A probabilistic framework for the diagnostic utility of tau‐PET

TL;DR

This study evaluates how useful tau-PET scans are for diagnosing Alzheimer's disease, showing they are highly accurate and even more so when combined with amyloid-PET results.

Contribution

The paper introduces a probabilistic framework to assess the diagnostic value of tau-PET in Alzheimer's disease, integrating age, amyloid status, and prior diagnostic certainty.

Findings

Tau-PET has high positive predictive value for Alzheimer's, especially in younger individuals with higher prior probability of the disease.

Combining tau-PET with amyloid-PET significantly improves diagnostic accuracy, particularly in older individuals with lower prior probability of Alzheimer's.

Tau-PET's negative predictive value is consistently high, effectively ruling out Alzheimer's across age groups.

Abstract

The tau‐PET radiotracer [18F]flortaucipir enables in vivo detection of tau pathology in Alzheimer's Disease (AD) and has recently been FDA‐ and EMA‐approved for clinical use. To support its implementation, we assessed tau‐PET's positive predictive value (PPV) and negative predictive value (NPV) for clinico‐pathological AD while accounting for age, amyloid‐status, and pre‐PET diagnostic certainty.

We computed the PPV and NPV of tau‐PET using a formula that considers two types of false‐positivity: clinico‐pathological (positive tau‐PET, yet tau does not significantly contribute to cognitive decline) and pathological (positive tau‐PET, yet no/low tau is found at autopsy; Figure 1C). A systematic review yielded a weighted sensitivity of 93.6% and specificity of 83.9% for [18F]flortaucipir to detect postmortem Braak V/VI tau pathology (N = 349; Figure 1A). PPV and NPV were calculated across age groups using previously derived tau‐PET positivity prevalence estimates in cognitively unimpaired individuals and hypothetical clinician‐estimated prior probabilities of clinico‐pathological AD (Figure 1B), resulting in the probability that cognitive impairment is primarily caused by AD‐related tau pathology. We evaluated the PPV and NPV of tau‐PET as a standalone biomarker and in combination with amyloid‐PET.

The PPV of standalone tau‐PET was highest in individuals with higher prior AD probabilities and younger ages (Figure 2A). For example, at a prior AD probability of 70%, the PPV for AD was 92% at ages 50‐55, declining to 87% at ages 85‐90. Tau‐PET NPV was consistently high across ages and prior AD probabilities, effectively ruling out AD (Figure 2B). Obtaining tau‐PET results after knowing amyloid‐status markedly increased PPV for AD compared to both standalone tau‐PET or amyloid‐PET. The greatest increases in PPV from combining amyloid‐ and tau‐PET, relative to amyloid‐PET alone, occurred with lower pre‐amyloid AD probabilities and older ages (up to a 37% increase; Figure 3A). Similarly, when tau‐PET was obtained after amyloid‐PET, strong increases in NPV were observed, particularly when prior AD probability was high (up to a 48% increase, Figure 3D).

Tau‐PET demonstrates high PPV and NPV for clinico‐pathological AD as a standalone marker, with added diagnostic value when amyloid‐status is already known. These findings underscore tau‐PET's value for optimizing the diagnostic process.

Figures