Results from a large autopsy validation study suggest that flortaucipir PET can help clinicians diagnose late-stage Alzheimer’s disease. In the April 27 JAMA Neurology, researchers led by Adam Fleisher at Avid Radiopharmaceuticals/Eli Lilly, Indianapolis, report that visual reads of brain scans predict tau pathology with a sensitivity above 92 percent.

The method’s specificity was equally good for some raters who inspected the scans but fell to as low as 52 percent for other raters in the study. All raters were trained to perform this test. Fleisher believes that small, noncontiguous bright spots on the scans led to false positives. “This can be mitigated by training,” he said.

Tau positivity also predicted AD neuropathology, meaning both plaques and tangles as outlined by the NIA-AA neuropathologic criteria for AD, with similar sensitivity and specificities. In a JAMA Neurology editorial, William Jagust, University of California, Berkeley, wrote that the study is a step toward validating methods crucial to improving the understanding of brain aging and AD. “The most important implication of this report is obvious: When visually interpreted by raters trained to recognize an AD pattern, 18F-flortaucipir PET performed in living patients is likely to detect advanced tau and AD pathologies,” he wrote.

Oskar Hansson and Ruben Smith of Lund University, Sweden, agreed. “The study is very well performed and supports the use of flortaucipir in the clinical diagnostic work-up of patients with cognitive impairment where AD is considered to be one of several possible etiologies,” they wrote to Alzforum (see comment below).

Senior author Mark Mintun, also from Avid, presented the bulk of the data at the 2018 CTAD meeting in Barcelona, Spain (Nov 2018 conference news). 

In a large collaboration called the A16 study, Fleisher and colleagues identified 156 volunteers who were terminally ill and who had agreed to a flortaucipir scan. They paper does not detail the types of illnesses. A16 bears no relation to the A3, A4, and A5 clinical trials for AD, but simply indicates that it was the 16th flortaucipir study conducted by Avid. Of the 156 volunteers, 67 died within six months and underwent autopsy. Three were evaluated as test cases, and the remaining 64 for comparison of neuropathology and tau PET. One of the volunteers had had mild cognitive impairment, 49 had had dementia, and 14 had been deemed cognitively normal. The mean time between the scan and autopsy was only 2.6 months, which Jagust called astonishing.

Five readers were trained in criteria to rate each scan as having a pattern consistent or not with AD. Avid did not attempt to quantitate the amount of tau in the brain for the main outcome. “The study responds to regulatory agency needs for visual interpretation of scans to reflect what is seen in neuropathology,” noted Val Lowe, Mayo Clinic, Rochester Minnesota. Some of the volunteers came from Mayo, and Lowe is a co-author on the paper. Lowe and colleagues also recently reported that flortaucipir scans correlate with tangle pathology in an autopsy study of 26 volunteers, of whom 11 had AD (Lowe et al., 2020). In that study, the scientists analyzed the scans quantitatively rather than with the naked eye. Visual reading of scans is thought to be less precise but more easily applicable in widespread clinical settings.

Fleisher and colleagues focused on diagnosis of advanced AD pathology, that is, people at Braak stages V and VI. In a particularly confusing twist of terminology, NIA-AA neuropathological criteria designate this “B3 level”, whereas B2 would correspond to Braak stages III to IV (Hyman et al., 2012). 

Reads by five raters, who were nuclear-medicine physicians or radiologists, predicted an AD B3 tau pattern with sensitivities of 92 to 100 percent, and specificities of 52 to 92 percent. The corresponding sensitivities and specificities for predicting AD neuropathology by NIA-AA criteria, namely a mix of plaques and tangles, were 94 to 100, and 50 to 93 percent, respectively.

To harmonize across the raters, Fleisher and colleagues calculated a “majority read.” The calculation for this was complex but basically boiled down to three or more raters interpreting a scan the same way. The sensitivity and specificity for majority reads for an AD tau pattern were 92 and 80 percent for flortaucipir, respectively, and 94 and 80 percent for AD NIA-AA neuropathology, respectively.

Will this data pass muster with the FDA? Lowe thinks so. Avid filed a new drug application last year, and Fleisher is hoping for a decision soon. An April 23 meeting of the FDA Medical Imaging Drugs Advisory Committee was cancelled because issues on which the agency was seeking input had been resolved.

How, or if, this scan may be used in the clinic remains unclear. The first PET ligand for amyloid, Avid’s florbetapir, was approved by the FDA eight years ago, and two more quickly followed suit (Apr 2012 news; Nov 2013 news; May 2014 news). Alas, they have not yet been cleared for reimbursement by the Centers for Medicare and Medicaid Services, despite data showing that amyloid PET scans help doctors tailor patient care (Aug 2017 news). “We leave it to the field to answer those questions and to researchers to let us know how best florbetapir should be used,” Fleisher told Alzforum.—Tom Fagan


  1. A crucial step in the validation of a PET tracer for tau pathology is the comparison of in vivo tracer retention and neuropathological verification of the pathology postmortem. Adam Fleisher and colleagues have now performed a relatively large end-of-life study, showing an accuracy of approximately 88 percent to detect Braak V/VI neurofibrillary tangle pathology using visual read of flortaucipir PET images. The sensitivity of flortaucipir was generally high (range: 92 to 100 percent), but the specificity was lower for some of the readers (range 52 to 92 percent). The relatively low specificity for some of the readers was explained by the fact that they overcalled small foci with signal in the temporal lobe.

    Importantly, quantitative measurement of the images revealed a higher specificity (of 100 percent), implying that a combination of visual read and quantitative assessment might improve the outcome. The study is very well performed and supports the use of flortaucipir in the clinical diagnostic work-up of patients with cognitive impairment where AD is considered to be one of several possible etiologies. It is interesting to note that all cases in the present study who had an elevated retention of flortaucipir (according to quantitative measurement) also exhibited significant amyloid pathology according to neuropathology. Therefore, it might not be necessary to also perform an amyloid-PET scan (or CSF Aβ analysis) in a patient with an abnormal tau-PET scan, except in rare cases where, e.g., svPPA is suspected.

    A weakness of the study might be that flortaucipir was mainly compared to Braak staging, considering that this staging system mainly reflects the distribution of neurofibrillary tangles in the brain and not the actual amount/density of tau pathology in different regions, which would better reflect the in vivo retention of flortaucipir. Further, the study implies that visual read of flortaucipir is not a suitable method to detect very early tau pathology in AD, and here P-tau181 and P-tau217 in cerebrospinal fluid as well as in plasma are likely to be more accurate markers.

    Overall, the study provides solid evidence that flortaucipir PET can be used in the diagnostic work-up of patients with cognitive impairment, and increased retention of flortaucipir according to quantitative assessment indicates with high certainty that a patient has AD. Flortaucipir PET will also be very useful to determine drug target engagement of certain anti-tau therapies, and to reveal potential downstream effects of anti-amyloid therapies.

Make a Comment

To make a comment you must login or register.


News Citations

  1. It’s Official: Tau PET Sees Tangles, and Staging Tangles Predicts Decline
  2. FDA Approves Amyvid for Clinical Use
  3. FDA Approves a Second Amyloid Imaging Agent
  4. Three’s Company: Florbetaben Approved, Excludes AD Diagnosis
  5. In Clinical Use, Amyloid Scans Change Two-Thirds of Treatment Plans

Paper Citations

  1. . Tau-positron emission tomography correlates with neuropathology findings. Alzheimers Dement. 2020 Mar;16(3):561-571. Epub 2020 Jan 4 PubMed.
  2. . National Institute on Aging-Alzheimer's Association guidelines for the neuropathologic assessment of Alzheimer's disease. Alzheimers Dement. 2012 Jan;8(1):1-13. PubMed.

Further Reading

No Available Further Reading

Primary Papers

  1. . Positron Emission Tomography Imaging With [18F]flortaucipir and Postmortem Assessment of Alzheimer Disease Neuropathologic Changes. JAMA Neurol. 2020 Jul 1;77(7):829-839. PubMed.