American Journal of Medical Genetics Part B (Neuropsychiatric Genetics) 127B:85–89 (2004) Meta-Analysis of the Association Between a Serotonin Transporter Promoter Polymorphism (5-HTTLPR) and Anxiety-Related Personality Traits Srijan Sen,1* Margit Burmeister,1,2 and Debashis Ghosh3 1Neuroscience Program, University of Michigan, Ann Arbor, Michigan 2Mental Health Research Institute, Departments of Psychiatry and Human Genetics, University of Michigan, Ann Arbor, Michigan 3Department of Biostatistics, University of Michigan, Ann Arbor, Michigan Anxiety-related personality traits, such as NEO KEY WORDS: neuroticism; harm avoidance; neuroticism and TCI/TPQ harm avoidance, have personality; variant; polymorph- been shown to have significant genetic compo- ism; depression nents. To date, however, no specific genetic vari- ants that contribute to these traits have been conclusively identified. At least 26 studies have INTRODUCTION investigated a putative association between a functional serotonin transporter promoter poly- Evidence from twin studies indicate that a substantial com- morphism (5-HTTLPR) and anxiety-related person- ponent of human personality trait variation is due to genetic ality traits. The results of these studies have been factors [Loehlin, 1993]. Finding the specific genetic variants inconsistent with some studies finding evidence responsible for the personality trait variation however has for an association, and others not. We performed proven difficult. A personality trait extensively studied in a meta-analysis of all applicable studies investi- behavioral genetics is neuroticism. Neuroticism, as measured gating this association. In the overall analysis through the NEO personality inventories (NEO-PI-R, NEO-PI, (N ¼ 5,629 subjects), we found suggestive evidence and NEO-FFI), is characterized by ‘‘negative emotionality’’ for an association between the 5-HTTLPR short such as anxiety, low mood, vulnerability, and hostility [Costa allele (s) and increased anxiety-related personal- and McCrae, 1997]. ity trait scores (P ¼ 0.087). However, we also found Lesch et al. [1996] reported an association between a strong evidence for heterogeneity. This hetero- serotonin transporter promoter polymorphism (5-HTTLPR) geneity is largely explained by substantial varia- and neuroticism. The serotonin transporter is located on the tion between the studies in the inventory used. presynaptic membrane of serotonergic neurons and acts to When the analysis was stratified by inventory resorb serotonin from the synapse. This protein is a target of type, there was a significant association between the SSRI class of anti-depressant/anti-anxiety medications. 5-HTTLPR and NEO neuroticism (P ¼ 0.000016), a 5-HTTLPR is a repeat polymorphism with two alleles, a non-significant association between 5-HTTLPR 14 repeat (s) and 16 repeat (l), predominant in most population and TCI/TPQ harm avoidance (P ¼ 0.166), and no samples studies to date [Nakamura et al., 2000]. There is association between 5-HTTLPR and other anxi- substantial evidence that the l allele is transcriptionally more ety-related personality traits (P ¼ 0.944). There active than the alternate s allele [Heils et al., 1995; Hanna was no evidence that these results were either et al., 1998]. due to publication bias or accounted for by any Since the initial report of an association between 5-HTTLPR one single study. We conclude that there is a and neuroticism, there have been numerous attempts to re- strong association between the serotonin trans- plicate the finding. Some of these studies followed Lesch and porter promoter variant and neuroticism as colleagues and used NEO neuroticism as their phenotype [Ball measured in the NEO personality inventory and et al., 1997; Nakamura et al., 1997; Gelernter et al., 1998; that non-replications are largely due to small Deary et al., 1999; Flory et al., 1999; Kumakiri et al., 1999; Du sample size and the use of different inventories. et al., 2000; Greenberg et al., 2000; Sen et al. (in press); ß 2004 Wiley-Liss, Inc. Stoltenberg et al., 2002; Umekage et al., 2003]. Another subset of studies investigated harm avoidance, a trait correlated with neuroticism measured by the temperament and character inventory (TCI)/tridimenstional personality questionnaire (TPQ) family of personality inventories [Ebstein et al., 1997; Mazzanti et al., 1998; Ricketts et al., 1998; Hamer et al., 1999; Katsuragi et al., 1999; Osher et al., 2000; Samochowiec et al., 2001; Tsai et al., 2002]. Harm avoidance is characterized by anxiety proneness and an aversion to risk taking [Zohar et al., 2003]. A third subset of studies employed other Grant sponsor: The Michigan Society of Fellows; Grant sponsor: correlated but distinct traits as outcome measures [Jorm Rachel Upjohn Clinical Scholars Program; Grant sponsor: NRSA; et al., 1998; Gustavsson et al., 1999; Murakami et al., 1999; Grant number: MH64299-01A1. Melke et al., 2001]. The results from these replication studies *Correspondence to: Srijan Sen, MHRI c/o Burmeister Lab, 205 have been inconsistent. Seven replication studies found Zina Pitcher Place, Ann Arbor, MI 48109. significant evidence for an association between the 5-HTTLPR E-mail: [email protected] s allele and higher anxiety-related personality trait scores Received 14 March 2003; Accepted 16 September 2003 while seventeen replication studies found no significant DOI 10.1002/ajmg.b.20158 evidence of this association. ß 2004 Wiley-Liss, Inc. 86 Sen et al. This inconsistency is typical for association studies involving necessary information from the publication or the authors complex traits [Hirschhorn et al., 2002]. One potential reason [Gelernter et al., 1998] or (2) the study design was incompatible for discrepancies between studies investigating the same as- with our analysis because only extreme values of neuroticism sociation is that with the sample sizes typically used in were genotyped [Ball et al., 1997; Deary et al., 1999]. association studies, it is difficult to distinguish between genes that exert small effects on a complex trait and genes that exert Statistical Analysis no effect at all. The situation can be further complicated by variation between studies in both the population samples We recorded the number of subjects, mean anxiety-related studied and the method used to evaluate the relevant trait. A personality trait score, and standard deviation for each of the technique that has proven useful in resolving discrepancies three genotype groups (s/s, s/l, and l/l) in each study included in between association studies is meta-analysis [Lohmueller our analysis. We also recorded the inventory used in each study et al., 2003]. Meta-analysis is a quantitative method of com- as well as the gender and ethnic compositions of the sample bining the results of independent studies and synthesizing used in each study. Studies were grouped into three inventory summaries and conclusions. This method increases power to categories: (1) NEO (NEO-PI, NEO-PI-R, and NEO-FFI), (2) distinguish between small effects and no effect. Furthermore, TCI/TPQ (TCI and TPQ), and (3) others (EPQ, KSP, and SRQ- it can help determine whether variation in effect between AD). For studies where all or part of this information was not studies is due merely to expected random statistical fluctua- available in the publication, the authors were contacted by tion, or also due to variation between studies in the sample email. used or trait assessment. Meta-analysis has been used suc- In order to have all studies on the same scale for our analysis, cessfully to both confirm [Altshuler et al., 2000; Faraone et al., the raw score from each study-genotype group was converted to 2001] and refute [Kluger et al., 2002; Lalovic and Turecki, a T-score so that each study had an overall mean score of 2002] putative associations between genetic variants and 50 Æ 10. A random effects meta-analysis was performed with complex traits. In this study, we perform a meta-analysis on the study-genotype T-score as the dependent variable and all available studies investigating the association between genotype as the independent variable. Ethnicity and gender 5-HTTLPR and anxiety-related personality traits. composition of the study were included as covariates and each study-genotype group was weighted according to the inverse of its variance. Three studies [Nakamura et al., 1997; Kumakiri MATERIALS AND METHODS et al., 1999; Osher et al., 2000] administered multiple inven- tories to the same sample. For these studies, the inventory Studies for which more subjects had valid trait scores was used in Studies were identified through PubMed at the National the overall analysis. For analyses specific to one inventory, the Library of Medicine using the search terms: (1) neuroticism studies using multiple inventories were included if the serotonin transporter, (2) harm avoidance serotonin transpor- relevant inventory was used in the study regardless of whether ter, (3) personality serotonin transporter. We subsequently the study used other inventories as well. checked the reference sections of the publications found To ascertain if the results of our analysis were strongly in- through our search to identify additional studies that may fluenced by any single study a sensitivity analysis was per- have been missed. We restricted the scope of our analysis formed. Both the overall significance and the inventory specific to studies that used adult samples. In total, 26 studies from significance of the analyses were recomputed after each study 24 publications were identified, and 23 studies were included was individually deleted from the analysis (Table II). in the analysis (Table I). Three of these studies were exclud- In order to determine if there is a significant publication bias ed from our analysis because: (1) we were unable to obtain the method described by Egger et al. [1997] was used. This TABLE I. Included Association Studies of 5-HTTLPR and Anxiety-Related Traits Study Inventory N Mean age Ethnicity Female (%)a Recruitment Jorm et al. [1998] EPQ 759 41.5 95% Caucasian 53 Volunteer Hamer et al. [1999] TCI 634 31.3 79% Caucasian 57 Volunteer Sen et al.