OBJECTIVES: Guidelines for economic evaluation from ISPOR and NICE recommend the use of mean values rather than medians, but do not offer guidance for situations when mean values are not available. This study evaluated the impact of different methods of estimating means from medians, within the context of estimating treatment duration for drugs used to treat mCRC. METHODS: Clinical trials and prescribing information for drugs used to treat mCRC were reviewed for information on the mean, median, and range of treatment cycles. Various approaches were used to estimate mean values, including direct use of the median, a published equation considering the median and range, confidence intervals and interquartile range, and the use of distributions commonly used in survival analysis. Where possible, the estimated means were compared with reported means from clinical trials. RESULTS: Very few studies reported both median and mean treatment duration; direct use of the median under-predicted the mean by 23-39% and the published equation over-predicted the mean by 19-28%. Simple assumptions about the distribution of treatment durations performed best, predicting the reported mean within ±12%. The use of progression-free survival as a proxy for treatment duration over-predicted treatment duration by 5-38%, although estimates were improved by accounting for early discontinuation. CONCLUSIONS: By only considering the 50th percentile, the median may not provide an accurate representation of the outcomes in a population. It is important that researchers and budget-holders are aware of the limitations in the use of medians, and that they consider multiple estimation methods to estimate mean values for economic analyses.