OBJECTIVES: The limitations of generic preference-based measures in disease areas such as oncology are widely recognised. Condition-specific measures offer more relevant assessments of health and can be used to derive utilities. The aim of this study was to use data collected with the European Organization for Research and Treatment of Cancer (EORTC) Quality of Life Questionnaire for Cancer (QLQC30) in a myelofibrosis clinical trial to derive utilities.
METHODS: QLQ-C30 data were collected over 48 weeks in an open-label trial of ruxolitinib (n_146) versus best-available therapy (BAT) (n_73). Two algorithms were used to map QLQ-C30 scores to utilities: the first mapped to EQ-5D utilities, the second to conditionspecific preference weights using a QLQ-C30 item subset (EORTC-8D). Changes from baseline (CFB) in utilities were calculated by treatment at week 48. Mean utilities by presence of constitutional symptoms (CS) (weight loss, fever or night sweats) and response (_35% reduction in spleen volume from baseline) were also derived.
RESULTS: Mean (SE) utility CFB from the EQ-5D algorithm was 0.082 (0.025) for ruxolitinib and 0.012 (0.040) for BAT. From the EORTC-8D algorithm, mean (SE) CFB was 0.038 (0.013) for ruxolitinib and 0.013 (0.021) for BAT. Patients without CS had higher mean (SE) utilities than patients with CS using both algorithms—EQ-5D, 0.730 (0.017) without and 0.539 (0.031) with CS; EORTC-8D, 0.818 (0.009) without and 0.719 (0.016) with CS. Similarly, patients defined as responders had higher mean (SE) utilities than nonresponders using both algorithms—EQ-5D, 0.754 (0.029) for responders and 0.670 (0.024) for nonresponders; EORTC-8D, 0.843 (0.015) for re- sponders and 0.785 (0.012) for nonresponders.
CONCLUSIONS: Utility values can be derived from condition-specific measures such as the QLQ-C30. Our analyses demonstrate that the presence of CS and splenomegaly in patients with myelofibrosis results in lower utility values.
