OBJECTIVES: To predict health state utilities in patients with advanced CTCL, adjusting for disease progression and skin-specific Skindex-29 scores. To consider the correlation between EQ-5D and Skindex-29.

METHODS: The phase III ALCANZA trial collected health-related quality of life (HRQL) data using two separate measures, EQ-5D-3L and Skindex-29, for patients treated with either brentuximab vedotin or physician’s choice of bexarotene or methotrexate. Skindex-29 is a dermatological measure with three domains (symptoms, emotions, and functioning) that results in scores from 0-100, with higher scores indicate a more severe impact on HRQL. Linear mixed effects regression models were fitted to model the relationship between disease progression, Skindex-29 total score and EQ-5D utilities whilst accounting for repeated measures.

RESULTS: A total of 713 EQ-5D observations collected from 91 patients with advanced CTCL were included in the analysis, of which 173 were from patients in the progressed disease state. The regression model included covariates for progression status and Skindex-29 score. Statistically significant coefficients indicated a utility decrement of 0.05 for a 10-point increase in Skindex-29 score and a utility decrement of 0.03 for disease progression. Analysis showed poor correlation between the instruments; patients with Skindex-29 scores close to 100 scored close to 1.0 on EQ-5D.

CONCLUSIONS: Progressed disease and increased Skindex-29 score were independent predictors of HRQL in patients with CTCL (p<0.05). These effects may be underestimated in this study due to limited trial follow-up (34-months). Correlation analysis showed that the EQ-5D measure may not be responsive to changes in skin-related symptoms. Clinical advice suggested that not all skin-related and physiological symptoms are captured by EQ-5D and Skindex-29. This was acknowledged by a NICE appraisal committee. Further work to develop a CTCL-specific measure or validated mapping algorithm may allow the impact of CTCL on patients’ HRQL to be more accurately measured.