rivaroxaban or aspirin for cryptogenic embolic stroke??

The treatment for patients who have embolic strokes of an undetermined source ("cryptogenic strokes") in order to prevent recurrence is not clear. It is usually felt these are from an arterial source, though may be from paradoxical emboli from a DVT in the setting of a patent foramen ovale (PFO). A recent study found that rivaroxaban was not superior to aspirin but had more bleeding associated (see stroke embolic rivarox not better than asa nejm2018 in dropbox, or DOI: 10.1056/NEJMoa1802686​), a drug company sponsored study.

Details:

-- 7213 people were enrolled at 459 international sites, from 2014 to 2017, with non-lacunar cryptogenic strokes

-- mean age 67, 62% male, 72% white/20% Asian, BMI 27, blood pressure 135/79, 78% on statins, 77% hypertensive, 25% diabetic, 21% current tobacco use, 17% previous stroke/TIA, 17% on aspirin prior to a qualifying stroke, median NIH stroke scale 1 (range 0-42, higher scores reflecting worse neurologic deficits)

-- all patients were required to have at least 20 hours of cardiac rhythm monitoring to rule out afib lasting 6 minutes or longer. The median duration of cardiac rhythm monitoring was 24 hours, with 34% having monitoring for more than 48 hours (not clear what the specific timeframes were in those with longer monitoring, though by the numbers it is unlikely that many had 30-day monitoring)

--patients all had cerebral imagining to document the ischemic stroke, did not have extracranial vessel atheroxclerosis causing >50% luminal narrowing in the area of ischemia or with identifiable risk factors for a cardiac source of embolism (atrial fibrillation, left ventricular thrombus, mechanical prosthetic heart valve, severe mitral stenosis), no history of atrial fibrillation, and no other cause could be found [intracranial imaging was optional, but if performed there was no atherosclerotic stenosis >50%]

-- half were randomized to receive rivaroxaban 15 mg a day, half to aspirin at 100 mg per day

-- primary efficacy outcome: 1^st recurrence of ischemic or hemorrhagic stroke or systemic embolism; primary safety outcome: the rate of major bleeding

Results:

-- the study was terminated early (at 11 months) because of evident lack of benefit regarding stroke risk and increase bleeding associated with rivaroxaban

-- the primary efficacy outcome (nonsignificant difference):

    --rivaroxaban: 172 patients, annualized rate of 5.1%

    --aspirin: 160 patients, annualized rate of 4.8 %

--recurrent ischemic stroke occurred in 158 patients on rivaroxaban and 156 on aspirin, both annualized rate of 4.7%

--major bleeding:

    --62 patients on rivaroxaban, annualized rate 1.8%

    --23 patients on aspirin, annualized rate 0.7%

    --and, there were 13 hemorrhagic strokes in the rivaroxaban group, vs 2 in the aspirin group

-- afib was found in 3% of patients during follow-up, at a median of 5 months after entry

Commentary:

--this a pretty common clinical scenario: approximately 20% of ischemic strokes are “cryptogenic”, some studies have reported up to 40% (the studies differ in the adequacy of their work-ups)

--One major concern regarding the etiology of cryptogenic stroke hinges on the definition of afib.

    ​--For example, in this study, they monitored patients for a median of 24 hours to exclude afib. However other studies suggest that a 30-day monitor may be necessary. For example one study of 239 patients with cryptogenic ischemic stroke found that 24% who had paroxysmal afib​ were diagnosed only in the last 10 days of 30-day monitoring (see afib parox 30-day monitoring stroke2012, or Flint AC. Stroke. 2012;43:2788-2790), though they considered a qualifying definition as afib lasting >5 seconds (1/2 of the patients had episodes lasting >30 seconds, but no further data supplied). The EMBRACE trial (see Gladstone DJ. New Engl J Med 2014; 370: 2467) of 572 patients with cryptogenic stroke defined afib as lasting 30 seconds or longer, finding it in 16% of those with 30-day event monitors, but only in 2.5% in those with 24-hour monitoring (using the criterion of afib lasting at least 2.5 minutes, the numbers were 9.9% vs 2.5%). In another study with 568 patients, 30-day monitoring vs 24-hour monitoring detected >20% more cases of afib lasting >5 minutes, and 15% more than were picked up by a one week monitor (see Botto GL. J CardiovascElectrophysiol 2009; 20: 241); though in this article there were several cases of stroke in those with AF episodes lasting <5 minutes in patients with higher CHADS₂ risk.

    --Another question is how long episodes need to last to be clinically significant, causing embolic phenomena. is 5 seconds enough? 30 seconds? 5 minutes? 6 minutes as in this study? And, if afib were detected, did that actually cause the stroke? A recent study found that the burden of afib (the percent of time in afib) correlated with the risk of ischemic stroke, though other studies have found similar risks in those with either paroxysmal or chronic afib (this study will be analyzed tomorrow)

    --and, the above study would not likely show benefit for anticoagulation even if a signficant minority of those with paroxysmal afib (who should benefit from anticoagulation) were included, since these patients might be statistically overwhelmed by patients with other nondefined etiologies where anticoagulation is not needed (but adverse efffects still happen)

--also, it seems that this was an unusually lucky population, with remarkably few stroke sequelae, as noted above: median NIH stroke scale 1 (range 0-42, higher scores reflecting worse neurologic deficits). raises questions about selection bias

--other potential deficiencies in the study:

    --not all of the patients in the study had full intracranial vascular imaging, and some studies have found unexpected occlusions in a significant minority of patients with cryptogenic stroke

    --patients with echocardiograms finding PFO were included (though treatment of this is typically antiplatelet drugs, unless their is evident venous thromboembolism/DVT or high risk of such).  including these patients might skew the results to lack of benefit for anticoagulation. Some studies suggest the PFO is a very frequent association with cryptogenic stroke. 

    --Some studies have also found high incidence (around 25%) of complex aortic plaques in stroke patients; the treatment is not so clear for that, but the current data suggest antiplatelet therapy. 

    --finding these problems would suggest more specific therapies

--they used a lower dose of rivaroxaban that often used in the US, though this lower 15mg dose "substantially overlaps the effect of the 20-mg dose" so is unlikely to alter the results (and may actually cause less bleeding than the 20-mg dose)

--and, for some reason, only 78% of the patients were on statins

so, how should this study affect our clinical practice???  i'm not so sure, largely because the role of afib is not so clear.

    --they seem to have done inadequate testing to see if the patients had paroxysmal afib, most only 24 hours of monitoring though there is pretty consistent evidence that longer monitoring picks up more cases (specifically, 30-day monitoring)

    --their definition of afib is on the stricter side: at least 6 minutes, which is more than i saw in several other studies (though loosening this criterion would likely bias the results to more efficacy with anticoagulation)

    --on the other hand, afib has been found as the "cause" of cryptogenic stroke in up to 40% of cases. but that means only that some degree of afib was found on monitoring, yet there is only a presumption that afib had actually caused the stroke

    --but, stroke can be quite devastating, and the real treatment to prevent a recurrent and perhaps more debilitating stroke is by anticoagulation in the setting of afib

--also, it might be prudent to check for:

    --hypercoagulable states, esp in patients with history of diseases that might be associated (eg lupus), other venous/arterial emboli, family history, etc

    --routinely to do transesophageal echocardiography: for example, being able to identify aortic plaques or atrial abnormalities leading to embolic stroke. 

    --routinely check for intracranial arterial occlusions. 

--Finding these would take the patient out of the "cryptogenic" category and have better defined specific treatments, minimizing the risks of anticoagulation unless more specifically indicated.

so, though this study did not find any benefit but actual harm in using anticoagulation over antiplatelet meds, i would be hesitant to do such a limited evaluation for afib, and will continue starting with a 24-hour monitor (since it still does have reasonable detection rates for paroxysmal afib), but then proceed to a 30-day monitor if that were negative. it also makes sense to me to get routine transesophageal echocardiograms and intracranial vascular imaging. And, of course, optimizing lifestyle changes (diet, exercise) and maximizing statin use to effectively lower LDL, preferably <70 (it is a bit unsettling that only 78% of these patients in this recently done study were on statins).