On the validity of reference tissue models for analysing (R)-[11C]PK11195 studies in traumatic brain injury

Introduction: Activated microglia in the brain can be imaged in vivo using (R)-[11C]PK11195 and PET. In theory, this activation can be quantified using a reference tissue approach, provided that an appropriate reference tissue is available. The purpose of the present study was to assess whether a reference tissue approach is valid for analysing (R)-[11C]-PK11195 studies in traumatic brain injury (TBI). Methods: Five TBI patients were included. About 10 days after TBI, subjects underwent a 60 minutes 3D dynamic (R)-[11C]PK11195 brain scan, including on-line and manual arterial sampling. Volumes of interest (VOI) were drawn on a co-registered T1-weighted MRI scan for frontal, occipital, parietal, temporal and cingulate cortex, and for thalamus and pons. In addition, VOI were defined in and near (penumbra) traumatic brain areas. A cerebellum VOI was used as reference tissue. Binding potential (BP-PI), volume of distribution (Vd), distribution volume ratio (DVR; Vd normalised to Vd cerebellum), blood volume fraction (Vb) and K1/k2 were derived for all VOI using a plasma input model. In addition, binding potential (BP-SRTM) was obtained using the simplified reference tissue model. Finally, simulations were performed to assess effects of varying Vb and K1/k2, the latter simulating blood-brain barrier (BBB) disruption, between reference and target tissues. Results: Plasma input analysis revealed a strong correlation (R2=0.81, slope=3.1) between Vd and K1/k2 over all regions and subjects, whilst no correlation was observed between BPPI and K1/k2. Furthermore, BP-PI did not significantly differ from those seen in healthy subjects. These findings indicate that, in TBI, variation in Vd is almost exclusively determined by variation of the K1/k2 ratio. In addition, both DVR and BP-SRTM correlated well (R2=0.74) with variation of K1/k2 between target and reference tissues. DVR and BP-SRTM therefore primarily reflect changes K1/k2 rather then specific binding. Areas in and near TBI (penumbra) showed large increases in DVR and BP-SRTM on parametric images, but this (false) positive signal could simply be explained by a corresponding large increase in K1/k2 compared with that of the reference region. On the other hand BP-PI showed no evidence for increased binding. Simulations confirmed that DVR and BP-SRTM may provide false positive Results: in case of changes (increases) in (bidirectional) BBB transport. For BP-SRTM this was also the case for increased Vb. Conclusion: Increases in (R)-[11C]PK11195 DVR and BP-SRTM, supposedly representing specific binding, in and near traumatic areas in TBI patients can be explained by differences in BBB transport and, in case of BP-SRTM, by changes in Vb. The large variation of K1/k2 across the brain of TBI patients prohibits the definition of an appropriate reference tissue. Therefore, in TBI patients, (R)-[11C]PK11195 studies should only be analysed using plasma input models.