Background and purpose: We compared the value of Alberta Stroke Program Early Computed Tomographic scoring using CT (CT ASPECTS), collateral score on CT angiography (CTA), ASPECTS using diffusion-weighted MRI (DWI ASPECTS), DWI lesion volume, and DWI volume with National Institute of Health Stroke Scale (NIHSS) in determining candidacy of patients who presented >6 hours from stroke onset.
Methods: Decision to treat was first determined for each test alone and then with knowledge from other tests. A dismantling design was used to determine the additive effects of each test. Any discrepancy between the first and subsequent decisions to treat, in terms of sensitivity and specificity, is the impact of gained knowledge and was assessed using a generalized mixed-model assuming a binary distribution with PROC GLIMMIX/SAS. Inter-rater reliability was examined using weighted-Kappa.
Results: We identified 39 patients between December 1st, 2015 and June 30th 2016. Median time from last-known normal to non-contrast CT was 492 minutes. Median interval between non-contrast CT and CTA was 7 minutes, and between CTA and MRI, 75.5 minutes. For sensitivity, effect of knowledge gained from successive tests was not significant (Table 1; Fig. 1). However, significant gains in specificity were observed from successive tests (63% to 84%; p<.01). In particular, specificity increased by 14% (p=.09), 18% (p=.02), and 12% (p=.07), for DWI ASPECTS, DWI Volume, and DWI Volume+NIHSS, respectively. Inter-rater reliability was between .34-1.0 for each test.
Conclusion: CT, CTA and MRI each have the ability to correctly determine ELVO patients who would be candidates for embolectomy. However, identification of poor candidates for endovascular therapy was significantly improved using diffusion-weighted MRI.