Performance characteristics of gas analysis systems: what we know and what we need to know.

It is important that sources of variation in gas analysis measurements are identified and described in an accurate and informative manner. In this paper, we discussed the potential sources of error, which should be considered in any measurement study on gas analysis systems. We then covered how errors in various terms associated with gas laws propagate to outcome measurements of gas exchange to help quantify the relative importance of sources of error. Finally, we performed a literature survey to explore the statistical methods researchers have employed to arrive at conclusions on the performance characteristics of gas analysis methods. We found examples of excellent practice in the literature, but there were also gaps in the knowledge of error in gas analysis systems. Consequently, we supplied guidelines for future method comparison studies. These guidelines included (i) a sample size of at least 40 participants and the citation of confidence intervals, (ii) a description of the relationships between systematic and random errors and the size of measured value, (iii) the parallel examination of test-retest error within a method comparison study, and (iv) an a priori-made judgement on how much systematic and random error between methods is acceptable for practical applications. We stressed that this judgement should be based on expert-agreed position statements about acceptable error, which unfortunately have yet to be formulated for gas analysis systems.