EUVIMEDEUVIMED entdecken
Metaanalyse

RESOLVE-IPD: Individual Patient Data Reconstruction and Uncertainty-Aware Subgroup Meta-Analysis

Lang Lang iD, Yao Zhao, Qiuxin Gao iD, Yanxun Xu iD

JCO Precision Oncology · 2026-11 · 10 (11)

DOI: 10.1200/po-26-00267 ↗

Kurz gefasst

Metaanalyse in der Zeitschrift JCO Precision Oncology, 2026-11. Verfasst von Lang Lang und 3 weiteren Autor:innen. Beteiligte Einrichtung: Department of Applied Mathematics and Statistics, Johns Hopkins University, Baltimore, MD. Hauptthemen laut MEDLINE: Clinical Trials as Topic, Data Science. Die Arbeit nennt 24 Literaturangaben.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Purpose

Individual patient data (IPD) from oncology trials are essential for reliable evidence synthesis but are rarely publicly available, necessitating reconstruction from published Kaplan-Meier (KM) curves. Existing reconstruction methods can be affected by digitization errors, strong uniform censoring assumptions, and the inability to recover subgroup-level pseudo-IPD when only aggregate statistics are available. We developed RESOLVE-IPD, a computational framework that combines vector-based KM extraction, censoring-informed survival reconstruction, and uncertainty-aware subgroup recovery for pseudo-IPD reconstruction and subgroup meta-analysis.

Methods

RESOLVE-IPD transforms published survival data into reconstructed pseudo-IPD. The framework uses VEC-KM for vector-graphic coordinate extraction and CEN-KM to incorporate censoring marks and reported risk tables into overall-population reconstruction. To address subgroup evidence recovery, we introduced MAPLE, an algorithm that identifies an ensemble of plausible subgroup labelings consistent with reported aggregate summaries. RESOLVE-IPD enables downstream IPD-based evaluations, including subgroup meta-analyses that quantify and propagate uncertainty inherent in the reconstruction process.

Results

RESOLVE-IPD was demonstrated through subgroup meta-analyses of four trials in advanced esophageal squamous cell carcinoma, focusing on patients with low programmed death-ligand 1 expression and patients age ≥65 years. VEC-KM and CEN-KM reconstructed overall survival pseudo-IPD that closely reproduced published KM curves and reported summary statistics. MAPLE produced plausible subgroup reconstructions consistent with reported hazard ratios and survival estimates. Uncertainty-propagating meta-analysis generated pooled survival and interval-specific hazard-ratio estimates while preserving subgroup reconstruction uncertainty.

Conclusion

RESOLVE-IPD provides a framework for reconstructing overall and subgroup pseudo-IPD from published survival information and for propagating reconstruction uncertainty, thereby supporting more robust and transparent secondary evidence synthesis in precision oncology.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Clinical Trials as TopicData ScienceHumansAlgorithmsUncertaintyData Analysis

Autor:innen und Institutionen

  • Department of Applied Mathematics and Statistics, Johns Hopkins University, Baltimore, MD.
  • Data Science & AI Institute, Johns Hopkins University School of Medicine, Baltimore, MD.
  • Division of Quantitative Sciences, Department of Oncology, Johns Hopkins University School of Medicine, Baltimore, MD.

ORCID-Profile: Lang Lang, Qiuxin Gao, Yanxun Xu

Zitieren

APA

Lang, L., Zhao, Y., Gao, Q., & Xu, Y. (2026). RESOLVE-IPD: Individual Patient Data Reconstruction and Uncertainty-Aware Subgroup Meta-Analysis. JCO Precision Oncology, 10(11). https://doi.org/10.1200/po-26-00267

Vancouver

Lang L, Zhao Y, Gao Q, Xu Y. RESOLVE-IPD: Individual Patient Data Reconstruction and Uncertainty-Aware Subgroup Meta-Analysis. JCO Precision Oncology. 2026;10(11). doi:10.1200/po-26-00267

BibTeX
@article{Lang2026,
  author = {Lang, Lang and Zhao, Yao and Gao, Qiuxin and Xu, Yanxun},
  title = {RESOLVE-IPD: Individual Patient Data Reconstruction and Uncertainty-Aware Subgroup Meta-Analysis},
  journal = {JCO Precision Oncology},
  year = {2026},
  volume = {10},
  number = {11},
  doi = {10.1200/po-26-00267}
}
Kostenlos starten →