Descriptive analysis of postmarket surveillance data for hip implants

Abstract Purpose Recent safety issues involving medical devices have highlighted the need for better postmarket surveillance (PMS) evaluation. This article aims to describe and to assess the quality of the PMS data for a medical device and, finally, to provide recommendations to improve the data gathering process. Methods A descriptive analysis of medical device reports (MDRs) on the use of MRA, a specific type of hip implant replacement submitted to the Food and Drug Administration Manufacturer and User Facility Device Experience database from 1 January 2008 to 31 December 2017. The number of reports was described as the number of MDRs per unique MDR number and stratified by different variables. The quality was assessed by the level of completeness of the collected PMS data. Results The total number of reports related to MRA was 2377, and the number of MDRs per year ranged between 84 in 2009 and 452 in 2017. Most of the reports were reported by manufacturer Depuy Johnson & Johnson and were reported by a physician. In 44.9% of the reports, the device problem was reported as “Unknown.” When the device problem was known, in the majority of cases, it was related to an implant fracture. The quality of the collected data was assessed as low due to missing information. Conclusion The underlying data should meet high quality standards to generate more evidence and to ensure a timely signal generation. This case study shows that the completeness and quality of the MDRs can be improved. The authors propose the development of tools to ensure a more dynamic complaint data collection to contribute to this enhancement.


| INTRODUCTION
An implantable medical device is a device that is partly or totally inserted into the human body or a natural orifice or is used to replace the surface of the body and is expected to stay in use for 30 days or more. Examples of implantable medical devices include dental implants, breast implants, hip implants and intraocular lenses. Surgical or medical procedures are used to insert, apply and remove implantable medical devices. To be classified as a non-active implantable medical device (NAIMD), the medical device should not have an integral power source; all devices with a power source are considered active implantable medical devices (eg, pacemaker, cochlear implants…). 1 Recent safety issues involving NAIMD have highlighted the need for better premarketing and postmarketing evaluation. 2,3 In the metalon-metal (MoM) hip safety issue, thousands of patients around the world may have been exposed to high levels of toxic metals from failing hip implants. The chromium and cobalt ions from the MoM hip implants could enter into the tissues of patients with this type of hip implants, leading to reactions that damaged the muscle and bone, and led to revision procedures or left some patients with long-term disability. [4][5][6][7] This safety issue was only identified by the Australian Health Authorities upon review of the Australian Orthopaedic Association National Joint Replacement Registry, and this finding was confirmed by the National Joint Replacement Registry of England, Wales, and Northern Ireland and the New Zealand Registry. This resulted in a worldwide recall of the MoM hip implants. The safety issue was highly publicized as MoM hip implants were approved for market use although lacking data derived from clinical trials. In addition, the manufacturers did not effectively review post-market clinical data (including device registries containing postmarket surveillance information) and thus failed to identify and report this risk to the health authorities. 8 A prior safety issue with Poly Implant Prothesis breast implant scandal 3 had also contributed to the emerging growing demand to improve the current passive-reactive postmarket surveillance (PMS) system of medical devices. An important part of this PMS system is the data collection of case (complaint) reports. To enhance the current surveillance system, it is important to measure and assess the quantity and quality of PMS data on medical devices.
Hip implants are NAIMD that are implanted during hip replacement surgery. Hip replacement surgery can be performed traditionally or by means of a minimally invasive technique. The main difference between the two procedures is the size of the incision and the type of prosthetic implant, either a total hip replacement or an MoM hip replacement. 9 With approximately 1.4 million hip implant surgeries performed every year around the world, it is the most common joint replacement procedure. In the United States, over 231 000 surgeries are performed annually. 10 Given the large use of hip implants and the need to improve medical device vigilance, we performed a case study and conducted a descriptive analysis of the PMS data from one of the most important publicly available spontaneous reports database, 11  Manufacturers must submit MDRs to the FDA "when they become aware of information that reasonably suggests that one of their marketed devices may have caused or contributed to a death or serious injury or has malfunctioned and the malfunction of the device or a similar device that they market would be likely to cause or contribute to a death or serious injury if the malfunction were to recur. Manufacturers must send MDRs of such deaths, serious injuries and malfunctions to the FDA" 12 when they become aware of any of the events described above reported from any country in the world. The definition of serious injury is described below.
The FDA provides access to MAUDE information through three different tools: (a) an online simple (single-parameter) interface, (b) advanced (multiparameter) search interface or (c) downloadable data files. These online search engines are extremely convenient; however, information obtained using these interfaces has some restrictions. [12][13][14] In our study, we used both the online search interface and the downloadable datasets.

| Outcomes
For this study, we were interested in reports related to the use of another type of hip implant (different from the MoM implant): the hip joint metal/ceramic/ceramic/metal semiconstrained cemented or uncemented prosthesis (FDA product code: MRA). 15 We considered

KEY POINTS
• The completeness and the quality of the data included in the medical device reports can be improved.
• New standards and safety tools should be developed to ensure a more dynamic complaint data collection process.
all events related to this type of device as events of interest. The FDA has a standardized vocabulary for adverse events and product problems. A total of 167 different event codes related to the use of the hip implant of interest (MRA) were analyzed. Malfunctions and serious injuries were classified according to the FDA regulatory definitions 16 ; a serious injury is an injury or illness that is life-threatening, results in permanent impairment/damage or necessitates medical/surgical intervention to preclude permanent impairment/damage. A malfunction stands for the failure of a device to meet its performance specifications or to perform as intended [Performance specifications include all claims made in the labeling of the device. The intended performance of a device refers to the intended use for which the device is labeled or marketed 16 ].
The type of reporter was classified as unknown or known (physician, nurse, patient, pharmacist, administrative and known others).
The type of reported adverse events was classified as malfunction and/or serious injury (for definitions, see above). The complaint sample availability and the corrective/remedial actions field were classified as Yes, No or NA.

| Data management and analysis
The study period comprised 10 years, and data for this period We obtained the following information from each of the downloadable datasets: • The MDR FOI Master dataset, filtering by the "manufacturer name" field for all the MRA Hip Implant Manufacturers available. The following variables were used: MDR report key, manufacturer name, type of event, report source, source type (country of origin; United States or foreign), reporter occupation, remedial actions and recalls.
• The Device Data dataset, filtering by "MDR report key." The following variables were used from this dataset: MDR report key number (to link), device availability and device evaluated by the manufacturer.
• The FOI Device Problem dataset, filtering by "MDR report key." The following information was used from this dataset: MDR report key Device Problem codes.
From these three datasets, one unique dataset was built using the "MDR report key," which was available in the three downloadable datasets.
The data were analyzed using descriptive statistics, reporting counts, proportions and stratifications. Absolute numbers and percentages were described by manufacturer, brand name, type of event (death, injury, malfunction, NA, other), reporter's occupation, type of reported adverse events and product problems, complaint sample availability (whether the device is available for further investigation) and corrective/remedial actions. The numerator was the number of reports with MRA hip implants for a specific brand name, and the denominator was the total number of reports for MRA hip implants during the study period. The proportion of reports with information on the type of device problem was 55.1% (in 44.9% of the reports, the device problem was reported as "Unknown"). The most frequently reported device problems included "implant fracture" (39.57%, 518 reports), "dislocation"    These results underline the need to obtain better postmarket complaint data for medical devices within the United States and beyond ( Figure 2). Improvements can be made in the reporting itself,  Event coding • Impossibility to identify patient harms and root causes associated with specific FDA device problem codes.
• FDA patient codes and FDA investigational codes (methods, results and conclusions) publicly available in MAUDE.
• FDA Patient exposure • Lack of information about frequency of device use does not allow to estimate patient exposure.
• The manufacturers could make their distribution/sales data available to the Health Authorities, upon request.
• Global Health Authorities (including FDA), Manufacturers

Root cause identification
• For the majority of reports the suspect sample is not available and cannot be evaluated by the manufacturer. Identifying the root cause of the event is especially difficult if the device in question has not been identified and directly evaluated by the manufacturer.
• Global adoption of UDI: in order to identify the device and link the device to a serial number, UDI needs to be present on the device and readily accessible in the medical record. • More guidance and training for the healthcare professionals on the importance of sending the device with all the adequate information (including UDI) to the manufacturer for evaluation, if the suspect device is explanted.
• Reporting facilities, FDA, Manufacturers Timely reporting • The MAUDE advanced search interface is updated monthly, and the search page reveals the date of the latest update. The FDA tries to include all reports received before the update, but the inclusion of some reports may be delayed.
• More guidance and training on the importance of timely reporting should be provided to the different stakeholders involved in the complaint handling process.
• Reporting facilities Report source • Most of the MDRs from MAUDE come from spontaneous reports received from the manufacturer. This type of report may be associated with reporting bias.
• Healthcare provider reports directly to the FDA need to be strongly encouraged via training and regulatory guidelines. • Enriching the FDA MAUDE PMS data with data from medical device registries. In order to be able to link the registry data with the manufacturer reports data, common standardized dataset including UDI should be created.

Scope
• MAUDE only includes FDA-reportable complaints. If the complaint does not meet the FDA reporting criteria, the complaint will not be in MAUDE. • MAUDE only includes complaints associated with medical devices that are marketed in the United States. If the medical device is not marketed in the United States, the complaint will not be in MAUDE.
• Exchange of PMS data (including FDAnonreportable complaints and trend reports for FDA-nonreportable complaints) between different Health Authorities. • Development of a global repository to store global PMS data for medical devices. can lead to quality improvement of the collected information. 17 Table 2 provides recommendations not only on how to improve MDR reporting but also recommendations on how to improve data collection in the MAUDE database. In addition to the limitations described in Table 2, our results also have additional limitations as postmarketing complaint data is prone to reporting bias. 18  example. The quality of postmarket complaint data and their timely collection are crucial for the validity of the complaint reports. It is time to face current challenges such as the lack of quality standards, lack of specificity in regulatory reporting, lack of harmonized coding and lack of engagement from reporters at the time to send samples back for analysis. We recommend that the different stakeholders in this process (manufacturers, health authorities, healthcare professionals and patients) work together to overcome these challenges.

ETHICS STATEMENT
The authors state that no ethical approval was needed.