OM1 has highlighted its role in the real-world evidence study that supported U.S. Food and Drug Administration approval of Hologic’s Aptima HPV Assay for primary cervical cancer screening. The submission drew on data from more than 650,000 women across multiple U.S. health systems, making it one of the largest real-world evidence studies used in support of an FDA medical device approval.
Why OM1’s real-world evidence role matters for FDA-grade diagnostics submissions
The significance of the OM1-supported study is not limited to the size of the patient population. Large datasets have become common in healthcare analytics, but regulatory-grade evidence requires something more demanding than scale. The data must be traceable, validated, clinically meaningful and structured in a way that can support decision-making by regulators who are assessing patient safety, diagnostic performance and intended use.
That is why this case matters for the diagnostics industry. The Hologic Aptima HPV Assay approval shows how real-world evidence can move beyond supplementary context and become part of the evidentiary backbone for a regulatory submission. For diagnostics developers, the study offers a practical example of how electronic health records, laboratory information systems, cytology reports and pathology records can be converted into a regulatory evidence asset when the process includes rigorous data cleaning, model validation and provenance controls.
The unresolved question is whether this model can be repeated across less standardized clinical settings. Cervical cancer screening produces structured and semi-structured data through established laboratory and pathology workflows, which may make it more suitable for large-scale evidence generation than areas with fragmented documentation. Regulators and sponsors will now be watching whether similar AI-enabled real-world evidence approaches can hold up in disease areas where patient pathways, coding practices and clinical documentation are less consistent.
How the Hologic Aptima HPV Assay study changes the debate around AI in regulatory evidence
AI in healthcare is often discussed in terms of diagnostic algorithms, clinical decision support or workflow automation. OM1’s role in the Hologic submission places AI in a different regulatory context: evidence generation. The platform was used to ingest data, de-identify and tokenize records, map datasets at scale, and extract key variables from unstructured clinical notes, cytology reports and pathology records.
That distinction is important. The AI models were not positioned as a patient-facing diagnostic product in this context. Instead, they were part of the infrastructure used to convert real-world clinical documentation into analyzable evidence. This could be a more scalable near-term use case for healthcare AI because it addresses a persistent bottleneck in medical device and diagnostics submissions: the cost and time required to manually abstract clinical records.
However, the bar for acceptance remains high. Automation does not remove the need for validation. In fact, it increases the importance of model documentation, audit trails and explainable data handling. If AI extracts variables from unstructured records, regulators must be able to understand how those variables were identified, how accuracy was tested, how errors were handled and whether bias could have entered the dataset through missing or inconsistent documentation.
Why unstructured clinical data is becoming a regulatory asset rather than a burden
One of the most important aspects of the study is that key variables were drawn from unstructured sources such as clinical notes, cytology reports and pathology records. Historically, that kind of information has been difficult to use at scale because manual abstraction is expensive, slow and difficult to standardize across sites. OM1 founder Rich Gliklich indicated that the study’s defining features were its scale and the fact that many important variables were embedded in unstructured clinical material.
For diagnostics, this is a meaningful shift. Laboratory and pathology reports often contain clinically rich information that is not always captured cleanly in structured fields. If AI and machine learning tools can extract those variables with sufficient validation, sponsors may be able to design evidence programs that are broader, faster and more reflective of routine clinical practice than traditional manually curated studies.
The risk is that unstructured data can be messy in ways that are not immediately visible. Different institutions may document findings differently, use different reporting conventions or apply different workflows around screening, follow-up and pathology confirmation. An AI-enabled evidence platform has to normalize those differences without flattening clinically important distinctions. That makes validation not just a technical requirement, but a core regulatory credibility issue.
What this reveals about the FDA’s evolving posture toward real-world evidence
The study’s participation in the National Evaluation System for health Technology Coordinating Center program gives the case broader relevance for medical device and diagnostics developers. NESTcc was designed to support the use of real-world evidence for medical devices, and this submission provides a visible example of how such evidence can be organized for regulatory review.
The FDA has encouraged the responsible use of real-world data and real-world evidence for years, but adoption has varied across product categories. In practice, sponsors still need to show that real-world datasets are fit for purpose. That means the data source, study design, endpoint definitions, patient population and statistical approach must match the regulatory question being asked.
The Hologic Aptima HPV Assay case suggests that regulators may be increasingly open to large-scale real-world evidence when the evidence package is built with sufficient traceability and quality controls. However, it does not mean that real-world evidence will automatically replace prospective trials or traditional validation studies. The more realistic takeaway is that real-world evidence may become a stronger component of hybrid evidence strategies, especially where large patient populations and established clinical workflows make data capture more reliable.
Why scale alone is not enough for AI-enabled real-world evidence submissions
The 650,000-patient scale is eye-catching, but the deeper story is whether scale can be converted into regulatory confidence. A massive dataset can reduce uncertainty, improve subgroup visibility and capture real-world variation across health systems. Yet large datasets can also amplify hidden data quality problems if extraction logic, coding practices or missing data patterns are not handled properly.
That is why OM1’s emphasis on quality control, real-time reporting dashboards, data provenance and model validation is commercially relevant. For medical device and diagnostics companies, the value of a platform like this is not simply that it can process more records. The value is that it can potentially make large-scale retrospective evidence generation operationally feasible while maintaining the kind of documentation regulators expect.
Still, sponsors will need to be cautious. A highly successful evidence program in one indication does not automatically create a template for all others. The strength of future submissions will depend on whether endpoints are clearly defined, whether source data are clinically reliable and whether AI-derived variables can be defended under regulatory scrutiny. In other words, automation may reduce manual burden, but it does not reduce accountability.
How this could influence diagnostics companies, medtech sponsors and evidence-generation vendors
For diagnostics companies, the OM1 and Hologic case could make real-world evidence planning more central to product strategy. Instead of treating evidence generation as a late-stage regulatory task, sponsors may increasingly design data strategies earlier, including site selection, data architecture, clinical variable mapping and validation protocols.
For medtech companies, this raises a competitive question. Firms that can generate robust evidence faster may gain an advantage in regulatory submissions, label expansions and post-market commitments. In screening diagnostics, where clinical utility and population-level performance matter, real-world evidence may help demonstrate how a test performs across diverse health systems rather than under narrow controlled conditions.
For evidence-generation vendors, the study also raises expectations. Healthcare AI platforms will need to compete not only on data volume, but on trust infrastructure. That includes documentation, data lineage, privacy-preserving linkage, site usability, audit readiness and regulator-facing transparency. The reported site Net Promoter Score of 87.5 also points to a commercial reality that matters in evidence generation: if sites find the process less burdensome, sponsors may be able to execute larger studies with fewer operational bottlenecks.
What clinicians and regulators may watch after the Hologic Aptima HPV Assay approval
Clinicians tracking cervical cancer screening will be most interested in how the FDA approval affects the role of the Hologic Aptima HPV Assay in primary screening workflows. Human papillomavirus testing has become central to cervical cancer prevention strategies because persistent high-risk HPV infection is a key driver of cervical cancer risk. A regulatory pathway supported by large-scale real-world evidence may strengthen confidence in how diagnostics are evaluated in routine practice settings.
Regulators, however, will likely focus on broader reproducibility. The next question is not whether one large study succeeded, but whether the model can be applied repeatedly with consistent quality. Future submissions using AI-enabled real-world evidence will need to show that variable extraction, cohort construction and endpoint analysis can withstand review across different institutions and clinical contexts.
Industry observers are also likely to watch whether this shifts investment priorities. Diagnostics sponsors may allocate more resources to evidence infrastructure, registry partnerships and longitudinal data access. At the same time, smaller developers may face a new competitive pressure if regulatory-grade real-world evidence becomes expected but remains expensive to build without platform partners.
Why this could become a turning point for regulatory evidence infrastructure
The OM1-supported Hologic Aptima HPV Assay submission looks less like a one-off data project and more like a signal that regulatory evidence generation is becoming an infrastructure problem. The industry has spent years talking about real-world evidence as a way to make trials more representative and submissions more efficient. This case shows the practical challenge behind that ambition: sponsors need systems that can extract, validate and document clinically meaningful data from the messy reality of healthcare records.
The opportunity is clear. AI-enabled evidence platforms could make large studies faster, less site-intensive and more reflective of everyday clinical practice. The caution is just as clear. Regulatory acceptance will depend on whether automation is paired with disciplined validation, transparent provenance and a study design that answers the right clinical question.
For OM1, the Hologic approval strengthens its positioning in AI-enabled real-world evidence at a time when regulators, medtech companies and diagnostics developers are searching for scalable evidence models. For Hologic, the Aptima HPV Assay approval gains an additional strategic layer because it is now tied to a high-profile example of real-world evidence use in FDA decision-making. For the broader diagnostics industry, the message is harder to ignore: the future of regulatory evidence may not be smaller or simpler, but it may become more automated, more traceable and far more data-intensive.
