The aim of this study was to assess the reliability of 2'-methoxyphenyl-(N-2'-pyridinyl)-p-18F-fluoro-benzamidoethylpiperazine (18F-MPPF) PET binding parameter's quantification via a test-retest study over a long-term period.
Ten healthy volunteers underwent 2 dynamic 18F-MPPF PET scans in an interval of 6 mo. As a methodologic control, 10 simulated datasets, including interindividual functional and anatomic variabilities, were also used to assess the measurement variations in the absence of intraindividual variability. Indices of tracer binding were computed using 2 different models: (a) the simplified reference tissue model (SRTM) and (b) the Logan graphical model. The SRTM allows computing the binding potential (BP) index and plasma-to-brain transport constants (R1, k2). The Logan model evaluates the distribution volume (DV). For both methods, cerebellum was taken as the reference region. From both models, binding indices were calculated with time-activity curves extracted from regions of interest, on one hand, and for each voxel to perform parametric images on the other hand.
Reliability indices--that is, bias, variability, and intraclass correlation (ICC)--indicated a good reproducibility: the BP percentage change in mean between test and retest is close to 1% in rich regions and 2% in poor regions. The typical error is around 7%. Mean ICC is over 0.70. The DV percentage change in the mean is +/-2.5%, with a typical error close to 6% and an ICC over 0.60.
Our results show a good reliability, with a reasonable level of intraindividual biologic variability that allows crossover studies with 18F-MPPF in which small percentage changes are expected between test and retest measurements, in group studies and for single subject assessment.