Fetal blood sampling by cordocentesis was performed in 85 cases of fetuses with either nonimmune hydrops fetalis, anomalies, intrauterine growth retardation, maternal immunological disorders, Rh incompatibility, hydronephrosis, polyhydramnios, maternal preeclampsia or fetal tumor. Fetal heart rate monitoring was also recorded before cordocentesis. We evaluated the usefulness of the nonstress test (NST) with regard to fetal hematologic gas profiles including pH, pO2 and pCO2. When the acceleration was classified into reactive and nonreactive group, pH and pO2 were significantly higher in the reactive group than in the nonreactive group. pCO2 was significantly lower in the former than the latter. However, the sensitivity and specificity of the reactive NST predicting fetal umbilical venous pO2 over 20 mm Hg was 79.7 and 68.7%, respectively. The false-positive rate of nonreactive NST predicting hypoxemia below 20 mm Hg pO2 was 44%. As a consequence, we cannot always rely on the NST, and should perform fetal blood sampling to assess fetal condition for further management.