Journal of Clinical Oncology, 2008 ASCO Annual Meeting Proceedings (Post-Meeting Edition).
Vol 26, No 15S (May 20 Supplement), 2008: 6529
© 2008 American Society of Clinical Oncology

This Article Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Download to citation manager Rights & Permissions Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Pickard, A. S. Articles by Cella, D. Search for Related Content PubMed Articles by Pickard, A. S. Articles by Cella, D.

Abstract

A patient-based utility measure of health for clinical trials of cancer therapy

University of Illinois at Chicago, Chicago, IL; Pfizer, Inc, New London, CT; The Netherlands Cancer Institute, Amsterdam, The Netherlands; Hines Veterans' Affairs Hospital, Hines, IL; Northwestern University, Chicago, IL

6529

Background: The EORTC QLQ-C30 is a widely used measure of quality of life in oncology clinical trials. The ability to generate utility scores and calculate quality-adjusted life years would further expand its use. The aim of this study was to develop utility mapping algorithms for the QLQ-C30 using utilities for own health elicited from cancer patients. Methods: Using data from 1,432 patients, two distinct approaches were applied. A conventional multi-attribute health state (MAHS) approach employed psychometric analysis and expert input to identify items from the functioning and symptom scales for model development. Second, using a conceptual model of health preference formation, models were developed using global health and quality-of-life (QoL) items. All models were fit using quantile regression and robust bootstrap variance estimators. Criteria for model selection included parsimony, statistical significance and logical consistency of parameter estimates, number of unique states, and range of scale. Predictive accuracy was evaluated using a 10-fold cross- validation approach. Results: The best-fitting MAHS model included one or more items from each scale except Emotional Functioning. The model included 14 parameters, described 2,304 unique states, and predicted values ranging from 0.61 to 1.00. However, it had poor predictive accuracy (cross-validation pseudo-R²=0.073), and most parameter estimates were insignificant. The best-fitting QoL approach-based model described 24 unique states using 9 parameters (all of which were significantly significant), predicted values ranging from 0.17 to 1.00, and exhibited better fit statistics than the MAHS model (e.g., cross-validation pseudo-R²=0.127). Conclusions: Using conventional and novel methodologies, utility-mapping algorithms were developed for the QLQ-C30. Each approach yielded algorithms with different strengths and weaknesses. Notably, the QoL-based model has more desirable statistical properties and a wider range of scale, while the MAHS-based model produces a far greater number of unique predicted values. Both algorithms may be used to predict patient preferences for own health to inform economic evaluations of cancer therapeutics.

Author Disclosure

Employment or Leadership Consultant or Advisory Role Stock Ownership Honoraria Research Expert Testimony Other Remuneration Pfizer Oncology

Abstract presentation from the 2008 ASCO Annual Meeting