When using only genomewide significant SNPs, our results were consistent with a possible causal effect of smoking initiation on schizophrenia risk. However, this result was somewhat hard to interpret as only four SNPs were used in this risk score, and all of these were from the same gene region (BDNF). These SNPs are all highly correlated, and although correlation was taken in to account in the MR model, the risk of biological pleiotropy (i.e., an effect of gene on the outcome not via smoking) was much harder to rule out; BDNF and its associated SNPs have reached genomewide significance in a number of other GWAS, including those of obesity and body mass index1920, and caffeine consumption21.
