In a genome-wide gene–smoking interaction analysis, although common SNPs on chromosomes 6 and 19 showed suggestive SNP–smoking interactions (p<5 × 10−7; appendix p 78), no gene–smoking interactions were detected at genome-wide significance (p<5 × 10−8). Three of the six variants associated with FEV1 extremes (table 2) showed weak evidence of interaction with smoking (Bonferroni correction for six tests p=0·0083; appendix pp 60–62), including common variants at the HLA-DQB1/HLA-DQA2 and TSEN54 loci and the rare variant at the RBM19/TBX5 locus. In a meta-analysis of the genome-wide association test statistics for low FEV1 versus high FEV1 across heavy and never smokers, motivated by our finding of shared genetic causes between heavy and never smokers and to increase the sample size, we identified an additional six novel genome-wide significant signals of association with FEV1 extremes. These included CCDC91, reported as a novel signal of association with lung function in the general population by a concurrent study,35 and SLMAP, for which there is corroborative evidence of association with lung function (appendix pp 79–80).8 Our pathway analysis identified a novel signal of enrichment of the
