\n\n

On October 20, 2026, the U.S. Food and Drug Administration (FDA) issued final guidance, Regulatory Considerations for Prescription Drug Use-Related Software (finalizing a September 2023 draft guidance on the topic). The guidance describes how the agency intends to apply its existing drug labeling authorities to prescription drug use-related software (PDURS) that is disseminated by or on behalf of a drug sponsor for use with a prescription drug product (including a biological product regulated as a drug). The guidance also applies to a prescription drug-led, drug-device combination product for which the Center for Drug Evaluation and Research (CDER) or the Center for Biologics Evaluation and Research (CBER) has primary jurisdiction. The guidance refers to content that the software presents to a patient, caregiver, or healthcare practitioner as “end-user output.” The FDA treats that output as prescription drug labeling when it supplements, explains, or otherwise relates textually to one or more of the sponsor’s drug products, regardless of whether the software is regulated as a device under the Federal Food, Drug, and Cosmetic Act (FDCA). Like other FDA guidance documents, the final guidance is nonbinding.

End-user output generally constitutes FDA-required labeling if it is essential to the safe and effective use of the product or if the FDA determines, based on evidence from the sponsor, that using the software with the drug leads to a meaningful improvement in a clinical outcome. Otherwise, the output generally constitutes promotional labeling. Whether and where the software is described in the Prescribing Information (PI) also depends on whether the function uses data transferred directly from the device constituent part of a combination product. The final guidance adds a labeling decision tree and illustrative PI language for device-connected and standalone software functions. When the PI describes a device-connected function that is optional, the FDA recommends stating that the software is unnecessary for safe and effective use. Certain postapproval changes to FDA-required output may require FDA review and approval under existing labeling rules. For products with traditional approval, applicants must submit promotional end-user output, including updates that change it, on Form FDA 2253 when first disseminated. Accelerated approval generally requires earlier submission of promotional materials.

Within the guidance’s scope, dissemination “by or on behalf of” a drug sponsor could include situations where a software developer distributes or directs end users to the software under an agreement with a drug sponsor. Accordingly, drug sponsors should consider contractual protections and obligations in such agreements to address compliance with drug labeling requirements, consistency with the PI, and applicable medical device requirements if the software function meets the device definition under the FDCA. Whether a software function meets the device definition and whether it requires premarket review are separate questions governed by the FDA’s device authorities. The FDA cautions that any required device marketing submission is separate from a Form FDA 2253 submission. The guidance recognizes that end-user output may be dynamic and indicates that the FDA may provide additional guidance on updates as it gains experience. It does not, however, specifically address generative artificial intelligence (AI). For example, it does not explain how sponsors should capture and submit output that a generative AI-enabled tool generates in response to individual users’ questions about the drug. In complex arrangements involving device software functions that provide outputs that affect the use of a drug product, such as a drug companion app, it may be important to coordinate clinical trials, labeling, quality controls, and marketing submissions for both the software product and the drug product throughout their respective life cycles and to clearly delineate the regulatory obligations between the parties.

In practice, the drug labeling and device frameworks apply in parallel when the software is also regulated as a device. The terms “device-connected” and “standalone” describe where the software obtains its data, not whether the software is a device. Consider the FDA’s example of a drug sponsor’s insulin autoinjector that records the dose and time of each injection and transfers the data to a mobile application. The application is optional, and the sponsor has not submitted evidence showing that it leads to a meaningful improvement in a clinical outcome. The FDA treats the application output as promotional labeling and recommends briefly describing the autoinjector and the software function in the PI. Assume the sponsor later updates the application to recommend doses based on that data. Before release, the sponsor should assess whether the dosing function is a device and, if so, whether it requires a device marketing submission. The sponsor should also check the recommendations against the approved PI and reassess how the new output should be classified and submitted under the drug labeling rules.