We examined the association between N-acetyltransferase 1 and 2 (NAT1 and NAT2) haplotype and risk of pancreatic cancer by genotyping eight NAT1 and seven NAT2 single nucleotide polymorphisms in 532 patients and in 581 healthy controls (all non-Hispanic white) who were recruited at M. D. Anderson Cancer Center from January 2000 to December 2006. Haplotypes were reconstructed by using an expectation-maximization algorithm. Odds ratios and 95% confidence intervals were estimated by using unconditional logistic regression models. Covariates included age (continuous variable), sex, pack-year of smoking (categorical), and history of diabetes when appropriate. NAT1 and NAT2 genotype was mutually adjusted. The prevalence of haplotype NAT1*10-NAT2*6A was 4.3% versus 2.7% (P=0.06) and NAT1*11-NAT2*6A was 1.2% versus 0.4% (P=0.05) in patients and controls, respectively. The diplotype NAT1*10/*10 or NAT1*10/*11 and NAT2*6A/any slow allele was associated with a higher risk of pancreatic cancer compared with other diplotypes (multivariate odds ratio, 4.15; 95% confidence interval, 1.15-15.00; P=0.03). NAT2 slow genotype were associated with increased risk of pancreatic cancer among heavy smokers and among individuals with a history of diabetes. We for the first time found that rare NAT1*10 or NAT1*11-NAT2*6A diplotype may be an "at-risk" genetic variant for pancreatic cancer. The NAT2*6A/any slow acetylation genotype may be a predisposing factor for pancreatic cancer among diabetics with smoking exposure. Our observations must be confirmed in larger independent studies.