A new Viewpoint article by Ludovic Trinquart and Martin Stockler, published in JAMA Oncology, is reigniting the conversation around data sharing in cancer research, arguing that the widespread withholding of individual patient data (IPD) from randomised clinical trials (RCTs) is no longer justifiable—especially in light of advancing technologies that allow highly accurate reconstruction of survival data from published figures.
The authors argue for a practical, privacy-conscious alternative: the routine publication of stripped-down, deidentified datasets alongside clinical trial results. They propose this as a stopgap that could immediately improve transparency, reproducibility and meta-analytical rigor, without undermining full IPD sharing in the long term.
Persistent Barriers to Access
Despite initiatives from the National Cancer Institute (NCI), NIH, and platforms like Vivli.org and Project Data Sphere, access to cancer trial IPD remains limited. As the authors note, even trials funded by public bodies are subject to complex access procedures involving detailed proposals, multi-party approval and restricted virtual desktop environments. These hurdles are thought to deter many researchers, slowing progress in a field where timely insights could save lives.
The paper critiques this status quo:
“In a field like oncology, where the stakes are high and the need for rapid progress is urgent, the inability to promptly and fully explore IPD from existing trials hampers secondary analyses, meta-analyses, translational science, and more generally, the advancement of treatment and management strategies.”
A Technological Turning Point
Importantly, the authors highlight that modern algorithms can now reverse-engineer pseudo–IPD from Kaplan-Meier survival curves published in trial results—with remarkable accuracy. By extracting time-to-event probabilities and combining them with at-risk numbers and censoring information, researchers can recreate datasets that approximate original IPD with high fidelity.
Although this reconstruction is currently labour-intensive, the process is rapidly being automated using AI and image-processing technologies. In other words, if stripped-down trial data can be recreated from figures, the rationale for withholding original data is becoming untenable.
A New Model for Transparency
To address this gap, the authors advocate for journals to require deidentified, stripped-down datasets—including randomisation groups and time-to-event data for key outcomes (e.g., progression-free and overall survival)—as supplementary files at the time of publication.
They emphasise that:
- The datasets would not include sensitive covariates (e.g., date of birth)
- They would support reproducibility, robust secondary analyses and meta-analysis integration
- Sharing would follow a universal data model to standardise the use
- This would not replace full IPD sharing, but complement it and lower the initial barrier to transparency
“The ability to reconstruct pseudo-data from Kaplan-Meier curves provides a compelling argument for why original trial data that underlie the published time-to-event outcome results of clinical trials should be made readily available at the time of publication,” the authors concluded.
They argue that this model could catalyse a culture shift—just as trial registration became standard after ICMJE enforced registry requirements.
Clinical and Research Impact
The availability of patient-level time-to-event data could enable:
- Reanalysis using alternative measures like restricted mean survival time, especially when proportional hazard assumptions do not hold
- Better modelling of informative censoring
- Integration into meta-analyses and trial design simulations
- Timelier validation of subgroup effects and methodological innovations
While the authors acknowledge challenges—including selective subgroup reporting and the risk of deprioritising full IPD sharing—they maintain that the benefits far outweigh the drawbacks.
This Viewpoint presents a compelling case that in an era of accelerating data science and digital tools, cancer research must move towards greater openness. By routinely releasing core survival data in a structured, deidentified format, medical journals and trial sponsors can help unlock the full scientific value of RCTs—without compromising patient privacy or investigator equity.
Paper: Trinquart L, Stockler MR. The Paradox of Data Sharing in Cancer Randomized Clinical Trials—A Call for Greater Transparency. JAMA Oncol. Published online July 03, 2025. Access online here.
