PPubMed24 Mar 2015Cited 51×

A Review and Comparison of Methods for Recreating Individual Patient Data from Published Kaplan-Meier Survival Curves for Economic Evaluations: A Simulation Study Comparison of Methods for Recreating Individual Patient Data

Xiaomin Wan, Liubao Peng, Yuanjian Li, Robert K Hills

Abstract

Decision analytic modeling is often applied as a vehicle for economic evaluations. A commonly used approach in decision analytic modeling is the Markov model. Transition probabilities between Markov states are one of the most important features of the stochastic model. There are several available methods for the estimation of transition probabilities. Currently, survival analysis has been frequently used to estimate transition probabilities. Recently, an algorithm was developed by Latimer that aimed to improve the quality of survival analyses included within economic evaluations. The framework

A figure from A Review and Comparison of Methods for Recreating Individual Patient Data from Published Kaplan-Meier Survival Curves for Economic Evaluations: A Simulation Study Comparison of Methods for Recreating Individual Patient Data — fig. from the paper

§ The Valyu brief

The paper, in five minutes

Read from the full text via Valyu Contents

Guyot and Hoyle‑Henley methods accurately recreate IPD from KM curves, while traditional methods overestimate mean survival.

What’s new

i.Reviewed four methods for recreating individual patient data from published Kaplan-Meier curves.
ii.Conducted a Monte‑Carlo simulation comparing the methods across sample sizes, censoring rates, and Weibull vs. lognormal distributions.
iii.Evaluated bias, mean‑square error, and uncertainty capture for each method.
iv.Provided guidance on method selection, showing traditional methods overestimate mean survival.

How they did it

The authors generated 1,000 simulated survival datasets for each scenario, varying sample size (100, 500, 1,000), censoring rates (26%, 42%, 76%) and using Weibull or lognormal distributions. They applied the four reconstruction methods and direct maximum‑likelihood estimation to the simulated IPD, then measured bias, MSE and uncertainty of the estimated mean survival time.

What they found

All methods gave accurate mean survival estimates when the sample size was 500 and a Weibull distribution was used.
With a sample size of 100 and Weibull distribution, the Guyot method was almost as accurate as the Hoyle‑Henley method, while traditional methods showed greater bias.
For lognormal distribution, the Guyot method produced noticeably less bias and more accurate uncertainty than the Hoyle‑Henley method.
Traditional least‑squares and graphical methods overestimated mean survival, especially at high censoring rates (up to 76%).

Caveats

✕Simulation only included Weibull and lognormal survival models, omitting other plausible distributions.
✕Only three censoring rates (26%, 42%, 76%) were examined, limiting assessment of extreme censoring scenarios.
✕A Monte‑Carlo simulation to directly test the accuracy of the proposed methods was not performed.
✕The study design favored the Guyot method, which may affect the generalizability of the findings.

§ The Valyu brief

Reading the full paper and taking notes. This takes a few seconds…

§ Ask this paper

Ask a question about this paper

Valyu reads the full text and answers from what the paper actually says.

§ Related work in other archives

A Review and Comparison of Methods for Recreating Individual Patient Data from Published Kaplan-Meier Survival Curves for Economic Evaluations: A Simulation Study Comparison of Methods for Recreating Individual Patient Data

The paper, in five minutes

Ask a question about this paper

Similar papers in all four archives

Similar papers in all four archives

The paper, in five minutes