OBJECTIVES: Reimbursement agencies require estimates of the long-term (i.e. lifetime) costs and benefits associated with each treatment option as part of the decision making process. As such, extrapolation of reported survival estimates is inevitable. Conventionally, information on all-cause mortality or disease progression is used. However, data for newly diagnosed CML may not be suited to such an extrapolation due to the paucity of observed events. One of solutions is to use a ‘surrogate approach’. METHODS: A 40 year Excel® based model was created to estimate overall (OS) and progression-free (PFS) survival in newly diagnosed CML patients receiving 1st or 2nd generation tyrosine kinease inhibitor (TKI) therapy through the use of a surrogate clinical endpoint (cytogenic response - CyR). Three response categories (complete, partial or no CyR at one year) were used. Long term response category specific OS and PFS data from IRIS clinical trial was used to inform the fitting of Weibull functions with goodness of fit assessed via the R2 statistic. CyR response rates were taken from a recently published network meta-analysis of first-line interventions. RESULTS: Using the conventional approach, CML patients were predicted to have an equivalent survival profile to the non-CML general population (31.6 versus 32.6 years), and the survival difference between 1st and 2nd generation drugs are 9 years (31.6 versus 22.8 years) (Botteman et al 2010). However, predicted OS estimates for 1st and 2nd generation TKI’s using the surrogate approach are 18.6 and 20.1 years respectively (R2 values 0.97, 0.94). Compared to 1st generation drugs the use of 2nd generation TKI’s results in approximately 1.5 additional years of survival. CONCLUSIONS: Extrapolating short term overall OS data in newly diagnosed CML results in inflated survival estimates. When a valid clinical surrogate is used there is a much smaller, and believable, difference in the predicted survival values.