Stroke Stop Study: editorial comment

This editorial refers to ‘Population screening of 75- and 76-year-old men and women for silent atrial fibrillation (STROKESTOP)’ by L. Friberg et al., on page 135

The paper by Friberg et al. in this issue of the journal describes an ongoing study to evaluate the use of intermittent external monitoring to detect ‘silent’ atrial fibrillation (AF) in a huge (25 000) group of 75 and 76 year olds in Sweden.¹ The hypothesis is that if silent AF is detected early and treated medically with oral anticoagulation (OAC), that strokes can be prevented, lives can be saved, and it will be proven that screening for silent AF is cost-effective.

Interest in silent AF began over 10 years ago with the publication of several scientific papers showing an association between implanted device detection of atrial arrhythmias and increased risks of death and more specifically stroke.^2,3 With the advent of sophisticated pacemakers and implantable cardioverter defibrillators (ICDs) with extensive memory capability, it became clear that many patients were having AF episodes that were previously unrecognized. Over the last 10 years a great effort has been made to try to determine the clinical significance of these atrial arrhythmia episodes (silent AF).^4,5 Specifically, how long does an AF episode have to last to increase the risk of stroke enough to warrant OAC with its inherent increased risks of bleeding? Do these silent AF episodes have the same clinical risk of stroke as manifest episodes of AF? Most of the prior studies in this field have been done in patients who have an implanted pacemaker or ICD that was implanted for primary indications, and the study of atrial arrhythmias was a secondary intention. The Stroke Stop Study is planned to look for AF episodes as a primary indication for application of external monitors.

The hypothesis that very brief episodes of AF detected by a limited monitoring period will have important clinical relevance is very intriguing. The amount of monitoring proposed for this study is only 0.07% of the 2-week ‘monitoring period’, which equates to monitoring only 0.0005% of the overall study duration. Presumably, a patient who has AF detected by this brief period of monitoring would be one who is having a high clinical burden of AF.

Several salient questions come to mind regarding this planned study. First, is 15 minutes of monitoring over a single 2-week period long enough to detect the burden of silent AF that exists in the population? Previous studies that have shown an association with silent AF and stroke used continuous monitoring with an implanted device as the monitoring agent.^4,5 The ASSERT (ASymptomatic atrial fibrillation and Stroke Evaluation in pacemaker patients and the atrial fibrillation Reduction atrial pacing Trial) study found newly detected AF in about 34% of patients, and newly detected AF was also found in 30% of patients in the TRENDS (The Relationship Between Daily Atrial Tachyarrhythmia Burden From Implantable Device Diagnostics and Stroke) trial. In the TRENDS trial, the median time from enrollment in the study to identification of new onset AF via the device was 72 (13–177) days, and more than 60% of these AF patients were identified beyond the initial 30 days of the study.⁶ Several other studies have shown that the longer the duration of monitoring, the more episodes of AF will be revealed.^7,8 If patients are only monitored intermittently for a total of 15 minutes over a single 2-week period, there is a high likelihood that many episodes of AF will be missed. In addition, if a patient has no AF detected during the 28, half minute recordings they transmit, they will be considered ‘no AF’ and will be followed as such. It is certainly possible that patients in that group will have AF events that were just not seen during the brief intermittent 15 minutes of monitoring, and will have stroke events which will be misclassified as occurring in the ‘no AF’ group.

Second, is there evidence to support recommending anticoagulation for episodes of AF that last only 30 seconds in duration in patients with a CHADS₂ score of 1? In the Task Force for the Management of Atrial Fibrillation of the European Society of Cardiology (ESC), OAC is recommended for primary prevention of thromboembolism in patients with non-valvular AF who have only one major stroke risk factor (age >= 75) as the patients in this study will have.⁹ Alternatively, Fuster et al.¹⁰ in the ACC/AHA/ESC 2006 guidelines for the management of patients with AF recommend as a Class IIA indication either OAC or aspirin for primary prevention of thromboembolism in patients with non-valvular AF who have only one of the classic CHADS₂ risk factors (age >= 75). However, in neither of these gold standard documents is a minimum duration of AF defined. Guideline documents were created from data obtained from patients who had clinically manifest AF that was presumably documented by an electrocardiogram, not the brief silent AF episodes to be studied here. It is not certain that recommending OAC summarily for AF episodes lasting only 15–30 seconds in duration is justified based on our current knowledge base. It is even possible that the increased bleeding risks of OAC for the patients in this study with very small amounts of AF, will outweigh the benefits of potential thromboembolism prophylaxsis.

Third, if silent AF is detected, what percent of the cardiologists at the treatment centres will prescribe OAC for this low AF burden? Since the decision to recommend treatment with OAC is not mandated by the study but rather is left up to the treating cardiologists, it is not clear what percentage of patients who have new silent AF detected will actually receive OAC. As outlined above, it is currently not known that prescribing OAC to pts with very brief episodes of AF is indicated for the prevention of thromboembolism. If a significant percent of pts with new silent AF do not receive OAC, this will affect the numbers needed to screen in order to reach a statistically significant result.

Fourth, have the authors accurately determined the necessary sample size and have they accurately estimated stroke risks? The sample size for Stroke Stop AF has been calculated on the following assumptions: ‘the annual stroke risk in patients with AF without OAC is 7% (5-year stroke risk 30%). The protective effect of OAC is 70%, thus the expected annual stroke rate in treated patients is 2% (5-year stroke risk 10%)’.¹ There are no references in the manuscript for these assumptions. In the five landmark randomized clinical trials of oral anticoagulants (OACs) among patients with non-valvular AF, control subjects had a mean 4.5% annual incidence of stroke (range, 3–7%).^11–15 An overview of these initial five randomized trials found that the incidence of ischaemic stroke was reduced from 4.5% per year to 1.4% per year (relative risk reduction 68%; 95% confidence interval 50–79%; P< 0.001 by treatment with OAC.¹⁶ These data have a somewhat lower incidence of stroke in both the treated and untreated groups which could have impact on the proposed sample size.

Finally, is the equipment being used to monitor 25 000 patients accurate for detection of true AF? The authors state (with the support of several references) that the one-lead electrocardiogram recorder from Zenicor (www.zenicor.com) is more sensitive than a Holter monitor. Perhaps it is more sensitive because the monitor has the ability to record a rhythm strip several times a day for several weeks whereas a Holter is typically limited to a single 24 h application. When doing a study of this magnitude and expense, the recorder must be highly sensitive and specific for accurate detection of true AF episodes.

In terms of follow-up, the Stroke Stop Study is being done in the very unique country of Sweden where the entire population is followed in a centralized health care system. Outcome events, such as stroke, death, and hospitalizations, are carefully recorded in a centralized database so that there is little chance of loss to follow-up or loss of data. The same principles of highly accurate calculations of costs for stroke management apply in Sweden, also because of the centralized database. In turn, records of patient medications will also be highly accurate. This comprehensive system, completeness of follow-up, and full disclosure of and accessibility to data are truly remarkable; it would be close to impossible to conduct a study like this anywhere else in the world.

As a scientist with a tremendous interest in this field of the clinical significance of silent AF, I very much look forward to the results of this study in 5 years time. It would be wonderful to learn that screening for silent AF could prevent strokes, save lives, and lower costs. Certainly, if this trial showed that screening was beneficial in a population of patients with minimum CHADS₂ scores of 1, one can only magnify that benefit in patients with higher CHADS₂ scores.

Conflict of interest: none declared.

References

Author notes