PRM20 - Chi-Square Automatic Interaction Detection (CHAID) Analysis: Understanding and Informing Treatment Choice of Second-Line Metastatic Urothelial Cancer Therapy

Abstract

To demonstrate the use of the Chi-square Automatic Interaction Detection (CHAID) analysis to identify drivers of second-line (2L) therapy for metastatic urothelial cancer (mUC). This study was a retrospective, non-interventional study conducted among n=295 physicians and n=1397 patients in Australia (n=18; n=71), Canada (n=25; n=112), France (n=52; n=247), Germany (n=49; n=244), Italy (n=50; n=248), Spain (n=50; n=241), and the UK (n=51; n=234). Patient characteristics, treatment patterns, and outcomes data were collected from medical charts of the 5 most recent patients who began and stopped 2L mUC treatment. 2L was defined as treatment after progression/recurrence after first-line (1L) treatment, or recurrence ≤12 months of neoadjuvant/adjuvant treatment. CHAID analysis was used to examine drivers related to receiving platinum versus non-platinum based 2L treatment. CHAID and other statistical anslyses were performed using SPSS v23. Physicians aged 45.0±8.2 years specializing primarily in oncology (93%) reported on data for 2L mUC patients. Most patients were male (74%) and aged 63.5±9.2 years at mUC diagnosis. Overall, 23.5% (n=328) of patients received platinum-based and 76.5% (n=1069) received non-platinum-based 2L mUC therapy. While, as expected, physician-reported patient platinum eligibility was correlated with treatment course (p<0.001), predictors of therapy type correlated differently within each eligibility groups (p≤0.02). For example, patients who were treated by younger physicians were less likely to receive platinum-based therapy (31.7% vs 45.9%, respectively, p≤0.001) indicating that physician characteristics are a potential source of variance associated with treatment choice. A small percentage of non-platinum eligible patients receive platinum-based 2L therapy (1.4%) which increases to 6.6% when the site of metastases was not in the lungs when first diagnosed with mUC (p=0.001). Multiple factors are involved in treatment choice for mUC related both to the treating physician and the patient. CHAID analysis is a useful tool for identifying associations between subgroups.