Optimizing the interval between G-CSF therapy and FDG PET imaging

Objectives To define the interval between granulocyte colony stimulating factor (G-CSF) administration and FDG PET imaging that limits interference by stimulated marrow.

Methods FDG PET scans performed in patients with Ewing sarcoma or rhabdomyosarcoma were retrospectively reviewed by two reviewers blinded to G-CSF administration. Reviewers subjectively scored marrow activity (1-4) with regard to interference with clinical interpretation and measured standardized uptake values in the spine.

Results 44 FDG PET exams were reviewed in 22 patients, 18 of whom received pegfilgrastim and 4 of whom received filgrastim. There was substantial agreement in scoring between reviewers (α=0.633). Stimulated marrow was scored as potentially or definitely interfering in 19/44 scans (43.2%). Of the 18 patients treated with pegfilgrastim, the median time between G-CSF administration and imaging was 10 days (range:1-19 days) for scans with potential or definite marrow interference and 19.5 days (range:7-55 days) for scans without marrow interference (score≤2). No scan >10 days after pegfilgrastim administration had marrow activity considered definitely interfering. Of the 4 patients treated with filgrastim, subjective interference by marrow stimulation was present in only 1 scan (1 day after G-CSF). SUVmax in the spine was ≥3 (high enough to interfere) in 17/44 scans (38.6%). Median time between pegfilgrastim administration and imaging in patients with SUVmax ≥3 in the spine was 16 days (range:1-20 days). In patients with SUVmax <3, median time between pegfilgrastim administration and imaging was 10 days (range:2-55 days). The only patient who received filgrastim that had SUVmax ≥3 in the lumbar spine was imaged on the day after filgrastim administration.

Conclusions An interval of 19-20 days between pegfilgrastim administration and FDG PET scanning appears sufficient to limit interference by stimulated marrow. In patients treated with filgrastim, shorter intervals may suffice but additional study is needed to confirm this.