ARTICLE IN BRIEF
In a small study, researchers reported that florbetapir-PET accurately identified vascular amyloid in patients with cerebral amyloid angiopathy (CAA)-related intracranial hemorrhage and accurately distinguished CAA–related intracranial hemorrhage from hypertensive intracranial hemorrhage.
Florbetapir, a Food and Drug Administration — approved positron emission tomography (PET) tracer — appears to highly accurately label vascular amyloid in patients with cerebral amyloid angiopathy (CAA)-related intracerebral hemorrhage (ICH), according to a report that appeared in the September 7 online edition of Neurology.
Moreover, the report indicated that florbetapir accurately distinguishes CAA–related ICH from hypertensive ICH (HTN-ICH), demonstrating 100 percent sensitivity for CAA and 89 percent specificity.
“After many years of research to identify a new diagnostic tool for CAA, florbetapir PET emerged as an extremely sensitive and highly specific method to diagnose this condition in cognitively normal patients with a primary hemorrhagic brain lesion,” said lead author M. Edip Gurol, MD, MSc, associate director of the Cerebrovascular Lab and assistant professor of neurology at Massachusetts General Hospital of the Harvard Medical School.
“CAA diagnosis is typically made using validated Boston criteria in people who have a lobar brain bleed and who also demonstrate one of more of lobar microbleeds or superficial siderosis on MRI scans,” Dr. Gurol told Neurology Today. “But there are many situations where diagnostic uncertainty exists — specifically patients with only one lobar brain bleed without any other hemorrhagic pathology, many older adults with only few microbleeds but no symptomatic bleed, patients with hemorrhagic lesions in both deep and lobar regions of the brain, and also people with a bleed on CAT scan who cannot get MRI.” [See sidebar, “Boston Criteria for CAA Diagnosis.”]
“Florbetapir not only binds the vascular amyloid but also the parenchymal amyloid plaques that form the basis of a pathologic diagnosis of Alzheimer's disease,” Dr. Gurol said. “For these reasons, we expect that florbetapir PET would be most helpful to make a diagnosis of CAA in patients who have at least one of the hemorrhagic manifestations of CAA, who do not exhibit cognitive impairment. A negative florbetapir scan in this setting might decisively rule out the presence of severe vascular amyloid as the cause of the patient's hemorrhagic brain pathology.”
He noted, however, that florbetapir will be unlikely to become a stand-alone diagnostic method, but rather its results will need to be combined with the clinical presentation as well as findings from head CT and particularly MRI in order to make an accurate diagnosis of CAA.
STUDY METHODS, FINDINGS
In the study, Dr. Gurol and colleagues enrolled 10 survivors of primary ICH related to probable CAA using Boston Criteria and nine survivors of HTN-ICH; both groups did not have dementia. All patients underwent florbetapir-PET and multimodal MRI, and patients with CAA had additional Pittsburgh compound B (PiB) PET — a demonstrated marker for CAA. Amyloid burden was assessed quantitatively and visually classified as positive or negative.
Florbetapir uptake and PiB retention were strongly correlated in patients with CAA both globally within cerebral cortex and regionally in occipital, frontal, temporal, and parietal cortices. These associations remained significant after adjustment for age and sex.
Average global cortical and occipital florbetapir uptake was significantly higher in CAA than HTN-ICH. Two trained neurologists, blinded to all other information, visually assessed the classification of positive florbetapir uptake and agreed with each other perfectly in all cases; their visual assessment of florbetapir-PET was positive for all 10 patients with CAA versus only one of nine HTN-ICH patients.
Dr. Gurol acknowledged the small sample size of the study, but said he believes the strength of the findings ensure generalizability. “Our study not only showed the significantly higher florbetapir binding in patients with CAA when compared to patients with hypertensive deep brain bleeds, but we also found a very strong correlation between florbetapir retention and PiB uptake in the CAA patients,” he said. “These results suggest that PiB and florbetapir similarly label vascular amyloid with high sensitivity. Our current paper that used florbetapir and the prior studies with PiB also show posterior predominance of amyloid tracer retention in the brains of patients with CAA, a well-known pattern for vascular amyloid.”
“The sample size is very small but the differences between lobar ICH and deep ICH were large and statistically significant, and the findings are plausible,” said Eric Smith, MD, MPH, the Kathy Taylor chair in vascular dementia at the Cumming School of Medicine at the University of Calgary.
“We know from prior clinical-radiopathological studies that patients with lobar ICH and probable CAA are highly likely to have vascular amyloid deposits while patients with deep ICH are much less likely to. Deep ICH is predominantly caused by arteriolosclerosis due to aging and conventional vascular risk factors, not CAA.
“Prior work, including autopsy validation, has shown that PiB binds to CAA,” Dr. Smith said. “By showing that florbetapir binding is elevated in lobar ICH and is highly correlated with PiB in the same subjects, Dr. Gurol and colleagues make a convincing case that florbetapir binds to CAA with high sensitivity and specificity. I see no reason to doubt the generalizability of their findings.”
Dr. Smith added that the radioactive atom in florbetapir is 18-fluorine, which has a much longer half-life than the radioactive 11-carbon atom in PiB — with a difference of 110 minutes versus. 20 minutes. “In both cases the compound must be made by a radiochemical productive facility and used on the same day,” he said. “But the longer half-life of florbetapir means that it can be made and transported over longer distances, making it much more practical for routine clinical use.”
Christopher D. Anderson, MD, associate director of the Acute Stroke Service at Massachusetts General Hospital, agreed. “I think the results are likely to be generalizable when comparing non-demented individuals with ICH in the setting of probable-CAA diagnoses with ICH in the setting of presumed hypertensive etiology,” he said. “Given the stark differences in presumed underlying pathology between groups, the small sample size was adequate to demonstrate a statistically significant result.”
(Dr. Anderson is a member of the same research group at Massachusetts General Hospital but was not involved in the design or execution of the present study).
Dr. Anderson added: “The results are, to me, fairly convincing that florbetapir performs similarly to PiB-PET in discerning probable-CAA related ICH from HTN-related ICH. This is real progress, as it suggests that a readily available PET-tracer could be a new approach to selecting appropriate patients for clinical trials in which you really want all subjects to share a similar underlying pathology.”
“Existing clinical criteria are very useful but there is a large pool of patients presenting with a single lobar hemorrhage that cannot be assigned into the probable CAA group, leaving their underlying pathology uncertain,” he said. “Future work will be needed to understand how reliably florbetapir might tease apart these finer diagnostic categories, but this current study suggests that such an approach holds significant promise.”
But James Burke, MD, assistant professor of neurology at University of Michigan Health System, offered an entirely different perspective. “The paper certainly has some strengths and this is an interesting test — making a CAA diagnosis could very credibly alter treatment recommendations — but the magnitude of the effect the authors found in this small study suggests that florbetapir PET likely does discriminate between CAA and hypertensive ICH. The statistical significance, though, only means that florbetapir is able to discriminate at all (that is. it's better than a random coin) and it only means that in this highly restricted sample.”
He noted that the claims that the test is “extremely sensitive and highly specific” should not be applied to a study where the 95 percent confidence intervals on sensitivity go down to 66 percent and specificity to 51 percent; the possibility that florbetapir is only marginally better than a coin flip is also consistent with these data due to the small sample size.
The biggest issue with this study, other than sample size, is spectrum bias, he said. The CAA group in this study was likely (although we don't really know this because the reported details are a bit sparse) to be obvious CAA, Dr. Burke said. This leads to two problems. First, sensitivity and specificity change based on the clinical context. If you applied this test to a study population where there are many more close-calls and many fewer slam dunks, you'd likely see worse performance on both sensitivity and specificity. Second, its also possible that the test only works well in the really obvious cases — patients with definite CAA may be precisely the people that have the large amyloid burdens that can be detected by florbetapir. Also, it's not clear when these tests were obtained — its not obvious, for example, that the tests would work as well if obtained in the acute setting.
In terms of clinical recommendations from this study, Dr. Burke said, the sample size fundamentally precludes anything resembling a clinical recommendation. The limited clinical spectrum this was tested on means that the sensitivity and specificity are likely overstated (possibly vastly overstated) relative to the clinical questions the authors allude to — it's not at all obvious that this test can help us find the right answer in the cases that are close calls. For a new test, the question isn't just whether it can discriminate one diagnosis from another, but how much value it adds to existing information. In this study it doesn't add anything because the groups were already clearly discriminated (apparently) on the basis of clinical and radiographic features in the first place.
“If larger studies in a more diverse patient population reproduce the findings of this study, then florbetapir PET could turn out to be a useful test,” Dr. Burke said. “For now, considerably more data is needed before this test should be applied clinically.”
Dr. Gurol told Neurology Today that his group is planning to perform further studies, specifically to test florbetapir binding to vascular amyloid in pathology specimens and also to validate the diagnostic value of this compound in other common scenarios such as patients with superficial microbleeds but no brain hemorrhages.
BOSTON CRITERIA FOR CAA DIAGNOSIS
The Boston criteria were first proposed in 1990 in order to standardise the diagnosis of cerebral amyloid angiopathy. They comprise combined clinical, imaging and pathological parameters.
The criteria are divided into four tiers:
- definite cerebral amyloid angiopathy:
- full post-mortem examination reveals lobar, cortical, or cortical/subcortical haemorrhage and pathological evidence of severe cerebral amyloid angiopathy
- probable cerebral amyloid angiopathy with supporting pathological evidence:
- clinical data and pathological tissue (evacuated haematoma or cortical biopsy specimen) demonstrate a haemorrhage as mendtioed above and some degree of vascular amyloid deposition
- doesn't have to be post-mortem
- probable cerebral amyloid angiopathy:
- pathological confirmation not required
- patient older than 55 years
- appropriate clinical history
- MRI findings demonstrate multiple haemorrhages of varying sizes/ages with no other explanation
- possible cerebral amyloid angiopathy:
- patient older than 55 years
- appropriate clinical history
- MRI findings reveal a single lobar, cortical, or cortical/subcortical hemorrhage without another cause, multiple hemorrhages with a possible but not a definite cause, or some hemorrhage in an atypical location
Source: Knudsen KA, Rosand ,Jarluk D, et al. Clinical diagnosis of cerebral amyloid angiopathy: Validation of the Boston criteria. Neurology 2001; 56(4): 537-539.