https://pubmed.ncbi.nlm.nih.gov/42842867/
(Chat GPT; article is open access.)
Former FDA Leader Woodcock Calls for National Effort to Close Precision Cancer Medicine’s Prediction Gap
New commentary argues that matching tumors to drugs is only a starting point—and that federal investment is needed to determine which patients will actually benefit.
Precision cancer medicine has delivered some remarkable treatments. But identifying a tumor’s molecular abnormality still too often leaves a crucial question unanswered: Will the drug selected for that abnormality help this particular patient?
A group led by former Food and Drug Administration official Janet Woodcock is calling for a national effort to close that gap. In a commentary published October 7 in the Journal of Clinical Oncology, the authors urge Congress to support a coordinated infrastructure linking cancer trials, molecular testing, patient specimens and treatment outcomes.
Woodcock spent approximately 38 years at the FDA, from 1986 to 2024, including more than two decades leading its Center for Drug Evaluation and Research. She also served as acting commissioner from January 2021 to February 2022 and subsequently as principal deputy commissioner. Her involvement brings a veteran drug regulator’s perspective to a proposal focused on strengthening the evidence behind treatment selection.
The commentary, coauthored by researchers, oncologists, industry representatives, policy specialists and patient advocates, presents a policy proposal rather than findings from a new clinical trial.
A Molecular Match Still Leaves Questions
The authors’ central argument is that medicine has invested heavily in developing cancer drugs while investing too little in understanding who benefits from them.
“Matching a detected alteration to a drug is not the same as predicting benefit for that patient,” they write.
They acknowledge the major successes of molecularly targeted treatment, including imatinib for certain leukemias, trastuzumab for HER2-positive breast cancer and osimertinib for EGFR-mutant lung cancer. In some cancers, a dominant molecular driver provides a particularly strong basis for choosing treatment.
Elsewhere, the connection is less dependable. A positive biomarker may establish eligibility for a drug without providing a reliable estimate of an individual patient’s benefit—or clarifying whether another treatment would be better.
The authors cite a 2017 analysis of 35 FDA oncology approvals requiring pharmacogenomic testing. Twenty-four relied entirely on trials enrolling biomarker-positive patients. Such studies can demonstrate that a treatment works in the selected population, but leave important questions about patients without the biomarker and differences among those who have it.
The problem becomes especially consequential when clinicians must weigh effectiveness against toxicity. In advanced melanoma, the authors note, available biomarkers do not reliably determine which patients need a more toxic combination of immunotherapies rather than a single agent.
Cancer Is a Moving Target
Cancer biology further complicates prediction. The same mutation can have different implications in different tissues. Tumors evolve during treatment, and a single biopsy may capture only part of their diversity. Molecular testing at one moment cannot fully describe a disease that continues to change.
More comprehensive approaches—including repeated molecular measurements, imaging and artificial intelligence—could improve prediction, the authors argue. But sophisticated analytical tools still need rigorous validation in the specific cancers and treatment settings where they will be used.
The challenge is therefore larger than discovering more biomarkers or developing more powerful algorithms. Researchers need connected biological and clinical information that allows them to test whether a proposed predictor actually improves treatment decisions.
Why the Market Underinvests in Better Prediction
The barriers are financial and organizational as well as scientific.
Drug developers have strong incentives to bring therapies to market. Better predictive tests, however, might reduce the number of patients eligible to receive those drugs. Diagnostic payment also frequently reflects the testing procedure rather than the downstream value of avoiding ineffective treatment, the authors contend.
Meanwhile, the laboratory investigations embedded in clinical trials to explain response and resistance are often optional, inconsistently funded and vulnerable to budget cuts. Patient records, molecular results, imaging and trial databases remain poorly connected.
The authors describe this as a collective problem that no single company, payer or research institution can resolve.
A Federal Infrastructure for Learning Who Benefits
The proposed federal effort would make those investigations a core part of cancer research. It would support standardized specimen collection, molecular profiling over time, interoperable data systems and prospective validation of predictive biomarkers.
The authors identify the Advanced Research Projects Agency for Health, or ARPA-H, as a particularly promising coordinating vehicle. Other options include expanding National Cancer Institute trial infrastructure, public-private partnerships and Medicare coverage arrangements that support evidence generation.
The effort would build on existing research networks. Initial projects could focus on settings where treatment choices remain uncertain and carry substantial toxicity or cost.
To encourage industry participation, the authors suggest tax credits or limited exclusivity extensions conditioned on rigorous validation, shared outcome data and governed access to research specimens.
The proposal grew from a November 2024 precision medicine town hall and a subsequent multidisciplinary think tank. Several authors disclose relationships with pharmaceutical, diagnostic or biotechnology companies.
Ultimately, the authors want progress measured by better treatment decisions and validated predictive tools, alongside new drug approvals. For patients facing several possible therapies, the practical goal is evidence that helps their oncologist choose the treatment most likely to help—and avoid treatment unlikely to do so.
Sidebar: Five Surprising Points
1. A companion diagnostic can be an eligibility gate without being a dependable personal forecast.
A positive result may identify a population in which a drug works, yet leave considerable uncertainty about which individuals will benefit. The authors distinguish these two functions throughout the article.
2. The same mutation can lead to different treatment outcomes in different cancers.
The authors contrast BRAF V600E–mutant melanoma with colorectal cancer. In colorectal cancer, feedback through another signaling pathway can undermine treatment aimed at BRAF. The mutation’s meaning depends on its biological surroundings.
3. A better diagnostic could shrink a drug’s market.
A test that identifies likely nonresponders might spare patients ineffective treatment while reducing the commercially eligible population. The authors argue that this creates an incentive problem for investment in better prediction.
4. Even innovative trial designs cannot compensate for missing biological data.
Adaptive trials and sophisticated statistics can improve research, but they cannot recover specimens or molecular measurements that were never collected. The proposal makes that collection a foundational part of clinical trials.
5. The authors would consider rewarding better diagnostics with longer drug exclusivity.
They propose exploring limited exclusivity extensions or tax credits tied to prospective biomarker validation and data sharing. The aim is to make identifying the right patients commercially rewarding, as well as clinically valuable.
