PRM196 - An Evaluation and Comparison of Methods Used in Survival Analysis To Fit Distributional Curves To Kaplan-Meier Data

Abstract

HTA bodies increasingly require accurate survival estimates in order to provide reliable recommendations. It is argued that access to individual patient data (IPD) can improve their accuracy. This paper aims to assess to what degree extracting IPD from published Kaplan-Meier curves helps improve extrapolated survival estimates. Some methods currently used for HTA submissions fit a survival curve directly to a published Kaplan-Meier curve, but does this lack accuracy? Two methods used to extract the IPD from Kaplan-Meier curves reviewed in this paper are by Guyot et al (Guyot, Ades, Ouwens, & Welton, 2012) and Hoyle and Henley (Hoyle & Henley, 2011) which were compared against the outcomes of the standard ‘least squares’ method. Comparisons were made for two situations: 1) when numbers at risk are available at different time points throughout the Kaplan-Meier curve and 2) when numbers at risk are only available at the start. The three methods resulted in the long-normal distribution showing the best fit, with all containing the true mean and median within their confidence intervals. However, the Hoyle and Henley method estimates a mean marginally closer to the true mean than the other methods in both situations. When many numbers at risk are provided, the Hoyle and Henley method gives narrower confidence intervals. Both extraction methods slightly outperformed the least squares method. The three methods give median estimates and resulting confidence intervals which are statistically equivalent to that of the IPD, except for the Guyot method when numbers at risk are not available. In conclusion, extraction methods can give marginally better results than the Least Squares method. However, these results may not be applicable to other examples. In addition, the extra time taken to run extraction methods could be too large to account for the small improvement in accuracy of results.