Fixed effects meta-analyses across the individual contributing studies of single variant associations were undertaken using the Cochran-Mantel-Haenszel method in RAREMETAL. Z-score statistics were used in the meta-analysis to ensure that the association results are robust against potentially different units of measurement in the phenotype definitions across studies [23]. We performed genomic control correction on the meta-analysis results. Variants with P < 1 × 10−6 in tests of heterogeneity were excluded. Variants with P ≤ 5 × 10−8 were taken forward for replication. In addition, rs12616219 was also taken forward for replication as its P-value was very close to this threshold (smoking initiation, P = 5.49 × 10−8). None of the rare SNVs were genome-wide significant, therefore we also took forward the rare variant with the smallest association P-value, rs141611945 (P = 2.95 × 10−7; MAF < 0.0001).
