ABSTRACT

Phosphodiesterase-4 (PDE4) is the primary enzyme responsible for hydrolysis of cyclic adenosine monophosphate (cAMP) in the brain. The PDE4B subtype has been implicated in inflammatory processes. The PDE4B-preferential PET radioligand [¹⁸F]PF-06445974 has shown promise for imaging PDE4B in humans but appears susceptible to time-dependent increases in total distribution volume (VT), raising concerns about potential radiometabolite effects. This study sought to determine a robust and practical imaging protocol for quantifying [¹⁸F]PF-06445974 in the human brain.

Methods: Twenty-seven healthy volunteers were studied across three cohorts with dynamic PET imaging: a 240-minute scan group (n=11) to assess VT time stability, a 120-minute test-retest group (n=11) to evaluate reproducibility, and a 90-minute baseline scan group (n=5). VT was estimated using both a two-tissue compartment model (2TCM) and Logan graphical analysis with arterial input function. Scan-duration effects on VT stability were evaluated using progressively truncated datasets. Test-retest reliability was assessed using intraclass correlation coefficients (ICC) and absolute test-retest variability (aTRV) for both region of interest (ROI) and voxel-wise analyses.

Results: During 240-minute acquisitions, regional VT values increased markedly after 120 minutes, with mean VT at 240 minutes 30% higher than at 120 minutes, suggesting late-time signal contamination most consistent with radiometabolite accumulation. In contrast, VT estimates derived from 60-, 90-, and 120-minute datasets exhibited substantially reduced drift and preserved regional distribution patterns. In the test-retest cohort, Logan-derived VT demonstrated good reproducibility for 120-minute scans (mean aTRV≈10%; ICC≈0.78), with similar ROI-level reproducibility observed for 60- and 90-minute acquisitions. Logan analysis demonstrated lower parameter uncertainty of VT estimation and greater ICC than 2TCM, particularly for 60-minute datasets. Voxel-wise reliability improved with increasing scan duration, reflecting reduced measurement noise, but longer acquisitions may also increase susceptibility to radiometabolite contamination.

Conclusion: When analyzed with Logan graphical methods, regional VT of [¹⁸F]PF-06445974 was reliably quantified using scan durations of  60-90 minutes. Although voxel-wise reliability improved with 120-minute acquisitions, longer scans may increase susceptibility to radiometabolite-related VT inflation. For most applications that rely on predefined regional outcome measures, shortened protocols (60-90 minutes) provide a practical and robust balance between quantitative stability and minimization of late-time bias.