Circulating tumor DNA, or ctDNA, is tumor-derived DNA that floats freely in the bloodstream after cancer cells shed fragments of their genetic material. A simple blood draw can capture these fragments and reveal mutations driving a patient’s cancer, track whether treatment is working, and even detect residual disease after surgery. The technology behind ctDNA testing has advanced rapidly over the past decade, and it is reshaping how oncologists make decisions. But the science comes with real limitations that matter if you or someone you know is navigating cancer care.
How Tumor DNA Ends Up in Your Blood
Every cell in your body releases bits of DNA into the bloodstream as it dies and gets broken down. This broad category of floating genetic material is called cell-free DNA, or cfDNA. Most of it comes from normal, healthy cells. But when a tumor is present, some of that cfDNA carries the specific mutations found in the cancer. That tumor-specific fraction is ctDNA, and it can be released both passively, when cancer cells die, and actively, through secretion by living tumor cells.
One of ctDNA’s defining features is how quickly the body clears it. Its half-life ranges from about 16 minutes to two and a half hours, which means the ctDNA circulating in your blood right now reflects what the tumor is doing right now, not what it was doing last month. That rapid turnover is both a strength and a constraint: it gives clinicians a near-real-time snapshot of tumor biology, but it also means the window for capturing those fragments after a blood draw is narrow, and sample handling matters enormously.
Finding the Needle in the Haystack
The central technical challenge is that ctDNA often makes up a tiny fraction of total cfDNA, sometimes less than one percent. Detecting a handful of mutant molecules against a massive background of normal DNA requires extremely sensitive technology.
Modern approaches fall into two broad camps. PCR-based methods, including digital PCR, are highly sensitive at finding specific known mutations. They work well when you already know what you’re looking for, such as tracking a particular mutation identified in a patient’s tumor tissue. Sequencing-based methods, particularly next-generation sequencing (NGS), cast a wider net and can scan hundreds of genes simultaneously, picking up mutations you might not have anticipated.
The sensitivity of NGS has been pushed further by techniques that tag individual DNA molecules with unique identifiers before amplification. These molecular barcodes let researchers trace every copy back to its original molecule, separating true mutations from errors introduced during the copying process. Duplex sequencing takes this a step further by requiring that both strands of a DNA molecule show the same change before calling it real. Combined with machine learning tools for filtering out background noise, these approaches can now detect mutations present at extraordinarily low frequencies, on the order of one mutant molecule per 100,000 normal ones.
Guiding Treatment Decisions
The most established clinical use of ctDNA today is matching patients with advanced cancer to targeted therapies. When a tumor biopsy is difficult, risky, or yields insufficient tissue, a blood-based test can identify actionable mutations that determine which drugs are most likely to work. This is particularly valuable in cancers like non-small cell lung cancer, where specific genetic alterations dictate whether a patient should receive a targeted therapy or standard chemotherapy.
The concordance between what a blood test finds and what a tissue biopsy finds varies depending on the stage of disease. In advanced cancers, the agreement is considerably better than in early-stage tumors. One study examining tumor stages found concordance rates of about 77 percent in stage IV cancers but essentially zero in stage I disease, reflecting the simple reality that early-stage tumors shed far less DNA into the bloodstream. In biliary tract cancer, ctDNA showed a sensitivity of about 85 percent compared to tissue profiling, and in some cases revealed targetable genetic changes that the tissue panel had missed entirely.
Data from the ROME trial, which enrolled patients with various advanced cancers, found that the overall concordance between tissue and liquid biopsies for clinically actionable mutations was about 49 percent. That number sounds discouraging until you look more closely: in roughly 16 percent of cases, the liquid biopsy found actionable changes that the tissue biopsy missed entirely. Patients whose mutations were confirmed by both methods and who received matched targeted therapy had improved survival compared to those on standard treatment, with median progression-free survival of about 5 months versus 3 months.
The takeaway for patients is that liquid and tissue biopsies are complementary rather than interchangeable. Each catches things the other misses, and using both together gives the fullest picture of what’s driving a cancer.
Detecting Leftover Disease After Surgery
Perhaps the most promising application of ctDNA is its ability to detect minimal residual disease, the invisible traces of cancer that can linger after surgery and eventually cause a relapse. Standard imaging cannot see microscopic clusters of cancer cells, and traditional blood markers are often too imprecise to be useful. ctDNA fills that gap.
In colorectal cancer, this has been studied extensively. A landmark study of 230 patients with stage II colon cancer found that among those not given chemotherapy after surgery, ctDNA was detected postoperatively in about 8 percent of patients. Of those ctDNA-positive patients, 79 percent went on to relapse. Among those who tested negative, fewer than 10 percent relapsed. The hazard ratio was 18, meaning ctDNA-positive patients had roughly 18 times the risk of recurrence compared to those who tested negative. A later meta-analysis pooling multiple studies in colorectal cancer confirmed a strong prognostic signal, with positive ctDNA after surgery associated with about a sevenfold increase in the risk of poor outcomes.
Similar findings have emerged in lung cancer. In a study of patients with early-stage and locally advanced non-small cell lung cancer, all four patients who remained ctDNA-positive one to two weeks after surgery eventually experienced disease progression, while only a third of ctDNA-negative patients relapsed during follow-up.
These results are driving serious clinical interest in using ctDNA to decide which patients truly need chemotherapy after surgery and which can safely skip it. If your ctDNA clears after an operation, the odds of recurrence are low, and the toxicity of additional treatment may not be worth it. If ctDNA persists, aggressive follow-up or adjuvant therapy becomes much more justified. Several ongoing trials are testing exactly this strategy.
Monitoring Treatment Response in Real Time
Because ctDNA turns over so quickly, serial blood draws during treatment can reveal whether a therapy is working well before imaging shows any visible change. In a study tracking ctDNA levels over time in patients with advanced cancers, rising ctDNA quantities predicted clinical or radiologic progression in about 73 percent of patients, with a median lead time of 23 days before the progression became apparent on scans. That three-week head start could allow clinicians to switch to a more effective therapy earlier rather than waiting for a tumor to grow large enough to show up on a CT scan.
Early ctDNA kinetics, meaning changes in ctDNA levels within the first days or weeks of starting a new treatment, are also being explored as rapid readouts of whether a drug is hitting its target. A steep early drop in ctDNA suggests the treatment is effective; a flat or rising trajectory suggests it is not. This kind of dynamic biomarker could eventually reduce the time patients spend on ineffective regimens.
The Problem of Clonal Hematopoiesis
One of the trickiest confounders in ctDNA testing is a phenomenon called clonal hematopoiesis of indeterminate potential, or CHIP. As people age, blood-forming stem cells naturally accumulate mutations. These mutations have nothing to do with a solid tumor, but they show up in cell-free DNA because blood cells are constantly dying and releasing their DNA into circulation. If a ctDNA test picks up one of these CHIP-associated mutations and mistakes it for a tumor-derived signal, the result is a false positive, which could lead to unnecessary imaging, biopsies, or even inappropriate treatment choices.
CHIP becomes more common with age and is found in a significant fraction of older adults, exactly the population most likely to be tested for cancer. Researchers have developed computational methods to distinguish CHIP mutations from true tumor mutations using patterns in the sequencing data itself. One approach applied to over 4,300 clinical samples demonstrated that it is possible to classify clonal hematopoiesis status directly from cell-free DNA sequencing, without needing a separate blood cell sample for comparison. Still, CHIP remains a meaningful source of noise, particularly in screening settings where the population being tested includes many people without cancer.
Early Cancer Screening in Healthy People
Using ctDNA to screen apparently healthy people for hidden cancers is the most ambitious and most uncertain frontier. The idea is seductive: a routine blood test that catches cancer before symptoms appear, potentially at a stage when treatment is most effective. Several multi-cancer early detection tests are in various stages of development and validation.
A prospective study of over 9,000 asymptomatic adults tested a ctDNA-based assay designed to detect multiple cancer types simultaneously. Of those who tested positive, about 40 percent were confirmed to have cancer, yielding a positive predictive value of roughly 40 percent. The test’s overall sensitivity was about 71 percent, meaning it caught roughly seven in ten cancers present in the cohort. Its specificity was high at about 99.7 percent, and among those who tested negative, over 99.9 percent were confirmed cancer-free over the following 12 months. However, the test’s ability to accurately predict which organ the cancer originated in was only about 53 percent.
A separate study evaluating a different multi-cancer detection test reported higher accuracy numbers in a controlled cohort, with sensitivity above 90 percent and specificity above 98 percent. But that study included known cancer patients alongside healthy controls, which tends to inflate performance compared to real-world screening of low-risk populations.
The gap between these results and the kind of reliability needed for population-wide screening is significant. A specificity of 99.7 percent sounds excellent until you consider that in a population of 100,000 screened people where perhaps 500 have undiagnosed cancer, roughly 300 healthy people would still receive false positive results. Each false positive triggers follow-up imaging, potential biopsies, anxiety, and medical costs. Whether the cancers caught early by screening lead to better outcomes than cancers found through conventional means is the question clinical trials are still working to answer.
When Blood Is Not Enough
For most solid tumors, a standard blood draw provides enough ctDNA to be clinically useful. Brain tumors are a major exception. The blood-brain barrier sharply limits how much tumor DNA crosses into the general circulation, meaning plasma ctDNA levels in brain cancer patients are often too low for reliable detection.
Cerebrospinal fluid, or CSF, offers a workaround. Because CSF bathes the brain and spinal cord directly, tumor DNA concentrations in CSF are substantially higher than in plasma for central nervous system cancers. Collecting CSF requires a lumbar puncture, which is more invasive than a blood draw, but it opens a window into brain tumor biology that blood tests largely cannot.
Beyond CSF, researchers have explored ctDNA in other body fluids. Urine contains cell-free DNA and can be collected entirely non-invasively, making it attractive for cancers of the urinary tract and potentially others. Pleural fluid from around the lungs and peritoneal fluid from the abdomen are also being studied as sources of ctDNA for cancers in those body cavities. Each alternative fluid has tradeoffs in terms of invasiveness, tumor DNA concentration, and how well-established the analytical methods are.
Why Sample Handling Matters More Than You Think
The analytical sensitivity of a ctDNA test is only as good as the sample it receives. If white blood cells in a blood tube burst open before the plasma is separated, they release a flood of normal genomic DNA that dilutes the tumor signal, sometimes to the point of being undetectable. This contamination, called genomic DNA leakage, is the single biggest pre-analytical threat to ctDNA testing.
Standard blood collection tubes without preservatives need to be processed quickly, ideally within a few hours. Specialized tubes containing preservative agents can stabilize cell-free DNA for longer, with manufacturers claiming stability up to 14 days, though independent testing has confirmed reliable preservation for up to about seven days. Even with preservatives, a recommended storage time of less than two days at room temperature is generally advised to avoid hemolysis and evaporation effects. Temperature extremes during shipping, delays at the lab, and even the choice of centrifuge speed during plasma separation can all influence whether a test produces an accurate result.
For patients, this means the logistics of getting your blood sample to the lab matter. If your oncologist orders a ctDNA test, the blood draw should go into the right type of tube and be handled according to the assay manufacturer’s instructions. It is a detail worth asking about, especially if you are at a facility that does not routinely process liquid biopsy samples.
Machine Learning and the Future of ctDNA Analysis
Raw sequencing data from a ctDNA test is enormously complex, and traditional statistical methods struggle to extract all the useful information buried within it. Machine learning algorithms are increasingly being applied to improve every stage of the process, from distinguishing true tumor fragments from background noise to identifying the tissue of origin when cancer is detected.
One promising area involves analyzing not just the mutations in ctDNA but the fragmentation patterns of the DNA itself. The way DNA wraps around proteins called nucleosomes differs between cancer cells and healthy cells, and those differences leave characteristic signatures in cell-free DNA fragment lengths and positions. Computational analysis of these patterns can reveal which tissue the DNA came from, even when the mutational signal alone is too faint to interpret.
Methylation-based approaches take a different angle. Chemical modifications to DNA vary by tissue type and disease state, and a targeted methylation sequencing panel has demonstrated the ability to correctly identify the tissue of origin in about 89 percent of cancer samples across 20 different cancer types. Combining mutation analysis, fragmentation patterns, and methylation signatures through machine learning creates a richer, more accurate picture than any single data type can provide alone.
The integration of artificial intelligence into ctDNA analysis is not a distant future; it is already embedded in several commercial assays. As these tools mature and are validated in larger patient populations, the sensitivity and specificity of ctDNA testing should continue to improve, gradually expanding its role from advanced cancer management into earlier disease stages and, potentially, routine screening.
Coverage and Access
Despite the clinical promise, coverage of ctDNA testing by insurance and health systems remains uneven. Regulatory approvals for specific liquid biopsy assays exist in several countries, particularly for matching advanced cancer patients to targeted therapies. But many emerging applications, including minimal residual disease detection and multi-cancer screening, are still considered investigational by most payers. Real-world economic analyses have begun to show that comprehensive genomic profiling through liquid biopsy can benefit payers not only through better treatment matching but also by facilitating clinical trial enrollment, which offsets drug costs. Still, the evidence base for clinical utility beyond a handful of well-studied tumor types remains an area where carefully designed studies are needed to establish broader adoption.