Meta-analysis of serotonin transporter gene promoter polymorphism (5-HTTLPR) association with antidepressant efficacy

European Neuropsychopharmacology (2012) 22, 239–258

www.elsevier.com/locate/euroneuro

REVIEW

Meta-analysis of serotonin transporter gene promoter polymorphism (5-HTTLPR) association with antidepressant efficacy Stefano Porcelli 1 , Chiara Fabbri 1 , Alessandro Serretti ⁎ Institute of Psychiatry, University of Bologna, Bologna, Italy

Received 5 July 2011; received in revised form 9 September 2011; accepted 21 October 2011

KEYWORDS Depression; Serotonin transporter gene promoter polymorphism; HTTLPR; Antidepressants; Treatment response; Meta-analysis

Abstract In the last decade the serotonin transporter gene promoter polymorphism (5-HTTLPR) was likely the most studied genetic variant as predictor of antidepressant response. Nevertheless results are not consistent across studies and previous meta-analysis, since various factors seem to modulate its effect on antidepressant response. With the aim of clarifying this issue, we systematically reviewed literature, selecting 33 studies for an exploratory analysis without any a priori hypothesis. Then we analyzed separately 19 studies performed on Caucasians and 11 on Asians. We tested two phenotypes – remission and response rates – and three genotype comparisons – ll versus ls/ss, ss versus ll/ls and ll versus ss – using the Cochrane review manager. Evaluations were performed separately for SSRIs and mixed/other drugs. Possible clinical modulators were investigated. In the exploratory analysis, we found an association between l allele and l/l genotype and remission. When the analysis was split for ethnic group, in Caucasians we found an association between l allele and both response (OR = 1.58, C.I. 1.16–2.16, p = 0.004), and remission (OR = 1.53, C.I. 1.14–2.04, p = 0.004) in the SSRI group. Only a marginal association between l allele and remission (OR = 1.41, C.I. 1.02–1.95, p = 0.04) survived pooling together mixed antidepressant treatments. In Asians, a small effect of 5-HTTLPR on remission for mixed antidepressants was detected (OR = 2.10, C.I. 1.15–3.84, p = 0.02). Gender, age and age at onset modulated the association in Caucasians. Gender, age and depression severity at baseline modulated the association in Asians. In conclusion, in Caucasians 5-HTTLPR may be a predictor of antidepressant response and remission, while in Asians it does not appear to play a major role. © 2011 Elsevier B.V. and ECNP. All rights reserved.

⁎ Corresponding author at: Institute of Psychiatry, University of Bologna, Viale Carlo Pepoli 5, 40123 Bologna, Italy. Tel.: +39 051 6584233; fax: +39 051 521030. E-mail address: [email protected] (A. Serretti). 1 Authors contributed equally to the manuscript. 0924-977X/$ - see front matter © 2011 Elsevier B.V. and ECNP. All rights reserved. doi:10.1016/j.euroneuro.2011.10.003

240

1. Introduction Nowadays depression has still an enormous economic burden due to both direct and indirect costs, in the order of tens of billions of dollars each year in the US alone (Halfin, 2007; Wang et al., 2003). Several factors are responsible of this economic burden. Among the most relevant ones, there are the still unsatisfactory rates of response and remission (Machado et al., 2006; Papakostas et al., 2006) (47 and 33%, respectively, in the STAR*D cohort) (Trivedi et al., 2006). Furthermore, initial side effects and failure to perceive benefit are common reasons of early discontinuation (Masand, 2003). The detection of factors that could predict response and minimize side effects of treatments therefore remains an essential goal. Clinical factors have been extensively investigated, but results are still controversial and cannot be translated into the clinical practice (Papakostas et al., 2008; Serretti et al., 2009). On the other hand, several lines of evidence suggested that genetic factors contribute for about 50% of the antidepressant response (Angst, 1965; Franchini et al., 1998; Kirchheiner et al., 2004; Maier and Zobel, 2008; Serretti et al., 1998), although so far the inconsistency of results has not allowed clinical applications. Despite this discouraging scenario, some polymorphisms were repeatedly associated to antidepressant response, particularly a polymorphism within the promoter of the serotonin transporter gene (SLC6A4), i.e. the 5-HTTLPR. It is a 44 bp insertion/deletion involving two units in a sequence of sixteen repeated elements and possibly influences serotonin transporter (SERT) expression, with the long (l) allele associated with a twice basal expression compared to the short (s) allele (Heils et al., 1996). 5-HTTLPR is a priori excellent candidate for genetic studies, since SERT is one of the main targets for antidepressant drugs and the primary target for SSRIs. Furthermore, it has been associated with several psychiatric disorders with affective symptomatology (i.e. bipolar disorder, anxiety disorders, eating disorders, substance abuse) and to pathological behaviors and personality traits related to anxiety, impulsivity and stress (Serretti et al., 2006). Concerning antidepressant pharmacogenetics, increasing evidence suggest that in Caucasian l allele is associated with better response, although negative findings have been reported as well (Table 1). On the other hand, in Asians an association between s allele and better response was initially reported, though several following studies reported no associations while only few studies found an association with l allele or l/l genotype (see Table 1 and our previous meta-analysis (Serretti et al., 2007b)). So far other ethnicities have not been investigated enough to reach any conclusion. These contradictory findings between Caucasian and other populations may be due to several reasons. Firstly, l allele is much less frequent in Asian compared to Western populations. Indeed: 1. the l/l genotype is present in 29–43% of Caucasians, but in 1–13% of East Asians (Goldman et al., 2010); 2. the s allele is present in 42% of Caucasians, but in 79% of Asians (Kunugi et al., 1997); 3. the s/s genotype varied from 21.6 to 28.3% in the studies including mainly Caucasian patients, while in studies on Asian patients these frequencies varied between 55.6 and 60.0% (Smits et al., 2004). Thus, the low frequency of l allele and l/l genotype in Asians weakens the association between

S. Porcelli et al. them and response found by some authors. Secondly, other genetic variants within the SERT gene or other related gene may represent further stratification factors. Indeed, a second polymorphism identified within the l allele (rs25531A/G) may determinate a reduced expression of the gene (lG allele), comparable with the expression due to the s allele (Hu et al., 2006). In the light of this finding all the studies performed before the detection of this mutation should be carefully re-examined and future investigations should provide a better covering of the gene (Dong et al., 2009). Finally some recent findings suggested possible interactions between 5-HTTLPR genotype and drug plasma concentration (Lotrich et al., 2008), augmentation strategies (Benedetti et al., 2008; Stamm et al., 2008), life events (Mandelli et al., 2009a) and gender (Walderhaug et al., 2007). Taken together, these data underline the need for further investigations in order to reach any definitive conclusion. Our previous meta-analysis exploring the association between 5-HTTLPR and antidepressant response, reported positive correlations between l allele and l/l genotype and both response and remission (Serretti et al., 2007b). Despite these results, the increasing amount of data about the topic, especially after the publication of STAR*D results, may be useful to further clarify the role of this relevant polymorphism. These considerations leaded to a new meta-analysis on 5-HTTLPR and antidepressant response. Conversely to our findings, authors did not find any association between 5-HTTLPR and both antidepressant response and remission (Taylor et al., 2010). Nevertheless, they did not perform separate meta-analyses on the basis of the ethnicity that, as stated above, is a well-known confounding factor in pharmacogenetic studies. The confounding role of ethnicity seems even more relevant for 5-HTTLPR, since the very different frequencies of its alleles among different populations (see above) and the increasing evidence of a different effect of 5-HTTLPR in Caucasian and Asian populations (Table 1). As a matter of fact, also Taylor and colleagues observed that ethnicity may be a stratifying factor, thus they included it as moderator in their metaregression, finding a negative result. Despite this, the limited statistical power of meta-regression approach (Baker et al., 2009) and the previously reported considerations suggest the need of further analysis to clarify this issue. Therefore the aim of the present paper was to perform a metaanalysis on the association between 5-HTTLPR and antidepressant response, firstly considering available studies until November 2010 both on Caucasian and Asian samples to perform an exploratory analysis without any a priori hypothesis. We then tried to dissect the obtained results by separated analysis in Caucasian and Asian populations, in order to reach a relatively definitive conclusion on the role of 5-HTTLPR on the antidepressant response in both the populations in exam. Despite the recent evidence of possible influence exerted by other polymorphisms within SLC6A4promoter, we focus this meta-analysis on the biallelic insertion/deletion polymorphism; indeed, it is the most known and studied so far, while data about the role of the other polymorphisms within the region are still poor and contradictory (Perroud et al., 2010) (see also Discussion). Further, possible modulators of the association were investigated.

Characteristics of the included studies.

Study

AD used

Study size (male/ female)

Diagnostic Inclusion criteria criteria

Smeraldi et al. (1998) Zanardi et al. (2000) Zanardi et al. (2001) Pollock et al. (2000)

Fluvoxamine

N = 53 (16/37)

DSM IV

Paroxetine

N = 58 (15/43)

Fluvoxamine

Remission criteria

Evaluation: study week

BP + MDD

HDRS ≤ 7

DSM IV

BP + MDD

HDRS ≤ 7

N = 88 (25/63)

DSM IV

BP + MDD

HDRS ≤ 7

paroxetine or nortriptyline

N = 57 (unknown)

DSM IV

MDD

Joyce et al. (2003)

Fluoxetine or nortriptyline

N = 169 (unknown) DSM III R

BP + MDD

Arias et al. (2003) Durham et al. (2004) Serretti et al. (2004)

Citalopram

N = 131 (31/100)

DSM IV

MDD

Sertraline

N = 106 (47/59)

DSM IV

MDD

DSM IV

BP + MDD

BP + MDD

N = 130 (69/61)

DSM IV ICD-10 DSM IV

MDD

N = 64 (unknown)

DSM IV

MDD

N = 163 (52/114)

ICD-10

MDD

Fluvoxamine or N = 221 (75/146) paroxetine +/− lithium Various N = 77 (22/55)

Kirchheiner et al. (2006) Bozina et al. Paroxetine (2008) Dogan et al. Sertraline (2008) Wilkie et al. Various (2009)

Result

Ethnicity

Reference

Remission rate: 6 w l ↑ response

Caucasian

Remission rate: 4 w l ↑ and faster response Remission rate: 6 w l ↑ response

Caucasian Caucasian

Response rate: 2 w

Caucasian

Smeraldi et al. (1998) Zanardi et al. (2000) Zanardi et al. (2001) Pollock et al. (2000)

MADRS reduction 60% or more HDRS 50% HDRS ≤ 7 reduction HDRS 50% reduction HDRS ≤ 7

Response rate: 6 w

l/l faster response to paroxetine l ↑ response

Response rate: 4 w

l ↑ response

HDRS 50% reduction HDRS 50% reduction HDRS 50% reduction HDRS 50% reduction

HDRS ≤ 7

Response rate: 3 w

No association Caucasian

Response rate: 6 w

l/l ↑ response

HDRS ≤ 7

Escitalopram N = 795 (413/382) DSM-IV/ or nortriptyline ICD-10

MDD

HDRS 50% reduction

Various

N = 103 (29/74)

DSM IV

BP + MDD

HDRS 50% reduction

Citalopram

N = 1074 (442/632)

DSM IV

MDD

Citalopram

N = 1655 (unknown)

DSM IV

MDD

QIDS-C 50% reduction

Response rate: l/l ↑ response 2, 4, 6, 8 w Remission rate: 6 w l ↑ response

HDRS ≤ 7 HDRS ≤ 7

Caucasian

Joyce et al. (2003)

Caucasian

Arias et al. (2003) Durham et al. (2004) Serretti et al. (2004)

Caucasian (95.4%) Caucasian

Caucasian

No association Caucasian Response and remission rate: 6 w; 18 w Response and remission rate: weekly up to 12 w Response rate: 4 w

QIDS-C 16 ≤ 5

Remission rate

QIDS-C 16 ≤ 5

Response and remission rates

No association Caucasian

l ↑ response

Caucasian

l ↑ response in Caucasian woman l/l ↑ remission Caucasian (White nonHispanic) No association 78% white, 16% black, 6% other

Kirchheiner et al. (2006) Bozina et al. (2008) Dogan et al. (2008) Wilkie et al. (2009) Huezo-Diaz et al. (2009) Gressier et al. (2009) Mrazek et al. (2009) Hu et al. (2007)

(continued on next page)

241

Huezo-Diaz et al. (2009) (GENDEP) Gressier et al. (2009) Mrazek et al. (2009) (STAR*D) Hu et al. (2007) (STAR*D)

Response criteria

5-HTTLPR and antidepressant efficacy

Table 1

242

Table 1 (continued) Study

AD used

Study size (male/ female)

Diagnostic Inclusion criteria criteria

Response criteria

Kraft et al. (2007) (STAR*D)

Citalopram

N = 1914 (735/1179)

DSM IV

MDD

QIDS-SR 50% QIDS-SR 16 ≤ 5 Response and reduction remission rates

Kraft et al. (2007)

Ng et al. (2006) Rausch et al. (2002) Ruhe et al. (2009)

Sertraline

N = 35 (17/18)

DSM IV

MDD

Fluoxetine

N = 51 (unknown)

DSM IV

MDD

Paroxetine

N = 42 (15/27)

DSM IV

MDD

HDRS 50% reduction HDRS 50% reduction HDRS 50% reduction

Ng et al. (2006) Rausch et al. (2002) Ruhe et al. (2009)

Escitalopram

N = 126 (43/83)

DSM-IV

N = 85 (36/49)

DSM-IV

N = 261 (89/172)

DSM-IV

N = 120 (42/78)

DSM-III-R

N = 54 (22/32)

DSM IV

HDRS 50% MADRS ≤ 11 reduction and HDRS ≤ 7 and CGI ≤ 2 MDD MADRS MADRS ≤ 7 reduction of at least 50% BP + MDD Maier–Philipp Remission rate: 8 w No association core mood severity subscale of the HAMD17 ≤ 4 and no item > 1 MD + BPI, II, HDRS 50% Response rate: 6 w s/s ↑ response dysthymia reduction MDD + BP MADRS 50% Response rate: 6 w s ↑ response reduction

Maron et al. (2009) Illi et al. (2010)

Citalopram, Fluoxetine. or paroxetine Reimherr Sertraline + et al. (2010) atomoxetine or sertraline + placebo

Fluoxetine or paroxetine Fluvoxamine

Evaluation: study week

Result

Ethnicity

No association 78.4% white, 15.6% AfricanAmerican, 1.1% Asian, 4.9% other Response rate: 6 w No association 67% Chinese, 33% Caucasian Response rate: 18 w l allele ↑ Unknown response Response rate: 6 w l/l ↑ SERT 69% Caucasian, occupancy and 17% Creole, ↑ response 14% Asian Response and No association Caucasian remission rate: (96% Estonian) 12 w Response and l/l ↑ remission Caucasian remission rate: 6 w

Reference

Maron et al. (2009) Illi et al. (2010)

Caucasian N = 205

Reimherr et al. (2010)

Asian

Kim et al. (2000) Yoshida et al. (2002)

Asian

S. Porcelli et al.

Kim et al. (2000) Yoshida et al. (2002)

MDD

Remission criteria

MADRS b 10

Response rate: 6 w

No association Asian

HDRS ≤ 7

Response rate: 4 w

l/l ↑ response

Asian

HDRS ≤ 7

Asian

MDD

HDRS 50% reduction

HDRS ≤ 7

l ↑ response Response rate: 2, 4, 6 w Remission rate: 6 w Response rate: 6 w s/s ↑ response

Asian

Kim et al. (2006)

DSM-IV

MDD + BP

HDRS ≤ 7

s/s ↑ response

Asian

DSM-IV

MDD + BP

l/l ↑ response

Asian

DSM-IV

MDD (not specified if BP) MDD (not specified if BP) MDD (not specified if BP)

HDRS 50% reduction HDRS 50% reduction Not specified

s ↑ response

Asian

HDRS 50% reduction

HDRS ≤ 7

Response rate: 4 w, 8 w

No association Asian

Kang et al. (2007) Min et al. (2009) UmeneNakano et al. (2009) Yoshimura et al. (2009)

HDRS 50% reduction

HDRS ≤ 7

Response and remission rate: 4 w

l ↑ response

Milnacipram

N = 80 (28/52)

DSM IV

MDD

Fluoxetine

N = 224 (93/131)

DSM IV

MDD

Paroxetine or fluvoxamine

N = 100 (56/44)

DSM IV

MDD

Kim et al. (2006)

Fluoxetine, sertraline or nortriptyline Mirtazapine

N = 208 (52/156)

DSM IV

Kang et al. (2007) Min et al. SSRIs or SNRIs (2009) Umene-Nakano Sertraline et al. (2009)

N = 101 (29/72) N = 567 (272/295) N = 59 (24/35)

Yoshimura et al. (2009)

Paroxetine

N = 60 (22/38)

DSM-IV

Lee et al. (2010)

Venlafaxine

N = 84 (40/61)

DSM-IV

HDRS ≤ 7

Response rate: 2 w, 4 w Response and remission rate: 6 w Response rate: 6 w

Asian

Yoshida et al. (2004) Hong et al. (2006) Kato et al. (2006)

5-HTTLPR and antidepressant efficacy

MADRS 50% reduction HDRS 50% reduction HDRS 50% reduction

Yoshida et al. (2004) Hong et al. (2006) Kato et al. (2006)

Lee et al. (2010)

BP = Bipolar Disorder. MDD = Major Depressive Disorder. HDRS = Hamilton Depression Rating Scale. MADRS = Montgomery–Asberg Depression Rating Scale. QIDS-C = 16-item Quick Inventory of Depressive Symptomatology clinician rating. QIDS-SR = 16-item Quick Inventory of Depressive Symptomatology Self-Report version. ↑ = Higher.

243

244

2. Experimental procedures 2.1. Search strategy An electronic search of the literature was performed to identify association studies investigating the link between 5-HTTLPR and antidepressant response. PubMed, PsycINFO and ISI Web of Knowledge databases were searched for articles published until November 2010 using any combination of the terms “SERT”, “5-HTT”, “SLC6A4”, “serotonin transporter gene”, “HTTLPR” with both “antidepressant”, “SSRIs”, “SSRI”, “SNRIs”, “SNRI”, “NRI”, “Tricyclic”, “response” and “remission”. References from retrieved papers were also considered.

2.2. Study selection and data extraction Two reviewers (C.F. and S.P.) independently screened searches to identify potentially relevant studies. The full text of selected studies was obtained and evaluated to detect pertinent studies. Association studies were included if: 1) these analyzed the relationship between 5-HTTLPR and antidepressant response; 2) the ethnicity of the sample was specified (only for the second-step analysis); 3) these were written in English. For each study, the following information was extracted: first author, publication year, location/ ethnicity, sample size, 5-HTTLPR genotypic frequencies, female percentage, mean age, clinical and socio-demographic variables, inventories and results concerning association of 5-HTTLPR and antidepressant response. However, not all data were available for all studies. When outcome data were not available in the article, we requested data from the author or we extracted data from the presented graphs by g3data program (https://github.com/pn2200/ g3data). We used response and remission rates as clinical outcomes. The initial search yielded 82 articles that have been revised in order to investigate if they fulfilled the inclusion criteria. In the first time, several of them were excluded because they were review (n = 21), meta-analysis (n = 5), they used other drugs besides antidepressants for augmentation (n = 3), assessed a sample overlapping with previous studies (n = 4), were not in English (n = 1), did not analyze the outcome under examination (n = 10) or data were not available (n = 2). In the case of studies with overlapping samples, only the study with the larger sample was included. Three studies were excluded because they used only CGI for the assessment of response. Indeed, the exclusive use of CGI to assess response seems to be a potential source of bias, since it is based on an overall opinion influenced by individual clinical experience. Despite this, in order to exclude that these studies may have a decisive influence on results, we tested if their inclusion changed results. Thus, 33 studies were included in the first-step analysis which aim was to detect any association between antidepressant response and 5-HTTLPR without any a priori hypothesis. Taking into account the obtained results, in the second-step analysis our aim was to value the role of ethnicity as source of heterogeneity across studies; thus we excluded the studies which ethnicity of the sample was not specified (n = 1), without homogeneous ethnic origin of the sample or analysis not performed for each ethnic group separately (n = 3). Therefore, 30 studies met the inclusion criteria for the second-step analysis (Fig. 1). Subjects with pindolol augmentation were excluded given the observed influence of pindolol on the 5-HTTLPR effect (Smeraldi et al., 1998).

2.3. Statistical analysis Data were entered and analyzed with the Cochrane Collaboration Review Manager Software (RevMan version 5). Individual and pooled 95% confidence intervals (CIs) were calculated. Heterogeneity between studies was assessed with χ2-test of fit and I2 measure. I2 values of 25%, 50% and 75% are generally considered low, medium and high heterogeneity, respectively (Higgins et al., 2003). The significance of the pooled effect size was determined using a Z test.

S. Porcelli et al. Given the lack of unequivocal data for 5-HTTLPR genotype pooling, we tested both dominant and recessive hypothesis: l/l versus s carriers, l carriers versus s/s and l/l versus s/s. Outcome was defined with two phenotypes: remission rate and response rate. Remission was defined as a final Hamilton Rating Scale for Depression (HAMD) total score of 7 or less. Response was defined as at least 50% decrease in Hamilton Depression Scale (HAM-D) or Montgomery and Asberg Depression Rating Scale (MADRS) total score. Data were first analyzed pooling together Caucasians and Asians. Then given the different distribution of 5-HTTLPR genotype frequencies among ethnic groups, we analyzed data pooling separately Caucasians and Asians. In both these steps we firstly analyzed all antidepressant classes together, then separately SSRIs and other antidepressants, in order to weight the effect size of different antidepressant classes. Sensitivity analysis was performed excluding each study one at a time (in order to exclude the result dependence by single study) and adding recent published works to the ones included in the meta-analysis by Taylor et al. (2010). In spite of we performed separate analyses in Caucasians and Asians, we could not exclude possible other sources of heterogeneity among the included studies. Indeed, stratification factors can derive from the inclusion of different sub-ethnic groups (e.g. Japanese, Korean and Chinese within Asians), different diagnosis (bipolar and unipolar depression) and different clinical-demographic features (e.g. age of onset, features of the episode, number of previous episodes). For these reasons, data were analyzed within a random effects framework both for Asians and Caucasians. A random effects framework assumes that between-study variation is due to both chance or random variation and an individual study effect. This model is more conservative than fixed effects models and generates a wider confidence interval. Furthermore, a funnel plot was created to assess potential ascertainment bias by plotting natural logarithm of individual study effect size against the standard error of the natural logarithm of individual study effect size. Ascertainment bias was also assessed using the Egger test (Egger et al., 1997) and the trim and fill test. We finally investigated any other source of heterogeneity in results across studies by using a multi-level model approach to metaanalysis, also known as meta-regression. This statistical approach allows an assessment of the independent impact of various study characteristics that might explain differences in the effect sizes obtained. We used a random-effect model to predict variation in effect size among studies with the following factors: age (average age of the sample); gender (percent of female participants in each study); mean age at onset, number of previous episodes (the mean number of depressive episodes reported); mean duration of current episode, severity of the current episode (mean HAMD score at baseline), bipolar status (percent of bipolar patients), psychotic status (percent of psychotic patients), education level (the mean number of education years), marital status (percent of married participants), employed/non employed status (percent of employed participants), and ethnic origin (percent of Caucasian subjects). Moreover, we used as moderators also the sample size of each study and the year of publication, in order to exclude publication bias. We considered the presence of extreme values in data distribution (> or bmean value ± 2SD), and we excluded them from the model. For data concerning Caucasian subjects extreme values were assessed both for the ratio responders/non-responders and remitters/non-remitters within the performed comparisons (l/l genotype versus s carriers and l carriers versus s/s genotype) and for moderator values. For data concerning Asians extreme values were assessed only for moderator values, since the low frequency of l allele in this population did not make the other evaluation appropriate. Anyway, a sensitive analysis also including outlier values was performed. We used the statistics environment R (http://www.Rproject.org) version 2.12.1, package “metafor”, to perform the meta-regression. Data pertaining to studies in Asians and Caucasians were always analyzed separately, ethnicity analysis apart.

5-HTTLPR and antidepressant efficacy

245

A Citations from literature N=1577 Citations excluded N=1496

Total potentially eligible studies N=82

Studies included in the first step N=33

Data from another study on the STAR*D sample (Kraft et al., 2007) were tested when results split for ethnic group were available

N=49 were excluded because: -were review (n= 21) -were meta-analysis (n=5) -there was an overlapping sample (n=4) -used other drugs besides antidepressants for augmentation (n=3) -used only CGI to assess response (n=3) -were not in English (n=1) or not pertinent to the analyzed issue (n=10) -data were not available (n=2)

N=4 were excluded because: -the ethinicity was not specified (n=1) -the ethnic origin was not homogeneous and results for each ethnic group were not available (n=3)

Studies included in the second step N=30

B

Reference (Kraft et al., 2005) (Smeraldi et al., 2006) (Mandelli et al., 2009a) (Yu et al., 2002) (Kato et al., 2005) (Bocchio-Chiavetto et al., 2008) (Stamm et al., 2008) (Bukh et al., 2009) (Popp et al., 2006)

(Lee et al., 2004) (Higuchi, 2010) (Murphy et al., 2004) (Baffa et al., 2010) Reference (Kraft et al., 2005) (Ruhe et al., 2009) (Ng et al., 2006) (Rausch et al., 2002) (Hu et al., 2007) (Kraft et al., 2007)

Figure 1

Reason for exclusion (First analysis) Only CGI as outcome measure Sample overlapped with the one of another study (Smeraldi et al., 1998) Sample overlapped with the one of another study (Smeraldi et al., 1998) Sample overlapped with the one of another study (Hong et al., 2006) Sample overlapped with the one of another study (Kato et al., 2006) Augmentation with other treatments (rTMS, antipsychotics) Augmentation with other treatments (antipsychotics, lithium) Augmentation with other treatments (antipsychotics, anticonvulsants, ECT) Only CGI as outcome measure and the evaluation of the association between SERT polymorphisms and side effects as primary outcome Only CGI as outcome measure Article in Japanese Data not available Data not available Reason for exclusion (Second analysis) Mixed ethnicity Mixed ethnicity Mixed ethnicity Ethnicity not specified Results not available for single ethnic group Mixed ethnicity (included in the second-step analysis only when data splited for ethnic group were available)

A. Study flow. B. Studies excluded from the first or second step of the analysis.

3. Results An overview of the selected studies on the association between 5HTTLPR and antidepressant response/remission is provided in Table 1. With regard to studies on the STAR*D sample, we included in the first step of our analysis both the study by Hu et al. (2007)

and Mrazek et al. (2009) one at a time, in order to test the impact on the effect size. In the second step, we included the study by Mrazek et al. (2009), since, unlike Hu et al. (2007), it reported results split for ethnic group. On the other side, Kraft et al. reported only partially split results (MAF in the responder and non responder groups) (Kraft et al., 2007); we tested its impact

246 on results including available data in a sensitivity analysis, but the result was not affected (data not shown). For significant results, we tested the influence of single studies by excluding them from the analysis one at a time. Studies were performed in a wide range of geographic locations; however, patients' diagnoses were made according to standard internationally accepted criteria in all cases (in most cases DSM-IV was used).

3.1. Exploratory analysis: Caucasian and Asian subjects pooled together A total of 33 studies (5479 subjects in all; 28 studies and 3866 subjects included in the “SSRIs sub-analysis”; 11 studies and 1613 subjects included in the “other antidepressants” subanalysis) were included. We found an effect of 5-HTTLPR only on remission rate. 5-HTTLPR l allele was associated with remission when we pooled the l/l genotype versus the s/s one (OR = 1.37, C.I. 95% 1.09–1.72, p = 0.007 for all antidepressant classes, OR = 1.48, C.I. 95% 1.12–1.96, p = 0.005 for SSRIs only and including Mrazek et al. (2009)) as well as when we considered together l/l and l/s genotypes versus the s/s genotype (OR = 1.28, C.I. 95% 1.02–1.61, p = 0.04 for all antidepressant classes and OR = 1.59, C.I. 95% 1.18–2.12, p = 0.002 for SSRIs only and including Mrazek et al. (2009)) (Supplementary Fig. 1). The effect size was higher when we considered SSRIs only, but it was still significant when all antidepressant classes were included. Moreover, the association was found both including data from the study by Mrazek et al. (2009) and Hu et al. (2007), suggesting the strength of result regardless of which STAR*D study was included. On the other side, no effect was found on response, also including the study by Kraft et al. (2005), which we have excluded for the use of CGI only for the assessment of response (data not shown). Heterogeneity among studies was mediumhigh for most comparisons. Then, we tried to dissect the source of the observed associations by performing a meta-analysis and meta-regression split for ethnic group.

3.2. Analysis in Caucasian subjects A total of 19 studies performed in Caucasians were included (3675 subjects in all; 16 studies and 2785 subjects included in the “SSRIs sub-analysis”; 6 studies and 890 subjects included in the “other antidepressants” sub-analysis). The pooled analyses and OR for response in the l/l and l/s genotypes versus the s/s genotype are presented in Fig. 2. A significant association was found between l allele and response rate for SSRIs (OR = 1.58, C.I. 95% 1.16–2.16, p = 0.004), and it survived removing single studies one at a time, the only exception being the study by Huezo-Diaz et al. (2009) (in this case p = 0.05). The result survived also adding data from the study by Popp et al. (2006) (p = 0.007), which we have excluded for the use of CGI only for the assessment of response. Considering all antidepressant classes taken together and only non-SSRI antidepressants no association was detected. On the other hand, pooling the l/l genotype versus the s/s one we found evidence of association both considering all antidepressant classes (OR = 1.62, C.I. 95% 1.22–2.16, p = 0.0008) and SSRIs only (OR = 1.71, C.I. 95% 1.20–2.45, p = 0.003), but not for other/mixed antidepressants (Fig. 3). These two associations were heavily influenced by one study (Huezo-Diaz et al., 2009), which exclusion made the effect size

S. Porcelli et al. on the edge of significance (p = 0.04 and p = 0.02, respectively for all antidepressants and SSRIs only). Finally, we pooled the l/s and s/s genotypes versus the l/l genotype (Supplementary Fig. 2) and no association with response was found, also including the study by Kraft et al. (2007) (data not shown). There was no evidence of heterogeneity across studies. With regard to remission, the pooled analyses and OR for the l/l and l/s genotypes versus the s/s genotype are presented in Fig. 4. As well as for response, the l carriers were the group with the higher probability of remission during SSRI treatment (OR= 1.53, C.I. 95% 1.14–2.04, p = 0.004); the association survived also removing all single studies one at a time. A minimal evidence of association persisted when we considered all antidepressant classes (OR = 1.41, C.I. 95% 1.02–1.95, p = 0.04). Nevertheless, this last result appeared very weak, since it disappeared by removing several single studies (Arias et al., 2003; Dogan et al., 2008; Huezo-Diaz et al., 2009; Serretti et al., 2004; Smeraldi et al., 1998; Zanardi et al., 2000) from the analysis. Pooling the l/l genotype versus the s/s one none association was found (Supplementary Fig. 3), as well as when we pooled the l/s and s/s genotypes versus the l/l one (Supplementary Fig. 4). There was no evidence of heterogeneity across studies on the base of Chi2 test (assuming a cut-off of 0.1), while I2 showed that 29% and 36% of variability for SSRIs and all antidepressants respectively were due to heterogeneity rather than sampling error. The meta-regression results are shown in Table 2. We found an effect of gender, mean age and age at onset on 5-HTTLPR association with antidepressant efficacy (for scatter plots see Supplementary Fig. 5). Particularly, the l/l genotype compared to s carriers seems to have better response rates when present age (p = 0.005) and age at onset (p = 0.007) increase. The effect of gender instead was more difficult to understand: indeed, when we compared l allele carriers to the s/s genotype, males seemed to have better response probability (p = 0.008), while remission rates resulted higher in females (p = 0.01). The sample size showed a minor influence on remission pooling l carriers versus ss (p = 0.0239), suggesting that remission rates in l carriers have the tendency to be higher in smaller studies. The year of publication had an impact on remission considering both all antidepressant classes (p = 0.0005) and only SSRIs (p = 0.0014). Given the risk of publication bias, we performed on these data also the trim and fill test and the Egger test. For all antidepressant classes, both these tests did not confirm the publication bias (p= 0.27 and p = 0.27, respectively; see Supplementary Fig. 6 for funnel plot) For SSRIs only, the trim and fill test did not suggest any bias (p = 0.44, see Supplementary Fig. 6 for funnel plot), while the Egger test did not allow to exclude it (p = 0.03). The moderator factors which had an effect on outcome did not substantially change also including the outlier values in the analysis (data not shown).

3.3. Analysis in Asian subjects A total of 11 studies performed in Asians were included (1429 subjects in all; 7 studies and 716 subjects included in the “SSRIs sub-analysis”; 5 studies and 713 subjects included in the “other antidepressants” sub-analysis). The only evidence of association between 5-HTTLPR and antidepressant efficacy was found pooling the l/l genotype versus l/s and s/s genotypes. Indeed, the l/l genotype showed higher remission probability (OR = 2.10, C.I. 95% 1.15–3.84, p = 0.02). This result seems to be very weak, since only four studies were includible in the analysis. Moreover,

5-HTTLPR and antidepressant efficacy

247

Figure 2 Response in Caucasians pooling the l/l and l/s genotypes versus the s/s genotype. A. SSRIs. B. Other/mixed antidepressant classes. C. All antidepressant classes.

removing the studies by Hong et al. (2006) and Min et al. (2009) one at a time the result disappeared. In Fig. 5 the results pertaining remission are shown. Regarding response, results for all pooling are included in supplementary materials (Supplementary Figs. 7, 8 and 9). The inclusion of data by Lee et al. (2004), which we have excluded for the use of CGI only for the assessment of response, did not change the result (data not shown). Heterogeneity across studies is greater than what was observed in Caucasians.

In Asians the meta-regression (Table 2) revealed an influence of age, gender and HAM-D score at baseline (for scatter plots see Supplementary Fig. 10). Particularly, younger age (p = 0.003), male gender (p = 0.005) and higher HAM-D score at baseline (p = 0.01) seem to predict higher probability of response in the shown comparisons. The moderator factors with an effect on outcome did not substantially change also including the outlier values in the analysis (data not shown). The year of publication did not show any influence on outcome. As well as for Caucasians, the

248

S. Porcelli et al.

Figure 3 Response in Caucasians pooling the l/l genotype versus the s/s one. A. SSRIs. B. Other/mixed antidepressant classes. C. All antidepressant classes.

sample size showed an effect on outcome, particularly the l/l genotype compared to s carriers showed higher response probability when the sample size increases (p = 0.0004).

3.4. Sensitivity analysis In order to clarify our different results with respect to the metaanalysis by Taylor et al., we tested the association between

remission/response during SSRI treatment and l allele by including all studies included by Taylor et al. (2010) and adding recently published studies. We chose these comparisons because our results suggested significant genotype effect, particularly in Caucasians. Just for this reason, the sensitivity analysis was completed by a meta-regression on the same studies and using the percent of Caucasian subjects as moderating variable. This analysis found an effect of 5-HTTLPR on remission (OR = 1.47, C.I. 1.10–1.95, p = 0.008) (Supplementary Fig. 11).

5-HTTLPR and antidepressant efficacy

249

Figure 4 Remission in Caucasians pooling the l/l and l/s genotypes versus the s/s genotype. A. SSRIs. B. Other/mixed antidepressant classes. C. All antidepressant classes.

The meta-regression suggested that ethnicity is likely a moderating factor (Table 2). Particularly, subjects carrying the l allele showed higher remission probability if the percent of Caucasians was higher. We then performed the same procedure to test the association between SSRI response and l allele. We included the same studies included by Taylor et al. (data for two studies (Joyce et al., 2003; Kim et al., 2006) are different since for unknown

reason Taylor et al. reported data pertained to total sample and not SSRI only). No evidence of association was found (Supplementary Fig. 11) and the heterogeneity across studies was high (Chi 2 test: p b 0.0001 and I 2 = 63%). The meta-regression was performed on the same studies except for Rausch et al. (2002) which could not be included because of lack of data about ethnic origin of the sample. Results were not significant (Table 2).

250

S. Porcelli et al.

Table 2

Results of meta-regression.

Moderator ethnicity HAMBSL Asian Caucasian Asian Caucasian Caucasian N previous episodes Asian Caucasian

Outcome

Pooling

Estimate

Response

l/l vs s car

Response

l car vs ss

Remission

l/l vs s car l car vs ss

0.37 −0.02 0.22 0.01 0.04 −0.14

Response

l car vs ss

Z

P

0.14 0.07 0.13 0.10 0.09 0.11

2.57 − 0.26 1.66 0.14 0.43 − 1.25

0.01 0.80 0.10 0.89 0.67 0.21

l/l vs s car

−0.11 −0.05 −0.22

0.60 0.44 0.47

− 0.18 − 0.11 − 0.47

0.86 0.91 0.64

l/l vs s car l car vs ss l/l vs s car l car vs ss

−0.14 4.46 −0.29 −0.36

3.96 4.92 0.39 0.44

− 0.04 0.91 − 0.76 − 0.83

0.97 0.36 0.45 0.41

Response

l/l vs s car

Response

l car vs ss

Remission

l car vs ss

Remission

l/l vs s car

−8.01 −0.28 −7.96 −4.92 8.71 5.27 4.55 −2.17

4.91 2.24 2.85 1.87 5.25 2.05 10.75 1.94

− 1.63 − 0.13 − 2.79 − 2.63 1.66 2.57 0.42 − 1.12

0.10 0.90 0.005 0.008 0.10 0.01 0.67 0.26

Remission

l/l vs s car

Remission

l car vs ss

Response

l car vs ss

Response

l/l vs s car

0.02 0.03 0.10 0.03 −0.10 −0.02 −0.15 0.03

0.08 0.02 0.05 0.03 0.04 0.04 0.05 0.01

0.24 1.24 2.09 0.89 − 2.63 − 0.45 − 3.00 2.78

0.81 0.21 0.04 0.37 0.008 0.66 0.003 0.005

Remission

l/l vs s car l car vs ss l car vs ss l/l vs s car

0.11 0.08 0.05 0.06

0.08 0.11 0.03 0.02

1.47 0.72 1.49 2.70

0.14 0.47 0.14 0.007

l/l vs s car l car vs ss l car vs ss l/l vs s car

−0.01 −0.01 −0.01 −0.01

0.01 0.02 0.02 0.01

− 1.03 − 0.44 − 0.67 − 0.81

0.30 0.66 0.50 0.42

l/l vs s car l car vs ss l/l vs s car l car vs ss

0.66 1.29 −3.53 4.41

0.86 0.75 6.05 5.69

0.77 1.70 − 0.58 0.78

0.44 0.09 0.56 0.44

% First episode Caucasian

Response

Caucasian

Remission

% Female Asian Caucasian Asian Caucasian Asian Caucasian Asian Caucasian Age Asian Caucasian Asian Caucasian Asian Caucasian Asian Caucasian Onset Caucasian

Response Duration of current episode Caucasian

Remission Response

% Bipolar Caucasian

Remission Response

SE

Education Caucasian

Remission

l/l vs s car l car vs ss

0.01 −0.05

0.04 0.04

0.22 − 1.36

0.82 0.17

Marital status Caucasian

Remission

l/l vs s car l car vs ss

−6.24 2.92

4.35 5.18

− 1.43 0.56

0.15 0.57

Employed/student Caucasian

Remission

l/l vs s car l car vs ss

2.67 −2.25

3.89 3.48

0.69 − 0.65

0.49 0.52

5-HTTLPR and antidepressant efficacy

251

Table 2 (continued) Moderator ethnicity

Outcome

Pooling

% Psychotic Caucasian

Remission

l/l vs s car l car vs ss

Year of publication Caucasian (all ADs)

Remission Response

Caucasian (SSRIs)

Response Remission

Asian (all ADs)

Remission Response

Asians (SSRIs) Number of subjects Caucasians (all ADs)

Response

Remission Response

Caucasian (SSRIs)

Response Remission

Asians (all ADs)

Remission Response

Asians (SSRIs)

Response

Estimate

SE

Z

P

1.03 0.91

0.98 0.74

1.05 1.22

0.29 0.22

l car vs ss ll vs s car ll vs s car l car vs ss l car vs ss ll vs s car ll vs s car l car vs ss l car vs ss ll vs s car ll vs s car l car vs ss l car vs ss ll vs s car

−0.12 −0.03 −0.07 −0.01 0.03 −0.03 −0.04 −0.11 −0.11 −0.12 0.19 0.15 0.10 0.10

0.03 0.03 0.04 0.07 0.06 0.08 0.04 0.03 0.12 0.14 0.17 0.10 0.14 0.29

−3.48 −0.94 −1.57 −0.21 0.49 −0.44 −1.03 −3.19 −0.94 −0.88 1.11 1.49 0.76 0.36

0.0005 0.35 0.12 0.83 0.62 0.66 0.30 0.001 0.35 0.38 0.27 0.14 0.45 0.72

ll vs s car l car vs ss ll vs s car l car vs ss l car vs ss ll vs s car ll vs s car l car vs ss l car vs ss ll vs s car ll vs s car l car vs ss l car vs ss ll vs s car

0.0002 −0.0004 −0.001 0.001 0.0005 −0.0004 −0.0001 −0.001 −0.002 −0.0002 0.004 0.001 0.006 0.02

0.0002 0.0003 0.001 0.001 0.001 0.001 0.0004 0.0003 0.001 0.001 0.003 0.002 0.01 0.01

1.11 −1.35 −1.30 0.70 0.57 −0.33 −0.15 −2.26 −2.08 −0.15 1.31 0.30 0.91 3.54

0.26 0.18 0.19 0.48 0.57 0.74 0.88 0.02 0.04 0.88 0.19 0.76 0.36 0.0004

l car vs ss ll vs ss ll vs s car l car vs ss ll vs ss ll vs s car

0.42 0.49 0.27 −0.05 0.18 0.18

0.16 0.19 0.15 0.19 0.27 0.19

2.63 2.58 1.74 −0.26 0.65 0.97

0.0084 0.0099 0.08 0.80 0.52 0.33

Percent of Caucasians Remission

Response

4. Discussion The aim of this meta-analysis was to elucidate the influence of 5-HTTLPR on antidepressant efficacy, since the studies published so far reached contradictory findings. We retrieved 33 studies to perform an exploratory analysis and test the impact of 5-HTTLPR on antidepressant efficacy without any a priori hypothesis. We found evidence of association between the l/l genotype and l allele and remission (Supplementary Fig. 1). Thus, we tried to dissect the association in order to identify the presence of stratification factors. In this second-step, 30 studies (5104 subjects), were included and analyzed separately for Caucasians and Asians. The results suggest an influence of 5-HTTLPR on SSRI efficacy in Caucasians. In particular, a higher probability of

response was detected in l carriers treated with SSRIs (p = 0.004), while no effect was detected neither for mixed antidepressant nor for non-SSRI antidepressants. Similar associations were found when we focused on remission: l carriers showed higher probability of remission during SSRI treatment (p = 0.004), while a very weak effect was found for mixed antidepressant (p = 0.04) and no effect was found for non-SSRI antidepressants. Our results are consistent with previous evidence of a selective influence exerted by 5-HTTLPR on SSRIs response in Caucasians (Serretti et al., 2007b). On the other hand, the negative results found by another recent meta-analysis (Taylor et al., 2010) may be an effect of pooling together studies performed in different populations (see Introduction). Nevertheless, they are probably due almost partially to the inclusion of more recent

252

S. Porcelli et al.

Figure 5 Remission in Asians considering mixed antidepressant classes. A. l/l genotype versus l/s and s/s genotypes. B. l/l and l/s genotypes versus s/s genotype.

data in our work, as the sensitivity analysis performed starting from the studies included by Taylor et al. showed (see Section 3.4). We provided a comparison of the present meta-analysis toward the previous ones on the topic in Table 3. Despite our results, some studies performed on Caucasian samples failed to find any effect of 5-HTTLPR on antidepressant efficacy (Dogan et al., 2008; Kirchheiner et al., 2006; Maron et al., 2009; Reimherr et al., 2010; Wilkie et al., 2009). For these studies, we need to underline some issues that could explain these contradictory results. Particularly, two studies were performed on patients selected for resistant depression (Reimherr et al., 2010; Wilkie et al., 2009). For these two studies, we could hypothesize the intervention of further stratification factors with respect to non-resistant depression. Further, both studies did not analyze separately the effect of 5-HTTLPR on patients treated with SSRIs, the antidepressant class most associated with the 5-HTTLPR effect, as stated above. Also Kirchheiner and colleagues studied a sample treated with mixed antidepressants (Kirchheiner et al., 2006), but when a sub-analysis on the SSRI sample was performed, a trend of association between the l allele and response was found. Nevertheless, this study was limited by the brief follow-up (3 weeks), the low number of clinical evaluations (only at the baseline and at the endpoint) and the selection of patients with severe – and probably resistant to first-line treatment – depression (only inpatients). Higher rates of side-effects in some groups of patients might represent an additional confounding factor. Indeed, Maron et al. failed to find any association between 5-HTTLPR and clinical response but they

showed that s carriers may have higher risk of side effects (Maron et al., 2009). The finding was confirmed by other authors (Kato and Serretti, 2010; Murphy et al., 2004; Perlis et al., 2003; Popp et al., 2006; Reimherr et al., 2010; Hu et al., 2007), even if negative findings exist as well (Huezo-Diaz et al., 2009; Wilkie et al., 2009). We could hypothesize that the high frequency of adverse events in s carriers might reduce the compliance to treatment, affecting at least partially the association between 5-HTTLPR and clinical outcomes. To avoid this confounding factor, a rigorous monitoring of patients' compliance is required, ideally through the antidepressant plasma level regular check. Also clinical-demographic variables might have an impact on antidepressant efficacy (Serretti et al., 2007a, 2009) and therefore they could constitute potential stratification factors per se or in interaction with 5-HTTLPR. Our metaregression (Table 2) in Caucasians showed that age, age at onset and gender may modulate some of the analyzed comparisons. Particularly, the l/l genotype compared to s carriers showed better response as age increases (p = 0.005). In regard to this, contradictory findings can be found in literature. Indeed, the l allele was associated to better response in subjects over 25 years (Joyce et al., 2003), but other authors did not confirm this result (Mandelli et al., 2007; Grigoriadis and Robinson, 2007). Good agreement exists instead concerning the effect of age at onset. Indeed, our analysis revealed that the l/l genotype compared to s carriers has increasing probability of response as age at onset increases (p = 0.0069), consistently with literature (Bagby et al., 2002; Serretti et al., 2007a; Souery et al., 2007). In regard to gender effect, the results are more difficult to

5-HTTLPR and antidepressant efficacy be interpreted. Indeed, l carriers compared to the s/s genotype showed higher response probability as the proportion of males increases, while an opposite effect was detected for remission. This may reveal a tendency of females to reach higher and more stable symptom decrease. Also in literature findings about the effect of gender are inconsistent. Indeed, in the GENDEP sample a better escitalopram response was found in l carriers only in males (Huezo-Diaz et al., 2009), consistently with the finding by Taylor et al. (2010). Nevertheless, other studies reached opposite results (Gressier et al., 2009; Grigoriadis and Robinson, 2007). In Asians, environmental factors and other genetic variants rather than 5-HTTLPR probably affect antidepressant response more strongly. Indeed, only when we pooled the l/l genotype versus l/s and s/s genotypes in four studies using different antidepressant classes we found an association between l/l genotype and remission (p = 0.02). In our previous meta-analysis (Serretti et al., 2007b), an association between the l allele and l/l genotype and response was found, but not with remission since only two studies were includible at that time. Anyway, our present result may be biased by the low number of studies includible and the very low number of l/l carriers in two of these studies. The last issue made even more questionable the analysis of data pooling together Asian and Caucasian samples. While in Asians the effect of 5-HTTLPR seems to be low (but with still uncertain direction) or absent, on the other hand clinical-demographic variables might have a significant influence. Indeed, age and gender were found to impact on response rates, since l carriers compared to the s/s genotype showed better response when the percent of males increases (p = 0.005) and age decreases (p = 0.008). Depression severity at baseline also resulted a modulator in Asians (p = 0.01): greater baseline HAM-D scores predict higher response probability for the l/l genotype compared to s carries (Table 2). The effect of these variables on antidepressant efficacy in Asians has been scarcely studied as far as now and no conclusion can be drawn. Only a previous study failed in finding any association between antidepressant response and baseline depression severity (Higuchi et al., 2008). Recent findings suggested that also other clinical stratification factors may influence 5-HTTLPR association with antidepressant response. For example, the s allele was found to predict poor response only in anxious depression (Baffa et al., 2010), while the l/l genotype and l allele were associated to melancholic depression (Willeit et al., 2003), especially in females (Baune et al., 2008). On the other side the s allele was associated to atypical depression (Willeit et al., 2003) and to seasonality (Rosenthal et al., 1998). Since 5HTTLPR was associated also to personality traits (Schinka et al., 2004; Serretti et al., 2006), such as neuroticism and anxiety scores, and they might impact on antidepressant efficacy, they should be considered as possible stratifying factors in association studies. Consistently, a previous report showed that harm avoidance may moderate the influence of SERT gene variants on treatment outcome in bipolar patients (Mandelli et al., 2009b), suggesting the usefulness to investigate depressive endophenotypes. Other than the reported issues, we must underline that new findings in regard to serotonin transporter polymorphisms are currently emerging. Indeed, the l variant is itself heterogeneous, encompassing multiple repeat variants

253 (Nakamura et al., 2000), each of them might have an impact on antidepressant efficacy. Among these variants only rs25531 (G/A) received special attention, even if its influence on gene expression and antidepressant response remains unclear (Zobel and Maier, 2010). We searched for available data to perform a meta-analysis on the association between this polymorphism and antidepressant response. Since the presumed functional effect of rs25531 on SERT transcription activity, a comparison based on the number of LA alleles seems an adequate choice. Unfortunately, few studies are available: only three studies could be included for response (Hu et al., 2007; Maron et al., 2009; Ruhe et al., 2009) and two studies for remission (Bukh et al., 2009; Hu et al., 2007). The obtained results did not reach the threshold for significance (data not shown). Further, recently rs25531 was found to lie 18 bp 5′ to the site of the l/s repeat polymorphism and not within it, so these polymorphisms should be considered as four alleles instead of a unique triallelic locus (Perroud et al., 2010). A better understanding of polymorphism localizations within 5HTTLPR and their functional effects would definitively clarify the role of serotonin transporter promoter in antidepressant response. Despite these unsolved issues, on the basis of the actual knowledge, our meta-analysis seems to confirm the effect of the biallelic serotonin transporter polymorphism on SSRI response and remission in Caucasians, with evidence of better outcome in l carriers versus s/s genotype. This effect seems to be dependent from ethnic origin, since 5-HTTLPR showed only low influence on remission in Asians during treatment with mixed antidepressant classes. Nevertheless, we cannot exclude that the emerging of new findings may reverse the results of this meta-analysis and eventually also exclude any effect of 5-HTTLPR on antidepressant response. The results of meta-regression about the effect of ethnicity are of not clear reading, but they did not exclude an effect of ethnic origin. Unfortunately, meta-regression is not a powerful statistical approach (Baker et al., 2009) and publication bias cannot be excluded. Since both our meta-regression and the Egger test suggested that publication bias may affect the results concerning remission in Caucasians, we could not exclude this bias when considering the whole sample, even more taking into account the findings by Taylor et al. (2010). Thus, we performed the trim and fill test also on the whole sample, considering the remission rates for the pooling l carriers versus s/s, but the result did not show evidence of publication bias (estimate = 0.07, se = 0.13, z = 0.55, p = 0.58). If our results will be confirmed, the clinical impact of 5HTTLPR genotyping in Caucasians is expected to be modest but useful in the perspective of inclusion in future multiallelic tests. Indeed, the cost-effectiveness of a singlelocus genotyping test was estimated to be quite-low (Perlis et al., 2009). E.g. concerning 5-HTTLPR, it might explain only the 3.2% of variance in antidepressant response but the cost-effectiveness of genotyping is increased after 2 depressive episodes (Serretti et al., 2011). Since the recurrent nature of major depression and the potential impact of genotyping on depression prognosis, the contribution of 5HTTLPR in determining treatment outcome still remains a “hot” topic for future investigations. If confirmations will be found it could represent one of the first examples of

254

Table 3

Comparison among recent meta-analysis on the topic.

Meta-analysis

Inclusion criteria

Present

–Analyze the (See Fig. 1) relationship between 5-HTTLPR and antidepressant efficacy in MDD patients –Written in English –Known ethnicity (for the second step of analysis)

33 studies, from 1998 to 2010

Taylor et al. (2010)

–Analyze the relationship between 5-HTTLPR and antidepressant efficacy in MDD patients

28 studies, from 1998 to 2009

Not specified

Included studies Statistical tests and time covered

–Outcome not 15 studies, from evaluated as 1998 to 2006 response or remission rate on a depression scale

Outcomes

Poolings

Significant results

–Mantel–Haenszel test, OR, random effect model –Meta-regression –Funnel plots, Egger test, trim and fill test, and meta-regression to ascertain publication bias

–Response: HDRS 50% reduction (or analogous scale) –Remission: HDRS ≤ 7 (or analogous scale)

–l/l versus l/s and s/s –l/l and l/s versus s/s –l/l versus s/s (all the studies, only Caucasian studies and only Asian studies)

–Mantel–Haenszel test, RR, fixed effect model –DerSimonian–Laird, RR, random effect model –Meta-regression –Funnel plots and trim and fill method to ascertain publication bias –OR, fixed effect model

–Response: HDRS 50% reduction (or analogous scale) –Remission: HDRS ≤ 7 (or analogous scale)

–l/l versus l/s and s/s –l/l and l/s versus s/s –l/l versus s/s

–Whole sample: –l/l and l associated with remission –Caucasian sample: –l/l and l associated with response and l with remission –Asian sample: –l/l associated with remission (but only 4 studies were includible) (for OR, C.I. and p values see the text) Higher remission rates for l carriers versus s/s: RR: 0.88 (95% CI: 0.79–0.98, p = 0.02)

–Response: HDRS 50% reduction (or analogous scale) –Response rate within 4 weeks

–l/l versus l/s and s/s –l/l and l/s versus s/s

–Whole sample: –l/l associated with response (OR = 2.01, C.I. 95% 1.39–2.89, p = 0.0002), response

S. Porcelli et al.

Serretti –Analyze the et al. (2007b) relationship between 5-HTTLPR and SSRI

Excluded studies

–Overlapping patient samples –Mixture of antidepressants (Kraft et al., 2005; Minov et al., 2001) –Use of CGI only (Lee et al., 2004) –Data not available (Murphy et al., 2004)

–Remission: HDRS ≤ 7 (or analogous scale)

within 4 weeks (OR = 2.57, C.I. 95% 1.70–3.88, p b 0.00001) –l allele associated with remission (OR = 2.21, C.I. 95% 1.53–3.21, p b 0.00001) and response within 4 weeks (OR = 1.72, C.I. 95% 1.20–2.47, p = 0.003) –Caucasian sample: –l/l associated with response within 4 weeks (OR = 1.75, C.I. 95% 1.07–2.88, p = 0.03) –l allele associated with remission (OR = 2.37, C.I. 95% 1.56–3.58, p b 0.00001) –Asian sample: –l/l associated with response (OR = 2.52, C.I. 95% 1.37–4.62, p = 0.003) and response within 4 weeks (OR = 5.96, C.I. 95% 2.70–13.17, p b 0.0001) –l allele associated with response within 4 weeks (OR = 1.85, C.I. 95% 1.22–2.79, p = 0.004)

5-HTTLPR and antidepressant efficacy

efficacy in MDD patients

255

256 quite-well theorized association within antidepressant pharmacogenetics. Supplementary materials related to this article can be found online at doi:10.1016/j.euroneuro.2011.10.003.

Role of the funding source This study was institutionally supported.

Contributors Porcelli S. and Fabbri C. performed the electronic search of the literature and carried out a dataset with study information. They equally contributed to the analysis of data and to the drafting of the review. Serretti A. assisted in the preparation of the manuscript and contributed to the drafting of the discussion.

Conflict of interest Dr. Serretti is or has been a consultant/speaker for: Abbott, Astra Zeneca, Clinical Data, Boheringer, Bristol Myers Squibb, Eli Lilly, GlaxoSmithKline, Janssen, Lundbeck, Pfizer, Sanofi, and Servier.

References Angst, J., 1965. The prognosis of antidepressive treatments: longitudinal and genetic studies. Anglo Ger. Med. Rev. 2, 733–751. Arias, B., Catalan, R., Gasto, C., et al., 2003. 5-HTTLPR polymorphism of the serotonin transporter gene predicts non-remission in major depression patients treated with citalopram in a 12weeks follow up study. J. Clin. Psychopharmacol. 23, 563–567. Baffa, A., Hohoff, C., Baune, B.T., et al., 2010. Norepinephrine and serotonin transporter genes: impact on treatment response in depression. Neuropsychobiology 62, 121–131. Bagby, R.M., Ryder, A.G., Cristi, C., 2002. Psychosocial and clinical predictors of response to pharmacotherapy for depression. J. Psychiatry Neurosci. 27, 250–257. Baker, W.L., White, C.M., Cappelleri, J.C., et al., 2009. Understanding heterogeneity in meta-analysis: the role of metaregression. Int. J. Clin. Pract. 63, 1426–1434. Baune, B.T., Hohoff, C., Mortensen, L.S., et al., 2008. Serotonin transporter polymorphism (5-HTTLPR) association with melancholic depression: a female specific effect? Depress Anxiety 25, 920–925. Benedetti, F., Barbini, B., Bernasconi, A., et al., 2008. Lithium overcomes the influence of 5-HTTLPR gene polymorphism on antidepressant response to sleep deprivation. J. Clin. Psychopharmacol. 28, 249–251. Bozina, N., Peles, A.M., et al., 2008. Association study of paroxetine therapeutic response with SERT gene polymorphisms in patients with major depressive disorder. World J. Biol Psychiatry 3, 190–197. Bukh, J.D., Bock, C., Vinberg, M., et al., 2009. Interaction between genetic polymorphisms and stressful life events in first episode depression. J. Affect. Disord. 119, 107–115. Dogan, O., Yuksel, N., Ergun, M.A., et al., 2008. Serotonin transporter gene polymorphisms and sertraline response in major depression patients. Genet. Test. 12, 225–231. Dong, C., Wong, M.L., Licinio, J., 2009. Sequence variations of ABCB1, SLC6A2, SLC6A3, SLC6A4, CREB1, CRHR1 and NTRK2: association with major depression and antidepressant response in Mexican-Americans. Mol. Psychiatry 14, 1105–1118.

S. Porcelli et al. Durham, L.K., Webb, S.M., et al., 2004. The serotonin transporter polymorphism, 5HTTLPR, is associated with a faster response time to sertraline in an elderly population with major depressive disorder. Psychopharmacology (Berl) 174, 525–529. Egger, M., Smith, G.D., Phillips, A.N., 1997. Meta-analysis: principles and procedures. BMJ 315, 1533–1537. Franchini, L., Serretti, A., Gasperini, M., et al., 1998. Familial concordance of fluvoxamine response as a tool for differentiating mood disorder pedigrees. J. Psychiatr. Res. 32, 255–259. Goldman, N., Glei, D.A., Lin, Y.H., et al., 2010. The serotonin transporter polymorphism (5-HTTLPR): allelic variation and links with depressive symptoms. Depress Anxiety 27, 260–269. Gressier, F., Bouaziz, E., Verstuyft, C., et al., 2009. 5-HTTLPR modulates antidepressant efficacy in depressed women. Psychiatr. Genet. 19, 195–200. Grigoriadis, S., Robinson, G.E., 2007. Gender issues in depression. Ann. Clin. Psychiatry 19, 247–255. Halfin, A., 2007. Depression: the benefits of early and appropriate treatment. Am. J. Manag. Care 13, S92–S97. Heils, A., Teufel, A., Petri, S., et al., 1996. Allelic variation of human serotonin transporter gene expression. J. Neurochem. 66, 2621–2624. Higgins, J.P., Thompson, S.G., Deeks, J.J., et al., 2003. Measuring inconsistency in meta-analyses. BMJ 327, 557–560. Higuchi, H., Sato, K., Yoshida, K., et al., 2008. No predictors of antidepressant patient response to milnacipran were obtained using the three-factor structures of the Montgomery and Asberg Depression Rating Scale in Japanese patients with major depressive disorders. Psychiatry Clin. Neurosci. 62, 197–202. Hong, C.J., Chen, T.J., Yu, Y.W., et al., 2006. Response to fluoxetine and serotonin 1A receptor (C-1019G) polymorphism in Taiwan Chinese major depressive disorder. Pharmacogenomics J. 6, 27–33. Hu, X.Z., Lipsky, R.H., Zhu, G., et al., 2006. Serotonin transporter promoter gain-of-function genotypes are linked to obsessive– compulsive disorder. Am. J. Hum. Genet. 78, 815–826. Hu, X.Z., Rush, A.J., Charney, D., et al., 2007. Association between a functional serotonin transporter promoter polymorphism and citalopram treatment in adult outpatients with major depression. Arch. Gen. Psychiatry 64, 783–792. Huezo-Diaz, P., Uher, R., Smith, R., et al., 2009. Moderation of antidepressant response by the serotonin transporter gene. Br. J. Psychiatry 195, 30–38. Illi, A., Poutanen, O., et al., 2010. Is 5-HTTLPR linked to the response of selective serotonin reuptake inhibitors in MDD? Eur. Arch. Psychiatry Clin. Neurosci. 261, 95–102. Joyce, P.R., Mulder, R.T., Luty, S.E., et al., 2003. Age-dependent antidepressant pharmacogenomics: polymorphisms of the serotonin transporter and G protein beta3 subunit as predictors of response to fluoxetine and nortriptyline. Int. J. Neuropsychopharmacol. 6, 339–346. Kang, R.-H., Wong, Ma-Li, et al., 2007. Association study of the serotonin transporter promoter polymorphism and mirtazapine antidepressant response in major depressive disorder. Progress in NeuroPsychopharmacology and Biological Psychiatry 31, 1317–1321. Kato, M., Serretti, A., 2010. Review and meta-analysis of antidepressant pharmacogenetic findings in major depressive disorder. Mol. Psychiatry 15, 473–500. Kato, M., Fukuda, T., Wakeno, M., et al., 2006. Effects of the serotonin type 2A, 3A and 3B receptor and the serotonin transporter genes on paroxetine and fluvoxamine efficacy and adverse drug reactions in depressed Japanese patients. Neuropsychobiology 53, 186–195. Kim, D.K., Lim, S.W., et al., 2000. Serotonin transporter gene polymorphism and antidepressant response. Neuroreport 11, 215–219. Kim, H., Lim, S.W., Kim, S., et al., 2006. Monoamine transporter gene polymorphisms and antidepressant response in Koreans with late-life depression. Jama 296, 1609–1618.

5-HTTLPR and antidepressant efficacy Kirchheiner, J., Nickchen, K., Bauer, M., et al., 2004. Pharmacogenetics of antidepressants and antipsychotics: the contribution of allelic variations to the phenotype of drug response. Mol. Psychiatry 9, 442–473. Kirchheiner, J., Nickchen, K., Sasse, J., et al., 2006. A 40-basepair VNTR polymorphism in the dopamine transporter (DAT1) gene and the rapid response to antidepressant treatment. Pharmacogenomics J. 7, 48–55. Kraft, J., Slager, S., McGrath, P., et al., 2005. Sequence analysis of the serotonin transporter and associations with antidepressant response. Biol. Psychiatry 58. Kraft, J.B., Peters, E.J., Slager, S.L., et al., 2007. Analysis of association between the serotonin transporter and antidepressant response in a large clinical sample. Biol. Psychiatry 61, 734–742. Kunugi, H., Hattori, M., Kato, T., et al., 1997. Serotonin transporter gene polymorphisms: ethnic difference and possible association with bipolar affective disorder. Mol. Psychiatry 2, 457–462. Lee, M.S., Lee, H.Y., Lee, H.J., et al., 2004. Serotonin transporter promoter gene polymorphism and long-term outcome of antidepressant treatment. Psychiatr. Genet. 14, 111–115. Lee, S.H., Choi, T.K., Lee, E., et al., 2010. Serotonin transporter gene polymorphism associated with short-term treatment response to venlafaxine. Neuropsychobiology 62, 198–206. Lotrich, F.E., Pollock, B.G., Kirshner, M., et al., 2008. Serotonin transporter genotype interacts with paroxetine plasma levels to influence depression treatment response in geriatric patients. J. Psychiatry Neurosci. 33, 123–130. Machado, M., Iskedjian, M., Ruiz, I., et al., 2006. Remission, dropouts, and adverse drug reaction rates in major depressive disorder: a meta-analysis of head-to-head trials. Curr. Med. Res. Opin. 22, 1825–1837. Maier, W., Zobel, A., 2008. Contribution of allelic variations to the phenotype of response to antidepressants and antipsychotics. Eur. Arch. Psychiatry Clin. Neurosci. 258 (Suppl 1), 12–20. Mandelli, L., Serretti, A., Zanardi, R., et al., 2007. Antidepressant response in the elderly. Psychiatry Res. 152, 37–44. Mandelli, L., Marino, E., Pirovano, A., et al., 2009a. Interaction between SERTPR and stressful life events on response to antidepressant treatment. Eur. Neuropsychopharmacol. 19, 64–67. Mandelli, L., Mazza, M., Martinotti, G., et al., 2009b. Harm avoidance moderates the influence of serotonin transporter gene variants on treatment outcome in bipolar patients. J. Affect. Disord. 119, 205–209. Maron, E., Tammiste, A., Kallassalu, K., et al., 2009. Serotonin transporter promoter region polymorphisms do not influence treatment response to escitalopram in patients with major depression. Eur. Neuropsychopharmacol. 19, 451–456. Masand, P.S., 2003. Tolerability and adherence issues in antidepressant therapy. Clin. Ther. 25, 2289–2304. Min, W., Li, T., Ma, X., et al., 2009. Monoamine transporter gene polymorphisms affect susceptibility to depression and predict antidepressant response. Psychopharmacology (Berl) 205, 409–417. Minov, C., Baghai, T.C., et al., 2001. Serotonin-2A-receptor and transporter polymorphisms: lack of association in patients with major depression. Neurosci. Lett. 303, 119–122. Mrazek, D.A., Rush, A.J., Biernacka, J.M., et al., 2009. SLC6A4 variation and citalopram response. Am. J. Med. Genet. B Neuropsychiatr. Genet. 150B, 341–351. Murphy Jr., G.M., Hollander, S.B., Rodrigues, H.E., et al., 2004. Effects of the serotonin transporter gene promoter polymorphism on mirtazapine and paroxetine efficacy and adverse events in geriatric major depression. Arch. Gen. Psychiatry 61, 1163–1169. Nakamura, M., Ueno, S., Sano, A., et al., 2000. The human serotonin transporter gene linked polymorphism (5-HTTLPR) shows ten novel allelic variants. Mol. Psychiatry 5, 32–38. Ng, C.H., Easteal, S., et al., 2006. Serotonin transporter polymorphisms and clinical response to sertraline across ethnicities. Prog. Neuropsychopharmacol. Biol. Psychiatry 30, 953–957.

257 Papakostas, G.I., Worthington III, J.J., Iosifescu, D.V., et al., 2006. The combination of duloxetine and bupropion for treatmentresistant major depressive disorder. Depress Anxiety 23, 178–181. Papakostas, G.I., McGrath, P., Stewart, J., et al., 2008. Psychic and somatic anxiety symptoms as predictors of response to fluoxetine in major depressive disorder. Psychiatry Res. 161, 116–120. Perlis, R.H., Mischoulon, D., Smoller, J.W., et al., 2003. Serotonin transporter polymorphisms and adverse effects with fluoxetine treatment. Biol. Psychiatry 54, 879–883. Perlis, R.H., Patrick, A., Smoller, J.W., et al., 2009. When is pharmacogenetic testing for antidepressant response ready for the clinic? A cost-effectiveness analysis based on data from the STAR*D study. Neuropsychopharmacology 34, 2227–2236. Perroud, N., Salzmann, A., Saiz, P.A., et al., 2010. Rare genotype combination of the serotonin transporter gene associated with treatment response in severe personality disorder. Am. J. Med. Genet. B Neuropsychiatr. Genet. 153B, 1494–1497. Pollock, B.G., Ferrell, R.E., et al., 2000. Allelic Variation in the Serotonin Transporter Promoter Affects Onset of Paroxetine Treatment Response in Late-Life Depression. Neuropsychopharmacology 23, 587–590. Popp, J., Leucht, S., Heres, S., et al., 2006. Serotonin transporter polymorphisms and side effects in antidepressant therapy—a pilot study. Pharmacogenomics 7, 159–166. Rausch, J.L., Johnson, M.E., Fei, Y.-J., et al., 2002. Initial conditions of serotonin transporter kinetics and genotype: influence on SSRI treatment trial outcome. Biol. Psychiatry 51, 723–732. Reimherr, F., Amsterdam, J., Dunner, D., et al., 2010. Genetic polymorphisms in the treatment of depression: speculations from an augmentation study using atomoxetine. Psychiatry Res. 175, 67–73. Rosenthal, N.E., Mazzanti, C.M., Barnett, R.L., et al., 1998. Role of serotonin transporter promoter repeat length polymorphism (5HTTLPR) in seasonality and seasonal affective disorder. Mol. Psychiatry 3, 175–177. Ruhe, H.G., Ooteman, W., Booij, J., et al., 2009. Serotonin transporter gene promoter polymorphisms modify the association between paroxetine serotonin transporter occupancy and clinical response in major depressive disorder. Pharmacogenet. Genomics 19, 67–76. Schinka, J.A., Busch, R.M., Robichaux-Keene, N., 2004. A metaanalysis of the association between the serotonin transporter gene polymorphism (5-HTTLPR) and trait anxiety. Mol. Psychiatry 9, 197–202. Serretti, A., Olgiati, P., et al., 2011. A model to incorporate genetic testing (5-HTTLPR) in pharmacological treatment of major depressive disorders. World J. Biol. Psychiatry 12, 501–515. Serretti, A., Franchini, L., Gasperini, M., et al., 1998. Mode of inheritance in mood disorder families according to fluvoxamine response. Acta Psychiatr. Scand. 98, 443–450. Serretti, A., Cusin, C., Rossini, D., et al., 2004. Further evidence of a combined effect of SERTPR and TPH on SSRIs response in mood disorders. Am. J. Med. Genet. B Neuropsychiatr. Genet. 129, 36–40. Serretti, A., Calati, R., Mandelli, L., et al., 2006. Serotonin transporter gene variants and behavior: a comprehensive review. Curr. Drug Targets 7, 1659–1669. Serretti, A., Calati, R., Oasi, O., et al., 2007a. Dissecting the determinants of depressive disorders outcome: an in depth analysis of two clinical cases. Ann. Gen. Psychiatry 6, 5. Serretti, A., Kato, M., De Ronchi, D., et al., 2007b. Meta-analysis of serotonin transporter gene promoter polymorphism (5-HTTLPR) association with selective serotonin reuptake inhibitor efficacy in depressed patients. Mol. Psychiatry 12, 247–257. Serretti, A., Chiesa, A., Calati, R., et al., 2009. Common genetic, clinical, demographic and psychosocial predictors of response to pharmacotherapy in mood and anxiety disorders. Int. Clin. Psychopharmacol. 24, 1–18.

258 Smeraldi, E., Zanardi, R., Benedetti, F., et al., 1998. Polymorphism within the promoter of the serotonin transporter gene and antidepressant efficacy of fluvoxamine. Mol. Psychiatry 3, 508–511. Smits, K.M., Smits, L.J., Schouten, J.S., et al., 2004. Influence of SERTPR and STin2 in the serotonin transporter gene on the effect of selective serotonin reuptake inhibitors in depression: a systematic review. Mol. Psychiatry 9, 433–441. Souery, D., Oswald, P., Massat, I., et al., 2007. Clinical factors associated with treatment resistance in major depressive disorder: results from a European multicenter study. J. Clin. Psychiatry 68, 1062–1070. Stamm, T.J., Adli, M., Kirchheiner, J., et al., 2008. Serotonin transporter gene and response to lithium augmentation in depression. Psychiatr. Genet. 18, 92–97. Taylor, M.J., Sen, S., Bhagwagar, Z., 2010. Antidepressant response and the serotonin transporter gene-linked polymorphic region. Biol. Psychiatry 68, 536–543. Trivedi, M.H., Rush, A.J., Wisniewski, S.R., et al., 2006. Evaluation of outcomes with citalopram for depression using measurementbased care in STAR*D: implications for clinical practice. Am. J. Psychiatry 163, 28–40. Umene-Nakano, W., Yoshimura, R., et al., 2009. Predictive factors for responding to sertraline treatment: views from plasma catecholamine metabolites and serotonin transporter polymorphism. J. Psychopharmacol. 24, 1764–1771. Walderhaug, E., Magnusson, A., Neumeister, A., et al., 2007. Interactive effects of sex and 5-HTTLPR on mood and impulsivity during tryptophan depletion in healthy people. Biol. Psychiatry 62, 593–599. Wang, P.S., Simon, G., Kessler, R.C., 2003. The economic burden of depression and the cost-effectiveness of treatment. Int. J. Methods Psychiatr. Res. 12, 22–33.

S. Porcelli et al. Wilkie, M.J., Smith, G., Day, R.K., et al., 2009. Polymorphisms in the SLC6A4 and HTR2A genes influence treatment outcome following antidepressant therapy. Pharmacogenomics J. 9, 61–70. Willeit, M., Praschak-Rieder, N., Neumeister, A., et al., 2003. A polymorphism (5-HTTLPR) in the serotonin transporter promoter gene is associated with DSM-IV depression subtypes in seasonal affective disorder. Mol. Psychiatry 8, 942–946. Yoshida, K., Ito, K., et al., 2002. Influence of the serotonin transporter gene-linked polymorphic region on the antidepressant response to fluvoxamine in Japanese depressed patients. Prog. Neuropsychopharmacol. Biol. Psychiatry 26, 383–386. Yoshida, K., Takahashi, H., et al., 2004. Prediction of antidepressant response to milnacipran by norepinephrine transporter gene polymorphisms. Am. J. Psychiatry 161, 1575–1580. Yoshimura, R., Umene-Nakano, W., et al., 2009. Rapid response to paroxetine is associated with plasma paroxetine levels at 4 but not 8 weeks of treatment, and is independent of serotonin transporter promoter polymorphism in Japanese depressed patients. Hum. Psychopharmacol. 24, 489–494. Zanardi, R., Benedetti, F., DiBella, D., et al., 2000. Efficacy of paroxetine in depression is influenced by a functional polymorphism within the promoter of serotonin transporter gene. J. Clin. Psychopharmacol. 20, 105–107. Zanardi, R., Serretti, A., et al., 2001. Factors affecting fluvoxamine antidepressant activity: influence of pindolol and 5-HTTLPR in delusional and nondelusional depression. Biol. Psychiatry 50, 323–330. Zobel, A., Maier, W., 2010. Pharmacogenetics of antidepressive treatment. Eur. Arch. Psychiatry Clin. Neurosci. 260, 407–417.