Journal of Psychiatric Research xxx (2014) 1e10
Contents lists available at ScienceDirect
Journal of Psychiatric Research journal homepage: www.elsevier.com/locate/psychires
Adipokines as emerging depression biomarkers: A systematic review and meta-analysis F. Carvalho a, *, Davi Q.C. Rocha a, Roger S. McIntyre b, c, Lucas M. Mesquita a, Andre € hler d, Thomas N. Hyphantis e, Paulo M.G. Sales a, Cristiano A. Ko Rodrigo Machado-Vieira f, g, h, Michael Berk i, j, k a
Translational Psychiatry Research Group, Faculty of Medicine, Federal University of Ceara, Fortaleza, CE, Brazil Departments of Pharmacology and Psychiatry, University of Toronto, Toronto, ON, Canada Mood Disorders Psychopharmacology Unit, University of Toronto, Toronto, ON, Canada d Memory Research Laboratory, Brain Institute (ICe), Federal University of Rio Grande do Norte (UFRN), Natal, RN, Brazil e Department of Psychiatry, Medical School, University of Ioaninna, Ioaninna, Greece f National Institute of Mental Health (NIMH), Bethesda, USA g ~o Paulo, USP, Brazil Laboratory of Neuroscience, LIM-27, Department and Institute of Psychiatry, University of Sa h Center for Interdisciplinary Research in Applied Neuroscience (NAPNA), USP, Brazil i IMPACT Strategic Research Centre, School of Medicine, Deakin University, Geelong, Vic., Australia j Florey Institute of Neuroscience and Mental Health, Australia k Orygen Youth Health Research Centre, University of Melbourne, Parkville, Vic., Australia b c
a r t i c l e i n f o
a b s t r a c t
Article history: Received 27 June 2014 Received in revised form 6 August 2014 Accepted 7 August 2014
Adiponectin, leptin and resistin may play a role in the pathophysiology of major depressive disorder (MDD). However, differences in peripheral levels of these hormones are inconsistent across diagnostic and intervention studies. Therefore, we performed meta-analyses of diagnostic studies (i.e., MDD subjects versus healthy controls) and intervention investigations (i.e., pre-vs. post-antidepressant treatment) in MDD. Adiponectin (N ¼ 1278; Hedge's g ¼ 0.35; P ¼ 0.16) and leptin (N ¼ 893; Hedge's g ¼ 0.018; P ¼ 0.93) did not differ across diagnostic studies. Meta-regression analyses revealed that gender and depression severity explained the heterogeneity observed in adiponectin diagnostic studies, while BMI and the difference in BMI between MDD individuals and controls explained the heterogeneity of leptin diagnostic studies. Subgroup analyses revealed that adiponectin peripheral levels were significantly lower in MDD participants compared to controls when assayed with RIA, but not ELISA. Leptin levels were significantly higher in individuals with mild/moderate depression versus controls. Resistin serum levels were lower in MDD individuals compared to healthy controls (N ¼ 298; Hedge's g ¼ 0.25; P ¼ 0.03). Leptin serum levels did not change after antidepressant treatment. However, heterogeneity was significant and sample size was low (N ¼ 108); consequently meta-regression analysis could not be performed. Intervention meta-analyses could not be performed for adiponectin and resistin (i.e., few studies met inclusion criteria). In conclusion, this systematic review and meta-analysis underscored that relevant moderators/confounders (e.g., BMI, depression severity and type of assay) should be controlled for when considering the role of leptin and adiponectin as putative MDD diagnostic biomarkers. © 2014 Elsevier Ltd. All rights reserved.
Keywords: Major depressive disorder Meta-analysis Adiponectin Leptin Resistin Biomarkers
1. Introduction
* Corresponding author. Department of Clinical Medicine, Federal University of , Faculty of Medicine, Rua Prof. Costa Mendes, 1608, 4 andar, 60430-040, Ceara Fortaleza, CE, Brazil. Tel./fax: þ55 8532617227. E-mail addresses:
[email protected],
[email protected] (A.F. Carvalho).
Current diagnostic practice for major mental disorders, including major depressive disorder (MDD) is based on the clustering of symptoms and other clinical features (Pizzagalli, 2011). MDD is frequently not properly recognized in diverse ‘real world’ clinical settings (Craven and Bland, 2013; Kessler et al., 2007; Lake and Baumer, 2010; Yan et al., 2013). This delay in diagnosis hinders early treatment intervention, leading to worse outcomes due to the
http://dx.doi.org/10.1016/j.jpsychires.2014.08.002 0022-3956/© 2014 Elsevier Ltd. All rights reserved.
Please cite this article in press as: Carvalho AF, et al., Adipokines as emerging depression biomarkers: A systematic review and meta-analysis, Journal of Psychiatric Research (2014), http://dx.doi.org/10.1016/j.jpsychires.2014.08.002
2
A.F. Carvalho et al. / Journal of Psychiatric Research xxx (2014) 1e10
dysregulation of several measureable pathophysiological pathways in the central nervous system, including but not limited to a dysfunctional activation of the hypothalamic-pituitary adrenal (HPA) axis (Pariante and Lightman, 2008), inflammation (Gold et al., 2013) and the generation of oxidative and nitrosative stress (O&NS) (Moylan et al., 2013). Despite this, there are currently no validated peripheral biomarkers for the diagnosis, treatment selection and response prediction in MDD (Breitenstein et al., 2014). The development of biomarkers for MDD and their incorporation into clinical practice promises to revolutionize the landscape of health care. Since the discovery of leptin by Zhang et al. (1994), its role in metabolism and homeostasis has been extensively investigated (Berman et al., 2013; Licinio et al., 2004; Paz-Filho et al., 2008; Zhang et al., 1994). It is increasingly recognized that the adipose tissue is not an inert tissue devoted to energy storage, rather being a metabolically active endocrine organ capable of secreting a number of bioactive products referred to as ‘adipokines’, which include adiponectin (Maeda et al., 1996), leptin, and resistin (Steppan et al., 2001). Adipokines are also secreted by diverse tissues, including but not limited to macrophages, myocytes, and pancreatic cells (Arnoldussen et al., 2014). The adipocyte-brain crosstalk is mediated to a large extent by adipokines and this circuit plays a pathophysiological role beyond obesity and cardiometabolic conditions (Nakamura et al., 2013; Paz-Filho et al., 2010). Leptin influences neurotransmitters such as dopamine (Ishibashi et al., 2012) and impacts gray matter plasticity (London et al., 2011). Consequently, a putative role for leptin, adiponectin and resistin in the pathophysiology of neuropsychiatric conditions associated with metabolic abnormalities, including MDD has emerged (vide infra) (Diniz et al., 2012; Liu et al., 2012; Lu et al., 2006; Weber-Hamann et al., 2007; Yamada et al., 2010). Adiponectin is a polypeptide that regulates glucose levels as well as fatty acid breakdown (Yildiz et al., 2004). It is exclusively secreted by adipocytes as an abundant adipose-derived serum protein. Adiponectin exerts insulin-sensitizing and either inflammatory or anti-inflammatory effects (Kwon and Pessin, 2013; Wan et al., 2014). AdipoR1 and AdipoR2, the cognate adiponectin receptors, are expressed in discrete brain areas related to mood regulation, including the hippocampus (Liu et al., 2012). Adiponectin exerts antidepressant-like effects in the social-defeat stress animal model of depression (Liu et al., 2012). Adiponectin haploinsufficiency blunts glucocorticoid-mediated negative feedback on the HPA axis (Liu et al., 2012). Notwithstanding that plasma levels of adiponectin are negatively correlated with obesity, waist circumference and visceral fat in humans, metabolically healthy obese subjects have peripheral levels of adiponectin similar to lean individuals (Arita et al., 2012; Cohen et al., 2011; Doumatey et al., 2012). Some reports point to higher adiponectin serum levels in MDD subjects compared to healthy controls (Jow et al., 2006), while other investigators found either lower levels (Cizza et al., 2010; Zeman et al., 2010) or unaltered (Kotan et al., 2012) peripheral levels in MDD individuals versus controls. Leptin circulates as a 16-kDa protein, and is the product of the ob gene. It is mainly synthesized by adipose tissue in proportion to percentage body fat (Zupancic and Mahajan, 2011). This peptide is transported across the bloodebrain barrier by a saturable process to exert its central effects (Zupancic and Mahajan, 2011). Leptin has antidepressant and anxiolytic activities in rodents (Liu et al., 2010). Diet-induced obesity in mice is associated with an impaired antidepressant response to leptin which is related to a blunted leptininduced increment in hippocampal BDNF levels compared to mice fed a standard diet (Yamada et al., 2011). The antidepressant activity of leptin may result from its modulatory effect upon the HPA axis. In food-deprived ob/ob mice, systemic administration of leptin
lowers corticosterone levels and prevents the induction of CRH synthesis in the paraventricular nucleus (PVN) (Huang et al., 1998). Leptin levels were associated with an elevated risk of depression onset in older men with a significant amount of visceral fat (Milaneschi et al., 2012). Several studies have found lower serum leptin levels in individuals with MDD compared to healthy controls (Jow et al., 2006; Kraus et al., 2001), whereas other studies in women with MDD found that plasma leptin levels were significantly increased (Esel et al., 2005; Rubin et al., 2002; Zeman et al., 2009). Similarly, some studies reported that leptin levels are variously increased (Esel et al., 2005; Kraus et al., 2002) or not changed by antidepressant treatment (Kraus et al., 2002; Schilling et al., 2013). Lastly, some studies suggest that leptin may be a biomarker of risk for de-novo depression (Pasco et al., 2008). The protein resistin is related to insulin resistance in rodents (Schwartz and Lazar, 2011). Some studies found elevated peripheral resistin levels in human obesity (Degawa-Yamauchi et al., 2003; Owecki et al., 2011), whereas other investigations found resistin down-regulated in obesity (Way et al., 2001). The findings relating resistin to MDD are inconsistent across studies. Lower resistin serum levels in individuals with MDD compared to healthy controls are reported (Aliyazicioglu et al., 2011a), but not replicated (Papakostas et al., 2013). In addition, resistin levels were positively correlated with cortisol levels in MDD patients (Weber-Hamann et al., 2007). Furthermore, there was a significant decrease in resistin serum levels in patients receiving antidepressant treatment who remitted from depression (Weber-Hamann et al., 2007). In order to clarify the inconsistent findings on the associations between peripheral levels of adiponectin, leptin and resistin both as putative diagnostic as well as treatment response biomarkers in MDD, we performed a meta-analysis of the available evidence. We hypothesized that there would be significant heterogeneity between studies (for example, related to age, body mass index and gender). Therefore, this review also aimed to identify potential confounders. As far as we know no previous meta-analyses had been published on the role of akipokines as depression biomarkers. 2. Material and methods 2.1. Study selection 2.1.1. Search strategy Articles for review were identified from the PubMed/MEDLINE, EMBASE, and Web of Science databases from inception to January 12, 2014. The standardized search algorithms are detailed in Supplementary Material S1. Search terms included ‘akipokines’, ‘leptin’, ‘adiponectin’ and ‘resistin’ cross-referenced with ‘depression’, ‘major depressive disorder’ and ‘depressive’. This search strategy was augmented by manual searches were performed on reference lists of included articles. We also tracked citations of included articles and relevant reviews using Google Scholar. Authors were contacted to provide additional data when necessary. We followed the Preferred Items for Reporting of Systematic Reviews and Meta-Analyses (PRISMA) statement (Moher et al., 2009). 2.1.2. Identification of eligible studies Eligible articles included original studies in any language which measured leptin, adiponectin and/or resistin levels in plasma or serum in patients with major depressive disorder and healthy controls (diagnostic studies), or measured these adipokines at baseline and after a trial with a standard antidepressant agent (referred hitherto as intervention studies). Eligible studies included participants who fulfilled Diagnostic and Statistical Manual of Mental Disorders (DSM) or International Classification of Diseases (ICD) criteria for MDD based on a validated structured or semi-
Please cite this article in press as: Carvalho AF, et al., Adipokines as emerging depression biomarkers: A systematic review and meta-analysis, Journal of Psychiatric Research (2014), http://dx.doi.org/10.1016/j.jpsychires.2014.08.002
A.F. Carvalho et al. / Journal of Psychiatric Research xxx (2014) 1e10
structured clinical interview (e.g., Structured Clinical Interview for DSM-IV [SCID-I] (Frist et al., 1996), Composite International Diagnostic Interview [CIDI] (Wittchen, 1994), Diagnostic Interview Schedule [DIS] (Robins et al., 1981)) or a screening instrument. Case series or case reports were excluded. Studies including MDD participants with co-morbid major mental disorders (e.g., substance use disorders or anxiety disorders) were excluded from this review. Finally, studies involving bipolar disorder participants on depressive episodes were also excluded from this review. Two investigators independently reviewed the articles for eligibility. If either deemed an article as potentially eligible based on title/abstract screening, then a full-text review was performed. Final decisions regarding eligibility were made by consensus following the full-text review. 2.2. Data extraction For each study, data were independently extracted by two authors (DQCR and LMM) and entered in a standardized form. Discrepancies were resolved by consensus. The following variables were extracted: (1) mean and standard deviation of leptin, adiponectin and/or resistin serum levels for each group; (2) demographic, clinical and treatment characteristics (including age, gender distribution, body mass index (BMI), scores on MDD rating scales, depression subtype, previous treatments, type of antidepressant treatment, depression severity, differences in mean BMI between MDD individuals and healthy controls); (3) type of assay (ELISA or RIA) and (4) study design (e.g., caseecontrol or trial). When an included study measured serum or plasma levels of adiponectin, leptin and/or resistin at different time points, we considered the last available observation. When an eligible study did not provide extractable data, corresponding authors were contacted. 2.3. Quantitative analyses Meta-analytical calculations were carried out using Comprehensive Meta-Analyses version 2.0 (CMA 2.0) with statistical significance set at P 0.05. Random effects models was used to estimate effect sizes using Hedge's g (±95% Confidence intervals; 95CI) on the difference between each adipokine (i.e., adiponectin, leptin and/or resistin) levels between MDD individuals and healthy controls (diagnostic studies), or the difference between each adipokine serum levels before and after antidepressant treatment (intervention studies). Heterogeneity between studies were assessed through the Q statistic. Publication bias was assessed by funnel plot asymmetry inspection and the Egger test (Egger et al., 1997). Whenever the heterogeneity between studies was not significant, fixed effects models were used to estimate Hedge's g. We evaluated the potential moderating effects of clinical variables (depression severity, whether the diagnosis was performed through a semi-structured interview, overall sample BMI, mean BMI difference between MDD individuals and controls), socio-demographic factors (% of female in overall sample; difference in % of females between MDD individuals and controls) and type of assay (ELISA versus RIA). We also conducted subgroup analyses. To investigate whether a particular study determined the summary measure in each metaanalysis, sensitivity analyses were performed. 3. Results 3.1. Search results and study characteristics Of 1185 unique references, 1120 were excluded after title/abstract screening. Sixty five references were selected for full-text
3
review. Finally, 23 original studies were included in the systematic review (Fig. 1). Please see Tables S1A and S1B for a description of primary reasons for exclusion (Supplementary Material). The authors of meeting abstracts were electronically contacted by the authors. Characteristics of the included diagnostic (i.e., MDD individuals vs. healthy controls) and intervention studies are presented in Supplementary Tables S2A and S2B (Supplementary Material), respectively. Primary data which could be extracted in meta-analyses are depicted in Tables S3A (diagnostic studies) and S3B (intervention studies). Please see supporting on-line material in the journal's web site for reference lists of excluded (i.e., after full-text review) as well as included studies. Diagnostic metaanalyses were performed for adiponectin (9 studies; N ¼ 1278), leptin (11 studies; N ¼ 893) and resistin (2 studies; N ¼ 298) (vide infra). Papakostas et al. (2013) reported a pilot and a replication study; for each study resistin data could be extracted (diagnostic meta-analysis). Standard meta-analyses techniques were performed to control for heterogeneity (meta-regression and subgroup analyses) for the adiponectin and leptin diagnostic meta-analyses. There were few included studies for the resistin diagnostic metaanalysis. Therefore, these results should be considered exploratory. 3.2. Publication bias After inspection of funnel plots of the diagnostic meta-analyses (Supplementary Fig. 1), there was no evidence of publication bias for adiponectin (Egger's intercept ¼ 6.15; 95% CI ¼ 14.09 to 1.79, P ¼ 0.10), leptin (Egger's intercept ¼ 3.07; 95% CI ¼ 7.10 to 13.2, P ¼ 0.51) or resistin (Egger's intercept ¼ 5.23; 95% CI ¼ 38.01 to 48.49, P ¼ 0.36). 3.3. Meta-analysis of diagnostic studies Significant heterogeneity between studies included for the adiponectin diagnostic meta-analysis was observed (Q ¼ 96.3; df ¼ 8; P < 0.001). A significant summary difference in adiponectin peripheral levels between MDD individuals and controls was not observed (Hedge's g ¼ 0.35, P ¼ 0.16) (Fig. 2). Significant heterogeneity between studies included for the leptin diagnostic meta-analysis was shown (Q ¼ 96.3; df ¼ 11; P < 0.001). No differences in leptin peripheral levels between subjects with MDD and healthy controls was verified (Hedge's g ¼ 0.018; P ¼ 0.93) (Fig. 2). As there was no significant heterogeneity between studies included in the resistin diagnostic meta-analysis, a fixed effects model was applied. Resistin levels were significantly lower in individuals with MDD compared to healthy controls (Hedge's g ¼ 0.25; P ¼ 0.03). 3.3.1. Meta-regression analyses Since there was significant heterogeneity observed between studies included in the adiponectin and leptin diagnostic metaanalyses, meta-regression analyses were performed to verify possible influence of moderator variables. For adiponectin, differences in peripheral levels between participants with MDD and healthy volunteers were significantly moderated by differences in the percentage of female participants across studies (i.e., MDD group minus control group) (Table 1). Greater differences in the percentage of females were reflected by a lower difference between adiponectin levels among MDD individuals and controls. Differences in adiponectin serum levels between MDD individuals and controls were also significantly moderated by depression severity (Table 1; Fig. 3). More severe depression was associated with higher differences in adiponectin peripheral levels between MDD subjects and healthy volunteers.
Please cite this article in press as: Carvalho AF, et al., Adipokines as emerging depression biomarkers: A systematic review and meta-analysis, Journal of Psychiatric Research (2014), http://dx.doi.org/10.1016/j.jpsychires.2014.08.002
4
A.F. Carvalho et al. / Journal of Psychiatric Research xxx (2014) 1e10
Fig. 1. PRISMA flow diagram of study selection process.
There was a significant moderator effect of overall sample BMI across leptin diagnostic studies (Table 1; Fig. 3). A higher overall sample BMI was associated with a significantly higher difference in leptin serum levels between MDD patients and healthy controls. A similar pattern was observed for differences in BMI between MDD individuals and healthy controls.
individual study had an independent influence in the summary effect size measure. These analyses revealed that the removal of each individual study did not alter the overall direction of the findings for each diagnostic meta-analysis (see Supplementary Fig. 2A and 2B). 3.4. Qualitative systematic review of intervention studies
3.3.2. Subgroup analyses Subgroup analyses were also performed to control for heterogeneity between studies for the adiponectin and leptin diagnostic meta-analyses. For adiponectin, there was a significant influence of the type of assay (RIA versus ELISA). When adiponectin serum levels were measured with RIA, significantly lower peripheral levels of adiponectin were observed in participants with MDD compared to healthy controls (Fig. 4). Importantly, significant heterogeneity between studies which had measured adiponectin levels with RIA was not verified (Q ¼ 2.13; df ¼ 3; P ¼ 0.51). Notwithstanding leptin peripheral levels did not differ between MDD individuals with severe depression versus healthy controls, but were significantly higher for participants with mild/moderate MDD compared to controls (Fig. 4). Significant heterogeneity was no longer observed for studies which had included participants with mild/moderate MDD (Q ¼ 2.13; df ¼ 4; P ¼ 0.30). 3.3.3. Sensitivity analyses Sensitivity analyses were carried out for the adiponectin and leptin diagnostic meta-analyses to determine whether each
Few intervention studies met inclusion criteria for systematic review (see Supplementary Table S2B) and meta-analyses could not be performed for either adiponectin or resistin. While a metaanalysis could be performed for leptin (Supplementary Table S3B), there was a high and significant degree of heterogeneity in the performed meta-analysis (data available upon request). Overall, there was no difference in leptin peripheral levels before or after antidepressant treatment (Supplementary Fig. 3). There were three intervention studies for adiponectin. A study included 34 participants with severe DSM-IV MDD (Pinar et al., 2008). These subjects received a 30-day maprotiline (150 mg/day) trial. By the end of the trial, participants had a significant increase in body weight and insulin resistance. Adiponectin levels significantly decreased compared to baseline (29.66 ± 5.64 to 24.00 ± 6.26, P < 0.001). In the second study, after receiving placebo for six days, MDD participants were randomized to either amitriptyline (n ¼ 24; up to 150 mg/day) or paroxetine (n ¼ 11; up to 40 mg/day) for 35 days. No significant differences in adiopnectin serum levels after antidepressant treatment were observed compared to baseline (Weber-Hamann et al., 2007). A third study in 43 hospitalized
Please cite this article in press as: Carvalho AF, et al., Adipokines as emerging depression biomarkers: A systematic review and meta-analysis, Journal of Psychiatric Research (2014), http://dx.doi.org/10.1016/j.jpsychires.2014.08.002
A.F. Carvalho et al. / Journal of Psychiatric Research xxx (2014) 1e10
5
Fig. 2. Forest Plot of the diferences in serum Adiponectin (A), leptin (B) and resistin (C) levels of major depressive disorder vs. healthy control individuals using Hedge's g in Random Effects Models (A and B) or Fixed Effects Model (C). The Forest Plot depicts the estimated difference in serum levels of depressed vs. control individuals (squares) and its 95% confidence intervals (horizontal black lines). *Pilot Study; **Replication Study; D Non-atypical major depressive disorder; ¶Atypical major depressive disorder.
patients with severe MDD administered venlafaxine, fluoxetine, maprotiline or mirtazapine for 4 weeks (Chen et al., 2010). Endpoint adiponectin serum levels were unaltered from baseline. A single intervention trial for resistin met inclusion criteria (Weber-Hamann et al., 2007). In this study, participants with MDD were randomized to either amytryptiline (n ¼ 24) or paroxetine (n ¼ 11). At the end of the trial, resistin levels significantly decreased only in MDD participants who had achieved clinical remission.
4. Discussion The goal of this study was to perform a systematic review and meta-analysis of diagnostic and intervention studies of adipokines, namely adiponectin, leptin and resistin in MDD. While there were no significant differences in adiponectin and leptin peripheral levels between participants with MDD and healthy controls, the study identified relevant moderators and potential confounders. Resistin serum levels were significantly lower among MDD
Please cite this article in press as: Carvalho AF, et al., Adipokines as emerging depression biomarkers: A systematic review and meta-analysis, Journal of Psychiatric Research (2014), http://dx.doi.org/10.1016/j.jpsychires.2014.08.002
6
A.F. Carvalho et al. / Journal of Psychiatric Research xxx (2014) 1e10
Table 1 Meta-regression of included studies. Variable Adiponectin % Females difference (MDD-controls) % Females in sample Sample age Age difference (MDD-controls) Sample BMI BMI difference (MDD-controls) Depression severity Semi-structured interview Type of assay Leptin % Females difference (MDD-controls) % Females in sample Sample age Age difference (MDD-controls) Sample BMI BMI difference (MDD-controls) Depression severity Semi-structured interview Type of assay
Variable type
d.f.
Point estimate (b)
95% CI of b
Continuous Continuous Continuous Continuous Continuous Continuous Categorical (Mild/Moderate vs. Severe) Categorical (No vs. Yes) Categorical (ELISA vs. RIA)
6 11 8 8 8 8 5
0.063 0.018 0.024 0.054 0.020 0.122 1.274
0.020 to 0.105 0.009 to 0.013 0.055 to 0.006 0.079 to 0.186 0.233 to 0.193 0.487 to 0.730 2.026 to 0.522
0.004 0.186 0.115 0.429 0.856 0.695 0.001
46.74 N/A N/A N/A N/A N/A 73.73
8
0.168
1.291 to 0.955
0.769
N/A
7
0.071
1.163 to 1.305
0.910
N/A
Continuous Continuous Continuous Continuous Continuous Continuous Categorical (Mild/Moderate vs. Severe) Categorical (No vs. Yes) Categorical (ELISA vs. RIA)
11 11 10 10 11 11 8
0.001 0.018 0.056 0.023 0.325 0.336 0.805
0.071 to 0.070 0.009 to 0.044 0.003 to 0.115 0.209 to 0.163 0.182 to 0.469 0.113 to 0.559 2.072 to 0.463
0.986 0.186 0.064 0.812
