A methodology for the extraction and analysis of real-world radiographic and electronic medical record data to reconstruct treatment arms of historical prostate cancer clinical trials.

Abstract

5033 Background: Real-world evidence (RWE), including synthetic comparator arms created from historical real-world data (RWD), has the potential to support the safety and efficacy evaluation of new medical products. However, many available RWD sources lack the details necessary to reliably identify patients comparable to clinical trial cohorts or to assess essential oncologic efficacy endpoints. This project demonstrates the ability to extract and analyze RWD to identify patients matching eligibility criteria to four historical clinical trials in metastatic castration-resistant prostate cancer (mCRPC), and calculate outcome measures. Methods: A total of 5,741 patients treated for prostate cancer at multiple institutions (2010-2020) were analyzed in two cohorts using data extracted from the EMR, Tumor Registry, Oncology Information System, and Picture Archiving and Communication System. Of 3,486 patients with prostate cancer in Cohort 1, 422 mCRPC patients were identified: those treated with ADT who achieved castration-level testosterone ( < 50 ng/dL), had evidence of metastatic disease, and exhibited rising PSA (PCWG2). These patients were further matched to four historical clinical trial treatment arms (COU-AA-301: 49, COU-AA-302: 143, AFFIRM: 30, PREVAIL: 79), based on prior chemotherapy and receipt of Abiraterone or Enzalutamide. Overall survival (OS) and time to skeletal related events (SRE) (pathological fracture, spinal compression, surgery to bone, and radiotherapy to bone) were calculated based on diagnosis and procedure codes using the Kaplan-Meier (KM) Estimator. Of 2,255 patients with prostate cancer in Cohort 2, 101 patients received Abiraterone or Enzalutamide and 59 patients had sufficient baseline and follow-up imaging to be scored. Radiographic progression-free survival (rPFS) was calculated from the start of treatment to the time of progression (RECIST 1.1) or loss to follow-up using the KM estimator. Results: In Cohort 1, median OS was 37.7 months (95% CI: 31.5-NR), and median time to SRE was 17.9 months (13.5-22.6). Median OS per patient cohort matched to historical trial treatment arm was COU-AA-301: 23.7 months (10.7-NR), COU-AA-302: 45.9 months (34.9-NR), AFFIRM: 35.3 months (6.34-NR), PREVAIL: 41.5 months (21.9-NR). In Cohort 2, median rPFS was 37.2 months (13.3-NR). Conclusions: The methodology employed in this analysis not only successfully identified a cohort of RWD patients similar to clinical trial-defined patients, but also curated sufficiently reliable data to calculate essential endpoints (e.g., rPFS). At scale, this methodology can be used to generate RWE, including synthetic comparator arms to support clinical trials with radiographic endpoints.