OBJECTIVES: Many company submissions received by Health Technology Assessment (HTA) Agencies evaluate the cost effectiveness of interventions which impact on survival. An accurate estimate of the survival benefit is required to calculate a reliable estimate of cost effectiveness.  Generally the relevant trial data is immature and must be extrapolated.  Many extrapolation models are available. Model choice is critical; different models can lead to different cost effectiveness results. The objectives were to review the methods/justification of the survival analysis used in company submissions to the National Centre for Pharmacoeconomics (NCPE). A further aim was to develop NCPE Guidance for the future handling of survival analysis. METHODS: Relevant submissions to the NCPE (economic evaluations which had dealt with advanced and/or metastatic cancer) were reviewed to determine the methods/justification of the survival analysis used. RESULTS: Twelve submissions were evaluated.  Appropriately, the mean overall survival (OS) had been estimated and used in eight cases (67%). Median OS estimates had been estimated/used in three (25%).  It was unclear which measure had been used in the remaining submission. The submissions which had used mean OS estimates were further investigated. Parametric model-based extrapolation techniques had been used to calculate the mean estimates in all eight. The most popular parametric models were the Weibull (n=3) and the loglogistic (n=3).  The methods used to fit the parametric models varied.  Most commonly the model was fitted using individual patient-level data. Some justification for the choice of extrapolation technique was offered in five submissions; AIC +/or BIC were estimated in three and visual inspection was reported in two. CONCLUSIONS: Survival analysis has not been conducted appropriately in all HTAs. Justification of the choice of model is not always offered.  Moving forward, NCPE Guidance is required to ensure that survival analysis using patient-level data is conducted appropriately. These will be presented.