Classification of Digital Therapeutics: A Practical Guide to Determining Your FDA Regulatory Pathway

Digital therapeutics (DTx) occupy an unusually complex corner of FDA jurisdiction. A software product that treats major depressive disorder, manages opioid use disorder, or delivers cognitive behavioral therapy for insomnia is genuinely novel — and the classification system FDA uses was built primarily around hardware devices. Understanding how that system applies to software-based treatments is the single most consequential regulatory decision you will make early in product development. Get it right, and you select the correct premarket pathway, build the right technical file, and allocate your timeline and budget accurately. Get it wrong, and you may spend 18 months pursuing a 510(k) clearance only to receive a letter from FDA explaining that your product requires a De Novo request — or, worse, a Premarket Approval (PMA). --- ## 1. What Classification Is and Why It Determines Everything Downstream Under 21 CFR Part 860, FDA assigns every medical device to one of three classes based on the level of regulatory control necessary to provide reasonable assurance of safety and effectiveness. * **Class I** — General controls only (labeling, registration, adverse event reporting). Most Class I devices are exempt from premarket notification. Risk is low. * **Class II** — General controls plus special controls (performance standards, post-market surveillance, patient registries). The standard premarket pathway is a **510(k)**, which requires demonstrating **substantial equivalence** — meaning your device is at least as safe and effective as a legally marketed **predicate device** (one already cleared or classified by FDA). * **Class III** — General controls plus PMA. These are devices that support or sustain human life, are of substantial importance in preventing impairment, or present unreasonable risk of illness or injury. PMA requires valid scientific evidence — typically clinical data — demonstrating reasonable assurance of safety and effectiveness. For digital therapeutics, classification is especially consequential because: 1. It determines whether you need clinical evidence before market entry or after. 2. It defines what special controls apply to your software specifically. 3. It establishes post-market surveillance obligations that affect your quality system design from day one. FDA's classification decisions for DTx products are not always intuitive. The same underlying technology — adaptive cognitive training — can be Class II when used to improve attention in ADHD and potentially Class III if the same algorithm is repurposed to treat acute psychosis without a predicate. --- ## 2. How FDA Has Actually Classified Digital Therapeutics ### The Product Code Framework FDA's product classification database assigns a **Product Code** (a three-letter identifier) to device types. For digital therapeutics, several product codes are directly relevant. | Product Code | Device Name | Class | Regulation | |---|---|---|---| | **QZG** | Prescription Software for Substance Use Disorder Treatment | II | 21 CFR 882.5801 | | **PWE** | Prescription Digital Therapeutic for Psychiatric Disorders | II | 21 CFR 882.5860 | | **QMD** | Software for Behavioral Therapy for Substance Use | II | De Novo classified | | **OZO** | Neurological Disorder Digital Therapeutic | II | De Novo classified | These codes did not exist before 2017. Each was created either through a **De Novo request** — FDA's pathway for novel, low-to-moderate risk devices with no predicate — or through subsequent 510(k) clearances that established the product type. ### Landmark Classifications You Should Know **EndeavorRx (Akili Interactive, 2020)** — The first prescription video game authorized by FDA, cleared via De Novo for attention-deficit/hyperactivity disorder (ADHD) in pediatric patients. FDA established Product Code **QMD** through this authorization and issued a special controls document specifying requirements including clinical performance testing, a patient decision checklist, and training materials for prescribers. If you are developing a DTx for pediatric cognitive or attentional disorders today, QMD is your most likely predicate product code. **reSET and reSET-O (Pear Therapeutics, 2017/2018)** — reSET received De Novo authorization as a prescription digital therapeutic for substance use disorder (SUD), creating Product Code **QZG** under 21 CFR 882.5801. reSET-O, targeting opioid use disorder specifically, was subsequently cleared via 510(k) using reSET as its predicate. This sequence illustrates the standard DTx commercialization path: one company absorbs the cost and complexity of De Novo to establish a new device type, and subsequent developers can use 510(k) with a lower evidentiary bar. **Somryst (Pear Therapeutics, 2020)** — A prescription DTx for chronic insomnia delivering cognitive behavioral therapy for insomnia (CBT-I). Cleared via De Novo, it established Product Code **OZO** under 21 CFR 882.5860. The special controls document requires that the device demonstrate clinical performance data showing improvement on the Insomnia Severity Index (ISI) and Patient Health Questionnaire-9 (PHQ-9) screening to rule out major depression as a confounding factor. These examples are not historical trivia. They define the specific technical and clinical requirements your device must meet if you intend to submit a 510(k) citing one of them as a predicate. ### Where DTx Most Commonly Lands by Indication * **Mental and behavioral health (anxiety, depression adjunct, PTSD):** Typically Class II via De Novo or 510(k) when a valid predicate exists. Watch for indications that shift the device from adjunct treatment to standalone treatment — standalone claims for moderate-to-severe depression without a predicate may require PMA-level evidence. * **Substance use disorder:** Class II, 21 CFR 882.5801, predicate available via reSET/reSET-O lineage. * **Chronic pain, musculoskeletal rehabilitation:** Often Class II if claims are limited to symptom management rather than disease modification; carefully scope your intended use. * **Neurological conditions (stroke rehabilitation, Parkinson's):** Variable. Adjunct cognitive rehabilitation tools may qualify as Class II; closed-loop neuromodulation software with autonomous control signals will likely face Class III scrutiny regardless of software-only architecture. --- ## 3. How to Determine Your Device's Class When the Answer Isn't Obvious When you cannot find a directly matching product code, use the following structured approach. ### Step 1: Define Your Intended Use and Indications for Use Precisely These two terms have distinct regulatory meanings. **Intended use** is the general purpose of your device (e.g., "to deliver cognitive behavioral therapy"). **Indications for use** specifies the patient population and clinical context (e.g., "adjunctive treatment of adults aged 22 and older with major depressive disorder receiving antidepressant therapy"). Small changes in indications language can move you between classes. "Adjunctive to antidepressant therapy" is fundamentally different from "as a standalone treatment." Write these statements before you search the classification database — the language you use will determine which predicate candidates are actually comparable. ### Step 2: Search 510(k) Clearances and De Novo Authorizations Use FDA's **510(k) Premarket Notification database** and the **De Novo database** at FDA.gov. Search by product code, device name, and applicant. For DTx specifically, also search the terms "prescription digital therapeutic," "cognitive behavioral therapy software," and your specific indication. Review the **Decision Summary** documents for any clearances you identify. These documents describe what FDA found substantially equivalent and what special controls apply. They are the clearest window into what FDA expects technically and clinically from devices in that product type. ### Step 3: Apply the Three-Part Substantial Equivalence Test Before committing to a 510(k) strategy, confirm that: 1. Your device has the **same intended use** as the predicate (or differences do not raise new safety or effectiveness questions). 2. Your device has the **same or different technological characteristics** — if different, those differences must not raise new questions of safety and effectiveness. 3. Performance data demonstrates that your device is **at least as safe and effective** as the predicate. If you cannot satisfy all three parts, 510(k) is not available. Your options become De Novo (if the device is Class II-appropriate but novel) or PMA (if it is genuinely Class III). ### Step 4: Evaluate SaMD Risk Category Using IMDRF Framework FDA's **Software as a Medical Device (SaMD): Clinical Evaluation** guidance (2017) incorporates the **International Medical Device Regulators Forum (IMDRF)** risk categorization framework. It evaluates software risk on two axes: * **State of healthcare situation:** Critical (life-threatening), serious, or non-serious. * **Significance of information provided by SaMD:** Treat or diagnose vs. drive clinical management vs. inform clinical management. DTx almost always sits in the "treat or diagnose" column. If the condition being treated is "serious" (significant irreversible health consequences without appropriate treatment), you are in IMDRF Category III — which corresponds directionally to Class II with robust special controls or potentially Class III. Use this framework as a cross-check, not a substitute for formal classification. ### Step 5: Request a Pre-Submission Meeting If classification remains ambiguous after Steps 1–4, submit a **Pre-Submission (Q-Sub)** meeting request under FDA's **Formal Meetings Between FDA and Sponsors of Medical Devices** guidance (2021). In the Q-Sub, present your intended use, proposed predicate (if any), and your classification rationale. Ask FDA directly whether they agree with your classification and pathway. Q-Subs are not binding on FDA, but they provide documented feedback that substantially de-risks a formal submission. Plan for approximately 90 days to receive written feedback. --- ## 4. Worked Example: Classifying a DTx for Generalized Anxiety Disorder Assume you are developing a smartphone application that delivers an eight-week structured program of exposure-based cognitive behavioral therapy (CBT) for adults with mild-to-moderate generalized anxiety disorder (GAD), intended as an adjunct to pharmacotherapy under clinician supervision. **Step 1 — Intended use and indications:** Adjunctive software-delivered CBT for adults (18+) with mild-to-moderate GAD receiving pharmacotherapy. Not standalone; not intended for severe GAD or panic disorder with agoraphobia. **Step 2 — Database search:** Searching the De Novo database for "anxiety" and "cognitive behavioral therapy" returns limited directly comparable authorizations as of April 2026. The Somryst (OZO) predicate applies to insomnia, not anxiety. No exact product code covers adjunctive CBT for GAD specifically. **Step 3 — Substantial equivalence analysis:** No predicate with the same indications for use currently exists in the cleared device database. A 510(k) is not available. The device is novel. **Step 4 — IMDRF risk category:** State of healthcare: GAD is serious (significant functional impairment, risk of escalation). Significance: treat. This places the device in IMDRF Category III — high-risk SaMD, but not necessarily Class III under U.S. law. Class II with adequate special controls is the appropriate target, consistent with similar mental health DTx classifications. **Pathway conclusion:** De Novo request under 21 CFR 860.257. You will need to propose special controls, which typically include: clinical performance testing demonstrating statistically significant improvement on a validated anxiety measure (e.g., GAD-7), a clinician training requirement, labeling specifying intended patient population and contraindications (severe GAD, active suicidality), and post-market data collection provisions. **Clinical data requirement:** FDA will expect at least one well-designed randomized controlled trial supporting the De Novo. Review the Somryst and EndeavorRx decision summaries for the clinical endpoints and study design elements FDA found sufficient in structurally similar De Novo packages. --- ## Key Takeaways * **Classification is not administrative — it is strategic.** The device class you land in determines your premarket evidence requirements, timeline, and cost structure. Determine classification before you finalize your clinical development plan. * **The De Novo pathway has created a growing predicate library for DTx.** Before assuming you need a De Novo, thoroughly review existing clearances for reSET (QZG), EndeavorRx (QMD), and Somryst (OZO) — and their subsequent 510(k) clearances — to identify whether a predicate already exists for your indication. * **Indications language drives classification outcomes more than technology.** "Adjunctive treatment" versus "standalone treatment," and "mild-to-moderate" versus "moderate-to-severe," are not just clinical distinctions — they can shift your device between pathway types entirely. * **Use the Q-Sub process when classification is genuinely ambiguous.** A 90-day investment in a Pre-Submission meeting is far less costly than 18 months on the wrong pathway. Document FDA's feedback and incorporate it explicitly into your formal submission. * **Review De Novo Decision Summaries — not just the authorization letter.** The Decision Summary documents for DTx authorizations contain the specific clinical endpoints, study design parameters, and labeling requirements FDA applied. They are the closest thing to a roadmap for your own submission.