Carbon dioxide analysers: accuracy, alarm limits and effects of interfering gases.

Six mainstream and twelve sidestream infrared carbon dioxide (CO2) analysers were tested for accuracy of the CO2 display value, alarm activation and the effects of nitrous oxide (N2O), oxygen (O2) and water vapour according to the ISO Draft International Standard (DIS)#9918. Mainstream analysers (M-type): Novametrix Capnogard 1265; Hewlett Packard HP M1166A (CO2-module HP M1016A); Datascope Passport; Marquette Tramscope 12; Nellcor Ultra Cap N-6000; Hellige Vicom-sm SMU 611/612 ETC. Sidestream analysers: Brüel & Kjaer Type 1304; Datex Capnomac II; Marquette MGA-AS; Datascope Multinex; Ohmeda 4700 OxiCap (all type S1: respiratory cycles not demanded); Biochem BCI 9000; Bruker BCI 9100; Dräger Capnodig and PM 8020; Criticare Poet II; Hellige Vicom-sm SMU 611/612 A-GAS (all type S2: respiratory cycles demanded). The investigations were performed with premixed test gases (2.5, 5, 10 vol%, error < or = 1% rel.). Humidification (37 degrees C) of gases were generated by a Dräger Aquapor. Respiratory cycles were simulated by manually activated valves. All monitors complied with the tolerated accuracy bias in CO2 reading (< or = 12% or 4 mmHg of actual test gas value) for wet and dry test gases at all concentrations, except that the Marquette MGA-AS exceeded this accuracy limit with wet gases at 5 and 10 vol% CO2. Water condensed in the metal airway adapter of the HP M1166A at 37 degrees C gas temperature but not at 30 degrees C. The Servomex 2500 (nonclinical reference monitor), Passport (M-type), Multinex (S1-type) and Poet II (S2-type) showed the least bias for dry and wet gases. Nitrous oxide and O2 had practically no effect on the Capnodig and the errors in the others were max. 3.4 mmHg, still within the tolerated bias in the DIS (same as above). The difference between the display reading at alarm activation and the set point was in all monitors (except in the Capnodig: bias 1.75 mmHg at 5 vol% CO2) below the tolerated limit of the DIS (difference < or = 0.2 vol%). The authors conclude that the tested monitors are safe for clinical used (except those failing the DIS limits). The accuracy of the CO2-reading (average of mean absolute bias) is better in the M-type than in the S1- or S2-type analysers although no statistical (nor clinical) significant differences could be detected. Most manufacturers work with stricter limits than those proposed by the DIS.