Quantitative Imaging-Biomarker in Prostate Cancer. Validation of Automated Bone Scan Index in the Context of Radiographic Disease Progression in Metastatic Castration Resistant Prostate Cancer

Background: Metastatic castration resistant prostate cancer (mCRPC) is a bone dominant lethal disease. Progression in bone scan of mCRPC disease is assessed with the Prostate Cancer Working Group (PCWG) two and three criteria, which relies on the appearance of two or more new lesions. There is an unmet need for a quantitative assessment that can build on the current progression criteria. This thesis has analytically and clinically validated automated Bone Scan Index (BSI) as an imaging-biomarker in the context of disease progression in mCRPC. Method: Analytical Validation – In a novel approach, a true analytical standard was developed using the XCAT phantoms and the SIMIND Monte Carlo program to simulate bone scan with pre-defined tumor burden. The simulated and patient bone scans evaluated the analytical and pre-analytical performance characteristics of BSI. Pearson’s correlation (r), linear regression, Coefficient of Variation (CV), Cohen’s kappa agreement (κ) was used to assess the performance characteristics of the automated BSI. Clinical Validation – The clinical utility of automated BSI was evaluated as a predictor of overall survival (OS) in mCRPC. Subsequently, the BSI increase at key PCWG landmarks was evaluated as the total quantitative increase in disease burden while waiting to meet the progression criteria. The time to BSI increase was associated with OS. Concordance index (C-index) and its standard error (SE) was used to evaluate the discriminatory strength of the automated BSI in predicting OS. Kendall’s Tau correlation was used to associate time to BSI increase with survival. Analytical Validation – The simulation study demonstrated high BSI accuracy (r=0.99; 95%CI: 0.99–0.99; p<0.0001) and precision (CV<20%). In the test/re-test study, the upper noise threshold of the BSI reproducibility was at 0.30. The inter-observer agreement, among independent readers, in assessing treatment-follow-up BSI change was strong (κ=0.96, P<0.0001). The pre-analytical studies demonstrated that the BSI accuracy and reproducibility were dependent on scanning speed, but not on the vendor-specific gamma-camera. In the subsequent clinical study, BSI was predictive of OS (C-index 0.72) and adding BSI to the blood-based model significantly improved the C-index from 0.67 to 0.72, p=0.017. Additionally, the change in BSI demonstrated an additive clinical value to that of relative change in the PSA, C-index improved from 0.73 to 0.77, p=0.041. Finally, significant BSI increase, relative to week 12 (1st follow-up scan), was observed at the meeting of PCWG criteria (median relative increase=109%[IQR:40-377%]; absolute increase=1.22[IQR:0.65-2.49]). Thresholds for the absolute increase in BSI from week 12 (1st follow-up) were explored for the time to BSI progression. The correlation of time to BSI progression with survival was observed to increase with the increase in magnitude of BSI thresholds. The time to increase in BSI, above the reproducibility threshold, was found to be associated with time to OS (Kendall’s Tau=0.41, p<0.0001). Conclusion: In conclusion, the thesis has demonstrated that the automated BSI is an accurate, precise and reproducible assessment of metastasis in bone scan images, and that it can build on the current progression criteria by quantitating the increase of both the existing lesions and the contribution of new lesions. The work here has laid a foundation towards the goal of qualifying the increase in automated BSI as a continuous quantitative radiographic progression biomarker.