Assessment of Data Sources That Support US Food and Drug Administration Medical Devices Safety Communications

Key Points

Question

What are the sources of initial safety signals that trigger the US Food and Drug Administration’s (FDA’s) Medical Device Safety Communications?

Findings

Of the 93 Medical Device Safety Communications published between 2011 and 2019, the source of 47% was direct reports to the FDA through the Medical Device Reporting program, and 34% were initiated by regulator-initiated assessments. Safety communications triggered by direct reports were associated with risk of death as the safety issue.

Meaning

This study found that Medical Device Safety Communications were mainly triggered by direct reports to the FDA, highlighting the importance of proactive identification of device-related safety issues to improve patient safety.

Abstract

Importance

Medical Device Safety Communications (MDSCs) are used by the US Food and Drug Administration (FDA) to convey important new safety information to patients and health care professionals. The sources of initial safety signals that trigger MDSCs have not been described previously.

Objective

To assess the sources of initial safety signals that trigger publication of MDSCs and the potential associations among MDSC data source, type of safety issue, and subsequent FDA action.

Design, Setting, and Participants

In this cross-sectional study, all MDSCs published on the FDA website between January 1, 2011, and December 31, 2019, were assessed. The MDSC characteristics, sources of initiating safety signals, regulatory approval or clearance pathways of the related medical devices, and subsequent FDA actions were collected from the FDA website.

Main Outcomes and Measures

The main outcome was the distribution of sources of initial safety signals that led to publication of MDSCs. Secondary aims included exploration of potential associations among safety signal sources (direct reporting vs other), type of safety issue (death vs other), and FDA action (withdrawal vs other).

Results

A total of 93 MDSCs were evaluated. Median time from device approval to MDSC posting was 10 years (interquartile range, 6-16 years). The most common data sources that triggered MDSCs were direct reports to the FDA through the Medical Device Reporting (MDR) program (44 of 93 [47%]) followed by regulator-initiated assessments (32 [34%]). Common safety issues included patient injury (25 [27%]), potential wrong diagnoses (19 [20%]), and death (18 [19%]). Frequent FDA action after MDSC posting included recommendation for increased vigilance and caution (47 [51%]), complete device withdrawal (12 [13%]), and warnings of specific lots or clinics (12 [13%]). There was a statistically significant correlation between direct reports of adverse events to the FDA through the MDR program and risk of death as a safety issue (14 of 44 [32%] for direct reporting vs 4 of 49 [8%] for any other data sources, P = .007).

Conclusions and Relevance

In this cross-sectional study, the most common source of initial safety signals that triggered MDSCs was direct reports of real-world adverse events to the FDA through the MDR program. The delayed detection of postmarketing adverse events highlights the importance of proactive identification of emerging device-related safety issues.

This cross-sectional study assesses the sources of initial safety signals that trigger publication of Medical Device Safety Communications and the potential associations among Medical Device Safety Communication data source, type of safety issue, and subsequent action by the US Food and Drug Administration.

Introduction

The US Food and Drug Administration (FDA) recognizes more than 6500 categories of medical devices, defined as “any instrument, machine, contrivance, implant, or in vitro reagent intended to treat, cure, prevent, mitigate, or diagnose disease in man.”¹ Devices are categorized into 3 classes based on perceived risks.² Class I devices are considered low risk and are mostly exempt from any premarketing review. Class II devices carry moderate risk and are mostly cleared through the 510(k) pathway, requiring manufacturers to demonstrate substantial equivalence in intended use and technological characteristics to current legally marketed devices. Class III devices include high-risk, life-sustaining devices that require premarketing approval.

Under the current framework, most new devices are not supported by any clinical efficacy or safety data because only a few are categorized as class III.² Although this approach can hasten approval rates and possibly fosters innovation and rapid uptake of new technologies, recent concerns regarding the emergence of postmarketing safety issues have led the FDA to strengthen the premarket clearance process of some medical devices (eg, comparison of new devices with recently cleared devices rather than those cleared >10 years ago and use of objective safety and efficacy criteria for clearing certain device types).³ However, even robust, clinically oriented premarket testing might fail to detect uncommon adverse events (AEs), long-term complications, AEs associated with off-label use, and postmarketing issues, such as manufacturing errors. A report⁴ presented to the Institute of Medicine that analyzed recalls designated by the FDA as involving the greatest safety concerns (ie, class I recalls) concluded that additional premarketing human clinical studies would not significantly impact identification of most safety issues.

The FDA has several postmarketing mechanisms aimed to identify emerging AEs. The Medical Device Reporting (MDR) program is an AE reporting system created to capture real-world device-associated deaths, serious injuries, and malfunctions.⁵ The Manufacturer and User Facility Device Experience (MAUDE) database houses MDRs submitted to the FDA by mandatory reporters, such as manufacturers, importers, and device user facilities, and voluntary reporters, such as health care professionals, patients, and consumers. The MDRs are publicly viewable in the FDA MAUDE database. Some device manufacturers are required to complete postapproval studies mandated at the time of initial approval or when concerns are raised for previously unsuspected safety issues.¹ The FDA also collects data from investigator-initiated postmarketing publications in the medical literature.

Despite these measures, there are increasing concerns regarding the current postmarketing surveillance system. Passive surveillance has been reported to underestimate malfunction of cardiac electronic devices that results in death.^6,7 A report⁸ published by the International Consortium of Investigative Journalists estimated that nearly 2 million injuries and 80 000 deaths associated with medical devices occurred between 2008 and 2017. Postmarketing trials were reported to often have design limitations that hindered their ability to be of clinical use.⁹ These increasing concerns have also reached the public media.^10,11 The FDA has begun developing a more proactive surveillance infrastructure,¹² but the extent of regulatory action informed by this effort is unclear.

Medical Device Safety Communications (MDSCs) are the FDA’s primary tool for communicating important emerging postmarketing safety information to patients, health care professionals, and the public.¹³ Data regarding MDSC characteristics and the distribution of different sources of initiating safety signals may inform practitioners and regulators about the current state of postmarketing AE monitoring and assist in directing future efforts to improve postapproval identification of previously unrecognized AEs. Such data are currently lacking. We aimed to describe MDSC characteristics and sources of initial safety signals and to explore potential associations between MDSC source and the interval between initial device approval to MDSC publication, type of safety issue, and subsequent FDA action.

Methods

Data Source

In this cross-sectional study, the FDA website was searched to identify all MDSCs published between January 1, 2011, and December 31, 2019.¹³ To reduce data heterogeneity, MDSCs that described updates of previously posted MDSCs were excluded from the final cohort. This study was performed in Israel. Under Israeli law, no ethical review/approval is required for analysis of publicly available nonpatient data.

Data Extraction

All MDSC data were extracted by one of us (N.T.), with adjudication and oversight by another (D.S.). The following data were extracted from each MDSC: publication date, device name and/or type, data source, main safety issue (grouped as death, injury, wrong diagnosis, infectious disease transfer, cybersecurity, inherent device futility for intended use, or other), and subsequent FDA action. Safety issues were categorized according to wording in the related MDSCs. Initial data sources of MDSCs were grouped as direct reporting of AEs to the FDA through the MDR program by patients, health care professionals, or the product’s manufacturer⁵; scientific publications in medical literature; regulator-initiated assessment (including studies mandated by the FDA or the European Medicines Agency); and notifications by the US Department of Homeland Security. Initial regulatory approval or clearance pathways were abstracted from the FDA website.^14,15 The FDA actions were divided into the following groups: complete device withdrawal (halting manufacturing and distribution), FDA recommendation for increased awareness and caution by health care professionals and/or medical device users, directive to avoid use of specific product lots or specific clinics, instruction to avoid using unauthorized devices, reminder to avoid using device for an unauthorized indication, and other FDA actions (eg, a warning of aircraft crashes caused by children playing with high-powered hand lasers). We also collected data regarding publication of specific guidelines to stakeholders on ways to mitigate risk, implementation of new guidance for manufacturers to proactively address the emerging safety issues, and device recalls (active removal of unsafe products from the market).^16,17 For class II and class III devices, the initial medical device clearance or approval date, if available, was retrieved from the 510(k) and premarket approval online public databases^14,15 or identified through a supplemental Google search.

Statistical Analysis

Data were descriptively reported for the entire MDSC cohort. These data include descriptions of MDSC initial data sources, safety issues, regulatory approval pathways, and subsequent actions undertaken by the FDA. The intervals between approval date and MDSC date were calculated for class II and class III devices (presented as median and interquartile range [IQR]). The trend for the number of MDSCs published annually was assessed using Spearman correlation coefficients. Associations between the initial trigger of MDSCs (grouped as MDR program vs other sources) and the interval between initial device approval to MDSC publication, type of safety issue (death vs other), and FDA action (withdrawal vs other) were explored using the Fisher exact test for categorical variables and the Wilcoxon rank sum test for continuous variables. Statistical significance was defined as 2-sided P < .05. Multivariable analysis was not planned because of the insufficient number of MDSCs to appropriately fit a multivariable model. Data analyses were conducted using SPSS software, version 21 (IBM Inc).

Results

From January 1, 2011, to December 31, 2019, 120 MDSCs were published, of which 27 were updates of prior MDSCs. Therefore, the final cohort comprised 93 safety communications (eTable in the Supplement). Common types of medical devices mentioned in MDSCs were operating room equipment (15 [16%]), imaging devices (13 [14%]), cardiac defibrillators (9 [10%]), and other implantable devices (eg, heart valves and breast implants) (10 [11%]) (Table 1). Twenty-one devices were classified as high-risk devices (class III) and required premarket approval, of which 2 were approved using an expedited review, 31 (33%) were cleared through the 510(k) program, 38 (41%) were never authorized by the FDA, and 3 (3%) were approved through other regulatory pathways (Table 1 and eTable in the Supplement). Most safety communications (57 [61%]) involved device types rather than a specific device or model. Representative examples of MDSCs, including those that target cardiac devices, duodenoscopes, programmable syringe pumps, and cranial perforators, are detailed in Table 2. The median time from initial device approval to safety communication posting for class II or class III devices was 10 years (IQR, 6-16 years). The number of safety communications published each year has remained stable over time (r = 0.44, P = .23).

Table 1. Medical Device Safety Communication Characteristics.

Characteristic	Finding (N = 93)a
Medical device type
Cardiac pacemakers or defibrillators	9 (10)
Other implantable devices	10 (11)
Operating room equipment	15 (16)
Endoscopy devices	8 (9)
Endovascular devices	7 (7)
Imaging devices	13 (14)
Infusion pumps	6 (6)
Laboratory tests	9 (10)
Other devices	16 (17)
Regulatory approval pathway
Premarket approval	21 (23)
Standard review	19 (21)
Expedited review	2 (2)
510(k)	31 (33)
Other	3 (3)
Not applicable	38 (41)
Risk classification
III	21 (23)
II	45 (48)
I	1 (1)
Not applicable	26 (28)
Safety communication source
Medical device reporting	44 (47)
Regulator-initiated assessment	32 (35)
Scientific publications	12 (13)
US Department of Homeland Security	5 (5)
Potential safety issue
Death	18 (19)
Injury	25 (27)
Wrong diagnosis	19 (21)
Disease transfer	13 (14)
Cybersecurity vulnerabilities	9 (10)
Inherent device futility	5 (5)
Other	4 (4)
FDA action
Withdrawal	12 (13)
Call for caution with approved usage	47 (50)
Publication of risk-mitigation guidelines	13 (28)
Implementation of new guidelines for industry	9 (19)
Warning
Specific lots or clinics	12 (13)
Unauthorized device	9 (10)
Unauthorized indication	10 (11)
Other	3 (3)
Recalls
Recalled devices	32 (34)
Recalled devices, median (IQR), No.	73 605 (4020-260 265)
Class I device recall	15 (47)
Class I recalled devices, median (IQR), No.	45 304 (17 875-500 000)

Abbreviations: FDA, US Food and Drug Administration; IQR, interquartile range.

^aData are presented as number (percentage) of devices unless otherwise indicated.

Table 2. Examples of Medical Device Safety Communications.

Medical device	Safety issue	Signal source	FDA action
St Jude Riata defibrillator leads	Lead failure, potentially resulting in death	MDR	Class I recall
Medtronic implantable cardiac devices	Cyber vulnerability of unknown significance	MDR	Implementation of new guidelines for industry to proactively address safety issue
ERCP duodenoscopes	Transmission of infectious agents	MDR	Publication of risk-mitigation guidelines for health organizations and implementation of new guidelines for industry to proactively address safety issue
Programmable syringe pumps	Inconsistent infusion rates, occlusions, and inadvertent boluses	MDR	Publication of risk-mitigation guidelines for health organizations and implementation of new guidelines for industry to proactively address safety issue
Cranial perforators with an automatic clutch mechanism	Failure to disengage, resulting in cranial injury	MDR	Reemphasis of correct surgical technique
Stryker Neptune surgical suction systems	Hemorrhage, soft tissue damage, and death	MDR	Recall and withdrawal of specific models

Abbreviations: ERCP, endoscopic retrograde cholangiopancreatography; FDA, US Food and Drug Administration; MDR, medical device reporting.

The most common data source that triggered MDSCs was direct reports to the FDA by patients, health care professionals, or the product manufacturer through the MDR program (44 of 93 [47%]; 25 patients or health care professionals [27%] and 19 company reports [20%]) followed by regulator-initiated assessments (32 of 93 [34%]) (eg, FDA warning of breast cancer screening using thermography¹⁸). A total of 14 regulator-initiated assessments (44%) were triggered by routine regulatory inspection of manufacturers, 9 (28%) were triggered by FDA awareness of widespread unauthorized marketing and use of certain medical devices, 7 (22%) were triggered by regulator-affiliated organizations, and 2 (6%) were triggered by FDA-mandated postmarketing studies (eTable in the Supplement). An additional 12 (13%) were triggered by nonmandated scientific publications (eg, MDSC targeting Stryker Wingspan Stent System for intracranial stenosis following the Stenting vs Aggressive Medical Management for Preventing Recurrent Stroke in Intracranial Stenosis [SAMMPRIS] study¹⁹) (Table 1). Twenty-one MDSCs (23%) mentioned secondary data sources, including the MDR program (8 [9%]) and scientific publications (6 [6%]). Common safety issues included patient injury (25 [27%]), potential wrong diagnoses (19 [20%]), death (18 [19%]), possible contagious disease transfer (13 [14%]), and cybersecurity vulnerabilities (9 [10%]). Eight cybersecurity-related MDSCs (89%) were published between 2015 and 2019. Some device types had distinctive safety issues: endoscopy devices and disease transfer (8 of 8 safety communications [100%]), imaging devices and potential wrong diagnoses (13 of 13 safety communications [100%]), implantable devices and injury (9 of 10 safety communications [90%]) or defibrillators and death (5 of 9 [56%]), or cyber-related issues (4 of 9 [44%]). Other device types had more variable safety-related issues.

The most common FDA action following safety communications was a recommendation for increased vigilance or a suggestion to apply caution when using the related medical device (47 [50%]). For some of these devices, the FDA provided stakeholders with specific guidelines to mitigate risks (13 [28%]), whereas for others, new guidelines for manufacturers were implemented to proactively address the emerging safety issues (9 [19%]). Other common actions were complete withdrawal of devices from the market (12 [13%]) and warnings of specific lots or clinics (12 [13%]) or of unauthorized indications (10 [11%]). The FDA requested recalls for 32 devices (34%), including 15 (47% of all recalls) class I device recalls (ie, highest risk), representing millions of potentially harmful devices (Table 1).

The preplanned analysis according to the source of MDSCs revealed a statistically significant associations between direct reports of AEs to the FDA through the MDR program and risk of death as a safety issue (14 of 44 [32%] for direct reporting vs 4 of 49 [8%] for any other data sources, P = .007). The main drivers of this association were the higher prevalence of direct reporting to the FDA for cardiac defibrillators (8 of 9 [89%]) and operating room equipment (11 of 15 [73%]), each associated with 4 safety issues that potentially resulted in patient death. No similar associations were found between MDSC trigger and market withdrawal (5 of 44 [11%] vs 7 of 49 [14%], P = .76), the time from initial approval to safety communication posting, or death as a safety issue and market withdrawal.

Discussion

The MDSCs are the FDA’s primary tool for conveying important emerging safety information to patients, health care professionals, and the public. We found that the 2 leading initial sources that trigger MDSCs were direct reporting of AEs to the FDA and regulator-initiated assessments. The MDSCs that involve risk of death were more likely to be triggered by direct reporting of AEs than by other data sources. We are not aware of previous reports that describe these findings.

Postmarketing surveillance is of considerable significance for medical devices because most are cleared without premarketing clinical testing. The FDA engages in passive and active postmarketing surveillance of medical devices. The most common source of important new safety data is an aggregation of passively collected AE reports. More than 90 000 reports are added to the MAUDE database each month,²⁰ and those reports were the trigger for only 44 MDSCs during a 9-year period. This low yield can be partly explained by the many limitations of the MAUDE database, which include underreporting, which can cause underestimation of AE prevalence; reports that may be incomplete, inaccurate, and untimely^5,21; reports sent almost solely by manufacturers²²; and lack of certainty regarding causality between the medical device and the reported AE.²³ Finally, completeness of the MAUDE database has been compromised by reporting millions of serious AEs associated with high-risk medical devices through summary AE reports rather than filling individual MDR reports, thereby bypassing the MAUDE database.^24,25 Active surveillance also has considerable limitations. Regulator-mandated postmarketing studies are often of nonrandomized design and enroll small patient cohorts that are followed up for short periods.^26,27,28 Results of some regulator-mandated studies remain unreported many years after device approval. Finally, active regulator-initiated surveillance is by nature focused in scope and is more effective in narrow, defined circumstances; therefore, such surveillance cannot be relied on as a mechanism for the early identification of emerging postmarketing AEs. The limitations of both postmarketing and active surveillance programs likely contribute to our findings of substantial intervals between the initial medical device approval and MDSCs. Delayed AE detection may be associate with preventable harm and death, underlining the need to proactively identify postmarketing device-related safety issues to expedite identification of serious AEs unrecognized at the time of initial device approval. This finding is emphasized by the large quantity of devices recalled after identification of safety issues.

Given the substantial role of medical devices in modern medicine, it would be expected that potential AEs would be tracked and identified through a robust, proactive postmarketing surveillance system with mandatory reporting by manufacturers rather than depend on the accumulation of voluntary selected AE reports, which may not represent the actual number of safety incidents. An efficient modern surveillance system would require automatic collection and continuous analysis of real-time, real-world data from multiple electronic sources (such as anonymized electronic patient records) allowing rapid identification of emerging safety issues and swift implementation of corrective actions by proactive regulators. Such a system should be transparent and open to free public access to improve trust and allow independent researchers to better identify undetected emerging safety issues. A potential model for a proactive surveillance program might be the FDA’s Sentinel Initiative, a national electronic system that uses administrative and insurance data from multiple sources to identify emerging real-world issues related to prescription drugs.²⁹ The medical device equivalent might be the National Evaluation System for Health Technology system, which aims to link and synthesize real-world evidence from different sources across the medical device landscape, including clinical registries, electronic health records, and medical billing claims.¹² The FDA requires that the labels of all class II and III devices include a Unique Device Identification number, making the manufacturer responsible for keeping a complete list of all end users for efficient notifications in case of an identification of an emerging safety issue and potentially enabling efficient analyses of large electronic health record and claims data sets.³⁰ A disproportional number of MDSCs were assocaited with the relatively uncommon, high-risk class III devices, making them a potential target for focused postmarketing scrutiny using Unique Device Identification–based electronic surveillance program. However, integration of Unique Device Identification numbers into electronic health records has been challenging, and implementation of the National Evaluation System for Health Technology in general has been slower than expected.²⁵

The importance of a proactive postmarketing surveillance program is highlighted by the recent implementation of the Breakthrough Devices Pathway Designation, enabling rapid FDA assessment and review of promising devices.³¹ Although the statutory premarket approval and 510(k) clearance standards are unchanged for this program, a recent study³² reported significantly higher recall rates for high-risk medical devices that underwent FDA priority review compared with those receiving standard review. This finding further accentuates the importance of robust postmarketing surveillance for rapid identification of emerging safety issues associated with these often novel and complex devices. The limitations of the current surveillance program are further emphasized by the association of MDSCs triggered by direct reports of AEs to the FDA with risk of death as a safety issue. The large proportion of MDSCs related to device type rather than specific problematic devices might also represent less granular data collected through anecdotal reports, making it difficult for regulators to assess the prevalence of emerging AEs related to specific devices.

A potential source of postmarketing safety data might have been published literature. Investigator-initiated postmarketing trials are a primary source of novel safety data for prescription drugs,³³ in contrast to the few MDSCs triggered by similar work (eTable in the Supplement). A potential explanation includes the diversity of device types and manufacturers, which make research more logistically challenging and the results less generalizable. This explanation is highlighted by FDA guidelines that call attention to recurrent safety issues with specific device types, instructing health organizations on ways to mitigate the reported risks (eg, disease transfer associated with duodenoscopes) or manufacturers to proactively address the identified safety issues (eg, cyber vulnerability of Medtronic cardiac devices).^34,35 Of interest, MDSCs related to cybersecurity issues have increased in recent years as device complexity has increased and more devices have involved computer networks and wireless data transmission. Because many of these devices are potentially vulnerable to cyberattacks that could disable or manipulate design settings, manufacturers and regulators should consider cybersecurity before marketing approval.^36,37

Limitations

This study has limitations. First, the source of the initial safety signal that triggers MDSCs might not necessarily represent the sources of other safety data received and processed by the FDA. Second, we chose to focus on a single source of the initial safety signal, although some MDSCs were supported by data from several sources of varying significance, which would have complicated the analysis. Different choice of methods may have resulted in different findings. Third, wording in the MDSCs regarding the triggering data sources might be inaccurate or incomplete. Fourth, inferential analysis was limited by the small number of MDSCs and by heterogeneity of the included devices.

Conclusions

In this study, most initial safety signals for MDSCs were based on data from direct AE reporting to the FDA or regulator-initiated assessments. The MDSCs triggered by direct AE reporting were associated with risk of death. These data highlight the importance of proactive identification of postmarketing device-related safety issues to provide health care professionals with more complete data regarding potential AEs and improve patient safety.

Supplement.

eTable. Characteristics of Medical Device Safety Communications