Atrial fibrillation screening during automated blood pressure measurement—Comment on “Diagnostic accuracy of new algorithm to detect atrial fibrillation in a home blood pressure monitor”

With the increasing longevity and prevalence of cardiovascular disease, such as hypertension and chronic heart failure, there is an emerging epidemic of atrial fibrillation (AF) in almost all communities. According to the worldwide estimation in 2010, the prevalence of AF was 0.596% in men and 0.373% in women, and the estimated number of patients with AF worldwide was 33.5 million.1 A major complication of AF is embolic stroke. In a recent registry, AF accounted for a third of all ischemic strokes.2 Anticoagulation therapy with warfarin, a vitamin K antagonist,3 or non‐vitamin K–dependent oral anticoagulants4 has proved to be highly effective in the prevention of stroke in AF. The challenge is that AF is usually asymptomatic or only mildly symptomatic and therefore significantly underdiagnosed. A quarter to a third of stroke cases related to AF occur as the first manifestation of AF.5 Nonetheless, screening is possible and with current technology can be cost‐effective, if anticoagulation is successfully implemented. Screening for AF in persons older than 65 years may identify a prevalence of 1.4% with a single time point examination,6 3% with patient‐activated ECG recordings over 2 weeks,7 and 10.4% with an even longer‐term continuous ECG recordings technique.8 The consensus therefore is that AF should be screened in high‐risk individuals, such as those older than 65 years, by all possible means.9

Current consensus on AF screening recommends various modalities for opportunistic or systematic screening of AF.9 In addition to a variety of ECG recording techniques, pulse measurement by manual palpation at the radial artery or by an automated oscillometric blood pressure (BP) monitor at the brachial or radial arterial sites is also recommended for AF screening.9 Pulse palpation is apparently cheap and simple but less accurate and more difficult for the elderly people, who have high prevalence of AF.10 On the other hand, automated oscillometric BP monitors may provide accurate information on irregularity of the cardiac rhythm by measuring the width and amplitude of pulse waves during BP measurement.10 Hypertension is a major risk factor for AF. Home BP monitoring with an automated device is recommended in patients with treated hypertension. AF screening using a BP monitor is becoming promising.

Several models of BP monitors from various manufacturers were developed with an added function of irregular heartbeat detection (Table).11, 12, 13, 14, 15, 16, 17 Some of those devices were tested for accuracy in the detection of AF. Wiesel and colleagues11, 13, 15 have conducted a series of studies on AF detection during BP measurement. In a pioneering study, Wiesel and associates11 used an algorithm for the detection of AF during BP measurement on the basis of the last 10 pulses during cuff deflation. The irregularity index was defined as the standard deviation of the time intervals between successive heartbeats divided by the mean of the intervals. In a validation study against a standard 12‐lead ECG, an irregularity index of 0.06 or greater was used as the threshold for the diagnosis of AF.11 This algorithm was integrated in a home BP monitor (modified Omron 712C, Omron). In 450 outpatients including 54 with AF diagnosed by ECG, the diagnostic sensitivity was 100% and specificity was 84% during a single BP measurement.11 The false‐positive results were mainly seen in patients with frequent ectopic beats.11 The investigators then modified the algorithm by discarding pulse intervals of 25% shorter or 25% longer than the average and using the remaining time intervals for the calculation of the irregularity index. In a later study in 405 patients including 93 with AF diagnosed by ECG, the newer algorithm with a single BP measurement had a sensitivity and specificity of 95% and 86%, respectively.13 When three measurements were performed with at least two of the three indicating AF, the sensitivity and specificity rates were 97% and 89%, respectively.13

Table 1.

Studies on the diagnostic accuracy of automated blood pressure monitors for the detection of atrial fibrillation

First author and year of publication	Study patients (No. of blood pressure measurements per patient)	Age, y	Prevalence of atrial fibrillation, %	Blood pressure–measuring devices	No. of blood pressure measurements		Accuracy against standard 12‐lead ECG
					Used	Needed for diagnosis	Sensitivity, %	Specificity, %
Wiesel 200411	450 Outpatientsa (2)	69	12	Modified Omron 712C	1 (any)	1	100	84
					2	2	100	91
Stergiou 200912	72 Outpatients (3)	71	37	Microlife BPA100 Plus	1 (any)	1	96	83
					2 (first)	1	100	76
					3	≥2	100	89
Wiesel 200913	405 Outpatients (3)	73	23	Microlife BPM P3MQ1‐2D	1	1	95	86
					3	≥2	97	89
Marazzi 201214	503 Outpatients (1 for Omron and 3 for Microlife)	67	20.1	Omron M6	1	1	100	94
				Microlife BP A200 Plus	3	Not reported	92	97
Wiesel 201415	183 Outpatients (1 for Omron and 3 for Microlife)	74	16.4	Omron M6	1	1	30	97
				Microlife BP A200 Plus	1 (first)	1	97	90
					3	≥2	100	92
Gandolfo 201516	207 Patients with strokea (3)	78	18	Microlife BPM P3MQ1‐2D	3	Not reported	89	99
Kearley 201517	999 Primary care patients (3?)	80	7.9	Microlife WatchBP	3 (?)	Not reported	95	90

The studies are listed according to the year of publication.

^aIncluding one¹¹ and four patients¹⁶ with atrial flutter, which was detected as nonatrial fibrillation by the device and considered as false‐negative by the researchers.

Several other studies12, 14, 15, 16, 17 investigated BP monitors with an algorithm similar to the abovementioned modified algorithm and showed similar accuracy in the detection of AF, except one monitor that had a sensitivity of 30%.15 In general, these BP monitors had higher sensitivity and specificity in the detection of AF. Nonetheless, there is still some room for improvement, especially in reducing false‐positive results.

In this issue of the Journal, Kabutoya and colleagues18 reported a new algorithm for the detection of AF using a BP monitor. The authors believe that the new algorithm may further improve accuracy for the detection of AF during BP measurement. They defined a new index, called irregular pulse peak (IPP) 25, as the absolute difference between each interval of pulse peak and the average of the intervals of pulse peak ≥25% of the average of the intervals of pulse peak. If the number of IPP beats was ≥20% of the total number of pulses on two of the three BP measurements, AF could then be diagnosed. The sensitivity and specificity was 88% and 100%, respectively. If 20% and 15% was used as the threshold for the definition of IPP, the sensitivity was 94% and 100%, respectively, and the specificity was 100% for both thresholds.

This new algorithm is different from the previous algorithm in several aspects.11 First, the new algorithm used all of the pulse waves during cuff deflation instead of the last 10 pulse intervals only. Second, pulse intervals of 25% shorter or longer than the average were included for the definition of the irregularity index IPP25 in the new algorithm but discarded in the previous algorithm. Third, for the diagnosis, the new algorithm counted the number of IPPs, whereas the previous algorithm used a fixed threshold of ≥0.06 of the irregularity index.

The new algorithm was developed from a new angle to improve specificity in the detection of AF. However, cardiac arrhythmias other than AF were not included in this study.18 Both the current and previous algorithms were designed by assessing pulse rate irregularity. False‐positive results can be present in patients with irregular heartbeats other than AF, such as various forms of premature beats and sick sinus syndrome. In the absence of other ECG‐proven cardiac arrhythmias, the current study does not allow any inference on specificity in the presence of other cardiac arrhythmias. Indeed, the new algorithm may have false‐positive results in patients with frequent premature complexes, because, as previously demonstrated, intervals of 25% shorter or longer than the average could also be attributable to premature beats.11

Whether adding another dimension of the pulse waves, such as pulse amplitude, for instance, would help improve accuracy in the detection of AF remains under investigation. We previously investigated the accuracy of devices with an algorithm accounting for both the variation of the pulse duration and the variation of the pulse amplitude.19 The sensitivity and specificity in the detection of AF and frequent ectopic beats were higher than devices with an algorithm accounting for the variation of the pulse duration only.

Not only the algorithm but also the number of BP measurements may play a role. In a previous study, increasing the number of BP measurements from one to three and applying a majority rule (at least two of the three measurements showing AF) increased sensitivity from 95% to 97% and specificity from 86% to 89% for the detection of AF.13 A similar improvement was observed in another study in sensitivity from 93% to 100% and specificity from 83% to 89%.12

The AF detection function should probably become available in all BP monitors for home use. Preferably, the detection can be automatically transmitted to a clinical setting for decisions of further AF screening and eventually for the diagnosis and treatment of AF. With these technical capabilities, large‐scale outcome studies on AF screening should be conducted for the identification of AF and the prevention of stroke.

Chen Y, Lei L, Wang J‐G. Atrial fibrillation screening during automated blood pressure measurement—Comment on “Diagnostic accuracy of new algorithm to detect atrial fibrillation in a home blood pressure monitor.” J Clin Hypertens. 2017;19:1148–1151. 10.1111/jch.13081