Errors in respiratory gas analysis

A series of experiments were carried out in which the consistencies and relative biases of Haldane gas analysers and Pauling oxygen analysers, when used for determining the oxygen and carbon dioxide contents of expired air, have been evaluated. The Pauling analyser, modified to eliminate instability of the potentiometer base line, is more consistent than the Haldane analyser for determining oxygen intake and metabolic rate but not respiratory quotient. Relative biases between the two types of analyser and between two analysers of the same type have negligible effects on the determination of oxygen intake and metabolic rate but could affect the respiratory quotient. The commonly accepted limit of about 0.05% for the difference between duplicate Haldane analyses of the contents of expired air, is unrealistically narrow and likely to produce biased results. A procedure for setting realistic limits, which also provides a check on operators' techniques is described.