Noninvasive Monitoring of Carbon Dioxide

To the Editor: We were interested to read the article by Tobias and Meyer [1], in which two noninvasive monitoring techniques of carbon dioxide (endtidal versus transcutaneous [TC]) were compared in toddlers and infants. Although we agree with their conclusions, we would like to comment on the statistical analysis, especially the Bland-Altman method. To compare two techniques of clinical measurement, the Bland-Altman analysis is a graphical method that is more appropriate than the correlation coefficient [2]. It plots the difference between the two methods against their mean and provides a bias (mean difference) and its standard deviation (SD). If the distribution is normal, 95% of the difference will lie between (bias + 1.96 SD) and (bias - 1.96 SD) (rounded to bias +/- 2 SD most of the time). These limits are defined as the limits of agreement. In their paper, Tobias and Meyer performed two Bland-Altman analyses (TC-CO2 versus PaCO2 and ETCO2 versus PaCO2). These analyses do not strictly follow the recommendations given by Bland and Altman [2]. First, the difference is plotted against the standard method (PaCO2) and not against the average of the standard and the new measurements ([TC-CO2 + PaCO2]/2 and [ETCO2 + PaCO2]/2). This, in fact, can be misleading, as proved in another paper by Bland and Altman [3]. Indeed, this will seem to show a relation between difference and magnitude (agreement varies with the size of the measurement) when there is none. A plot of the difference against the average does not mislead in this way. Second, Tobias and Meyer use "precision" (1 SD) instead of agreement (2 SD). Precision is an inappropriate term because only 68% of the difference lies into the interval (bias + SD, bias - SD). Graphically, it gives a narrower interval around the bias. Consequently, the limits of agreement in the Tobias and Meyer study are (4.02, -5.38) for TC-CO2 versus PaCO2 and (3.34, -16.7) for ETCO2 versus PaCO2. Nevertheless, the practical conclusions of their paper (i.e., TC-CO2 monitoring provides a more accurate estimation of PaCO2 than ETCO (2) monitoring) are unchanged because precision was the tool used in both graphs. Bruno Grenier, MD Marc Dubreuil, MD Department of Anesthesiology; Hopital Pellegrin; 33076 Bordeaux Cedex, France