Primary Care Corner with Geoffrey Modest MD: The elusive search for afib in stroke patients; and an app

​​​​by Dr Geoffrey Modest

Atrial fibrillation is an important risk factor for current ischemic strokes, but may be hard to diagnose in those presenting in sinus rhythm. A reasonably large German study found that prolonged Holter monitoring picked up many more cases of atrial fibrillation than standard monitoring, the Find-AFRANDOMISED trial (see Wachter R. Lancet Neurol 2017; 16: 282–90).


–398 patients were recruited from 2013-2014 in 4 German centers, all with acute ischemic stroke and symptoms for 7 days or less, aged 60 years or older, in sinus rhythm and no history of atrial fibrillation (AF).

— Mean age 73, 40% women, 80% hypertension/27% diabetes/41% hyperlipidemia/18% current smoker/29% previous smokers/20% previous ischemic stroke/8% previous TIA/5% heart failure/10% MI/15% CAD/7% with ejection fraction <50%

— lacunar lesion on brain imaging found in 40%, cardioembolism 20%/small vessel disease 30%/stroke of unknown cause 50%, mean CHA2DS2-VASC score 4.8 (most in the 4-6 range), mean CHADS2 score 3.5 (50% in the 4-6 range). 197 patients were classified as having cryptogenic stroke; 201 as non-cryptogenic, mostly small vessel occlusion (118 pts) and cardioembolic stroke (75 pts)

— Those with severe ipsilateral carotid or intracranial artery stenosis were excluded

— patients were randomized into standard monitoring (at least 24 hours of rhythm monitoring: 188 of 198 patients had stroke unit telemetry for a median duration of 73 hours, and 149 of the 198 patients received additional Holter monitoring for a median of 24 hours) versus 10-day Holter monitoring at baseline, at 3 months, and at 6 months of follow-up. The initial Holter was done at a median of 3.5 days after symptom onset

— primary endpoint was the occurrence of atrial fibrillation or atrial flutter (lasting 30 seconds or longer) within 6 months after randomization and before stroke recurrence.

— secondary endpoints included: the detection of AF within 12 months, recurrence of stroke, systemic embolism or death within 12 months.


— after 6 months, 13.5 % were found to have atrial fibrillation in the enhanced monitoring group versus 4.5% in the standard group, absolute difference 9.0%, p=0.002, number needed to screen=11

— no patient with detected atrial fibrillation had a recurrent stroke or systemic embolization before the detection of atrial fibrillation within 6 months [by the way, this and another recent study I saw challenged the prior conventional wisdom that recurrrent strokes were much more common within the first week or two after the initial event]

— one of 27 patients in the enhanced monitoring group had atrial flutter

— the median duration of the longest AF episode during Holter monitoring was 5 hours, though one third lasted more than 24 hours and slightly less than one third < 6 minutes, and the number of episodes of atrial fibrillation detected ranged from 1 to 12

— review of their graph shows that the 1st 10 day Holter monitor picked up 18 patients, about ½  were picked up in the 1st 5 days; the 2nd  10-day monitor picked up an additional 6 with 2 picked up in the 1st 5 days; and the 3rd picked up one on the 8th day

— oral anticoagulation was given to all of the 39 patients who developed AF, more in the intervention group since more AF was picked up there

–clinical sequelae were found in 8 patients in the intervention group (5 recurrent strokes and 3 TIAs) and 14 in the control group (9 recurrent strokes and 5 TIAs), for rates of 3.7% vs 5.4%, nonsignificant (though this trial was underpowered for clinical outcomes, this finding does mirror that of the CRYSTAL-AF trial, which used an implantable cardiac monitor to pick up AF, finding 21% fewer events after 12 months). No cases of systemic embolization. No difference in picking up AF by age, sex, CHADS2, NIH Stroke Scale, symptoms at admission, or if the stroke was considered “cryptogenic”)​


— The rationale for looking aggressively for atrial fibrillation is that strokes from AF can be more severe, there is a high risk of recurrent strokes, and the detection of AF really changes therapy from antiplatelet drugs to oral anticoagulants, the latter decreasing the risk of recurrent strokes by 60 to 70%.  Since there are significant adverse events associated with these anticoagulants, it seems that their indications need to be pretty clear.

— The European Society of Cardiology recommends at least 72 hours of to monitoring, and also gives a Class IIa recommendation for implantable cardiac monitors (see Eur Heart J 2016; 37: 2893–962.)

— Review of the timing of AF pickups in the above study found that most (18/25, 72%) happened on the first 10-day cycle, and the pickup was reasonably evenly spread throughout the 10-day period; 6/25, (24%) were picked up in the second 10-day monitoring, again spread throughout the 10-day period; and one (4%) was near the end of the third 10-day period. This suggests to me that the monitoring should be for the entire 10-day periods, and that it is unlikely that a 4th 10-day period would be useful. The researchers in the above study suggested 7-10 days of monitoring within the first few days of symptom onset, and then repeating if higher risk (repeated cryptogenic strokes or embolic stroke of unknown source, frequent supravenrtricular ectopies, elevated natriuretic peptides, left atrial enlargement, or reduced atrial contractility).

–Holter monitoring has the advantage of being cheap, noninvasive, available, and able to be done within days of a cerebrovascular event.

so, very interesting study finding a significant number of patients having a stroke do in fact have AF on monitoring, and the more monitoring , the higher the pickup rate.  But hard to come to firm conclusions without a larger study powered sufficiently to assess clinical outcomes in order to see if AF pickup and treatment mattered (eg, is AF causative, or is it an innocent bystander which we know is common as age increases? and we also know that strokes themselves can cause cardiac arrhythmias, so which came first?) The other issues the larger trial could assess include:

​– what defines risky AF: eg, do really short episodes of AF matter (and what length does seem to matter?), and is this age-dependent?

— is there a number of AF episodes per 10-day monitoring that increase risk of stroke/TIA (and does that number vary depending on the length of AF episodes)? and, is this age-dependent?

— at what age should we do more aggressive monitoring (and should there be scaled amounts of monitoring based on different age groups, since AF is more common with increasing age)? is there an age where monitoring stops being clinically useful (either the AF doesn’t really increase risk that much, or the risks start to outweigh the benefits)?

the bottom line to me is that if we can show that picking up AF leads to improved clinical outcomes,  I would support more aggressive monitoring than the recommendations of the study authors: even though there were only 1 pickup during the third 10-day period, given how devastating a recurrent stroke can be, my inkling would be to support the 3 monitoring periods.

See here which argues for enhanced screening for atrial fibrillation overall (not just in people with strokes)

and  there are many blogs on atrial fibrillation treatment ( type atrial fibrillation in the search window)


As an aside, there is a free app for iphones called Cardiio which displays one’s pulse (just place a finger lightly on the camera on the back of the iphone). In Europe, it is approved to diagnose AF, but the FDA has not approved it in the US at this point. But one can see one’s rhythm, and patients could be shown how to use it and assess for abnormalities which might be AF. Basically, a study found that in 1013 patients with hypertension, diabetes, and/or aged >65, the sensitivity for the full Cardiio (Cardio Rhythm) was 92.9% and the specificity was 97.7%, as compared to single-lead ECG tracings reviewed by 2 cardiologists (see Chan P-H. J Am Heart Assoc. 2016;5:e003428, or  doi: 10.1161/JAHA.116.003428), though the positive predictive value in this study was only 53.1%. I have played with the app a little and seems pretty impressive to me (ie, I can see a clear waveform, documentation of the pulse, and, at least for the few times I’ve done it, I seem to be in normal sinus rhythm. Though not sure what I’d find with three 10-day Holter monitor recordings…)