# Development and Validation of a 18F-Flortaucipir PET Visual Stratification Method

## TL;DR

A new method for visually assessing tau levels in Alzheimer's patients using PET scans was developed and validated, showing high agreement among readers.

## Contribution

A novel visual stratification method for 18F-flortaucipir PET scans was developed and validated without requiring quantitation.

## Key findings

- The visual method achieved median positive and negative percent agreements of 83.4% and 88.9% against quantitation-based standards.
- Inter-reader and intra-reader agreements were nearly perfect with Fleiss κ of 0.8882 and Cohen κ of 0.9599.
- The method successfully stratified Alzheimer's patients based on tau levels with high reliability.

## Abstract

Tau PET quantitation methods have been used in research settings and clinical trials to measure tau burden for diagnostic, staging, and prognostic purposes. However, these methods require specialized software, skilled analysts, and advanced image processing. We developed a novel 18F-flortaucipir PET (FTP, or Tauvid) visual read method enabling stratification of patients with Alzheimer disease (AD) according to the tau level without the need for quantitation. An independent reader study (I7E-AV-A26) was conducted to test this method against a quantitation-based high-tau standard of truth. Methods: A total of 140 baseline or screening FTP scans were randomly selected from the TRAILBLAZER-ALZ 2 phase 3 trial (NCT04437511). Five qualified imaging physicians were trained for the FTP visual stratification method, using previously identified thresholds and cortical regions of interest thought to optimally stratify high-tau and non–high-tau scans. Positive and negative percent agreement (PPA and NPA, respectively) between visual stratifications and quantitation-based high tau (AD-signature SUV ratio > 1.46) were calculated. Predefined success criteria were met if the lower bounds of a 2-sided 95% CI for PPA and NPA were 50% or greater for at least 3 of the 5 readers. Inter- and intrareader reliability were assessed using Fleiss κ (n = 140) and Cohen κ (n = 20 test–retest scans) metrics. Results: The median PPA and NPA were 83.4% and 88.9%, respectively, with lower bounds of 2-sided 95% CIs being 50% or greater for all readers. The Fleiss κ-point estimate was 0.8882 (95% CI, 0.8356–0.9409) and the Cohen κ-point estimate was 0.9599 (95% CI, 0.9049–1.000) for all readers, indicating almost perfect inter- and intrareader agreement. Study I7E-AV-A26 successfully validated the feasibility of the FTP visual stratification method, possibly supporting AD staging and prognosis with high inter- and intrareader agreements, confirming the reliability of the method. Conclusion: Future investigations may include expanding the validation dataset, including real-world clinical data from diverse populations, using autopsy confirmation, exploring alternative regions and thresholds for other tau PET stratifications, and assessing differences in treatment response among visually stratified participants enrolled in disease-modifying therapy trials.