Estimating the Relationship between Preference-Based Generic Utility Instruments and Disease-Specific Quality-of-Life Measures in Severe Chronic Constipation

Abstract

Background: Many clinical trials that generate evidence on the quality-oflife (QOL) improvements provided by new health technologies do not incorporate a preference-based generic measure, but generate only diseasespecific data. However, in order to meet the information needs of regulators such as the UK National Institute for Health and Clinical Excellence (NICE), such disease-specific data need to be converted into a broader generic measure; for NICE, the preferred instrument is the EQ-5D. The process of converting QOL data from one instrument to another is known as ‘mapping’.

Objective: The objective of this study was to examine the extent to which disease-specific measures generated in the clinical trials for a new treatment for chronic constipation (prucalopride) can be ‘mapped’ onto a preferencebased generic measure (EQ-5D and SF-6D) to generate robust and reliable utility estimates.

Methods: Disease-specific QOL data generated in the clinical trials of prucalopride (PAC-QOL scores) were converted into utility values estimated using the preference-based generic measure EQ-5D. SF-36 data were also collected in the clinical trials and used to generate SF-6D estimates for comparative purposes. Regression analysis was used to derive a range of mapping functions to identify the extent to which increasing the complexity of the hypothesized underlying mapping function enhanced the robustness and reliability of the obtained mapping relationship.

Results: The mean utility observed at baseline for chronic constipation, based on SF-36 data, was 0.813 with the EQ-5D and 0.723 with the SF-6D. An examination of the differences between predicted and observed values generally found that the mapping functions generated were robust and reliable, with little evidence of bias across the range of the dependent variable. However, the nature of the symptoms explored in the PAC-QOL measure was, in general, less severe than those explored in the EQ-5D. For example, the condition-specific measures explored the degree to which patients experienced ‘discomfort’, rather than ‘pain’ as evaluated in the EQ-5D. Given this limitation in the severity range covered in the disease-specific measures, it is perhaps not surprising that a floor effect was identified, with certain health dimensions mapping only to the upper range of the EQ-5D measure.

Conclusions: In circumstances where direct utility measurement is not available, mapping provides a valuable method by which to estimate utility data for incorporation into cost-effectiveness analyses. Our findings emphasize the importance of the structure and nature of the mapping analysis undertaken as being a fundamental determinant of the utility estimates generated. Unfortunately, the theoretical guidance available to steer such analyses is still comparatively underdeveloped and this remains an area of health economic analysis in which empiricism largely rules. Ensuring that such mapping is undertaken and interpreted in as transparent and robust a manner as possible is therefore crucial in allowing regulators to accurately compare the clinical and cost effectiveness of new drugs across therapeutic areas.