A multicenter study used 756 samples from 251 patients in 12 institutions to compare arterial (PaO2, PaCO2) with transcutaneous (PsO2, PsCO2) oxygen and carbon dioxide tensions, measured usually at 44 degrees C. Of these samples, 336 were obtained from 116 neonates, 27 from 25 children with cystic fibrosis, and 140 from 40 patients under general anesthesia. Ninety-one patients were between 4 weeks and 18 years of age, 32 were between 18 and 60 years, and 12 were over 60. The ratio of transcutaneous to arterial P(s/a)CO2 was 1.01 +/- 0.11 with PaCO2 less than 30 mm Hg, increasing to 1.04 +/- 0.08 at PaCO2 greater than 40 mm Hg. Mean bias and its standard deviation (PsCO2 - PaCO2) were + 1.3 +/- 3.9 mm Hg in the entire group, + 1.8 +/- 4.2 mm Hg in neonates (NS). Bias was + 0.2 +/- 2.7 mm Hg when PaCO2 was less than 30 mm Hg (N = 175, NS), 1.0 +/- 3.4 with 30 less than PaCO2 less than 40 (n = 329, p less than 0.001), and + 2.04 +/- 4.00 mm Hg with 40 less than PaCO2 less than 70 (n = 229, p less than 0.001). These data suggest that, using transcutaneous PCO2 monitors with inbuilt temperature correction of 4.5%/degrees C, the skin metabolic offset should be set to 6 mm Hg. The linear regression was PsCO2 = 1.052(PaCO2) - 0.56, Sy.x = 3.92, R = 0.929 (n = 756); and PsCO2 = 1.09(PaCO2) - 1.57, Sy.x = 4.17, R = 0.928 in neonates (n = 336).(ABSTRACT TRUNCATED AT 250 WORDS)