Cameron, James D.
In this issue of the Journal of Hypertension two separate groups of investigators report on the precision and accuracy of new devices designed to noninvasively assess central SBP (cSBP). Both groups (Pucci et al. and Ding et al.) present carefully performed and thoroughly analyzed work comparing the effects of the different inherent approaches of the specific devices for the estimation of central blood pressure (BP). They also report on the impact of the different methods of calibration that can be used in these noninvasive devices for estimating cSBP. Whereas most scrutiny and discussion is usually, and understandably, of the inherent ‘accuracy’ of these new devices, it is also important that they be considered in the broader context of what a user might expect of a particular device if it is used clinically. This type of wider discussion is in general lacking and is often hampered by lack of suitable and appropriate basic data – the work presented in this issue by the two groups is a welcome addition and provides very relevant performance data on three devices proposed for use in estimating cSBP [SpygmoCor (AtCor Medical, Sydney, Australia), Vicorder (Skidmore Medical, Bristol, UK) and Omron HEM 9000AI (Omron Healthcare, Kyoto, Japan)]. The SphygmoCor device represents well known early technology for this application with the other two being relative newcomers.
It is important to divorce issues around accuracy from those around applicability and clinical or research usefulness. Full consideration of clinical usefulness is outside the topic of this commentary and whereas the jury is still out on the additional advantage of general assessment of central BP, the applicability of any clinical measurement must be assessed in the context of patient benefit.
The ever increasing ease of constructing novel systems for noninvasive assessment of human central haemodynamics in compact computing environments has led to the burgeoning availability of new devices that report parameters associated with central BP (e.g. [3–10]) (or of aortic stiffness and mechanics, to which similar comments apply). The multiplicity and ubiquity of these devices run the risk of convincing the casual observer, just by continual exposure, that they have become part of a standard armamentarium in cardiovascular assessment and therefore must, by virtue of availability, be well authenticated and clinically useful. Without device-specific data this may be an unwarranted assumption.
It is highly likely that the busy clinician or researcher, and especially the occasional user, will take for granted that parameters reported under the same name, albeit by different technologies, must be interchangeable. This later point is perhaps not unreasonable on cursory consideration, but studies such as that reported by Pucci et al. and Ding et al. in this issue of Journal of Hypertension clearly support the likelihood that as many different estimated values of cSBP or other descriptive central BP parameters can be obtained as there are devices designed to obtain them. It is also clear that because of different inherent technologies and algorithms and differing means of calibration these values will not in general be interchangeable across devices .
A major issue often forgotten or ignored, especially by early adopters, is that in fact all proprietary devices and approaches to assessment of central haemodynamics are different and, importantly, that newer devices should not necessarily be accepted as ‘substantially equivalent’ to existing or better known devices. Both studies [1,2] highlight the fact that results from different devices, even if given the same name (i.e. ‘central BP’), are not interchangeable and therefore results obtained by one practitioner using a given device and relating to a given patient are not transportable with that patient if they should subsequently attend a different practitioner using an alternate favourite device for assessing the same parameter. Both studies [1,2] (reasonably on their results) conclude that specific ranges of ‘normal values’ will be required for different devices if noninvasively obtained central pressure parameters are to be employed in practice. This device dependence is potentially a severe limitation for applicability of all devices; if results are not interchangeable this will be a point of confusion for patients and physicians.
Figure 1 shows modelled variance over a physiological pressure range between the Sphygmocor and Omron devices using the regression equations given in Ding et al. for device minus catheter values of the three central BP parameters studied. Absolute concordance with catheter measurements aside the devices are clearly not equivalent in numerical result as demonstrated by a reasonably constant (non-zero) difference in the estimation of the dimensionless central-to-brachial pulse pressure (PP) ratio between the two devices and an obvious pressure-dependent variation in central-to-brachial PP difference and central-to-brachial SBP difference between the devices at the same brachial BP.
A further issue, worthy of increased critical consideration, is what should be the basic (gold) standard in comparison studies on the accuracy and applicability of noninvasive assessment of central BP? Should it be the obvious one of simultaneously measured invasive pressures (presumably with solid-state transducers) or is there an acceptable noninvasive standard of comparison and if so should the usual British Hypertension Society/American Association for the Advancement of Medical Instrumentation (AAMI) standards apply? By default, and by ease of application, the more well known SphygmoCor device is being used for comparison in putative validation studies without real justification [1–3,5,6,9]. With the rapid expansion of alternative devices and a wider knowledge of the concept of central BP there is a tendency to be more forgiving in applying rigorous assessment to newer devices (the ‘if it reports a value it must be correct’ approach) and thus newer devices have tended to avoid the considerable attention and investigation that is accorded to devices, such as SphygmoCor, that appears first in the market. This is in spite of un-allayed concerns regarding the transfer function approach .
The concept of ‘substantial equivalence’ or acceptance based on agreement with better known existing devices is not without risk in this field. Whereas noninvasive comparisons are useful and certainly easier to perform, it should perhaps be mandatory that all new entrants to the market perform at least fundamental invasive comparison. Sequential comparison against previous noninvasive devices will undoubtedly, with the passage of time, suffer from a ‘creep’ phenomenon as new devices run the risk of gradually varying further and further from earlier devices and an original invasive comparison.
Both groups of authors in the present edition of the Journal present and discuss their results in terms of comparisons, one between SphygmoCor and Vicorder but also including an invasive subgroup , the other using SphygmoCor and Omron HEM 9000AI including a central and brachial invasive comparison . An established confounder of comparison studies and indeed of the technology in general is the significant effect that the method of calibration has on derived noninvasive results . Inaccuracies related to the noninvasive assessment of the brachial cuff BP used in calibration of noninvasive devices are well known to have a substantial influence on the generated value of the derived central pressure, and this has been further demonstrated and discussed by both groups [1,2], as well as previously by others [12,13]. It would now seem clear that issues around using brachial SBP to calibrate radial pressure waveforms is a substantial cause of underestimation of cSBP.
The most relevant issue, however, may not be the absolute level of agreement of the various approaches or even how accurate they are in estimating group cSBP or other characteristics. What may turn out to be of most significance is clear establishment of the use to which any estimated values can appropriately be put and the ongoing development of knowledge in the medical community as to their meaning and applicability. For instance, it would appear from the data of Pucci et al. that in their invasively measured sub-population, if noninvasive techniques are acceptably accurate, then there is no significant peripheral amplification of BP (invasive central 141 ± 19 mmHg versus noninvasive brachial 143 ± 17 mmHg). In fact, the noninvasive, directly measured brachial BP (143 mmHg; Table 1)  seems just as good an estimate of the invasive cSBP in this cohort as any noninvasively calibrated central estimate (Fig. 2). On this basis it might be reasonable to ask why bother with any device at all, highlighting the type of issue often not considered when judging only concordance.
In the study of Ding et al. (their Table 2) it seems that the Omron device does not provide estimates of central BP significantly different from brachial values or central-to-peripheral BP ratio much different from unity (i.e. 1.6 ± 8.1, 1.6 ± 8.1, 1.06 ± 0.17 for SBP difference, PP difference, ratio, respectively). This again suggests that in some groups cSBP estimation may not add much additional useful information. It is also interesting to note that for both devices used in comparison by Ding et al. the difference between noninvasively determined central-to-brachial SBP appears essentially constant irrespective of true difference (as determined invasively), seemingly around 14 mmHg for SphygmoCor and of the order 1–2 mmHg for Omron (Fig. 1, Table 2) . It is also clear from the Bland–Altman plots (Fig. 2)  that the errors in noninvasive estimation are proportional to the mean of the values. In this context trends in differences between invasive and noninvasive values are a very important issue for the correct understanding of group results and are also reported by Pucci et al. as present in some of their comparisons.
Both devices studied by Pucci et al. evidently meet the AAMI standard when invasively calibrated; however, noninvasive calibration showed considerably less agreement. It could be argued that the accepted standard for noninvasive brachial BP devices is not the appropriate standard in the case of central BP estimates. It may also be that accuracy in terms of exact concordance with invasive measures may not be the practical requirement, an acceptable ‘accuracy’ is dependent on the need for precision – it is an unanswered question as to how precise central BP estimation needs to be.
The other perennial issue is that of group agreement versus individual agreement. When considering accuracy it is clear from the Bland–Altman plots from the study by Pucci et al. that even if group means are similar there are, as usually observed , wide differences in individual cases between measured and estimated cSBP. Again this may not be an issue if absolute accuracy is not necessarily an absolute requirement as long as appropriate classification enabling stratification into accepted management groups is available. The presence of obvious trends in some of the Bland–Altman graphs, however, suggests that this might be problematic in many cases .
We are informed by Ding et al. that the studied Omron device incorporates in its inbuilt algorithm a correction to allow known under-estimation of brachial SBP by cuff BP measurement. This highlights the ‘buyer beware’ principle that, since no two proprietary devices are identical, and that not all approaches are the same, it is therefore not reasonable to expect the same results from all devices . In the device reported by Ding et al., apart from causing consequent device-dependent differences in central-to-brachial BP ratios and pressure amplification as highlighted by the authors, this type of in-built adjustment would seem to have implications for the use of the obtained brachial BP values for traditional risk factor assessment.
The type of data presented by Pucci et al. and Ding et al. is clearly essential for the proper assessment of this type of technology in general and also to inform choice of device for those who choose to adopt it. It is, however, equally important that comparison results be judged in the context of clinical utility, not only as an esoteric issue of concordance. It is necessary that more work be done on assessing the issue of device-specific results, the continuation of device specificity will hamper general application, and also in assessing the true practical benefit of this type of noninvasive technology in the clinic. Just because something can be quantified and recorded does not make it useful or necessary to do so. Continued comparisons of novel technology, all of which may agree within reasonable limits to earlier generations, still means that with the inevitable sequential creep that the cumulative magnitude of absolute errors propagating through successive iterations of new devices may be substantial.
If, as suggested by the data provided in the two studies in this edition of the Journal[1,2], traditional brachial BP is in some cohorts as good an estimate of central BP as that derived by proprietary estimation approaches, this obviously questions the need for departing from the well understood and standard practice of measuring brachial BP in the clinic. Whereas lack of peripheral amplification may be true in some patient groups, there could equally be groups in which cSBP, as assessed by a (possibly specific) estimating method, would provide useful information (e.g. nondimensional central–peripheral amplification as is increasingly suggested ). This is not to say that all available approaches to noninvasive estimation of central BP will demonstrate the same appropriateness in specified patient groups. It might, as the saying goes, ‘be horses for courses’ as far as devices for estimating central BP are concerned.
There is clearly a rapidly expanding pool of devices for assessing central BP parameters. These are compact and cheap and thus appealing to practitioners. Until there is a consensus or perhaps guideline recommendation, users whether in a research or clinical situation, need to seek out the available data on a particular technology and balance its depth and appropriateness to assure themselves of the applicability before settling on their method of choice. Whereas it is difficult to imagine the development of substantial consensus or even less of a single mandated device, the myriad of different approaches employed (transfer function, regression, some not given) and calibration issues imply that results will remain device-specific. This drawback would seem likely to moderate general uptake and impede design and establishment of the appropriate large trials that are required to establish the place of assessment of central BP in practice. The vital requirement to overcome some of these issues is the full publication of basic accuracy and precision data relating to all devices of interest, data including that such as provided by Pucci et al. and Ding et al.. It could be that the biggest risk to the rapidly expanding interest in central BP is the unquestioning adoption of new devices just because they are new, rather than they are the most appropriate candidate for the job. Work such as those reported in these publications [1,2] allows those interested in the use of given devices to form reasoned conclusions.
In conclusion, those investigating central BP need to make an informed decision about what form of technology to employ and when to employ it; users should also consider what comparisons they require to inform their choice and when, how and what criteria should be employed in any comparisons performed.
Conflicts of interest
There are no conflicts of interest.
1. Pucci G, Cheriyan J, Hubsch A, Hickson SS, Gajendragadkar PR, Watson T, et al. Evaluation of the Vicorder, a novel cuff-based device for the noninvasive estimation of central blood pressure. J Hypertens 2013; 31:77–85.
2. Ding F-H, Li Y, Zhang R-Y, Zhang Q, Wang J-G. Comparison of the SphygmoCor and Omron devices in the estimation of pressure amplification against the invasive catheter measurement. J Hypertens 2013; 31:86–93.
3. Weiss W, Gohlisch C, Harsch-Gladisch C, Tölle M, Zidek W, van der Giet M. Oscillometric estimation of central blood pressure: validation of the Mobil-O-Graph in comparison with the SphygmoCor device. Blood Press Monit 2012; 17:128–131.
4. Brett SE, Guilcher A, Clapp B, Chowienczyk P. Estimating central systolic blood pressure during oscillometric determination of blood pressure: proof of concept and validation by comparison with intra-aortic pressure recording and arterial tonometry. Blood Press Monit 2012; 17:132–136.
5. Climie RE, Schultz MG, Nikolic SB, Ahuja KD, Fell JW, Sharman JE. Validity and reliability of central blood pressure estimated by upper arm oscillometric cuff pressure. Am J Hypertens 2012; 25:414–420.doi: 10.1038/ajh.2011.238.
6. Ding FH, Fan WX, Zhang RY, Zhang Q, Li Y, Wang JG. Validation of the noninvasive assessment of central blood pressure by the SphygmoCor and Omron devices against the invasive catheter measurement. Am J Hypertens 2011; 24:1306–1311.
7. Weber T, Wassertheurer S, Rammer M, Maurer E, Hametner B, Mayer CC, et al. Validation of a brachial cuff-based method for estimating central systolic blood pressure. Hypertension 2011; 58:825–832.
8. Horvàth IG, Németh A, Lenkey Z, Alessandri N, Tufano F, Kis P, et al. Invasive validation of a new oscillometric device (Arteriograph) for measuring augmentation index, central blood pressure and aortic pulse wave velocity. J Hypertens 2010; 28:2068–2075.
9. Zuo JL, Li Y, Yan ZJ, Zhang RY, Shen WF, Zhu DL, et al. Validation of the central blood pressure estimation by the SphygmoCor system in Chinese. Blood Press Monit 2010; 15:268–274.
10. Cheng HM, Wang KL, Chen YH, Lin SJ, Chen LC, Sung SH, et al. Estimation of central systolic blood pressure using an oscillometric blood pressure monitor. Hypertens Res 2010; 33:592–599.
11. Rezai MR, Goudot G, Winters C, Finn JD, Wu FC, Cruickshank JK. Calibration mode influences central blood pressure differences between SphygmoCor and two newer devices, the Arteriograph and Omron HEM-9000. Hypertens Res 2011; 34:1046–1051.
12. Hope SA, Meredith IT, Cameron JD. Effect of noninvasive calibration of radial waveforms on error in transfer function-derived central aortic waveform characteristics. Clin Sci (Lond) 2004; 107:205–211.
13. Cloud GC, Rajkumar C, Kooner J, Cooke J, Bulpitt CJ. Estimation of central aortic pressure by SphygmoCor requires intra-arterial peripheral pressures. Clin Sci 2003; 105:219–225.
14. Hope SA, Meredith IT, Tay D, Cameron JD. ‘Generalisability’ of a radial-aortic transfer function for the derivation of central aortic waveform parameters. J Hypertens 2007; 25:1812–1820.
15. Cameron JD. Assessment of central blood pressure waveforms: let the buyer beware: different approaches result in different results. Hypertens Res 2011; 34:994–995.
16. Benetos A, Gautier S, Labat C, Salvi P, Valbusa F, Marino F, et al. Mortality and cardiovascular events are best predicted by low central/peripheral pulse pressure amplification but not by high blood pressure levels in elderly nursing home subjects: The PARTAGE (Predictive Values of Blood Pressure and Arterial Stiffness in Institutionalized Very Aged Population) Study. J Am Coll Cardiol 2012; 60:1503–1511.