The impact of statistical choice on variability and error in maximal respiratory mouth pressure testing

Various statistics have been used to assess repeatability and reproducibility of maximal inspiratory and expiratory mouth pressure measurements (PImax and PEmax, respectively). Few studies have examined the impact of rater experience on these values. The aims were to assess: (1) how choice of statistic impacted repeatability of PImax and PEmax; and (2) the impact of rater experience on reproducibility of PImax and PEmax. Thirty-one healthy adult males completed one familiarisation and two experimental visits. The PImax and PEmax were assessed using a hand-held respiratory pressure meter by three raters with different testing experience. The intraclass correlation coefficient, coefficient of variation, standard error of measurement, smallest detectable change and limits of agreement (LoA) were determined across experimental visits for a given rater and between raters to determine repeatability and reproducibility of PImax and PEmax. For PImax and PEmax, the intraclass correlation coefficient was ≥0.97 for reproducibility and 0.89-0.95 repeatability; coefficient of variation was ≤4% for reproducibility and ≤6% for repeatability; standard error of measurement was 6-9 cmH₂O for reproducibility and 8-15 cmH₂O for repeatability; smallest detectable change was 16-26 cmH₂O for reproducibility and 22-40 cmH₂O for repeatability; LoA ratio bias was -2% to 5% for reproducibility and -4% to 2% for repeatability; and LoA ratio error was 9%-17% for reproducibility and 19%-25% for repeatability. If test administrators were trained, rater experience did not impact replicate values. However, choice of statistic impacted interpretation of PImax and PEmax variability, and measurement error and reproducibility were better than repeatability.