Performance characteristics of methods for quantifying spontaneous intracerebral haemorrhage: data from the Efficacy of Nitric Oxide in Stroke (ENOS) trial

Abstract

BackgroundPoor prognosis after intracerebral haemorrhage (ICH) is related to haemorrhage characteristics. Along with developing therapeutic interventions, we sought to understand the performance of haemorrhage descriptors in large clinical trials.MethodsClinical and neuroimaging data were obtained for 548 participants with ICH from the Efficacy of Nitric Oxide in Stroke (ENOS) trial. Independent observers performed visual categorisation of the largest diameter, measured volume using ABC/2, modified ABC/2, semiautomated segmentation (SAS), fully automatic measurement methods; shape, density and intraventricular haemorrhage were also assessed. Intraobserver and interobserver reliability were determined for these measures.ResultsICH volume was significantly different among standard ABC/2, modified ABC/2 and SAS: (mean) 12.8 (SD 16.3), 8.9 (9.2), 12.8 (13.1) cm3, respectively (p<0.0001). There was excellent agreement for haemorrhage volume (n=193): ABC/2 intraobserver intraclass correlation coefficient (ICC) 0.96–0.97, interobserver ICC 0.88; modified ABC/2 intraobserver ICC 0.95–0.97, interobserver ICC 0.91; SAS intraobserver ICC 0.95–0.99, interobserver ICC 0.93; largest diameter: (visual) interadjudicator ICC 0.82, (visual vs measured) adjudicator vs observer ICC 0.71; shape intraobserver ICC 0.88 interobserver ICC 0.75; density intraobserver ICC 0.86, interobserver ICC 0.73. Graeb score (mean 3.53) and modified Graeb (5.22) scores were highly correlated. Using modified ABC/2, ICH volume was underestimated in regular (by 2.2-2.5 cm3, p<0.0001) and irregular-shaped haemorrhages (by 4.8-4.9 cm3, p<0.0001). Fully automated measurement of haemorrhage volume was possible in only 5% of cases.ConclusionsFormal measurement of haemorrhage characteristics and visual estimates are reproducible. The standard ABC/2 method is superior to the modified ABC/2 method for quantifying ICH volume.Clinical trial registrationISRCTN9941422.