• Grace T S Law, C Y Wong, C W Kwan, K Y Wong, F P Wong, and H N Tse.
• Department of Medicine and Geriatrics, Kwong Wah Hospital, 25 Waterloo Road, Kowloon, Hong Kong.
• Hong Kong Med J. 2009 Dec 1;15(6):440-6.

ObjectiveTo explore the correlation and concordance between end-tidal carbon dioxide and arterial carbon dioxide partial pressure, and confirm the experience of the general consensus among service environments.DesignA prospective cross-sectional analysis.SettingTwo respiratory service units in Hong Kong.ParticipantsTwo hundred respiratory patients were recruited, in whom 219 sets of observations were recorded. Patients deemed to require arterial blood gas determination also had their end-tidal carbon dioxide partial pressure measured at that time, using two LifeSense LS1-9R Capnometers.Main Outcome MeasuresThe agreement of end-tidal carbon dioxide partial pressure and arterial carbon dioxide partial pressure was studied by correlation coefficients, mean and standard deviation of their difference, and the Bland-Altman plot.ResultsOverall, the correlation was low and insignificant (r=0.1185, P=0.0801). The mean of the difference was 7.2 torr (95% confidence interval, 5.5-8.9) and significant (P<0.001). The limits of agreement by Bland-Altman analysis were -18.1 to 32.5 torr, which were too large to be acceptable. In the sub-group on room air, the mean difference was reduced to 2.26 torr, the correlation between end-tidal carbon dioxide partial pressure and arterial carbon dioxide partial pressure was 0.2194 (P=0.0068), though statistically significant, the extent of correlation was still low.ConclusionEnd-tidal carbon dioxide partial pressure did not show significant correlation or concordance with arterial carbon dioxide partial pressure, especially when supplemental oxygen was used. End-tidal carbon dioxide partial pressure currently cannot replace arterial blood gas measurement as a tool for monitoring arterial carbon dioxide partial pressure. Possible reasons for the discrepancy with previous studies include small sample size in previous studies, lack of research facilities in service settings, and publication bias against negative studies.

20 keywords...

hide…