(P-270) Real-world Algorithms in Defining Castration Resistant Prostate Cancer in Retrospective Observational Studies

Background: Prostate cancer is the second common cancer among men. Castration resistant prostate cancer (CRPC) occurs with high prostate specific antigen (PSA) level despite androgen deprivation therapy (ADT). Defining CRPC in real-world data (RWD) sources is challenging as it is dependent on treatment response and not solely on tumor characteristics.

Objectives: The objective was to characterize operational definitions of CRPC in the existing literature using RWD sources.

Methods: A systematic literature review was conducted using Embase, PubMed, and Medline according to PRISMA guidelines (2010-2020). Peer-reviewed articles of retrospective RWD studies on CRPC patients were included. Case reports, review articles, non-English and single-institution chart review studies were excluded. Operational definition and algorithms used to define CPRC were characterized by study design, data source, sample size, metastatic vs. non-metastatic, country/region, and study year.

Results: Among 1815 articles identified, 96 studies were eligible in the final review. The included studies were based in 18 different countries, with more than half (n=62) conducted in North America. The majority (n=41) were electronic medical records (EMR) followed with claims data (n=18) and registries (n=20). The rest used multiple sources of data. 45 articles focused exclusively on metastatic CRPC. CRPC-related medication (n=89) were the most frequently used to define CRPC followed by diagnostic codes (n=64), PSA levels (n=43), CRPC-related procedure, e.g. surgical/medical castration (n=24). In most cases, combination of information was used. In claims, 14 studies used a combination of diagnostic coding with CRPC medications. There were variations in CRPC medication used in the algorithms over time. In other cases, the combination of medication and diagnosis codes were used along with CRPC-related procedure or PSA levels. When diagnostic codes were used, International Classification of Diseases (ICD) ninth or tenth revision coding was the most commonly utilized.

Conclusions: The treatment landscape in CRPC is evolving rapidly with potential novel treatments on the horizon. This literature review identified several algorithms from various RWD sources. Future studies are needed to validate these real-world algorithms in defining CRPC.