Mapping HRQoL questionnaires into utilities for prostate cancer patients: Epic, UCLA-PCI, SHIM, IPSS, SF36 to PORPUS.

Background: Patients diagnosed with prostate cancer (PCa) have similar survival rates across treatments, making treatments choices based on health related quality of life (HRQoL) more important. It is unknown how the general and disease specific HRQoL measures relate to one other. We developed functions (mapped) to predict Patient Oriented Prostate Utility Scale (Porpus) utilities from 5 different HRQoL questionnaires: Expanded Prostate Cancer Index 26 (EPIC), UCLA Prostate Cancer Index (PCI), Sexual Health Inventory for Men (SHIM), International Prostate Symptom Score (IPSS) and Short Form 36v2 (SF36). Methods: 1,720 CaPSURE patients completed the six questionnaires, at the same time. The full cohort was randomly split into a model and validation cohort. Each utility was mapped using individual questions (IPSS, SHIM) or total scores (EPIC, PCI, SF36). We performed identity, log, square and logit-transformations of the dependent variable (Porpus utility). Linear and Lowess regression models were fitted to the model dataset for each dependent variable and compared. The predicted and actual scores were compared with t-tests. R2and root mean square errors (RMSE) were computed and models further examined within specific disease severity level subsets. Results: For EPIC and SHIM, a logit transformed model fit best as: log(PorpusU/(1-PorpusU))= 0.53 + 0.00080 * EPIC(Urinary Incontinence)score + 0.00045* EPIC(Urinary Irritation)score + 0.0011 * EPIC(Bowel)score + 0.0015 * EPIC(Sexual)score + 0.0014 * EPIC(Hormonal)score). A linear model was preferred for IPSS and a log transformed linear model for PCI and SF36. The difference between Porpus predicted and observed utilities mapped from EPIC, PCI, SHIM, IPSS and SF36 were respectively 0.007, -0.004, 0.012, 0.001 and -0.004 (all p<0.01). IPSS and SF36 had the lowest R2(0.22 and 0.25) and others ranged from 0.41-0.66. Conclusions: We were able to adequately predict utility scores using different general and disease specific questionnaires. These mapping equations are useful for determining the relationship of HRQoL measures commonly used in prostate cancer. Mapping to utility scores also allows use of HRQoL in cost-effectiveness analyses.