From the author:
- S. E. Rees
- S. E. Rees, Center for Model-based Medical Decision Support, Aalborg University, Niels Jernes Vej 144-313, DK 9220 Aalborg East, Aalborg, Denmark. E-mail: sr{at}hst.aau.dk
In his letter, G.S. Zavorsky provides a summary of the results of the study of Rees et al. 1 and two criticisms, the first related to the correctness of the statistical methods and the second related to the novelty of the results.
In summarising the results, he incorrectly states that the study of Rees et al. 1 compares arterial and peripheral values of oxygen tension, carbon dioxide tension and pH. In fact this 1 and a similar previous 2 study evaluate a method for calculating arterial gas tension values from peripheral venous blood 3, the comparisons performed being between measured and calculated values, as illustrated in figure 2 of our manuscript 1.
In response to G.S. Zavorsky's criticism concerning statistical methods, it is important to understand that the standard statistical method for comparing two measures of the same variable is Bland–Altman analysis 4, and it is this which had been used to compare calculated and measured values 1, 2. For the critics of this method 5, standard regression analysis was also included in the article by Rees et al. 1. In the Bland–Altman method, one uses the mean difference between the variables as a measure of accuracy, i.e. the systematic variation, and the standard deviation of the difference as a measure of the precision, i.e. the random variation. Therefore, if one subject has a measured arterial oxygen tension (Pa,O2) 1 kPa higher than the calculated arterial oxygen tension (Pca,O2) and another subject has a Pa,O2 1 kPa less than Pca,O2, the mean difference indeed is zero, and this statistic is a useful representation of little systematic variation in the data, with random variation being characterised by the standard deviation of the difference. Calculation of the absolute mean difference, as proposed by G.S. Zavorsky, is not helpful as it would not effectively separate systematic and random variation.
G.S. Zavorsky comments that the results of the study 1 are of no surprise, relating his comment to his recent meta-analysis comparing arterial and capillary blood analysis 6. In many regards it is satisfying that the results are not surprising, as this indicates that mathematical arterialisation is as good as mechanical arterialisation. In short, the capillary blood samples included in the meta-analysis have been mechanically arterialised 6. This involves either administration of local vasodilation cream to, or warming of, the sampling site. Typically, there is a need to wait 10 min after application of the cream or for the sampling site to reach the correct temperature 7, 8, a procedure which has lead to development of special devices 9 and may indeed make clinical application rather cumbersome. In contrast, the previously presented 3 and evaluated 1, 2 method is a mathematical arterialisation where peripheral venous blood sampled in normal clinical practice is mathematically transformed into arterial values. For patients presenting in departments of emergency or pulmonary medicine, peripheral venous punctures are usually made for other blood samples, meaning that this technique could be readily integrated into current practice. In addition, the mathematical algorithm could be built into commercially available blood gas analysers, with the only extra overhead of the method being measurement of arterial oxygen saturation with a pulse oximeter (Sp,O2), which is required as input to the algorithm.
The need for Sp,O2 as input to the algorithm can be seen as beneficial, in that it eliminates the conundrum postulated by G.S. Zavorsky. As shown by Rees et al. 1, useful values of Pa,O2 can be calculated in all situations in which Sp,O2 is ≤96%. In patients with Sp,O2 values ≥97%, the method can still effectively be used to calculate arterial carbon dioxide tension and pH, and it is safe to assume a clinically adequate Pa,O2, despite the large errors in Sp,O2 readings. While large scale studies are required, G.S. Zavorsky's recommendation against using venous blood to predict Pa,O2 levels does not, therefore, appear justified, at least for the case of peripheral venous blood values transformed using mathematical arterialisation.
Statement of interest
A statement of interest for S.E. Rees can be found at www.erj.ersjournals.com/misc/statements.dtl
- © ERS Journals Ltd