Embolic Stroke, No Source Found: Why Empiric Anticoagulation Keeps Losing Its Trials
A stroke that looks embolic but has no confirmed source tempts the reflex to anticoagulate against a hidden cause — three separate randomized trials, including one that specifically pre-selected for the biomarker profile this patient carries, have now tested that reflex and found nothing to show for it.
L.A., a 59-year-old woman who has run her own accounting practice out of a converted garage office for fifteen years, noticed her words "coming out scrambled" while on a call with a client, bad enough that the client called her husband directly to check on her. She arrived at the emergency department within an hour, NIHSS 5, with mild expressive aphasia and subtle right facial weakness that had mostly resolved by the time of her neurologic exam. She has hypertension on losartan and no prior stroke or known cardiac disease; she does recall occasional "flutters" in her chest over the past year that she attributed to caffeine and never mentioned to anyone.
MRI confirmed a small cortical infarct in a distribution consistent with an embolic mechanism rather than a small-vessel lacunar pattern. Her full ESUS workup was otherwise unrevealing: 48-hour inpatient telemetry showed no atrial fibrillation, transthoracic echocardiogram showed no intracardiac thrombus or significant valvular disease, and vessel imaging showed no significant carotid or intracranial stenosis on either side. Two findings did stand out, though: her NT-proBNP came back elevated at 310pg/mL, and her echocardiogram noted mild left atrial enlargement — both proposed markers of what's been termed atrial cardiopathy, a theory that silent structural or electrical atrial disease can generate emboli even without confirmed atrial fibrillation. NAVIGATE ESUS and RE-SPECT ESUS both tested empiric anticoagulation against aspirin in unselected ESUS patients and both found no benefit, by different routes: NAVIGATE ESUS was halted early for futility with excess bleeding on rivaroxaban, while RE-SPECT ESUS ran to completion over a median nineteen months and simply failed to show dabigatran superior. ARCADIA went further, specifically enrolling ESUS patients with atrial cardiopathy markers — elevated NT-proBNP, abnormal P-wave terminal force, or left atrial enlargement, the same profile she carries — to test whether pre-selecting for exactly this biomarker signature would finally show a benefit to empiric anticoagulation. It did not. The trial that seemed built to succeed where the first two failed produced the same negative result.
At discharge planning, with a negative workup and two odd biomarkers
I'd add empiric anticoagulation here rather than stopping at aspirin. Her NT-proBNP is elevated and her echo shows left atrial enlargement — both real structural correlates of what's been called atrial cardiopathy, and it's a genuinely plausible mechanism for a silent embolic source even without confirmed fibrillation. Add her reported palpitations and this looks like exactly the patient that theory was built to describe.
That theory has actually been tested directly, and not just once. NAVIGATE ESUS and RE-SPECT ESUS both randomized empiric anticoagulation against aspirin in ESUS patients generally and neither found a benefit — NAVIGATE ESUS stopped early for futility, RE-SPECT ESUS ran its full course and still came back negative, which is the more informative of the two failures. But the case you're describing, atrial cardiopathy markers specifically, is exactly what ARCADIA was built to test: it enrolled ESUS patients pre-selected for elevated NT-proBNP, abnormal P-wave terminal force, or left atrial enlargement — the same profile she has — to see if biomarker selection would finally show a benefit empiric anticoagulation could deliver.
It didn't. That's not a smaller or less relevant negative trial than the first two — it's the one built specifically around her situation, and it came back negative too.
Which reframes the actual question worth asking here: not whether to anticoagulate empirically on a biomarker proxy, which three trials have now answered, but whether a longer look at her rhythm could still find a real, confirmable arrhythmia a 48-hour inpatient telemetry strip is too short to catch — especially given her own reported palpitations, which were never worked up before today.
An implantable loop recorder gives weeks to months of monitoring instead of two days. If it catches genuine atrial fibrillation, that's real evidence that changes the anticoagulation decision on its own terms — not an inference from a biomarker, an actual arrhythmia caught on a strip.
Agreed: aspirin monotherapy continues, and an implantable loop recorder will be placed for extended cardiac rhythm monitoring, targeting the possibility of a genuine, confirmable arrhythmia that her brief inpatient telemetry was too short to catch, particularly given her previously unreported palpitations.
Not agreed: exactly how long surveillance should run before the workup is considered genuinely complete if nothing is ever caught — left open as a judgment call for her outpatient follow-up rather than fixed in advance tonight.