Cyclophosphamide is a known bladder carcinogen, with cumulative dose directly related to increased risk. There is no consensus, however, on which major cyclophosphamide metabolite (i.e., acrolein or phosphoramide mustard) drives bladder carcinogenesis. We examined 19 cyclophosphamide-related bladder tumors to test the hypothesis that they might contain somatic mutations in the p53 tumor suppressor gene that could link a specific metabolite to the etiology of these cancers. Forty-three % (9 of 19) of the cases had a mutation in p53, with a predominance at G:C bp (7 of 9, 77%), a preference for non-CpG sites (6 of 7, 86%), and frequent G:C-->A:T transitions (5 of 7, 71%). The p53 mutation spectrum of these cyclophosphamide-associated bladder cancers differed significantly from patterns reported for sporadic (P = 0.020), smoking-related (0.043), and schistosomiasis-linked (P = 0.002) tumors but not arylamine-associated neoplasms (P = 0.860). Differences between the cyclophosphamide and arylamine-associated spectra included an unusual degree of clustering of exon 6 mutations (43% versus 17%, respectively) and an absence of multiple mutations in the former. Notably lacking in our series were G:C-->T:A transversions, the principal mutation associated with acrolein. Instead, the mutation spectrum matches the phosphoramide mustard adduction sequences determined by a repetitive primer-extension assay (P = 0.024), indicating that this metabolite might be a key mutagen in cyclophosphamide-related bladder cancer.