Technical Analysis of MAUDE Data to Discover Insights

Introduction

In the last few decades, vascular surgery has made a dramatic shift from traditional open surgeries to percutaneous endovascular interventions that require internal access to an artery. Once the procedure is completed, the hole in the artery is closed by the traditional approach of applying manual pressure on the puncture site, or alternately, a Vascular Closure Device (VCD) is used to stop the bleeding. The traditional procedure is time consuming, painful and requires up to eight hours of bedrest, whereas VCDs stop bleeding quickly, thereby increasing patient comfort. However, complications such as rebleeding, hematoma formation, pseudoaneurysm, or retroperitoneal bleeding may arise during vascular closure procedures. It is mandatory for manufacturers and importers to submit to the USFDA, reports containing information on their marketed devices that cause serious injury or death on account of malfunctioning of their devices. All such reports are stored in a MAUDE (Manufacturer and User Facility Device Experience) database.

MAUDE

MAUDE is an online searchable database that is a repository of incidents caused by approved medical devices. It is mandatory for medical device manufacturers to submit safety and effectiveness data to the FDA, disclosing adverse effects on patients, and also submit a summary of the clinical studies. In addition to such reports that are submitted to the FDA during pre-market approval, manufacturers must also submit adverse events that happen during actual use of the device.

SciTech Patent Art (SPA) conducted a study to assess the safety and effectiveness of vascular closure devices (VCDs) by analyzing the adverse event data available in MAUDE database. While many studies on interpretation of MAUDE data are available in literature, they are based on the auto classification provided by  database and are not accurate. In the current study, the adverse event reports were downloaded, read and reclassified manually. This approach has helped bring out better insights on device problems.

Study on Perclose ProGlide SMCS

The key market players in the vascular closure devices area are Terumo, Cardiva Medical, Abbott Laboratories, Cordis, Vivasure Medical, Teleflex, Vasorum Ltd and others. Abbott Laboratories has three different products, namely Perclose ProGlide Suture Mediated Closure (SMC) System, StarClose SE Vascular Closure System (VCS) and Prostar XL Percutaneous Vascular Surgical System (PVSS). Perclose ProGlide SMCS was selected for this study, as it is the most commonly used device for performing vascular closure procedures.

Study details

A search was conducted in MAUDE database using the Advanced Search Interface. The manufacturer’s name used was “Abbott”, and product code “MGB” (Device, Hemostasis, Vascular). The study period selected was   1^st January 2011 to 10^th September 2020 (date of receipt of report received by USFDA). During the process of retrieving information, it was observed that MAUDE database has several limitations. It returns only the latest 500 search hits and provides an export option for these 500 hits only. The records of all the adverse events reported over a period of one year are available as zipped files which are  not provided in a single zipped file. Therefore, for each selected year, multiple zipped files have to be downloaded from MAUDE.

For the current study, approximately 26 zipped files were downloaded. The downloaded datasets were large, with a few files containing 10 million rows of data. As such large datasets cannot be processed using conventional methods, SPA leveraged Big Data Analytics to consolidate the data into simpler form. Information for the following fields was extracted from the dataset and compiled into a single spreadsheet:

1. MDR Report Key
2. Date Received
3. Brand Name
4. Manufacturer Name
5. Patient Problem Code
6. Device Problem Code
7. Event Type
8. Event Description
9. Manufacturer Narrative
10. Report Number

The original MAUDE database contained duplicate reports that were generated due to multiple reports being filed by stakeholders such as doctor, hospital, device manufacturer etc. for the same faulty device or incident. Therefore, once the data was compiled, duplicate reports were removed. Further, reports that were not relevant and also those containing improper or insufficient information were removed.

For the study, SPA analysts addressed the following questions related to the device:

1. What are the frequently occurring problems associated with the device?
2. What are the user related problems?
3. What are the critical problems that could have led to the patient’s death?
4. Are there new problems that have been reported in the recent past?
5. Do opportunities exist for improving the design of the device?

Need for reclassification

The MAUDE database provides classification of all the adverse events that are reported to the USFDA. Based on the data under “Event Description”, which is provided by the reporting authority, all the reports are given problem codes related to the device or patient. The device-related problem codes are the ones describing the problem faced on account of a faulty medical device during the surgical procedure, whereas the patient problem code describes how the patient is affected because of the faulty device. For example, during the vascular puncture closure procedure, the VCD may have failed to deploy the sutures, because of which there could have been bleeding. Also, the procedure could have taken additional time to complete, since the doctor would have had to discard the first VCD and perform the puncture closure procedure with a second VCD. In such a case, during data reclassification, failure to deploy has been considered as the device problem, and bleeding as the patient problem.

Fig. 1 depicts the top 15 problems related to device and patient as classified by MAUDE database.

Fig. 1: MAUDE classification of top 15 problems related to device and patient.

In Fig.1, the terminology used for describing the adverse event is general in nature. For example, the most frequently occurring problem with VCD is identified as “Device operates differently than expected”. One cannot understand the exact nature of the problem associated with the device by reading the classification provided by MAUDE.

Further, similar device problems have been classified under different headings in MAUDE database as shown below.

It can be understood from “Event Description” in the above examples that in both cases, the device problem is suture detachment. However, in the MAUDE database, while one report has been classified under the device problem code “Device Operates Differently   Expected”, the other one has been classified under “Detachment of Device Component”. Such improper classification can be seen in patient problems too, wherein the same patient effect is classified under different patient problems. Hence, there is a definite need for reclassification of adverse event reports appearing in the MAUDE database, so as to understand the precise problem existing with the device/patient.

For purposes of this study, adverse event reports were downloaded and collated for Abbott’s Perclose ProGlide SMCS, for the period 2011 to 2020.  Duplicate reports and those that were lacking in complete information, and reports based on literature review were deleted. The number of adverse event reports thus retrieved was 12,637. The adverse event reports were reclassified using VantagePoint, a powerful desktop text-mining tool that helps the user analyze and visualize search results into patterns and relationships. VantagePoint also helps one rapidly navigate through a large set of results.

Insights from the study

The reclassification was performed using keywords obtained from the “Event Description” text data. The number of adverse events categories after reclassification is approximately 15, whereas there are approximately 100 categories in MAUDE database. Presented below are the problems identified based on reclassified data obtained using VantagePoint.

Table 1: Reclassified MAUDE data pertaining to device problems in Abbott’s Perclose ProGlide.

Table 2: Reclassified MAUDE data pertaining to patient problems in Abbott’s Perclose ProGlide.

Table 3: List of device problems classified under “Others” in Table 1.

After identifying the adverse events associated with Perclose ProGlide SMCS, the reports were analyzed to identify those that were caused due to user error. Some of the adverse events reported due to user error are as follows:

• Failure to position and maintain the device at 45 degrees throughout the deployment and retraction of plunger.
• Footplates not aligned properly on the inner surface of the vessel wall.
• Removal of device even though a slight resistance is felt resulting in damage to the tissue.
• Rotating the device during usage.
• Physician not following the correct sequence of deployment as per instructions for use.
• Usage of device even after expiration date.
• Physician not trained in the use of Perclose ProGlide SMCS.

Though, only few user error-related adverse events were reported to FDA in this area, there  could be a major research opportunity to identify better designs, since these errors can be mitigated through design enhancements. One such example could be to increase the strength of the black sheath or have an indication that footplates are properly aligned to the vessel wall.

Challenges:

Challenges encountered by the analyst during the course of this study are listed below:

1. Identifying the product code for hemostatic devices:
   • The list of product codes in MAUDE database is very long and therefore, it is difficult to identify the relevant codes.
   • Sometimes, there are multiple product codes for the same device.
2. Downloading the entire database since 2011 and compiling the sheets:
   • MAUDE database provides an option to download only 500 results from the search interface. So, the analyst needs to download multiple zip folders for each year and compile all the data in a single spreadsheet.
   • In some files, the data is so large (some files have more than 15 million rows) that Excel cannot handle the data.
3. Filtering the results with respect to product code and assignee:
   • Excel filters would not work on account of the large dataset and would often keep crashing.
   • Duplicate reports pose a major problem while compiling/analyzing the data. MAUDE does not identify adverse event reports filed by multiple stakeholders for the same device. Hence, analysts have to take Event Date, LOT Number, Catalogue Number etc. into consideration in order to identify duplicate reports.
   • Further, in the adverse event reports, the company names are not normalized, and so one needs to consider all possible variations of a company’s name.
4. Compiling patient problem code, device problem code, event description, manufacturer narrative etc.
   • As the entire data is not available in a single file, there is a need to compile data related to patient problem code, device problem code, event description etc., from multiple files.
   • Extracting definitions for patient problem code and device problem code poses a problem.
5. Analysis and interpretation
   • Even after all the data is compiled in a single file, the data cannot be analyzed or interpreted meaningfully, as in MAUDE, the same problem is categorized under different codes. No standard categorization protocol is followed in MAUDE.

For reasons mentioned above, there is a need for reclassification of adverse event data in the MAUDE database. To bring clarity, SciTech Patent Art’s analysts have manually read and reclassified the MAUDE adverse event reports.

Discussion:

Even though there are a large number of adverse events reported for this device, 96% of the time there has been no severe complications with the patient indicating that Perclose ProGlide is one of the safest devices available in the market. Also, Abbott has been constantly striving to improve its existing products in order to overcome the above-mentioned adverse events. This statement is being made based on the study of Abbott’s PMA’s (Premarket Approval) for Perclose ProGlide. Abbott received an approval to modify the angle tolerance of the plunger component in the year 2015. It received approvals for making changes to the sheath in the year 2017 and 2020. Additionally, Abbott has received premarket approval for incorporating new design changes to the deployment device of Perclose ProGlide SMCS. Abbott has been continuously filing continuation-in-part (CIP) patent applications claiming enhancements to the existing product portfolio. The company has filed CIPs for improved cuff link and new deployment techniques.

A medical product can be called as successful only if it is working well during the actual surgical procedure and causes minimum problem to the physician as well as the patient. By conducting studies such as this, companies can keep track of their own products’ performance, and can also track that of their competitors. Such studies will help identify design improvements that can be incorporated into their own future products. Such a study  can be conducted on any type of medical device as MAUDE provides data for all medical devices such as infusion pumps, robotic devices, cardiovascular devices, endoscopic devices, injectors etc.

Conclusion:

As surgical systems continue to evolve with the onset of new technologies, critical analysis of adverse events will enhance the understanding of medical devices and offer opportunities for device enhancements. Combining big data analytics and domain knowledge will help bring out insights from the adverse events in the MAUDE database, resulting in reduced incident rates that will ultimately result in improved patient comfort.