Effect of benzene exposure on (A) white blood cell (WBC) and granulocyte counts; (B) colonies from the colony-forming unit–granulocyte-macrophage (CFU-GM) and burst-forming unit–erythroid (BFU-E); and (C) colonies from the colony-forming unit–granulocyte, erythroid, macrophage, megakaryocyte (CFU-GEMM). Erythropoietin (EPO) was added to half of the cultures (9). Trends with benzene were tested by linear regression (WBC, granulocytes), negative binomial regression (CFU-GM, BFU-E), and unconditional logistic regression [CFU-GEMM, categorizing subjects into 0 or more than 0 colonies (92%, 74%, and 40% of subjects had >0 colonies among controls, <10 ppm, and ≥10 ppm, respectively)]. Models were adjusted for age and sex, and additionally for smoking, alcohol, recent infections, and body mass index (BMI) if significant (9). Strong, inverse trends between benzene and all cell types were present (P_trend shown). There was a greater proportional decrease in colonies in workers exposed to ≥10 ppm versus controls for CFU-GM, BFU-E, and CFU-GEMM compared to the decline in WBCs (P < 0.011, 0.048, and 0.0078, respectively) and for CFU-GM and -GEMM compared to the decline in granulocytes (P = 0.026 and 0.0094, respectively) (9).