Looking for a needle in a haystack

Atrial fibrillation (AF) is an important and preventable cause of stroke. If AF is detected then a patient will usually be advised to go onto anticoagulation which reduces the risk of stroke by about 70%.

Some patients will have paroxysmal AF (PAF) which comes and goes interspersed by normal heart rhythm. PAF is defined as 30 seconds or more of AF and the risk of stroke is the same as with persistent AF. The ASSERT trial tried to determine how much PAF is necessary to increase the stroke risk and the results suggest that as little as 6 minutes of AF over a period of 3 months increases stroke risk by 2.5 fold.

After a stroke it is routine to do an 12 lead ECG to assess the cardiac rhythm and if this is normal then a longer period of heart monitoring would be performed to look for evidence of PAF. The question is however how long should the monitoring be done for to stand a good chance of detect PAF? Is 24h enough, or should it be a week or a month or even longer.

Two papers published this week in the New England Journal of Medicine have tried to address this question. The EMBRACE study compared a standard approach with 24h ECG monitor to a 30 day cardiac event recorder. 572 patients with apparently normal heart rhythm with a previous stroke in the last 6 months were randomly assigned to either 24h or 30 days of monitoring. In the group monitored for 24h just 3.2% of people had AF detected whereas in the group monitored for 30 days AF was detected in 16.1%. This meant that for every 8 people screened for the longer period 1 extra case of PAF was detected. Once AF was detected it led to a change of treatment for the patients with antiplatelet drugs being switched to anticoagulants which are much more effective at reducing recurrent stroke.

In the CRYSTAL AF study patients were randomly assigned to either having a loop recorder implanted or standard care. After 6 months PAF has been detected in 8.9% of patients with the ILR compared to 1.4% in the control group and by 12 months this had increased to 12.4% in the ILR and 2% in the control group.  There is a difference in the AF detection rate between the two studies which is probably due to the EMBRACE trial having an older population with more patients suffering from hypertension and diabetes. What is clear however is that the longer the period of monitoring the more cases of undiagnosed AF are detected. Since this has a profound effect on management of the patient these findings are very important.

There are some practical problems with monitoring patients for 30 days due to the ability to comply with the need for electrodes of the chest. The EMBRACE study used a dry electrode chest belt which has better tolerability and less skin irritation than traditional electrodes. The ILR technique is attractive particularly and with the advent of virtually injectable devices such as the REVEAL LINQ  but the cost of these devices might be seen as prohibitive if the technology was applied to every patient with a stroke. Perhaps though when a cost-effectiveness analysis is performed and the number of recurrent strokes prevented is factored in this type of monitoring device would pay for itself. If you take the parallel situation of a patient with a heart attack we think nothing of implanting several drug eluting stents which cost thousands of pounds in order to prevent a recurrent admission to hospital with chest pain or a non-fatal heart attack. What then of spending a similar amount to prevent a stroke?

These new studies are challenging the current practice of accepting a short period of monitoring when looking for AF. In this situation absence of evidence of AF should not be taken as evidence of absence and it looks as though a more prolonged period of monitoring is likely to be beneficial.

---