Plasma from a patient without bulky disease contains only small quantities of cell-free DNA (cfDNA), nearly all of it released by blood cells. In localized cancers the tumor-derived share is far smaller than in metastatic disease and is undetectable in a proportion of patients (Bettegowda et al., 2014). After resection it can fall well below one mutant molecule in a thousand. At such levels two limits apply. One is sampling: a tube of blood holds a finite number of genome copies, and a mutation absent from the tube cannot be found by any chemistry. The other is error, since polymerase and sequencer mistakes occur at rates comparable to the signal sought (Wan et al., 2017).

A third difficulty is biological. With age, blood stem cells acquire somatic mutations and expand as clones, a process termed clonal hematopoiesis (Jaiswal et al., 2014). Deep sequencing of plasma has shown that a large share of the variants found in cfDNA originate from these clones and not from tumor (Razavi et al., 2019). Unless they are filtered, they produce false positive results.

Assays for residual disease are of two kinds, compared in Table 1. A tumor-informed assay begins by sequencing the resected tumor, selects a set of that patient's own somatic variants and builds a personalized panel to track them in plasma, an approach used in the studies of Reinert et al. (2019) and Tie et al. (2016). A tumor-agnostic, or plasma-only, assay applies one fixed panel to every patient and commonly adds methylation patterns characteristic of colorectal cancer to raise sensitivity (Parikh et al., 2021). Neither is superior on every count, and the choice depends on tissue availability, turnaround and laboratory capacity.

Prospective cohorts agree that postoperative ctDNA is a strong prognostic marker (Tie et al., 2016; Reinert et al., 2019; Henriksen et al., 2022). A large Japanese observational study added that patients with detectable ctDNA after surgery fared better with adjuvant chemotherapy than without it, while no such difference was apparent in those who tested negative (Kotani et al., 2023). The randomized DYNAMIC trial in stage II colon cancer showed that a ctDNA-guided strategy reduced the use of adjuvant chemotherapy without worsening recurrence-free survival (Tie et al., 2022). Guidance has nonetheless remained cautious. A joint review by two American professional societies found the evidence for clinical utility in early-stage disease insufficient at the time (Merker et al., 2018), and a European working group later judged that residual disease assays, although clinically valid, should not yet direct routine treatment (Pascual et al., 2022). The workflow below should be read in that light.

A pathologist selects a block from the resection specimen with adequate tumor cellularity, and DNA is extracted from it and from a normal sample, usually the buffy coat of a blood draw. Sequencing both allows germline variants and clonal hematopoiesis to be subtracted, leaving the somatic variants of the tumor. Variants present in most tumor cells are preferred as targets, because a subclonal variant may be absent from the cells that seed a recurrence. The record should state the block used, the estimated tumor fraction, the sequencing depth and the variants chosen.

If a tumor-informed approach is taken, a panel covering the selected variants is designed and its performance checked on the patient's own tumor and normal DNA. Tracking several variants at once raises the chance that at least one mutant fragment is present in a given tube. If a fixed panel is used, this step reduces to confirming that the panel is suitable for colorectal cancer and that a method for excluding clonal hematopoiesis is in place. In either case the assay and its version are fixed before the first plasma sample is tested.

Errors made before the sample reaches the laboratory are among the most frequent and the least recoverable. Plasma is used in preference to serum, because clotting releases DNA from white cells and dilutes the tumor fraction. Blood drawn into EDTA tubes must be separated within a few hours; tubes containing a cell-stabilizing preservative allow longer transport at ambient temperature. Two centrifugation steps remove cells and debris, and plasma is stored frozen at −80 °C until extraction (Merker et al., 2018). The time of draw, tube type, time to processing and plasma volume are recorded for every sample.

Timing relative to treatment matters as much as handling. Surgery injures tissue and releases normal cfDNA for some weeks, which can mask a small tumor signal, and chemotherapy may suppress shedding without eliminating disease. The first postoperative sample is therefore best taken about a month after resection and before adjuvant treatment begins, as in the DYNAMIC trial (Tie et al., 2022), with the schedule agreed in advance (Dasari et al., 2020).

cfDNA is extracted with a method validated for short fragments, and the yield is measured. The quantity matters for interpretation and is more than a quality check. The number of genome copies entered into the assay sets the lowest tumor fraction that could have been detected in that sample, and a negative result from a low-input sample is weaker than one from a generous input. A fragment size profile helps to reveal contamination by high-molecular-weight DNA from lysed white cells.

Detection at very low allele fractions depends on separating true mutations from artifacts. The standard solution is to tag each original DNA molecule with a unique molecular identifier before amplification, so that reads descending from one molecule can be grouped and a consensus taken. Errors that appear in only some reads of a family are then discarded (Kinde et al., 2011). Hybrid-capture methods combine such barcoding with statistical models of background error at each position (Newman et al., 2014). Whatever the chemistry, every run should include a negative control and a positive control at a known low allele fraction.

The criteria for a positive sample are written down before clinical testing starts and are not adjusted case by case. For a tumor-informed assay a typical rule requires that a minimum number of tracked variants be detected above the background established in healthy donors. For a fixed panel, any variant also found in the patient's white cells is removed as clonal hematopoiesis (Razavi et al., 2019). The output is categorical, detected or not detected, accompanied by a quantitative estimate and by the input amount from step 4.

The report states the assay type, the result, the quantity of cfDNA analyzed and a plain statement of what a negative result does and does not exclude. Results are discussed alongside stage, pathology, carcinoembryonic antigen and imaging. A single negative sample soon after surgery leaves a real chance of later relapse, and serial testing during surveillance raises sensitivity for recurrence (Reinert et al., 2019). The interval between draws and the action to follow a positive result should be set by the team beforehand, since a result with no agreed consequence causes anxiety without benefit.