Data-sharing, reanalysis and sensitivity analysis of overall survival and progression-free survival of pivotal randomized controlled trials in oncology.

Clinical trial data sharing has improved in recent years, but many trials still do not make their data fully available. When data are not shared, it becomes difficult for independent researchers to check whether the results can be reproduced. Reproducibility means that when another research team uses the same information, they should reach the same conclusions. This process is essential for building trust in the evidence used to approve new medicines.

Cancer affects millions of people worldwide, and new cancer medicines are regularly reviewed by the European Medicines Agency (EMA). To decide whether to grant marketing authorisation for medicines, the EMA relies on evidence from clinical studies, specifically randomised controlled trials (RCTs). These trials compare a new treatment to a standard treatment or a placebo to see whether the new medicine helps people live longer or slows the growth of their cancer. Because many patients rely on these decisions, the results of these trials must be reproducible and based on solid methods.

In this project, we will study how often data from these cancer trials are shared and whether the results can be reproduced. We will focus on requesting 60 RCTs in oncology that led to positive opinions from the EMA’s Committee for Medicinal Products for Human Use (CHMP) between January 1, 2020, and December 31, 2024 to reproduce their results.

For each randomly selected trial, we will request the Individual Participant Data (IPD), protocols and all the relevant documentation on how the study was conducted, and we will use it to reanalyse, according to the methods used in the original analysis, the two major outcomes used in cancer research:

• Overall Survival (OS), which measures how long patients live after joining the trial.

• Progression-Free Survival (PFS), which measures how long patients live before their cancer worsens.

We will repeat the original analyses to see whether we reach the same findings. We will then examine how censoring may affect the results. Censoring occurs when information about a participant stops being collected, such as when a participant leaves the study early; censoring rules define on which occasions and from which time point data about a participant stops being included in the analysis.
We will use a method called reverse Kaplan–Meier analysis, which switches the roles of events and censoring, to help us understand whether one treatment group presented more censoring than the other. Imbalances in censoring can introduce attrition bias, which means the treatment groups are not being compared fairly.

Finally, we will reanalyse the IPD using different censoring rules and compare all results. This will allow us to assess how robust, or stable, the study conclusions are when censoring is modified. Our goal is to evaluate whether data is actually shared and how reproducible the evidence supporting new cancer medicines is with the information available. This work will support greater transparency and strengthen trust in clinical trial findings that guide cancer treatment decisions.