health · digital · family: the rulebook was written for a world that no longer exists
software trapped in a hardware rulebook
Prescription digital therapeutics forced through hardware device regulatory pathways
Problem statement
Prescription digital therapeutics (PDTs) — software-based interventions that deliver evidence-based therapeutic treatments — are reviewed using FDA regulatory pathways designed for physical medical devices. These pathways (510(k), De Novo, PMA) have, in the words of a 2023 regulatory analysis, "not always been aligned with the rapid, dynamic, and iterative nature of treatments delivered as software": PDTs can be frequently updated after authorization, pose unique placebo-control challenges in clinical trials, and may be delivered on platforms different from those tested in pivotal studies. A 2025 review in npj Mental Health Research identified four FDA-authorized mental health software devices whose pivotal trials tested prototypes delivered on different digital platforms than the final marketed products, and argued that the platform evaluated in trials should match the version later offered to patients. No regulatory pathway exists that is specifically designed for the characteristics of software-as-therapy.
Why this matters
A January 2022 Research and Markets analysis, cited in the 2023 Frontiers in Digital Health paper, valued the 2021 global digital therapeutics market at $3.35 billion and estimated it would reach $12.1 billion by 2026. The npj review found five mental health software devices classified through the De Novo pathway and 14 cleared through 510(k), for conditions including substance use disorders, insomnia, depression, anxiety, ADHD, and lower back pain. The review argues that the study design behind an authorization shapes the decisions of healthcare stakeholders: it reports that UnitedHealthcare's pharmacy benefit manager, OptumRx, announced it would deny coverage for computer-based CBT devices, calling the approach "unproven until additional studies are available." Pear Therapeutics' reSET, classified by the FDA in September 2017, was the first FDA-authorized PDT to make treatment claims.
What’s been tried and why it hasn’t worked
The De Novo pathway has been used for first-in-class PDTs such as reSET (substance use disorders) and EndeavorRx (ADHD), but the npj review reports that the FDA itself acknowledges its current medical device framework is not well-suited to software. The FDA's Software Precertification Pilot Program, announced in 2017, looked first at the developer rather than solely at the product, but it remained a pilot; the FDA's final report in September 2022 concluded that some changes would need legislation. In January 2026, the FDA updated its Clinical Decision Support and general wellness guidance documents, adopting a more permissive approach to some digital health tools, and withdrew its SaMD clinical evaluation guidance. Those updates concern decision-support and wellness products; the Latham & Watkins commentary notes that the FDA's legal authorities are unchanged and that therapeutic uses are still likely to bring a product under device regulation. Many 510(k)-cleared digital mental health devices lack direct evidence of effectiveness, relying on equivalence to predicates — a concern for software, where, the npj review notes, modifying the design or delivery platform of an intervention can significantly affect how users engage with and respond to it. Developing credible placebo controls for digital interventions remains difficult because user interaction, design features, and engagement are integral to the therapy, and full blinding is rarely achievable.
What would unlock progress
A regulatory pathway specifically designed for software-based therapeutics — one that accommodates iterative development, platform-agnostic evidence standards, and adaptive trial designs — would address the structural mismatch. This could draw on models from software certification in other safety-critical domains (aviation, automotive). The npj review points to Germany, which introduced a new regulatory framework for digital health interventions through formal policymaking, and reports the FDA's own conclusion that legislative changes are needed to regulate software as a medical device more effectively. Consensus on appropriate clinical trial designs for PDTs, including standardized approaches to placebo controls and outcome measurement for software interventions, would provide the evidentiary foundation.
Entry points for student teams
A student team could conduct a systematic comparison of regulatory pathways for software-based therapeutics across jurisdictions (FDA, EMA, PMDA, TGA) to identify which elements of non-U.S. frameworks address the gaps in the current FDA approach. Another entry point would be designing and testing a sham digital therapeutic control condition for a specific therapeutic area, addressing the placebo-control methodology gap. Relevant skills include regulatory science, clinical trial design, software engineering, and behavioral science.
Genes
Structural cousins: stuck the same way, in other fields
Sources
Herpertz, J., Stern, A. D., Opel, N., Reininghaus, U. & Torous, J., "FDA-authorized software as a medical device in mental health: a perspective on evidence, device lineage, and regulatory challenges," npj Mental Health Research 4, 62 (2025), accessed 2026-02-19 go to source ↗
Working notes
Notes made while writing and checking this brief, shown as written. They use working shorthand.
This brief draws on a 2025 Nature npj Mental Health Research study of FDA-authorized SaMD, a 2023 Frontiers in Digital Health regulatory analysis, FDA guidance updates from January 2026, and Latham & Watkins regulatory commentary. Related to health-digital-therapeutics-outcome-measurement (addresses measurement rather than pathway) and health-510k-predicate-creep (shared concern about 510(k) evidence standards for software). Possible overlap with health-digital-therapeutics-reimbursement-gap — this brief focuses on the FDA regulatory pathway problem while the existing brief addresses the payer/reimbursement gap. Tagged `domain:digital` in addition to `domain:health` because the software development lifecycle is central to the problem. Tagged `tractability:research-contribution` because the pathway design question requires policy research rather than prototyping.
Reconciliation 2026-09-30: The whole write-up was read against the four sources it names. (a) The npj Mental Health Research article on the Source line was fetched in full on 2026-09-30 as the publisher's PDF from https://www.nature.com/articles/s44184-025-00174-2.pdf, with its citation record from https://api.crossref.org/works/10.1038/s44184-025-00174-2. (b) The "2023 Frontiers in Digital Health regulatory analysis" is named above without a title or link; it was read as Watson, Chapman, Shafai and Maricich, "FDA regulations and prescription digital therapeutics: Evolving with the technologies they regulate," Frontiers in Digital Health 5, 1086219 (2023), which the npj article cites and which is the only paper matching that description found; full text fetched 2026-09-30 from Europe PMC (PMC10150093). (c) The Latham & Watkins commentary was read in full at https://www.lw.com/en/insights/fda-issues-updated-guidance-loosening-regulatory-approach-to-certain-digital-health-tools (client alert dated January 15, 2026). (d) The FDA's own pages for the two guidance documents, and its list of withdrawn guidance at https://www.fda.gov/medical-devices/guidance-documents-medical-devices-and-radiation-emitting-products/withdrawn-or-expired-guidance, were fetched the same day; the second reader also opened the two guidance PDFs (General Wellness issued January 6, 2026; Clinical Decision Support issued January 29, 2026, superseding a January 6, 2026 version). Changes: (1) The Source line gave the journal as "Nature npj Mental Health Research" and the title as "FDA-authorized SaMD in mental health." The record gives the title "FDA-authorized software as a medical device in mental health: a perspective on evidence, device lineage, and regulatory challenges," by Herpertz, Stern, Opel, Reininghaus and Torous, npj Mental Health Research, volume 4, article 62, 2025. The Source line now carries those details copied from the record; the link is unchanged. (2) The text said "In January 2025" the FDA updated its Clinical Decision Support and general wellness guidance and withdrew its SaMD clinical evaluation guidance. The FDA pages date both updated guidances January 2026, the Latham & Watkins alert gives January 6, 2026, and the FDA's withdrawn-guidance list gives a withdrawal date of 01/06/2026. The body and the first paragraph of these notes now say January 2026. The phrase "signaling a shift in strategy" is in none of the sources and was removed. (3) The text said "The FDA has found that several authorized PDTs were tested on prototype platforms." That finding is the npj authors', not the FDA's, and the count is four; the sentence now says so, and gives the authors' argument that the platform tested should match the one offered to patients in place of "undermining the clinical evidence basis." (4) "Projected to reach $13 to $15 billion by 2028" is in no source. The Frontiers paper cites a January 2022 Research and Markets analysis that valued the 2021 global market at $3.35 billion and estimated $12.1 billion by 2026; the sentence now gives those figures. (5) "Hundreds of products in development" and diabetes are in no source; the sentence now gives the npj review's count of five De Novo and 14 510(k) mental health software devices and the conditions it names, including the virtual-reality devices for lower back pain. (6) "Regulatory pathway misalignment increases development costs, extends time-to-market, and creates uncertainty that deters investment" is in no source. It was replaced by a consequence the npj review does report: OptumRx's announcement that it would deny coverage for computer-based CBT devices. (7) The failure of Pear Therapeutics in 2023, and the reason given for it, are in none of the sources (the Frontiers paper's authors give Pear Therapeutics affiliations, and the Europe PMC record dates its publication 17 April 2023). The sentence now says only what the sources say: that Pear's reSET, classified in September 2017, was the first FDA-authorized PDT to make treatment claims. (8) The sentence on the FDA's Breakthrough Device Designation is in no source and was removed; in its place is the Software Precertification Pilot Program, which both papers describe. (9) "Reducing regulatory burden also risks reducing evidentiary standards" is in no source; it was replaced by what the Latham & Watkins alert says about the scope of the updates. (10) Three phrases were brought to the strength of the sources: the pathways "assume a static, manufactured product" now quotes the Frontiers paper's wording; "small UI/UX changes can dramatically alter therapeutic engagement" now follows the npj review's statement that modifying design or delivery platform can significantly affect engagement, and "digital health devices" became "digital mental health devices," which is what the review covers; "extremely difficult" became "difficult"; and "it was not designed for software products" now reports the FDA's own acknowledgment as the review gives it. (11) The statement that aviation and automotive software updates "are managed through structured change-impact assessment rather than full re-certification" is in no source and was cut back to the suggestion alone; a sentence on Germany and on the FDA's call for legislation, both from the npj review, was added. Confirmed in the sources and left as written: reSET and EndeavorRx as De Novo products; many 510(k)-cleared devices relying on predicates without direct evidence of effectiveness; the placebo-control difficulty and that full blinding is rarely achievable; that no dedicated pathway for software treatments exists; the more permissive direction of the January 2026 guidance updates; and the definition of PDTs as delivering "evidence-based therapeutic treatments," which the Frontiers paper's abstract supports ("access to evidence-based therapies") and which a second reader restored after the first reader had replaced it. A second reader went over all of these repairs against the same sources on the same day. Not checked: the long title and display title use "hardware," a framing word the sources do not use (they say "traditional medical devices"); titles were left unchanged. The regulators listed in the entry points (EMA, PMDA, TGA) are suggestions and are not discussed in the sources. Genome tags untouched.