The purpose of this study was to analyze the quantitation of the human immunodeficiency virus type 1 RNA (HIV-1 RNA) in the genital tract of HIV-1-infected pregnant women and to evaluate a possible correlation with the viral load in blood plasma (Spearman's rank correlation coefficient). A total of 38 each of cervical, vaginal, and blood samples from 38 women were obtained during the third trimester of pregnancy for quantitation of the HIV-1 RNA load. Viral loads were determined by reverse transcription-polymerase chain reaction. The HIV-1 RNA viral load was detectable in 29 of the 38 (76.3%) blood samples, in 6 of the 38 (15.7%) cervical secretion samples, and in 8 of the 38 (21%) vaginal secretion samples. Overall, the correlation between the HIV-1 RNA viral load in the blood plasma and in cervical secretion samples was 0.51 ( P<0.001). However, the correlation disappeared ( r=0.27) when three patients with high blood plasma viral loads were eliminated from the statistical study. The viral load in the vaginal secretions did not correlate with that in the blood samples ( r=0.26). There were two cases in which HIV-1 RNA was undetectable in the blood and cervix but was detectable in vaginal secretions: one woman had 220 copies/ml and the other 68 copies/ml. These results suggest that pregnant women with undetectable viral loads in blood plasma are still at risk of transmitting the virus vertically during vaginal delivery. Because of this, antiretroviral prophylaxis during vaginal delivery must be administered to HIV-1-infected women and their newborns, regardless of the mother's viral load in plasma. In conclusion, quantification of cervicovaginal levels of HIV-1 may represent a useful tool for assessing the individual risk associated with a vaginal delivery and for guiding decisions about whether a scheduled caesarean should be recommended.