Comparison of Magnetic Resonance Imaging Mismatch Criteria to Select Patients for Endovascular Stroke Therapy Diffusion and Perfusion Imaging Evaluation for Understanding Stroke Evolution 2 Investigators

perfusion–diffusion (perfusion-weighted imaging–diffusion-weighted imaging, PWI–DWI) mismatch: patients with mismatch have a favorable clinical response to reperfusion, whereas patients without mismatch do not. This study examined whether alternative mismatch criteria can also differentiate patients according to their response to reperfusion. Methods —Patients from the DEFUSE 2 study were categorized according to vessel occlusion on magnetic resonance angiography (MRA) and DWI lesion volume criteria (MRA–DWI mismatch) and symptom severity and DWI criteria (clinical–DWI mismatch). Favorable clinical response was defined as an improvement of ≥ 8 points on the National Institutes of Health Stroke Scale (NIHSS) by day 30 or an NIHSS score of ≤ 1 at day 30. We assessed, for each set of criteria, whether the association between reperfusion and favorable clinical response differed according to mismatch status. Results —A differential response to reperfusion was observed between patients with and without MRA–DWI mismatch defined as an internal carotid artery or M1 occlusion and a DWI lesion <50 mL. Reperfusion was associated with good functional outcome in patients who met these MRA–DWI mismatch criteria (odds ratio [OR], 8.5; 95% confidence interval [CI], 2.3–31.3), whereas no association was observed in patients who did not meet these criteria (OR, 0.5; 95% CI, 0.08–3.1; P for difference between the odds, 0.01). No differential response to reperfusion was observed with other variations of the MRA–DWI or clinical–DWI mismatch criteria. Conclusions —The MRA–DWI mismatch is a promising alternative to DEFUSE 2’s PWI–DWI mismatch for patient selection in endovascular stroke trials. ( Stroke . 2014;45:1369-1374.) analysis were retained in the model. We evaluated the difference in the response to reperfusion between patients with and without mismatch using regres- sion models that included the significant covariates and the interaction between reperfusion and mismatch status as independent variables: where C is a constant, R is the presence of reperfusion, and is the presence of mismatch. All analyses were conducted with and Stats Direct.