AD's p-tau217 Results and the Road Ahead for Parkinson's PET tracers

Ideas and Inspiration
,
1

A major highlight of the recent AAIC 2026 (London, July 12–15), was the pooled longitudinal analysis of 2,684 cognitively unimpaired participants from major Alzheimer’s disease research cohorts including A4/LEARN, HABS, ADNI, WRAP, and HABS-HD by Buckley et al 2026 (JAMA). The study found that cognitively healthy older adults with very high p-tau217 had an approximately 78% risk of developing cognitive impairment over 10 years, and about a 1-in-3 chance within five years, while those with slightly elevated p-tau217 had an absolute risk of 15% and 45% over 5 and 10 years, respectively. A separate real-world study presented at the same conference, covering more than 1,300 patients and 165 physicians, found that primary care physicians’ diagnostic accuracy rose from 65% before receiving blood test results to 93% after, while specialists’ accuracy rose from 74% to 89%.

What made this possible was infrastructure, not just the biomarker itself: p-tau217 could only be validated at this scale because ADNI, HABS, and A4/LEARN had already spent years to decades collecting harmonized, longitudinal data on the same participants. Parkinson’s disease (PD) is attempting the same “detect before symptoms” goal but starting from a different infrastructure position. To date, no equivalent validated blood biomarker has been identified for PD; current PD imaging research is instead focused on developing PET tracers for alpha-synuclein, the misfolded protein underlying PD pathology. Progress here has been recent and incremental. An earlier Merck candidate, MK-7337, bound alpha-synuclein with a dissociation constant of 0.4 nM but was not advanced further due to off-target binding; its successor, MK-0947, was designed to remove that off-target binding and to be labeled with fluorine-18, with a trial starting later this year and readout expected in 2026. Separately, a Harvard-designed tracer, SY08, showed evidence of binding more tightly to the brainstem of sporadic PD patients than controls in early human PET scans — notable because earlier tracers such as AC Immune’s ACI-12589 detected alpha-synuclein in multiple system atrophy but failed to detect deposits in sporadic PD patients, who carry a lower burden of pathology.

Structurally, this is the same gap the bolded sentence points at: the p-tau217 findings drew on cohorts running for years to decades, enabling a pooled sample large enough to produce prognostic estimates. The alpha-synuclein PET tracer programs, by contrast, are currently running as separate, smaller efforts, Merck, AC Immune, MODAG, and the Harvard group each with their own candidate and cohort, without a shared reference dataset or common acquisition protocol reported to date. Larger strides in the field may require pooling these efforts together, following the cohort-harmonization model that made the AAIC 2026 p-tau217 results possible.

2

Amgad, I was at AAIC too bummer I didn’t run into you! I did see @malosco though :slight_smile: Also, always admire people who go to conferences and put together such neat summaries. My conference takes are always “there’s some new marker but can’t remember what it actually was” because I have zero memory skills… I’m basically there to build up my network and enhance my photo library with so many new slides/posters I take pics of and then forget :rofl:

3

So great to see you @ecebayram!

4

Very interesting summary of the current state-of-the-art; thanks @AmgadDroby! I tend to agree that infrastructure to gather, share, and harmonize large datasets is the way to make significant progress in heterogenous diseases like PD, which requires a lot of sustained collaboration.

One of the most substantial challenges is in the harmonization of data across different sites, investigators, and studies, especially if they were not initially designed to be integrated. Generally speaking, the larger the dataset in terms of participants, the fewer the datapoints in terms of biomarkers or read-outs, in order to ensure a combined dataset is comparing like with like.

In my opinion, one of the major strengths of the PPMI dataset is that it is broad in biomarkers, deep in longitude, and wide in cohort numbers. Continued enrollment and protocol updates help to keep expanding on these axes, too.

But I get the sense you are suggesting that there could be harmonization of data across studies, to truly supercharge the numbers for PD. Do you have any studies in mind that could consider combining forces? Or do you think new studies are required that begin with the assumption of integration/harmonization with others?

5

@vcatterson I do agree that PPMI is currently the broadest and richest dataset in the field. Yet, several independent cohorts running alpha-synuclein PET trials each use a different tracer, different reference regions, and different recruitment criteria, so a full merge of that data isn’t really feasible without losing what makes each one useful.
That said, I think there’s a middle ground short of full integration: closer coordination and partial integration between these independent trials and ongoing large cohorts like PPMI, so existing data and infrastructure can be validated and reused rather than each trial building its own comparator from scratch. Concretely, that could look like:

  • Shared reference/comparator subset: a small group of PPMI participants gets scanned with each new tracer, giving every trial a common, already-deeply-phenotyped comparison group instead of a bespoke control cohort.
  • Common analysis conventions: even without merging raw imaging data, agreeing on shared reference regions and quantification methods (e.g., SUVR thresholds) would let results across tracers be compared indirectly, similar to how ADNI’s imaging core enabled cross-calibration of different amyloid/tau tracers.
  • Ancillary-study model: new tracer programs run as PET sub-studies nested within PPMI (or AMP PD) rather than as fully standalone trials, so they inherit PPMI’s existing clinical, genetic, and fluid-biomarker data on the same participants.
    I think any of these would preserve each program’s independence while still letting the field build toward the kind of pooled, cross-validated evidence base that made the p-tau217 result possible.
6

These seem like really great suggestions! I agree that it is much easier to think about these studies as “federated” instead of fully integrated, in which case identifying some reference participants who are scanned with all tracers provides the comparators across studies. And the depth and breadth of PPMI makes it the ideal candidate to provide those comparator participants. Here’s hoping that we can influence future studies in this direction!