August 10th, 2018
As technology keeps improving, it has become easier and easier for clinicians to detect disease and risk factors for disease. But just because we can detect disease does not mean we should always do these tests, especially for asymptomatic people. The recent debate over atrial fibrillation screening is a prime example.
Findings from a recent randomized controlled trial published in JAMA suggested that we can find more cases of atrial fibrillation (Afib) by screening asymptomatic high-risk patients. But just because screening helps identify more arrhythmia and increases use of anticoagulant drugs does not mean that it provides more benefit than harm for those who are screened.
We examined the issue in a blog on the Lown website, calling out the lack of evidence for treating patients with irregular or “low-burden” Afib with anticoagulant medication and the potential for overdiagnosis from screening.
Now, in an editorial in JAMA Internal Medicine, cardiologists Dr. John Mandrola, Dr. Andrew Foy, and Dr. Gerard Naccarelli use known probabilities to quantify the potential costs and benefits of Afib screening.
They first calculated the number of people who would need to be screened to prevent one stroke, called the “number needed to treat.” According to data from the large STROKESTOP study, it takes 200 initial ECGs to find one new case of Afib, and it takes 50 people treated with anticoagulants to prevent one stroke. Therefore, it takes (200*50)= 10,000 people to be screened to prevent one stroke. An electrocardiogram generally costs $50 (though it can cost more in many places), making the financial cost of preventing one stroke at least $500,000.
Would using an ECG patch with continuous monitoring get a better cost/benefit result? This question wasn’t addressed in the JAMA editorial, so we took a crack at it here at Lown headquarters. The answer is… sort of. According to the mSTOPS trial, 3.9% of people in the screened group were diagnosed with Afib after wearing the patch after four months, which means about 26 people needed to be screened to find one case of Afib. Multiplied by 50, that’s 1,300 people needed to be screened to prevent one stroke, a much lower NNT than doing just one ECG. However, wearable patches cost several hundred dollars, far more than a single ECG test, making the cost of preventing one stroke around $400,000.
As the authors point out, the financial cost of Afib screening is just the tip of the iceberg. The specificity of ECGs is not perfect, meaning there are often false positives — 10% of the time, to be exact. So out of the 10,000 people needed to screen to prevent one stroke, 800 people would get a false positive result (If Afib is present in 2% of patients, then 8000 do not have Afib, and 10% of these get a false positive result).
This needs to be taken very seriously if we are considering screening a broad population. The authors calculate that even if the test were 95% accurate instead of 90% accurate, for every million people screened, 49, 000 people would be wrongly told they had Afib and 250 would suffer a major bleed from the anticoagulants unnecessarily.
Mandrola and colleagues’ editorial emphasizes the need for more information before we subject all high-risk asymptomatic people to Afib screening. They point out that we only partially understand the relationship between Afib and stroke, especially the risk of stroke in patients with “silent” Afib, which could change our risk/benefit calculation.
At the same time that Mandrola and colleagues’ piece was published, the US Preventive Services Task Force also came out with a recommendation statement on whether or not we should screen for Afib with ECG. The USPSTF concluded similarly that “the current evidence is insufficient to assess the balance of benefits and harms of screening for atrial fibrillation with ECG.”