Temporal patterns of association between cigarette smoking and leukemia risk

Objectives To evaluate variation in smoking-related leukemia risk with time-since-exposure. Methods We analyzed data from a population-based case-control study conducted in Germany. Odds ratios were estimated by applying conditional logistic regression methods to 470 incident leukemia cases and 1,009 controls. Cases were classified as acute non-lymphocytic leukemia (ANLL), acute lymphocytic leukemia (ALL), chronic myeloid leukemia (CML), and chronic lymphocytic leukemia (CLL). Temporal variation in the impact of smoking on leukemia risk was assessed via exposure time-windows and a spline latency function. Results Current smokers were at greater risk of ANLL than those who never smoked (OR = 1.65 95% CI: 0.95, 2.87) and a positive trend was observed in ANLL risk with cumulative pack-decades smoked, under a 2-year exposure lag assumption (OR/pack-decade = 1.11 95% CI: 0.96, 1.30). This was primarily due to the association between ANLL and smoking in the period 2 to < 10 years prior (OR/pack-decade = 2.72 95% CI: 0.93, 7.99). There was minimal evidence of association between ANLL risk and packs smoked 10 or more years prior. CML and ALL exhibited minimal evidence of association with smoking status. CLL exhibited a positive association with smoking in the periods 2 to < 10 years and 10 to < 20 years prior to diagnosis although estimates of association were highly imprecise. Conclusions The temporal pattern of smoking-induced ANLL risk appears to follow a prompt peak in excess incidence that diminishes with time since exposure.