGB, SG 12 7 Abbreviations: BPD, biliopancreatic diversion; GB, gastric banding; GG, gastrogastrostomy; GP, gastric partitioning; NA, not available; RYGP, Roux-en-Y gastric bypass; SG, sleeve gastrectomy; VG, vertical gastroplasty.

The pooled mean ± SD age of the subjects was 40.3 ± 8.0 years. In the entire cohort, 79.4% were women. There was a significant reduction of BMI from 47.4 ± 3.2 kg/m2 at baseline to 34.5 ± 4.5 kg/m2 post-operatively, with a pooled d+ of −13.3 kg/