A transcription error has been discovered in the article “Serotonin Transporter (5-HTTLPR) Genotype and Amygdala Activation: A Meta-Analysis” by Munafò et al., which appeared in Biological Psychiatry, Volume 63, Number 9 (Biol Psychiatry 2008;63:852–857). This transcription error resulted in the erroneous inclusion of standard errors for one study instead of standard deviations, leading to an inflated effect size estimate for that study. Although it does not change the conclusions substantially, it does alter some of the values originally reported in the results of the article. Here, the authors report the correct results when the true effect size for this study is included;
        Our main results are largely unchanged—in particular, the effect in the main analysis remains statistically significant, although the effect size is slightly reduced (Z = 5.64, p < .001, d = .57, 95% CI .37 – .77). There is evidence of significant between-study heterogeneity (χ2 [13] = 32.85, p = .002), but the effect continues to remain nominally significant within a random-effects framework (d = .56, 95% CI .24 – .89).
        Two results are modified slightly – the comparison of the pooled effect size for studies using fMRI methods (d = .51), compared to other methods (d = 1.12), now indicates a trend towards a greater effect size among studies using other methods (p = .07), while Egger's test now indicates only a trend towards evidence of publication bias (p = .09) when three unpublished datasets are excluded.
        Two results are changed: the comparison of the effect size indicated by the first published study (d = 1.00) compared to the pooled effect size for subsequent studies (d = .54) now does not indicate a significant difference (p = .26), and meta-regression does not indicate association between year of publication and study effect size (p = .76).
        A power analysis, on the basis of our revised effect size estimate, now indicates that a sample of over 110 participants (assuming a 2:1 ratio of Short and Long genotypes) would be required to achieve 80% power to detect association with amygdala activation at an alpha level of .05, somewhat larger than in our original report.
        We are grateful to Jonathan Roiser and Elizabeth O'Nions for identifying the error in the original manuscript and bringing it to our attention.

        Linked Article