BACKGROUND Endtidal (ET) measurement of carbon dioxide is well established for intraoperative respiratory monitoring of adults and children, but the method's accuracy for intraoperative use in small newborn infants has been less extensively investigated. OBJECTIVE The aim of this study was to compare carbon dioxide from ET measurements with arterialized capillary blood samples in newborn infants during general anesthesia and surgery. METHODS Endtidal carbon dioxide was continuously measured during anesthesia and surgery and compared with simultaneous blood gas analyses obtained from capillary blood samples. Fifty-nine sample sets of ET to blood gas carbon dioxide were obtained from 23 prospectively enrolled infants with a gestational age of 23-41 weeks and a birth weight of 670-4110 g. RESULTS Endtidal levels of carbon dioxide were considerably lower in all sample sets and only 4/23 individual ET-blood gas sample pairs differed <7.5 mm Hg (1 kPa). Bland-Altman analysis indicated a poor agreement with a bias of -13 ± 7 mm Hg and a precision of ±14 mm Hg. The performance of ET measurements was particularly poor in infants weighing below 2.5 kg, in infants in need of respiratory support prior to anesthesia, and when the true (blood gas) carbon dioxide level was high, above 45 mm Hg. CONCLUSIONS Main-stream capnography during anesthesia and surgery correlated poorly to blood gas values in small and/or respiratory compromised infants. We conclude that caution should be exercised when relying solely on ET measurements to guide mechanical ventilation in the OR.