Design: Derivation: retrospective analysis of collated clinical stroke trial data (Virtual International Stroke Trials Archive). Validation: retrospective analysis of a prospective multicenter cohort study (Prediction of haematoma growth and outcome in patients with intracerebral haemorrhage using the CT-angiography spot sign [PREDICT]).
Setting: Neurocritical and ICUs.
Patients: Patients with acute intracerebral hemorrhage presenting less than 6 hours. Derivation: 552 patients; validation: 275 patients.
Measurements and Main Results: We generated a receiver operating characteristic curve for the association between 24-hour National Institutes of Health Stroke Scale change and clinical outcome. The primary outcome was a modified Rankin Scale score of 4–6 at 90 days; secondary outcomes were other modified Rankin Scale score ranges (modified Rankin Scale, 2–6, 3–6, 5–6, 6). We employed Youden’s J Index to select optimal cut points and calculated sensitivity, specificity, and predictive values. We determined independent predictors via multivariable logistic regression. The derived definitions were validated in the PREDICT cohort. Twenty-four–hour National Institutes of Health Stroke Scale change was strongly associated with 90-day outcome with an area under the receiver operating characteristic curve of 0.75. Youden’s method showed an optimum cut point at –0.5, corresponding to National Institutes of Health Stroke Scale change of greater than or equal to 0 (a lack of clinical improvement), which was seen in 46%. Early neurologic change accurately predicted poor outcome when defined as greater than or equal to 0 (sensitivity, 65%; specificity, 73%; positive predictive value, 70%; adjusted odds ratio, 5.05 [CI, 3.25–7.85]) or greater than or equal to 4 (sensitivity, 19%; specificity, 98%; positive predictive value, 91%; adjusted odds ratio, 12.24 [CI, 4.08–36.66]). All definitions reproduced well in the validation cohort.
Conclusions: Lack of clinical improvement at 24 hours robustly predicted poor outcome and showed good discrimination for individual patients who would do poorly. These findings are useful for prognostication and may also present as a potential early surrogate outcome for future intracerebral hemorrhage treatment trials.