Repeatability of ^sup 18^F-FDG PET in a Multicenter Phase I Study of Patients with Advanced Gastrointestinal Malignancies

^sup 18^F-FDG PET is often used to monitor tumor response in multicenter oncology clinical trials. This study assessed the repeatability of several semiquantitative standardized uptake values (mean SUV [SUV^sub mean^], maximum SUV [SUV^sub max^], peak SUV [SUV^sub peak^], and the 3-dimensional isocontour at 70% of the maximum pixel value [SUV^sub 70%^]) as measured by repeated baseline ^sup 18^F-FDG PET studies in a multicenter phase I oncology trial. Methods: Double-baseline ^sup 18^F-FDG PET studies were acquired for 62 sequentially enrolled patients. Tumor metabolic activity was assessed by SUV^sub mean^, SUV^sub max^, SUV^sub peak^, and SUV^sub 70%^. The effect on SUV repeatability of compliance with recommended image-acquisition guidelines and quality assurance (QA) standards was assessed. Summary statistics for absolute differences relative to the average of baseline values and repeatability analysis were performed for all patients and for a subgroup that passed QA, in both a multi- and a single-observer setting. Intrasubject precision of baseline measurements was assessed by repeatability coefficients, intrasubject coefficients of variation (CV), and confidence intervals on mean baseline differences for all SUV parameters. Results: The mean differences between the 2 SUV baseline measurements were small, varying from -2.1% to 1.9%, and the 95% confidence intervals for these mean differences had a maximum half-width of about 5.6% across the SUV parameters assessed. For SUV^sub max^, the intrasubject CV varied from 10.7% to 12.8% for the QA multiand single-observer datasets and was 16% for the full dataset. The 95% repeatability coefficients ranged from -28.4% to 39.6% for the QA datasets and up to -34.3% to 52.3% for the full dataset. Conclusion: Repeatability results of doublebaseline ^sup 18^F-FDG PET scans were similar for all SUV parameters assessed, for both the full and the QA datasets, in both the multi- and the single-observer settings. Centralized quality assurance and analysis of data improved intrasubject CV from 15.9% to 10.7% for averaged SUV^sub max^. Thresholds for metabolic response in the multicenter multiobserver non-QA settings were -34% and 52% and in the range of -26% to 39% with centralized QA. These results support the use of ^sup 18^F-FDG PET for tumor assessment in multicenter oncology clinical trials. [PUBLICATION ABSTRACT]