AN OVERVIEW OF DIFFERENT MAPPING TECHNIQUES TO DERIVE HEALTH STATE UTILITY VALUES IN MULTIPLE MYELOMA FOR DECISION-ANALYTIC MODELING

Purpose:

The National Institute for Clinical Excellence (NICE) recommends quality-adjusted life years (QALYs) as main measure for health outcomes in cost-utility analyses. In the absence of utilities, NICE recommends to use “mapping” functions to link utilities to other outcomes. The goal of our study is to give an overview on published utility data and mapping techniques that can be applied to derive utility values from non-preference based (e.g. disease-specific) patient reported outcome measures in multiple myeloma (MM).

Method(s):

We performed a systematic literature search in PubMed/Medline to identify studies reporting on health-related quality of life (HRQoL) in patients with MM derived from the EQ-5D and EORTC questionnaires (QLQ-C30 or QLQ-MY20). To meet our inclusion criteria, studies were required to evaluate treatment strategies for patients with MM, to be published in English as full text, and to report utilities or results from the EQ-5D and EORTC questionnaires. In the absence of sufficient data on utilities, we conducted a search (systematic and hand search) on studies describing algorithms for mapping health utilities from non-preference based questionnaires (EORTC) to a generic preference-based questionnaire EQ-5D. We extracted and summarized the published algorithms, their predictive performance, statistical models, and validation techniques based on the ISPOR Good Research Practice Task Force for mapping.

Result(s):

We identified 39 studies reporting on HRQoL from EQ-5D and EORTC. Only seven studies reported utilities. However, these utilities did not sufficiently cover current MM treatment options and disease stages. In our search for mapping algorithms, we identified eight studies applying mapping approaches to derive utilities. Only four studies reported mapping algorithms for MM; the remaining studies included different cancer diseases. The most frequently applied statistical model for mapping was the ordinary least square model. Furthermore, ordered probit regression, Tobit models, two-part models, splining models, response mapping models, limited depended variable mixture models, and multiple linear regression analysis were applied. Only two studies reported on validation of the mapping algorithms.

Conclusion(s):

We identified several studies reporting on HRQoL data from EQ-5D and EORTC. Based on the extracted mapping algorithms, we identified several approaches on how to translate the extracted data to utility values. The quality of reporting the mapping techniques varies considerably and in some cases the applied algorithms were not sufficiently described to be replicated.