Simulating the FDA De Novo classification pathway for a novel-risk-profile digital therapeutic with no existing predicate device
When a digital therapeutic has no legally marketed predicate, the sponsor cannot use the 510(k) substantial-equivalence pathway. The De Novo classification pathway (created by the FDA Modernization Act of 1997, streamlined by the 21st Century Cures Act) is the route for novel low-to-moderate risk devices. Before committing years of clinical work, sponsors almost always request a Q-Submission (Q-Sub) pre-submission meeting to pressure-test their regulatory strategy against FDA's Digital Health Center of Excellence reviewers.
A 510(k) clearance requires demonstrating "substantial equivalence" to an already-legally-marketed predicate device with the same intended use and technological characteristics. A genuinely novel digital therapeutic — say, a prescription video game that treats pediatric ADHD by training attention networks — has no predicate. Two options exist: PMA (Premarket Approval, the most burdensome pathway, reserved for high-risk Class III devices) or De Novo, which is designed for novel devices where general and special controls can provide reasonable assurance of safety and effectiveness without the full PMA burden.
The De Novo pathway, formalized under 21 CFR 860.220 and streamlined by Section 3038 of the 21st Century Cures Act (2016) to allow direct De Novo filing without a prior 510(k) Not Substantially Equivalent (NSE) determination, has become the primary route for first-in-class digital therapeutics, novel AI/ML diagnostic software, and connected sensor platforms.
A Q-Submission is a formal mechanism for sponsors to obtain FDA feedback before a marketing application is filed. For DTx sponsors, a Pre-Submission (Pre-Sub) meeting typically addresses:
• Whether De Novo is the appropriate pathway given the device's risk profile and absence of predicates • Adequacy of the proposed pivotal trial design — primary endpoint selection, control arm design (often a "digital placebo" — a visually similar but therapeutically inert app), sample size justification • Human factors / usability engineering plan (IEC 62366-1) given the device is patient-operated software • Cybersecurity documentation expectations under the FDA's 2023 premarket cybersecurity guidance • Whether the sponsor qualifies for Breakthrough Device Designation, which grants priority review and more frequent interactive engagement with FDA staff
FDA aims to provide written feedback within roughly 70-75 calendar days of Q-Sub receipt, and feedback is strictly non-binding — the agency can and does change its position at formal review — but a well-run Q-Sub dramatically de-risks trial design, since restarting a pivotal RCT after an FDA rejection can cost 18-24 months and tens of millions of dollars.
Akili Interactive ran multiple Q-Sub meetings with FDA's Digital Health Center of Excellence before finalizing the STARS-ADHD pivotal trial design for what became EndeavorRx — aligning the primary endpoint (TOVA Attention Performance Index) with FDA expectations up front avoided a costly redesign.
Unlike a drug, a digital therapeutic's "active ingredient" is an interactive software mechanism — game mechanics, cognitive-behavioral content modules, adaptive difficulty algorithms — that must be tested with the same rigor as a pharmacologic intervention: randomization, blinding where feasible, and a pre-specified primary endpoint powered to detect a clinically meaningful effect.
Akili Interactive's EndeavorRx (AKL-T01) received FDA De Novo authorization in June 2020 for children ages 8-12 with primarily inattentive or combined-type ADHD. The pivotal STARS-ADHD trial (2020, published in The Lancet Digital Health) randomized 348 children to either the active game intervention or a digital control app for four weeks. The prespecified primary endpoint was change in TOVA (Test of Variables of Attention) Attention Performance Index, an objective, validated computer-administered attention measure — not a subjective parent questionnaire, which strengthened the evidentiary claim.
A follow-on trial, STARS-2, replicated the effect in a broader population and supported label claims around improvement in objectively measured attention function. Critically, EndeavorRx was authorized as an adjunct to, not replacement for, a clinician-managed ADHD treatment plan — the label explicitly scopes the claim to attention function improvement rather than broader symptom remission.
Pear Therapeutics' reSET, granted De Novo authorization in September 2017, was the first prescription digital therapeutic device authorized by FDA for treating a disease — substance use disorder (SUD) as an adjunct to outpatient treatment. Its pivotal trial enrolled patients receiving standard outpatient SUD treatment, randomized to treatment-as-usual versus treatment-as-usual plus the reSET cognitive-behavioral-therapy software module. The primary endpoint was abstinence, measured by urine toxicology and self-report over 12 weeks; the reSET arm showed a statistically significant increase in abstinence rate versus control.
Both EndeavorRx and reSET illustrate the standard DTx evidentiary template: a randomized, controlled, objectively measured, prespecified-endpoint trial embedded in real clinical care pathways rather than a lab setting — evidence that closely mirrors what FDA expects of a Class II therapeutic drug-device combination, scaled down proportionate to the device's risk classification.
Common pitfalls FDA reviewers flag at Q-Sub and formal review:
• Subjective-only endpoints (patient-reported symptom scales) without an objective corroborating measure • Underpowered trials that cannot distinguish true effect from digital-placebo response — digital interventions show unusually large placebo effects (often 20-30% symptom improvement in control arms) because novel, engaging interfaces themselves carry non-specific benefit • Missing engagement/dose-response analysis — FDA increasingly expects sponsors to show adherence correlates with outcome, evidence the therapeutic mechanism (not mere study participation) drives benefit • Real-world evidence (RWE) supplementing but not replacing the pivotal RCT — post-market registries and payer claims data are used to support durability claims after initial authorization
Once the formal De Novo request is filed, it enters FDA's Center for Devices and Radiological Health (CDRH) queue. The Medical Device User Fee Amendments (MDUFA V) performance goals target roughly 150 FDA-review days to a decision — but the clock stops every time FDA issues an Additional Information (AI) request and restarts only once the sponsor responds, so wall-clock time to authorization is often considerably longer.
The De Novo review proceeds through several checkpoints:
1. Acceptance Review (within 15 days): FDA checks the submission against the Acceptance Checklist — is it administratively complete? Missing elements (e.g., a required cybersecurity risk assessment, incomplete human factors report) trigger a Refuse-to-Accept (RTA) decision that halts the clock entirely until resubmission — RTA affects a meaningful share of first-time De Novo filings and is the single most common source of avoidable delay.
2. Substantive Review: once accepted, a multidisciplinary team (biomedical engineers, clinical reviewers, statisticians, human factors specialists, and for DTx specifically the Digital Health Center of Excellence) evaluates the classification request against the statutory De Novo criteria — the device presents low-to-moderate risk and general controls, or general and special controls, provide reasonable assurance of safety and effectiveness.
3. Interactive Review / Additional Information (AI) requests: FDA issues written AI requests when data gaps or ambiguities remain — a request could ask for reanalysis of a secondary endpoint, additional cybersecurity penetration testing documentation, or clarification of the algorithm's adaptive-difficulty logic. The MDUFA clock pauses ("stops the clock") upon an AI request and resumes only when FDA receives a complete response.
4. Benefit-Risk Determination: reviewers apply FDA's benefit-risk framework (weighing probable benefit against probable risk, patient tolerance for risk, and availability of alternative treatments) — for a Class II DTx with a favorable safety profile (software carries no direct physical harm mechanism analogous to an implant), the bar centers heavily on demonstrated clinical benefit magnitude and durability.
FDA's clock-stop convention means a "150-day" review routinely spans 9-15 wall-clock months once AI-request cycles are included — sponsors budget accordingly and often keep parallel workstreams (manufacturing scale-up, commercial launch prep, payer engagement) running during the interactive review period.
Sponsors of DTx addressing serious or life-threatening conditions with no adequate existing alternative can request Breakthrough Device Designation, which grants:
• Priority review — the submission jumps the queue relative to non-designated devices • More frequent, interactive engagement with a assigned FDA review team throughout development, not just at formal review • Sprint discussions to resolve specific technical questions on an accelerated timeline • Flexibility in clinical study design discussions, including potential use of adaptive trial designs or reliance on a broader evidence package including real-world data
Breakthrough designation does not lower the evidentiary bar for safety and effectiveness — it changes the tempo and collaborativeness of the interaction, which for a resource-constrained DTx sponsor often matters as much as the underlying trial data quality.
A De Novo grant is not simply an authorization for one product — it establishes a new device classification regulation and generic device type, along with the special controls that all future devices of that type must satisfy. Drafting these special controls is FDA's way of encoding, in regulation, everything it learned scrutinizing this first-of-kind device.
Special controls are device-type-specific requirements FDA layers on top of general controls (registration, listing, good manufacturing practice / quality system regulation, labeling, adverse event reporting) to mitigate the risks specific to that device category. For a novel prescription digital therapeutic, a typical special controls package includes:
• Clinical performance testing requirements — future devices of this type must demonstrate effectiveness via adequately powered clinical investigation with objective, clinically meaningful endpoints, mirroring what the original sponsor demonstrated • Software verification, validation, and hazard analysis consistent with IEC 62304 software lifecycle processes, including documented risk-based classification of software items • Human factors / usability validation under IEC 62366-1, since the device is operated directly by patients (often children, or individuals with the very cognitive or psychiatric condition being treated) without clinician mediation during use • Cybersecurity controls — threat modeling, a software bill of materials (SBOM), a vulnerability disclosure and patching plan, consistent with FDA's premarket cybersecurity guidance, since these are connected, update-capable software products • Labeling requirements specifying the intended use population, adjunctive-use limitations (e.g., "as an adjunct to clinician-managed treatment," not a replacement), and any algorithm-adaptation limits • Algorithm change control plan — for devices with adaptive or updateable logic, a predetermined change control plan (PCCP) describing what modifications can be made post-market without a new submission
The special controls become the compliance checklist for every subsequent device claiming to be of the same generic type. This is the mechanism by which a De Novo grant creates competitive and regulatory infrastructure: a second company building a similar attention-training game for pediatric ADHD no longer needs its own De Novo request — it can pursue 510(k) clearance by demonstrating substantial equivalence to the newly created predicate and compliance with the same special controls, which is materially faster and cheaper than the original path.
This is precisely why De Novo grants are strategically significant for first movers: EndeavorRx's De Novo grant created the "Adjunctive Game-Based Digital Health Therapeutic Device for use in Patients with ADHD" generic type (21 CFR 882.5801), and reSET's grant similarly created a generic type for substance-use-disorder behavioral DTx (21 CFR 882.5801-class analog for SUD) — subsequent competitors now target 510(k) against these predicates rather than repeating the full De Novo evidentiary burden.
The De Novo grant order is the culmination of the pathway: FDA issues a classification order, publishes the new special controls in the Federal Register / eCFR, and the device — and its generic type — become part of the permanent regulatory landscape. For the sponsor, this converts years of clinical and regulatory investment into a marketable Class II medical device claim and, often, a durable competitive moat.
Upon granting the De Novo request, FDA issues a classification order that:
• Classifies the device into Class I or Class II (De Novo devices are almost never Class III, since Class III implies PMA is required and the whole point of De Novo is to avoid the PMA burden for low/moderate-risk novel devices) • Establishes a new product code and regulation number in Title 21 of the Code of Federal Regulations, with the accompanying special controls now legally binding on all future entrants • Grants the requesting sponsor marketing authorization for their specific device, subject to compliance with the newly created special controls plus general controls • Publishes a public De Novo decision summary describing the review basis — a valuable transparency artifact competitors and clinicians can review
From this point, the authorized device can be legally marketed in the U.S. as a prescription (or, if the special controls and label permit, over-the-counter) medical device, subject to ongoing post-market obligations: adverse event (MDR) reporting, post-market surveillance if ordered, and periodic quality system inspections.
Perhaps the most consequential downstream effect of a De Novo grant is structural, not just commercial: it converts an expensive, novel-risk pathway into a fast, cheap, well-trodden one for everyone who follows. A second-mover DTx company building a conceptually similar product now argues 510(k) substantial equivalence to the newly minted predicate — same intended use, same technological characteristics, no new questions of safety or effectiveness — which the FDA can typically clear in a matter of months rather than the 12-18+ months (plus preceding years of trial work) the original sponsor required.
This "predicate flywheel" is exactly how the broader medical device ecosystem accumulates precedent over decades: roughly 90% of medical devices reach market via 510(k) precisely because so few require an original De Novo or PMA. EndeavorRx's 2020 grant and reSET's 2017 grant each function as category-defining events — every subsequent adjunctive game-based ADHD therapeutic or CBT-based SUD app now has a clearer, faster, and cheaper regulatory map to follow, which is part of why the digital therapeutics sector has been able to scale its total authorized-product count meaningfully faster in the years since these first grants than in the years before them.
The De Novo pathway is deliberately self-limiting in its novelty burden: each grant makes the next comparable product easier to clear, concentrating the hardest, most expensive regulatory lift on true first movers — which is precisely why Q-Sub strategy, endpoint selection, and evidence package rigor in Stages 1-2 disproportionately determine whether a DTx company survives long enough to become the predicate, rather than chasing one.