Meta-analysis of the COMT Val158Met polymorphism in major depressive disorder: the role of gender.

Abstract Objectives: Many studies have reported an association of the COMT Val158Met polymorphism and major depressive disorder (MDD), although with conflicting results. The role of gender is a possible modulator. To overcome the problem of poor sample size detecting genes of small effect, we perform a meta-analysis of the current literature, investigating the influence of the COMT Val158Met polymorphism on the pathogenesis of MDD, with a major focus on the effect of gender. Methods: Out of 977 retrieved articles, 21 included case–control studies allowed the analysis of 9005 patients with MDD and 12,095 controls. Allelic and genotypic pooled odds ratios (OR) were calculated for the total sample and gender-subgroups. Results: In the absence of publication bias, allelic and genotypic analyses showed no significant association in the total sample, as well as in gender-specific subgroups. Sensitivity analysis did not alter the ORs. Conclusions: The results imply a complex nature of the genotype × phenotype interaction. Further studies of the COMT gene or the locus remain to be justified given the important positional and functional relevance and the plethora of gender-specific findings. A possible way to further dissect this topic is shifting the focus to gene-based or genome-wide analyses of intermediate phenotypes.


Introduction
Major depressive disorder (MDD) is a common psychiatric disorder, being predicted to be the second leading cause of disability worldwide by 2020 (Murray and Lopez 1996). It is a heterogeneous disorder caused by a large number of genetic and environmental factors with complex interactions, each with a relatively small contribution (Ebmeier et al. 2006). The heritability -the contribution of genetic factors to the onset of MDDhas been estimated from twin studies at about 40% (Sullivan et al. 2000) and is higher in women than in men (Jansson et al. 2004;Kendler et al. 2006) pointing towards a genetic basis for gender differences.
Despite the large focus on the serotonergic system, the role of dopamine in the aetiology of MDD has already been described in 1965 (Schildkraut 1995), and since then the idea of disturbances in dopamine signalling as one of the main contributors to the pathology of MDD has been underlined in literature. Many symptoms of depression or changes in certain behavioural aspects like mood, motivation, attention, decision making or psychomotor speed are regulated by dopamine (Opmeer et al. 2010). In the prefrontal cortex, the inactivation of dopamine and thus the termination of action in the synapse are mainly regulated by extracellular degradation by catechol-O-methyltransferase, a major catecholamine-degrading enzyme encoded by a single gene on the chromosome 22q11.1.

The COMT gene
The COMT gene consists of six exons. Two promoter regions enable the production of two COMT isoforms, the membrane-bound (MB-COMT) and the 50 amino acids shorter soluble (S-COMT) form encoded by the COMT gene (Bertocci et al. 1991). Most of the human tissues express both the short and the long COMT transcripts except of the adult human brain, where only the long transcript has been found (Tenhunen et al. 1994;Hong et al. 1998) making it the predominant variant in the brain. MB-COMT is assumed to be more important in the prefrontal cortex and in the striatal neurones than in other brain regions. It is responsible for over 60% of dopamine degradation in the prefrontal cortex and 15% in the striatum (Karoum et al. 1994).
The activity of the COMT enzyme is influenced mainly by a functional single nucleotide polymorphism, a G to A substitution at codon 158 (rs4680) of the MB-COMT sequence (corresponding to position 108 of S-COMT) resulting in a valine to methionine exchange (Lachman et al. 1996). This single nucleotide polymorphism is broadly referred to as Val158Met (Lachman et al. 1996;Malhotra et al. 2002). A 3-4-fold reduction in enzymeactivity and thermal stability is the consequence (Lachman et al. 1996), making this gene one of the most studied candidate genes in mental disorders. The alleles are co-dominantly causing a trimodal phenotype distribution (Floderus et al. 1981)

COMT in depression
Since genotyping became an accessible method and with the knowledge that the Val158Met polymorphism largely influences the COMT enzymatic activity, the hypothesis whether the high-activity Val allele and the resulting increased dopamine degradation determine the diagnosis of a depressive disorder was tested in several studies, although with contradictory results (Opmeer et al. 2010): some studies found an association between the high-activity COMT Val allele, particularly the COMT Val/Val genotype and early-onset MDD in adults (Massat et al. 2005(Massat et al. , 2011 and between the Val/Val genotype and depressive symptoms in children (Sheikh et al. 2013). On the contrary, other studies found that the presence of the low-activity Met allele was significantly associated with depression in adult-onset MDD (Ohara et al. 1998) and that symptoms of depression in pregnancy and post-partum-depression increase in carriers of the Met allele (Doornbos et al. 2009, Comasco et al. 2011. Predominantly, studies found no association of the Val158Met polymorphism with the diagnosis of MDD (Kunugi et al. 1997;Frisch et al. 1999;Cusin et al. 2002;Funke et al. 2005;Garriock et al. 2006;Serretti et al. 2006;Jabbi et al. 2007;Baune et al. 2008;Huuhka et al. 2008;Ising et al. 2009;Potter et al. 2009;Huang et al. 2010;Kocabas et al. 2010;Lewis et al. 2010;Muglia et al. 2010;Rietschel et al. 2010;Utge et al. 2010;Bosker et al. 2011;Demirkan et al. 2011;Shi et al. 2011;Soronen et al. 2011) or with scores on depression scales (Henderson et al. 2000;Anttila et al. 2008;Baekken et al. 2008;Wray et al. 2008;Illi et al. 2010;Luciano et al. 2010;Terracciano et al. 2010).

COMT and gender
As many conditions in human life, most of the psychiatric disorders show gender differences regarding incidence, clinical features, age of onset or outcome (Piccinelli and Wilkinson 2000;Harrison and Tunbridge 2008;Diflorio and Jones 2010;Essau et al. 2010;McLean et al. 2011;Ochoa et al. 2012). The influence of sex hormones, sex chromosome genes and epigenetic mechanisms such as DNA methylation and chromatin modifications (Kaminsky et al. 2006) as well as influences of autosomal genes have been postulated to result in noteworthy sexual dimorphism (Harrison and Tunbridge 2008).
The sexually dimorphic effect of COMT was already ascribed to oestrogenic regulation in 1971 by Cohn and Axelrod (1971). One of the possible mechanisms has been described by Xie et al. (1999) -the oestrogenmodulated regulation of COMT transcription by interaction of the oestrogen receptor complex with a response element in the COMT promoter region, where oestrogen can inhibit COMT gene transcription, although it seems unlikely to be the only mechanism responsible for gender-based dimorphism.
A number of clinical studies reported a lower COMT activity in women than in men (Floderus et al. 1981;Boudikova et al. 1990;Chen et al. 2004) and in psychiatric research, studies show a significant association between gender-associated differences in psychiatric phenotypes according to COMT genotype. For instance, the functional Val(158)Met polymorphism in COMT is associated with obsessive-compulsive disorder in men (Pooley et al. 2007), with anxiety phenotypes in women (Eley et al. 2003;Enoch et al. 2003;Stein et al. 2005), and has a greater impact on cognitive function in boys than in girls (Barnett et al. 2007).
When considering gender as a covariate, there are more results pointing toward a significant role of the COMT Val158Met polymorphism in the aetiology of major depression. In a large population-based study, Aberg et al. (2011) found that depressed individuals displayed a higher frequency of Met/Met and Met/Val genotypes compared to controls. This association was found among men only. Further, depressed men homozygous for the Val-allele, had a higher motivation level than depressed men with a Met-variant.
In a study by Baekken et al. (2008), the Met/Met genotype and Met allele were significantly less common among depressed men compared to controls. The COMT Val allele was found to be associated with higher pre electro-convulsive therapy (ECT) severity of depression and better treatment response to ECT -these findings were restricted to the female subgroup (Domschke et al. 2010). Nyman et al. (2011) examined aetiological factors for the development of depression in a large Finnish birth cohort. An association of the COMT genotype (rs4680) with depression was detected particularly in male individuals at high developmental risk (low birth weight, late motor development, late development of speech).
In suicide research, one of the major contributors to mortality in depression, several gender-specific results occur: for male suicide attempters, there was a significant difference in COMT Val158Met genotype distributions and allele frequencies compared to controls. The Val/Val genotype and Val carrier status were more frequent in suicide attempters than in control subjects (Lee and Kim 2011). Another study found opposite results in suicide completers, the high activity Val/Val genotype being significantly less common in male suicide completers, suggesting a protective factor against suicide in males (Ono et al. 2004).
Animal studies show further evidence towards sexual dimorphism in COMT function (Gogos et al. 1998): dopamine levels of the frontal cortex of COMT knockout mice are increased almost threefold in male COMT -/mice (and twofold in +/-mice) compared to wild-type mice, this effect has not been found in female knockout mice, where the tissue dopamine levels remained unchanged. This effect may occur presumably because of sex-specific compensatory mechanisms (Harrison and Tunbridge 2008).
The results of these multiple studies imply the importance of including gender as a covariate in analyses examining genetic influences on personality, behaviour and more broadly mental states including its pathologies.

Aim of the study
Many studies have reported on an association of the Val158Met polymorphism and psychiatric disorders including MDD, although with conflicting results. The role of gender has been postulated to be a possible modulator, although with the extent of its effect not yet been fully elucidated.
Due to the large sample sizes needed to detect genes of small effect as implicated in aetiology of psychiatric disorders such as MDD, the majority of genetic studies published so far are without doubt statistically underpowered. Therefore, we perform a meta-analysis of the current literature, investigating the influence of the COMT Val158Met polymorphism on the pathogenesis of MDD, with a major focus on the effect of gender as well.

Search strategy
The literature research was conducted in the PubMed database for genetic association case-control studies on COMT Val 158 Met polymorphism published before 25 March 2014, using a broad definition of keywords to minimise the search bias. The following search terms were applied: ''(depression OR depressive disorder OR depressed) AND (COMT OR rs4680 OR catechol-omethyltransferase OR candidate gene*)''; ''genomewide AND depress*''; ''(Val158Met OR rs4680) AND (depress* OR mood disorder* OR major depress*)''. All retrieved abstracts were reviewed to identify studies which examined the relationship between COMT and depression. After retrieving potentially relevant studies, references cited in these publications were reviewed to identify further publications not obtained by PubMed.

Inclusion criteria
Based on this literature research, genetic association studies examining the COMT Val158Met polymorphism were included if they were case-control studies investigating adult (418 years) unrelated patients with MDD and healthy control subjects, where the case status was defined as having a current diagnosis of MDD, diagnosed by a trained professional in an established psychiatric interview fulfilling standard diagnostic criteria (Diagnostic and Statistical Manual of Mental Disorders) for MDD. In the majority of studies, the control-status was established as absence of any psychopathology. In some studies, the screening process of the healthy control subjects was not described explicitly (Kunugi et al. 1997;Arias et al. 2006;Serretti et al. 2006;Huuhka et al. 2008;Illi et al. 2010;Rietschel et al. 2010) which was not considered an exclusion criterion. The comorbidity with an anxiety disorder as a secondary diagnosis in some MDD cases was present in one study (Bosker et al. 2011). This was not considered an exclusion criterion and was examined in the sensitivity analysis.
Different age groups of patients were included (adults, geriatric patients), children and youth were not included in this analysis.
Studies were excluded if: (1) the cases were selected solely by a self-report questionnaire assessing symptoms of depression or reporting on sub-syndromal forms of depression or depression-related phenotypes such as personality traits as well as pure postpartum depression patients, (2) study participants were related, (3) the case or control group contained also bipolar patients or other primary Axis I comorbidity, (4) the data were used in another study on the same polymorphism, (5) the study used others than Caucasian subjects, and (6) was written in another language than English or German.

Data extraction
After identifying potentially relevant studies reporting on a case-control design and assessment of COMT rs4680, the full versions of all articles were obtained. In case more diagnoses were examined in one study, only the data on MDD was used.
For each study, the following information was extracted: first author, year of publication, diagnostic system (DSM-III, DSM-IV, ICD-10) and diagnostic tools for determining case status, sample size, ethnicity, gender and mean age of cases and controls, rs4680 allele and genotype frequencies stratified by gender. In case that not all data were published for a study fulfilling the inclusion and exclusion criteria, the authors were contacted to provide the unpublished data (mainly genotype distributions by gender). All contacted authors replied to our requests; however, not all unpublished data were retrievable.

Statistical analysis
Meta-analysis was performed for case-control studies reporting on COMT Val158Met polymorphism in MDD in adult Caucasian subjects fulfilling all inclusion and none of the exclusion criteria. Allele frequencies and genotype distribution were extracted for cases and controls. The present meta-analysis includes an overall analysis including all available data and subgroup analyses by gender. Overall, the total sample as well as females and males separately were analysed for both allelic and genotypic associations.
The allelic association of the Val allele with the risk of MDD, relative to the Met allele (OR Val vs. Met) was analysed. Furthermore, genotypic ORs were calculated separately for the homozygous Val/Val genotype with the Met/Met genotype as reference, as well as for heterozygous Val/Met genotype, with the Met/Met and the Val/Val genotype as reference. The odds ratios (OR) were pooled according to the methods of DerSimonian and Laird (1986). The significance of the pooled OR was determined by z-test, the degree of heterogeneity was quantified with the I 2 statistics (Higgins and Thompson 2002) which describes the proportion of total variation in study estimates due to heterogeneity. Case-control studies were analysed by random effects meta-analysis in case that at least moderate heterogeneity occurred (I 2 430%), and fixed effects meta-analysis in absence of heterogeneity.
The publication bias was evaluated by visual inspection of the funnel plot (Egger et al. 1997). To evaluate the influence of individual studies on the pooled OR, sensitivity analyses were performed by removing studies not reporting on Hardy-Weinberg equilibrium (HWE; Frisch et al. 1999;Illi et al. 2010) or deviating from HWE (Kocabas et al. 2010), studies not explicitly reporting on the screening process of healthy controls (Kunugi et al. 1997;Arias et al. 2006;Serretti et al. 2006;Huuhka et al. 2008;Illi et al. 2010;Rietschel et al. 2010), studies not reporting on screening for psychiatric comorbidities (Kunugi et al. 1997;Frisch et al. 1999;Serretti et al. 2006;Illi et al. 2010) or the presence of comorbidity-secondary anxiety disorder in one study (Bosker et al. 2011). Furthermore, a sensitivity analysis was conducted excluding studies which examined late life depression (Pan et al. 2009), major depression mixed with dysthymia (Demirkan et al. 2011), patients receiving ECT  or suicidal patients , to rule out potential bias due to different aetiology of depression or more severe manifestation. The OR and 95% confidence interval were recalculated. Cochrane Review Manager Version 5.1 was used for all statistical analyses. P50.05 (two-tailed) were considered statistically significant.

Results
The literature search resulted in 977 articles fulfilling the search criteria of which 58 described case-control studies on COMT Val158Met in unrelated adult subjects. We excluded 34 studies for different reasons described in detail in Figure 1. A total of 24 studies fulfilled all inclusion and none of the exclusion criteria. The raw data for further analysis of three studies were not retrievable after contacting the authors (Cusin et al. 2002;Muglia et al. 2010;Shi et al. 2011), resulting in 21 studies included in the final meta-analysis ( Figure 1). For 15 out of 21 included studies, gender-specific data was provided by the authors. One study consisted of two samples, which were analysed separately (Frisch et al. 1999); hence, reported results include 22 samples of 21 studies with a final sample size of 9005 patients with MDD and 12,095 controls (Table 1).

Heterogeneity between studies
Heterogeneity in meta-analyses indicates, that included studies differ considerably in one or several important aspects, which may have caused differences in results and affected their comparability (Tak et al. 2010). In the majority of our meta-analytic tests, no significant heterogeneity was observed. Merely in the Val/Met vs. Met/Met condition heterogeneity was found in the total sample (I 2 ¼ 42%, P ¼ 0.03, df ¼ 19) and the female subgroup (I 2 ¼ 55%, P ¼ 0.004, df ¼ 15). Results of heterogeneity tests are shown in Table 2.

Publication bias
If trials with statistically significant results are more likely to be published, the so-called publication bias arises. In meta-analyses, this may lead to inflated effect estimates in the hypothesised direction (Tak et al. 2010). A funnel plot -a scatter graph plotting a trial's effect estimate against a measure of prediction (standard error of the effect size SE) -was used to asses publication bias. After visual inspection, the funnel plot shown in Figure 2 is symmetric showing no evidence of publication bias.

Association of COMT Val158Met and MDD
The results of the different tests for association of the genetic polymorphism and risk for MDD in the total sample and the gender subgroups are shown in Table 2 and Supplementary Tables 1-12 (available online at http://dx.doi.org/10.3109/15622975.2015.1083615). If no heterogeneity was observed, fixed effects model was applied, in case of observed heterogeneity, random effects model was used.
In all allelic (Val vs. Met) and genotypic (Val/Val vs. Met/Met, Val/Met vs. Val/Val; Val/Met vs. Met/Met) association analyses neither fixed nor random pooled OR were significant both in the total sample, as well as in females or males.

Sensitivity analyses
To evaluate the influence of individual studies on the results of the meta-analytic tests, sensitivity analysis (selectively excluding studies and recalculating the pooled OR) was conducted. Sensitivity analysis was performed with respect to Hardy-Weinberg equilibrium (HWE) deviations, screening of controls, comorbidity and severity of depression. Studies reporting on genotype frequencies deviating significantly from HWE (Kocabas et al. 2010) or not explicitly stating results of HWE analyses (Frisch et al. 1999;Illi et al. 2010) were removed. Furthermore, studies were removed if the control subjects were not screened for depression (Rietschel et al. 2010) or if screening for depression in controls was not reported (Kunugi et al. 1997;Arias et al. 2006;Serretti et al. 2006;Huuhka et al. 2008;Illi et al. 2010). Moreover, studies allowing comorbidity (Bosker et al. 2011) or not explicitly stating the screening process for comorbidities (Kunugi et al. 1997;Frisch et al. 1999;Serretti et al. 2006;Illi et al. 2010;) were excluded. Finally, a sensitivity analysis was performed excluding studies which examined late life depression (Pan et al. 2009), major depression mixed with dysthymia (Demirkan et al. 2011) or patients with severe depression -patients who underwent ECT  or attempted suicide . Excluding these studies did not substantially alter the ORs as reported above (data not shown).

Discussion
In the present study, we aimed at elucidating an association between the COMT Val158Met functional polymorphism and MDD with a special focus on the role of gender as a moderator. Given the large sample sizes required to asses even small effects of polymorphisms, a meta-analysis represents a crucial technique. As far as we know, this is the largest meta-analysis to date that investigates the role of COMT Val158Met in the aetiology of MDD, and the first one to examine the effect of gender as a variable on a meta-analytic level. Some of the strengths of our study are the wide definition of search terms used to minimise search bias, a conservative approach in study inclusion enhancing the homogeneity of the sample, the acquisition of unpublished (especially gender-specific) data from included studies to   Schosser et al. 2013;Song et al. 2013); diagnosis performed prior to time of examination (Conway et al. 2010;Hatzimanolis et al. 2013); Axis I comorbidities (Russ et al. 2000); non-Caucasian (Ohara et al. 1998;Lee and Kim 2011;Wong et al. 2012;Seok et al. 2013;Wang et al. 2013;Chiesa et al. 2014), unpublished data not retrievable (Cusin et al. 2002;Muglia et al. 2010;Shi et al. 2011). enhance the power and a high level of data transparency, a quality criterion as suggested by Huf et al. (2011).
We analysed data from 9005 patients with MDD and 12095 controls from 21 studies (22 samples from 21 studies). Our results showed no association of the Val158Met polymorphism of the COMT gene with MDD, which is in line with a previously published smaller meta-analysis (Lopez-Leon et al. 2008) that included six studies. Even though there is a growing body of evidence for a sexual dimorphism in MDD due to COMT polymorphisms, the subgroup-analyses according to gender did not show an association on allelic or genotypic level. A simple explanation of the relationship between COMT genotype and MDD seems to be unlikely, as underlined by our results. The precise mechanism of action of a genotype Â phenotype interaction is complex and not fully clear, but published studies point towards multiple possible contributors, outline directions for further research and help to interpret the lack of association in our study despite of multiple positive findings in the literature.
One possible explanation for the lack of association could be the rather broad definition of the MDD diagnosis with a variety of symptoms making one SNP   Antypa et al. (2013) observed that clinical phenotypes such as depression severity, the diagnosis of depression or behavioural endophenotypes were less reliably associated with COMT genetic variation, which might contribute to the negative findings of our study. A more pronounced gene effect was seen in emotion processing systems and antidepressant efficacy, pointing towards a more biological model of action of this gene-phenotype interaction. Specifically, the Met allele was associated with increased activity in limbic areas and prefrontal cortex and was also more likely to show a better response to antidepressant treatment, compared to the Val allele.
In a fMRI study by Swart et al. (2011), COMT Met homozygotes reported more difficulties in verbalising their feelings and the Met allele was associated with attenuated brain activation in posterior cingulate gyrus, supporting the hypothesis that the Met allele modulates neural activation in regions associated with emotional awareness. Similarly, Williams et al. (2010) showed in an fMRI study, that a larger number of Met alleles predicted a hypersensitivity toward negative information and an attenuated processing of positive cues. These patterns of activation had correlates in self-reported negativity bias (perceiving and expecting negative events and outcomes) reflecting risk for depressive disorders. These effects were more apparent for females.
In a study by Weiss et al. (2007), Val homozygosity, as compared to Met homozygosity, was associated with better and faster recognition of negative facial expressions such as anger and sadness.
Underlining the complexity of the issue, contradictory results were found: Wichers et al. (2008) examined the experience of reward in daily life as a relevant factor in the development of depression, finding significantly increased ability to experience reward in everyday life with increased number of Met alleles.
A consistent COMT-by-sex interaction effect on affectrelated personality traits was found by Chen et al. (2011), where males with the Val/Val Genotype showed significantly higher negative emotionality and significantly lower positive emotionality scores compared with Met/ Met Males. Females, however, showed an opposite but non-significant pattern.
In a recent fMRI study, Domschke et al. (2012) examined neural activation correlates of emotional face processing as a neural underpinning for depression and anxiety-related intermediate phenotypes and the influence of COMT Val158Met variant with significant and gender-specific results: the more active Val158 variant increased predominantly left-sided amygdala activity in response to fearful/angry facial stimuli (allele-dose effect) and the influence of the Val allele was strictly female-specific, providing further support for the gender-specific effects of COMT Val158Met on emotional processing. Kempton et al. (2009) also reported on a significant effect of gender on brain activation in an affective processing task with Val/Val carriers showing larger signal magnitude in limbic and paralimbic regions compared to Met/Met carriers, particularly in females.
Another contributor to the diversity of findings might be the gene Â gene interaction as well as the role of haplotypes and epigenetic mechanisms strengthening the association between MDD and genes. Doornbos et al. (2009) reported on a significant role of COMT in the development of peripartal depression, this effect was even stronger in combination with a MAO-A variant. Mandelli et al. (2007) described a significant influence of the COMT Met allele on higher depression scores after adverse life events, which was more pronounced in the presence of the short allele of the serotonin transporter protein gene (SERTPR). Hatzimanolis et al. (2013) found two polymorphisms (rs2020917, rs737865) in the MB-COMT promoter region being significantly associated with female depressive symptomatology. Funke et al. (2005) reported on a significant association between rs2097603 (COMT promoter region) and MDD. A high-risk haplotype of the Val allele of rs4680 with the A allele of rs165599 was significantly predictive of differences in neuroticism and risk for several anxiety disorders and major depression in females but not in males (Hettema et al. 2008). Nackley et al. (2006) described certain haplotypes in COMT exhibiting the largest difference in COMT enzymatic activity due to a reduced amount of translated protein not to an altered amount of COMT mRNA. The effect of COMT on depression might also be masked by epigenetic processes such as the genotypedependent methylation of the COMT Val158 allele modulating lifetime stress perception (Ursini et al. 2011), a significant contributor to the development of depressive states.
Considering the complexity of mental disorders including MDD, the neurochemical pathways involved and the genetic underpinnings, it is not surprising that the interrelation between genetic markers and mental states interacts with other biological factors altering the final effect. Studies examining COMT's effect on frontal lobe function demonstrate that the relationship between genotype and phenotype, such as emotion regulation networks via the prefrontal cortex, illustrates a similar phenomenon. Winterer et al. (2004) suggested an inverted U-shaped model of the relationship between dopamine levels and prefrontal cortex (PFC) function with both super-and sub-optimal PFC dopamine levels impairing cognitive function and thus the presentation of symptoms. COMT activity influences the position of a given individual on this inverted U curve. This phenomenon highlights the difficulty to differentially asses the function of any individual SNP or partially explains the conflicting results regarding COMT.
It is important to view this study in the scope of its limitations.
In general, the results of a meta-analysis always depend on the quality of each included study. By setting conservative inclusion criteria we aimed at solving this issue, nevertheless, between-study heterogeneity (variability between studies) was still partially present. Possible sources of heterogeneity are various clinical and methodological factors in the conducting of the studies, e.g. screening of the healthy participants, assessing comorbid disorders, differences in interviewing styles, etc. In case of heterogeneity, we used randomeffects meta-analyses as a more conservative method with a wider confidence interval. Performing sensitivity analysis and sequentially excluding separate studies to check for the effect of suspected divergences in the conducting was used to further dissect this question. No clear confounding factors were found.
Population stratification could still be a problem due to unknown heterogeneity even though we tried to minimise this risk by including Caucasian samples only. It is known, that even among European countries, allelic frequencies differ remarkably (Palmatier et al. 1999) but a more pronounced selection of participants could lead to loss of power. In the future, single well-powered studies on distinct populations could overcome this problem.
To enable comparisons between ethnic groups and further generalizability of results, future research should also take inclusion of other ethnic groups into account.
To obtain a homogeneous and thus comparable sample, we considered comorbid Axis I disorders an exclusion criterion except in one case (Bosker et al. 2011) with anxiety disorders (comorbidity between depression and anxiety-spectrum disorders is highly prevalent and points towards a possibly shared pathogenesis (van Veen et al. 2012)). The role of the comorbidity as a confounder was assessed in sensitivity analysis excluding studies not explicitly reporting on screening for comorbidities and exclusion of one study with comorbid anxiety disorders, showing no effect on the OR. An overall exclusion of comorbid patients seems partially problematic regarding the high level of occurrence. This issue can be solved or minimised by addressing common pathological pathways or shared phenotypes.
For the purpose of maximising the power, we did not focus on other COMT polymorphisms or gene Â gene interactions and concentrated on Val158Met, the best studied polymorphism in the COMT gene. This might be considered a limitation.
In conclusion, our meta-analysis did not show an association between COMT Val158Met and MDD in the overall analysis as well as in the gender-specific subgroups.
Further study of the COMT gene or the locus remains to be justified given the important positional and functional relevance and the plethora of gender-specific findings. A possible way to further dissect this topic is shifting the focus to gene-based or genome-wide analyses of intermediate phenotypes related to the disease of interest, which provide a homogeneous dimensional measure, can be defined more precisely and are considered to be more closely related to genotype (Gottesman and Gould 2003).