Insights
July 26, 2026
Douglas Fridsma, MD, PhD · Chief Medical & Science Officer, Health Universe
The Cancer Data Divide: Why a 5-Year-Old Has Better Data Rights Than a 65-Year-Old
Childhood cancer survival has soared from under 30% to roughly 85% in five decades. The secret is not just better drugs; it's a data-sharing system that adult oncology still hasn't built. This is Part 1 of a two-part series.
Two cancer patients, two completely different futures
Imagine two cancer patients diagnosed this week — one who is five years old, and one who is 65.
The five-year-old's data, including tumor genomics, imaging, biospecimens, and clinical history, will be harmonized, multi-omic, and queryable by researchers around the world within months. The data will be shared by default across dozens of institutions under a single governance framework; it will contribute to a dataset that has helped lift pediatric cancer survival from under 30% in the 1970s to roughly 85% today.
The 65-year-old's data will likely sit in an institutional tumor bank, governed by a registry use agreement, accessible only after a months-long data-use negotiation — if it's accessible at all. Their cancer center may be licensing it to a data broker, or powering an AI model they'll never benefit from. Most likely, it is simply not flowing to the researchers who could use it.
- 85% — 5-year survival for childhood cancer today, up from under 30% in the 1970s
- 7.1% — of adult cancer patients enrolled in treatment trials nationally
- 86%+ — pediatric trial enrollment rates, for many tumor types
The asymmetry isn't biology. It's an infrastructure choice.
The difference between how a child's data is handled and how an adult's data is handled seems crazy. Why don't we treat all cancer data the same? Understanding how it happened is the first step to fixing it.
How pediatric oncology made data sharing the default
At the most recent Ci4CC Summit, Adam Resnick, co-director of the Center for Data-Driven Discovery in Biomedicine (D3b) at Children's Hospital of Philadelphia (CHOP), laid out in precise detail how his team built a working pediatric cancer data commons. This was not a vision presentation, but a running system, with funded budgets, published governance documents, and measurable outcomes.
The pediatric community got here by making a series of critical choices that started decades ago, in which patient data was treated as a shared research commons rather than an institutional asset. Four principles drove every decision:
- Data is shared by default. The exception requires justification, not the other way around.
- One governance framework across many institutions, not dozens of bilateral agreements.
- Data is released and coded early and continuously, not after embargo.
- Patient families are treated as partners. Many of the foundational partnerships were initiated by parent advocacy organizations.
There was also a structural forcing function: pediatric cancer is rare. No single institution sees enough cases to learn alone. Siloing data isn't just inefficient — it's a survival risk.
As Resnick put it: "Every child, every time, everywhere." Three words that encode an entire governance philosophy. The result is that children's cancer data is rarely monetized, and the incentives for sharing outweigh any institutional benefit to keeping it. Those incentives drive sharing and the use of children's data for research and to improve care.
What that philosophy built: a working data commons
D3b runs four interlocking systems. Together they form one of the most functional cancer data-sharing ecosystems in medicine. The breakdown below defines what each piece does and why it matters for anyone building health AI.
The Children's Brain Tumor Network (CBTN)
Thirty-two member institutions across four countries, operating under one shared regulatory governance architecture at CHOP. Over 4,700 subjects enrolled, more than 65,000 biospecimen aliquots collected, and multi-omic data on 1,000+ tumors openly accessible — forming the largest open-access pediatric brain tumor dataset in existence. CBTN unites 32 institutions under one governance framework, rather than 32 bilateral agreements.
The Gabriella Miller Kids First Data Resource Center
A cloud-based, open-access ecosystem holding harmonized data from more than 34,000 participants. Roughly 1.3 petabytes of genomic and clinical data are queryable across pediatric cancer and rare disease cohorts, available to any qualified researcher, anywhere.
CAVATICA
A cloud analysis platform that lets investigators run bioinformatic work in the browser without requiring a private HPC cluster. Crucially, it lets pediatric data interoperate with adult cancer datasets and NIH resources, extending the reach of every tool in the ecosystem.
OpenPBTA
A global open-science analysis of 1,074 pediatric brain tumors from 943 patients, all coded on GitHub and processed on PedcBioPortal. The analytic notebooks and the manuscript were written in the open as the work happened. This is the only large-scale cancer atlas I know of where that's true.
RADIANT (and what comes next)
In October 2024, D3b received an ARPA-H award of up to $10 million to extend this model to real-time clinical decision support across cancer and rare disease. The distance between scientific discovery and patient care is being shortened deliberately, systematically, with funding attached.
Data for the Common Good — the Pediatric Cancer Data Commons
Beyond CHOP, other institutions are contributing too. The Pediatric Cancer Data Commons is a University of Chicago platform harmonizing pediatric, AYA, and adult cancer clinical data from hundreds of collaborators across forty-plus countries. More than a dozen disease-specific consortia — neuroblastoma (INRG), soft tissue sarcoma (INSTRuCT), and others — operate under one shared governance model with NCI-aligned data dictionaries, turning siloed trial data into a cross-disease, queryable resource for the rarest cancers.
When sharing is the default, science compounds
Health AI is only as good as the data it can access. The D3b model demonstrates something that should reshape how we think about oncology data governance: when sharing is the default, discoveries build on each other. When restriction is the default, science fragments — and patients pay the price.
Pediatric oncology patients cannot consent for themselves; their parents do it on their behalf. The pediatric community designed an entire ecosystem around the assumption that those parents want their child's data to help every other child. Adult patients can consent for themselves — and when asked, 70 to 97% say yes.
The pediatric community built a system that listens to what patients want. It's time for adult oncology to do the same.
In Part 2, we'll look at exactly what's blocking adult cancer data sharing — data monetization, DUA gridlock, EHR vendor lock-in — and what federal policymakers, cancer center leadership, and patients themselves can do to break it open.
About the author — Douglas Fridsma, MD, PhD, is Chief Science Officer & Chief Medical Officer at Health Universe. He previously served as Chief Science Officer at the American Medical Informatics Association (AMIA) and as Director of the Office of Standards and Interoperability at ONC.