Advertisement

Screening for Asymptomatic Atrial Fibrillation While Monitoring the Blood Pressure at Home: Trial of Regular Versus Irregular Pulse for Prevention of Stroke (TRIPPS 2.0)

      Asymptomatic atrial fibrillation (AF) is a common cause of strokes. Physician screening for AF has been recommended. Home screening for AF may increase the likelihood of detecting asymptomatic AF in patients at risk for stroke because of AF. The aim of this study was to assess the feasibility and accuracy of screening for AF when taking home blood pressure (BP) measurements using an automatic AF-detecting BP monitor. Subjects aged >64 years or those with hypertension, diabetes, congestive heart failure, or previous stroke were enrolled by their primary physicians and given the AF-BP monitor and an electrocardiographic event monitor to use at home for 30 days. The AF-BP monitor reading was compared with the electrocardiographic reading to calculate the sensitivity and specificity of the device for detecting AF. A total of 160 subjects were enrolled, with 10 withdrawing, 1 excluded, and 10 with no AF-BP monitor logs or electrocardiographic recordings. Of the 139 subjects included, 14 had known AF. There was a total of 3,316 days with AF-BP monitor readings and electrocardiographic readings. On the basis of the initial daily AF-BP monitor readings, the AF-BP monitor demonstrated sensitivity of 99.2% and specificity of 92.9% for detecting AF. Two subjects with no histories of AF had AF-BP monitor readings of AF that were confirmed by the electrocardiographic monitor. One of these subjects was started on warfarin. In conclusion, home screening for asymptomatic AF while taking BP measurements can be performed accurately. This can be used to detect new AF, allowing treatment with anticoagulation to reduce the future risk for stroke.
      Approximately 2.3 million Americans have been diagnosed with atrial fibrillation (AF), and the prevalence is expected to double over the next 40 years.
      • Go A.S.
      • Hylek E.M.
      • Phillips K.A.
      • Chang Y.
      • Henault L.E.
      • Selby J.V.
      • Singer D.E.
      Prevalence of diagnosed atrial fibrillation in adults: national implications for rhythm management and stroke prevention: the Anticoagulation and Risk Factors in Atrial Fibrillation (ATRIA) study.
      Strokes due to undiagnosed AF can also be expected to increase in number. Periodic home blood pressure (BP) monitoring has been recommended for patients with hypertension in the United States and Europe.
      • Chobanian A.V.
      • Bakris G.L.
      • Black H.R.
      • Cushman W.C.
      • Green L.A.
      • Izzo Jr., J.L.
      • Jones D.W.
      • Materson B.J.
      • Oparil S.
      • Wright Jr., J.T.
      • Roccella E.J.
      Seventh report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure.
      • Parati G.
      • Stergiou G.S.
      • Asmar R.
      • Bilo G.
      • de Leeuw P.
      • Imai Y.
      • Kario K.
      • Lurbe E.
      • Manolis A.
      • Mengden T.
      • O'Brien E.
      • Ohkubo T.
      • Padfield P.
      • Palatini P.
      • Pickering T.
      • Redon J.
      • Revera M.
      • Ruilope L.M.
      • Shennan A.
      • Staessen J.A.
      • Tisler A.
      • Waeber B.
      • Zanchetti A.
      • Mancia G.
      ESH Working Group on Blood Pressure Monitoring
      European Society of Hypertension guidelines for blood pressure monitoring at home: a summary report of the Second International Consensus Conference on Home Blood Pressure Monitoring.
      By incorporating an algorithm that can detect AF in a home BP monitor, patients can be automatically screened for AF whenever they measure their BPs. Microlife Corporation (Taipei, Taiwan) developed a BP monitor that incorporates an algorithm designed specifically to detect AF. Previous studies performed at medical clinics demonstrated that the AF-BP monitor has sensitivity of 93% to 100% and specificity of 86% to 93% for detecting AF, a significant improvement over pulse self-examination.
      • Wiesel J.
      • Fitzig L.
      • Herschman Y.
      • Messineo F.C.
      Detection of atrial fibrillation using a modified Microlife blood pressure monitor.
      • Stergiou G.S.
      • Karpettas N.
      • Protogerou A.
      • Nasothimiou E.G.
      • Kyriakidis M.
      Diagnostic accuracy of a home blood pressure monitor to detect atrial fibrillation.
      However, these results were based on taking all the readings during 1 clinic visit. Home monitoring, which involves taking multiple readings in the same patient over many days, has the potential for increased sensitivity and decreased specificity. By taking many readings over a prolonged time period, new-onset AF and paroxysmal AF may be detected earlier, increasing the sensitivity for detecting AF. However, if other arrhythmias develop that could cause false-positive readings, specificity could be reduced. Therefore, this study was designed to determine the feasibility, sensitivity, and specificity of home screening for AF using the AF-BP monitor.

      Methods

      The Microlife BP monitor (model BPM BP3MQ1-2D; Microlife USA, Inc., Dunedin, Florida) has an algorithm that detects AF by analyzing the standard deviation and mean of the pulse beat intervals.
      • Wiesel J.
      • Fitzig L.
      • Herschman Y.
      • Messineo F.C.
      Detection of atrial fibrillation using a modified Microlife blood pressure monitor.
      To determine the applicability of home monitoring for AF in a general medical population at risk for stroke because of AF, subjects were recruited from general internists' offices. Included patients met ≥1 of the following criteria: age ≥65 years, hypertension, diabetes mellitus, congestive heart failure, and a history of a stroke. Subjects with pacemakers or implantable defibrillators were excluded. Subjects with known non-AF arrhythmias were not excluded from the trial. After providing informed consent, subjects were given the AF-BP monitor device to take home, to use daily for 30 days and to chart the readings on a log. An electrocardiographic event monitor (Heartrak 2; Mednet Healthcare, Ewing, New Jersey) was also provided to the subjects to obtain 60-second electrocardiographic recordings before the AF-BP monitor readings and to transmit the electrocardiograms to the monitoring center daily. If the daily AF-BP monitor reading indicated AF, the subject was to take 2 additional sequential readings. Using the “majority rule,” if either 2 or all 3 of the readings indicated AF, the subject was to wait 1 hour and obtain a fourth reading. If this last reading indicated AF, the subject was to record another electrocardiogram and transmit it at that time.
      A printout of each subject's electrocardiographic recordings was reviewed by a board-certified cardiologist, blinded to the results of the AF-BP monitor readings, who determined if the rhythm was AF or not AF. The study was approved by the New York Hospital Queens institutional review board. Written informed consent was obtained from all patients before participation in the study.
      The AF-BP monitor readings were analyzed using 2 methods. First, the first daily AF-BP monitor reading was compared to the electrocardiographic recording. This was used to determine the accuracy of individual AF-BP monitor readings. The first reading was classified as true-positive or true-negative or as false-positive or false-negative by comparing it with the contemporaneous electrocardiographic reading, which was considered the diagnostic standard. If the subject did not obtain an electrocardiographic recording or an AF-BP monitor reading on a given day, that day was excluded from the analysis.
      Second, the results of all AF-BP monitor readings taken on a given day were assessed (daily AF-BP monitor status). If most of the initial 3 AF-BP monitor readings and the fourth reading indicated AF, AF-BP monitor status was considered positive for AF on that day. If either the first or second electrocardiographic recording on that day showed AF, positive AF-BP monitor status was classified as a true-positive. If neither electrocardiogram showed AF, positive AF-BP monitor status was classified as a false-positive. Similarly, negative AF-BP monitor status was considered a true-negative or a false-negative by comparing the results with that day's electrocardiographic reading.
      If a subject had AF documented by electrocardiography on any day during the 30-day trial, he or she was considered to have an electrocardiographic diagnosis of AF. If the subject had no AF by electrocardiography on all the days of the trial, he or she was considered to have an electrocardiographic diagnosis of no AF. Similarly, if a subject had any day with positive AF-BP monitor status, he or she was considered to have an AF-BP monitor diagnosis of AF. If a subject was AF-BP monitor status negative on all the days of the trial, he or she was considered to have an AF-BP monitor diagnosis of no AF.
      Subjects were excluded from the analysis if they did not record any electrocardiograms at home or did not log any of the results of the AF-BP monitor readings. Only days on which an AF-BP monitor reading was logged and an electrocardiogram was recorded were included in the analysis. If a subject documented a positive AF-BP monitor reading on a given day but did not take the multiple readings as per the protocol, he or she was considered noncompliant with the protocol.
      Any subject found to have AF by the electrocardiographic event monitor who was not known to have AF previously was considered to have newly diagnosed AF. If this was documented as AF detected by the AF-BP monitor on the subject's log, it was considered to be new AF diagnosed by the device. This information was given to the subject's physician, and further evaluation and treatment were at the physician's discretion.
      On the basis of the protocol, every day that was included in the analysis included ≥1 AF-BP monitor reading and 1 electrocardiographic recording. Days with only negative AF-BP monitor readings required no additional device readings, but on days on which the first AF-BP monitor reading was positive, multiple additional device readings were required. Therefore, the most accurate assessment of the sensitivity and specificity of the AF-BP monitor for AF for individual readings could be obtained by comparing the first daily AF-BP monitor reading with the daily electrocardiographic recording. Confidence intervals were calculated by using the method of Jung and Ahn
      • Jung S.H.
      • Ahn C.
      Estimation of response probability in correlated binary data: a new approach.
      in the estimation of the variance of the sensitivity and specificity estimates for correlated binary data, which takes into account the number of positive and negative readings per subjects, as well as the intraclass correlation of the repeated measures per subject.
      The sensitivity and specificity of the AF-BP monitor for making the diagnosis of AF were calculated using the daily device status, which included all of the daily device readings. However, not all subjects complied with the requirement for multiple AF-BP monitor readings when the first reading was positive. Those subjects who were noncompliant and had electrocardiographic diagnoses of no AF could have had all false-positive, some false-positive, or no false-positive AF-BP monitor diagnoses of AF. The lower range for the specificity of the AF-BP monitor for diagnosing AF was calculated on the basis of the assumption that all the noncompliant subjects had false-positive diagnoses of AF. The upper range for specificity was calculated assuming that none of the noncompliant subjects had false-positive diagnoses of AF. The likely number of false-positive AF-BP monitor diagnoses of AF was calculated using a logistic regression model. In the compliant subjects, the likelihood of a subject having a false-positive AF-BP monitor diagnosis of AF was directly related to the number of days the subject had an initial false-positive AF-BP monitor reading. The logistic regression model calculated the probability of a false-positive AF-BP monitor diagnosis for each noncompliant subject on the basis of the number of first daily false-positive AF-BP monitor readings for that subject compared with the number of first daily false-positive AF-BP monitor readings in the compliant subjects.
      • Cohen J.
      • Cohen P.
      • West S.G.
      • Aiken L.S.
      Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences.
      • Fox J.
      Applied Regression Analysis, Linear Models, and Related Methods.
      The probability of a true-positive AF-BP monitor diagnosis of AF in those with AF who were noncompliant was calculated using the individual reading probability of detecting AF, accounting for the requirement of 3 of 4 positive readings and also accounting for the number of days in which AF was present for that subject.

      Results

      A total of 160 subjects were enrolled in the trial. Ten subjects withdrew from the study before recording any electrocardiographic or AF-BP monitor readings. One subject did not record any electrocardiographic readings, and 1 subject with a pacemaker was erroneously enrolled in the trial. Nine subjects did not record logs of their AF-BP monitor readings. Data analysis was based on the remaining 139 subjects who had ≥1 AF-BP monitor reading documented with a comparative electrocardiographic recording. Subject demographics and medical histories are listed in Table 1. By design, the study population is typical of those patients who are at risk for developing a stroke because of AF.
      Table 1Subject demographics and medical history (n = 139)
      VariableValue
      Mean age (yrs)67
      Age range (yrs)26–89
      Men51 (37%)
       White99 (71%)
       Black7 (5%)
       Hispanic7 (5%)
       Asian6 (4%)
       Not listed21 (15%)
       Hypertension118 (85%)
       Diabetes mellitus17 (12%)
       Congestive heart failure9 (6%)
       Previous stroke4 (3%)
      History of AF
       Chronic8 (6%)
       Paroxysmal8 (6%)
      CHADS2 score1.4
       Medications
      Angiotensin-converting enzyme inhibitor38 (27%)
      Angiotensin receptor blocker22 (16%)
      Calcium channel blocker21 (15%)
      β blocker37 (27%)
      Diuretic39 (28%)
      Warfarin14 (10%)
      There were 3,316 days on which an electrocardiographic recording and ≥1 AF-BP monitor reading was taken, for an average of 24 daily readings per subject (range 1 to 32). For the daily initial individual AF-BP monitor readings, the overall sensitivity of the AF-BP monitor for AF was 99.2% (confidence interval 93.7% to 100%), and the specificity was 92.9% (confidence intervals 92.3% to 93.4%) . There were 251 days on which the electrocardiogram showed AF in the 14 subjects with AF documented on electrocardiography. All subjects with AF had AF on >1 day of the trial. Of the 16 subjects with histories of AF, 4 did not have any episodes of AF documented by the electrocardiographic monitor during the course of this study. The 12 subjects with histories of AF, who also had documented AF on the electrocardiographic monitor, all had AF-BP monitor readings showing AF. Eight of these subjects had AF on all the electrocardiographic recordings and, presumably, had chronic AF. The other 4 subjects had paroxysmal AF, with electrocardiographic recordings that showed AF on some days and no AF on other days. Two subjects without histories of AF had AF-BP monitor readings showing AF that were confirmed by the electrocardiographic monitor to be true-positive AF readings. These subjects were not known by their physicians to have had AF previously and were asymptomatic at the time the device detected AF.
      Of the 139 subjects with AF-BP monitor readings and electrocardiographic recordings, 117 complied with the protocol on all days and took multiple daily readings when required. Of the compliant subjects, all 8 subjects with electrocardiographically documented AF had true-positive AF-BP monitor diagnoses of AF and 8 subjects had false-positive diagnoses of AF, for a sensitivity of 100% and a specificity of 93%. However, these results are the best-case results, because the noncompliant subjects were excluded. Inclusion of the noncompliant subjects would likely affect the specificity without a significant effect on the sensitivity. Because AF was present on >1 day in all subjects with AF and the sensitivity for AF for individual readings was 99.2%, the calculated probability of an AF-BP monitor diagnosis of AF in all subjects with AF results in sensitivity >99%. Therefore, the 6 subjects with AF who were noncompliant with the requirement for multiple readings would all have had AF-BP monitor diagnoses of AF had they been compliant with the protocol.
      There were 16 noncompliant subjects without AF on electrocardiography. If all these subjects had false-positive diagnoses of AF, the specificity would be 81%, and if they were all true-negative diagnoses of AF, the specificity would be 94%. To assess the most likely specificity, a logistic regression analysis was performed. Of the 125 subjects who had no episodes of AF by electrocardiographic monitoring during this study, 109 were compliant with the protocol, 70 with no false-positive individual AF-BP monitor readings and 39 with ≥1 individual false-positive reading. Of those 39 with ≥1 false-positive individual reading, only 8 had false-positive diagnoses of AF. This is because multiple sequential false-positive readings need to occur for a false-positive diagnosis of AF. For the 16 subjects who were noncompliant with the required multiple readings and had ≥1 false-positive individual AF-BP monitor reading, a logistic regression model was used to determine the best estimate of the specificity. On the basis of this model, 5 of the noncompliant subjects (95% confidence interval 2 to 8) would have had false-positive diagnoses. Using the logistic regression model, the total number of subjects with false-positive AF-BP monitor diagnoses of AF would be 13 (95% confidence interval 10 to 16). This would result in the AF-BP monitor having sensitivity of 100% and specificity of 90% (range 87% to 92%) for the diagnosis of AF (Table 2).
      Table 2Atrial fibrillation–blood pressure monitor diagnoses of atrial fibrillation versus electrocardiographic readings
      Positive ECG AF DiagnosisNegative ECG AF Diagnosis
      Positive AF-BP monitor AF diagnosis1413 (95% confidence interval 10–16)
      Negative AF-BP monitor AF diagnosis0112 (95% confidence interval 109–115)
      Results are based on a logistic regression model.
      ECG = electrocardiographic.

      Discussion

      Home BP monitoring with the Microlife BP monitor with a novel algorithm designed to detect AF was able to detect new AF in this trial. New AF was found in 2 subjects, 1 of whom was started on warfarin by her physician. The AF-BP monitor was found to have high sensitivity and specificity for diagnosing AF. In addition, the sensitivity and specificity for individual AF-BP monitor readings at home were comparable with the results obtained by the device when used in medical clinics. Although prolonged home monitoring increases the likelihood of individual false-positive AF-BP monitor readings, the requirement for multiple positive readings to diagnose AF maintains the overall diagnostic specificity at an acceptably high level. On the basis of the results of our 30-day study, monitoring for AF every other week or even monthly over the course of 1 year may provide an acceptably low false-positive rate. The total number of subjects with AF in this study was relatively small. A larger trial with more subjects with AF would be helpful to confirm the results of this study.
      Use of this device to screen for AF is more practical than using an electrocardiographic monitor at home. Most patients at risk for AF have hypertension, and home monitoring for hypertension is already recommended. Including AF screening in a home BP monitor avoids the burden of these patients' having multiple devices and reduces the likelihood of noncompliance when they are told to use the 2 devices. Although electrocardiographic monitoring is accurate when read by a trained physician, long-term home monitoring would have to rely on automated electrocardiographic readings, which, at this point, are not very accurate even for 12-lead electrocardiography.
      • Mant J.
      • Fitzmaurice D.A.
      • Hobbs F.D.R.
      • Jowettt S.
      • Murray E.T.
      • Holder R.
      • Davies M.
      • Lip G.Y.H.
      Accuracy of diagnosing atrial fibrillation on electrocardiogram by primary care practitioners and interpretative diagnostic software: analysis of data from Screening for Atrial Fibrillation in the Elderly (SAFE) trial.
      The cost-effectiveness of home screening for AF depends on the prevalence of undiagnosed AF in the screened population. Home screening for AF using a BP monitor would be reasonable in those at high-risk for AF, such as those aged >65 years with hypertension.
      Compliance with multiple readings and charting was an issue in this study. Methods of improving compliance by contacting the subjects regularly were not used. One method of improving compliance could be implemented in future studies. The majority rule for multiple AF-BP monitor readings was used in this trial on the basis of improved sensitivity and specificity compared with single readings in medical clinics. A post hoc analysis of the results of this study suggested that by changing the protocol from the majority rule to an “all positive” rule, whereby AF is diagnosed only if all 4 of the AF-BP monitor readings are positive, the specificity would be increased with only a small decrease in the sensitivity. Using this protocol and modifying the monitor so that all readings are performed automatically could improve specificity and increase the compliance rate.

      Acknowledgment

      We thank Cheryl Fruchter for assistance with data collection.

      Disclosures

      Dr. Wiesel has a patent for the AF algorithm, which is licensed to Microlife Corporation.

      References

        • Go A.S.
        • Hylek E.M.
        • Phillips K.A.
        • Chang Y.
        • Henault L.E.
        • Selby J.V.
        • Singer D.E.
        Prevalence of diagnosed atrial fibrillation in adults: national implications for rhythm management and stroke prevention: the Anticoagulation and Risk Factors in Atrial Fibrillation (ATRIA) study.
        JAMA. 2001; 285: 2370-2375
        • Chobanian A.V.
        • Bakris G.L.
        • Black H.R.
        • Cushman W.C.
        • Green L.A.
        • Izzo Jr., J.L.
        • Jones D.W.
        • Materson B.J.
        • Oparil S.
        • Wright Jr., J.T.
        • Roccella E.J.
        Seventh report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure.
        Hypertension. 2003; 42: 1206-1252
        • Parati G.
        • Stergiou G.S.
        • Asmar R.
        • Bilo G.
        • de Leeuw P.
        • Imai Y.
        • Kario K.
        • Lurbe E.
        • Manolis A.
        • Mengden T.
        • O'Brien E.
        • Ohkubo T.
        • Padfield P.
        • Palatini P.
        • Pickering T.
        • Redon J.
        • Revera M.
        • Ruilope L.M.
        • Shennan A.
        • Staessen J.A.
        • Tisler A.
        • Waeber B.
        • Zanchetti A.
        • Mancia G.
        • ESH Working Group on Blood Pressure Monitoring
        European Society of Hypertension guidelines for blood pressure monitoring at home: a summary report of the Second International Consensus Conference on Home Blood Pressure Monitoring.
        J Hypertens. 2008; 26: 1505-1526
        • Wiesel J.
        • Fitzig L.
        • Herschman Y.
        • Messineo F.C.
        Detection of atrial fibrillation using a modified Microlife blood pressure monitor.
        Am J Hypertension. 2009; 22: 848-852
        • Stergiou G.S.
        • Karpettas N.
        • Protogerou A.
        • Nasothimiou E.G.
        • Kyriakidis M.
        Diagnostic accuracy of a home blood pressure monitor to detect atrial fibrillation.
        J Hum Hypertens. 2009; 23: 654-658
        • Jung S.H.
        • Ahn C.
        Estimation of response probability in correlated binary data: a new approach.
        Drug Inform J. 2000; 34: 599-604
        • Cohen J.
        • Cohen P.
        • West S.G.
        • Aiken L.S.
        Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences.
        Lawrence Erlbaum, Mahwah, New Jersey2003 (736)
        • Fox J.
        Applied Regression Analysis, Linear Models, and Related Methods.
        Sage, Thousand Oaks, California1997 (624)
        • Mant J.
        • Fitzmaurice D.A.
        • Hobbs F.D.R.
        • Jowettt S.
        • Murray E.T.
        • Holder R.
        • Davies M.
        • Lip G.Y.H.
        Accuracy of diagnosing atrial fibrillation on electrocardiogram by primary care practitioners and interpretative diagnostic software: analysis of data from Screening for Atrial Fibrillation in the Elderly (SAFE) trial.
        BMJ. 2007; 335: 380-382