In this study, we found that 18F-flortaucipir uptake in the entorhinal cortex was the best predictor for discriminating PET-Aß-positivity, and those in the amygdala or parahippocampal cortex also showed similar accuracy. This suggests that tau PET can serve as a predictor for PET-Aß-positivity in MCI patients.
For predicting PET-Aß-positivity in MCI, several clinical or MR-based methods have been developed, including the use of nomograms based on neuropsychological tests , informant-based reporting of cognitive symptoms , and MR-based brain morphometry or hippocampal volumetry . However, while morphometry which provided up to 80% of the sensitivity and specificity for predicting PET-Aß-positivity without clinical variables, other methods were suboptimal. Cerebrospinal fluid (CSF) Aß is also a powerful biomarker for predicting Aß pathology in AD . Previous studies have consistently shown a negative relationship between the CSF Aβ42 or Aβ42/Aβ40 ratio and Aß burden measured by PET and high agreement in Aß-positivity status between CSF studies and amyloid PET [14-17]. Nevertheless, the invasiveness required to obtain CSF limits its usage in clinical practice.
Like the previous 18F-flortaucipir studies that showed greater uptake in the medial and inferior temporal regions in MCI patients when compared to healthy controls [4, 5, 18], the MCI-Aβ+ patients in our study showed greater 18F-flortaucipir uptake in the parietal, lateral and medial temporal regions when compared to the MCI-Aβ- patients. 18F-flortaucipir uptake in these regions was also useful for discriminating AD from other neurodegenerative diseases.14 We therefore believe it is reasonable to select an optimal area from the regions with greater uptake in MCI-Aβ+ patients for predicting PET Aβ-positivity.
In our study, the highest sensitivity was achieved with 18F-flortaucipir uptake in the entorhinal (73.5%) and parahippocampal (67.4%) cortices, as expected by the early appearance of NFT pathology in these regions in AD . Expansion of Aβ throughout the neocortex without tau burden in the entorhinal cortex sufficient to exceed the cut-off threshold in 18F-flortaucipir PET might reduce the sensitivity. Similarly, 18F-flortaucipir uptake in the entorhinal, amygdala and parahippocampal cortices provided 85.7 to 94.6% of the specificity. Small false positivity for predicting PET-Aβ-positivity might be attributable to primary age-related tauopathy (PART) . However, there was no interaction between age and 18F-flortaucipir uptake in the entorhinal and parahippocampal cortices.
Although the greatest AUC value was achieved with the entorhinal cortex, followed by the amygdala and parahippocampal cortex, the post-hoc comparison of AUC values between the entorhinal cortex and the other two regions did not exhibit superiority of one region over any others for predicting PET-Aβ-positivity. However, compared to the entorhinal cortex, the amygdala and parahippocampal cortex provided lower sensitivity (< 70%) and higher specificity (> 90%). When focusing on the screening of PET-Aß-positivity, the entorhinal cortex may be the best region for predicting PET-Aβ-positivity.
This study was limited by the visual assessment for deciding Aβ-positivity, although this has been validated . In addition, absence of an external validation can be a limitation. Nevertheless, our study first added the usefulness of tau PET for predicting PET-Aß-positivity in MCI patients.