Evaluation of publication bias suggested that only 9 studies reporting a null effect would be sufficient to shift the observed effect to a non-significant level. A fail-safe N of 9 is substantially lower than the number of studies needed for a robust effect (n = 120) according to Rosenthal (1991). The funnel plot is presented in Figure 3. The plot appears to be roughly symmetrical by visual inspection. Additionally, there was a significant moderating effect of year of publication (β = -0.04, SE = 0.02, p < .05), reflecting a greater number of studies showing null results published in recent years. Indeed, the effect size across studies published since the Feinn et al., (2005) meta-analysis did not reflect a significant association between 5HTTLPR and alcohol dependence (OR = 0.93, 95% CI = 0.76, 1.14, p = .48).
