A Four-Tenths-of-a-Point Gap Decides a Product Line

Two thresholds, one test, and four-tenths of a percentage point

The colorectal cancer screening determination effective 8 June 2026, with implementation on 4 January 2027, sets a condition that a non-invasive biomarker test must satisfy to be covered. It offers two routes. The test must demonstrate either at least 90 percent sensitivity with at least 87 percent specificity, or at least 79 percent sensitivity with at least 90 percent specificity.

Shield, the only blood-based colorectal screening test with a device approval, reports 83.1 percent sensitivity and 89.6 percent specificity in its FDA labelling. Place those four numbers against the two routes and the outcome is uncomfortable. Against the first route, sensitivity is 6.9 percentage points short. Against the second, sensitivity clears comfortably — 83.1 against a floor of 79 — and specificity falls 0.4 points short.

On the face of it, a test that has been through premarket review meets neither threshold. That conclusion is available, it is what a careless reading produces, and it is not what the liquid biopsy market report says. The reason it does not is the most important thing in this article.

Why the obvious conclusion is not the safe one

The determination specifies specificity for colorectal cancer against colonoscopy, per FDA labelling. Shield’s 89.6 percent is specificity for advanced neoplasia — a broader category that includes precancerous lesions as well as cancer. Those are different denominators, and the label does not let an outside reader convert one into the other.

Specificity is the share of people without the target condition who test negative. Widen the target from cancer to advanced neoplasia and the population of true negatives shrinks, because some of the people who do not have cancer do have an advanced adenoma. A specificity figure computed against the wider definition is therefore not directly comparable to a threshold written against the narrower one, and the direction of the adjustment is not knowable without the underlying data.

So the honest statement is not that the test fails. It is that a coverage decision governing an entire product line turns on a definitional question which a fraction of a percentage point can decide, and that neither document available to a reader resolves it. That is a stranger and more useful finding than either a pass or a fail, because it describes how coverage actually gets made: not by comparing a number to a threshold, but by agreeing what the number measures.

Coverage is revocable in a way approval is not

The same determination contains a second condition that is easier to miss and harder to satisfy. A covered test must meet the FDA-required post-approval study requirements in order to continue to be covered.

That is a different kind of obligation from the one attached to approval. A device approval is a permission that persists unless the agency withdraws it, and withdrawal is a formal action with a process. A coverage condition is a payment decision that lapses by default if a study milestone is missed, and the party harmed is the manufacturer rather than the patient. For a business whose revenue depends on a screening population measured in millions of people, a lapsed coverage determination is indistinguishable from a product withdrawal.

This is the structural asymmetry that the piece on coverage as the product argues from the profit-and-loss side. Approval is permission to sell. Coverage is permission to be paid. In a market where the buyer and the payer are different parties, the second is the binding constraint, and it can move without any change to the product.

The same mechanism caps MRD by design

The clearest demonstration that coverage design shapes market size rather than merely gating it sits in minimal residual disease. MolDx coverage for MRD testing applies where the result precludes other surveillance — that is, where a negative result replaces imaging or another monitoring modality rather than being added to it.

That is a sensible rule for a payer, and it is a volume cap written into the coverage language. A test that must displace an existing test cannot be ordered alongside it, so its addressable volume is bounded by the surveillance it replaces rather than by the number of patients who might benefit. This is why the liquid biopsy forecast sets MRD at 20 percent, well below the 46 to 51 percent growth the fastest participants reported in 2025. Those reported rates are measured off a small base inside a coverage envelope that has not yet been tested at scale.

An analyst who extrapolates a 51 percent test-volume growth rate into a 2033 market has made an assumption about coverage policy without stating it. The report states it, and the sensitivity is direct: MRD at 15 percent instead of 20 percent takes the category compound rate from 16.20 percent to 14.21 percent. Screening at 20 percent instead of its base case gives 14.53 percent.

What this means for the size of the market

Screening is the segment where a single determination can move a category, because screening is sold to healthy people at population scale. A diagnostic test ordered for a symptomatic patient has a physician advocating for it. A screening test ordered for an asymptomatic population has to justify itself against the cost of the programme, and the threshold is set accordingly.

That is also why the segment’s growth carries more policy risk than its clinical promise suggests. The clinical case for blood-based colorectal screening is straightforward — colonoscopy is unpleasant, uptake is poor, and a blood draw is easy to obtain. The commercial case depends on a specificity definition and a post-approval study milestone, neither of which is a scientific question.

The piece on the clinical laboratory fee schedule documents the other half of the same mechanism: even where coverage exists, the rate is administered and can be revised in an administrative cycle. Coverage decides whether a test can be ordered. The rate decides whether ordering it is worth doing.

What to watch, and what would settle it

Three things. The first is whether the specificity definition in the final implementation language is written against colorectal cancer or against advanced neoplasia — that single clause determines whether the only approved blood-based screening test meets either threshold, and it will be visible in a document rather than in a press release.

The second is the post-approval study timetable. A coverage condition tied to study completion is a deadline, and deadlines in this market have historically been met late and absorbed by the manufacturer.

The third is whether the larger participants treat screening as a coverage problem or a marketing problem. The ones that have succeeded in this category — the two-test company described in the companion piece — built the health-economic dossier before the launch rather than after it, and treated the coverage determination as the product specification. The ones that did not are still waiting for a definition to be clarified.