Profile and Prognosis of Spontaneous Lobar Intracerebral Hemorrhage: Comparison of 6-month Survival with STICH II and the MISTIE III Lobar Hemorrhage Subset

Abstract

Abstract Background Randomized trials on spontaneous lobar intracerebral hemorrhage (ICH) provided no convincing evidence of the superiority of surgical treatment. Since recruitment in the trials was under the premise of equipoise, a selection bias toward patients who did not need surgery or were in hopeless condition must be suspected. The aim of the actual analysis was to compare outcome and patient profile of an unselected hospital series with recent randomized trials and to develop a prognostic model. Methods Of 821 patients with spontaneous ICH managed at the neurosurgical department of the University Hospital Düsseldorf between 2013 and 2018, 159 had lobar bleedings. Patient characteristics, hematoma volume, treatment modality, and 6-month survival were compared with STICH II and the subset of lobar hemorrhage in the MISTIE III trial. In addition, a prognostic model for 6-month survival in our patients was developed using a random forest classifier. Results One hundred and seven patients were managed by surgical evacuation of the hematoma and 52 without surgical evacuation. Median hemorrhage volume in our surgical cohort was 66 and 42 mL in the conservative cohort, compared with 38 and 36 mL in the STICH II trial, and 46 and 47 mL in the surgical and conservative MISTIE III lobar hemorrhage subset. Median initial Glasgow Coma Scale (GCS) score was 12 in our surgical group and 11 in the conservative group, compared with 13 in the STICH II cohorts and 12 in the MISTIE III lobar hemorrhage subset. Median age in our surgical and conservative cohorts was 73 and 74 years, respectively, compared with 65 years in both STICH II cohorts and 68 years in the MISTIE II subsets. Twenty-nine percent of our surgical cohort and 55% of our conservatively managed patients deceased within the first 6 months, compared with 18 and 24%, respectively, in STICH II and 17 and 24% in the MISTIE III subset. Our prognostic model identified large hemorrhage volumes and low admission GCS score as main unfavorable prognostic factors for 6-month survival. The random forest classifier achieved a predictive accuracy of 78% and an area under curve (AUC)- value of 88% regarding survival at 6 months, on a test set independent of the training set. Conclusions In comparison with our surgical group, the STICH II and MISTIE III cohorts, recruited under the premise of physician equipoise, underrepresented patients with large ICHs. The cohorts in the randomized trials were therefore biased toward patients with a favorable perspective under conservative management. Initial hematoma volume and admission GCS were the main prognostic factors in our patients.