Using genetic variants associated with smoking initiation at a less stringent significance threshold, 21 additional independent SNPs associated with smoking initiation (at p < 10−6) provided no evidence of a causal effect of smoking status on risk of schizophrenia (OR per doubling in odds of being a smoker including one BDNF SNP (rs6265) OR 1.03; 95% CI 0.97–1.09, p = 0.32). Heterogeneity in this analysis was somewhat large, (I2 = 61.3, 95% CI 38–76%, p < 0.001). However, MR Egger regression suggested no evidence of pleiotropy (intercept OR 1.03, 95% CI 1.00–1.06, p = 0.061), and weak evidence of causality (slope OR 0.69, 95% CI 0.43–1.08, p = 0.118). Results were similar if we excluded rs6265.
