What ctDNA Minimal Residual Disease Can and Cannot Do

Circulating tumor DNA, or ctDNA, consists of tumor-derived DNA fragments found within cell-free DNA in blood. After surgery, radiation, or other curative-intent treatment, a sufficiently specific ctDNA signal can indicate molecular residual disease even when routine imaging shows no tumor.

That makes ctDNA a powerful risk marker. A positive result after treatment is often associated with a much higher recurrence rate. But prediction is not treatment proof. A test can identify risk accurately while the best action remains unknown, and a negative result can miss disease below its detection limit.

What is circulating in the tube#

Most cell-free DNA in plasma comes from normal blood-forming and other cells. The tumor-derived fraction can be tiny, especially after apparently complete treatment. Assays search for mutations, methylation patterns, fragment features, copy-number changes, or combinations that distinguish tumor DNA from background.

“Minimal residual disease” is established language in blood cancers, but solid-tumor blood testing is often called molecular residual disease or measurable residual disease. “Minimal” does not specify a fixed number of cells. A positive plasma signal means a validated tumor-linked feature was detected above an assay threshold in a stated context.

The test does not tell you whether one microscopic cluster or many sites released the DNA. It usually cannot locate disease. It also cannot prove that every detected fragment came from viable cells capable of producing recurrence, although persistent postoperative positivity is strongly prognostic in several cohorts.

Tumor-informed and tumor-naive approaches#

A tumor-informed assay first sequences the removed tumor, often with matched normal material, and selects patient-specific alterations to track in plasma. Following multiple known tumor variants can improve specificity and sensitivity at very low signal levels. It requires adequate tumor tissue, extra processing, and time to build the personalized panel.

A tumor-naive, tumor-uninformed, or plasma-only assay does not require prior tumor sequencing. It may use a fixed panel, methylation, fragment patterns, genome-wide methods, or combined signals. It can be faster and available when tissue is limited, but it must distinguish tumor signal from inherited variants and blood-cell clones through other methods. Neither architecture is inherently best for every use: performance depends on the cancer, the stage, the desired turnaround, the number and type of tracked features, the blood volume, the background correction, the laboratory process, and the decision threshold.

Why timing matters#

Surgery and tissue injury release abundant cell-free DNA, which can dilute a small tumor fraction. Blood collected too soon can therefore have lower analytical sensitivity. Waiting longer can improve signal-to-background ratio, but it can delay an adjuvant-treatment decision.

Chemotherapy and radiation can change shedding and reduce ctDNA. A sample during treatment answers a different question from one after treatment, and the FDA's 2024 final guidance emphasizes prespecified collection timing, assay details, handling, and context of use in early-stage solid-tumor trials.

Serial testing can detect a signal missed on one draw and can distinguish transient from persistent findings, though it also creates repeated opportunities for false-positive results and requires an interpretation rule set before you start. “Ever positive,” “positive at a landmark,” and “two consecutive positives” are different tests statistically.

Why a negative result cannot certify cure#

Sensitivity is constrained by how much tumor DNA reaches the sampled blood and how many informative molecules are actually collected. A small blood tube contains only a sample of circulation. If the target fraction is extremely low, no tumor molecule may be present even when residual cells exist.

Tumors differ in shedding. Disease confined to the brain, peritoneum, or other sites may release less detectable DNA into peripheral blood. Small volume, indolent biology, treatment effect, and low cell turnover can reduce the signal. A tumor may also evolve and lose a tracked alteration, although multi-variant panels reduce that problem.

An analytically negative result therefore means “not detected under this assay and specimen's conditions.” What it means for you depends on sensitivity, time point, tumor, stage, and baseline probability. Negative predictive value rises or falls with recurrence prevalence in the tested group.

Why a positive result needs specificity safeguards#

Sequencing errors can mimic very-low-frequency variants, so assays use molecular barcodes, replicate evidence, error models, and quality controls. Contamination, sample mix-up, and index misassignment require laboratory safeguards.

Clonal hematopoiesis is a major biological issue. Blood-forming stem cells can acquire mutations with age or after therapy, and their DNA enters plasma and can be mistaken for tumor DNA, particularly in genes commonly altered in both blood clones and solid cancers. Matched white-blood-cell sequencing or tumor-informed selection can help separate sources. An inherited variant can also appear in plasma if the assay is not designed to classify germline findings, so a positive test's specificity is tied to its algorithm and context, not just to the sequencing machine.

Analytical validity, clinical validity, and utility#

Analytical validity asks whether the assay accurately and reproducibly detects its target near the claimed limit; it includes precision, blank and detection limits, interfering substances, specimen stability, failure rate, and bioinformatics version control.

Clinical validity asks whether the result predicts a clinical state or outcome. Many ctDNA studies show that postoperative positivity is associated with recurrence-free and overall outcomes. A large hazard ratio can make ctDNA a strong prognostic classifier.

Clinical utility asks whether using the result to make a decision improves outcomes compared with a strategy that does not use it. This is the harder question, and it is the one you should be asking. A positive result may identify high risk, but an available additional therapy may not eradicate the residual clone. A negative result may identify low risk, but withholding therapy could still harm a small group with undetected disease. Those three claims should not be collapsed into “the test works.”

Prognostic is not automatically predictive#

A prognostic biomarker identifies people with different outcomes regardless of a particular treatment. A predictive biomarker identifies different relative benefit from a treatment. Positive ctDNA is clearly prognostic in several settings. Showing that positivity predicts added benefit from one therapy requires a treatment comparison with an interaction analysis or a randomized biomarker strategy.

Observational data may show ctDNA-positive patients who received chemotherapy did better than those who did not. Treatment selection, health status, tumor features, and timing can confound that comparison. A randomized strategy is stronger. And even when a marker is predictive for treatment benefit, the best regimen, duration, and tradeoff may remain uncertain, because biomarker enrichment in a drug trial answers a narrower question than routine management of every positive result.

What DYNAMIC showed in stage II colon cancer#

The randomized DYNAMIC trial assigned people with resected stage II colon cancer to a ctDNA-guided adjuvant strategy or conventional management. The ctDNA-guided group used chemotherapy less often while meeting the trial's noninferiority criterion for two-year recurrence-free survival.

That is genuine clinical-utility evidence for a strategy, not merely a prognostic association. The five-year report supported durability of the main conclusion. Still, the trial tested a particular cancer stage, assay, postoperative timing, algorithm, and health-system setting. It did not prove that any commercial MRD assay can guide any cancer decision.

The trial also did not prove zero risk after a negative result. Recurrences occurred in ctDNA-negative groups. The value was risk refinement and a tested treatment-selection strategy.

DYNAMIC-III shows why extension is difficult#

The later randomized DYNAMIC-III trial studied resected stage III colon cancer. ctDNA-guided management reduced oxaliplatin use and adverse events in the de-escalation pathway, but outcomes only approached the prespecified standard rather than creating a simple universal permission to reduce therapy. Escalating conventional chemotherapy in ctDNA-positive patients did not improve recurrence-free survival.

Persistent ctDNA after treatment marked very poor prognosis. That finding reinforces prognostic validity while demonstrating the utility gap: identifying people at extreme risk does not guarantee that more of the available chemotherapy changes their outcome. Novel strategies need direct testing. Read the stage II and stage III findings together and the lesson is the same: ctDNA can help allocate treatment in some settings, but the action attached to each result must earn its own evidence.

Lead time is not the same as added life#

ctDNA can become positive months before radiographic recurrence. Earlier knowledge may create time for confirmatory imaging, trial enrollment, or intervention. It can also create a longer period of anxiety without an effective earlier treatment.

Lead-time bias makes survival measured from molecular detection appear longer even if the date of death is unchanged. To prove benefit from earlier detection, a trial should compare a ctDNA-triggered surveillance or treatment strategy with standard follow-up and measure patient-centered outcomes. Starting therapy solely because a blood test turned positive can also expose people to toxicity while the disease remains unlocalized, and whether that tradeoff helps depends on tumor biology and therapy, not on detection speed alone.

ctDNA does not replace pathology or imaging#

Pathology establishes histology, grade, margins, lymph-node findings, and many tumor features. Imaging localizes anatomy, measures lesions, detects complications, and guides biopsy or surgery. ctDNA supplies a molecular signal that may precede imaging but usually cannot show where to operate or radiate.

Plasma genomic profiling for a treatment target is also a different use from MRD. In metastatic disease, a liquid biopsy may search for an actionable alteration. In a post-treatment MRD assay, the goal is ultra-sensitive detection of a small known or inferred tumor signal. One test design may not be validated for the other purpose. Discordance should be expected occasionally: positive ctDNA with negative imaging, negative ctDNA with a suspicious lesion, and different results across assays each require the relevant disease pathway rather than a universal tie-breaker.

Regulation and recommendations are context-specific#

FDA's final guidance addresses use of ctDNA as a biomarker in early-stage solid-tumor drug development. Guidance for trial sponsors is not blanket approval of every laboratory test for routine treatment decisions, nor does it declare ctDNA a validated surrogate for all cancer outcomes.

The ESMO Precision Medicine Working Group concluded in 2022 that molecular relapse detection had strong clinical validity in many cancers but insufficient evidence at that time for routine treatment direction. Since then, randomized colon-cancer evidence has advanced. The appropriate conclusion is not that all prior caution vanished; it is that utility is developing setting by setting. So check the current labels, the regulatory status, the specialty guidance, and the trial evidence for the exact tumor and the exact decision you are facing.

Sources#

1. FDA final guidance on ctDNA in early-stage solid-tumor drug development
2. FDA regulatory-science project on liquid biopsy and residual disease
3. Randomized DYNAMIC stage II colon-cancer trial
4. Five-year outcomes of the randomized DYNAMIC trial
5. Randomized DYNAMIC-III stage III colon-cancer trial
6. ESMO Precision Medicine Working Group ctDNA recommendations

Any test-guided cancer decision requires disease-specific specialist interpretation.*