For any report of research data the appropriate use of methodology and the accurate interpretation of the findings are a fundamental requisite before accepting the final conclusions. Therefore, we appreciate the efforts by Cheng et al., who performed additional analyses of our published review and meta-analysis of studies on automated blood pressure (BP) measurement in atrial fibrillation . This reply is an opportunity to discuss the potential criteria applied for the validation of BP monitors and the meta-analysis of such data.
In the last 25 years, the US Association for the Advancement of Medical Instrumentation (AAMI), the British Hypertension Society (BHS), the European Society of Hypertension (ESH), the International Organization for Standardization (ISO) and other national organizations published extensive protocols for the validation of BP monitors [3–6]. These protocols have several differences, but have the same philosophy and essential criteria for the validation of BP monitors [3–6] and allow the standardization of the validation procedure and reliable comparisons among different studies and devices.
Correlation coefficients have been occasionally used for the comparison of repeated or different measurements. However, these coefficients reflect concordance rather than agreement and strong correlations do not exclude a systematic difference . Therefore, the correlation coefficients are not regarded as reliable criteria of similarity and have never been adopted by any of the established validation protocols [3–6].
The Bland–Altman approach based on the calculation of the mean difference and its SD has been the basis of all the established validation protocols for the validation of BP monitors [3–6]. The US AAMI  and more recently the ISO protocol  have based their validation criteria on the mean BP difference (which should be within 5 mmHg) and the SD of the differences (which should be within 8 mmHg for individual BP readings and dependent on the mean difference for average BP of triplicate readings of each individual). On the contrary, the BHS  and the ESH-IP protocol  are based on the proportion of absolute BP differences within 5, 10 and 15 mmHg. The Bland–Altman scatter plot is an additional requirement in the ESH protocol, not only to demonstrate the distribution of the BP differences, but also to display the range of BP values included in the analysis . The 95% limits of agreement (mean difference ± 1.96 × SD of differences) have not been formally applied in any of the validation protocols, yet they represent another approach of the Bland–Altman method.
Unfortunately, none of the studies on the accuracy of automated BP measurement in atrial fibrillation included in our analysis  have faithfully followed the established validation protocols. Therefore, we meta-analyzed all the validation parameters reported by most of the published studies . In this context, pooled estimates of SD of differences and 95% limits of agreement might indeed provide important additional information. Unfortunately, of the six studies that provided mean BP differences only four reported the SD of the differences. As stated in our study, for the two studies not providing the SD of the differences [8,9], the highest reported SD value of the ‘mean’ BP values was used in order to allow meta-analysis of the mean BP differences . This assumption gave only a crude and very strict estimate of the SD of differences. Indeed, the SD of differences which were arbitrarily added for the two studies with missing data, were 50% higher than the next highest SD of the studies that reported such values. Interestingly, in the meta-analysis of the SD of differences by Cheng et al. the results were better for systolic BP (satisfying the validation criterion of within 5 ± 8 mmHg) compared with diastolic BP. However, this finding was reversed after excluding the study by Farsky et al. and the SD was higher than the 8 mmHg threshold for both systolic and diastolic BP.
In our meta-analysis of mean BP differences, the abovementioned ‘SD of differences’ assumption is expected to mainly affect the precision (uncertainty) of the pooled estimate . However, in the meta-analysis of the SD of differences performed by Cheng et al. this assumption is expected to have a much larger impact on the pooled estimate value (accuracy), which questions the validity of this analysis based on four reported and two arbitrary and high values. Thus, it is not a surprise that the findings regarding the SD of differences were not satisfactory for both systolic and diastolic BP . Given the small number of studies reporting SD of differences and their heterogeneity (particularly the study by Farsky et al. that differs considerably in device type and results), and the major assumptions required to meta-analyze the SD of differences, we believe that these results should be regarded as inconclusive.
As we stated in our original article , there is limited evidence and significant heterogeneity in the studies that validated automated BP monitors in atrial fibrillation, and further technological improvement is required. Our conclusion that the measurement of systolic BP appears to be more accurate than that of diastolic BP was based on the collective review of all the available data, which did not include the SD of differences due to very limited data .
Conflicts of interest
There are no conflicts of interest.
1. Cheng H-M, Tufanaru C, Pearson A, Chen C-H. Automated blood pressure measurement in atrial fibrillation: a systematic review and meta-analysis. J Hypertens
2. Stergiou GS, Kollias A, Destounis A, Tzamouranis D. Automated blood pressure measurement in atrial fibrillation: a systematic review and meta-analysis. J Hypertens
3. Association for the Advancement of Medical Instrumentation. The national standard of electronic or automated sphygmomanometers
. Arlington, VA: AAMI; 1987.
4. O’Brien E, Petrie J, Littler W, De Swiet M, Padfield PL, O’Malley K, et al. The British Hypertension Society protocol for the evaluation of automated and semi-automated blood pressure measuring devices with special reference to ambulatory systems. J Hypertens
5. Noninvasive sphygmomanometers: clinical validation of automated measurement type. International Organization for Standardization (ISO) 81060-2, 2009. http://www.iso.org
. [Accessed 23 September 2011]
6. O’Brien E, Atkins N, Stergiou G, Karpettas N, Parati G, Asmar R, et al. Working Group on Blood Pressure Monitoring of the European Society of Hypertension. European Society of Hypertension International Protocol revision 2010 for the validation of blood pressure measuring devices in adults. Blood Press Monit
7. Bland JM, Altman DG. A note on the use of the intraclass correlation coefficient in the evaluation of agreement between two methods of measurement. Comput Biol Med
8. Jani B, Bulpitt CJ, Rajkumar C. Blood pressure measurement in patients with rate controlled atrial fibrillation using mercury sphygmomanometer and Omron HEM-750CP device in the clinic setting. J Hum Hypertens
9. Vazquez-Rodriguez B, Pita-Fernandez S, Regueiro-Lopez M, García-Pedreira D, Carro-Rodriguez MJ, Perez-Rivas G, et al. Concordance between automatic and manual recording of blood pressure depending on the absence or presence of atrial fibrillation. Am J Hypertens
10. Farsky S, Benova K, Krausova D, Sirotiakova J, Vysocanova P. Clinical blood pressure measurement verification when comparing a Tensoval duo control device with a mercury sphygmomanometer in patients suffering from atrial fibrillation. Blood Press Monit