With the advent of pulse oximetry, there has been a general decrease in the use of transcutaneous (Tc) blood gas monitoring in intensive care environments. The available data, however, suggest that arterial carbon dioxide pressure (PCO2) levels are best estimated by Tc methods. In this study, we report our experience using routine Tc PCO2 monitoring in 32 consecutive infants less than 2 weeks of age with birthweights less than 1500 g. A total of 644 simultaneous pairs (Tc PCO2 versus arterial PCO2) were obtained. Pairs were categorized according to a 2 x 2 matrix design based on sensor temperature (40 degrees or 43 degrees C) versus site of arterial sampling (umbilical [UAC] or peripheral artery catheter [PAC]). Sampling via the UAC resulted in excellent correlation between sample pairs at both sensor temperatures with similar regressions between groups. Sampling via the PAC, however, yielded poor correlation between sample pairs and a significantly different regression from both UAC groups. Based on these findings, we advocate the use of a sensor temperature of 40 degrees C in very low birthweight infants for tracking Tc PCO2 values. In addition, we suggest that inaccuracies in PAC sampling may lead to erroneous PCO2 determinations. We conclude that routine monitoring of Tc PCO2 is accurate and serves a useful and continuing role in the neonatal intensive care environment.