When Does ctDNA Actually Add Clinical Value?

Circulating tumour DNA has attracted a great deal of interest, much of it warranted. It is also frequently discussed in terms broad enough to obscure where it actually helps.

A more useful approach is to ask what question is being asked, and then whether ctDNA can answer that particular question better than the alternatives.

Where it is most clearly useful

When tissue is insufficient. A small biopsy may not yield enough material for comprehensive profiling. Plasma can be sampled without an additional invasive procedure. This is often the most immediately practical application.

When re-biopsy is difficult and the molecular profile may have changed. At progression, the relevant question is frequently what has changed since baseline. Where the disease site is hard to biopsy, plasma may be the realistic route to that information.

When capturing heterogeneity matters. A single-site biopsy samples one location. Plasma receives contributions from multiple sites, which may better represent a heterogeneous disease — though it also cannot say which site any given finding came from.

Serial monitoring. Repeated sampling permits trajectory rather than snapshot. Change over time is often more informative than any single measurement.

Where interpretation requires the most care

A negative result. Shedding varies enormously between tumour types, disease burdens and anatomical sites. Some tumours shed very little. A negative plasma result in a low-shedding tumour is weak evidence of absence, and in most settings prompts tissue testing rather than reassurance.

Clonal haematopoiesis. Variants in blood cell precursors — particularly in genes such as DNMT3A, TET2 and ASXL1 — can be detected in plasma and are not tumour-derived. Distinguishing them requires either awareness of the characteristic genes and allele fractions, or paired sequencing of white cells.

Low allele fractions near the limit of detection. The confidence in a call at very low fraction differs substantially from one at high fraction, and reports do not always foreground this.

Where the evidence is still developing

Molecular residual disease detection after curative-intent treatment is an area of substantial activity, with the strongest data so far in colorectal cancer. The prognostic association is well described. Whether acting on a ctDNA result improves outcomes — rather than identifying earlier who will do badly — is the question trials are addressing, and the distinction between prognostic and predictive matters here.

Screening asymptomatic populations is a different problem again, with the usual screening considerations of specificity, over-diagnosis and downstream investigation burden applying with full force.

What it does not replace

ctDNA does not replace imaging. It gives no anatomical information — it cannot say where disease is, how large it is, or whether it threatens a structure.

It does not replace tissue pathology. Histological diagnosis, grade, architecture and the assessment of transformation all require tissue. Histological transformation in lung cancer is a clear example: a plasma assay may show no new alteration while the tumour has changed its histology entirely.

These are complementary modalities answering different questions, and framing ctDNA as a replacement for either sets it up to disappoint.

A practical framing

Before ordering, it is worth being explicit: what would a positive result change? What would a negative result change? If neither answer is anything, the test is producing a number rather than information.

Where the answers are substantive, ctDNA can be genuinely valuable. The value comes from asking a specific question, not from the modality itself.