Evaluating health-related quality of life:: A case-study of differential item functioning analysis in small trials

Clinical trials Usually focus on endpoints that traditionally are biomedical measures. However, such endpoints do not reflect patient's perception of his or her well-being and satisfaction with therapy. Health-Related Quality of Life (HRQoL) may be an important health outcome. The use of a questionnaire with documented validity and reliability is recommended. Psychometric properties of HRQoL questionnaires are often investigated using Cronbach's alpha, inter-item correlations. However, these psychometric tests do not take into account the intentional use of an instrument (e.g., specific clinical context). Item bias also called differential item functioning (DIF) concentrates on potential bias resulting from the construction of items of a questionnaire. DIF investigates whether items behave in the same way whatever the subgroup studied, and whether the global score can be considered as a good representation of the items. We investigated DIF in a disease-specific instrument, the Minnesota Living With Heart Failure LiHFE questionnaire, in a small trial involving congestive heart failure patients. We evaluated the extent of the problems encountered during analysis of DIF in small studies and with the choice of score categorization when using the contingency table approach. Potential DIF problems were found in three items and in one item of the physical and emotional dimension., respectively, with the NYHA classification (physician's assessment of functional capacity) and with age. However, owing to our small sample size, results should be taken cautiously. The use of DIF analysis should be extended to clinical trials because if treatment DIF is present in some HRQoL questionnaires, direct comparisons of such questionnaires might lead to badly misleading results.