Evaluation of the performance of algorithms mapping EORTC QLQ-C30 onto the EQ-5D index in a metastatic colorectal cancer cost-effectiveness model
BACKGROUND: Cost-effectiveness models require quality of life utilities calculated from generic preference-based questionnaires, such as EQ-5D. We evaluated the performance of available algorithms for QLQ-C30 conversion into EQ-5D-3L based utilities in a metastatic colorectal cancer (mCRC) patient population and subsequently developed a mCRC specific algorithm. Influence of mapping on cost-effectiveness was evaluated. METHODS: Three available algorithms were compared with observed utilities from the CAIRO3 study. Six models were developed using 5-fold cross-validation: predicting EQ-5D-3L tariffs from QLQ-C30 functional scale scores, continuous QLQ-C30 scores or dummy levels with a random effects model (RE), a most likely probability method on EQ-5D-3L functional scale scores, a beta regression model on QLQ-C30 functional scale scores and a separate equations subgroup approach on QLQ-C30 functional scale scores. Performance was assessed, and algorithms were tested on incomplete QLQ-C30 questionnaires. Influence of utility mapping on incremental cost/QALY gained (ICER) was evaluated in an existing Dutch mCRC cost-effectiveness model. RESULTS: The available algorithms yielded mean utilities of 1: 0.87 ± sd:0.14,2: 0.81 ± 0.15 (both Dutch tariff) and 3: 0.81 ± sd:0.19. Algorithm 1 and 3 were significantly different from the mean observed utility (0.83 ± 0.17 with Dutch tariff, 0.80 ± 0.20 with U.K. tariff). All new models yielded predicted utilities drawing close to observed utilities; differences were not statistically significant. The existing algorithms resulted in an ICER difference of €10,140 less and €1765 more compared to the observed EQ-5D-3L based ICER (€168,048). The preferred newly developed algorithm was €5094 higher than the observed EQ-5D-3L based ICER. Disparity was explained by minimal diffences in incremental QALYs between models. CONCLUSION: Available mapping algorithms sufficiently accurately predict utilities. With the commonly used statistical methods, we did not succeed in developping an improved mapping algorithm. Importantly, cost-effectiveness outcomes in this study were comparable to the original model outcomes between different mapping algorithms. Therefore, mapping can be an adequate solution for cost-effectiveness studies using either a previously designed and validated algorithm or an algorithm developed in this study.
|Keywords||Colorectal cancer, EQ-5D-3L, Mapping algorithm, QLQ-C30, Quality of life, Utility|
|Persistent URL||dx.doi.org/10.1186/s12955-020-01481-2, hdl.handle.net/1765/129136|
|Journal||Health and Quality of Life Outcomes|
Franken, M.D. (Mira D.), de Hond, A. (Anne), Degeling, K. (Koen), Punt, C.J.A, Koopman, M, Uyl-de Groot, C.A, … van Oijen, M.G.H. (2020). Evaluation of the performance of algorithms mapping EORTC QLQ-C30 onto the EQ-5D index in a metastatic colorectal cancer cost-effectiveness model. Health and Quality of Life Outcomes, 18(1). doi:10.1186/s12955-020-01481-2