We maximised the power of our study by sampling from the extremes of a large biobank. No other similar resources of a comparable size were available for replication studies. Nevertheless, the novel genome-wide significant signals of association with FEV1 extremes in NPNT and KANSL1 in never smokers were also significantly associated in the independent set of heavy smokers. Furthermore, to corroborate the new signals we identified in TET2 and TSEN54, we present evidence of association with FEV1 in a previously reported large study of FEV1 in the general population.8 However, the rare SNP on chromosome 12 for which we found association with extremes of FEV1 was exclusive to the recently released UK10K Project component of the imputation panel and has not yet been measured in suitably large studies. Our comparison of smokers and never smokers represents a powerful approach because of the restriction to heavy smokers rather than ever smokers. For novel signals for smoking behaviour, we present additional evidence of association with smoking behaviour for NCAM1, TEX41/PABPC1P2, and NOL4L in independent populations.17,19 Although these independent datasets have limited power, they provide corroboration of key genome-wide significant findings in UK BiLEVE.
