Reproducibility and clinical correlations of post-treatment changes on CT of prostate cancer bone metastases treated...

The objective of this study was to determine whether, in patients with prostate cancer (PCa) bone metastases receiving chemotherapy, early post-treatment changes on CT are reproducible and associated with clinical outcomes.

Blinded to outcomes, two radiologists with 1 year and 5 years of experience independently reviewed CTs obtained before and 3 months after chemotherapy initiation in 38 patients with bone metastases from castration-resistant PCa, recording the size, matrix and attenuation of ≤5 lesions; presence of new lesions, extraosseous components, periosteal reactions and cortical thickening; and overall CT assessment (improved, no change or worse). Kappa statistics were used to assess inter-reader agreement; the Kruskal-Wallis test and Cox regression model were used to evaluate associations.

Inter-reader agreement was low/fair for size change (concordance correlation coefficient=0.013), overall assessment and extraosseous involvement (κ=0.3), moderate for periosteal reaction and cortical thickening (κ=0.4-0.5), and substantial for CT attenuation (κ=0.7). Most metastases were blastic (Reader 1, 58%; Reader 2, 67%) or mixed lytic-blastic (Reader 1, 42%; Reader 2, 34%). No individual CT features correlated with survival. Readers 1 and 2 called the disease improved in 26% and 5% of patients, unchanged in 11% and 21%, and worse in 63% and 74%, respectively, with 64% interreader agreement. Overall CT assessment did not correlate with percentage change in prostate-specific antigen level. For the more experienced reader (Reader 2), patients with improved or unchanged disease had significantly longer median survival (p=0.036).

In PCa bone metastases, interreader agreement is low in overall CT post-treatment assessment and varies widely for individual CT features. Improved or stable disease identified by an experienced reader is statistically associated with longer survival.