Performance of pulse palpation compared to one‐lead ECG in atrial fibrillation screening

Abstract Background The 2020 European Society of Cardiology atrial fibrillation guidelines recommend opportunistic screening for atrial fibrillation by pulse taking or ECG rhythm strip in those aged over 65 years. Hypothesis We aimed to compare the diagnostic accuracy of pulse palpation to ECG rhythm strip when screening for atrial fibrillation. A secondary aim was to investigate whether participants with palpitations were more likely to be diagnosed with new atrial fibrillation. Methods The study population were 75/76 year old individuals that participated in the STROKESTOP II study, a Swedish screening study for atrial fibrillation. Pulse palpation of the radial pulse for 30 sec was performed by healthcare professionals and recorded as regular or irregular. Thereafter a 30‐sec single‐lead ECG was registered. Patients were asked also if they had a history of palpitations. Results Of the 6159 participants included in the study, 461 (7.5%) had irregular pulse. Twenty‐two (4.8%) of those with irregular pulse were diagnosed with atrial fibrillation on single‐lead ECG rhythm strip. Among those with regular pulse, 6 (0.1%) cases of new atrial fibrillation were found. The sensitivity of the pulse palpation test was 78.6% and positive predictive value 4.8%. The proportion of newly diagnosed atrial fibrillation was not different between those with and without history of palpitations. Conclusion Pulse palpation was inferior to single‐lead ECG when screening for atrial fibrillation. We therefore advocate the use of single‐lead ECG rather than pulse palpation when screening for atrial fibrillation. Palpitations did not predict atrial fibrillation.


| Study design and participants
This is a sub-study to The STROKESTOP II study, a prospective cohort study using NT-proBNP as a risk-stratifying tool in a systematic AF screening program for purposes of stroke prevention. The design and baseline results of the STROKESTOP II study have been published. 14,15 In brief all 75/76 year olds in the Stockholm region were identified by Statistics Sweden by their 10-digit personal identification number assigned to all citizens in Sweden and randomized 1:1 gender-and age-based to be invited to participate in the STROKESTOP II study or to serve as control group. There were no inclusion criteria other than year of birth and residence in the Stockholm region and no exclusion criteria.
Included in this study were all participants from the STROKESTOP II study who had no history of known AF. In the case where no pulse palpation had been performed, or if a single-lead ECG was deemed uninterpretable and no 12-lead ECGs were recorded as back up, the participant was excluded from further analysis. Data was collected between April 2016 and February 2018.

| Patient and public involvement
The research questions were developed by the study team when designing the STROKESTOP II study. The patients were not involved in designing the study, choice of outcomes or the recruitment of the study. All participants received oral and written information and signed informed consent documents.

| Test methods
Participants were asked to self-report their medical history with regard to prior thromboembolic risk factors according to the CHA 2 DS 2− VASc score, and if they had a history of palpitation symptoms.
Pulse palpation of the radial pulse for 30 s was performed by trained healthcare professionals. Regular pulse was defined as regular in rhythm and force. The pattern of irregularity was not especially defined. Irregular pulse was considered to be a positive test and recorded as such in the eCRF form.
Following the pulse palpation, a 30 s, handheld single-lead ECG recording with the Zenicor ECG device (Zenicor Medical Systems, Stockholm, Sweden) was obtained and evaluated by the same health care professionals. The ECG analysis with this singlelead ECG used a computerized algorithm marking all ECGs as normal or abnormal including those with poor signal quality or possible AF. The algorithm has been extensively validated, showing a 100% sensitivity of the system in identifying AF on an individual level and a negative predictive value of 99.99%. 16 If the algorithm classified a recording as sinus rhythm, a 12-lead ECG was not deemed necessary. If the algorithm indicated AF or if the signal quality was too poor for interpretation, a 12-lead ECG was obtained for rhythm confirmation and a cardiologist from the study team was contacted for interpretation. AF was defined as irregular rhythm with no organized or regular atrial activity on an ECG. For the single-lead ECGs, a minimum duration of 30 sec was required for diagnosis. The results from the single-lead ECG (and/or 12-lead ECG when deemed necessary) was used as reference standard as this is the case for diagnosing AF according to the ESC AF-guidelines. 5 Palpitation symptoms were self-reported as yes/no to the question "Do you have palpitations?" by patients in a questionnaire before the visit. There was no request for a specification of character, duration or temporal context of palpitation symptoms.
Post hoc, all single-lead ECGs were scrutinized by three cardiologists to confirm whether sinus rhythm could be clearly identified or whether a 12-lead ECG should have been taken at the time. A sensitivity analysis was performed including participants from whom a 12-lead ECG had not been obtained and the results from 2 weeks of intermittent (4 times daily) single-lead screening following the singletime point visit were imputed. Participants that had not concluded the additional 2 weeks of screening were excluded from the sensitivity analysis.

| Analysis
Based on previous studies on pulse palpation, for reliable sensitivity and specificity, we needed at least 4984 participants to achieve 80% power to show significant difference between the two screening tests. 7 Categorical variables were expressed as numbers and percentages. The demographics of individuals with and without new AF were compared using the Pearson's chi-squared test for categorical variables.
Diagnostic accuracy parameters of the screening test were derived from a 2 × 2 contingency table. We reported the number of true positives, true negatives, false positives and false negatives. We calculated sensitivity and specificity, accuracy, positive and negative predictive values, likelihood ratios as well as the post-test predictive values. Diagnostic accuracy parameters were expressed as means with 95% confidence intervals. We conducted subgroup analyses on the accuracy of the screening test for symptomatic versus asymptomatic participants and compared their area under the curve.
McNemar's test for dependent proportions was used to test whether the pulse palpation test was equivalent to the handheld-ECG test. For all statistical comparisons a p-value <.05 was considered significant.
We reported the results of this study, according to the Standards for Reporting of Diagnostic Accuracy Studies (STARD) statement. All analyses were performed using STATA/MP 15.1.

| Ethics
The study complies with the Declaration of Helsinki, and the protocol was approved by the regional ethics committee in Stockholm (DNR 2015/2079-31/1). Written informed consent was obtained from all participants in the screening program. ClinicalTrials.gov identifier:

| Participants and data analysis
In total, 6868 participants were included in the STROKESTOP II study.
Previously known AF was found in 553 (8.1%) of the participants and those were excluded from analysis leaving 6315 participants eligible for this study. In addition, 19 (0.3%) participants had no pulse palpation performed and in 137 (2.2%) participants the single-lead ECG was considered to be of insufficient signal quality with no 12-lead ECG obtained, hence these were excluded, leaving 6159 participants, see Figure 1. Baseline characteristics of participants with regular and irregular pulse are shown in Table 1. Female sex, congestive heart failure, hypertension and history of palpitations were significantly more frequent in participants with irregular pulse. The test performance of pulse palpation compared to single-lead ECG is shown in Table 2

| DISCUSSION
In this cohort study comprising over 6000 participants, pulse palpation showed a low positive predictive value and low sensitivity for AF detection when compared to a single-lead ECG rhythm strip. For each positive pulse palpation test, more than 20 ECGs were needed to confirm one case of new AF. It is worth noting that about 30% of participants reported history of palpitations, but this did not predict new AF detection.
The European Society of Cardiology Atrial Fibrillation Guidelines from 2020 recommend pulse palpation in opportunistic screening for AF. 5 Pulse palpation is a low-cost and highly available, although nondiagnostic test for AF, making verification with ECG necessary in case of irregular pulse. In our study, we found a lower sensitivity than that of Hobbs et al. found in the SAFE study resulting in a high false negative rate, regarded as an undesirable effect in a triage test. 17  similarly low positive predictive value for the pulse palpation test (5.1%) in the age group 75-84 years. 18 In a multicenter primary care study in Sweden, where patients were instructed to palpate their own pulse and afterwards record a single-lead ECG three times daily for 2 weeks, sensitivity was only 25% per measurement occasion. 19 In the editorial on that paper, pulse palpation was called "unsuitable" in modern day, resource-constrained health systems. 20 Tests of irregular pulse are triage tests, whereas single-lead ECG devices have shown good diagnostic accuracy. 9,21 A review from 2018 examined the evidence supporting the use of some of the newer technologies 8 and in an era of rapid technological advances and the many emerging new instruments for detecting irregular pulse as well as easy to use single-lead ECG recordings, it seems that there might be room for more modern and precise methods than pulse palpation. Observer variability between the study team cardiologist that confirmed single-lead and 12-lead ECG diagnosis is also possible.
The prevalence of the condition affects the performance of the test. Our study population was elderly, which means that both previously known AF as well as undiagnosed AF could be anticipated