Background: To improve the precision of health economics analyses in oncology, reference datasets of health utility (HU) scores are needed from cancer survivors across different disease sites. These data are particularly sparse amongst Canadian survivors.
Methods: A survey was completed by 1759 ambulatory cancer survivors at the Princess Margaret Cancer Centre which contained demographic questions and the EuroQol-5D (EQ-5D) instrument. Clinical information was abstracted from electronic records and HU scores were calculated using Canadian health state valuations. Construct validity was assessed through correlation of HU and visual analog scale (VAS) scores (Spearman) and by comparing HU scores between performance status groups (effect size). The influence of socio-demographic clinical variables on HU was analyzed by non-parametric between-group comparisons and multiple linear regression.
Results: Mean EQ-5D HU scores were derived for 26 cancers. Among all survivors, the mean ± standard error of the mean EQ-5D utility score was 0.81 ± 0.004. Scores varied significantly by performance status (p < 0.0001) and correlated with VAS (Spearman r = 0.61). The cancer sites with the lowest mean HU scores were acute lymphoblastic leukemia (0.70 ± 0.03) and pancreatic cancer (0.76 ± 0.03); testicular cancer (0.89 ± 0.02) and chronic lymphocytic leukemia (0.90 ± 0.05) had the highest mean scores. A multiple regression model showed that scores were influenced by disease site (p < 0.001), education level (p < 0.001), partner status (p < 0.001), disease extent (p = 0.0029), and type of most recent treatment (p = 0.0061).
Conclusions: This work represents the first set of HU scores for numerous cancer sites derived using Canadian preference weights. The dataset demonstrated construct validity and HU scores varied by general socio-demographic and clinical parameters.