We attempted to replicate the findings of two recent meta-analyses that personality inventory moderates the association between the serotonin transporter gene and anxiety-related traits. A total of 24 studies contributed to the meta-analysis, of which three reported genotype frequencies that deviated from Hardy–Weinberg (HW) equilibrium. We found some support for the view that results depend on the type of questionnaire used, although in a direction opposite to that previously reported. Contrasts between the S/S and L/L groups were significant for TCI/TPQ harm avoidance studies (P=0.0024) but not NEO neuroticism (P=0.9757). When studies not in HW equilibrium were excluded the TCI/TPQ result for the S/S genotype still exceeded our 5% threshold, although with reduced significance (P=0.0082), and the NEO result remained nonsignificant (P=0.9109). While we cannot rule out an association between the 5HTT gene and anxiety-related traits, particularly for TCI/TPQ harm avoidance, our findings do indicate that the effect, if present, is small. Our results emphasise the importance of complete ascertainment of studies and the identification of relevant sources of heterogeneity.
It is now accepted that a substantial proportion of variability in human personality is attributable to genetic variation.1 As theories of personality propose that individual differences in specific traits are associated with individual differences in specific neurotransmitters, a number of candidate genes with known functional effects on neurotransmitter pathways have been investigated.2 One pathway that has received considerable attention is the serotonin system, which is known to play an important role in emotional behaviour.3 However, attempts to demonstrate an association between variation in genes involved in serotonin function and neuroticism or harm avoidance, a personality trait believed to reflect inherited differences in emotional response, have yielded inconsistent results.
Even three independent meta-analyses of the large number of relevant publications do not come to a consensus: while Munafò et al4 found little evidence for an association, both Schinka et al5 and Sen et al6 reported that 5HTT-LPR was significantly associated with neuroticism as measured using the class of personality questionnaires derived from Costa and McCrae's7 five-factor model of personality (NEO-PI, NEO-PI-R, NEO-FFI), but was not associated with harm avoidance as measured using the class of questionnaires derived from Cloninger's8 tridimensional theory of personality (TCI, TPQ).
The findings of the Schinka and Sen meta-analyses are an important methodological and conceptual challenge to personality genetics. NEO neuroticism demonstrates substantial psychometric form equivalence with TCI/TPQ harm avoidance9 and, assuming that questionnaire equivalence reflects the measurement of the same underlying biological substrate, each assessment of neuroticism should be associated with the same genetic variants. Failure to provide comparable genetic associations questions the view that different personality measurements measure the same personality trait.
There are a number of differences between the three reports that might explain the inconsistent findings. First, the meta-analyses did not include the same studies for analysis; failure to ascertain all relevant literature for inclusion may seriously undermine any conclusions.10 The Sen meta-analysis omitted five published studies and one unpublished study that were included in the Munafò meta-analysis, all of which used the TCI/TPQ measure of harm avoidance, while the Schinka meta-analysis omitted three published studies and one unpublished study that were included in the Munafò meta-analysis, all of which also used the TCI/TPQ measure. Since the publication of the Munafò meta-analysis, three relevant studies have been published (which were included in the Sen meta-analysis), while the Schinka meta-analysis included two studies not reported in either the Munafò or the Sen meta-analyses. Second, trait personality and psychiatric case status may have been confounded. Both the Sen and Schinka meta-analysis included studies that recruited participants from psychiatric populations, whereas the Munafò meta-analysis explicitly excluded studies that recruited from psychiatric populations only and, when both psychiatric and control samples were recruited, only included data from normal, healthy controls. Third, the Schinka meta-analysis also only made a comparison across two genotype groups (L/L vs S/L+S/S), thereby precluding the investigation of different models of gene action (ie additive, dominant, recessive, etc). Fourth, and finally, the Sen meta-analysis transformed the raw score from each study genotype group into a T-score so that each study had an overall mean score of 50±10, in order to ensure that data from all studies was scaled equally. However, these data were then combined using inverse variance methods, which may result in each study contributing near equal weight irrespective of individual study size due to this standardisation. Given that the accuracy of an effect size estimate is proportionate to sample size, and that meta-analytic methods tend to give greater weight to larger studies, it is more appropriate to weigh individual studies accordingly.
We therefore attempted to replicate the findings of the Sen and Schinka meta-analyses that personality inventory moderates the association between the serotonin transporter gene and anxiety-related traits using a more comprehensive ascertainment of studies for inclusion while also enforcing a more stringent exclusion criterion to include only data from nonpsychiatric populations and employing a statistical analysis that enabled a comparison across all genotype groups, weighted by individual genotype group sample size.
Materials and methods
Selection of studies for inclusion
Studies were drawn primarily from the three existing meta-analyses of the association between genetic polymorphisms and personality traits.4, 5, 6 In addition, the search strategy used in our previous meta-analysis4 was re-run (24 June 2004) to capture any additional studies published subsequent to this.
Association studies of genetic polymorphisms and personality in healthy adults (aged 16 or over) were included. Studies reporting data on male and female participants of any ethnic origin were included. Personality studies were only included if they used a measure of either NEO neuroticism (ie NEO-PI, NEO-PI-R, NEO-FFI) or TCI/TPQ harm avoidance. Data from psychiatric populations (including alcoholism and organic disease) were excluded; in cases where a single study included data on both psychiatric patients and normal controls only the data from normal controls was included. In addition, data that appeared in more than one published study was only included once in the analysis.
Data were extracted from papers using a standardised data extraction form by two authors (MM and TC). Discrepancies were resolved by mutual consent. Data extracted included number of participants, mean neuroticism/harm avoidance trait score and standard deviation for each of the three genotype groups (L/L, S/L, S/S) in each study eligible for inclusion. The male/female ratio and predominant ethnicity of the sample was also extracted from included studies. Genotype frequencies were used to calculate whether or not they deviated significantly from Hardy–Weinberg (HW) equilibrium. In cases where data was not available in the published report in this format authors were contacted directly.
Data were analysed using the S-Plus (Version 6.1) statistical software package. The raw score from each study genotype group was converted to a T-score so that each study had an overall mean score of 50±10. This conversion ensured that data from all studies were scaled similarly. We estimated the pooled effects of genotype using a weighted mixed (linear) model, with individual study genotype T-scores as the dependent variable, study as a random effect and genotype group as a fixed effect. Sex ratio and ethnicity were included in the model as covariates. Weighting was by the individual genotype group sample sizes. Genotype group levels were S/S and S/L, with L/L as the comparison group. A dominant genetic model was tested by combining the S/S and S/L groups (S/S+S/L vs L/L), while a recessive model of genetic action was tested by combining the S/L and L/L groups (S/S vs S/L+L/L).
Where a single study included data on both questionnaires this was included in each analysis separately. Separate analyses were conducted for NEO neuroticism and TCI/TPQ harm avoidance. Tests of genetic association using a case–control design assume that allele frequencies will be in HW equilibrium. Departures from equilibrium may reflect the presence of population stratification, inbreeding or genotyping error, which will prejudice the association test result. We therefore conducted sensitivity analyses excluding studies with genotype frequencies that deviated from HW equilibrium, as indicated by a χ2 test. Studies were only excluded where there was evidence of departure from HW equilibrium. If there was insufficient data to test for departure from HW equilibrium the study was retained. The power of each individual study to detect departure from HW equilibrium was calculated using simulation methods.
Description of studies
A total of 26 studies met our inclusion criteria.11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36 In total, 13 studies reported data on NEO neuroticism and 16 studies reported data on TCI/TPQ harm avoidance. This included three studies that reported data on both NEO neuroticism and TCI/TPQ harm avoidance14, 20, 21 and two studies not reported in either the Munafò, Schinka or Sen meta-analyses.33, 35 In four studies genotype frequencies deviated significantly from HW equilibrium.11, 18, 24, 31 In one study21 there was insufficient data to test for departure from HW equilibrium. In two studies data were not reported in a format that enabled their inclusion in the subsequent meta-analysis28, 31 and attempts to contact the authors of these studies to request data in an appropriate format were unsuccessful. A total of 24 studies therefore contributed to the meta-analysis, of which three reported genotype frequencies that deviated from HW equilibrium. The characteristics of these studies are described in Table 1.
Since we tested two groups of studies (NEO and TCI/TPQ) and tested both dominant and recessive models of genetic action, we used P-value of 0.0125 for the 5% significance threshold. We obtained this threshold by using a Bonferroni correction for four independent tests.
A χ2 test did not indicate the presence of between-study heterogeneity in the NEO studies (P>0.99); between-study effects of sex ratio (P=0.1485) and ethnicity (P=0.8001) were nonsignificant. A weighted mixed (linear) model for NEO neuroticism did not indicate a significant effect of genotype on neuroticism. With L/L genotype as the comparison group neither the effects of S/S (P=0.5749) nor S/L (P=0.5288) genotype were significant. The tests of dominant (P=0.8224) and recessive (P=0.9180) models of genetic action were nonsignificant. When one study not in HW equilibrium was excluded, the results did not change substantively, with the effects of S/S (P=0.6665) and S/L (P=0.6147) genotype, and the tests of dominant (P=0.9540) and recessive (P=0.8336) genetic action, all remaining nonsignificant. The effects of sex ratio (P=0.1534) and ethnicity (P=0.8721) remained nonsignificant.
Analysis of the TCI/TPQ harm avoidance studies yielded similar findings. Again we found no evidence of between-study heterogeneity (P>0.99); between-study effects of sex ratio (P=0.0813) and ethnicity (P=0.2119) were nonsignificant. A weighted mixed (linear) model for TCI/TPQ harm avoidance indicated marginal evidence for a significant effect of genotype on neuroticism. With L/L genotype as the comparison group the effect of S/S genotype (P=0.0024) was significant but the effect of S/L (P=0.1936) genotype was not. The test of a dominant (P=0.2534) model of genetic action was not significant, while the test of a recessive (P=0.0003) model of genetic action was significant. When two studies not in HW equilibrium were excluded the results did not change substantively, with the effect of S/S (P=0.0082) genotype remaining significant, and the effects S/L (P=0.3535) genotype remaining nonsignificant. The test of a dominant (P=0.2620) model of genetic action remained nonsignificant, while the test of a recessive (P=0.0021) model of genetic action remained significant. The effects of sex ratio (P=0.0992) and ethnicity (P=0.2629) remained nonsignificant.
The weighted mean NEO neuroticism and TCI/TPQ harm avoidance scores across genotype groups for included studies are presented in Table 2.
Our meta-analysis offers some support for the view that genetic association between 5HTT-LPR and anxiety-related traits depends on the type of questionnaire used, although in a direction opposite to that previously reported. Contrasts between the S/S and L/L groups were significant for TCI/TPQ harm avoidance (P=0.0024) but not NEO neuroticism (P=0.9757) studies. While these results support the hypothesis of a relationship between 5HTT-LPR and anxiety-related traits measured using TCI/TPQ harm avoidance instruments, they are marginal. When studies not in HW equilibrium were excluded the effect for TCI/TPQ harm avoidance still exceeded our 5% threshold, although with reduced significance (P=0.0082), while the result for NEO neuroticism remained nonsignificant (P=0.9109). The test of a recessive model of genetic action for TCI/TPQ measures also remained significant when studies that departed from HW equilibrium were omitted (P=0.0021). Departures from HW equilibria probably reflect genotyping errors, but are not a very sensitive measure of this source of error. If, as we suspect, our meta-analysis includes additional studies with genotyping errors, then the results we report here are still likely to be biased, although the 5HTT-LPR is no more vulnerable to genotyping error than other markers that are used to detect size differences, rather than sequence differences. At the least, this observation cautions against drawing firm conclusions from the analysis.
Discrepancies between our earlier meta-analysis3 and two others5, 6 and the finding that results depend on the questionnaire used (and the direction of any effect), appear to arise from the inclusion of different studies and the employment of different meta-analytic methods. For the present meta-analysis, we have combined the studies captured by three separate meta-analyses3, 5, 6 and re-searched the published literature. We are confident that we have identified all relevant studies available at the time of writing, and we do not believe our findings are biased by ascertainment or heterogeneity, at least due to sex and ethnicity. The explicit exclusion of data collected on psychiatric samples reduces the possibility of confounding an association between the 5HTT gene and other psychiatric diseases with anxiety-related traits. The inclusion of male/female ratio and predominant ethnicity of individual study samples did not indicate significant main effects of either of these variables, and there was no evidence of significant between-study heterogeneity. The inclusion of three genotype groups in the primary analysis allows the testing of additive, dominant and recessive models of genetic action, while weighting by individual genotype group sample sizes ensures that greater weight is given to larger (ie more accurate) studies.
In summary, while we cannot rule out an association between the 5HTT gene and anxiety-related traits, particularly for TCI/TPQ harm avoidance, our findings do indicate that the effect, if present, is small. Our results emphasise the importance of complete ascertainment of studies and the identification of relevant sources of heterogeneity.
Loehlin JC . Nature, nurture, and conservatism in the Australian twin study. Behav Genet 1993; 23: 287–290.
Depue RA, Collins PF . Neurobiology of the structure of personality: dopamine, facilitation of incentive motivation, and extraversion. Behav Brain Sci 1999; 22: 491–517.
Lesch KP, Zeng Y, Reif A, Gutknecht L . Anxiety-related traits in mice with modified genes of the serotonergic pathway. Eur J Pharmacol 2003; 480: 185–204.
Munafò MR, Clark TG, Moore LR, Payne E, Walton RT, Flint J . Genetic polymorphisms and personality: a systematic review and meta-analysis. Mol Psychiatry 2003; 8: 471–484.
Schinka JA, Musch RM, Robichaux-Keene N . A meta-analysis of the association between the serotonin transporter gene polymorphism (5-HTTLPR) and trait anxiety. Mol Psychiatry 2004; 9: 197–202.
Sen S, Burmeister M, Ghosh D . Meta-analysis of the association between a serotonin transporter promoter polymorphism (5-HTTLPR) and anxiety-related personality traits. Am J Med Genet 2004; 127B: 85–89.
Costa Jr PT, McCrae RR . Stability and change in personality assessment: the revised NEO personality inventory in the year 2000. J Pers Assess 1997; 68: 86–94.
Cloninger CR . A unified biosocial theory of personality and its role in the development of anxiety states. Psychiatric Dev 1986; 3: 167–226.
De Fruyt F, Van De Wiele L, Van Heeringen C . Cloninger's psychobiological model of temperament and character and the five-factor model of personality. Pers Individual Diff 2000; 29: 441–452.
Munafò MR, Flint J . Meta-analysis of genetic association studies. Trends Genet 2004; 20: 439–444.
Cruz C, Camarena B, Mejia JM, Paez F, Eroza V, De F et al. The dopamine d2 receptor gene TaqI a1 polymorphism and alcoholism in a Mexican population. Arch Med Res 1995; 26: 421–426.
Lesch KP, Bengel D, Heils A, Sabol SZ, Greenberg B, Petri S et al. Association of anxiety-related traits with a polymorphism in the serotonin transporter gene regulatory region. Science 1996; 274: 1527–1531.
Ebstein RP, Nemanov L, Klolz I, Gritsenko I, Belmaker RH . Additional evidence for an association between the dopamine D4 receptor (D4DR) exon III repeat polymorphism and the human personality trait of Novelty Seeking. Mol Psychiatry 1997; 2: 472–477.
Nakamura K, Muramatsu T, Ono Y, Matshushitu S, Higuchi S, Mizushima H et al. Serotonin transporter gene regulatory region and anxiety-related traits in the Japanese. Am J Med Genet Neuropsychiatric Genet 1997; 74: 544–545.
Mazzanti CM, Lappalainen J, Long JC, Bengel D, Naukkarinen H, Eggert M et al. Role of the serotonin transporter promoter polymorphism in anxiety-related traits. Arch Gen Psychiatry 1998; 55: 936–940.
Ricketts MH, Hamer RM, Sage JI, Manowitz P, Feng F, Menza MA . Association of a serotonin transporter gene promoter polymorphism with harm avoidance behaviour in an elderly population. Psychiatric Genet 1998; 8: 41–44.
Flory JD, Manuck SB, Ferrell RE, Dent KM, Peters DG, Muldoon MF . Neuroticism is not associated with the serotonin transporter (5-HTTLPR) polymorphism. Mol Psychiatry 1999; 4: 93–96.
Hamer DH, Greenberg BD, Sabol SZ, Murphy DL . Role of the serotonin transporter gene in temperament and character. J Pers Dis 1999; 13: 312–328.
Katsuragi S, Kiyota A, Tsutsumi T, Isogawa K, Nagayama H, Arinami T et al. Lack of association between a polymorphism in the promoter region of the dopamine D2 receptor and personality traits. Psychiatry Res 2001; 105: 123–127.
Kumakiri C, Kodama K, Shimizu E, Yamanouchi N, Okada SI, Noda S et al. Study of the association between the serotonin transporter gene regulatory region polymorphism and personality traits in a Japanese population. Neurosci Lett 1999; 263: 205–207.
Benjamin J, Osher Y, Kotler M, Gritsenko I, Nemanov L, Belmaker RH et al. Association between tridimensional personality questionnaire (TPQ) traits and three functional polymorphisms: dopamine receptor D4 (DRD4), serotonin transporter promoter region (5-HTTLPR) and catechol O-methyltransferase (COMT). Mol Psychiatry 2000; 5: 96–100.
Comings DE, Gade-Andavolu R, Gonzalez N, Wu S, Muhleman D, Blake H et al. A multivariate analysis of 59 candidate genes in personality traits: the temperament and character inventory. Clin Genet 2000; 58: 375–385.
Du L, Bakish D, Hrdina PD . Gender differences in association between serotonin transporter gene polymorphism and personality traits. Psychiatric Genet 2000; 10: 159–164.
Greenberg BD, Li Q, Lucas FR, Hu S, Sirota LA, Benjamin J et al. Association between the serotonin transporter promoter polymorphism and personality traits in a primarily female population sample. Am J Med Genet Neuropsychiatric Genet 2000; 96: 202–216.
Herbst JH, Zonderman AB, McCrae RR, Costa PT . Do the dimensions of the temperament and character inventory map a simple genetic architecture? Evidence from molecular genetics and factor analysis. Am J Psychiatry 2000; 157: 1285–1290.
Hu S, Brody CL, Fisher C, Gunzerath L, Nelson ML, Sabol SZ et al. Interaction between the serotonin transporter gene and neuroticism in cigarette smoking behavior. Mol Psychiatry 2000; 5: 181–188.
Osher Y, Hamer D, Benjamin J . Association and linkage of anxiety-related traits with a functional polymorphism of the serotonin transporter gene regulatory region in Israeli sibling pairs. Mol Psychiatry 2000; 5: 216–219.
Schmidt NB, Storey J, Greenberg BD, Santiago HT, Li Q, Murphy DL . Evaluating gene × psychological risk factor effects in the pathogenesis of anxiety: a new model approach. J Abnorm Psychol 2000; 109: 308–320.
Strobel A, Brocke B, Ebstein RP . Interaktionseffekte monoamin-relevanter genetischer polymorphismen mit traits des TPQ. (Interaction effects of monoamine-relevant genetic polymorphisms with traits of the TPQ.) Ze Differ Diagno Psychol 2000; 21: 194–199.
Samochowiec J, Rybakowski F, Czerski P, Zakrzewska M, Stepien G, Pelka-Wysiecka J et al. Polymorphisms in the dopamine, serotonin, and norepinephrine transporter genes and their relationship to temperamental dimensions measured by the temperament and character inventory in healthy volunteers. Neuropsychobiology 2001; 43: 248–253.
Cohen H, Buskila D, Neumann L, Ebstein RP . Confirmation of an association between fibromyalgia and serotonin transporter promoter region (5-HTTLPR) polymorphism, and relationship to anxiety-related personality traits. Arthritis Rheum 2002; 46: 845–847.
Tsai SJ, Hong CJ, Cheng CY . Serotonin transporter genetic polymorphisms and harm avoidance in the Chinese. Psychiatr Genet 2002; 12: 165–168.
Brummett BH, Siegler IC, McQuiod DR, Svenson IK, Marchuk DA, Steffens DC . Associations among the NEO Personality Inventory, Revised and the serotonin transporter gene-linked polymorphic region in elders: effects of depression and gender. Psychiatr Genet 2003; 13: 13–18.
Umekage T, Tochigi M, Marui T, Kato C, Hibino H, Otani T et al. Serotonin transporter-linked promoter region polymorphism and personality traits in a Japanese population. Neurosci Lett 2003; 337: 13–16.
Lang UE, Bajbouj M, Wenicke C, Rommelspacher H, Danker-Hopfe H, Gallinat J . No association of a functional polymorphism in the serotonin transporter gene promoter and anxiety-related personality traits. Neuropsychobiology 2004; 49: 182–184.
Sen S, Villafuerte S, Nesse R, Stoltenberg SF, Hopcian J, Gleiberman L et al. Serotonin transporter and GABA(A) alpha 6 receptor variants are associated with neuroticism. Biol Psychiatry 2004; 55: 244–249.
Marcus Munafò is funded by a Cancer Research UK Research Fellowship. Taane Clark is funded by an NHS R&D Research Training Fellowship. Jonathan Flint is supported by the Wellcome Trust.
About this article
Effect of genetic polymorphism of brain-derived neurotrophic factor and serotonin transporter on smoking phenotypes: A pilot study of Japanese participants
Testing bidirectional relationships between alcohol use and depressive symptoms: What is the role of the serotonin transporter gene?
Dating violence victimization and substance use: The role of a serotonin transporter gene polymorphism (5HTTLPR)
Drug and Alcohol Dependence (2018)
The association between COMT rs4680 and 5-HTTLPR genotypes and concussion history in South African rugby union players
Journal of Sports Sciences (2018)
Reaction to norm transgressions and Islamization threat in culturally tight and loose contexts: a cross-cultural comparison of Germany versus Russia
Culture and Brain (2018)