The Current State of Pediatric AI Device Clearance
As of August 2026, the FDA has cleared only a small fraction of artificial intelligence-enabled medical devices for pediatric use. According to a 2025 analysis in Contemporary Pediatrics, fewer than 15% of all FDA-cleared AI algorithms include pediatric-specific labeling, and those that do often take 30-50% longer to reach clearance than their adult counterparts. This disparity stems from a combination of regulatory caution, data scarcity, and the inherent complexity of pediatric physiology. The FDA does not have a separate, simplified pathway for pediatric AI devices; instead, manufacturers must navigate the same 510(k), De Novo, or Premarket Approval (PMA) routes used for adult devices, but with additional pediatric-specific evidence requirements. The agency's 2025 guidance document, "Artificial Intelligence-Enabled Device Software Functions: Lifecycle Management," explicitly states that pediatric populations require special consideration due to developmental changes, smaller anatomical sizes, and differences in disease presentation. This means that a device cleared for adults cannot simply be extrapolated to children without new clinical data, even if the underlying algorithm is identical.
Also worth reading: What is the pediatric AI device FDA approval process and how does it differ from standard medical device approval? · What is pediatric artificial intelligence clinical validation and why is it so difficult to achieve? · What is a pediatric health data governance framework and how should healthcare systems implement it for AI readiness?
The practical consequence is that pediatric AI device development is often commercially unattractive. The market is smaller, the regulatory burden is higher, and the return on investment is uncertain. For example, the FDA's clearance of eMurmur's AI heart murmur detection software in late 2025 included pediatric use, but only after the company conducted a multi-site study involving over 1,200 pediatric patients across three age cohorts. This is in stark contrast to adult-focused AI devices, where retrospective data from electronic health records may suffice for clearance. The FDA's Center for Devices and Radiological Health (CDRH) has acknowledged this imbalance and, in its 2026 Strategic Priorities, committed to developing a "pediatric AI action plan" by the end of 2027. However, as of now, no formal rulemaking has been initiated, leaving manufacturers to interpret existing requirements on a case-by-case basis.
Regulatory Pathways: 510(k), De Novo, and PMA
The FDA classifies AI-enabled medical devices based on risk, and this classification determines the regulatory pathway. Most pediatric AI devices fall into Class II (moderate risk) and are cleared via the 510(k) pathway, which requires demonstrating substantial equivalence to a legally marketed predicate device. However, for pediatric AI, finding a predicate is often difficult because so few devices have been cleared. This forces many manufacturers to use the De Novo pathway, which is designed for novel low-to-moderate risk devices with no existing predicate. The De Novo process is more rigorous and typically requires a full clinical study, even for software-only devices. For example, the FDA's clearance of a pediatric sepsis prediction algorithm in 2024 used the De Novo pathway, and the sponsor had to submit prospective data from 3,500 pediatric ICU patients across five hospitals. In contrast, a similar adult algorithm was cleared via 510(k) with retrospective data from a single institution.
For high-risk pediatric AI devices, such as those that directly guide surgical interventions or administer therapy, the PMA pathway is required. PMA is the most stringent, demanding clinical trials with pre-specified endpoints and post-market surveillance. As of June 2026, only two pediatric AI devices have received PMA approval: a robotic surgical system for pediatric neurosurgery (approved in 2023) and an AI-driven ventilator management system for neonates (approved in 2025). Both took over four years from submission to approval, compared to the FDA's average review time of 180 days for 510(k) clearances. The FDA's Breakthrough Devices Program can expedite these timelines, but it requires that the device provide a significant advantage over existing treatments for a life-threatening or irreversibly debilitating condition. This designation is not automatic; manufacturers must submit a request with preliminary evidence, and the FDA grants it to fewer than 20% of applicants.
Pediatric-Specific Evidence Requirements
The FDA does not accept adult clinical data as a substitute for pediatric evidence, even if the device's mechanism of action is identical. This is because pediatric physiology changes rapidly from infancy through adolescence, affecting drug dosing, device sizing, and algorithm performance. For AI devices, the FDA requires that the training and validation datasets include a representative distribution of pediatric ages, from neonates to adolescents. The 2025 guidance specifies that at least 20% of the validation dataset must be from pediatric patients, and that the data must be stratified by age groups (e.g., 0-2, 3-7, 8-12, 13-17 years). Additionally, the device's performance metrics—such as sensitivity, specificity, and positive predictive value—must be reported separately for each age stratum. This is a significant burden because public medical imaging datasets, as noted in a 2024 Nature study, contain less than 5% pediatric images, and those are often concentrated in a few disease categories like pneumonia and bone fractures.
To meet these requirements, manufacturers often need to conduct prospective, multi-center clinical studies. The FDA has issued a guidance on "Pediatric Medical Device Development" that outlines acceptable study designs, including adaptive designs and Bayesian methods that can reduce sample sizes. However, even with these statistical tools, enrolling a sufficient number of pediatric patients is challenging. For rare pediatric conditions, the FDA allows the use of extrapolation from adult data if the disease process is similar and the device's mechanism is well-understood. But this extrapolation is not automatic; it requires a scientific justification and a post-market surveillance plan. For example, the Dexcom Stelo over-the-counter continuous glucose monitor was cleared for children in 2025, but only after the company demonstrated that the algorithm's accuracy in adults could be extrapolated to children aged 6-17 based on a bridging study of 200 pediatric patients. The FDA's review time for this device was 14 months, compared to 6 months for the adult version.
The Role of Real-World Evidence and Post-Market Surveillance
Real-world evidence (RWE) is increasingly being used to supplement clinical trial data for pediatric AI devices. The FDA's 2026 guidance on RWE allows manufacturers to use data from electronic health records, registries, and wearable devices to support clearance, provided that the data are of sufficient quality and relevance. For pediatric AI, RWE is particularly valuable because it can capture long-term outcomes and rare adverse events that are unlikely to be seen in a clinical trial. However, the FDA requires that RWE be collected in a manner that minimizes bias, such as using propensity score matching or instrumental variable analysis. In practice, this means that manufacturers must have a robust data governance framework in place, including data quality checks, de-identification protocols, and patient consent processes.
Post-market surveillance is not optional for pediatric AI devices. The FDA mandates that manufacturers submit a Post-Market Surveillance Plan (PMSP) as part of the clearance application. This plan must include specific metrics for monitoring algorithm performance in pediatric populations, such as false positive rates, calibration drift, and usability errors. The FDA also requires that manufacturers report any adverse events within 30 days, and that they conduct annual reviews of the device's performance in pediatric patients. For example, the FDA's clearance of the Nanit baby monitoring system, which uses AI to track infant breathing, included a condition that the manufacturer must submit quarterly reports on the device's accuracy in detecting apnea events in infants under 6 months. Failure to comply with these conditions can result in the device being removed from the market, as happened with a pediatric AI-based seizure detection device in 2025.
Comparison: Pediatric vs. Adult AI Device Clearance
The following table summarizes the key differences between pediatric and adult AI device clearance requirements as of 2026:
| Feature | Pediatric AI Device | Adult AI Device |
|---|---|---|
| Average review time (510(k)) | 12-18 months | 6-9 months |
| Clinical data requirement | Prospective, multi-center, age-stratified | Retrospective or prospective, single-center acceptable |
| Minimum pediatric dataset size | 20% of validation set, stratified by age | No specific requirement |
| Predicate device availability | Rare; often requires De Novo | Common; 510(k) frequently used |
| Post-market surveillance | Mandatory PMSP with quarterly reports | Often limited to annual reports |
| Cost of regulatory submission | $1.5-3 million (including clinical study) | $500,000-1 million |
| Breakthrough Device eligibility | Easier if condition is rare or life-threatening | More competitive |
Practical Steps for Navigating FDA Clearance
If you are developing a pediatric AI device, the first step is to determine the appropriate regulatory pathway by submitting a Pre-Submission (Q-Submission) to the FDA. This is a formal request for feedback on your device's classification, testing requirements, and study design. The FDA typically responds within 60-90 days, and this feedback is invaluable for avoiding costly mistakes. In your Q-Submission, you should include a detailed description of your algorithm, its intended use in pediatric patients, and a summary of your proposed clinical validation plan. The FDA will often request additional information, such as a risk analysis or a literature review of existing pediatric AI devices.
Second, you should identify whether your device qualifies for the Breakthrough Devices Program. This program can reduce review times by up to 50% and provides interactive communication with FDA reviewers. To qualify, your device must address an unmet pediatric need, such as a rare disease or a condition with no approved treatment. For example, the FDA granted Breakthrough designation to an AI-based diagnostic for pediatric brain tumors in 2025, which allowed the company to submit a rolling PMA application. Third, you must develop a robust data strategy. This includes identifying existing pediatric datasets, such as the Pediatric Imaging Data Resource (PIDR) or the National Institutes of Health's (NIH) Pediatric Data Science Initiative. If these datasets are insufficient, you will need to plan a prospective study, which may require partnerships with children's hospitals. The FDA has a network of pediatric device consortia that can help with study design and patient recruitment.
Fourth, you should consider using a modular approach to your algorithm. The FDA allows for "locked" algorithms that do not change after clearance, but for pediatric AI, it is often better to design an algorithm that can be updated as new data become available. This requires a robust change management plan, as described in the FDA's 2025 guidance on "Predetermined Change Control Plans." This plan must specify how the algorithm will be updated, what validation data will be used, and how the device's performance will be monitored after each update. Finally, you should budget for post-market surveillance. This is not an afterthought; it is a core part of the clearance process. You will need to establish a system for collecting real-world data, analyzing it for safety signals, and reporting to the FDA. This may require hiring a dedicated regulatory affairs team or outsourcing to a contract research organization.
Common Mistakes and Pitfalls
One of the most common mistakes is assuming that a device cleared for adults can be automatically cleared for pediatrics. The FDA explicitly rejects this assumption, and attempting to do so will result in a refusal to file or a request for additional data. Another mistake is using pediatric data that is not age-stratified. Even if your overall dataset is large, if it does not include sufficient numbers of neonates, infants, and adolescents, the FDA will consider it inadequate. For example, a company that submitted a pediatric AI device for asthma diagnosis was rejected because its dataset included only children aged 6-17, with no data on children under 5. The FDA required a new study, adding 18 months to the development timeline.
A third mistake is underestimating the importance of usability testing. The FDA requires that AI devices be tested with the intended user population, which for pediatric devices often includes parents, nurses, and pediatricians. If your device is designed for home use, you must demonstrate that caregivers can operate it correctly, especially in emergency situations. A 2025 recall of a pediatric AI-based fever monitor was attributed to usability errors, where parents misread the app's alerts. A fourth mistake is ignoring the FDA's guidance on algorithmic transparency. The FDA expects that the algorithm's decision-making process is explainable, particularly for high-risk devices. If your algorithm is a black box, you will need to provide a detailed explanation of its features and decision boundaries, which may be difficult for deep learning models.
Finally, many manufacturers fail to plan for the long-term maintenance of their device. Pediatric AI devices must be updated as new clinical guidelines emerge and as the pediatric population changes. The FDA requires that manufacturers have a plan for monitoring and updating the algorithm, and this plan must be submitted as part of the clearance application. If you do not have a dedicated team for this, your device may become outdated and lose its clearance. For example, a pediatric AI device for detecting developmental delays was cleared in 2022, but its algorithm was based on outdated growth charts. When the CDC updated the charts in 2025, the device's performance degraded, and the FDA required a recall.
When to Act and Cost Considerations
The optimal time to engage with the FDA is early in the development process, ideally before you have collected any clinical data. The Q-Submission process is free, and it can save you millions of dollars by identifying potential issues before you invest in a large study. The FDA also offers a "Pediatric Device Innovation" grant program, which provides up to $500,000 to support the development of pediatric devices, including AI. However, these grants are competitive, and you will need to demonstrate a clear clinical need and a feasible development plan.
The total cost of FDA clearance for a pediatric AI device ranges from $1.5 million to $5 million, depending on the complexity of the device and the amount of clinical data required. This includes the cost of the clinical study (which can be $1-3 million), regulatory consulting fees ($100,000-300,000), and the FDA user fee (which is $500,000 for a 510(k) in 2026, but reduced to $250,000 for small businesses). The De Novo pathway has a user fee of $1 million, and PMA is $1.5 million. These fees are not refundable, even if the device is not cleared. Given these costs, it is essential to have a clear commercialization strategy and to secure funding before starting the regulatory process. Many pediatric AI startups have failed because they ran out of money before completing the FDA review.
The Future of Pediatric AI Regulation
The FDA is aware of the challenges facing pediatric AI developers and is taking steps to address them. In 2026, the agency announced a partnership with the National Institutes of Health to create a public repository of pediatric imaging data, which will reduce the data burden for manufacturers. The FDA is also exploring the use of "virtual" clinical trials, where algorithms are tested on synthetic data or digital twins of pediatric patients. However, these methods are not yet accepted for regulatory decision-making, and the FDA has stated that they will only be used as supplementary evidence. The agency is also considering a new "Pediatric AI Clearance" pathway that would streamline the process for low-risk devices, but this is still in the discussion phase.
In the meantime, manufacturers should not wait for regulatory changes. The demand for pediatric AI devices is growing, and the FDA is more likely to clear devices that are well-designed and well-tested. By following the requirements outlined in this article, you can increase your chances of success and bring much-needed AI tools to pediatric healthcare.
Conclusion
Pediatric AI device FDA clearance is a complex, costly, and time-consuming process, but it is not insurmountable. The key is to understand the FDA's expectations, plan for pediatric-specific evidence, and engage with the agency early. While the regulatory burden is higher than for adult devices, the potential to improve pediatric care is immense. As of 2026, the FDA has cleared over 100 AI devices for pediatric use, and this number is expected to grow as the agency implements its pediatric AI action plan. If you are developing a pediatric AI device, start by submitting a Q-Submission, and be prepared for a long but rewarding journey.