A clinician’s perspective on what we’re learning from ctDNA and what it means for patients with cancer
When circulating tumor DNA (ctDNA) for molecular residual disease (MRD) first emerged as a promising biomarker in oncology, it was largely seen as a research tool – one with significant, but largely unknown, clinical utility. That’s changing fast. Today, ctDNA is beginning to reshape how we think about disease monitoring, recurrence risk and personalized therapy decisions. It’s no longer just a window into the past – it’s a real-time lens into the future of patient care.
Unlike traditional liquid biopsies for next-generation sequencing (NGS), which search for known or actionable mutations using a fixed gene panel, MRD-focused ctDNA testing is “tumor-informed,” meaning it identifies an individual’s unique tumor signature and then tracks those markers over time. This distinction is what allows ctDNA to serve as a real-time indicator of recurrence risk rather than a fixed genetic snapshot at the time of testing.
When I first started working with ctDNA as a clinician and researcher, I found myself asking: What does this actually look like in the real world? How are clinicians using ctDNA in practice – not just in trials? Are patients with positive results being treated differently? Can we learn anything about outcomes by analyzing testing patterns across Flatiron’s network?
That curiosity has since sparked something bigger. With Flatiron’s commitment to fostering innovation – and the support of incredible teammates – we developed a machine learning (ML) model that could extract ctDNA testing details from unstructured electronic health record (EHR) data. What started as a proof-of-concept quickly grew into a scalable AI capability, a company-wide collaboration, and some of the largest real-world studies of ctDNA testing to date.
Since then, my focus has been on understanding not just the technical implementation, but the clinical implications: What drives a clinician to act on a positive result? Is testing being ordered as part of routine surveillance, in response to symptoms, or based on patient profile, practice setting, or payer coverage? And most importantly – how can real-world data clarify ctDNA’s role in cancer care today while shaping where the field goes next?
The answers aren’t always straightforward, but the signal is growing stronger.
From potential to practice: how ctDNA is being used today
We’re now seeing ctDNA used to guide treatment decisions in ways that weren’t possible just a few years ago. In the adjuvant setting (i.e., after surgery but before signs of recurrence), ctDNA is helping clinicians identify patients who may still harbor microscopic disease even if scans look clear. This can support more confident decisions about whether to safely escalate or even de-escalate therapy, helping prevent both under- and over-treatment, which can cause toxicities and unnecessary adverse events.
We’re also seeing anecdotal evidence of clinicians initiating treatment based on ctDNA results alone – even in the absence of radiographic progression. This represents both a meaningful practice shift but also a new level of complexity. While not yet widespread, this emerging behavior highlights the tension between traditional endpoints and molecular signals – a dynamic we’re studying closely in the real world
At Flatiron, we’re uniquely positioned to explore these questions at scale. Our ML and large language model (LLM) tooling allow us to extract nuanced clinical details from unstructured data, giving us insight into how ctDNA is being used across the care continuum in real-time. We’re actively analyzing and researching real-world ctDNA to better understand utilization patterns, correlations with outcomes, and how this emerging technology is being integrated into clinical pathways ahead of formal guideline adoption.
Why it matters: a real-time, actionable biomarker
As with any new technology, it takes time to understand not only how to use ctDNA but how to ensure it has the greatest impact for patients. ctDNA is not yet part of standard treatment guidelines, but already we’re seeing signs in the data that it’s influencing real-world decisions. And these aren’t just anecdotes – they’re observable signals that the field is making progress towards integrating this new technology into practice.
Because of this, we’re able to detect patterns and generate insights before clinical trial results fully mature. When something new enters the care landscape – whether it’s a drug, a biomarker, or a testing modality – Flatiron can be among the first to understand how it’s actually being used in practice and what that might mean for patients and providers.
What makes ctDNA so compelling isn’t just its ability to detect residual disease or early recurrence – it’s the timing. ctDNA can often provide a signal of molecular recurrence well before clinical symptoms or imaging results. And in some cancers, it may be the only early signal available. That early signal is what makes ctDNA so promising – it creates the possibility to act sooner, at a time when interventions may matter most.
For patients, that can mean earlier answers. For clinicians, an added layer of information to guide difficult decisions. And for researchers and developers, a foundation for new models of risk stratification, trial design, and treatment adaptation. But to realize this potential, we also need clarity.
I’ll never forget a conversation I had at ASCO 2024 with a fellow clinician who wanted to discuss our ctDNA research in colorectal cancer (CRC). She had recently been diagnosed with CRC herself and told me she had positive ctDNA results but her scans showed no evidence of disease.
“We’re terrified,” she said. “We don’t know what to do.”
Her words have stayed with me – a reminder of both the promise and the weight of these results, and how we still have so much to learn about them. Her story really gets at the heart of what makes ctDNA so exciting and yet so challenging. For patients, an early positive test can be empowering but also deeply unsettling when clinical guidance is unclear. For clinicians, it forces hard conversations about whether – and how – to act before traditional testing modalities show visible recurrence.
These tensions are exactly why generating scalable, accurate real world evidence quickly is so critical. Despite promising applications, ctDNA still raises important considerations. A positive result in early-stage disease can be difficult to act on, while a negative result may not rule out recurrence in low-shedding tumors. That’s why at Flatiron, we’re working to connect ctDNA signals with real-world clinical decisions and build the evidence to guide how results are interpreted and used in practice.
Accelerating integration: Flatiron and the expanding ctDNA ecosystem
There’s growing recognition that ctDNA has the potential to transform how we monitor cancer, personalize treatment, and make earlier, more informed decisions. This shift is reflected across real-world practice, prospective trial design, and even policy – such as the recent expansion of Medicare coverage for Natera’s Signatera assay, signaling institutional trust in ctDNA as a meaningful clinical tool.
But to fully integrate ctDNA into clinical pathways and treatment guidelines, there’s still work to be done. To get there, we’ll need:
- Prospective validation of survival benefit from acting on ctDNA
- Clarity on actionability vs. prognostic value (i.e., when and how to intervene)
- Guidance on testing frequency, timing, and thresholds
- Greater equity in access, as real-world uptake remains highly variable across regions, providers, and payers
At Flatiron, our shared goal is to generate the evidence needed to bring ctDNA from an emerging biomarker to an embedded clinical standard. That means leveraging real-world data to understand how ctDNA is being used today and partnering with MRD vendors to power studies that can shape tomorrow’s guidelines.
We’re also building infrastructure to support longitudinal ctDNA tracking across diverse tumor types and sample formats, and we’re exploring how newer modalities - like urine or pleural fluid - might expand possibilities as the science evolves. And through scaled ML- and LLM-based extraction, we’re unlocking deeper insights into testing rationale, utilization patterns, and clinical actionability across solid and hematologic cancers.
This isn't just happening in our data – the entire ctDNA for MRD ecosystem is gaining momentum. For example, Quest Diagnostics’ Haystack MRD test recently earned Breakthrough Device designation from the FDA for stage II colorectal cancer, signaling regulatory confidence in real-world utility. Meanwhile, Natera’s Signatera assay now has expanded Medicare coverage, further validating its role as a clinical tool. And the Myriad Precise MRD test was recently featured in The Lancet Oncology, showing promising findings in oligometastatic renal cell carcinoma – highlighting ctDNA’s potential to guide treatment de-escalation.
Taken together, these advances show how quickly the field is evolving – and why Flatiron’s role in generating real-world evidence is so critical to ensure this momentum translates into practice.
Shaping the future of ctDNA
We’re at a turning point. ctDNA is no longer a future-facing concept. It’s here, it’s influencing care, and it’s reshaping how we monitor cancer, define recurrence, and make decisions about treatment.
And yet, there’s still so much we don’t know.
That’s where real-world data can help. By learning from the lived experience of every person with cancer – across settings, timepoints, and decisions – we can translate signals into strategy, and insight into action.
In an era defined by AI and precision oncology, we don’t just need to keep pace with change. We have the opportunity to be at the forefront – generating insights, driving understanding, and helping shape the real-world practice paradigm.
Let’s keep building – together.
To learn more about how real-world evidence can support your research priorities, reach out to us.
