Background Two non-invasive techniques are available for monitoring carbon dioxide (CO₂) in neonatal intensive care unit (NICU) patients. Transcutaneous monitoring (TcCO₂) is widely used, whereas the application of end-tidal CO₂ (capnography) has been limited in newborn infants, due to accuracy problems.

Objectives To determine in preterm infants the accuracy of end-tidal CO₂ (EtCO₂) monitoring, and to assess the influence of alveolar plateau duration; to compare EtCO₂ and TcCO₂ for detecting low and high values of PCO₂.

Materials and Methods In 38 preterm infants (birth weight <1500 g) undergoing mechanical ventilation, recordings of EtCO₂ (capnography using sidestream device: Capnostream) were performed. Results were compared with those of CO₂ venous (umbilical venous catheter) pressure measured from one blood sample (PvCO₂), and with simultaneous TcCO₂ recordings. In EtCO₂/PvCO₂ measurement pairs, the bias (EtCO₂/PvCO₂) and the intraclass correlation coefficient (ICC) were calculated. EtCO₂ and TcCO₂ techniques were compared for the detection of low (PvCO₂ <45 mm Hg) and high PCO₂ (PvCO₂ >60 mm Hg), using receiver operator characteristic curves (area under the curve, AUC).

Results 104 EtCO₂/PvCO₂ pairs obtained from 38 patients were analysed. The bias was −1 ± 7.8 mm Hg. It decreased when the alveolar plateau increased. The ICC was 0.28. AUC for EtCO₂, compared with TcCO₂, were: 0.82 versus 0.89 for high PCO₂ (p = 0.18) and 0.89 versus 0.90 for low PCO₂ (p = 0.71), respectively.

Conclusion In ventilated preterm infants, the correlation between EtCO₂ and PvCO₂ was low. Capnography may, however, be useful in the monitoring of CO₂ trends.