Program Information
Anatomic Heterogeneity in Metastatic Castrate Resistant Prostate Cancer
A Roth1*, S Harmon1 , T Perk1 , G Liu1 , R Jeraj1,2 , (1) University of Wisconsin, Madison, WI, (2) University of Ljubljana, Ljubljana, Slovenia
Presentations
WE-F-201-10 (Wednesday, August 2, 2017) 1:45 PM - 3:45 PM Room: 201
Purpose: Bone metastatic prostate cancer (mPC) is characterized by tumor burden throughout the skeleton, but the impact of spatial distribution of disease (heterogeneity) on clinical assessment hasn’t been explored. This study assesses clinically relevant aspects of baseline anatomic disease by correlating them to Progression-Free Survival (PFS) and comparing to similar global metrics.
Methods: Fifty-four mPC patients received [F-18]-NaF PET/CTs before treatment with chemotherapy (N=16) or androgen-receptor inhibitors (N=38). An automatic PET/CT technique identified and segmented 3,466 bone lesions. The skeleton was divided into 11 anatomic regions. Number of lesions (NL), SUVtotal, SUVmax, SUVmean, and number of diseased regions (NR) were calculated for each patient from all lesions (global) and for each region. The regional binary presence of disease was also calculated. Cox proportional-hazard analyses were performed to relate global and regional metrics with PFS. Independent predictors of PFS were determined using multivariate Cox proportional-hazards regression analysis.
Results: PFS was captured for all patients (median 7.3, range 0.07-30.3 months). NL was previously shown to significantly predict PFS (p<0.008, HR=1.42) and six regions (cervical spine, thoracic spine, sacrum, sternum, arms, and pelvis) had similarly performing significant hazard ratios. Global SUVtotal wasn’t a significant predictor of PFS, but SUVtotal in pelvic lesions was (p<0.003, HR=1.49). Global SUVmax significantly predicts PFS (p=0.008, HR=1.51) and performed similarly in arm lesions. NR significantly predicts PFS (p<0.02, HR=1.48). Binary presence of disease in the cervical spine (p=0.04, HR=1.38) and arms (p<0.03, HR=1.45) were significant predictors. Two anatomic measures (sacral and arm NL) remained significant in the multivariate model exhibiting their independence of global measures (baseline PSA, SUVtotal, and SUVmean).
Conclusion: Univariate analysis of anatomic metrics showed some perform as well as global metrics in patients with extensive disease. The combination of anatomic and global metrics in a multivariate model revealed that anatomic metrics may provide additional information.
Contact Email: