Model. Reversible two-tissue compartment model with 5% blood volume, the model Barret et al. found fits flortaucipir over 210 minutes (the one-tissue model fit poorly). Rate constants are their group means from arterial-input scans (Table 2, 130 min of data, 16 people). k3 and k4 were the least well determined (errors up to about 25%); their ratio k3/k4 is more reliable. Type your own values into K1–k4 to test other estimates.
Assumption. The paper shows the measured plasma curve only as a figure, so this uses a generic plasma shape (Feng et al. 1993) times a parent fraction set to the reported 17.3% at 90 minutes. With the Alzheimer's · lateral temporal defaults it gives a +16% SUVR rise from 80–100 to 110–130 min and +36% by 160–180 min; Barret reported about +20% and +30%. Treat absolute values as illustrative.
What the numbers mean. True DVR is VT(target) / VT(reference) at equilibrium; BPND = DVR − 1. In Alzheimer's the SUVR keeps rising for over three hours (no equilibrium), so it depends on scan timing and on blood flow even when tau is unchanged; that is why trials fix the window. SUVRs also depend on the target and reference regions and the processing pipeline, so values from different studies are not interchangeable; the CenTauR scale (Villemagne et al. 2023; Leuzy et al. 2024) is the effort to put tracers on one scale.
What tau PET cannot show. It is an aggregate, tracer-dependent signal: it does not identify which cells or cell compartments hold tau, or which molecular form. Off-target binding (for flortaucipir, basal ganglia and choroid plexus; Barret found different putamen kinetics, with k4 about double) and partial-volume loss in atrophied cortex both distort regional values.
Clinical context. In the 2024 Alzheimer's Association criteria, tau PET is a later-changing Core 2 biomarker used for staging: stage A (amyloid positive, tau PET negative), B (tau limited to medial temporal lobe), C (moderate neocortical uptake), D (high neocortical uptake), without fixed numeric cut-points (Jack et al. 2024). Among 1,325 cognitively unimpaired people, progression to mild cognitive impairment was much more likely with amyloid plus neocortical tau (HR 19.2, 95% CI 10.9–33.7) or medial temporal tau (HR 14.6, 8.1–26.4) than with neither (Ossenkoppele et al. 2022). TRAILBLAZER-ALZ 2 (1,736 people) enrolled by tau PET level; donanemab slowed decline at 76 weeks in the low/medium tau group (iADRS change −6.02 vs −9.27 on placebo; Sims et al. 2023). In cognitively unimpaired people with high amyloid (≥ 68 Centiloid), flortaucipir signal in an Alzheimer's meta-region rose about 0.02–0.03 SUVR per year faster than in lower-amyloid groups (Knopman et al. 2021).
Sources. Barret O et al. J Nucl Med 2017;58:1124–31 · Feng D et al. Int J Biomed Comput 1993;32:95–110 · Jack CR Jr et al. Alzheimers Dement 2024;20:5143–69 · Ossenkoppele R et al. Nat Med 2022;28:2381–7 · Sims JR et al. JAMA 2023;330:512–27 · Knopman DS et al. JAMA Neurol 2021;78:217–28 · Villemagne VL et al. Alzheimers Dement (Amst) 2023 · Leuzy A et al. 2024 (CenTauR). Background reading: compartmental modeling in PET (Radiology Key). Teaching tool, not for clinical or research decisions.