die of lung cancer compared with 1.0 percent of the wounded control, which while elevated was not significant. Compared with the mortality patterns of the general population, however, the gas cohort had 39 observed deaths from respiratory cancer compared with 26.6 expected. This statistically significant excess is even more impressive if one considers that the gas cohort had a deficit of smokers. Significant lung cancer excess was not observed for the pneumonia or the wounded cohorts. Beebe concluded that, since there was no substantial difference in lung cancer mortality between sociologically similar groups of soldiers, yet there was a substantial difference with the general population, this study is "quite suggestive" but "unproved."

Norman (1975) followed Beebe's cohorts for an additional 10 years. The mortality ratios for the mustard-exposed, pneumonia, and wounded populations were 0.99, 0.93, and 0.92, respectively, which may reflect an excess in that soldiers must pass a screening examination before induction. The mortality ratio of the gas-exposed cohort compared to wounded controls was 1.3, with 95 percent confidence intervals extending from 0.9 to 1.9. With "a more sensitive test," this difference in lung cancer deaths in the mustard-exposed cohort was even higher. A further case control analysis, for which little methodological detail was provided, indicated that mustard exposure added little excess risk compared with cigarette smoking. The authors recognized that this comparison might not accurately reflect the cumulative amount of cigarette smoking in each group; indeed, they provided data that smokers who were gassed quit substantially earlier than those in the other two groups. This study provides some evidence supporting the conclusion that a battlefield exposure associated with hospitalization is later associated with an excess mortality from lung cancer. This study also demonstrates the degree of information that would be essential to disentangle unequivocally the effect of mustard exposure from cigarette smoking.

In the early 1970s, reports began appearing of acute nonlymphocytic leukemia (ANL) in patients with multiple myeloma and lymphoma who had been treated with nitrogen mustards (Kyle et al., 1970; Rosner and Grunwald, 1975). Initially, it was thought that this occurrence might represent a consequence of the late natural history of these diseases. However, ANL was also found to occur in other types of cancer patients who had been treated with nitrogen mustards and other alkylating agents (Einhorn, 1978; Fisher et al., 1985; Greene et al., 1982). In particular, patients with ovarian cancer treated with melphalan and chlorambucil developed this complication. The reason why the initial