Original Article: Downsides of Detecting Atrial Fibrillation in Asymptomatic Patients [Editorials]
Issue Date: March 15, 2019
See additional reader comments at: https://www.aafp.org/afp/2019/0315/p354.html
To the Editor: Stroke is rightly one of the most feared diseases of our times, having the devastating potential of leaving a person severely disabled within minutes. Furthermore, strokes brought on by emboli from atrial fibrillation (AF) are more likely to cause disability or death compared with other causes of stroke. Fortunately, AF-related strokes can be prevented if a person is aware of the disease. In their editorial, Drs. Mandrola and Foy argued that practical difficulties occur when trying to find patients with asymptomatic AF to start prescribing stroke-protective therapy. But are the difficulties practical?
Low Prevalence in the Screened Population. Drs. Mandrola and Foy state in the editorial that 0.5% of new AF was detected in the STROKESTOP study, but this is misleading. We conducted a large screening study and detected 3% of patients with new AF, and in total 5% of patients had untreated AF. Hence, the number needed to screen is 33 to detect one patient with new AF or 20 to detect one patient with untreated AF.1 For comparison, 2,451 women would need to be screened with mammography for five years to prevent one cancer-related death.2
Excess Costs Associated with Screening. In the Swedish STROKESTOP study, health economists estimated the cost of avoiding one stroke based on screening at $8,315.3 The social lifetime present value cost for stroke in 2009 was estimated at $86,902.4 Indeed, screening for AF in patients who have already had a stroke has been shown to save money.5
Poor Specificity for AF Screening. In a recent health technology assessment, specificity ranged from 97% with 12-lead electrocardiography to 94% using single-lead electrocardiography.6 The merit of a screening test is rarely solely based on specificity because a false-negative test would commonly be of greater importance, making a missed diagnosis of AF a greater worry than specificities in the range of 94% to 97%.
Lack of Direct Evidence. No study has yet shown that screening for AF can reduce stroke. We agree that the lack of direct evidence is an acceptable argument against screening. Physicians with concerns about screening their patients for this potentially life-threatening condition can wait for the results of ongoing trials.
In Reply: We thank Drs. Svennberg, Engdahl, and Rosenqvist for their interest in our editorial on AF screening.1
We agree that stroke is a devastating disease and that anticoagulant therapy reduces stroke risk when used in patients similar to those in clinical trials.
Regarding their point on the low prevalence of AF, we disagree that our use of the 0.5% new AF detection from the STROKESTOP study was misleading.2 It is true that STROKESTOP detected AF in 3% of patients overall; we chose 0.5% specifically because it measured the rate of detecting AF with a single screening electrocardiogram. Most of the AF detection in the STROKESTOP study (140 of 218 new cases) occurred with serial intermittent electrocardiography recordings. Our editorial addressed the use of electrocardiography, so we believed that using the detection rate of the index electrocardiogram was most appropriate.
Regarding cost, we consider our estimates to be conservative given that we assigned yearly costs of oral anticoagulants of $100, and we excluded the downstream costs of incidental findings and misdiagnoses. Downstream testing in the fee-for-service U.S. health care system could be massive. Also, it is not possible to estimate the cost effectiveness of an initiative without first knowing that it is useful. Namely, we don’t know whether oral anticoagulation will provide net clinical benefit to patients with subclinical screen-detected AF.
We respectfully disagree with Drs. Svennberg, Engdahl, and Rosenqvist on the matter of specificity. Given the low overall prevalence of AF in the general population, even in the elderly, incorrectly telling hundreds of thousands, perhaps millions, of people that they have AF could cause direct harm from anticoagulant-related bleeding and indirect harm from the anxiety of being labeled with a heart condition.
The only way to sort the benefits and harms of AF screening is with a randomized screening trial. As Wilson and Jungner stated in 1968: “The object of screening for disease is to discover those among the apparently well who are in fact suffering from disease…In theory, therefore, screening is an admirable method of combating disease...In practice, there are snags.”3
As practicing physicians, we are deeply concerned about the snags of AF screening.