Turning FDA Data Into Regulatory Advantage

Regulatory data from the FDA is an underused strategic asset in MedTech. Beyond supporting compliance, it offers deep insights into how competitors position their devices, what evidence they needed for clearance, and how regulators view emerging technologies. When analyzed properly, this information can guide smarter predicate selection, reduce regulatory burden, and highlight opportunities to accelerate time‑to‑market. The challenge is that much of this data is fragmented across PDFs, APIs, and databases that were never designed for easy exploration. At MD squared, we’ve built tools that turn this raw regulatory data into actionable intelligence—giving our clients a clearer, faster path through the US regulatory landscape. In this article, we highlight three use cases that show the types of information that can be derived from this data. This information allows clients to make more informed choices

Available data sources

Key data sources include 510(k), De Novo, and PMA decision letters. These contain precise information about intended use, indications for use, and the evidence that supported the regulatory clearance or approval. This is critical for companies developing new products: if a suitable Class II predicate device can be identified, the new device may follow the lighter 510(k) pathway and only needs to provide evidence comparable to that of the predicate. Choosing the right predicate can save significant time and effort when clearing a novel device.

The MAUDE database contains Medical Device Reports filed by users or manufacturers to document potential device failures. These reports describe types and causes of failures as well as their impact on patients. This information is valuable input for risk management during development. Products in the EU must also include post‑market surveillance, which covers not only complaints about their own device but also similar devices on the market. Until the EUDAMED module for problem reporting is fully active, the FDA MAUDE database remains the main publicly available source of information about competitor devices.

Both decision letters and MAUDE data can be searched through the FDA’s public website. However, not all relevant information is easy to access, and the search tools have significant limitations in query types and record limits. Fortunately, the FDA also provides an API and regular downloadable database dumps. The site https://open.fda.gov/apis/downloads/ details the available datasets.

Case I: Identification and Analysis of Potential Predicates

The standard FDA 510(k) database only supports searches based on company or product name, product code, or 510(k) number. Predicate identification typically involves searching for competitor products or browsing relevant product codes. Once a potential predicate is found, the PDF must be manually reviewed to confirm the indications for use. From there, previous predicates can also be identified, allowing the 510(k) tree to be traced backward.

The MD squared 510(k) database enables searches directly on the text of the indications for use. After identifying interesting predicates, it is also possible to see which subsequent 510(k)s used that predicate, allowing the 510(k) tree to be built forward as well.

 

Example of a 510(k) tree built by tracing both forward and backward from the initially identified 510(k) K163622 by Lung Vision. Nodes are colored by manufacturer.

When a set of potential predicates has been identified, the full text of the 510(k) letters can be analyzed by a local LLM to answer specific questions, such as whether certain evidence or standards were included in the submission. This is highly relevant because in principle a new 510(k) device must provide the same types of evidence as its predicate. Identifying predicates that required no or limited clinical data, or that did not include usability testing, can significantly reduce the regulatory burden.

Example set of 510(k)s—those in red were cleared without clinical data, those in green used clinical data as part of their clearance.

Full‑text search of the indications for use allows us to identify predicates that might not be obvious when relying only on known competitors. Deeper analysis supports informed decision‑making about the optimal regulatory pathway, helping streamline the US clearance process.

Case II: Benchmarking Regulatory Timelines

As noted in previous article, actual 510(k) review timelines can vary significantly. The nominal 90‑day review period quoted by the FDA hides substantial complexity, and companies often struggle to understand how their timelines compare to others. Benchmarking against all devices within a product code—and against specific peers—can provide meaningful insights.

The figure below shows an example benchmarking a fictional client against two peers across several product categories. As expected, there is broad variation in clearance times across the client’s products. For most categories, their timelines fall within the overall spread of the category and align with their selected peers. However, for product category AAA, the last three 510(k)s took notably longer than expected based on both peers and the rest of the category. This insight can prompt deeper investigation: Is a specific team struggling? Is there a type of evidence required in this category that does not align well with internal processes? Only by analyzing the data can a meaningful root‑cause discussion begin.

Box plots for the distribution of regulatory approval times across product codes. Colored dots indicate individual 510(k) submissions by the client and their peers.

In‑Depth Analysis of MAUDE Data

The MAUDE complaint database offers a wealth of information about typical problems observed in the field. The FDA website provides basic search options and the ability to download results to Excel, but only up to 500 records at a time. This means that broader searches must be split to retrieve all data. Our internal search interface supports much faster querying, has no record limits, and allows full‑text search of the complaint narrative. Because the data is not always consistently encoded (manufacturer and product names often appear in many variants), normalization can be applied to enable more accurate analysis.

One of the most powerful outputs is the ability to generate a matrix of product problems against patient problems. This reveals relationships between failure modes and clinical harm, which is valuable for both risk management and post‑market surveillance reporting.

Co‑occurrence matrix showing relationships between product problems and patient problems. Results shown here include only cases reporting clinical data.

Conclusion

The FDA provides a wealth of information about competitors’ products that can be used for regulatory strategy optimization and post‑market surveillance. However, the public search tools have limitations. MD squared has developed proprietary tools that allow us to extract far more valuable insights. Feel free to reach out to discuss any questions or challenges where we might be able to support you.

Erik 02

Erik van Dijk

Managing Partner

Don’t hesitate to ask us your questions.

Drop an e-mail or contact us via LinkedIn

This website uses cookies to ensure you get the best experience on our website. If you want to get more information about this, please view our privacy policy page.