Adoption of methods for reconstructing individual patient data from Kaplan-Meier curves: a bibliometric study of 390 publications

Background. Methods for reconstructing individual patient data (IPD) from published Kaplan-Meier curves have increasingly been used to perform secondary survival analyses when original patient-level datasets are unavailable. Several reconstruction algorithms have been developed, including the Guyot method and, more recently, IPDfromKM, which provides a freely accessible implementation. The extent to which these methods have been adopted in the medical literature has not been extensively characterized. Objective. To assess the adoption of methods for reconstructing IPD from published Kaplan-Meier curves and to describe their distribution across major therapeutic areas. Methods. PubMed was searched on 28 August 2026 using the query ‘reconstruct* AND Kaplan AND curve*’. The search covered PubMed from inception. Publications from the preceding 10 years were screened by two reviewers to determine whether they used a documented method to reconstruct patient-level data from Kaplan-Meier curves, together with any supplementary information reported in numbers-at-risk tables or in the text and tables of the original publication. Eligible studies were classified into three broad therapeutic areas: oncology/hematology, cardiology/cardiovascular medicine, and other clinical areas. Results. The PubMed search identified 999 records, of which 867 had been published during the preceding 10 years. Among these, 390 publications met the eligibility criterion. Of the 390 studies, 211 (54%) concerned oncology or hematology, 137 (35%) cardiology or cardiovascular medicine, and 42 (11%) other clinical areas. The annual number of eligible publications increased markedly from 2022 onward and reached its highest level in 2025. Conclusions. Reconstruction of IPD from published Kaplan-Meier curves has become widely used in the medical literature, particularly in oncology and cardiology. The marked recent increase in publications employing this approach indicates that reconstructed patient-level survival data have become an increasingly established tool for secondary analyses of published time-to-event evidence.