OBJECTIVES: Reporting of survival outcomes from clinical trials is often limited to median survival times, hazard ratios, Kaplan-Meier curves and numbers at risk. The numerical results are not always sufficient for meta-analysis and cost-effectiveness analysis. Further information can be obtained by digitizing and analysing the Kaplan-Meier curves. The most basic analysis approach is to fit a non-linear model to the Kaplan-Meier curve and use this to estimate parameters such as the mean survival time. Methods have recently been developed for estimating individual patient data (IPD) from Kaplan-Meier curves. Once individual patient data is estimated, standard survival analysis approaches can be used to estimate parameters and also provide estimates of uncertainty in the curve fits.  The objective of this study was to review the methods commonly used and assess the impact of the improved methods, where IPD is estimated, on the inferences drawn. METHODS: We conducted a systematic review of the methods that have been used in NICE HTAs to obtain data from published Kaplan-Meier curves. We examined the frequency of each method, how results were used and any feedback from Evidence Review Groups. Improved methods, estimating IPD, were applied to a selection of studies where this was not conducted in the original analysis.  The impact of the improved methods on the conclusions of the studies was assessed. RESULTS: The review showed that most HTAs used non-linear models to approximate the Kaplan-Meier curves.   It also showed that the improved methods, estimating IPD, can have a significant impact on conclusions drawn from survival results.    CONCLUSIONS: The estimation of IPD from Kaplan-Meier curves is a valuable method that is currently underutilised. It has the potential to provide better estimates of survival parameters and to improve the characterisation of uncertainty in such estimates. This is especially important when survival curves are extrapolated.