Noninvasive monitors of blood pressure in the critically ill: what are acceptable accuracy and precision?

Abstract

Editor, We read with interest the study by Drs Ilies et al.,1 comparing blood pressure (BP) measurements by the noninvasive CNAP technology (for Continuous Noninvasive Arterial Pressure) with intra-arterial measurements, in the aftermath of cardiovascular surgery. The authors analysed more than 11 000 measurements of BP collected in more than 100 patients and they should be congratulated for this effort. The authors concluded that, for mean BP (but not for systolic BP and only ‘partially’ for diastolic BP), CNAP measurements showed acceptable agreement with intra-arterial measurements and these two techniques were then considered interchangeable. In this work, the authors chose to define acceptability of CNAP measurements by two methods, first the calculation of the percentage error, as frequently used in the assessment of cardiac output monitors;2 and second the calculation of the agreement/tolerability interval (ATI) ratio. For this latter approach, the choice of quite a wide range of BP values supposedly not requiring a therapeutic intervention made the ATI ratio more often favourable and then contributed to consider CNAP measurements as acceptable.3 Apart from this quite complex approach to define acceptability, at a glance, interpretation of results of Ilies et al.1 may be quite different, that is interchangeability may be questioned. Indeed, the mean bias ± SD, that is, the mean difference between CNAP and intra-arterial measurements of mean BP (i.e. the accuracy) ± the random fluctuations around this mean (i.e. the precision) were 6.1 ± 7.6 mmHg.1 In other words, overall, CNAP measurements overestimated the mean BP by 6.1 mmHg. Furthermore, the limits of agreement (including 95% of the pairs of measurements) were wide: intra-arterial BP could be overestimated of as much as 21 mmHg or underestimated of up to 9 mmHg.3 This overestimation or underestimation of the mean BP appeared quite unpredictable. The agreement was even worse for systolic BP and diastolic BP measurements. As decision-making in the critically ill may require more accurate and precise measurements of BP,4 the interpretation of CNAP performance may differ from that of the authors. This point is of utmost importance: how to define clinical acceptability of a new BP monitor? In other works assessing continuous noninvasive BP monitoring devices, definitions of acceptability vary markedly from one study to another, yielding conflicting conclusions about the performance of the technology under test.5 Official guidelines currently exist for the evaluation of intermittent BP monitoring devices (oscillometric automated brachial cuff for instance),6 but reluctance to use these guidelines may arise, as they do not mention continuous noninvasive BP monitoring devices.6 However, continuous measurements of BP should be expected to be as reliable as intermittent measurements. Furthermore, these guidelines are dedicated to devices used in hospitalised patients as well as in outpatients. Nevertheless, in the critically ill, BP measurements should be even more accurate than in the outpatient. For this reason, the criteria of acceptability of international guidelines are sometimes deemed too liberal.4 In other words, the boundaries recommended in these guidelines should not be considered too stringent to be applied in the critically ill and should rather be seen as maximal limits of acceptability. Of note, these international guidelines have already been applied in several studies assessing continuous BP monitoring devices.5 Among these guidelines, the Association for the Advancement of Medical Instrumentation recently developed the ISO Standard 81060-2, adopted as the US National Standard.6 When applying the ISO criteria (the way they are frequently applied in operating room and ICU studies,5 i.e. maximum mean bias and SD of 5 and 8 mmHg, respectively,6 without taking into account intra-arterial measurements variability) to the studied population by Ilies et al.,1 CNAP measurements of BP (systolic, diastolic and even mean BP) did not pass the test. It is very likely that this was also the case when applying the International Protocol of the European Society of Hypertension.7 Of note, a recent meta-analysis found that, even in the operating room, the inaccuracy and imprecision of the CNAP technology are larger than what was defined as acceptable by international guidelines.5 Hence, the performance of the CNAP technology reported by Ilies et al.1 was qualified as acceptable according to the authors’ approach but may appear disappointing according to official standards.6 Furthermore, in some instances, one device may even fulfil the criteria of one standard but not those of another standard. Importantly, in the literature, the lack of consistency in the way acceptability of noninvasive BP monitoring devices is defined stresses the importance of adopting, in clinical studies in the critically ill, a consensual standard of acceptability of commercialised or future BP monitors in order to avoid conflicting interpretations of their performance. Acknowledgements relating to this article Assistance with the letter: none. Financial support and sponsorship: none. Conflicts of interest: none.