1. Himon Table 7 Association between HCL and benzene exposure as a function of temporal exposure characteristics;Clavel, Conso; Limasset, Mandereau; Roche, Flandyin
2. There was no evidence of an association with any of the temporal variables, time since first exposure, time since the end of exposure, and duration of exposure (table 7). Moreover, if only those exposures occurring at least five years before diagnosis are considered, the ORs remain close to unity;(OR; for benzene exposure irrespective of exposure source or level
3. The data available to the experts were insufficient to enable assessment of the exposure of painters before the dearomatisation of the mineral spirits used in the paint trade. The analysis in table 6 was thus repeated with exposure values 10-then 100-fold greater than the initial levels for painters. The estimated ORs remained close to 1. No association or increasing trend was detected, although subject distribution as a function of exposure level varied substantially from estimation to estimation
4. Restriction of the analysis to non-smokers and adjustment on the pack-years variable (rather than on smoker, ex-smoker, or nonsmoker) yielded similar results. Restriction of subjects to those never having farmed, or to those from regions in which little agriculture is practised had no impact on the results. Similarly, no change in the results was found when the population was restricted to those taken into hospital after 1984
5. Due to the rarity of the disease, retrospective recruitment was conducted. For practical reasons, the study only examined living cases. If benzene exposure is related to duration of survival, this design could miss an effect. However, the results were unchanged when the analysis was restricted to pairs in which the cases were diagnosed after 1984, when survival was threefold higher because interferon-a became available for therapeutic use. Replacement of Hairy cell leukemia and occupational exposure to benzene non-responding controls by other eligible con-We are also indebted to Dr Victoria Mora, occupational trols did not seem to induce a bias as estimates were very similar when analysis was restricted health physician, who conducted most of the specific interviews. We are grateful to AG Mullarky for his skillful translation of the manuscript