OBJECTIVES: The Guyot method is a widely used algorithm to generate pseudo individual patient data (IPD) from published Kaplan-Meier survival curves. Lack of guidance on the preparation of input data for this method drives high variability in user implementation due to the nuances in user tracing of published curves and manual edits of digitizer application output. This study aimed to provide standardized guidance and a data preparation function to use alongside the Guyot algorithm, and to test its performance for reducing errors and improving accuracy.

METHODS: A guidelines document and checklist were created based on research experience to outline steps for curve tracing, data processing, reporting standardization, and quality control. The data preparation function was developed and subsequently tested using mock data to eliminate recurring errors.

RESULTS: The data preparation function modifies the input by: adding the number at risk table information at time zero as the number of patients allocated to the trial arm, if not already provided; excluding survival probabilities above one, ensuring all curves begin with a reference point at time zero with 100% probability of survival; interpolating the survival probability vector to ensure a survival probability is entered each time there is information about the number of patients at risk; correcting digitization errors caused by imperfect tracing by removing outliers and ensuring the curve is monotonically decreasing. The guidelines document and checklist provide detailed instructions for curve tracing, reconstruction, and validation to reduce variability across results.

CONCLUSIONS: The guideline documents and data preparation function reduced errors and improved standardization when using Guyot to generate pseudo IPD. It also reduced time spent to learn the method and the need for troubleshooting issues. Without manual modifications and error mitigation, this function improved replicability of pseudo IPD output across iterations.