DEFUSE-3 is DAWN All Over Again

A couple months ago the stroke community was presented with the results of DAWN – an acute stroke trial leaving us longing wistfully of the simple days where we just bickered over tPA in the 3-hour window. With DAWN, the authors evaluated patients with symptoms of 6 to 24 hour duration and found, with specific perfusion mismatch criteria, revascularization was likely beneficial. DEFUSE-3 is the second member of this club.

This is an open-label, blinded-assessor, randomized trial of patients presenting 6 to 16 hours following “last known well”. Dynamic randomization stratified based on age, core infarct, time from symptom onset, and NIHSS. Proximal middle cerebral artery and internal carotid occlusions were required, as was perfusion imaging demonstrating an ischemic core of less than 70 mL and a clinically important penumbra of at least 15 mL. The primary outcome was the clinically and statistically flawed ordinal shift in the modified Rankin scale.

As is become virtually routine in these endovascular trials, this one was stopped early – after 182 of 476 originally planned. These authors halted enrollment after the DAWN publication, and applied their protocol for interim analysis for early stoppage – which was met. Ignoring their ordinal shift measure and going straight for the typical mRS 0-2, we find 45% meeting that outcome with endovascular intervention, and 17% with medical therapy. Adverse events, including intracranial hemorrhage, were modestly increased in the endovascular arm, although mortality in the endovascular cohort are halved. Revascularization rates at 24 hours, interestingly, were only 79% in the endovascular cohort, and they still had a fair bit of infarct growth – but, sample size limitations notwithstanding, clinical outcomes obviously vastly favor the endovascular cohort. The American Stroke Association has wasted no time updating their guidelines to incorporate new time windows, up to 24 hours, using the DAWN and DEFUSE criteria.

Obviously, in an open-label trial, the scales can be tilted by the intensity of follow-up care – for instance, in IST-3, more stroke patients receiving tPA ended up in intensive care units. ICUs, with their more favorable nursing ratios, are virtually guaranteed to be more likely to be on top of all the other components of high-quality post-stroke care. In open-label trials, as well, it is reasonable to suggest different conversations may be had about comfort care measures, depending on whether an intervention were performed with the hopes of potential downstream improvement. These are just a couple examples of the many ways these sorts of trials may exaggerate the magnitude of potential benefit.

All that said, the Bayesian position – colored by sponsored trials as it is – still suggests these data are on the right track. The key to stroke care is tissue viability, not arbitrary time metrics. Either the collateral flow is there, or it isn’t – with the caveat that, yes, eventually those collaterals can collapse and complete the infarction. How these new tissue-based considerations fit into systems of stroke care – and who needs to be transferred for specialized imaging – is an open question/headache we’ll all be struggling to figure out over the next months/years.

“Thrombectomy for Stroke at 6 to 16 Hours with Selection by Perfusion Imaging”
http://www.nejm.org/doi/full/10.1056/NEJMoa1713973