Presentation (CTI)

OBJECTIVES: To establish imaging-anchored real-world evidence (RWE) as an alternative to claims-based approaches in prostate cancer research, and to demonstrate the feasibility of constructing longitudinal cohorts from radiology report-derived findings linked to administrative claims.
METHODS: We analyzed a subset of Avandra’s federated network of U.S. health systems, linking Avandra’s proprietary DICOM-indexed radiology reports to licensed administrative medical and pharmacy claims at patient level. Prostate cancer patients were identified using ICD-10 codes (C61; D075). Imaging cohorts were defined by the presence of prostate/pelvic MRI or PET reports; report text was queried for clinical indicators (PI-RADS, PSA, PSMA, Gleason). Longitudinal patients were defined as those with ≥2 imaging studies separated by ≥30 days. Linked claims provided aggregate measures of treatment patterns (J-codes for ADT, chemotherapy), biochemical recurrence (ICD-10 R97.21), and metastatic progression (ICD-10 C79.51). Analyses were performed by Avandra on aggregated cohort data; no patient-level results are reported.
RESULTS: Within the analyzed subset, 14,301 patients had linked prostate/pelvic MRI or PET radiology reports. Of these, 10,115 (70.7%) had report text containing extractable clinical indicators, and 3,725 unique patients met longitudinal imaging criteria (≥2 studies ≥30 days apart), forming a serial-imaging cohort linked to claims-derived treatment and outcome measures. The linked, aggregate infrastructure enables characterization of imaging-modality distribution, longitudinal follow-up duration, treatment-pattern prevalence, and report-derived biomarker extraction rates — metrics that inform RWE study design.
CONCLUSIONS: Radiology reports capture what claims-based data cannot: lesion-level findings, biomarker trends, and imaging-defined progression. Linking these imaging-derived findings to claims data adds treatment and outcome context, completing the full patient picture. This challenges the status quo—prostate cancer RWE grounded in imaging, not billing codes—and extends to any disease where imaging informs clinical decisions. This approach enables label expansion, comparative effectiveness, and external control arm studies grounded in imaging-defined populations.