Adult Cancer Data Is Being Locked Away. Here's How to Fix It.

Seventy percent of adult cancer patients want their data shared for research. So why are only 7% enrolled in clinical trials? The answer isn't biology. This is Part 2 of a two-part series.

In Part 1, I described what Adam Resnick and the D3b team built at Children's Hospital of Philadelphia (CHOP): a working pediatric cancer data commons with 34,000+ participants, 1.3 petabytes of queryable data, and clinical-trial enrollment rates that hit 86% or higher for many pediatric cancers. Over five decades, that infrastructure helped lift pediatric 5-year survival from under 30% to roughly 85%.

Now for the inverse picture, and a direct question: if the model works for pediatric patients, why haven't we replicated it in adult oncology?

The numbers don't add up

Joseph Unger and colleagues at Fred Hutch published the most recent national estimate of adult cancer trial enrollment in the Journal of Clinical Oncology in April 2024. Drawing on accreditation data from roughly 1,200 Commission on Cancer programs that covered over 70% of all U.S. cancer diagnoses, their finding was this:

That 7.1% was reported as good news — roughly double the historical 2.3% estimate. But the structural inequity buried in that headline is stark: where you receive care almost entirely determines whether you have access to a trial. And for AI-powered cancer research tools, data that never enters a trial or a commons is data that never trains a model, evaluates a treatment path, or measures outcomes — further slowing the advancement of cancer care.

Now compare that to what adult cancer patients actually say they want:

Patients overwhelmingly want their data used. But the current system makes that hard.

Four categories of friction — none of them the patient

If patients want to share, and trial enrollment is structurally constrained, what is actually blocking the flow? I've identified four key blockers, and none of them are patient reluctance.

1. Data as a profit center

Many of the largest academic and NCI-designated cancer centers have, over the past five to seven years, built out data commercialization functions. Pathology slide archives, tumor genomic datasets, longitudinal clinical registries, and imaging libraries are increasingly governed by institutional licensing programs. Sometimes these organizations participate with equity stakes in AI spinouts; sometimes through tiered-access consortia that share among members but charge or restrict outsiders.

The defense is that these arrangements fund research. The cost is that the data your patient consented to contribute to "research" now lives behind a paywall for most researchers.

2. EHR vendors building data-monetization platforms

Many EHR vendors and oncology-focused health IT firms have announced explicit data-licensing products through 2024 and 2025, positioning de-identified patient data as a recurring revenue line for health systems. The architecture is straightforward: the vendor aggregates de-identified data from health-system customers and sells access to life sciences and pharma. The health system gets a revenue share. The patient is left behind, and the data is now more restricted than before, because the vendor's commercial terms control downstream redistribution.

3. Data brokers and real-world-data consolidators

A growing class of intermediaries aggregates de-identified oncology data — claims, EHR, genomic, and mortality data — and licenses it under exclusive or semi-exclusive terms. The data exists, but isn't accessible at academic research prices, or under terms that allow open-science publication of derived models. Health AI companies building on this data inherit its restrictions.

4. Data-use agreement friction between academic centers

Even when two academic centers want to share, the contractual machinery is challenging. Data use agreements (DUAs) routinely take months to negotiate. Each institution's Office of Sponsored Programs has its own template, indemnification language, publication clauses, and IP terms. The literature describes cases that "fall apart after months or years of discussions."

A pediatric brain tumor sample collected at CHOP today is harmonized, multi-omic, and queryable on PedcBioPortal within months. An adult glioblastoma sample collected three blocks away at a comparable academic center may sit in a tumor bank behind a months-long DUA negotiation. The result is two patients in the same city with fundamentally different futures.

Policy and sub-regulatory levers that can actually move the needle

The pediatric community proved the model. Now the question is which levers move adult oncology fastest toward the same standard. Here's how to prioritize them.

A direct call to action

This argument has been discussed for years and still hasn't moved the needle. Below are my recommendations for ecosystem stakeholders.

To federal policymakers

Add a research purpose-of-use to TEFCA, and go beyond auditing DMS plans — make them scorable, tangible, and part of successful grants. Condition Cancer Center Support Grant renewals on demonstrated cross-institutional sharing. Pursue information-blocking enforcement against research-data hoarding the same way you'd pursue it against clinical-record hoarding.

To cancer center CIOs, CMIOs, CROs, and general counsel

Adopt model DUAs by default and publish your data-sharing posture. Audit your institution's licensing arrangements and ask, honestly, whether they accelerate or hinder cures. Clinical care and research outcomes are the metrics by which data value should be measured — not monetization.

To EHR and oncology IT vendors

Move toward an open philosophy and stop perpetuating data-monetization platforms that lock your customers' data into your commercial terms. The market for genuinely interoperable, research-friendly oncology platforms is wide open. Someone is going to fill it — it should be you.

To patients and advocates

Ask the right questions of your cancer center: Where does your data go after de-identification? Do they contribute to open research commons? Is your data being licensed to brokers? Demand an opt-in to a research donation registry. You have the right of access — use it, and then share your data.

To Adam Resnick and the D3b team

You showed, at Ci4CC and over the past decade, what the rest of oncology should look like. The pediatric community proved the model. Now the adult side must follow.

Every cancer patient I've ever met, when asked about sharing their data to benefit research and other patients, said yes. The pediatric community built a system that listens. It's time for adult oncology to catch up.

Sources

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.