This case-control study of tobacco smoking and acute myeloid leukemia (AML), emphasizing specific associations with morphologic and cytogenetic subtypes, comprised smoking histories for 333 cases and 351 controls. Smoking status (ever smokers versus life-long non-smokers) showed no evident effect on AML risk. However, an effect of smoking was indicated at high cumulative smoking doses (pack-years), e.g. 40 pack-years was associated with an odds ratio (OR) of 1.5 [95% confidence interval (CI) 1.0-2.3]. Among morphologic subtypes, the smoking associated OR for acute erythroleukemia was 8.9 (95% CI 1.0-76). No clear associations between smoking and cytogenetic subtypes of AML were observed.