The Current State of Pediatric AI Approvals: A Reality Check
As of August 2026, the FDA has not established a separate, dedicated regulatory pathway exclusively for pediatric AI devices. Instead, pediatric AI devices are reviewed under the same general frameworks that govern all AI-enabled medical devices—primarily the 510(k) clearance, De Novo classification, and Premarket Approval (PMA) pathways. However, the agency has issued specific guidance and has internal review considerations that apply when a device is intended for pediatric use. The most important document remains the 2021 "Artificial Intelligence/Machine Learning (AI/ML)-Based Software as a Medical Device Action Plan," which outlines a total product lifecycle approach but does not carve out pediatric-specific requirements. In practice, this means that a pediatric AI device must meet the same evidence standards as an adult device, but the FDA expects additional scrutiny on data representativeness, age-specific performance, and long-term safety.
Also worth reading: What are the compliance requirements for AI medical devices designed for pediatric populations? · How does the pediatric medical device software clearance process work under FDA regulations? · How to design a pediatric AI device clinical trial: key considerations and regulatory pathways?
The reality is stark: pediatric AI devices are rare. A study published in RAPS in early 2026 found that among all FDA-cleared AI-enabled medical devices, only about 3% had a pediatric indication or included pediatric patients in their validation cohorts. Radiology dominates the AI device landscape, accounting for over 75% of all FDA AI clearances since 1995, but pediatric-specific algorithms are a tiny fraction of that. For example, in a recent batch of 68 new FDA-cleared radiology algorithms announced in mid-2026, only two included pediatric patients in their training or validation data. This underrepresentation is not accidental—it reflects a combination of scientific, ethical, and economic barriers that make pediatric AI development less attractive to industry.
The FDA's stance, articulated by former Commissioner Scott Gottlieb and echoed in current leadership, is that pediatric devices should not be held to a lower standard of evidence. However, the agency has also acknowledged that the same level of evidence may not be feasible for rare pediatric conditions. In 2025, the FDA introduced the RAPID (Real-World Analytics for Pediatric Innovation and Development) pathway, which was initially designed to accelerate adult device reviews using real-world evidence. As of June 2026, the RAPID pathway has not been extended to pediatric devices, despite advocacy from pediatricians and device manufacturers. This has created a two-tier system where adults benefit from faster, real-world-data-driven approvals, while children remain stuck in traditional clinical trial requirements that are often impractical.
Why Pediatric AI Devices Face Longer FDA Review Times
The FDA review process for pediatric AI devices is not just slower—it is fundamentally more complex. One of the primary reasons is the lack of pediatric training data. A 2024 Nature study highlighted that children are severely underrepresented in public medical imaging datasets, with some datasets containing less than 1% pediatric images. This data scarcity forces developers to either collect new data from scratch, which is expensive and logistically difficult, or to use adult data with the assumption that the algorithm will generalize to children—an assumption that the FDA has repeatedly rejected. The agency requires that any AI device intended for pediatric use be validated on a pediatric-specific dataset, unless the developer can provide a robust scientific justification for why adult data is sufficient. In practice, this justification is rarely accepted, especially for imaging devices where anatomical and physiological differences are significant.
Another factor is the ethical and regulatory requirement for pediatric clinical trials. The FDA's Pediatric Research Equity Act (PREA) and the Best Pharmaceuticals for Children Act (BPCA) apply to drugs, but for devices, the requirements are less codified. However, the FDA still expects that pediatric AI devices undergo clinical validation in the intended age group. This is particularly challenging for AI devices because they often require large, diverse datasets to achieve acceptable performance. For a rare pediatric disease, enrolling enough patients to demonstrate statistical significance can take years. The FDA has tried to mitigate this by allowing extrapolation of adult data in some cases, but for AI algorithms, extrapolation is rarely accepted because the algorithm's behavior is data-driven and may not be predictable across age groups.
Additionally, the FDA has been increasingly concerned about algorithmic bias and fairness. In 2025, the agency issued a draft guidance on "Transparency in AI/ML-Enabled Devices," which requires that developers disclose the demographic composition of their training and validation datasets. For pediatric devices, this means that developers must not only show that the algorithm works in children but also that it works across different pediatric age subgroups (e.g., infants, toddlers, adolescents). This level of granularity is often impossible to achieve with existing data, leading to longer review times and higher rejection rates. A 2026 analysis by Contemporary Pediatrics found that the average time from submission to FDA decision for pediatric AI devices was 14 months, compared to 8 months for adult AI devices. This delay is not due to FDA inefficiency but rather to the iterative back-and-forth required to address data gaps and performance questions.
The 2026 Regulatory Framework: What Has Changed and What Hasn't
As of August 2026, there have been several notable regulatory developments that affect pediatric AI devices, but none have created a dedicated pathway. The FDA's Digital Health Center of Excellence continues to oversee AI device reviews, and in early 2026, the agency released a new guidance titled "Considerations for the Use of Real-World Data in Pediatric Medical Device Development." This guidance clarifies that real-world data (RWD) can be used to support pediatric AI device approvals, but only if the data is of sufficient quality and relevance. The guidance also introduces a new concept called "pediatric data bridging," which allows developers to use adult data to inform pediatric model development, but only if they can demonstrate that the underlying disease or condition has similar pathophysiology across age groups. This is a pragmatic approach, but it has been criticized by some experts who argue that it could lead to unsafe devices if the bridging assumptions are incorrect.
Another important change is the FDA's increased focus on total product lifecycle (TPLC) for AI devices. In 2026, the FDA began requiring that all AI device submissions include a plan for ongoing monitoring and updates, including how the device will be re-evaluated if it is used in a pediatric population. This is particularly relevant for pediatric AI devices because children grow and develop, and an algorithm that works for a 5-year-old may not work for a 10-year-old. The FDA now expects that developers specify the age range for which the device is intended and provide evidence that the algorithm's performance is stable across that range. If a developer wants to expand the age range after initial approval, they must submit a new 510(k) or PMA supplement, which adds time and cost.
Despite these changes, the fundamental challenge remains: the FDA's regulatory framework is designed for devices that are used in a relatively homogeneous adult population. Pediatric AI devices are inherently more complex because they must account for growth, development, and age-specific disease presentations. The FDA has acknowledged this by creating a "Pediatric Device Review Committee" that provides additional scientific and clinical input during the review process. However, this committee does not have the authority to waive evidence requirements; it only offers advice. As a result, developers of pediatric AI devices must navigate a patchwork of requirements that are often unclear and inconsistent. The lack of a clear, predictable pathway is a major deterrent for small companies and startups, which may choose to focus on adult devices instead.
Practical Steps to Achieve FDA Approval for a Pediatric AI Device in 2026
If you are developing a pediatric AI device, the first step is to determine the appropriate regulatory pathway. Most AI devices are classified as Class II and can go through the 510(k) pathway, but if your device is truly novel and has no predicate, you will need to use the De Novo pathway. For high-risk devices that support or sustain human life, a PMA is required. In 2026, the FDA has streamlined the De Novo process, but it still requires substantial clinical evidence. For pediatric devices, you should expect that the FDA will require at least one prospective clinical study that includes a sufficient number of pediatric patients. The exact sample size depends on the device's intended use and the prevalence of the condition, but a common benchmark is at least 100 patients per age subgroup.
Before you start collecting data, you should engage with the FDA through the Pre-Submission (Q-Sub) process. This is a formal mechanism to get feedback on your development plan, including your proposed study design, data collection methods, and statistical analysis plan. The FDA encourages Q-Subs for pediatric devices because they help avoid costly mistakes later. In your Q-Sub, you should specifically address how you will handle the challenges of pediatric data collection, such as obtaining informed consent from parents and assent from children, and how you will ensure that your dataset is representative of the intended pediatric population. The FDA will also want to know how you plan to handle missing data, which is common in pediatric studies due to patient dropout or inability to complete certain procedures.
Another critical step is to leverage real-world data (RWD) from electronic health records (EHRs), registries, and wearable devices. The FDA has been increasingly accepting RWD for regulatory decisions, and in 2026, it is more open than ever to using RWD to supplement traditional clinical trial data. For pediatric AI devices, RWD can be particularly valuable because it can provide large, diverse datasets that would be impossible to collect in a clinical trial. However, you must ensure that the RWD is of high quality and that you have a clear plan for validating the accuracy of the data. The FDA has published several guidance documents on RWD, and you should familiarize yourself with these before submitting your application. Additionally, you should consider using a "pediatric-specific" data standard, such as the Pediatric Medical Device Data Standard, which was developed by the FDA and the National Institutes of Health (NIH) to facilitate data sharing and interoperability.
Finally, you should plan for post-market surveillance. The FDA requires that all AI devices have a post-market monitoring plan, but for pediatric devices, this is even more important because the long-term effects of AI algorithms on children are unknown. You should design a plan that includes active surveillance for adverse events, as well as periodic re-evaluation of the algorithm's performance as new data becomes available. The FDA has also introduced a "Predetermined Change Control Plan" (PCCP) that allows developers to make certain updates to their AI algorithms without requiring a new submission, but this is only available for devices that have been approved with a PCCP. For pediatric devices, you should consider including a PCCP in your initial submission to allow for age-specific updates as your algorithm is used in real-world settings.
Comparison of Regulatory Pathways for Pediatric AI Devices
| Feature | 510(k) Clearance | De Novo Classification | Premarket Approval (PMA) |
|---|---|---|---|
| Typical Class | Class II | Class II (novel) | Class III |
| Evidence Required | Substantial equivalence to a predicate | Moderate clinical evidence, often a pivotal study | High-quality clinical trial, often randomized |
| Average Review Time (2026) | 6-9 months | 10-14 months | 12-18 months |
| Pediatric-Specific Considerations | Must demonstrate that the predicate's pediatric use is valid; often requires additional data | Must provide age-specific performance data; FDA may require a pediatric study | Must include pediatric subgroup analysis; often requires a dedicated pediatric trial |
| Cost (Estimated) | $50,000 - $200,000 | $200,000 - $500,000 | $1 million - $5 million |
| Post-Market Requirements | Standard adverse event reporting | May require post-market study | Strict post-market surveillance and periodic reports |
| Best For | Devices with a clear predicate, including adult AI devices being extended to pediatrics | Novel AI algorithms with no predicate, but moderate risk | High-risk AI devices that guide critical clinical decisions |
Common Mistakes and How to Avoid Them
One of the most common mistakes that developers make is assuming that an AI algorithm trained on adult data will work in children. The FDA has repeatedly rejected such submissions, and in 2026, it is even less likely to accept them. To avoid this mistake, you should collect pediatric-specific data from the outset, even if it means delaying your development timeline. Another mistake is failing to account for age subgroups within the pediatric population. Children are not a homogeneous group; a 2-year-old is vastly different from a 16-year-old. The FDA expects that you will analyze your algorithm's performance across different age groups and report any significant differences. If you do not do this, you may be asked to conduct additional studies, which will delay your approval.
A third mistake is underestimating the importance of data quality. Pediatric data is often messier than adult data, with more missing values, inconsistent measurements, and variations in how data is recorded. The FDA has become more sophisticated in detecting data quality issues, and if your dataset has obvious problems, your submission will be rejected. To avoid this, you should invest in robust data curation and validation processes, and you should consider using external data quality audits. A fourth mistake is neglecting to engage with the FDA early. Many developers wait until they have a complete submission before contacting the FDA, which is a mistake. The FDA is more willing to provide feedback during the development process, and a Q-Sub can save you months of time and significant costs. Finally, do not ignore the ethical and social implications of your device. The FDA is increasingly considering the impact of AI devices on health equity, and if your device is only tested on a narrow demographic, you may face additional scrutiny. You should aim to include a diverse pediatric population in your studies, including children from different racial, ethnic, and socioeconomic backgrounds.
When to Act: Timing Your Submission in 2026
The optimal time to submit a pediatric AI device to the FDA in 2026 depends on several factors, including the maturity of your technology, the availability of data, and the regulatory environment. If you are just starting development, you should plan for a timeline of 3-5 years from concept to approval. This includes 1-2 years for data collection, 6-12 months for algorithm development and validation, and 6-18 months for FDA review. If you have already collected data and have a working algorithm, you should submit a Q-Sub as soon as possible, ideally before the end of 2026. The FDA's review times have been stable over the past year, but there is always the risk of changes in policy or staffing. The current political environment, with the FDA under scrutiny from conservative think tanks and some members of Congress, could lead to changes in the approval process. For example, there have been proposals to reduce the FDA's authority over AI devices, which could either speed up or slow down approvals depending on how they are implemented.
Another consideration is the availability of funding. Pediatric AI devices are often less profitable than adult devices, and many venture capital firms are hesitant to invest in this space. However, there are government grants and nonprofit funding sources, such as the FDA's Pediatric Device Consortia (PDC) grants, which provide funding and regulatory assistance to developers of pediatric devices. If you are relying on external funding, you should time your submission to align with your funding milestones. For example, if you need to raise a Series A round, having a successful Q-Sub meeting can be a strong signal to investors. Conversely, if you are running low on funds, you may want to submit your application earlier to avoid running out of money before approval.
Finally, you should monitor the FDA's regulatory agenda. In 2026, the FDA is expected to release a new guidance on AI-enabled devices that will specifically address pediatric considerations. This guidance is currently in draft form, and the FDA is accepting public comments until September 2026. If you are planning to submit a device in the next 12 months, you should review the draft guidance and consider how it might affect your submission. The guidance is likely to require more detailed reporting on pediatric performance, which could increase your evidence burden. However, it may also provide more clarity on acceptable study designs, which could reduce the risk of rejection. In any case, you should stay informed and be prepared to adapt your strategy as the regulatory landscape evolves.
Cost and Pricing Considerations for Pediatric AI Devices
The cost of obtaining FDA approval for a pediatric AI device in 2026 varies widely depending on the pathway, the complexity of the device, and the amount of clinical evidence required. For a simple 510(k) submission, you can expect to spend between $50,000 and $200,000 on regulatory fees, data collection, and consulting services. However, this does not include the cost of developing the algorithm itself, which can be significant. For a De Novo submission, the cost is higher, typically ranging from $200,000 to $500,000, because you will need to conduct a more rigorous clinical study. For a PMA, the cost can exceed $1 million, and in some cases, it can reach $5 million or more. These costs are often underestimated by startups, leading to budget overruns and delays.
In addition to the direct costs of FDA approval, you should also consider the ongoing costs of post-market surveillance and algorithm maintenance. The FDA requires that you monitor your device for adverse events and report them to the agency. This can be done in-house, but it requires dedicated staff and systems. You will also need to update your algorithm periodically to maintain its performance, which requires access to new data and computational resources. These ongoing costs can add 10-20% to your annual operating budget. To offset these costs, you should consider pricing your device appropriately. Pediatric AI devices are often priced lower than adult devices because the market is smaller, but you can still achieve profitability if you target high-volume conditions, such as pediatric asthma or diabetes. You should also explore reimbursement options, such as CPT codes for AI-based diagnostic tools, which have been expanding in recent years. In 2026, Medicare and many private insurers cover AI-based diagnostic tools, but coverage for pediatric-specific devices is less consistent. You may need to work with payers to establish coverage policies, which can be a lengthy process.
The Future of Pediatric AI Regulation: What to Expect Beyond 2026
Looking beyond 2026, the regulatory landscape for pediatric AI devices is likely to evolve in several ways. First, there is growing pressure from pediatricians and patient advocacy groups for the FDA to create a dedicated pediatric AI pathway. In June 2026, a coalition of pediatric medical societies submitted a citizen petition to the FDA, requesting that the agency establish a "Pediatric AI Device Accelerated Approval" program, similar to the RAPID pathway for adults. While the FDA has not yet responded, it is likely that this petition will be considered in the coming months. If such a program is created, it could significantly reduce the time and cost of pediatric AI approvals, but it would also require the FDA to accept more real-world evidence and post-market data, which carries its own risks.
Second, the FDA is likely to increase its focus on health equity and algorithmic fairness. In 2025, the FDA established a new Office of Health Equity, which has been tasked with ensuring that medical devices are safe and effective for all populations, including children. This office is expected to issue new guidance on how to evaluate AI devices for bias, and it may require that pediatric AI devices include subgroup analyses by race, ethnicity, and socioeconomic status. This will add to the evidence burden but will also make devices more trustworthy and equitable.
Third, there is a trend toward international harmonization of AI device regulations. The International Medical Device Regulators Forum (IMDRF) has been working on a common framework for AI devices, and in 2026, it released a draft document on "Good Machine Learning Practice" that includes a section on pediatric considerations. If this framework is adopted by the FDA and other regulators, it could simplify the approval process for companies that want to market their devices globally. However, harmonization also means that the FDA may adopt stricter requirements from other countries, such as the EU's Medical Device Regulation (MDR), which has been criticized for being overly burdensome for AI devices.
Finally, the role of real-world evidence is likely to expand. The FDA has been increasingly using RWD to monitor devices after approval, and it is possible that in the future, RWD could be used to support initial approvals for pediatric AI devices, especially for rare conditions where clinical trials are impossible. This would be a major shift, and it would require the FDA to develop new methods for validating RWD and ensuring its reliability. In the meantime, developers should stay informed about these developments and be prepared to adapt their regulatory strategies accordingly. The key takeaway is that while the current environment is challenging, there are opportunities for those who are willing to invest in high-quality pediatric data and engage with the FDA early and often.
Conclusion: Navigating the Pediatric AI Approval Maze in 2026
In summary, the FDA pediatric AI device approval requirements in 2026 are not a separate set of rules but rather a set of heightened expectations within existing pathways. The agency demands robust pediatric-specific data, careful consideration of age subgroups, and a clear plan for post-market monitoring. The review process is longer and more expensive than for adult devices, but it is not insurmountable. By understanding the regulatory landscape, engaging with the FDA early, and investing in high-quality data, developers can successfully bring pediatric AI devices to market. The rewards are significant, both in terms of improving children's health and in terms of establishing a foothold in a market that is underserved and likely to grow. As the regulatory environment continues to evolve, staying informed and adaptable will be the key to success.