NEW YORK (GenomeWeb) – A University of Edinburgh-led team has developed a model that combines computed tomography scans and genotyping to predict whether someone who has had a stroke is more likely to have another one.
About half of strokes can be attributed to intracerebral hemorrhage (ICH), and may be caused by cerebral amyloid angiopathy (CAA) in which amyloid proteins build up in blood vessels of the brain. As CAA-associated strokes are linked to a higher risk of additional strokes and dementia, distinguishing them from other types of strokes could affect therapeutic decisions and prognosis estimates.
The Edinburgh team used APOE genotyping — APOE is linked to CAA — and brain scans from 110 adults who presented with ICH to develop their prediction model. The team reported in Lancet Neurology that their model was able to discriminate among people with high, medium, or low risk of CAA-associated ICH.
"Identifying the cause of a brain hemorrhage is important to planning patient care," first author Mark Rodrigues from Edinburgh said in a statement. "Our findings suggest that the combination of routine CT scanning with APOE gene testing can identify those whose ICH has been caused by CAA — a group who may be more at risk of another ICH or dementia."
Rodrigues and his colleagues developed a multivariable prediction model for CAA-linked stroke based on two features gleaned from patients' CT scans and their APOE genotyping status. They enrolled 110 people into the Lothian IntraCerebral Haemorrhage, Pathology, Imaging and Neurological Outcome (LINCHPIN) study after their first ICH.
The participants died shortly thereafter, but first underwent non-contrast CT scans that were then analyzed by two neuroradiologists. At the same time, the researchers used either peripheral blood or cerebellar tissue samples to determine participants' APOE status.
Three features — subarachnoid hemorrhage and finger-like projections from CT scans and being APOE ε4 positive — were independently associated with moderate or severe CAA, and the researchers combined these to make their predictive model.
The researchers reported that their model had excellent discrimination, based on a c-statistic analysis.
They further used the model to develop low-, medium-, and high-risk groups. If no predictors were present, that person's probability of moderate or severe CAA was 7 percent, while if someone had subarachnoid hemorrhage or was APOE ε4 positive, that person's had moderate, 44 percent to 64 percent risk of CAA. But, if someone had subarachnoid hemorrhage and another feature, that person had a 95 percent risk of moderate or severe CAA.
Rodrigues and his colleagues noted that CT scans are widely available and inexpensive diagnostic tests that can be performed on acutely ill patients. However, they noted that genotyping isn't as common. Leaving APOE genotyping out of the model and making it a CT-only model decreased its sensitivity.
The study, however, was limited by its small sample size and a post-hoc power calculation of 71 percent, leaving it open to type II errors, the researchers cautioned.
Additionally, University College London's David Werring added in a related commentary that the model needs to undergo external validation to ensure that to performs just as well in other patient populations and that the model needs to be compared to the more common MRI-based Boston criteria. Rodrigues and his colleagues added that they are planning such an external validation study.
Werring, too, noted that APOE genotyping isn't always available. "This requirement will limit immediate application in clinical practice, but if the criteria are externally validated and are shown to add important clinical value, it would make a strong argument for allocating resources to make APOE testing more widely available," he wrote.