Waiv

Both products are algorithms operating on slides the laboratory already produces, with no assay, reagent, scanner, or sequencing step of their own. That is the entire commercial thesis: extract prognostic and biomarker signal from routine H and E rather than from an additional molecular test. Remove the model and there is no product and no test.

Both products are structurally positioned to inform a clinician rather than replace a test, but they do so differently and the distinction matters. RlapsRisk BC produces a risk score used alongside clinico-pathological factors, explicitly as an additional layer of information rather than a standalone determination.

MSIntuit CRC is a pre-screen that rules OUT microsatellite stable cases so pathologists concentrate confirmatory testing on the rest, which means a false negative removes a patient from confirmatory MSI testing entirely. A rule out design placed upstream of confirmatory testing carries a different risk than a flag placed alongside it, and the two are not interchangeable in deployment.

Development and validation cohorts are disclosed for the lead product, and the research models are public. For RlapsRisk BC the company has described training on more than 1,800 patients and blind, single shot validation on independent cohorts of more than 1,200 slides, reporting a concordance index of 0.81 for the AI score combined with clinicopathological factors against 0.76 for those factors alone, which measures what the model adds rather than its standalone accuracy; the product page adds TAILORx and NSABP B-28 validations.

In July 2026 Waiv released two robustified research foundation models, Phaet and Mascaret, on Hugging Face, fine tuned to reduce sensitivity to scanner, laboratory and staining variation. Architecture details for the clinical products, version history and update practice are not published, and MSIntuit CRC's development is not described.

Waiv's own models on a cloud service whose providers are described but not named. Waiv develops its tests and foundation models itself, so no outside model provider sits behind a result; in July 2026 it released two research foundation models, Phaet and Mascaret, built on the public Phikon-v2 and Midnight-12k models, on Hugging Face. Destra, the platform where customers upload slides and run the tests, is described as SaaS infrastructure, and the security FAQ says Waiv uses cloud providers in the same region as its research partners, HDS certified for European data. No provider, subprocessor or region for Destra customer data is named, and no retention or training term governs uploaded slides.

A structurally short chain at the model layer, with the hosting parties unnamed.

Large trial cohort validations reported by the company, with publication status and method not given on the page. The RlapsRisk BC page reports validation in 7,584 node negative patients from the TAILORx trial, where the histology component was an independent prognostic signature incremental to the 21-gene Recurrence Score, and in 731 node positive patients from NSABP B-28 (hazard ratio 3.6 between high and low risk groups), plus a prespecified pooled analysis (hazard ratio 7.42; negative predictive value 88 to 97.4 percent across cohorts).

The same page notes that one supporting cohort contained only endocrine therapy patients and was enriched with patients who relapsed. RlapsRisk BC was developed with Gustave Roussy. The page links TAILORx and NSABP B-28 research without stating whether each is peer reviewed, and no comparable figures for MSIntuit CRC are published.

Research data handling is described; customer data in Destra is not. Waiv's FAQ on patient data security says that for its research, Waiv and its partners act as independent controllers, use only patient data pseudonymized or de-identified before they access it, keep the key with the providing center, store European data in Europe with HDS certified providers and North American data mainly in North America, and follow privacy by design. The privacy policy covers only data for which Waiv is controller and excludes processing it does for others, which is where a laboratory's slides uploaded to Destra fall. Nothing published states how long uploaded slides are kept, whether they are de-identified, or whether customer cases may train Waiv's models, which matters for a company whose positioning rests on proprietary foundation models and a large patient data network. Safety engineering, meaning how a wrong result is handled as a safety event, is not published.

General assurances and research data practices that do not answer for customer data.

A French company that does not invoke HIPAA, with its processor role implied and no document behind it. The privacy policy is written to the GDPR and French law, covers personal data for which Waiv is controller, and expressly excludes processing Waiv carries out for others, which is the position a laboratory uploading slides to Destra would be in; no data processing agreement is published for that processing. The patient FAQs explain research data rights and security.

No HIPAA statement, business associate position or agreement appears, and no product is offered for clinical use in the United States, though the company reports North American research data and a Daiichi Sankyo collaboration. Nothing says whether Waiv will sign a business associate agreement.

Nothing independent is published for Waiv in its own right. The site carries terms of use, cookie and privacy policies, a patient information page and a password protected instructions for use page, and its security FAQ describes research data practices, stating that European research data is stored by cloud providers certified under the French health data hosting standard (HDS). That certification is the hosting providers', not Waiv's, and the FAQ's "robust technical and organizational measures" are asserted without an outside check. Destra is described as "clinically compliant" on scalable SaaS infrastructure. No ISO 27001 certificate, SOC 2 report, trust center, penetration testing statement or vulnerability reporting route is published. Waiv separated from Owkin in March 2026, and Owkin's own certifications do not transfer to a new company.

Nothing independent published on security.

A European regulatory position stated for the lead product, research use only stated for another, and no US clearance. Waiv's products page separates research products from "IVD solutions that are regulatory-approved and ready for use in clinical routine", and labels RlapsRisk BC, its breast relapse risk test, CE-IVDR and BRCAura, a germline BRCA prescreen, research use only.

RlapsRisk BC and the MSIntuit CRC prescreen were first CE-IVD marked in September 2022, when the business was Owkin's diagnostics arm; the MSIntuit Suite page carries no regulatory label, and the instructions for use page, which would state each product's marking, is password protected. FDA's database lists no clearance under Waiv or Owkin for these products, and nothing published offers them for clinical use in the United States.

Validation limits are disclosed openly; per group results are not published. The RlapsRisk BC page discloses that one supporting cohort was restricted to endocrine therapy patients and enriched with relapses, and reports validation separately in node negative (TAILORx) and node positive (NSABP B-28) patients, which are two clinical populations the product serves, though not compared on a common measure.

No results by age, menopausal status, ethnicity, site or scanner are published, and no governance framework or monitoring commitment is named. The company markets these tests as widening access to precision testing, and the MSIntuit page notes that endometrial cancer mortality is rising disproportionately in non-White groups, which makes per group performance the evidence that matters.

Discrimination figures are published, and nothing stands behind a result. The RlapsRisk BC page reports that in a prespecified pooled analysis the test separated high and low risk groups with a hazard ratio of 7.42 (95% CI 4.32 to 12.75) and a negative predictive value of 88 to 97.4 percent across cohorts, a hazard ratio of 3.6 (2.3 to 5.5) between groups in the NSABP B-28 node positive cohort, and independent prognostic value incremental to the 21-gene Recurrence Score in TAILORx.

For a prognostic score the error that matters is calibration, predicted against observed relapse with the event counts behind it, and the page gives neither the counts nor calibration; it also notes that one development cohort was enriched with patients who relapsed. MSIntuit CRC, a rule out prescreen, has no published error figure. No warranty, indemnity or remediation commitment attaches to a result, and Destra lets users review results with interactive visualizations.

Integration is claimed through an API, with no system named. The Destra page describes the platform as interoperable with image management systems and laboratory information systems, platform agnostic and available across existing viewers, and offers a Destra API for integrating the tests into a laboratory's image management system; results can be exported as PDF, CSV and JSON.

Integration into routine digital pathology workflows was earlier pursued through a partnership with the Cerba Path laboratory network. No image management system, LIS, standard or production integration is named.

A hosted platform with location stated only for research data. Destra is the SaaS platform through which Waiv's tests are used: slides are uploaded singly or in batches, viewed in the browser, analyzed, organized into folders and exported, or the tests are reached through the Destra API from a laboratory's image management system. The security FAQ says Waiv's research data is kept by cloud providers in the region of the partner that supplied it, in Europe with HDS certified providers for European centers, but it describes research processing, not Destra. No hosting provider, region, residency commitment or tenant isolation is published for customer slides, and no named customer site or implementation process is described.

A single hosted option with location implied rather than committed.

No price is published; the value case is framed against the tests the products replace. Waiv presents RlapsRisk BC as faster and cheaper than gene expression tests, which it describes as expensive and not always accessible, and MSIntuit CRC as a pre-screen that reduces the volume of confirmatory MSI testing, and its products page lists cost saving against traditional tests and minutes rather than weeks. That lets a buyer frame value against a known line item, the molecular test avoided, but no fee, per test rate, subscription or tier is published for either product or for the Destra platform they run on.

The sales posture is discoverable and no price is published.

Deliberately narrow indications, stated precisely, with validation behind the lead one. RlapsRisk BC predicts distant relapse risk in ER positive, HER2 negative early breast cancer from H&E slides, and the product page reports validation in 7,584 node negative patients from the TAILORx trial and 731 node positive patients from the NSABP B-28 trial, in addition to earlier cohorts; MSIntuit CRC prescreens colorectal tumors for microsatellite status, and the MSIntuit Suite page extends the approach to gastric and endometrial cancer. The company states intent to expand to further disease areas. Scanners and staining protocols validated for each product are not listed.