that log odds ratios and their standard errors were calculated. The association between cotinine levels and lung cancer risk is shown in Supplementary Figure 1 (available online). Weighted least squares regression was then carried out on log odds ratios, weighted by the inverse of the standard error squared, with the intercept forced to be 0 to represent the odds ratio of 1 for the lowest cotinine interval. The resulting regression coefficient was then multiplied by the per-allele increase in cotinine estimated from this study and converted to an expected odds ratio for lung cancer risk associated with each additional copy of the rs1051730–rs16969968 risk allele. Lower and upper limits of the confidence interval from the per-allele increase were also calculated in the same way.
