Researchers Found Limited Code Sharing in Cancer Studies

A new study reveals that most clinical prediction model research fails to provide reproducible underlying code.

Updated on Oct. 13, 2026 in Cancer

Isometric editorial illustration featuring three clustered rectilinear volumes in teal and mustard, representing the structure of clinical research data transparency. — Only 12% of clinical prediction model research papers shared their underlying code, hindering scientific reproducibility and the verification of patient care findings. AI Illustration. Upload story photo >

Live Poll

Should medical researchers be required to share their underlying code to verify clinical study findings?

Researchers from the Dana-Farber Cancer Institute and MIT have published findings showing that only 12% of reviewed clinical prediction model studies shared their underlying code. The study analyzed approximately 4,000 published research papers to assess code availability and reproducibility.

Why it matters

The lack of accessible and reproducible code hampers the ability of the scientific community to verify findings in clinical research. This transparency gap can slow down the development and adoption of accurate predictive models for patient care.

A review of 4,000 published studies found that only 12% shared underlying code. Even when shared, the code was rarely found to be reproducible.

The players

Dana-Farber Cancer Institute

This is a world-renowned cancer treatment and research center based in Boston that focuses on patient care and innovative scientific study.

MIT

The Massachusetts Institute of Technology is a leading research university known for its advancements in technology, science, and interdisciplinary collaboration.

National Institutes of Health

This is the primary agency of the United States government responsible for biomedical and public health research.

The details

Using an AI-assisted approach, researchers examined clinical prediction models to check for transparency and scientific rigor. The study, which received funding from the National Institutes of Health, found that code sharing prevalence varied significantly depending on the journal and the country of origin.

Study results were published in Nature Medicine on October 13, 2026.

The Big Picture

The findings reflect a ongoing struggle within the reproducibility crisis in clinical research, where software transparency continues to lag behind established academic standards. By quantifying the lack of accessible code, this study reinforces the necessity for more rigorous documentation in computational medical science.

The lack of reproducible models means that clinicians may have difficulty validating new diagnostic tools before they are used in real-world patient settings. As the scientific community addresses these gaps, future medical technologies may undergo more rigorous scrutiny to ensure safety and reliability.

The takeaway

Researchers and institutions should prioritize open-science practices to improve the reliability of clinical modeling. Ensuring that code is both accessible and reproducible is essential for transforming research findings into safe and effective clinical tools.