5-HTTLPR and anxiety-related personality traits meta-analysis revisited: response to Munafo and colleagues

Article metrics

Sir—In 2004, two meta analyses of studies investigating the association between the serotonin transporter length polymorphism repeat (5-HTTLPR) and anxiety-related personality traits were published and came to similar conclusions: a significant association between 5-HTTLPR and Neuroticism among studies using the NEO personality inventories, and a non-significant association between 5-HTTLPR and Harm Avoidance among studies using the TCI/TPQ inventories.1, 2 Recently, Munafo et al3 performed another meta-analysis on essentially the same association and reported markedly different results: no association between 5-HTTLPR and NEO Neuroticism (P=0.9757) and a significant association between 5-HTTLPR and TCI/TPQ Harm Avoidance (P=0.0024). Munafo and colleagues attribute this startling difference in results primarily to the different sets of studies and analytic method used by the three meta- analyses. In this manuscript, we show that these differences in method do not account for the discrepant results between the Sen and Munafo meta-analyses, confirm the results of the original Schinka and Sen meta-analyses and provide a possible alternative explanation of the discrepant results.

Munafo and colleagues attribute the difference in results between the Sen and Munafo meta-analyses specifically to three deficiencies in the Sen meta-analysis: (1) the absence of weighting studies by sample size; (2) the omission of studies that were either missed or published too recently to be included in our analysis; (3) the inclusion of studies either not in Hardy–Weinberg equilibrium or using samples drawn from populations with significant pathology.

The criticism that we did not weigh studies based on sample size is incorrect. In our analysis, studies were weighted by sample size/variance, which is proportional to the weighting factor, sample size, used by Munafo and co-workers. To assess whether the inclusion of additional studies and the stricter exclusion criteria account for the disparate results, we performed our meta-analysis again with the same set of studies included by Munafo and co-workers. One TCI/TPQ study4 could not be included because we were unable to obtain the necessary raw data from the paper or from the authors. The results we found were essentially unchanged from our original analysis: a highly significant association between 5-HTTLPR and Neuroticism among studies using the NEO personality inventories (P=0.0000053), and a nonsignificant association between 5-HTTLPR and Harm Avoidance (P=0.181) among studies using the TCI/TPQ inventories. Thus, it seems highly unlikely that differential study inclusion/exclusion accounts for the different results of our meta-analyses.

Another potential reason for the discrepancy in our results is the difference in analytical method between our studies. The Munafo method differs from our method through (1) weighting studies by sample size rather than sample size/variance; (2) using the l/l genotype group as a comparison group for the other two genotypes rather than assuming an additive effect; (3) performing the analyses in S-Plus rather than SPSS. We applied the Munafo parameters to our analysis to assess whether they account for the disparate results. With these parameters, among NEO studies, there was a significant difference between the l/l genotype group and both the s/s (P=0.0004) and the s/l groups (0.0006). For the TCI/TPQ group, there was a significant difference between the l/l and s/s groups (P=0.0160), but no significant difference between the l/l and s/l groups (0.4313). Thus, difference in analytic method does not account for discrepancy between our meta-analyses.

The last possible reason for the discrepancy that we consider is a difference in data collection and conversion. In Table 2 of the Munafo paper, the authors present the weighted standardized mean trait score and standard deviation across 5-HTTLPR genotypes for each of the two inventory groups (NEO and TCI/TPQ). We reproduced the data of this table and added the analogous data from our analysis (Table 1).

Table 1 Study specific trait T-scores organized by inventory group and genotype

In examining the tables, the calculated weighted means are similar in the two analyses. There is, however, an approximately three-fold difference in the calculated weighted standard deviations. We find the weighted standard deviation among T-scores in each genotype group to be substantially lower among NEO studies and substantially greater among TCI/TPQ studies than reported by Munafo and co-workers. Given its scale, this difference in standard deviation is likely to be the primary reason for the divergent results found by the two analyses: when means are similar, a larger SD will result in a less significant P-value. In looking at the genotype-specific T-scores in each of the two inventory groups (Table 1), it is clear that the variance in T-scores within each genotype group is much smaller among NEO studies than among TCI/TPQ studies. To ascertain if there was an error on our part in recording information or computing T-scores, we checked the recorded raw data against the original publications and recalculated the T-scores and standard deviations for each study. We found no discrepancies from our original data collection and analysis. Given these results, we believe that the results of our original meta-analysis hold: there is a highly significant association between 5-HTTLPR and Neuroticism among studies using the NEO personality inventories and a nonsignificant association between 5-HTTLPR and Harm Avoidance among studies using the TCI/TPQ inventories.

References

  1. 1

    Sen S et al. Meta-analysis of the association between a serotonin transporter promoter polymorphism (5-HTTLPR) and anxiety-related personality traits. Am J Med Genet B Neuropsychiatr Genet 2004; 127: 85–89.

  2. 2

    Schinka JA et al. A meta-analysis of the association between the serotonin transporter gene polymorphism (5-HTTLPR) and trait anxiety. Mol Psychiatry 2004; 9: 197–202.

  3. 3

    Munafo MR et al. Does measurement instrument moderate the association between the serotonin transporter gene and anxiety-related personality traits? A meta-analysis. Mol Psychiatry 2004; 10: 415–419.

  4. 4

    Benjamin J 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.

Download references

Author information

Correspondence to S Sen.

Rights and permissions

Reprints and Permissions

About this article

Further reading