FDA approved the August 2017 pacemaker firmware fix as a recall, to be installed patient by patient
On August 29, 2017, the FDA announced firmware for 465,000 U.S. pacemakers to close wireless security flaws, approved as a corrective action. In our reading, the recall label put a security patch under device-correction rules and a clinical risk-benefit decision for each patient.
Looking Back revisits past weeks one at a time, up to the most recent, and asks what it meant for the banks, health systems and public agencies we serve. The series was written in 2026, with the benefit of hindsight, and each article says which week it looks back on.
Two-minute summaryThe dek and the first sentence of each section, quoted from the article
On August 29, 2017, the FDA announced firmware for 465,000 U.S. pacemakers to close wireless security flaws, approved as a corrective action. In our reading, the recall label put a security patch under device-correction rules and a clinical risk-benefit decision for each patient.
- The week On Tuesday, August 29, 2017, the Food and Drug Administration issued a safety communication about cybersecurity vulnerabilities in radio frequency-enabled pacemakers from Abbott, which the agency described as formerly St. Jude Medical [1].
- The classification: why this fix was a correction, not a routine update The FDA’s December 2016 postmarket guidance divided cybersecurity changes into two groups [3].
- The clinical decision: a risk-benefit choice for each patient A routine IT patch is applied to every system unless there is a reason not to.
- The delivery: an in-person procedure through the device clinic The update could not be pushed remotely.
- The record: what each party had to document, and who could find the patients A correction spans several parties.
- How the recall reached different organizations These carried the core work: identifying patients, discussing the update, performing it and documenting it.
- What the classification did not settle The 2017 recall resolved one set of vulnerabilities in one product line.
- Objections, in order The label could alarm patients, and the assessed exploitation risk was low.
- What a health system could check The 2017 recall left questions that apply to any future correction of a connected implant.
- With hindsight The FDA’s framework for device cybersecurity changed after 2017, though the postmarket guidance at the center of this case did not.
The week
On Tuesday, August 29, 2017, the Food and Drug Administration issued a safety communication about cybersecurity vulnerabilities in radio frequency-enabled pacemakers from Abbott, which the agency described as formerly St. Jude Medical [1]. It announced a firmware update to address them. The agency said it had approved the update on August 23, 2017, and that the update was intended as a recall, specifically a corrective action.
The communication named six device families: Accent, Anthem, Accent MRI, Accent ST, Assurity and Allure [1].
The agency described the risk in clinical terms. It said it had confirmed that the vulnerabilities, if exploited, could allow an unauthorized user to access a patient’s device using commercially available equipment. Such access could be used to modify programming commands, which could harm patients through rapid battery depletion or inappropriate pacing [1].
It also gave the scale and the record. The agency said there were no known reports of patient harm related to the vulnerabilities in the 465,000 affected devices in the United States [1]. The firmware update would be available beginning August 29, 2017. Pacemakers manufactured beginning August 28, 2017 would have the update preloaded.
The Department of Homeland Security’s Industrial Control Systems Cyber Emergency Response Team, known as ICS-CERT, published a medical advisory, ICSMA-17-241-01, with technical detail [2]. It said it was releasing the advisory in response to the FDA’s August 29 communication. It said a security research company had identified the vulnerabilities, and that a third-party security research firm had verified that the new firmware mitigated them. The advisory is now hosted by the Cybersecurity and Infrastructure Security Agency, in a version last revised on April 16, 2018.
The advisory listed three vulnerabilities. The first was weak authentication: the pacemaker’s authentication algorithm, involving a key and time stamp, could be compromised or bypassed by a nearby attacker. It carried a CVSS v3 base score of 7.5. The second was a lack of limits on correctly formatted “RF wake-up” commands, which could let an attacker drain the battery. It had a base score of 5.3 [2].
The third affected only the older Accent and Anthem devices. They transmitted patient information without encryption to clinic and home equipment, and stored optional patient information unencrypted. Its base score was 3.1. The advisory said exploitation required an attacker close enough for wireless communication. Exploitation had been publicly demonstrated, but exploit code was not publicly available [2].
Both documents described an update applied by a health care provider. The FDA said it required an in-person visit and could not be installed from home through the company’s remote monitoring network. The agency said the process took about three minutes, during which the device ran in a backup mode at 67 beats per minute [1].
The agency did not tell every patient to get the update. It recommended that patients and providers discuss the risks and benefits at the next regularly scheduled visit. It and Abbott did not recommend prophylactic removal and replacement of affected units [1]. The decision to install was left to the physician’s judgment for each patient.
The communication followed the agency’s December 2016 final guidance on postmarket management of cybersecurity in medical devices [3]. That guidance set out when a manufacturer’s cybersecurity change was a routine update and when it was a correction the FDA expected to hear about.
A security patch for a server is an IT task. A security patch for an implanted pacemaker is a medical procedure. The August 2017 communication showed what happens when the two meet. The FDA approved the firmware and called it a recall, a corrective action [1]. That label carried rules about reporting and records. The clinical recommendation carried a risk-benefit decision for each patient.
The classification was not inevitable. FDA’s own postmarket guidance of December 2016 said the majority of manufacturers’ cybersecurity actions were generally considered routine updates that did not need to be reported as corrections [3]. This one was treated differently. The reasons, and the consequences, tell device owners how to read later fixes.
Our argument is that treating the 2017 pacemaker fix as a recall moved a security patch into the device-correction regime, with its reporting duties for the manufacturer and its patient-by-patient decisions for clinicians. We take that regime in four parts:
- The classification: why this fix was a correction, not a routine update.
- The clinical decision: a risk-benefit choice for each patient.
- The delivery: an in-person procedure through the device clinic.
- The record: what each party had to document, and who could find the patients.
The first part concerns the regulatory category. The second concerns who decides whether a given patient gets the fix. The third concerns how the fix is physically applied. The fourth concerns the evidence trail that shows it happened. Together they cover the path from an approved firmware file to a patched device in a patient. After the four parts we look at how the recall reached different organizations, at objections, and at what a health system could check.
A note on sources and our reading. We describe the 2017 event from the FDA’s communication, which we read in an archived copy, and the ICS-CERT advisory. We describe the reporting framework from the FDA’s 2016 guidance and the regulation at 21 CFR part 806. Later developments come from the FDA’s current cybersecurity page. Statements about how a rule applies are our reading. Manufacturers and providers should confirm their obligations with counsel and regulatory staff.
The classification: why this fix was a correction, not a routine update
The FDA’s December 2016 postmarket guidance divided cybersecurity changes into two groups [3]. The first it called cybersecurity routine updates and patches. These were upgrades to increase device security, and updates or patches for vulnerabilities associated with controlled risk. The guidance said the FDA would not require advance notification or reporting of them under 21 CFR part 806.
The second group covered vulnerabilities with uncontrolled risk. The guidance defined uncontrolled risk as unacceptable residual risk of patient harm due to inadequate compensating controls and risk mitigations [3]. For those, it said manufacturers must report under part 806 unless certain conditions were met.
What part 806 covers
Part 806 requires device manufacturers and importers to report certain corrections and removals to the FDA [4]. A report is required when a correction or removal was initiated to reduce a risk to health posed by the device. It is also required when one was initiated to remedy a violation of the act that may present a risk to health. The report is due within ten working days of starting the action.
The regulation defines a correction as the repair, modification, adjustment, relabeling, destruction or inspection of a device without physically removing it from its point of use [4]. A firmware change to an implanted pacemaker fits that definition naturally. The device stays in the patient. Its software is modified.
The regulation also exempts some actions. Among them are actions to improve a device’s performance or quality that do not reduce a risk to health or remedy a violation [4]. In our reading, that exemption is what made the guidance’s routine update category possible. A patch that only improves security, without addressing a risk to health, falls outside the reporting duty.
Why the pacemaker fix fell on the other side
The FDA’s communication explained the risk to health directly. Exploitation could lead to rapid battery depletion or inappropriate pacing [1]. Those are harms to the patient, not to data. The firmware was developed, in the FDA’s words, to reduce the risk of patient harm. In our reading, those statements placed the fix in the category of corrections initiated to reduce a risk to health.
The conditions that could have avoided a report
The guidance described circumstances in which the FDA did not intend to enforce part 806 reporting for an uncontrolled-risk vulnerability [3]. There had to be no known serious adverse events or deaths. The manufacturer had to communicate with customers within 30 days of learning of the vulnerability, with interim controls and a remediation plan. It had to fix and distribute the fix within 60 days. And it had to participate actively in an information sharing and analysis organization.
The documents we cite do not say which of those conditions were met or missed in this case, or when Abbott first learned of each vulnerability. We do not speculate. What they show is the outcome: the FDA approved the update and described it as a recall [1]. The firmware also needed the agency’s approval, which the guidance said manufacturers should consider when changes might require a premarket submission [3].
Approval as well as reporting
Reporting and approval are separate questions. The guidance told manufacturers to evaluate whether a device change needed a premarket submission, such as a supplement to a premarket approval application or a 510(k) [3]. It said remediation of devices with annual reporting requirements, such as class III devices, should be included in the annual report. In this case the FDA approved the firmware before release [1]. A fix that needs agency approval cannot move faster than the review.
A responsibility the guidance called shared
The guidance described medical device cybersecurity as a shared responsibility among health care facilities, patients, providers and manufacturers [3]. The correction rules in part 806 put the reporting duty on the manufacturer alone. In our reading, this case showed both statements at work. The manufacturer reported and fixed. Providers and patients decided and applied. Facilities held the records and settings that made the fix possible.
Recall language and its signals
The word recall has weight with patients. It suggests a defective product. In this case, the FDA used it for a software fix to devices that remained implanted and working. The communication was careful to say there were no known reports of harm and that removal was not recommended [1]. The label described the regulatory mechanism, not a judgment that devices should come out.
That distinction matters for how organizations explain such events. A recall in the device sense can mean a correction applied in place. Clinicians and patient communication teams need to say so clearly, or the label will cause alarm the facts do not support.
What classification brought with it
Once a fix is a correction, several things follow. The manufacturer reports it. The manufacturer must keep records of corrections and removals whether or not they are reportable [4]. The guidance pointed to the FDA’s recall strategy provisions and effectiveness checks as references for remediation plans [3]. In our reading, those references bring an expectation that the manufacturer will track how far the fix has reached.
Reading the next notice
For a provider, the classification is a signal. A manufacturer notice that describes a routine security update, with no reduction of a risk to health, belongs to the guidance’s first group. A notice that describes a corrective action to reduce a risk of patient harm belongs to the second. The second kind brings the reporting and record expectations described here, and in this case a clinical decision as well. Reading which kind a notice is should be the first step in any response plan.
The clinical decision: a risk-benefit choice for each patient
A routine IT patch is applied to every system unless there is a reason not to. The FDA’s communication reversed that default. It recommended that patients and providers discuss the risks and benefits at the next regularly scheduled visit, and that providers determine whether the update was appropriate for each patient [1].
The risks on each side
On one side was the cybersecurity risk. Exploitation required an attacker within radio range and, according to the advisory, high skill [2]. There were no known reports of harm [1]. On the other side was the risk of the update itself.
The FDA reported Abbott’s previous experience with firmware updates. Reloading of the previous firmware due to an incomplete update had occurred at a rate of 0.161 percent. Loss of programmed settings had occurred at 0.023 percent. Loss of diagnostic data had not been reported. Complete loss of device function had occurred at 0.003 percent [1].
Those are small numbers. Applied to a population of 465,000 devices, they are not zero. In our reading, the FDA’s approach recognized that for a single patient, the chance of an update malfunction might exceed the chance of an attack. The right answer could differ from one patient to the next.
Factors the FDA named
The communication listed factors to weigh: pacemaker dependence, the age of the device and patient preference [1]. A patient whose heart relies on the pacemaker for every beat faces higher stakes if the update fails. A device near the end of its battery life may be replaced soon anyway. A patient may prefer to accept a small security risk over any procedure risk.
For pacing-dependent patients, the FDA advised considering the update in a facility where temporary pacing and generator replacement could be readily provided [1]. In our reading, that turned some updates from a clinic task into a hospital-capable procedure. The location of the update became part of the clinical decision.
Older devices and the encryption gap
The advisory drew a line between device families. The older Accent and Anthem devices sent patient information unencrypted and stored optional patient information unencrypted. The Assurity and Allure devices did not have that weakness. The update would stop unencrypted transmission on Accent and Anthem. The advisory also said Accent and Anthem were no longer sold in the United States [2].
For a clinic, that meant the benefit of updating differed by device family. A patient with an older device gained protection for both commands and data. A patient with a newer device gained protection mainly for commands. The age of the device, which the FDA listed as a factor, also bore on how close it was to replacement.
Devices near replacement
A pacemaker’s battery runs down over time, and the device is then replaced. A patient whose device was close to replacement could reasonably wait. The replacement device, if manufactured on or after August 28, 2017, would carry the new firmware [1]. For that patient, the update added procedure risk to protect a device that would soon be gone. The FDA’s mention of device age covered this case.
Consent and conversation
The FDA asked providers to give patients Abbott’s patient guide as part of the discussion [1]. That is a consent-style conversation about a software change. Clinicians needed to explain radio-frequency access, authentication and battery depletion in terms a patient could weigh.
Declining is a valid outcome
A patient and physician could reasonably decide not to update. The communication allowed for that. It means some devices stayed on the old firmware by deliberate choice. That is an unusual feature of this kind of recall. Its success could not be measured as a share of devices patched, because not every patient was meant to be.
The decision as a record
A risk-benefit decision should leave a record. If a patient declines, the chart should say so and why. If a patient accepts, the chart should record the update and its outcome. Without those records, a later question about whether a patient was offered the fix has no answer. The fourth part returns to this.
The delivery: an in-person procedure through the device clinic
The update could not be pushed remotely. The FDA said it required an in-person patient visit with a health care provider, and could not be done from home through the remote monitoring network [1]. The advisory said it was applied through the company’s clinic programmer [2].
Why remote delivery was not offered
The sources do not explain why the update required a clinic visit. In our reading, the procedure itself suggests reasons. The device entered a backup pacing mode during installation. Providers were asked to save settings and diagnostics beforehand and check the device afterward [1]. Those steps need a clinician present. A failed update at home would leave a patient without help at hand.
The procedure
The FDA described the steps. Providers should print or store the programmed settings and diagnostic data before updating, in case of loss. The update took about three minutes, with the device pacing at 67 beats per minute in backup mode. Essential, life-sustaining features remained available. After the update, providers should confirm the device worked, was not in backup mode and kept its programmed parameters [1].
Each step is a clinical task with a checklist. It takes clinic time, staff trained on the programmer and a room equipped to respond if something goes wrong. For a busy device clinic, adding a discussion and possible update to each affected patient’s visit adds load. Abbott’s letter to physicians of August 28, 2017 described the process, the FDA said [1].
The next regularly scheduled visit
The FDA tied the discussion to the next regularly scheduled visit [1]. That choice spread the work over the normal follow-up cycle instead of calling every patient in at once. In our reading, it reflected the absence of known harm and the high skill the advisory said an attack required. It also meant the fix would roll out over months as patients came in, not in a single campaign.
Staff who can run the update
The update ran on the clinic’s programmer under a clinician’s control [2]. Staff had to know the update path, recognize backup mode, restore settings if needed and confirm the result. A clinic with one trained person on that procedure had a single point of failure. Training more staff before the visits started spread the load and reduced the chance that an update was done by someone unfamiliar with recovery steps.
Scheduling as triage
A clinic could not see every affected patient at once, and the FDA did not ask it to. It could still order its work. Pacing-dependent patients needed a suitable setting, so their visits had to be planned with that facility. Patients with older device families had more to gain from the update. Patients whose next visit was far off might warrant an earlier call if they asked. In our reading, a clinic that sorted its list this way turned a passive schedule into a managed rollout.
New devices
Pacemakers manufactured from August 28, 2017 had the update preloaded [1]. For devices implanted after that date, the question did not arise. Clinics had to distinguish between preloaded devices and older ones, which required knowing each device’s manufacture date or firmware version.
The programmer and home monitor in the system
The update changed what the pacemaker would accept. After it, any device trying to communicate with the pacemaker had to provide authorization. The FDA said the clinic programmer and the company’s home transmitter would provide it [1]. The fix therefore depended on the surrounding equipment as well as the implant.
The FDA also said it would keep assessing the cybersecurity of Abbott’s implantable cardiac devices and its home transmitter [1]. In our reading, that signaled the implant was one component of a system. Clinics had to keep their programmers current, and patients’ home units had to work with the updated firmware, for the fix to function as intended.
The clinic as the patch channel
In IT, the patch channel is a network. Here, the patch channel was the device clinic’s schedule. Its capacity, its staffing and its ability to identify affected patients set the pace of remediation. In our reading, that made device clinic operations part of the security response, whether or not the clinic thought of itself that way.
The record: what each party had to document, and who could find the patients
A correction spans several parties. The manufacturer issues it. Providers apply it. Hospitals and practices hold the patients’ records. Each party documents a different piece, and no single party sees the whole.
The manufacturer’s record
Part 806 required a correction report with specified content, and records of corrections and removals [4]. The regulation’s definitions included consignees, meaning persons or firms that received, purchased or used a device subject to the correction [4]. For an implant, the consignee chain runs from manufacturer to hospital or clinic to patient.
The guidance referenced the FDA’s recall effectiveness check provisions in its discussion of remediation [3]. In our reading, a manufacturer tracking the effectiveness of this recall would want counts of devices updated, devices reviewed and declined, and patients not yet seen. Providers held the bulk of that information.
The provider’s record
The provider’s record holds the clinical facts: the discussion, the decision, the update if performed and the post-update checks. The FDA asked providers to save settings and diagnostic data before updating [1]. Those saved files, and the post-update confirmation, are the evidence that the procedure was done safely.
That record also protects the provider. If a patient later experiences a device problem, the chart shows whether the update was offered, accepted and verified. If an attack were ever alleged, the chart shows whether the device had been patched.
Saved settings as clinical data
The instruction to print or store programmed settings and diagnostic data before updating created a new record [1]. If an update lost settings, the saved copy let the clinician restore them. If it lost diagnostics, the saved copy preserved the history. Those files belong in the patient’s record, retrievable later, not on a programmer’s local storage where they may be overwritten.
Finding the patients
Before any discussion can happen, someone must identify which patients carry affected devices. A device clinic following its own patients has that list. A health system whose patients see cardiologists in several practices may not have a single list. An emergency department seeing a patient with an implanted pacemaker may not know the model.
The recall exposed an information gap. The implanted device is part of the patient’s record, but its model and firmware may sit in an operative note, a device clinic database or the manufacturer’s registry. Pulling them into a usable list is a data task that the recall did not assign to anyone in particular.
Firmware as a tracked attribute
Before 2017, a clinic may have treated a pacemaker’s firmware version as mainly a manufacturer’s concern. After the update, it became clinically relevant. A patient on the old firmware with no recorded decision was a gap. A clinic that recorded firmware versions could see its remaining exposure at a glance. One that did not had to interrogate each device at the next visit to find out.
Adverse event reporting
The FDA asked anyone who suspected or experienced a problem with these devices to file a voluntary report through MedWatch. It said health care personnel at facilities subject to the FDA’s user facility reporting requirements should follow their facilities’ procedures [1]. A failed update, a loss of settings or a device left in backup mode was a reportable kind of problem. Clinics needed staff who knew how to route such events.
Communication with patients outside the clinic
The FDA directed patients and caregivers to consult their physicians and gave a hotline and website [1]. Patients who heard the word recall in the news might call their cardiologist, their primary care physician or their hospital. Each of those needed a consistent answer. Without a shared script, patients could receive conflicting advice.
How the recall reached different organizations
Device clinics and electrophysiology practices
These carried the core work: identifying patients, discussing the update, performing it and documenting it. The FDA’s audience list named cardiologists, electrophysiologists and cardiothoracic surgeons [1]. A practice with a large population of the affected devices faced a sustained increase in visit complexity.
Hospitals and health systems
Health systems that implant pacemakers held the implant records. They also held the facilities the FDA mentioned for pacing-dependent patients. A system could coordinate across its practices by building one list of affected patients, setting a standard script and deciding where higher-risk updates would happen.
Ambulatory and primary care networks
Primary care physicians were in the FDA’s audience too [1]. They would field patient questions without performing updates. An ambulatory network needed a route to refer patients to the right device clinic and a consistent message on what the recall meant.
Clinical engineering and security teams
Hospital clinical engineering and information security teams typically manage equipment the hospital owns. Implanted pacemakers sit in patients, not on a hospital’s inventory. The update did not touch hospital networks. That put security teams in an advisory role. Their contribution was explaining the vulnerabilities and helping build the patient list, not running the patch.
Remote monitoring services
Remote monitoring was the channel that could not carry this update. The FDA said it could not be done from home through the company’s remote monitoring network [1]. Teams that review remote transmissions could still help. They see device data between visits and could flag patients still on old firmware, or patients whose device showed problems after an update.
Patients and caregivers
Patients faced a decision about a risk they could not see. The FDA addressed them directly and told them to consult their physicians [1]. For a patient, the practical questions were when their next visit was, whether they were pacing-dependent and where the update should happen.
Other device manufacturers
Manufacturers of other connected implants could watch how the FDA handled this case. It showed that a cybersecurity fix could be approved and described as a recall, with a clinical recommendation attached. In our reading, it gave the industry one model for how a serious device vulnerability might be resolved under the 2016 guidance.
What the classification did not settle
The 2017 recall resolved one set of vulnerabilities in one product line. It left several questions open for the next case.
First, it covered pacemakers only. The communication said it did not apply to implantable defibrillators or cardiac resynchronization defibrillators [1]. A patient with a defibrillator from the same manufacturer had no update under this action. Clinics serving both device types had to explain why one patient was offered a fix and another was not.
Second, it did not set a pattern for timing. The guidance’s conditions for avoiding a report involved 30-day and 60-day steps [3]. The documents we cite do not show how this case measured against them. Other manufacturers could not read from this case how fast they were expected to move.
Third, it did not say how a fix would be handled if exploitation began. The FDA’s recommendation rested on the absence of known harm [1]. If attacks had been reported, the balance between update risk and attack risk would shift. A clinic plan built for a routine pace might need to become a call-in campaign.
Fourth, it did not assign the patient-finding task. The manufacturer knew which devices it had shipped. Clinics knew their own patients. No document we cite gave anyone the job of reconciling the two across a region. In our reading, that gap is where patients could be missed.
Fifth, it did not show how the FDA would treat a fix that carried no procedure risk. This update needed approval, a clinic visit and a clinical decision. A future fix that could be installed safely and remotely might be handled differently. The 2017 case set one model, not the only one.
These open questions are not criticisms of the 2017 decision. They mark the limits of a single case as precedent, which matters for anyone building a response plan on it.
Objections, in order
1. Calling it a recall overstated the risk
The label could alarm patients, and the assessed exploitation risk was low. But the label followed the mechanism, not the severity. In our reading, part 806 turns on whether a correction was initiated to reduce a risk to health, not on how large that risk is [4]. The FDA paired the label with clear statements that removal was not recommended and no harm had been reported. The remedy for alarm is explanation, not a different label.
2. Leaving the decision to physicians slowed the fix
It did. A blanket instruction would have patched devices faster. It would also have exposed every patient to the update’s malfunction risk, including those for whom that risk outweighed the attack risk. The FDA chose individual judgment. In our reading, that was consistent with treating the fix as a medical intervention rather than an IT task.
3. Remote delivery would have been safer and faster
Remote updates avoid clinic load and reach patients who miss visits. They also remove the clinician from a procedure that carries a small risk of device malfunction. The 2017 update required in-person delivery [1]. Whether a future update can be delivered remotely depends on its risk profile and the device’s design. The objection argues for designing devices that can be updated safely at a distance, not against this decision.
4. Hospitals had no role because the devices sit in patients
The implants sit in patients, but health systems hold the records and the facilities. Finding affected patients, setting a standard message and providing a safe setting for higher-risk updates are institutional tasks. A health system that left them to individual practices accepted inconsistent outcomes.
The FDA’s audience list reflected this. It addressed cardiologists, electrophysiologists, cardiothoracic surgeons and primary care physicians, who practice across clinics and hospitals [1]. Only an institution that spans those settings can make their answers consistent. In a system with several cardiology groups, one patient list and one message are institutional products, not tasks any single practice can complete.
5. The vulnerabilities required skill and proximity, so the fix was unnecessary
The advisory rated the attack as requiring high skill and nearby access [2]. Those conditions limit risk; they do not remove it. Exploitation had been publicly demonstrated. A fix that is available and approved lets each patient and physician decide. Without it, there was no choice to make.
6. Security teams should own all device risk
One view holds that every connected device belongs in the security program. For hospital-owned equipment, that may be right. For implants, the risk decision is clinical, as the FDA’s recommendation made plain [1]. Security teams can assess vulnerabilities and inform clinicians. The decision to update a device inside a patient belongs to the patient and the physician.
What a health system could check
The 2017 recall left questions that apply to any future correction of a connected implant.
- Check whether the organization can produce, within a day, a list of patients with a given implant model and firmware version.
- Confirm device clinic records capture the firmware version of each implant at each interrogation.
- Confirm there is a documented process for recording a patient’s decision on a recommended device update, including a decision to decline.
- Identify which facilities can support updates for pacing-dependent patients, with temporary pacing and generator replacement available.
- Prepare a shared patient script for device recalls that explains a correction in place, for use by cardiology, primary care and call centers.
- Agree who in the organization receives manufacturer and FDA device cybersecurity notices and who decides on the clinical response.
- Review whether clinical engineering and security teams have a defined advisory role for implanted device vulnerabilities.
These checks treat an implant correction as a joint clinical and information task. The 2017 recall showed that the fix itself may be quick. The work around it, finding patients, deciding with them, applying the update safely and recording the result, sets the pace. A health system that has rehearsed that work will move faster when the next notice arrives.
With hindsight
The FDA’s framework for device cybersecurity changed after 2017, though the postmarket guidance at the center of this case did not. As checked on November 15, 2026, the FDA’s cybersecurity page lists the postmarket guidance dated December 2016 as final guidance [5]. The page describes it as providing recommendations for remediating and reporting vulnerabilities for controlled and uncontrolled risks of patient harm.
The larger change came from Congress. The FDA’s page says Congress granted the agency explicit cybersecurity regulatory authority in 2022 by adding section 524B to the Federal Food, Drug, and Cosmetic Act. The authority took effect on March 29, 2023 [5].
Section 524B applies to cyber devices. The FDA’s page gives the statutory definition in three parts. A cyber device includes software validated, installed or authorized by the sponsor, and has the ability to connect to the internet. It also contains characteristics that could be vulnerable to cybersecurity threats [5]. Manufacturers submitting such devices for premarket review must include information showing reasonable assurance that the device and related systems are cybersecure.
That information includes a plan to address postmarket vulnerabilities and exploits, with coordinated vulnerability disclosure. It includes processes to provide reasonable assurance of cybersecurity, including postmarket updates and patches. It includes a software bill of materials [5]. The page says the requirements do not apply to submissions made before March 29, 2023, but do apply to a new premarket submission for a change to a previously authorized device.
In our reading, those requirements answer the 2017 case from the design end. In 2017, the update path was decided after the vulnerability was found, and it ran through the clinic. Under section 524B, a manufacturer must describe its update and patch processes before the device is authorized. The question of how a fix will reach patients is asked at the start.
The FDA’s page also addresses the classification question this article began with. Asked whether a cybersecurity patch is a design change the FDA must approve, it says changes made solely to strengthen cybersecurity, with no other impact on the device, likely will not require a premarket submission. Manufacturers should assess whether a change could significantly affect safety or effectiveness [5]. In our reading, the 2017 pacemaker update, which the FDA approved before release, sat on the other side of that line.
The premarket guidance has kept moving. The FDA’s page lists Cybersecurity in Medical Devices: Quality System Considerations and Content of Premarket Submissions, updated to address section 524B, as a version dated February 3, 2026 that supersedes one issued on June 27, 2025 [5].
The page adds a note on health care delivery organizations. It says the FDA recognizes they may need to patch or update devices or supporting infrastructure, and recommends working closely with manufacturers on such changes [5]. For implanted devices, that collaboration still runs through the clinic and the patient’s physician.
What held up is the two-track structure. A cybersecurity fix can be a routine update or a correction, and the classification decides what follows. What changed is that, for devices submitted since March 2023, the manufacturer must plan for updates and patches before authorization. The clinical side has not been automated away. A fix for an implant can still require a conversation, a visit and a record for each patient.
Spotted an error? Tell us and we will check it and correct the article where needed.