Response by Demeestere et al to Letter Regarding Article, “Alberta Stroke Program Early CT Score Versus Computed Tomographic Perfusion to Predict Functional Outcome After Successful Reperfusion in Acute Ischemic Stroke”

We thank Harston et al. for their interest in our article, which describes the associations between baseline imaging parameters (ASPECTS, automated ASPECTS and ischemic core volume on CTP) and clinical outcome, indeed aimed to advance the field (1). In contrast to other studies, our analysis was focused on patients presenting in the extended time window. We reported the AUC only for ischemic core on CTP, because ischemic core on CTP was the only imaging modality associated with good and poor functional outcome in this selected population (2). As requested by Harston et al. we performed a ROC curve analysis for conventional ASPECTS and automated ASPECTS which revealed non-significant AUCs (conventional ASPECTS: 0.56 95%CI {0.47–0.66} for good outcome and 0.48 {95%CI 0.38 – 0.59} for poor outcome; automated ASPECTS: 0.59 {95%CI 0.5–0.68} for good and 0.52 {95%CI 0.41 – 0.63} for poor functional outcome).

Harston et al. bring up a few important methodological issues that were not described in sufficient detail in our manuscript. CTP data was indeed reviewed by the Stanford core imaging laboratory, and artefacts were removed where applicable. We performed a similar review for the non-contrast CT data: 14 patients were excluded based on insufficient quality of the non-contrast CT as judged by 3 ASPECTS raters and the automated ASPECTS results were reviewed for artefacts by one rater. We agree that if ASPECTS ratings had been performed unblinded to the CTP results, this could have induced bias. It is therefore important to note that raters were blinded to the CTP data and the automated ASPECTS when rating conventional ASPECTS. Finally, as pointed out by Harston et al. automated ASPECTS outperformed conventional ASPECTS for prediction of functional outcome but did not outperform core volume on CTP in this study.

CRISP Investigators were not blinded to baseline imaging, but they were instructed not to use the results from the RAPID maps for treatment decisions (2). We agree this approach may have induced selection bias, reflected in the low number of patients with large ischemic cores and low ASPECTS scores as acknowledged as a limitation of our study. We excluded patients with incomplete recanalization as continued infarct growth in these patients cannot be predicted by the baseline imaging and would bias the outcome analysis. Moreover, since reperfusion rates in current clinical practice are very high (>80%), clinicians are mainly interested in prediction of outcome in the setting of successful recanalization.

We do not agree with Harston et al. that our conclusions are not justified by our results. Only ischemic core volume measured by CTP was correlated with good and poor functional outcome in uni- and multivariable analysis, whereas automated ASPECTS was associated with good, but not poor, functional outcome. Our results and conclusions reflect this. We acknowledge that additional studies with larger sample sizes and more variability in stroke volumes are required to address some of the limitations of our study.