Accurately identifying high tau burden in patients with Alzheimer disease (AD) is essential for diagnosis, prognosis and therapeutic decisions. While positron emission tomography (PET) quantitation techniques offer detailed insight, they rely on complex processing and specialised software. To streamline this process, researchers developed and validated a novel visual interpretation method using 18F-flortaucipir (FTP) PET scans. This method aims to distinguish between high tau and non–high tau burdens in a more accessible and reliable manner, without the need for quantitation. The approach was validated in study A26, conducted on scans from the TRAILBLAZER-ALZ 2 phase 3 clinical trial.
Validation of the Visual Stratification Method
The newly developed FTP visual stratification method builds upon an existing 3-tier read system that classifies scans as consistent or not consistent with AD. The method introduces a higher visual signal threshold—2.80 times the mean cerebellar count (MCC)—specifically for assessing tau accumulation in the frontal lobes. This adjustment increases specificity, as previous thresholds lacked the ability to distinguish high tau from moderate or low levels.
In study A26, five trained imaging physicians assessed 140 FTP scans using a two-step visual process. The first step involved classifying the scan’s consistency with AD and the second applied the higher threshold to classify tau burden. Among these scans, 70 were confirmed to have high tau and 70 were non–high tau based on quantitative SUVr analysis, with a cut-off value of 1.46. The primary endpoints were the method’s agreement with quantitation-based classifications, assessed via positive and negative percent agreement (PPA and NPA).
Must Read: Biomarker-Based Stratification For Alzheimer’s Disease Diagnosis
The results were promising. The median PPA and NPA were 83.4% and 88.9%, respectively. Each reader met the predefined success criteria, with lower bounds of 95% confidence intervals for both PPA and NPA exceeding 50%. Moreover, Fleiss k-statistic values for interreader agreement reached 0.8882, and Cohen k-statistics for test–retest reliability peaked at 0.9599, demonstrating near-perfect reliability.
Clinical Relevance and Stratification Accuracy
Visual stratification revealed significant alignment with quantitative results. Among the 140 cases, 124 were interpreted with complete agreement across all five readers. Disagreements occurred in 16 cases, of which 10 were high tau and 6 non–high tau. Notably, many discordant cases had SUVr values near the 1.46 threshold, highlighting the difficulty in categorising borderline scans.
The method also demonstrated robust agreement for scans that failed quantitation. These results support the practical utility of the visual approach in real-world scenarios where quantitation may be unavailable or unreliable. Further, the method's reliance on visual signal intensity in the frontal cortex is consistent with the known progression of tau pathology in AD, which typically affects this region in later disease stages.
Post hoc analyses evaluated the method’s applicability to other brain regions, including the parietal, temporal and occipital lobes. High inter- and intrareader agreement in these areas suggests the method could be adapted for stratifying patients with low or medium tau burdens using alternate thresholds or regions. Such flexibility may prove valuable in future clinical and research applications.
Strengths, Limitations and Future Directions
One of the method’s greatest strengths is its accessibility. It does not require advanced image processing or proprietary software, allowing trained imaging physicians to perform reliable assessments using standard tools. This simplicity could significantly enhance AD diagnostics, particularly in clinical settings where resources are limited.
Despite strong validation outcomes, some limitations were noted. Accuracy, while high, was not perfect. Discordant cases mostly involved SUVr values near the quantitation threshold or atypical patterns of tau distribution. These edge cases may challenge even experienced readers, particularly when the signal is subtle or concentrated in non-frontal regions. Additionally, study A26 included only a subset of trial participants, and while statistically powered, results might not generalise to broader or more diverse populations.
The visual read method is specific to FTP and requires separate validation for other tau PET tracers. It was also developed under controlled imaging protocols within a clinical trial, which may not reflect the variability found in everyday clinical practice. Nonetheless, ongoing tau PET harmonisation efforts, such as the CenTauR scale, may support future adaptations of this method across tracers and settings.
Looking ahead, further studies could explore the application of this visual method in routine clinical environments, with diverse populations and less standardised imaging conditions. Autopsy confirmation, expanded datasets and comparative analyses with other stratification techniques would strengthen its clinical utility. Moreover, the method may help identify patients more likely to benefit from disease-modifying therapies, based on their visual stratification into high tau categories.
The FTP visual stratification method provides a practical, reproducible and reliable approach to identifying high tau burden in early symptomatic AD patients. Validated through robust agreement with quantitation-based classifications and strong reader reliability metrics, it represents a significant step forward in simplifying tau PET interpretation. While additional validation in diverse settings is warranted, this method holds promise for broader clinical use, enabling more accessible AD staging and supporting advances in therapeutic decision-making and clinical trial design.
Source: The Journal of Nuclear Medicine
Image Credit: iStock
