OBJECTIVES: To estimate the effect of Bortezomib on utility values in multiple myeloma patients. METHODS: A Phase III trial (APEX) compared Bortezomib with High Dose Dexamethasone. Utility scores (EuroQol-5D) were collected together with a comprehensive range of clinical and serological outcomes. Changes in utility scores were analysed separately for survivors and non-survivors using non-parametric regression methods. A maximum likelihood method that adjusts for the latent, underlying risk of death as a utility-driver was employed to take account of unobservable effects of declining health on utility scores. RESULTS: Data for 655 out of a total of 669 patients in APEX were included in the analysis, including 129 patients for whom information was available on time to death. Mean utility scores before progression were similar between Responders (0.65) and Non-responders (0.61; mean diff. 0.045, 95% CI –0.01, 0.10); the respective estimates for the after-progression phase (0.67 vs. 0.58) had a mean difference of 0.10 (95% CI –0.00, 0.19). No differences were found between treatment groups. A steep decline in utility over time, following a stable period, began approximately 100 days before death. Estimates from a two-step regression model of utility scores are consistent with informative censoring or sample selection bias in the estimating sample; adjusting for such bias reduced the estimated post-progression difference between responders and non-responders. CONCLUSIONS: The last days of life in patients with multiple myeloma are associated with a steep decline in health related quality of life. Evaluations of utility outcomes conducted alongside clinical trials of salvage therapies in haematology and cancer are likely to require adjustment for the unavailability of representative samples for assessment as a result of early trial termination. We propose an approach commonly used in other fields to deal with this issue and discuss its advantages and limitations.