Why MRD flags recurrence but can't guide treatment

A blood-based MRD test can often tell you cancer is coming back months before a scan can. Ask it the next question — what should we treat it with? — and it usually goes quiet. That gap isn't a bug in any one product; it falls out of how much tumor signal is physically in the tube.

What MRD actually measures

Measurable (or minimal) residual disease is the small amount of cancer left after treatment — below what imaging can see, detectable only by sensitive molecular methods. In solid tumors, the dominant approach reads circulating tumor DNA (ctDNA): fragments shed into the blood by tumor cells.

As a recurrence alarm, it works. After surgery, ctDNA positivity carries roughly a ten-fold higher risk of recurrence in colorectal cancer (Kotani et al., 2023), and it often turns positive months before a scan would — a median of ~5 months in localized lung cancer (Chaudhuri et al., 2017) and a mean of ~8.7 months in colorectal (Reinert et al., 2019). But strong prognosis is not the same as a treatment plan. The ESMO Precision Medicine Working Group put the ceiling plainly: molecular-residual-disease detection "cannot be recommended in routine clinical practice, as currently there is no evidence for clinical utility in directing treatment" (Pascual et al., 2022). It tells you whether, not what.

The tumor-fraction wall

Here's the physical constraint. To detect that disease is present, an assay needs to distinguish a faint tumor signal from background. To genotype — to read out a druggable alteration you can act on — it needs enough tumor molecules to characterize them confidently. Those are very different bars, and tumor burden decides which one you can clear.

• Advanced / metastatic disease — ctDNA is relatively abundant (tumour fraction commonly around 1% or more): plenty of signal to detect and genotype. This is where liquid-biopsy therapy selection works today.
• Localized / early disease — tumour fraction is often below 0.1%, and lower still in the earliest cancers; ctDNA levels fall by orders of magnitude as stage and burden drop (Bettegowda et al., 2014).
• MRD / remission — after treatment, tumour fraction typically sits far below 0.1% (roughly 0.001–0.05%): enough to detect presence, too dilute to genotype for therapy.

The gap is measurable. Tumour-agnostic assays generally need a variant fraction around 0.1% or higher to call a specific mutation (Chan et al., 2023; Wan et al., 2017), while methylation-based detection can register that disease is present far below that (Parikh et al., 2021). So a test can confirm cancer is back long before it has the signal to say what to treat — and by the time the burden has grown enough to genotype, the early-warning lead time has largely been spent.

The sensitive assays keep detecting — but give up the target

Below the mutation-genotyping floor, detection doesn't stop; it changes register. Three families of assays push the detection floor down toward ~0.001% tumor fraction:

• Methylation — reads tens of thousands of methylation regions rather than a handful of mutations, so it holds up at low signal-to-noise. But it flags presence, not druggable targets.
• Tumor-informed (tissue-informed) assays — build a bespoke panel from the patient's tumor tissue, reaching ~0.01% VAF (the deepest to a few parts-per-million). They need tissue, and at MRD burden they still report presence, not a therapy.
• Fragmentomics — reads the physical fragmentation pattern of cell-free DNA, explicitly independent of specific mutations or methylation marks. Again: a presence signal.

And there's a hard trade-off underneath all of them: the few tumor fragments you recover at MRD burden can be spent on depth (confirming presence) or breadth (surveying many genes), not both. More sensitivity does not convert into more actionability. So the transition as burden falls is not detectable → undetectable. It's "genotype-able" → "detectable, but only as presence" — the amber band the timeline drops into during remission, and climbs back through at recurrence.

Why this is structural, not an engineering gap

You can't sequence molecules that aren't in the tube. At MRD burden the tumor DNA is diluted a thousand-fold or more by normal cell-free DNA, so broad genotyping simply runs out of material. Methylation and fragmentomics restore detection by asking easier, aggregate questions — not by recovering a druggable variant. And none of these read protein at all: proteins are not encoded in cell-free DNA, so targets like HER2 or PD-L1 are invisible to any plasma-DNA method, at any burden. Guidelines note the same for structural variants — ctDNA is less reliable for gene fusions and copy-number changes, and tissue testing remains preferred for many patients (Pascual et al., 2022).

A different starting point: whole cells

This is the problem LMNTIC works on, and it starts by changing the analyte. A circulating tumor cell reframes the question: tumor content becomes discrete cells rather than a fraction of a molecular soup. Capture even a few, and each one carries a complete genome and proteome — so in principle a positive can be genotyped and phenotyped for predictive markers regardless of the overall burden. The constraint moves from "is there enough tumor fraction to profile?" to "did we capture an informative cell?"

We're honest that this trades one hard problem for another: capturing enough cells at minimal-residual burden is an open challenge, and it's exactly what our work targets. The point isn't that cells replace ctDNA — the two are complementary — but that a cell-based readout is where an actionable answer at low burden could come from. That whole cells carry information plasma can't is already documented: the AR-V7 variant read from CTCs predicts resistance to enzalutamide and abiraterone in prostate cancer (Antonarakis et al., 2014), and ESR1 mutations turn up more often in CTC-derived DNA than in matched plasma (Smilkou et al., 2025). LMNTIC's L:Biopsy isolates circulating tumor cells from a whole-blood draw and is built to yield DNA, RNA, and protein from the same cells.

Get in touch

If you're working on MRD, actionability at low tumor burden, or CTC-based approaches, we'd like to compare notes — reach us at lmntic.com.

References

The scientific claims here are grounded in the sources below; the interactive figure's specific values are illustrative, as noted. Peer-reviewed papers link to their DOI; assay specifications link to the manufacturer.

1. Kotani D, et al. Molecular residual disease and efficacy of adjuvant chemotherapy in patients with colorectal cancer (CIRCULATE-Japan GALAXY). Nat Med. 2023;29:127–134. doi:10.1038/s41591-022-02115-4
2. Chaudhuri AA, et al. Early detection of molecular residual disease in localized lung cancer by circulating tumor DNA profiling. Cancer Discov. 2017;7:1394–1403. doi:10.1158/2159-8290.CD-17-0716
3. Reinert T, et al. Analysis of plasma cell-free DNA by ultradeep sequencing in patients with stages I–III colorectal cancer. JAMA Oncol. 2019;5:1124–1131. doi:10.1001/jamaoncol.2019.0528
4. Pascual J, et al. ESMO recommendations on the use of circulating tumour DNA assays for patients with cancer. Ann Oncol. 2022;33:750–768. doi:10.1016/j.annonc.2022.05.520
5. Bettegowda C, et al. Detection of circulating tumor DNA in early- and late-stage human malignancies. Sci Transl Med. 2014;6:224ra24. doi:10.1126/scitranslmed.3007094
6. Wan JCM, et al. Liquid biopsies come of age: towards implementation of circulating tumour DNA. Nat Rev Cancer. 2017;17:223–238. doi:10.1038/nrc.2017.7
7. Chan HT, et al. Tumor-informed or tumor-agnostic circulating tumor DNA as a biomarker for risk of recurrence in resected colorectal cancer patients. Front Oncol. 2023;12:1055968. doi:10.3389/fonc.2022.1055968
8. Parikh AR, et al. Minimal residual disease detection using a plasma-only circulating tumor DNA assay in patients with colorectal cancer (Guardant Reveal). Clin Cancer Res. 2021;27:5586–5594. doi:10.1158/1078-0432.CCR-21-0410
9. Coombes RC, et al. Personalized detection of circulating tumor DNA antedates breast cancer metastatic recurrence. Clin Cancer Res. 2019;25:4255–4263. doi:10.1158/1078-0432.CCR-18-3663
10. Cristiano S, et al. Genome-wide cell-free DNA fragmentation in patients with cancer (DELFI). Nature. 2019;570:385–389. doi:10.1038/s41586-019-1272-6
11. Liu MC, et al. Sensitive and specific multi-cancer detection and localization using methylation signatures in cell-free DNA. Ann Oncol. 2020;31:745–759. doi:10.1016/j.annonc.2020.02.011
12. Klein EA, et al. Clinical validation of a targeted methylation-based multi-cancer early detection test. Ann Oncol. 2021;32:1167–1177. doi:10.1016/j.annonc.2021.05.806
13. Antonarakis ES, et al. AR-V7 and resistance to enzalutamide and abiraterone in prostate cancer. N Engl J Med. 2014;371:1028–1038. doi:10.1056/NEJMoa1315815
14. Smilkou S, et al. Detection rate for ESR1 mutations is higher in circulating-tumor-cell-derived genomic DNA than in paired plasma cell-free DNA. Mol Oncol. 2025;19:2109–2119. doi:10.1002/1878-0261.13787
15. Signatera (Natera). Assay specifications — tumor-informed design, limit of detection ~0.01% VAF. natera.com — Signatera FAQ