Global Liquid Biopsy Market Size, Share, Trends and Forecast 2026-2033

Liquid biopsy in 2025: a category that paid for its own growth

Two disclosures, read side by side, describe this market more honestly than any forecast can. Guardant Health grew oncology revenue 26 per cent in 2025 on 34 per cent more tests, and took non-GAAP gross margin from 62 to 66 per cent. Natera processed 800,800 oncology tests, up 51.6 per cent, and lifted full-year gross margin from 60.3 to 64.7 per cent. At the largest participants the category is expanding and its unit economics are improving.

The other disclosure is GRAIL. The company reported USD 147.2 million of revenue for 2025, up 17 per cent, against a net loss of USD 408.4 million, a gross loss of USD 62.6 million and negative adjusted EBITDA of USD 320.6 million. Galleri is the most recognised product in early cancer detection, and in 2025 it turned USD 147.2 million of revenue into a USD 408.4 million loss.

Both statements are true at once, and the tension between them is this report's subject. It resolves into a single question: what turns a test into a protocol, because that is the only thing in this market that has ever produced a margin.

What is inside the perimeter, and what is not

Liquid biopsy is a method, not a market. Pheonix Research defines the market here as end-market spend on molecular testing of plasma or other body fluids for cancer indications, across three segments: therapy selection (comprehensive genomic profiling from circulating tumour DNA, used to choose a drug), molecular residual disease and recurrence monitoring (serial testing after treatment), and blood-based cancer screening and multi-cancer early detection.

Four things are deliberately outside it. Tissue-based profiling, which is a separate and larger market. Germline hereditary testing, which answers a different clinical question. Non-invasive prenatal testing, which is cell-free DNA but is reproductive rather than oncology. And biopharma research services and data licensing, which are real revenue for several companies in this report but are sold to drug developers rather than against a diagnostic indication.

The perimeter matters more here than in most markets, because the participants' reported revenue is mostly not liquid biopsy. The four largest listed companies in this space reported USD 4,707.1 million of total revenue in 2025. Of that, USD 1,814.6 million sits inside this report's perimeter. The balance is women's health, organ health, tissue profiling, data and biopharma services — businesses that are adjacent to liquid biopsy, growth-linked to it in places, and not the same thing.

A report that treated those headline numbers as the size of the market would put women's health, organ health, tissue profiling, data and biopharma services inside liquid biopsy, and would then forecast them all at the growth rate of cancer diagnostics. That is not a perimeter error. It is a different market, with a different buyer and a different growth rate.

The market: USD 3,757.6 million in 2025 to USD 12,492.2 million in 2033

Pheonix Research sizes the global liquid biopsy market at USD 3,757.6 million in 2025, rising to USD 12,492.2 million in 2033. That is 3.32 times the base and a compound annual growth rate of 16.20 per cent, with USD 8,734.6 million of incremental value created across the window.

The base is built segment-first, because the three segments have different payers, different evidence burdens and different ceilings, and any one of them forecast at the blended rate would be wrong in an interesting way.

Weighted growth in the first forecast year is 14.51 per cent; the compounded rate across the window is 16.20 per cent. The 169 basis point gap is arithmetic rather than optimism. Molecular residual disease moves from 34.6 per cent of the market to 44.7 per cent and screening from 7.4 to 20.5 per cent, so the faster segments compound upward in weight. A reader who checks the endpoints will find they agree. A reader who averages the three segment rates will get 20.3 per cent, which is not the answer, because therapy selection is more than half the base.

Therapy selection is the mature business, and it grows at 9 per cent

Plasma CGP is where liquid biopsy started and it is now largely a replacement market. Guardant's oncology revenue grew 26 per cent, but that growth came from 34 per cent more tests against a 5.97 per cent decline in realised revenue per test, from roughly USD 2,634 to USD 2,477. Foundation Medicine's liquid offering and Tempus's liquid panels compete for the same ordering physicians. In China the same clinical question is sold at a fraction of the US price, and Burning Rock's central laboratory revenue fell 8.9 per cent to RMB 160.0 million as volume shifted into hospitals at lower realised price.

Nine per cent sits above the incumbent giants' disclosed core growth and below the pure-plays' reported growth. The difference is mix: guideline expansion into new tumour types, and a rising share of therapy selection performed on blood rather than tissue, partly offset by price.

Molecular residual disease is where the margin is

MRD is the only segment in this market where a test is ordered repeatedly by protocol rather than once at a physician's discretion. That single structural difference explains most of the rest. Natera's gross margin rose as oncology volume grew 51.6 per cent, because a serial test amortises its own fixed cost. Adaptive Biotechnologies delivered 105,587 clonoSEQ tests, up 39 per cent, recognised USD 212.3 million of MRD revenue, up 46 per cent, and reported that its MRD business reached positive adjusted EBITDA and positive cash flow.

Twenty per cent for the segment is below both of those observed rates, deliberately. It is the rate that survives the coverage constraint set out in the regulatory and risk chapters, not the rate implied by the fastest adopter.

Screening is 7.4 per cent of the revenue and most of the narrative

Blood-based cancer screening attracts almost all of the public attention given to liquid biopsy and a small fraction of its revenue. GRAIL reported USD 136.8 million of US Galleri revenue for the year, up 26 per cent. Guardant reported USD 79.7 million of screening revenue from approximately 87,000 Shield tests. Between them that is USD 216.5 million of product revenue inside a category of USD 3,757.6 million.

The forecast has screening reaching 20.5 per cent of the market by 2033 at 32 per cent compound growth. That is the most aggressive number in this report, and it is concentrated in the outer years, where screening volume depends less on assay performance than on whether a national coverage determination admits the test and whether the follow-on diagnostic pathway can absorb the positives. Both are examined in the regulatory and risk chapters.

Only two of the eight pools are disclosed without judgement

Guardant's USD 763.3 million is disclosed (USD 683.6 million oncology plus USD 79.7 million screening). Adaptive's USD 212.3 million is disclosed MRD revenue. GRAIL's USD 147.2 million and Burning Rock's RMB 539.6 million are disclosed in total but required attribution to this perimeter. The remainder are estimates built from disclosed test volumes, segment revenue and stated splits.

The residual — unlisted and regional laboratories, in-hospital MRD programmes and hospital laboratory-developed tests — is USD 1,050.0 million, or 27.94 per cent of the base. That is a larger disclosure gap than any other molecular diagnostics market Pheonix Research covers, and the reason is structural: most tumour profiling and most recurrence monitoring happen in hospital laboratories attached to health systems, and those laboratories publish revenue to nobody.

No participant discloses global volume

Where a company discloses test counts they are company-specific — Guardant's 276,000 oncology tests, Natera's 800,800, Adaptive's 105,587, Guardant's 87,000 Shield tests — and they cannot be aggregated without assuming one average selling price across three different clinical products. Pheonix Research therefore reports value only and leaves the volume fields empty rather than deriving a unit count from a blended price that does not exist.

Table of Contents

• Liquid biopsy in 2025: a category that paid for its own growth
• Competitive landscape: two businesses wearing one label
• Value chain: where the margin sits and who collects it
• Investment activity: capital as the input, not the output
• Technology and innovation: from tissue substitute to protocol component
• Market risk: the criteria are tighter than the label

Competitive Landscape

Competitive landscape: two businesses wearing one label

Every participant in this market sells a blood test for cancer, and they are not in the same business. One group runs a centralised laboratory and sells its own test to a physician at a price a payer sets. The other sells a platform, a kit or a companion diagnostic claim into somebody else’s laboratory, and earns a royalty on a menu. The first group carries the fixed cost of a commercial organisation and a bioinformatics stack. The second carries the fixed cost of validation and regulatory clearance, then scales at software-like incremental cost.

That split explains most of the apparent contradictions in the reported numbers, and it is why a single market-share table has to be read with the business model attached.

The 2025 base, by participant pool

The first thing the table says is that this is a two-company market at the top and an extremely long tail underneath. Guardant and Natera are each about a fifth of it. Neither has ever reported an annual profit. The single largest line, at 27.94 per cent, is not a company at all — it is the hospital laboratories nobody measures.

Guardant: the scale of a category leader and the cost of being one

Guardant reported USD 982.0 million of total revenue in 2025, up 33 per cent. Oncology revenue was USD 683.6 million, up 26 per cent, on approximately 276,000 oncology tests, up 34 per cent. Screening contributed USD 79.7 million from approximately 87,000 Shield tests in its first full year of commercial scale. Biopharma and data revenue was USD 210.1 million, up 18 per cent, and sits outside this report’s perimeter.

Non-GAAP gross margin reached 66 per cent against 62 per cent in 2024, and free cash flow burn improved to USD 233 million from USD 275 million. Improving margin, improving burn, and still a loss. The company also received the first Guardant360 CDx companion diagnostic approval in colorectal cancer, launched in-house testing in Italy at Policlinico Gemelli on Guardant360 CDx technology, expanded tissue-free Guardant Reveal into late-stage therapy response monitoring, and submitted a Reveal chemotherapy monitoring data package to MolDx for Medicare reimbursement.

That last item is the strategic tell. Reveal’s next increment of revenue does not depend on an assay improvement or a sales hire. It depends on a coverage determination.

Natera: the annuity that works

Natera reported USD 2,306.1 million of revenue in 2025, up 35.9 per cent, with product revenue of USD 2,295.8 million. It processed approximately 3,525,500 tests, including 800,800 oncology tests, up 51.6 per cent, and lifted gross margin from 60.3 to 64.7 per cent. Net loss was USD 208.2 million, and — uniquely among the pure-plays in this report — it generated positive cash inflow of approximately USD 107.6 million for the year.

Most of Natera’s revenue is women’s health and organ health and is outside this perimeter. The oncology line is not disclosed as revenue, which is why it appears here as an estimate: 800,800 tests at an assumed USD 950 realised. That single assumption is the largest judgement in this report’s competitive table, and a reader who prefers a different realised price should move the whole base by the difference rather than adjust nothing.

Roche and Foundation Medicine: the toll collector

Roche reported Diagnostics revenue of CHF 13,847 million in 2025, up 2 per cent at constant exchange rates, with Molecular Lab at CHF 2,527 million, up 4 per cent CER, and Pathology up 14 per cent. Roche does not disaggregate a liquid biopsy line, so this report attributes 12 per cent of Molecular Lab to plasma testing, converted at the 2025 annual average of 1.2062.

Roche’s structural position is different from the pure-plays in a way the revenue line does not show. It sells the sequencer, the chemistry and the cleared companion diagnostic, and it earns on a hospital’s menu whether or not that hospital sends the sample out. When therapy selection becomes routine, Roche’s exposure is to the number of laboratories running plasma panels; Guardant’s exposure is to the number of tests it personally performs. Those are different markets on the same clinical question.

Tempus: growth that cannot be decomposed

Tempus reported USD 1,271.8 million of revenue in 2025, up 83.4 per cent, of which Diagnostics was USD 955.4 million, up 111.5 per cent. Oncology volume grew 26 per cent and hereditary volume 29 per cent. Diagnostics revenue cannot grow 111.5 per cent while oncology volume grows 26 per cent unless price, mix or acquired revenue is doing most of the work, and the release attributes the growth to the acquisitions of Paige AI and OneOme without quantifying their contribution.

The arithmetic is stated here and not resolved, because it cannot be resolved from disclosure. This report takes 15 per cent of Diagnostics revenue as the liquid share and treats the remainder as tissue and hereditary testing outside the perimeter.

GRAIL: the most famous product, the smallest revenue line

GRAIL’s total revenue was USD 147.2 million in 2025, up 17 per cent, with US Galleri revenue of USD 136.8 million, up 26 per cent. Fourth quarter US Galleri revenue was USD 41.3 million, up 31 per cent, which is the strongest growth rate in the company’s disclosed series and implies the year accelerated into its close.

Galleri detects more than 50 cancer types and has no national coverage pathway. Multi-cancer early detection has no equivalent of the colorectal screening benefit, and that absence — not the test’s performance — is what caps this revenue line at a fraction of the segment it dominates by reputation.

Adaptive, Burning Rock, and the shape of the tail

Adaptive Biotechnologies reported USD 277.0 million of total revenue in 2025, up 55 per cent, with MRD revenue of USD 212.3 million, up 46 per cent, and 105,587 clonoSEQ tests delivered, up 39 per cent. It received expanded Medicare coverage for clonoSEQ in recurrence monitoring for mantle cell lymphoma and reported its MRD business at positive adjusted EBITDA and positive cash flow.

Burning Rock reported revenue of RMB 539.6 million, or USD 77.2 million, up 4.6 per cent. The composition matters more than the total: central laboratory revenue fell 8.9 per cent to RMB 160.0 million, in-hospital revenue was flat at RMB 224.1 million, and pharma research and development services grew 34.5 per cent to RMB 155.5 million. Net loss narrowed to RMB 55.3 million, and overseas revenue reached RMB 101.0 million.

That is the Chinese model in one set of numbers: the laboratory business is shrinking, the hospital-embedded business is subsidising access, the growth is in services to drug developers, and the export business is the only line with a price. A market-share table that treated RMB 539.6 million as clinical testing revenue would be wrong by about 29 per cent.

What actually competes

Not sensitivity. Every credible plasma panel in this market detects what it claims to detect at the limit it claims. What competes is the evidence package that supports a coverage decision, the guideline recommendation that makes an ordering physician’s default, and the installed relationship with the laboratory or health system that controls the sample.

That is why the participants with the largest revenue in this table are also the ones spending most on evidence generation, and why the ones without that spending are either selling to drug developers or exiting.

Value Chain

Value chain: where the margin sits, and who collects it

Every liquid biopsy test in this market passes through the same seven steps, and the economics of the market are decided by which step holds pricing power. Reading the chain from the blood draw to the payment is the fastest way to understand why two companies with similar assays can report 66 per cent and negative 42 per cent gross margins in the same year.

The chain

1. Collection. A phlebotomy draw, a kit, a courier and a consent. Low value, high logistics. Where the sample is drawn determines which laboratory can test it, which is why the collection network is a strategic asset rather than an administrative one.

2. Cell-free DNA extraction and library preparation. Where the technical difficulty begins. Circulating tumour DNA is a small fraction of total cell-free DNA, and the lower the tumour fraction the more the library preparation decides whether a signal exists at all. This is the step that separates a screening test from a monitoring test: a screening test is looking for a signal in a person who may have no detectable disease, so specificity is set here.

3. The assay. Three families compete, and each has a different cost structure and a different regulatory posture. Mutation-based comprehensive genomic profiling interrogates a fixed gene panel — CMS descriptors in this market run from 55 to 74 genes for Guardant360, 311 or more genes for FoundationOne Liquid CDx, and 648 genes for Tempus xT CDx. Methylation-based testing interrogates epigenomic patterns: Shield’s descriptor covers more than 2,500 differentially methylated regions. Personalised tumour-informed MRD assays are built for one patient from that patient’s own tumour sequencing, which makes each test a small bespoke project — Signatera’s descriptor specifies assays personalised to each patient based on prior next generation sequencing of the patient’s tumour and germline DNA.

4. Sequencing. A consumable toll paid to a platform supplier. This is the only step in the chain where a single vendor’s pricing affects every competitor simultaneously, and it is why the reagent line is the one place where a vertically integrated participant can gain a structural cost advantage.

5. Bioinformatics and the algorithm. Increasingly the actual product. Two laboratories running the same chemistry on the same sequencer produce different results if their variant calling, their noise model or their classifier differs. This is where gross margin is created or destroyed, because the compute is cheap and the validation is not.

6. The ordering physician and the care pathway. The gate that decides volume. A test that answers a question a guideline already asks gets ordered by default; a test that answers a new question needs a physician to change behaviour. Veracyte’s tissue business demonstrates the difference outside this perimeter — a single test answering a guideline question grew 27 per cent to USD 310.7 million on about 102,000 tests, which is a price and a volume that a broad menu cannot match.

7. The payer. Where the price is set. Not negotiated with the laboratory in any conventional sense, but administered, and administered by a process with dates attached.

Step seven is the whole chain

The administered price mechanism in the United States works like this. A new advanced diagnostic laboratory test is paid during an initial period of roughly nine months at the laboratory’s own list price. After the initial period, payment is set from the median of actual Medicare claims for that test. That is a designed transition from launch pricing to revealed pricing, and every product in this market passes through it.

The documented cases in this market show both directions of travel, which is why the mechanism has to be read rather than assumed. Signatera’s code, 0340U, entered its initial period on 1 July 2021 at USD 3,500.00 and that period ran to 31 March 2022. Shield’s code, 0537U, entered its initial period on 1 April 2025 at USD 1,495.00, with the period running to 31 December 2025. Guardant360’s code, 0242U, entered at USD 5,000.00, and FoundationOne Liquid CDx, 0239U, at USD 3,500.00.

Then, in the 2025 clinical laboratory fee schedule, code 0340U was first posted at USD 2,919.60 and revised ten days later to USD 3,920.00 in the file modified 8 January 2025. Both figures are on the record. The point is not which one is correct; the point is that the number a laboratory is paid per test is an administrative output that moved 34 per cent inside ten days without any change in the product.

For a business whose revenue is volume multiplied by an administered price, that is the single most important fact in the value chain. A five per cent change in the administered price of a test with a 60 per cent gross margin is larger than the contribution of several quarters of volume growth.

Where the margin actually lands

Gross margin among the participants that disclose it is compressed into a narrow band, which tells you that the laboratory step is not where differentiation lives. Guardant reported non-GAAP gross margin of 66 per cent, Natera 64.7 per cent, and Adaptive reached positive adjusted EBITDA in its MRD business. These are respectable gross margins for a testing business and they are achieved by companies reporting net losses, because the cost that destroys the operating margin sits below the gross line: sales, evidence generation and the bioinformatics organisation.

The recurring test is where a fixed cost base can be absorbed. A therapy selection test is ordered once per patient per line of therapy, so its commercial cost is amortised over one transaction. A monitoring test is ordered serially for years, so the same commercial cost is amortised over many. That is the mechanical reason Natera’s margin expands with volume while Guardant’s revenue per test falls, and it is the reason this report gives MRD twenty per cent growth and therapy selection nine.

The payer as the customer of record

For most of the market the payer is the customer of record and the patient is the user. That inverts the usual direction of product feedback: a laboratory cannot raise price by improving the product, and it cannot cut price to win volume without destroying the administered rate that its competitors also depend on.

The consequence is that the value chain rewards a specific kind of investment — evidence that changes a coverage policy — and punishes a different kind — capacity built ahead of coverage. GRAIL’s USD 62.6 million gross loss in 2025 on a test with no national coverage pathway is the clearest available illustration of what building ahead of the payer costs.

Where the chain is likely to change

Two shifts would move margin along the chain during the forecast window. The first is the migration of monitoring volume from centralised laboratories into hospital laboratories, which is already visible in Burning Rock’s numbers: central laboratory revenue down 8.9 per cent, in-hospital revenue flat, services to drug developers up 34.5 per cent. If that pattern spreads, the value migrates from the test to the platform and the royalty, and the participant that owns the cleared menu collects a smaller share of a larger volume.

The second is the arrival of a screening test with a national coverage pathway. That would move the payer from a per-test adjudicator to a programme buyer, which is a different commercial relationship entirely, and it would create the only genuinely volume-driven revenue line in this market. It is also the scenario with the widest range of outcomes, which is why the forecast assigns screening 20.5 per cent of 2033 rather than a share that assumes it succeeds.

Investment Activity

Investment activity: capital is the input, not the output

In most markets covered by Pheonix Research, capital intensity is a property of the plant: a fab, a refinery, a fleet. In liquid biopsy it is a property of the income statement. The laboratories are not capital-hungry in the conventional sense — sequencers are bought, not built — and yet this is one of the most capital-consuming categories in diagnostics, because the money goes into evidence generation, coverage submissions and a commercial organisation that has to call on oncologists before the payer has agreed to pay.

What the 2025 filings actually show

Three of these companies — Natera, Tempus and GRAIL — reported combined net losses of USD 861.6 million in 2025. Guardant’s free cash flow burn adds a further USD 233 million, though it is a different measure and should not simply be added to the others. Against that, Natera generated USD 107.6 million of positive cash inflow and Adaptive’s MRD business turned cash-positive at the segment level.

The distribution is the finding. The participant with the largest instalment of recurring monitoring revenue is cash-generative; the participant with the most famous screening product and no coverage pathway has the worst ratio of loss to revenue in the table. That is not a verdict on either company’s science. It is a verdict on which business model the payment system currently rewards.

GRAIL is the clearest case study in what a payer gap costs

GRAIL’s 2025 result included amortisation of Illumina acquisition-related intangible items of USD 138.3 million and an intangible asset impairment of USD 28.0 million. The company reported a gross loss of USD 62.6 million, which means the direct cost of delivering Galleri exceeded the revenue it generated, and an adjusted gross profit of USD 73.6 million once those items are excluded.

The structure of that result is instructive beyond the numbers. A strategic acquirer bought a screening business on the strength of a clinical proposition, carried the intangible on its own balance sheet, absorbed the operating loss, impaired part of the asset, and then returned the company to the public market as an independent listing under the ticker GRAL. Whatever the strategic rationale at each step, the financial sequence is that a loss-making screening programme was moved between three sets of owners — private investors, a platform company, and public shareholders — without ever being funded by a payer.

That is the central capital fact of this market. Screening revenue in 2025 was USD 277.6 million against a category of USD 3,757.6 million, and the largest participant in it was loss-making at the gross line.

Tempus shows the cost of being early, and the value of a data asset

Tempus reported a net loss of USD 245.0 million in 2025, which included USD 136.3 million of stock compensation expense and related employer payroll taxes. Nearly all of the operating improvement came from gross profit growth of 109.4 per cent, with adjusted EBITDA improving USD 97.3 million to negative USD 7.4 million even after absorbing two acquisitions.

The acquisitions — Paige AI and OneOme — are the other half of the story. Paige is a computational pathology business and OneOme is pharmacogenomics; neither is liquid biopsy. A company growing Diagnostics revenue 111.5 per cent while oncology volume grows 26 per cent is buying revenue and capability rather than growing them, and the release does not decompose which.

What supports the multiple here is not the test margin. It is the remaining total contract value of over USD 1.1 billion and net revenue retention of 126 per cent in Data and Applications — a software-like revenue line attached to a testing business. That is a different investment thesis from the pure-play laboratories, and it is why Tempus’s capital story does not read like GRAIL’s.

Guardant’s position is the most interesting in the table

Guardant narrowed free cash flow burn to USD 233 million from USD 275 million while growing total revenue 33 per cent and lifting non-GAAP gross margin four points. On any operating trajectory that is a company converging on self-funding.

It is also a company whose next two revenue catalysts are administrative rather than commercial. The Reveal chemotherapy monitoring package has been submitted to MolDx for Medicare reimbursement, and Shield’s initial payment period ran only to 31 December 2025. Both of those are decisions taken by somebody else.

Where the money goes

Very little of the capital consumed in this sector is fixed asset. The recurring investments are three: clinical evidence — prospective trials large enough to support a coverage decision, which for a screening indication means tens of thousands of participants; coverage and health economics — the submissions, the guideline work, the payer meetings; and commercial infrastructure — a field force calling on oncologists and pathologists in advance of reimbursement.

All three are operating expenses with a multi-year payback and no salvage value if a coverage decision goes the wrong way. That is why the sector’s capital intensity should be read as high even though its balance sheets look light, and why the cost of capital is the most important single governor of this market’s growth rate.

The consolidation logic

Two forces point the same direction. First, scale is required to fund evidence generation, and evidence generation is what unlocks coverage, so a sub-scale laboratory cannot afford the input that determines its own revenue. Second, the recurring monitoring franchise is the only line with a margin structure that improves with volume, which means owning one is worth paying for.

The visible consequence is that the participants with the strongest balance sheets are buying capability — Tempus buying AI and pharmacogenomics, Roche and Illumina holding platform positions across the chain — while the sub-scale participants are either narrowing losses by shrinking their laboratory businesses, as Burning Rock is, or leaving the field. That is a consolidating market with a long tail of independent hospital laboratories, and the tail is not consolidating because it is not for sale.

Technology & Innovation

Technology and innovation: from tissue substitute to protocol component

The technology in this market is easy to describe and easy to misread. Three assay families do almost all of the commercial work, and the difference between them is not sophistication. It is what question each one is built to answer, and therefore who pays for the answer.

Three families, as the payment system actually sees them

The clearest way to read the technology is through the reimbursement descriptors, because a descriptor is a technology that has been reduced to something a payer will price.

Mutation-based comprehensive genomic profiling interrogates a fixed panel of genes for variants, copy number changes and rearrangements. Guardant360’s descriptor specifies plasma analysis of 55 to 74 genes, interrogation for sequence variants, gene copy number amplifications and gene rearrangements. FoundationOne Liquid CDx specifies 311 or more genes, including substitutions, insertions, deletions, select copy number variations and rearrangements. These are genotyping instruments: they tell a physician which alteration is present so that a drug can be selected.

Methylation and epigenomic testing interrogates patterns rather than sequences. Shield’s descriptor specifies analysis of cell-free DNA for epigenomic patterns, next generation sequencing of more than 2,500 differentially methylated regions, with an algorithm reported as positive or negative. This is a classifier, not a genotyper. It does not tell you what is wrong; it tells you whether something is wrong.

Personalised tumour-informed MRD is a different construction entirely. Signatera’s descriptor specifies analysis of minimal residual disease from plasma, with assays personalised to each patient based on prior next generation sequencing of the patient’s tumour and germline DNA, reported as absence or presence of MRD with disease-burden correlation. Each patient’s test is built from that patient’s own tumour.

Those three descriptors are the whole competitive technology map. Everything else in this market is an implementation detail on one of them.

The personalisation trade-off nobody prices

A tumour-informed MRD assay has the best analytical properties in the market and the worst operational ones. Because the panel is derived from the patient’s own tumour, the assay does not have to detect unknown alterations; it looks only for the variants it already knows are there. That is why the sensitivity can reach a limit of detection that a fixed panel cannot match, and it is the scientific basis for the claims in this segment.

The operational cost is that the test cannot be run until the tumour has been sequenced, and it cannot be repurposed if the patient is treated elsewhere. Every order requires a prior tissue sequencing event that somebody has to have paid for. Natera has built its position on this dependency and it is a genuine moat — the tumour sequencing business and the monitoring business reinforce each other — but it also means MRD growth is bounded by the number of patients whose tumours have already been sequenced, not by the number who could benefit.

Natera reported over 650 issued or pending patents and over 350 peer-reviewed publications as of 31 December 2025. That is what the personalisation approach costs to defend, and it is why the MRD segment has fewer credible participants than therapy selection.

The performance thresholds are different in each segment, and they are not interchangeable

A therapy selection test has to be broad and accurate at the level of a single alteration, because the clinical decision is binary: this variant, this drug. Sensitivity of a few per cent tumour fraction is sufficient.

A monitoring test has to detect disease at the limit of measurement, because the clinical question is whether a tumour is returning before imaging can see it. Molecular residual disease at the limits claimed by these assays is orders of magnitude below imaging resolution, which is the entire value proposition.

A screening test has to do neither. It has to be specific at population scale, because the cost of a false positive is a colonoscopy in a healthy person and the reputational cost of a bad positive rate is a coverage decision. Shield’s FDA labelling reports 89.6 per cent specificity for advanced neoplasia and 83.1 per cent sensitivity for colorectal cancer, with sensitivity for advanced precancerous lesions at 13.2 per cent.

That last number deserves attention. A blood test that detects established cancer eight times in ten and advanced precancerous lesions about once in eight is a different product from a colonoscopy, and the market has consistently described it as a screening option rather than a screening replacement. The technology’s clinical value and its commercial value are set by different numbers.

The real technology shift: from genotype to surveillance

The transition that matters commercially is not a chemistry transition. It is a change in what the test is priced against.

A therapy selection test is priced against tissue profiling. It substitutes for a tissue biopsy, so its ceiling is the cost and risk of a tissue biopsy — and that is a price ceiling that has been falling for a decade as tissue panels became routine and cheap.

A monitoring test is priced against imaging and clinical follow-up. It substitutes for a CT scan or a surveillance visit, and it can be repeated far more often than either. Guardant expanded tissue-free Reveal into late-stage therapy response monitoring in 2025, which is the same move down the same curve: from answering a question once to answering it repeatedly.

A screening test is priced against nothing, because it does not replace an existing test for most of the population it would address. That is simultaneously why its potential is largest and why its reimbursement is hardest: a payer has to create a new benefit rather than substitute within an existing one.

What would actually change the technology race

Three developments would move value between participants during the forecast window. The first is a guideline recommendation for monitoring in a high-incidence tumour type, which is a clinical-policy event rather than a technical one, and which the coverage framework explicitly requires. The second is the migration of multi-cancer detection from a rule-out tool for symptomatic patients to a population screening instrument, which requires accumulated outcome evidence rather than better classifiers. The third is a cost step-change in sequencing, which would move value from the test to the algorithm and favour participants with a validated classifier over participants with a broad panel.

None of those is a laboratory breakthrough. All three are the kind of change that reallocates a market rather than expanding it, and a forecast that assumed a technical breakthrough would be forecasting the wrong variable.

Market Risk

Market risk: the coverage criteria are tighter than the label

The largest single risk to this forecast is not competitive, technological or financial. It is arithmetic inside a coverage determination, and it can be stated in two sentences.

Risk one: a tenth of a percentage point

The colorectal screening determination effective 8 June 2026 requires a covered non-invasive biomarker test to meet either threshold A — sensitivity of at least 90 per cent with specificity of at least 87 per cent — or threshold B — sensitivity of at least 79 per cent with specificity of at least 90 per cent.

Shield’s FDA labelling reports 83.1 per cent sensitivity for colorectal cancer and 89.6 per cent specificity for advanced neoplasia. Against threshold A, the test is 6.9 percentage points short on sensitivity. Against threshold B, sensitivity clears the 79 per cent floor comfortably, and specificity falls 0.4 percentage points short of the 90 per cent floor.

On the face of the numbers, the only blood-based colorectal screening test with a device approval meets neither threshold. That is a strong statement and it comes with an important qualification, which the reader should weigh rather than skip: the determination specifies specificity in the detection of colorectal cancer compared with colonoscopy based on the FDA labelling, whereas Shield’s 89.6 per cent figure is specificity for advanced neoplasia — colorectal cancer or advanced precancerous lesions. Those are different denominators. A test can be more specific against cancer than against the combined endpoint, or less, and the labelling numbers do not let an outside reader compute which.

So the honest position is not that Shield fails coverage. It is that a coverage decision worth several hundred million dollars of screening revenue turns on a definitional question that a fraction of a percentage point can decide, and that neither the determination nor the label resolves the question on its face. This report treats it as the single largest binary in the forecast and does not assume a resolution in either direction.

The determination compounds the exposure with a second condition: the test must achieve the requirements of the FDA-required post-approval study to continue coverage. Coverage is therefore revocable in a way that a device approval is not.

Risk two: every launch price in this market has an expiry date

New advanced diagnostic laboratory tests are paid at the laboratory’s list price for an initial period of roughly nine months, then at a rate set from the median of actual Medicare claims. Shield’s initial period ran from 1 April 2025 to 31 December 2025 at USD 1,495.00. That period has ended.

What the post-initial rate is cannot be assumed in either direction, and the market has a documented example of the mechanism moving upward. In the 2025 clinical laboratory fee schedule, code 0340U was first posted at USD 2,919.60 and revised to USD 3,920.00 in the file modified 8 January 2025. A laboratory’s administered price moved 34 per cent in ten days with no change to the product.

For a business with a 60 to 66 per cent gross margin, a five per cent change in administered price outweighs several quarters of volume growth. That is not a tail risk; it is the operating environment.

Risk three: the monitoring determination caps volume by design

The MolDx minimal residual disease policy covers a test only where it is medically acceptable that the test precludes other surveillance or monitoring tests intended to provide the same information. That is not an incidental condition. It means the payer will fund monitoring that replaces something, and will not fund monitoring that adds to something.

The ceiling on MRD volume is therefore the volume of surveillance it can substitute for, adjusted by guideline recognition. As imaging intervals lengthen or shorten for a given tumour type, the addressable monitoring volume moves with them, and not in the direction a laboratory would choose.

This is the reason the forecast assigns MRD 20 per cent rather than the 46 to 51 per cent observed at the two fastest-growing participants in 2025. The observed rates are early-adoption rates on a small installed base within covered tumour types. They cannot persist once the covered indication set is the constraint.

Risk four: the growth is partly funded by equity

Natera, Tempus and GRAIL reported USD 861.6 million of combined net losses in 2025, and Guardant burned USD 233 million of free cash flow. A category growing at double digits while its largest participants consume capital is dependent on capital markets remaining open, and that dependence is a forecasting input whether or not it appears in a demand model.

The mitigating evidence is real and should be given equal weight: Natera generated USD 107.6 million of positive cash inflow, Adaptive’s MRD business turned cash-positive at segment level, Tempus improved adjusted EBITDA by USD 97.3 million, Guardant cut burn by USD 42 million, and Burning Rock narrowed its loss to RMB 55.3 million. Every participant in the table is converging. The question is the slope, not the direction.

Risk five: geography and pricing power outside the United States

This market is priced in the United States and performed everywhere. US administered prices for these tests run from USD 1,495 to USD 5,000 a test in the documented initial periods. Outside the US there is no comparable price.

Burning Rock’s 2025 results show what that means operationally: central laboratory revenue down 8.9 per cent to RMB 160.0 million, in-hospital revenue flat at RMB 224.1 million, growth coming from services to drug developers at RMB 155.5 million, and overseas revenue of RMB 101.0 million out of a total of RMB 539.6 million. The Chinese clinical testing business is not growing; the export and services businesses are. A forecast that extrapolated US revenue per test into Asian volume would be wrong by an order of magnitude.

Risk six: the false positive is a capacity problem

A 10.4 per cent false positive rate at population scale is a gastroenterology capacity question before it is a diagnostics question. For every thousand people screened, roughly a hundred need a follow-on colonoscopy. The coverage determination requires that the patient be informed of exactly that, and it requires the test to be used only in average-risk asymptomatic people aged 45 to 85.

If screening volume grew at the forecast rate, the constraint on it would move from payer approval to endoscopy capacity within a small number of years. That is a real ceiling on the outer years of this forecast and it is not modelled separately.

Sensitivities

Two variables dominate. If molecular residual disease grows at 15 per cent rather than 20 per cent, the compound annual growth rate falls from 16.20 to 14.21 per cent. If blood-based screening grows at 20 per cent rather than 32 per cent, the rate falls to 14.53 per cent. If both happen, the market still grows in low double digits, which is the point worth holding on to: this forecast is not fragile to any single segment, only to the reimbursement framework as a whole.

The asymmetry that matters is the opposite one. A favourable colorectal coverage decision and a guideline recommendation for monitoring in a high-incidence tumour would together justify a rate above 16.20 per cent, and would do so without any change in the underlying technology.

Assessment

Overall market risk is high, and it is high for a specific reason rather than a general one. The demand is not in question: cancer incidence is not going to fall, oncologists are not going to stop wanting molecular information, and the analytical performance of these tests is established. What is in question is the price, the covered indication set, and the frequency — and all three are administrative variables set by payers rather than commercial variables set by competition.

A reader who wants one sentence: this market’s growth rate is a policy outcome, and the policy is being rewritten in 2026.

Regulatory Landscape

Regulatory landscape: coverage is granted test by test

This market has two approval routes and one payment route, and it is the payment route that governs everything. That is an unusual configuration, and misunderstanding it is the most common error made about liquid biopsy.

Route one: the device pathway

A blood test sold as a diagnostic product needs marketing authorisation. Shield received it as a premarket approval, P230009, approved on 26 July 2024 — the first blood test authorised as a primary screening option for colorectal cancer in adults aged 45 and over at average risk.

The approval rested on a clinical validation cohort of 10,258 people, of whom 7,861 were eligible and evaluable. The published performance is 83.1 per cent sensitivity for colorectal cancer, with a 95 per cent confidence interval of 72.2 to 90.3 per cent, and specificity for advanced neoplasia of 89.6 per cent, with a confidence interval of 88.8 to 90.3 per cent. Sensitivity for stage I, II or III colorectal cancer was 87.5 per cent. Sensitivity for advanced precancerous lesions was 13.2 per cent. In plain terms, as the FDA’s own summary puts it, one in ten people without advanced abnormal growths will have a false positive result.

Route two: the laboratory-developed test

Most tests in this market are not devices. They are laboratory-developed tests, built and performed inside a single CLIA-certified laboratory, and their regulatory basis is the laboratory’s own validation rather than a marketing authorisation.

That basis was contested and then restored. The FDA issued a rule on 6 May 2024 that would have phased laboratory-developed tests into device regulation. The rule was vacated on 31 March 2025, and on 19 September 2025 the agency reverted to the pre-2024 framework. The practical consequence for this market is that the device pathway remains optional for most therapy selection and monitoring tests, and the participants that hold device authorisations hold them for strategic reasons rather than because they are compelled to.

What is not optional is coverage. A laboratory-developed test that no payer covers is a research service with a price list, and the coverage decision is taken by a different body using different criteria from the ones a device approval tests.

The national coverage determination for colorectal screening

Medicare’s colorectal cancer screening benefit, NCD 210.3, was extended to blood-based biomarker tests effective 19 January 2021. The current version took effect on 8 June 2026, with claims processing implementation on 4 January 2027, and it is more specific than what it replaced.

The conditions are cumulative. The test must be ordered by the physician, physician assistant, nurse practitioner or clinical nurse specialist who will use the result in managing the patient. The patient must be aged 45 to 85, asymptomatic, and at average risk — which excludes anyone with a personal history of adenomatous polyps, colorectal cancer or inflammatory bowel disease, and anyone with a relevant family history. The patient must be given information about test performance and the importance of a follow-on colonoscopy after a positive result. Coverage is once every three years.

Then the test itself. It must be FDA market authorised and indicated for colorectal cancer screening. It must achieve the requirements of the FDA-required post-approval study as specified in the safety and effectiveness data to continue coverage. It must be processed in a CLIA-certified laboratory. And it must demonstrate performance meeting either of two thresholds: sensitivity of at least 90 per cent with specificity of at least 87 per cent, or sensitivity of at least 79 per cent with specificity of at least 90 per cent, in the detection of colorectal cancer compared with colonoscopy, based on the FDA labelling.

Two features of that construction matter commercially. The post-approval study condition means coverage is conditional on continued evidence generation after approval, which converts a one-off regulatory cost into a permanent one. And the two-threshold structure means a test can fail on specificity while passing comfortably on sensitivity, which is what the risk chapter examines.

The local determination that governs monitoring

Molecular residual disease has no national coverage determination. It is covered, where it is covered, by Medicare administrative contractor policies, of which the MolDx programme is the most consequential because it covers a large share of the laboratories that perform these tests.

The MolDx local coverage determination for minimal residual disease testing in cancer provides limited coverage, and the conditions are demanding. Where next generation sequencing is used, the conditions of NCD 90.2 must be met — broadly, that the patient has advanced cancer, plans to be treated for it, and has not previously been tested for the same genetic content. The patient must have a personal history of cancer within the test’s intended use. The identification of recurrence or progression must be recognised in National Comprehensive Cancer Network or equivalent guidelines as a condition requiring a definitive change in patient management. The test must demonstrate that it identifies molecular recurrence before clinical, biological or radiological evidence, at sensitivity and specificity comparable with or superior to the imaging standard for that tumour type. And the test must be medically acceptable on the basis that it precludes other surveillance or monitoring tests intended to provide the same information, unless those are needed to confirm or follow up the finding.

That structure is why monitoring revenue is concentrated in a small number of tumour types, and why companies in this segment invest so heavily in guideline inclusion. A test cannot be covered for a use that a guideline does not already require, which makes the clinical guideline a regulatory instrument in everything but name.

What does not exist

There is no national coverage pathway for multi-cancer early detection. Galleri detects more than 50 cancer types and has no equivalent of the colorectal screening benefit. The absence is not a gap in the evidence base that a better study would close; it is an absence of a benefit category, and creating one requires legislation or a coverage determination that does not currently exist.

That single fact explains the shape of the screening segment: USD 277.6 million of revenue inside a USD 3,757.6 million market, with the largest participant loss-making at the gross line, in the category that attracts the most attention. Attention and reimbursement are not correlated in this market, and the regulatory structure is why.

The approval pathway, summarised

Liquid biopsy in the United States runs on a hybrid structure: premarket approval or clearance where a product claim is needed, laboratory-developed test validation where it is not, local coverage determinations for most monitoring, national coverage for a small number of screening indications, and an administered price set through the clinical laboratory fee schedule with an initial period at list price followed by claims-median pricing. Regulatory complexity is high, and it is high in a specific way: the binding constraint is rarely whether a test may be sold, and almost always whether it will be paid for, at what price, for which patients, and how often.

Further Reading

Two published analyses examine the two variables this forecast is most exposed to: a coverage determination that turns on a fraction of a percentage point, and the administered price that resets a test’s revenue without any change to the product.

• A Four-Tenths-of-a-Point Gap Decides a Product Line – The colorectal screening determination offers two routes, 90% sensitivity with 87% specificity or 79% with 90%. Shield’s labelling reports 83.1% and 89.6%, which is 6.9 points short on the first route and 0.4 points short on the second, and the article explains why the label does not let a reader resolve the definitional difference that decides it.
• The Administered Price Is the Real Forecast Variable – Every test in this market launches at its own list price under an ADLT initial period and then reprices to the median of actual claims. Code 0340U was posted at USD 2,919.60 and revised to USD 3,920.00 in the file modified 8 January 2025, a 34.26% move in ten days against gross margins of 60% to 66%.