OBJECTIVES: This study examined the responsiveness of preference-based measures based on the anchor of self-reported change in general health condition of patients with colorectal cancer (CRC). METHODS: A baseline sample of 333 patients was recruited at the specialist outpatient clinic of academic teaching hospital in Hong Kong between September 2009 and July 2010, and was surveyed prospectively at 6-month follow-up. SF-6D preference-based indices were derived from the generic SF-6D measure (SF-6DDirect), from the Short Form-12 Health Survey (SF-6DSF-12) and also mapped from the condition-specific Functional Assessment of Cancer Therapy-Colorectal (SF-6DFACT-C). The responsiveness of three measures was assessed using the internal responsiveness and external responsiveness. The 95% bias-corrected and accelerated bootstrapping confidence intervals were performed to compare the internal responsiveness statistics measured by standardized effect size, standardized response mean, and responsiveness statistic. External responsiveness was evaluated by receiver operating characteristic (ROC) curve analysis that examined the ability to detect score changes with global health condition changes or discriminate between the worsened and unchanged/improved groups. RESULTS: Over half of patients reported no change in global health condition based on the self-reported anchor, whilst 15.1% and 32.9% of patients rated better and worse in current health condition compared to baseline respectively. In worsened group, internal responsiveness was satisfactory for the SF-6DDirect and SF-6DFACT-C preference-based indices. The SF-6DSF-12 and SF-6DFACT-C indices were significantly more responsive to detect positive changes than the SF-6DDirect index in improved group. The SF-6DDirect and SF-6DFACT-C indices were more externally responsive based on ROC curve. The SF-6DFACT-C index was generally more responsive to changes in health status compared with other indices. CONCLUSIONS: Direct SF-6D measure was more responsive than mapped preference-based measures in improved group but the direction was reversed in worsened group. Use of a preference-based index mapped from a condition-specific measure captures both negative and positive important changes in HRQOL score among CRC.