Relationship between commonly used HRQOL measures: Correlations and mapping between the UCLA-PCI and EPIC scales

Background: Treatments available for prostate cancer (PCa) provide similar survival outcomes, so that quality of life becomes more important in treatment decisions. Different general and disease specific questionnaires measure health related quality of life (HRQoL) and it is unknown how they relate. We investigated correlations among the Expanded Prostate Cancer Index 26 (EPIC), UCLA Prostate Cancer Index (PCI), Sexual Health Inventory for Men (SHIM) and International Prostate Symptom Score (IPSS) questionnaires in PCa patients and generated functions to map PCI to the newer EPIC. Methods: 1,853 CaPSURE patients completed HR QoL questionnaires (EPIC, PCI, SHIM, IPSS). Pearson's correlations compared the relationship between domains. To map PCI items to EPIC scores, the full cohort was randomly divided into a model and a validation cohort. We considered identity, log, square and logit-transformations of the dependent variable. We performed linear and Lowess regressions for each dependent variable transformation. Predicted and observed scores were compared, and R2 and root mean square error (RMSE) calculated. We examined models within disease severity subsets. Results: Correlations between EPIC and PCI were highest for urinary function (0.92) and sexual bother (0.84). Correlation between EPIC bowel and PCI bowel function was 0.64 and with bother 0.68. Correlation between bowel bother scores of EPIC and PCI was 0.69. SHIM's correlations with PCI sexual function (SF) were higher than EPIC SF, (0.92 and 0.76). Correlation between IPSS (lower is better) and EPIC urinary irritative score was -0.71. The linear model predicted EPIC scores best with EPIC(UB) = 51 + (0.44 ∗ PCI question34). The difference between observed and predicted scores was -0.56 to 1.68, RMSE, 8 to 26% and R2,0.46 to 0.93. Predicted values for severity subsets were overestimated in some ranges with small sample sizes (low scores). Conclusions: High correlations between function and bother scores suggest good agreement between PCI and EPIC scores. Predicted EPIC scores give good predictions across the population, but are less reliable for the few low score patients. The mapping equations allow scores to be updated from PCI to EPIC.