The Role of Repeat CT for Predicting Fatal Outcome Following Massive Middle Cerebral Artery Territory Infarction
Background Large MCA ischemic stroke when associated with extensive mass effect can result in brain herniation and death. As yet there are no guidelines to aid the selection of patients for decompressive therapies, such as hemicraniectomy. Methods This was a multicentre retrospective study of large MCA infarction requiring neurocritical care. The repeat CT scans performed within 120 hours of stroke onset were assessed for horizontal displacement of septum pellucidum and pineal gland, the size of MCA infarction, involvement of other vascular territories and hydrocephalus. The primary outcome measure was death within 30 days. Results A total of 251 patients fulfilled entry criteria into the study of which 201 received conventional medical therapy alone. Univariate analysis identified the following predictors of early death: NIHSS >16 (P<0.03, OR 2.17 95% CI 1.12–4.2), anteroseptal shift >7 mm (P<0.001, OR 9.2 95% CI 4.1–20.63), pineal shift >3 mm (P<0.001, OR 12.1 95% CI 4.74–30.8), ischemia involving additional vascular territories (P<0.001, OR 7.02 95% CI 3.35–14.7), hydrocephalus (P<0.02, OR 2.13 95% CI 1.15–3.94), and temporal lobe involvement (P<0.001, OR 5.66 95% CI 2.58–12.4). Multivariate analysis was performed but no independent variables were identified because the CT variables were highly correlated. Anteroseptal shift dichotomised into < 7 and > 7 mm had sensitivity 53%, specificity 93%, positive predictive value (PPV) 88%, negative predictive value (NPV) 69% for neurological death. Pineal shift dichotomised <3 and = 3 mm had sensitivity 45%, specificity 94%, PPV 87% and NPV 66%. Conclusions We identified the role of follow-up CT in predicting early death following massive MCA infarction. Anteroseptal shift >7 mm and pineal shift >3 mm if present were strongly predictive of fatal outcome. These follow-up CT parameters require validation before they impact the decision to perform surgical intervention.