Cancer-specific utility: clinical validation of the EORTC QLU-C10D in patients with glioblastoma

Abstract

Introduction: Many health economic evaluations rely on the validity of the utility measurement for health-related quality of life (HRQoL). While generic utility measures perform well in HRQoL assessments of many diseases and patient populations, appropriateness for cancer-specific disease burdens needs attention and condition-specific measures could be a viable option. This study assessed the clinical validity of the cancer-specific EORTC QLU-C10D, a utility scoring algorithm for the EORTC QLQ-C30, in patients with glioblastoma. We expect the EORTC QLU-C10D to be sensitive and responsive in glioblastoma patients. Furthermore, we compared its statistical efficiency with the generic utility measure EQ-5D-3L.

Methods: We used data from a multi-center randomized controlled trial (NCT00689221) with patients from 146 study sites in 25 countries. Both, the QLQ-C30 and the EQ-5D-3L, had been administered at seven assessment points together. Utilities of both measures were calculated for four country value set (Australia, Canada, UK, USA). Ceiling effects, agreement (Bland-Altman plots (BA), intra-class correlation (ICC)), were calculated to analyze construct validity. Sensitivity to known-groups (performance status; global health) and responsiveness to changes (progressive vs. non-progressive; stable vs. improved or deteriorated HRQoL) were investigated for clinical validity. Relative Efficiency (RE) was calculated to compare statistical efficiency of both utility measures.

Results: 435 patients were included at baseline and six subsequent time points (median timeframe 497 days). QLU-C10D country value set showed negligible ceiling effects (< 6.7%) and high agreement with EQ-5D-3L (ICC > 0.750). BA indicated that differences between both utility measures increased with deteriorating health states. While the QLU-C10D was more sensitive to global health groups (RE > 1.2), the EQ-5D-3L was more sensitive to performance status groups (RE < 0.7) than the other utility measure. Statistical efficiency to detect differences between change groups and within HRQoL deterioration group (RE > 1.4) favored QLU-C10D in 18 of 24 (75%) and 20 of 24 (83%) comparisons with the EQ-5D-3L respectively. Responsiveness to overall HRQoL change (RE > 3.4) also favored the QLU-C10D.

Conclusion: Our results indicate that the QLU-C10D is a valid utility measure to assess HRQoL in patients with glioblastoma. This facilitates the investigation of HRQoL profiles and utilities in this patient population by administering a single questionnaire, the EORTC QLQ-C30. Efficiency analyses point to higher statistical power of the QLU-C10D compared to the EQ-5D-3L.

Keywords: EQ-5D-3L; Glioblastoma; Health-related quality of life; QLU-C10D; Utility; Validity.

Fig. 1

Original Study Procedures with assessment timepoints for HRQoL. Abbreviations: SD = Standard Deviation; HRQoL = Health-related Quality of Life; retr. = retrospective; Q = QLU-C10D; E = EQ-5D-3L; Country value set: AUS = Australia, CAN = Canada, UK = United Kingdom

Fig. 2

Bland–Altman Plots comparing QLU-C10D and EQ-5D-3L Country value sets at baseline. Dotted horizontal line = 0 at y-axis; LOA = Level of Agreement (= mean ± 1.96 * SD); Proportional bias: Results of Linear regressions, predicting differences with means (df = 1, 433): AUSTRALIAN Utilities, r = .09, R² = .008, F = 3.54, p = .061; CANADIAN Utilities, r = .229, R² = .052, F = 23.96, p < .001; UK Utilities, r = .444, R² = .197, F = 106.10, p < .001; USA Utilities, r = .001, R² = .000, F = 0.0, p = .992; Regression lines are not displayed

Fig. 3

Relative Efficiency (RE) of Country value sets for detecting known-groups at baseline. Abbreviations Country value sets: AUS = Australia, CAN = Canada, UK =United Kingdom, US =United States of America; Known Groups: ECOG Performance status (PS) 0 vs. > 0; QLQ-C30 Global Health Status Score (GHS)  ≤ 50 vs. > 50; RE results > 1 favor QLU-C10D, < 1 favor EQ-5D-3L. Descriptive & detailed results see Appendix A

Fig. 4

Responsiveness Analysis QLU-C10D vs. EQ-5D-3L. Grey arrows signify values outside the scope of x-axis; Abbreviations: Country value sets: AUS = Australia, CAN = Canada, UK = United Kingdom, US = United States of America; RE = Relative Efficiency; GHS = Global Health Scale (QLQ-C30); A RE = quotient of t-values of paired t-tests (Tx—Baseline); B RE = quotient of f-values of one-way ANOVA, comparing QLQ-C30 GHS groups (> 6 points decrease vs. ≤ 6 points change vs. > 6 points increase); C & D RE = quotient of t-values calculated with paired t-tests (T_x—Baseline); E & F Responsiveness Index (RI) standardizes mean Decrease or Increase of GHS groups with the standard deviation of the stable group; Difference RI = RI QLU-C10D minus RI EQ-5D-3L; Details of calculations and results of A are presented in Appendix B, B-F in Appendix C