Mirvie

The assay and the model are inseparable here, which is why this sits at B rather than A. Sequencing cell free RNA produces a very high dimensional measurement, and turning that into a binary risk statement is a machine learning problem: the signature is discovered rather than specified in advance, and the company describes combining transcriptomic analysis with machine learning to do exactly that.

What holds it below A is that the laboratory work is equally load bearing. Without the assay there is nothing for the model to read, and the accumulated cohort of sequenced pregnancies is itself the asset, which is the moat is the dataset case this index applies elsewhere.

Low autonomy and appropriately so. The output is a risk classification delivered to a pregnant woman and her clinician, and the response it points toward, low dose aspirin and closer monitoring, is an established guideline supported intervention that a clinician prescribes.

One feature deserves scrutiny rather than credit by default. The product is sold with a virtual assistant alongside the result, which means an automated system is in conversation with a pregnant woman about her own risk of a condition that can become an emergency. Nothing published describes what that assistant will and will not say, how it handles a woman reporting symptoms, or when it directs her to seek care rather than answering her question.

The mechanism is named at a level that lets an outside reader evaluate it: cell free RNA measured from maternal blood, interpreted through transcriptomic analysis combined with machine learning, with the biological rationale stated as reading placental health rather than inferring risk from maternal characteristics.

Performance figures are published with the gestational window specified and with peer reviewed papers behind them. One important qualification is recorded on the governance axis: the headline sensitivity figure is stated for a defined subgroup rather than for the whole tested population. No model architecture, feature count or classifier description is published.

Nothing identifies any party in the chain, and no retention schedule, biobanking position or statement on whether samples and sequence contribute to future model development was located. A distinctive feature of the sample deserves naming because this index has not encountered it before.

Cell free genetic material in maternal blood is substantially of placental and fetal origin, so the material analysed carries molecular information about a second individual who cannot consent and who does not yet exist as a patient.

That is a different stewardship object from an ordinary blood test on one adult: the consent obtained covers the person who gave the sample, the information derived describes two, and one of them may later become a person with their own interest in what was retained about them before birth.

It sits alongside rather than inside the genomic permanence problem recorded elsewhere in this index, because permanence is compounded by the subject having had no capacity to participate in the decision at all. The training question is the material one here given that the company's stated advantage is a cohort of many thousands of sequenced pregnancies, which is precisely the asset that would improve future models and precisely the material whose reuse nobody has described.

Ask for the retention schedule on sample and sequence separately, the biobanking position, whether stored material trains models, and what a person can request about a sequence taken before they were born.

A strong B and worth explaining precisely, because the evidence is genuinely good and sits one rung below this index's top bar.

What exists is prospective, multi centre and peer reviewed: a validation drawing on more than 9,000 pregnancies from a prospective study, published in Nature Communications, with prior work in Nature and in a leading obstetrics journal, and further findings presented at a specialty meeting. Very few records here can show prospective validation published in journals of that standing.

What is absent is the two things this index treats as A. There is no randomised comparison against standard care, so nothing yet shows that acting on the result changes outcomes rather than that the prediction is accurate. And the test has not been adopted into the national guidelines it is positioned against, which remain based on maternal characteristics. The underlying study is also company sponsored, which does not diminish peer review but is worth stating.

Graded on an honest basis, with a distinctive feature of the sample worth naming.

Cell free RNA in maternal blood is substantially of placental and fetal origin, so the material analysed carries molecular information about a second individual who cannot consent and who does not yet exist as a patient. That is a different stewardship object from an ordinary blood test on one adult, and it sits alongside the genomic permanence problem this index recorded on the liquid biopsy records rather than being the same thing.

No retention schedule, biobanking position or statement on whether samples and sequence contribute to future model development was located, and for a company whose stated advantage is a cohort of nearly 11,000 sequenced pregnancies, that last question is the material one.

Graded on an honest basis. No compliance documentation or agreement posture was located.

As with the other laboratory records in this lane, the business associate frame fits awkwardly. A laboratory receiving a sample on a clinician's order and reporting a result is generally a covered entity in its own right rather than a business associate of the ordering practice, so the instrument this axis usually turns on is not the governing one, and consent sits in the requisition rather than in a vendor contract.

Recorded honestly: the dedicated trust and security search this index requires was not run in this pass, so the grade is provisional and should not be quoted until it has been. No attestation was encountered incidentally.

No clearance or approval was located and none is claimed. A test of this kind is ordinarily offered as a laboratory developed test under clinical laboratory certification, which permits a laboratory to validate and offer its own assay without a device submission.

That route is legitimate and it is also the reason a buyer should read the evidence rather than rely on a regulatory shorthand: the validation here rests on the company's published studies rather than on an agency having reviewed them. The status of federal oversight of laboratory developed tests has been contested and in flux, so a purchaser should confirm the current position rather than assume it. The company states nothing that overstates its regulatory standing, which is to its credit.

This is the axis that matters most on this record and the disclosure is partial in a specific and consequential way.

The headline sensitivity figure of 91 percent is reported for women aged 35 and over without pre existing high risk conditions. That is a defined subgroup, not the whole tested population, and a figure quoted for a favourable subgroup in a headline is a different claim from one quoted overall. A reader should establish the performance across all ages before treating 91 percent as the product's sensitivity.

The condition makes this sharper than it would be elsewhere. Preeclampsia and maternal mortality show some of the widest and best documented racial disparities in United States medicine, and this product proposes to reassign risk in exactly the population where existing risk assessment is contested. A test that performs unevenly across groups here would not merely be inaccurate; it would redirect preventive treatment away from the women who most need it. No breakdown by race or ethnicity was located.

The mechanism is named at a level that lets an outside reader evaluate it, and the biological rationale does more work than it appears to. Cell free genetic material measured from maternal blood is interpreted through transcriptomic analysis combined with machine learning, and the stated rationale is that the test reads placental biology rather than inferring risk from maternal characteristics. That distinction is the substantive one for this indication.

Obstetric risk prediction has a long history of models built on demographic and historical variables, which encode the disparities in who develops complications rather than measuring the process that produces them, and a test grounded in a biological signal is making a categorically different claim that can be tested on its own terms.

Performance figures are published with the gestational window specified, which matters because a prediction is only actionable if there is time left to act, and peer reviewed papers sit behind them. Held below the top grade on two points. The headline sensitivity figure is stated for a defined subgroup rather than for the whole tested population, which is a materially different claim and should be read as such wherever it is quoted.

And no model architecture, feature count or classifier description is published, and no warranty, indemnity or remediation commitment attaches. Ask for performance across the full tested population, the false positive rate, and what a clinician is expected to do differently on a positive result.

Not described. No integration, result delivery mechanism or ordering workflow was located.

The practical question for an obstetric practice is whether a result arrives in the record where a clinician will see it at the right visit, since the actionable window is narrow and the intervention it points to, aspirin prophylaxis, has to begin early to be effective. Nothing published addresses it.

A centralised laboratory service rather than deployed software, so the usual questions of hosting and customer controlled infrastructure do not apply in their normal form. Samples travel to the company; results come back.

What is not published is where sequencing and analysis occur, how long samples and derived data are retained, and whether the resulting sequence joins the research cohort the company describes as the foundation of its platform.

No price, no coverage position and no patient cost was located.

That gap is more consequential than usual because the buyer here is substantially the patient. A pregnant woman offered a test her guidelines do not yet require needs to know what it costs her, whether her insurer covers it and what happens if it does not, and none of that is published. Reimbursed diagnostics in this index are systematically the most price transparent segment precisely because a published rate exists; this product does not yet appear to have one, and a buyer should ask whether it is billed to insurance, offered cash pay, or sold through the practice.

Narrow by design and coherent. One specialty, obstetrics, one moment in the pathway, the second trimester between roughly 17.5 and 22 weeks, and at present one commercial indication, preeclampsia risk.

The platform is described as extending to preterm birth and fetal growth restriction with research published or presented on both, so the pipeline is broader than the product. The setting is prenatal care rather than the hospital, which distinguishes it from the labour and delivery surveillance systems elsewhere in this index: this acts months before the events those systems watch for. Coverage appears to be United States only.