Use of Bayesian Markov Chain Monte Carlo Methods to Estimate EQ-5D Utility Scores from EORTC QLQ Data in Myeloma for Use in Cost-Effectiveness Analysis
Restricted accessResearch articleFirst published online April, 2015
Use of Bayesian Markov Chain Monte Carlo Methods to Estimate EQ-5D Utility Scores from EORTC QLQ Data in Myeloma for Use in Cost-Effectiveness Analysis
Background. Patient-reported outcome measures are an important component of the evidence for health technology appraisal. Their incorporation into cost-effectiveness analyses (CEAs) requires conversion of descriptive information into utilities. This can be done by using bespoke utility algorithms. Otherwise, investigators will often estimate indirect utility models for the patient-reported outcome measures using off-the-shelf utility data such as the EQ-5D or SF-6D. Numerous modeling strategies are reported; however, to date, there has been limited utilization of Bayesian methods in this context. In this article, we examine the relative advantage of the Bayesian methods in relation to dealing with missing data, relaxing the assumption of equal variances and characterizing the uncertainty in the model predictions. Methods. Data from a large myeloma trial were used to examine the relationship between scores in each of the 19 domains of the European Organisation for Research and Treatment of Cancer (EORTC) QLQ-C30/QLQ-MY20 and the EQ-5D utility. Data from 1839 patients were divided 75%/25% between derivation and validation sets. A conventional ordinary least squares model assuming equal variance and a Bayesian model allowing unequal variance were estimated on complete cases. Two further models were estimated using conventional and Bayesian multiple imputation, respectively, using the full data set. Models were compared in terms of data fit, accuracy in model prediction, and characterization of uncertainty in model predictions. Conclusions. Mean EQ-5D utility weights can be estimated from the EORTC QLQ-C30/QLQ-MY20 for use in CEAs. Frequentist and Bayesian methods produced effectively identical models. However, the Bayesian models provide distributions describing the uncertainty surrounding the estimated utility values and are thus more suited informing analyses for probabilistic CEAs.
MeadsDBrowneCEdlinRMcCabeC. A systematic review of the consideration of uncertainty in indirect utility (mapping) models. Working Paper 12_05. Leeds, UK: Academic Unit of Health Economics, University of Leeds; 2012.
2.
BrazierJEKolotkinRLCrosbyRDWilliamsGR. Estimating a preference-based single index for the Impact of Weight on Quality of Life (IWQOL-Lite) Instrument from the SF-6D. Value Health. 2004;7:490–8.
3.
TsuchiyaABrazierJEMcCollEParkinD. Deriving preference based single indices from non-preference based condition-specific instruments: converting AQLQ into EQ5D indices. Sheffield Health Economics Group Discussion Paper Series 02/1. Sheffield, UK: University of Sheffield;2002.
4.
BrennanDSSpencerAJ. Mapping oral health related quality of life to generic health state values. BMC Health Serv Res. 2006;6:96–106.
5.
LeQADoctorJN. Probabilistic mapping of descriptive health status responses onto health state utilities using Bayesian networks: an empirical analysis converting SF-12 into EQ-5D utility index in a national US sample. Med Care. 2011;49(5):451–60.
6.
WijeysunderaHCTomlinsonGNorrisCMGhaliWAKoDTKrahnMD. Predicting EQ-5D utility scores from the Seattle Angina Questionnaire in coronary artery disease: a mapping algorithm using a Bayesian framework. Med Decis Making. 2011;31(3):481–93.
7.
McCabeCEdlinRMeadsDBrowneCKharroubiS. Constructing indirect utility models: some observations on the principles and practice of mapping to obtain health state utilities. Pharmacoeconomics. 2013;31(8):635–41.
8.
National Institute for Health and Clinical Excellence. Guide to the Methods of Technology Appraisal. London: NICE; 2008.
9.
DolanP. Modelling valuation for EuroQol health states. Med Care. 1997;35:351–63.
10.
CrottRBriggsA. Mapping the QLQ-C30 quality of life cancer questionnaire to EQ-5D patient preferences. Eur J Health Econ. 2010;11:427–34.
11.
KontodimopoulosNAletrasVHPaliourasDNiakasD. Mapping the cancer-specific EORTC QLQ-C30 to the preference-based EQ-5D, SF-6D, and 15D instruments. Value Health. 2009;12(8):1151–7.
12.
SteadMBrownJVelikovaG. Development of an EORTC questionnaire module to be used in health-related quality-of-life assessment for patients with multiple myeloma. Br J Haematol. 1999;104:605–11.
13.
BasuAMancaA. Regression estimators for generic health-related quality of life and quality-adjusted life years. Med Decis Making. 2012;32(1):56–69.
14.
Hernandez AlavaMWailooAWolfeFMichaudK. A comparison of direct and indirect methods for the estimation of health utilities from clinical outcomes. Med Decis Making. 2014;34(7):919–30.
15.
LittleRJRubinDB. Statistical Analysis with Missing Data. New York: John Wiley; 2002.
16.
GomesMDiaz-OrdazKGrieveRKenwardMG. Multiple imputation methods for handling missing data in cost-effectiveness analysis that use data from hierarchical studies: an application to cluster randomized trials. Med Decis Making. 2013;33(8):1051–63.
17.
RubinDB. Multiple Imputation for Nonresponse in Surveys. New York: John Wiley; 1987.
RobinsJMWangN. Inference for imputation estimators. Biometrika. 2000;87:113–24.
20.
WangNRobinsJM. Large sample theory for parametric multiple imputation procedures. Biometrika. 1998;85:935–48.
21.
SpiegelhalterDJThomasABestNGLunnD. WinBUGS User Manual: Version 1.4. Cambridge, UK: MRC Biostatistics Unit;2003.
22.
GelmanARubinDB. Inference from iterative simulation using multiple sequences. Stat Sci. 1992;7:457–72.
23.
BriggsAClarkTWolstenholmeJClarkeP. Missing . . . presumed at random: cost-analysis of incomplete data. Health Econ. 2003;12(5):377–92.
24.
ClaxtonKPalmerSLongworthL. Informing a decision framework for when NICE should recommend the use of health technologies only in the context of an appropriately designed programme of evidence development. Health Technol Assess. 2012;16(46):1–323.
25.
DoubiletPBeggCBWeinsteinMCBraunPMcNeilBJ. Probabilistic sensitivity analysis using Monte Carlo simulation: a practical approach. Med Decis Making. 1985;5(2):157–77.
26.
ClaxtonKSculpherMMcCabeC. Probabilistic sensitivity analysis for NICE technology assessment: not an optional extra. Health Econ. 2005;14(4):339–47.
27.
BriggsAHGoereeRBlackhouseGO’BrienBJ. Probabilistic analysis of cost-effectiveness models: choosing between treatment strategies for gastroesophageal reflux disease. Med Decis Making. 2002;22(4):290–308.
28.
ParmigianiG. Measuring uncertainty in complex decision analysis models. Stat Methods Med Res. 2002;11(6):513–37.
29.
GriffinSClaxtonKHawkinsNSculpherM. Probabilistic analysis and computationally expensive models: necessary and required?Value Health. 2006;9(4):244–52.
30.
O’HaganAStevensonMMadanJ. Monte Carlo probabilistic sensitivity analysis for patient level simulation models: efficient estimation of mean and variance using ANOVA. Health Econ. 2006;16:1009–23.
31.
AustinPC. Bayesian extensions of the Tobit model for analyzing measures of health status. Med Decis Making. 2002;22(2):152–62.
32.
HuangICFrangakisCAtkinsonMJ. Addressing ceiling effects in health status measures: a comparison of techniques applied to measures for people with HIV disease. Health Serv Res. 2008;43(1, Pt 1):327–39.
33.
ChuangLHKindP. Converting the SF-12 into the EQ-5D: an empirical comparison of methodologies. Pharmacoeconomics. 2009;27(6):491–505.
34.
Rivero-AriasOOuelletMGrayAWolstenholmeJRothwellPMLuengo-FernandezR. Mapping the Modified Rankin Scale (mRS) measurement into the generic EuroQol (EQ-5D) health outcome. Med Decis Making. 2010;30(3):341–54.
35.
JiaHZackMMMoriartyDGFrybackDG. Predicting the EuroQol Group’s EQ-5D index from CDC’s “Healthy Days” in a US sample. Med Decis Making. 2011;31(1):174–85.
36.
MortimerDSegalLSturmJ. Can we derive an “exchangerate” between descriptive and preference-based outcome measures for stroke? Results from the transfer to utility (TTU) technique. Health Qual Life Outcomes. 2009;7:33.
