The performance of TAUVID imaging to estimate the density and distribution of aggregated tau neurofibrillary tangles (NFTs) was evaluated in two clinical studies: Study 1 (NCT02516046) and Study 2 (NCT03901092). In each study, TAUVID imaging was interpreted by 5 independent readers who were blinded to clinical information. Readers interpreted TAUVID imaging as positive or negative
Study 1 enrolled 156 terminally ill patients who agreed to undergo TAUVID imaging and to participate in a postmortem brain donation program. In 64 of these patients, reader interpretation of the TAUVID scan was compared to tau pathology based on scoring provided by independent pathologists, who evaluated the density and distribution of NFTs in the post-mortem brain (see
Image reader performance for distinguishing B3 (positive) from B0-B2 (negative) tau pathology is shown in
The performance of the five TAUVID readers for sensitivity (95% CI) ranged from 92% (80, 97) to 100% (91, 100) and for specificity (95% CI) ranged from 52% (34, 70) to 92% (75, 98). Exploratory analysis evaluated how the same TAUVID interpretations distinguished B2-B3 from B0-B1 tau pathology, a threshold used in integrating tau and amyloid pathology for the neuropathological diagnosis of AD. In this analysis, the performance of the five TAUVID readers for sensitivity (95% CI) ranged from 68% (55, 79) to 86% (74, 93) and for specificity (95% CI) ranged from 63% (31, 86) to 100% (68, 100)
Study 2 included the same terminally ill patients as in Study 1 (plus 18 additional terminally ill patients) and 159 patients with cognitive impairment being evaluated for AD (the indicated population). Inter-reader agreement for five new TAUVID readers was evaluated using Fleiss' kappa statistic (95% CI) and found to be 0.87 (0.83, 0.91) across all 241 patients. Exploratory analysis evaluated inter-reader agreement in two subgroups. In this analysis, Fleiss' kappa (95% CI) was 0.82 (0.75, 0.88) in the terminally ill patients and 0.90 (0.85, 0.95) in the indicated population.