Limited Diagnostic Utility of Amyloid PET in Distinguishing Cerebral Amyloid Angiopathy from Alzheimer Disease

Cerebral amyloid angiopathy (CAA) is a major cause of spontaneous intracerebral hemorrhage (ICH) and is also a critical contributor to cognitive decline in older patients.¹ According to a prior meta-analysis published in Neurology, investigators observed that amyloid-PET had moderate to good diagnostic accuracy in differentiating patients with probable CAA from cognitively normal healthy controls or patients with deep ICH.²These previously published findings suggest that a negative scan could be useful to rule out CAA in a clinical setting given amyloid-PET labeled both cerebrovascular and parenchymal amyloid.

In a recent study presented at the 2024 International Stroke Conference (ISC), held February 7-9, in Phoenix, Arizona, a regional uptake pattern specific to CAA was unable to be identified among patients despite stringent patient selection towards minimizing the odds of overlap between CAA and Alzheimer disease (AD).³ Conducted by senior author Hsin-Hsi (Cynthia) Tsai, MD, PhD, and colleagues, patients with probable CAA (“pure CAA”) or probable AD (“pure AD”) were identified retrospectively. In the study, amyloid PET was expressed as absolute SUVR or as ROI/whole cortex ratios and diagnostic utility was assessed by 2 experienced nuclear physicians.

Tsai, a neurologist at National Taiwan University Hospital and assistant professor of neurology at National Taiwan University College of Medicine, recently had a conversation with NeurologyLive® to discuss how the study operationally defined pure CAA and pure AD to exclude between-condition overlap. Tsai also shared the hypothesis that she and her colleagues had regarding occipital lobe uptake in CAA compared with AD, and their findings. Additionally, she spoke about the challenges and results that emerged from the investigation into amyloid PET diagnostic utility using visual analysis.

Top Clinical Takeaways

Amyloid PET scans, despite efforts to differentiate, showed limited diagnostic accuracy in distinguishing CAA and AD, indicating challenges in clinical application.
The study's findings challenge use of visual analysis by experienced specialists, revealing low sensitivity and moderate specificity in distinguishing between AD and CAA.
The research suggests the need for alternative approaches, such as voxel-wise analysis or combining amyloid PET with other biomarkers, to enhance diagnostic certainty.

NeurologyLive: Can you elaborate on the criteria used to define "pure CAA" and "pure AD" in patient selection, and how you decided to use these criteria for the study?

Cynthia Tsai, MD, PhD: AD typically presents with both brain amyloid and tau pathologies, while showed limited/no vascular insults. CAA is typically characterized by vascular amyloid-ß deposition but not tau pathology. Both the amyloid plaque in AD (mostly parenchyma) and CAA (mostly vascular) could be imagined using amyloid PET scans.

Therefore, we operationally defined pure CAA as a group of patients without tau pathology who show CAA-related vascular lesions as those in the Boston criteria, have positive amyloid PET but negative tau PET, and pure AD as a group of patients with typical A(+)T(+) on PET scans but show no vascular insults on MRI. This is our maximal effort to exclude patients with between-conditions overlap as these 2 age-related diseases (CAA and AD) commonly overlap.

Between patients with CAA and AD, what were the key findings regarding regional uptake patterns in the occipital, frontal, and posterior cingulate cortex?

As CAA tends to involve occipital area most frequently and probably most pronouncedly, we hypothesized that occipital lobe may show a higher relative uptake (to the whole cortex) in CAA than in AD, as suggested by several previous works by others. However, despite our strict patient selection, we still could not detect a significant difference (but only a trend) showing high relative occipital uptake in CAA.

This could be related to a small effect size that requires larger sample size to confirm, while it also implicates limited clinical diagnostic utility applying amyloid PET in differentiating CAA and AD. As for frontal and posterior cingulate cortex uptake, the amyloid burden was higher in AD than in CAA as these regions are affected by Alzheimer pathology early in the disease stage.

In the assessment of diagnostic utility, were there any unexpected challenges or limitations encountered in determining cutoffs for SUVR and visual assessments?

The most unexpected challenges and results probably came from our investigation into amyloid PET diagnostic utility using visual analysis to differentiate CAA and AD. We have included 2 experienced nuclear medicine specialists who had more than 5 years of amyloid scan reading experiences (mostly for identifying patients with AD clinically). They have been instructed to categorize each scan as typical or atypical AD pattern. However, using visual scans could only yield low sensitivity (22-30%) and moderate specificity (52-62%) in differentiating AD from CAA.

SUVR cutoffs yielded slightly better, but still limited accuracy. This suggested the fact that during clinical practice, the visual pattern or regional uptakes characteristics of amyloid PET may have little role in improving our diagnostic certainty when determining the diagnosis CAA or AD. We may need other approaches to enhance the clinical utility of this tool (for example, a voxel-wise approach, combining with other biomarker, etc.)

Given the prevalence of incipient AD pathology in older individuals, how did the study address potential confounding factors or overlaps that might influence the identification of a specific regional uptake pattern for CAA?

This is of course our major limitation as we don’t have available pathological data in our samples. We have maximized our efforts using dual amyloid and tau PET scan, and adopted a strict criteria selecting negative tau PET patients in CAA—to minimize our possibility of selecting patients with overlap AD. In addition, we have performed subgroup analysis for patients CAA who had ICH and patients with CAA who had cognitive impairment, as those who presented with cognitive impairment may have a higher probability of concomitant AD. The results were not presented at ISC—but will be in our submitted paper thats currently under review—basically these findings were unchanged even in the comparisons between AD and CAA-cognitive impairment.

What implications does the study have for the clinical utility of amyloid PET in differentiating between CAA and AD, and are there potential areas for further research?

Overall, amyloid PET has limited diagnostic utility in differentiating CAA and AD in our study, not to mention probably even lower in those with overlap conditions (AD plus CAA). We are exploring other approaches. Future area of research could include a voxel-wise approach to identity CAA-specific brain area/amyloid uptake pattern, or to combine with other biomarker (CSF, plasma, other imaging modality, etc.) to enhance our diagnostic ability.