Thiazolidinediones have gained widespread use for the treatment of type 2 diabetes mellitus and other insulin resistance states, including polycystic ovary syndrome (PCOS). In thiazolidinedione-treated patients a small reduction in hemoglobin and hematocrit levels often is observed, and this generally has been attributed to fluid retention. Because testosterone is a hematopoietic hormone, we investigated whether a reduction in plasma free testosterone concentration was associated with the decrease in hemoglobin and hematocrit levels in 22 nondiabetic women (9 with normal glucose tolerance and 13 with impaired glucose tolerance; mean age, 29 +/- 5 years; mean body mass index, 35.6 +/- 5.8 kg/m2) with PCOS who were treated with pioglitazone, 45 mg/d. Before treatment and after 4 months, subjects underwent an oral glucose tolerance test and measurement of total body water content with bioimpedance. Plasma testosterone, androstenedione, dehydroepiandrosterone sulfate, hemoglobin, and hematocrit levels were evaluated at baseline and every month for 4 months. The fasting plasma glucose concentration (98 +/- 9 mg/dL) was unchanged after pioglitazone treatment, whereas the 2-hour plasma glucose concentration declined from 146 +/- 41 to 119 +/- 20 mg/dL (P = .002). Both the free androgen index and the free testosterone levels calculated according to Vermeulen et al decreased significantly (from 14.4 +/- 7.1 to 10.6 +/- 7.8 [P = .02] and from 59.4 +/- 23.4 to 46.6 +/- 23.3 [P = .03], respectively). The plasma androstenedione level declined from 259 +/- 134 to 190 +/- 109 ng/dL (P = .01), whereas the dehydroepiandrosterone sulfate level did not change significantly (from 139 +/- 90 to 127 +/- 84 mug/dL, P = .2 [not significant]). The levels of both hemoglobin (from 13.6 +/- 1.0 to 12.8 +/- 1.1 g/dL, P = .0002) and hematocrit (from 39.7% +/- 2.2% to 37.9% +/- 2.7%, P = .002) fell slightly after 4 months of pioglitazone administration. Collectively, before and after pioglitazone administration, the plasma free testosterone level according to Vermeulen et al correlated positively with the levels of hemoglobin (r = 0.49, P < .0001) and hematocrit (r = 0.40, P < .0001), as well as the free androgen index (r = 0.38 [P < .0003] with hemoglobin and r = 0.29 [P < .006] with hematocrit); the decrement in plasma free testosterone level and free androgen index also correlated with the decrements in the levels of both hemoglobin (r = 0.51 [P = .01] and r = 0.54 [P = .01], respectively) and hematocrit (r = 0.42 [P = .05] and r = 0.50 [P = .02], respectively). Body weight increased from 90.5 +/- 17.3 to 92.4 +/- 18.8 kg after pioglitazone administration (P = .05), as did body fat content (from 42.7 +/- 15.3 to 44.8 +/- 17.1 kg, P = .03), which could explain the increase in weight, because edema did not develop in any of the subjects. Total body water content did not change significantly after pioglitazone administration (from 37.7 +/- 5.0 to 37.8 +/- 4.9 L, P = .68 [not significant]). In summary, pioglitazone treatment is associated with a mild decline in hematocrit or hemoglobin level, which is correlated with the reduction in plasma testosterone level. These results suggest that increased body water content cannot explain the reduction in hematocrit or hemoglobin level in women with PCOS. Further studies are necessary to evaluate whether the same scenario is applicable to normoandrogenic women and individuals with type 2 diabetes mellitus.