OBJECTIVES:   In 2009, the cost-effectiveness of drugs for chronic myeloid leukaemia for use in the UK NHS was evaluated by the National Institute of Health and Clinical Excellence (NICE).  Two questions were considered: a) dasatinib vs. nilotinib vs. high-dose imatinib vs. interferon-alpha for imatinib-resistant patients, and b) dasatinib vs. nilotinib vs. interferon-alpha for imatinib-intolerant patients.  Here, three methodological issues are discussed which strongly influenced the cost-effectiveness of these drugs.  These issues are also important in estimation of the cost-effectiveness of many other drugs and other health technologies. METHODS:   1)   Overall survival:  Several methods were considered for estimating overall survival, including those used by the drug sponsors, Novartis and Bristol-Myers Squibb; 2)   Sources of mortality:  Two approaches were considered: a) split by mortality due to chronic myeloid leukaemia and general mortality or b) both combined; 3)   Treatment duration:  This was reported in none of the trials.  Several methods of estimating treatment duration were considered. RESULTS:      1)   Overall survival:  It was not possible to extrapolate overall survival because it was very immature in the trials.  Instead, the preferred method was estimation via a surrogate relationship using major cytogenetic response; 2)   Sources of mortality:  Option (a) was preferred; and 3)   Treatment duration:  the preferred method was by reference to mean progression-free survival, adjusted for treatment cessation due to adverse events. CONCLUSIONS:   To estimate the cost-effectiveness of health technologies for a variety of conditions, it is recommended that 1) if overall survival from a trial is immature, it can be estimated by surrogate relationships; 2) for chronic conditions, the analyst should consider modelling separately disease-specific mortality and general mortality; and 3) for drugs, the mean number of doses in clinical trials should be reported so that it is not necessary to estimate this important information using indirect methods.