Most tau PET tracers detect only the type of misfolded tau found in Alzheimer’s disease. In the September 18 JAMA, researchers led by Rik Ossenkoppele at VU University Medical Center, Amsterdam, and Oskar Hansson at Lund University, Sweden, make the case that this selectivity can be exploited for the differential diagnosis of AD dementia versus other neurodegenerative disorders. In a cross-sectional study of 719 patients at several memory clinics, the tracer flortaucipir picked out AD cases with a sensitivity and specificity of around 90 percent. At the dementia stage of disease, the marker outperformed both amyloid imaging and structural MRI. However, flortaucipir missed cases at the prodromal stage, when tau pathology is sparse. The data suggest tau imaging could aid diagnosis, Hansson said.

  • Flortaucipir signal distinguishes AD dementia from non-AD disorders with high accuracy.
  • It outperforms both amyloid PET and MRI.
  • At the MCI stage, flortaucipir misses AD cases.

Other researchers expressed enthusiasm. “The diagnostic discrimination was impressive,” William Klunk at the University of Pittsburgh wrote to Alzforum (full comment below). Melissa Murray at the Mayo Clinic in Jacksonville, Florida, said the consistency of the findings across multiple centers was exciting.

At the moment, tau tracers remain a research tool, while amyloid PET agents are in widespread clinical use. Amyloid imaging allows clinicians to detect very early signs of AD pathology, but because amyloid plaques are present during a 20- to 30-year prodromal phase, a positive amyloid signal does not always explain a patient’s cognitive symptoms, particularly in elderly people. For example, a person might be in the preclinical phase of AD, and be experiencing memory impairment due to depression or some other co-morbidity. Tau tangles, on the other hand, appear later in the disease and correlate more closely with cognitive impairment, suggesting that tau imaging might be a better diagnostic for AD at symptomatic stages (May 2016 news). Flortaucipir, aka AV1451, has been the most-used tau tracer to date. While it produces a strong signal in AD, it binds only weakly in other tauopathies (Feb 2016 conference news). 

Tau PET Detects AD Dementia. AD patients (left) accumulate tau tangles throughout the brain (color shows intensity versus healthy control); this signal is much lower at the MCI stage (right). [Courtesy of Ossenkoppele et al., ©2018, American Medical Association. All rights reserved.]

To explore its diagnostic potential, Ossenkoppele and Hansson collaborated with researchers at the University of California in San Francisco and the Yonsei University College of Medicine, Seoul, South Korea. The study combined data gathered between 2014 and 2017 from the Swedish BioFINDER study at Lund and from patients who visited the Memory Disorder Clinic of Gangnam Severance Hospital in Seoul or the University of California at San Francisco Alzheimer Disease Research Center. Participants had an average age of 69. All underwent clinical and neuropsychological evaluations, MRI structural imaging, and flortaucipir imaging. Most were evaluated for brain amyloid by PET or cerebrospinal fluid Aβ42. In total, 179 people were diagnosed with AD dementia, 83 with mild cognitive impairment due to AD, and 43 with MCI from other causes. One hundred and sixty controls scored in the normal range on cognitive tests. The remaining 254 participants were diagnosed with other disorders, including Parkinson’s disease, progressive supranuclear palsy (PSP), behavioral variant frontotemporal dementia, dementia with Lewy bodies (DLB), corticobasal syndrome, primary progressive aphasia (PPA), and vascular dementia. The AD dementia and MCI due to AD groups were all amyloid-positive, but controls and non-AD disorders included a mix of amyloid-positives and -negatives. Tau PET was not used for diagnosis.

Compared with the control group, tau signal was elevated in people diagnosed with MCI due to AD, and higher yet in AD dementia (see image above). Flortaucipir signal did not differ significantly between any of the non-AD disorders and controls. The tau signal did discriminate well between AD dementia and non-AD cases. Because there are no recognized flortaucipir binding values that constitute tau positivity, the authors used two different methods to establish cutoffs for deeper analysis. In the first, they set the bar as two standard deviations above the average PET signal in the control group, or a standard uptake value ratio (SUVR) of 1.34. This cutoff distinguished AD from non-AD cases with 90 percent sensitivity and 91 percent specificity. In the second method, they selected an SUVR that divided AD patients from controls in one cohort, and then tested that cutoff in the other two cohorts. A cutoff of 1.27 derived in the Seoul cohort identified AD in the other two cohorts with 97 percent sensitivity and 88 percent specificity, while a BioFINDER-derived cutoff, also 1.27, gave 92 percent sensitivity and 85 percent specificity in the other two. The authors did not use the UCSF cohort to derive a cutoff because there were very few controls in that group.

While these cutoffs worked well for discriminating AD versus non-AD disorders as a group, they were not quite as effective for DLB and semantic variant PPA. A few individuals with these conditions did show moderate flortaucipir binding. Alzheimer’s-like plaques and tangles are often present in DLB, and in the case of svPPA, previous studies have reported flortaucipir signal in the temporal lobe, which may reflect non-specific binding to something other than tau (Makaretz et al., 2017). As a result, the SUVR cutoffs had lower specificity for these disorders—67 and 64 percent, respectively.

For all of these analyses, the authors assessed tracer signal in five prespecified regions of interest: entorhinal cortex, inferior temporal cortex, a temporal cortex composite, temporoparietal cortex, and neocortex. All the regions except entorhinal cortex discriminated well between AD dementia and non-AD disorders. The temporal cortex composite and temporoparietal cortex performed the best. Breaking it down further, tau signal in inferior and middle temporal cortex, medial temporal lobe, and posterior cingulate was the most predictive of AD. Cerebellar gray matter served as the reference region in all cases.

Tau PET Beats MRI. Tau binding in the inferior temporal cortex, temporal composite, and neocortex (blue, navy, and gold lines) discriminates AD better than do three MRI measures (red, green, and brown). [Courtesy of Ossenkoppele et al., ©2018, American Medical Association. All rights reserved.]

Notably, tau PET outperformed amyloid PET for distinguishing AD dementia from other disorders. The former identified AD over FTD, DLB, CBD, and other non-AD dementias with 90 percent accuracy, compared with 74 percent for amyloid PET. For dementia patients older than 69, tau PET had an even bigger advantage over amyloid, outperforming it by 20 percentage points or more. This is because amyloid plaques are so common at older ages, yet correlate poorly with symptoms, Hansson noted. Although tau pathology increases with age in normal controls, age-related tangles are usually confined to the entorhinal cortex. Thus, widespread tangles are much more predictive of symptomatic Alzheimer’s disease than amyloid is, he said.

Tau PET also performed better than MRI measures of hippocampal volume and cortical thickness, which varied from 63 to 75 percent accuracy (see image above). Combining MRI measures with tau PET did not improve the latter’s performance.

The picture changed at the MCI stage, however. For this group, tau PET distinguished AD from non-AD disorders with just 62 percent sensitivity, although still 91 percent specificity. Many people in this amyloid-positive MCI group were still tau-negative. “For early detection, amyloid is a better marker,” Ossenkoppele noted. Others agreed that the ideal diagnostic marker will depend on the disease stage. Tau imaging may work best to distinguish people with early AD from those with other dementias, while markers of neurodegeneration such as MRI might improve differential diagnosis in late-stage disease, suggested Beau Ances at Washington University in St. Louis.

Will tau PET be useful in the clinic? Researchers noted that the answer to this depends not just on these tracers being approved for such use, but also on whether insurance will reimburse for them. Amyloid imaging still needs to prove its clinical value to receive blanket coverage (Apr 2015 newsAug 2017 conference news). For tau imaging, one of the first questions will be whether it has better diagnostic accuracy than CSF biomarkers, the authors noted. They plan to analyze this next. Lumbar punctures are low-cost and covered by Medicare and Medicaid. Moreover, high CSF total tau can distinguish rapidly progressive tauopathies from AD, Ances said.

The authors are also following patients longitudinally to find out how the tau PET signal changes over time, and whether it predicts cognitive decline. In addition, participants in these studies have agreed to donate their brains at death. So far, six participants have come to autopsy, and their clinical diagnoses were confirmed in all cases. Postmortem evaluations will be particularly helpful for figuring out pathology in a handful of discrepant cases, Ossenkoppele noted. For example, 10 percent of amyloid-positive people diagnosed with AD dementia had no tau signal, while some diagnosed with non-AD disorders did.

Another priority is to examine tau tracers in a head-to-head study, Hansson said. Several tracers are in research use, and some of these may not be specific for AD tau (Aug 2016 conference newsMay 2018 news). For example, Piramal’s PI-2620 appears to bind to tau aggregates in PSP brain, suggesting it could have distinct clinical uses (Apr 2017 conference news).—Madolyn Bowman Rogers


  1. This is a large, well-done study of 719 patients overall and 254 with suspected non-AD disorders from a diverse group of three sites around the world. The tau-PET data is supplemented by amyloid-PET data. The sensitivity/specificity (90/91 percent and 97/88 percent by two different measures) were impressive and substantially better than volumetric MR measures. The study produced volumes-of-interest and cutoffs that have extrinsic merit as well. While the diagnostic discrimination was impressive for the main purpose of distinguishing non-AD disorders from clinical AD, the performance was not as impressive in MCI due to AD (i.e., Aβ + MCI). This is likely a reflection of the later nature of tau deposition in the pathophysiological sequence of events in AD and points to the fact that different biomarkers have different “sweet spots” along this sequence of events.

Make a Comment

To make a comment you must login or register.


News Citations

  1. On Multiple Marker Analysis, Tangles Track Best With Functional Decline
  2. At HAI, Researchers Explore Diagnostic Potential of a Tau Tracer
  3. $100M IDEAS: CMS Blesses Study to Evaluate Amyloid Scans in Clinical Practice
  4. In Clinical Use, Amyloid Scans Change Two-Thirds of Treatment Plans
  5. Improving Tau PET: In Search of Sharper Signals
  6. Move Over, Flortaucipir? New Tau Tracers Tested in People
  7. Next-Generation Tau PET Tracers Strut Their Stuff

Paper Citations

  1. . Flortaucipir tau PET imaging in semantic variant primary progressive aphasia. J Neurol Neurosurg Psychiatry. 2017 Oct 6; PubMed.

Further Reading

Primary Papers

  1. . Discriminative accuracy of [18F] flortaucipir positron emission tomography for Alzheimer disease vs other neurodegenerative disorders. JAMA. 2018 Sep 18.