Experimental design: Retrospective cohort study in 91 patients with prostate-specific antigen (PSA) relapse following prostatectomy, imaged with 2-[18F]fluoro-2-deoxyglucose positron emission tomography (FDG-PET) in a tertiary care cancer center between February 1997 and March 2003. Comparison was made with magnetic resonance imaging (n = 64), bone scan (n = 56), and computed tomography (n = 37). The standard of reference included biopsy or clinical and imaging follow-up. We calculated sensitivity and specificity of PET and correlated PET findings with PSA values, other clinical parameters, and conventional imaging, when available.
Results: PET was true positive in 28 of 91 (31%) patients, showing isolated disease in the prostate bed (n = 3) or metastatic disease with (n = 2) or without (n = 23) simultaneous disease in the prostate bed. In detail, PET identified lesions in the prostate bed (n = 5, all true positives), bones (n = 22; 20 true positives, 2 false positives), lymph nodes (n = 7; 6 true positives, 1 likely false positive), and one liver metastasis. Mean PSA was higher in PET-positive than in PET-negative patients (9.5 +/- 2.2 versus 2.1 +/- 3.3 ng/mL). PSA of 2.4 ng/mL and PSA velocity of 1.3 ng/mL/y provided the best tradeoff between sensitivity (80%; 71%) and specificity (73%; 77%) of PET in a receiver operating curve analysis. Combination with other clinical parameters in a multivariate analysis did not improve disease prediction. There were only two patients in whom other imaging studies showed isolated local recurrence or metastatic disease.
Conclusions: FDG-PET detected local or systemic disease in 31% of patients with PSA relapse referred for this test. There is a link to tumor burden and tumor biology in that the probability for disease detection increased with PSA levels.