Deploy First, Validate Later: FDA Lets Digital Chronic-Disease Tools Be Tested in Real-World Care

Digital tools for chronic diseases face a long-standing challenge: without large-scale use, it is difficult to obtain real-world evidence; without sufficient evidence, it is also difficult to enter routine care. The U.S. Food and Drug Administration’s (FDA) TEMPO pilot program seeks to break this cycle through supervised real-world deployment. Its latest list of participants covers two distinct approaches: hypertension medication management and AI-assisted metabolic health.

Cadence Solutions’ platform is supervised by clinicians and used for medication management in patients with hypertension. The core of this type of tool is not merely recording blood pressure, but connecting at-home measurements, care follow-up, and medication-adjustment workflows. Whether it can consistently improve blood pressure control and prevent inappropriate medication adjustments remains to be answered by the routine-care data collected through the program.

Another solution comes from continuous glucose monitoring company Dexcom, which was also the first TEMPO participant announced by the FDA on July 22. Dexcom said its Glucose Health Program will integrate glucose data from G7 and Stelo sensors with information on nutrition, activity, sleep, and stress to provide real-time data and AI-generated personalized prompts; its clinical software will be used for population monitoring and treatment adjustments.

The solution is intended to assist in screening for prediabetes and type 2 diabetes, and to explore whether it can improve glucose control and glycated hemoglobin in people with prediabetes. However, currently available information does not specify the scale of the AI model’s training data, external validation results, or error performance across different populations, nor has it demonstrated that prompts generated from multiple data sources can improve clinical outcomes. Its selection should therefore be understood as the beginning of evidence generation, rather than an endorsement of the algorithm’s performance.

TEMPO is linked to the ACCESS payment model of the U.S. Centers for Medicare & Medicaid Services. When eligible devices are used in ACCESS care settings, the FDA may exercise enforcement discretion regarding certain premarket authorization and investigational device requirements; in exchange, companies must continuously collect, monitor, and report real-world data related to the intended use. This arrangement is not a blanket exemption, nor does it mean that a product has received general marketing authorization.

The FDA has explicitly cautioned that it has not yet evaluated the effectiveness of participating devices for their pilot uses. For the hypertension platform, key questions include whether medication adjustments are safe and whether clinicians can intervene promptly. For Dexcom’s solution, the issues to be examined include screening accuracy, whether AI prompts lead to misjudgments, and whether sustained user participation can produce measurable health improvements.

The FDA expects to select approximately ten U.S. companies in each of ACCESS’s four clinical areas and will continue accepting expressions of interest. What TEMPO is truly testing is not merely two products, but a regulatory approach: whether patients can gain earlier access to digital care while payment, clinical oversight, and real-world evidence together form a safety net capable of constraining risk.