To evaluate the robustness of these findings, we explored the sensitivity of the results to alternative codings of CPD and age at onset. The 2 alternative CPD codings used included (1) heavy smoking (CPD >20) vs light smoking (CPD ≤10), excluding the intermediate group, and (2) CPD as an ordinal variable coded 0 (CPD ≤10), 1 (CPD of 11–20), 2 (CPD of 21–30), and 3 (CPD >30), as described previously herein. The 3 alternative codings of age at onset included (1) dichotomized as 16 years or younger or older than 16 years (primary variable); (2) categorized as younger than 15, 15 to 16, 17 to 18, and older than 18 years; and (3) categorized by within-study quartile. The results of the secondary analyses are presented in detail in eAppendix 3.
