Post-market clinical follow-up within post-market surveillance
Post-market clinical follow-up (PMCF) collects clinical data on a CE-marked device for as long as it is on the market, to confirm its safety and performance, to keep its identified risks acceptable and to detect emergent risks and off-label use. It is the proactive part of a manufacturer's post-market surveillance, beside the reactive vigilance system that runs on a different clock, and it draws on sources that range from the literature and user surveys to what users post on social media.
What post-market clinical follow-up is
Post-market clinical follow-up is part of post-market surveillance: the process of collecting clinical data on a CE-marked device to confirm its safety and performance throughout its expected lifetime after market approval. It is also a means of determining the acceptability of identified risks and of detecting emerging risks, by gathering long-term data from a larger patient population than is possible during device development. It is similar to the post-approval studies for medicines.
MEDDEV 2.12/2 defines post-market clinical follow-up as proactive, organised methods and procedures to collect clinical data, with the objective of confirming the clinical performance and safety of a device throughout its expected lifetime and the acceptability of identified risks, and of detecting emergent risks on the basis of factual evidence. The EU Medical Devices Regulation (MDR) gives it much the same aims: to proactively collect and evaluate clinical data, to confirm safety and performance, to ensure the continued acceptability of identified risks and to detect emergent risks. Set side by side, almost all of the points in the two texts are more or less the same.
The follow-up is carried out through two documents, both part of post-market surveillance: the PMCF plan, a document the MDR provides for, describes the methods and procedures for collecting clinical data on the CE-marked device, and the PMCF evaluation report presents, analyses and evaluates the results.
Proactive and reactive surveillance
Post-market clinical follow-up should be covered in the manufacturer's post-market surveillance system, where it represents all the proactive surveillance activities and vigilance represents mainly the reactive ones. Both post-market surveillance and post-market clinical follow-up are covered by the manufacturer's quality management system. The reactive part covers complaints, nonconformities, findings and recalls, each of which needs to be documented and taken forward.
Safety reporting, or vigilance, and post-market surveillance work to different timescales once a device is on the market. Safety reporting is done on an urgent, emergency basis. Its trigger is usually a single occurrence that meets serious criteria, most often a serious health issue or a death, with a cause in the device, and the response serves both to protect the patient and to report to the authorities so that further harm is not done. Post-market surveillance looks at serious and non-serious events alike over the whole lifespan of the device, and then reports. It leaves time to act as things happen and to report at a later phase, and it runs continuously throughout the year and throughout the life of the device. Its authorities are most often the notified bodies and sometimes other health institutions, and its documents, such as the PMCF or post-market surveillance reports, depend on the device's classification or risk class.
The documents of this reporting, among them the PMCF documents, post-authorisation safety studies, the post-market surveillance report and the clinical evaluation report, come with timelines that vary with the risk classification and the authority, from one or two days to 15 or 20 days.
From the directive to the regulation
Post-market clinical follow-up was already required under the Medical Devices Directive, through its amendment 2007/47/EC, with guidance in MEDDEV 2.12/2 revision 2, and a PMCF plan and report were already required then. The clinical requirements did not change with the MDR, but the procedure for demonstrating compliance did. MEDDEV 2.12/2 bases the requirement for PMCF studies on the identification of possible residual risks, unclarity on long-term clinical performance, or both, that may affect the benefit-risk ratio, and it cites examples of when a PMCF study might be justified. The guidance is not legally binding, though, and non-compliance with it could not be a reason for non-compliance with the directive, so manufacturers could decide that a PMCF study was not necessary and that their approach to PMCF was acceptable. The MDR makes clear that post-market clinical follow-up is a requirement, not an option.
What sets the MDR apart from the directive and from MEDDEV 2.12/2 is its focus on post-market clinical follow-up as a continuous process. Annex XIV Part B of the MDR treats it as a continuous process that is used to update the clinical evaluation and should be addressed in the manufacturer's post-market surveillance plan. It is not something a manufacturer does once, but data collection throughout the device's lifetime and the time it is on the market.
Under the directives, complaint monitoring and literature searches were often sufficient, and under the MDR a more proactive approach is required. Manufacturers that were perhaps reactive under the directive are asked to be a lot more proactive in collecting PMCF data.
In keeping with its continuity, the process is cyclical: the manufacturer plans how PMCF data will be collected, collects them, analyses them and prepares the reports, then uses what was learned to update clinical documentation such as the risk management, design, labelling and clinical evaluation report, and feeds the same learning back into the PMCF plan for the next round of data collection.
Sources of proactive data
Proactive surveillance draws on literature searches, searches for clinical studies, real-world data, technical support, customer surveys and social media, which the regulations ask for. For example, a post on LinkedIn, Facebook or TikTok saying that a continuous glucose monitor is causing redness is surveillance data, and picking it up lets the next version of the monitor avoid the problem. The law is slowly starting to ask for collection at that level.
User surveys, data collection from registries and PMCF studies are examples of how manufacturers can collect clinical data for CE-marked devices. No one data-collection approach works for all devices, and many different methods can be taken.
Manufacturers can pay for data sets from electronic health records, provided by companies that gather, process and analyse health data from international and local markets, and the data sets can include adverse events, patient feedback, social media reporting, product performance and information about competitors. They are an option for collecting safety and performance data for devices with a long market history, for which data from the scientific literature or from clinical investigations are often limited. The data can be difficult and cumbersome to collect, may not be easy to analyse together and may need a per-case analysis. Before unique device identifiers are adopted, a device may be hard to identify unambiguously, and the data sets have to include information relevant to the device's safety and performance parameters, which may not be possible depending on the device and the information available.
Surveys, especially when distributed online, can be a good way of quickly reaching large numbers of patients or healthcare professionals, and a well-structured online survey can quickly generate useful data. Like post-market studies, user surveys should be based on a predefined endpoint and a statistical rationale. Not all users will complete a survey, so the data may be incomplete and not representative of all users.
Social media listening
The EU MDR 2017/745 made clear that post-market surveillance would require significant changes, and white papers and conferences showed the need for surveillance activities different from the traditional ones. As a result, social media first appeared as a term under proactive surveillance, in an industry that had used it mainly for digital marketing and recruiting staff.
For post-market surveillance, social media means social media listening: the monitoring of public digital conversations on the internet to retrieve and understand customers' opinions about a brand, a product or both. Social media is not only platforms such as Facebook or Instagram but also forums, blogs, portals, community and microblogging sites, digital magazines and newspapers, online television and virtually every website on the internet.
Listening is not a new concept: as people began using social media more often to discuss their personal lives, health and illness, social media gained recognition as an essential source of information in the health sector. Boston Children's Hospital's HealthMap, for example, was founded in 2006 as a platform that uses informal online sources for disease outbreak monitoring and real-time surveillance of emerging public health threats. The pharmaceutical industry, notably ahead of the device industry, recognises listening as a pharmacovigilance tool to detect adverse events and adverse drug reactions. Projects such as WEB-RADR and Vigi4Med have explored the value of social media for identifying adverse events, the FDA sponsors projects with specific patient forums to detect them, and data mining has been used to detect adverse drug reactions from Google and Yahoo search logs. So regulatory authorities expect listening as a proactive surveillance activity in the device industry. One post-market surveillance specialist knew of few or no publications applying it to devices.
Several software packages, among them Talkwalker, Brandwatch, Hootsuite, Sprout Social and Synthesio, support device manufacturers in making listening less time-consuming, and some are also used for pharmacovigilance and can report adverse events to competent authorities. They work on one principle: a query on keywords such as a brand and product name pulls posts and comments, called mentions, from the internet. Data can be gathered for a set period or in real time, mentions can be limited to particular countries and languages, alerts can be created, and frequent posters or influencers can be identified. Mentions can be classified by sentiment analysis, and sentiment analysis algorithms may help identify adverse events, but many mentions are still classified as neutral.
Experts on pharmacovigilance describe advantages and challenges of using social media to detect adverse events that apply to the device industry. The advantages are:
- large amounts of real-time, real-world usage information
- geographic and population diversity
- direct access to patient perspectives
- the opportunity for niche studies
- identification, through healthcare professionals' posts, of institutions for potential PMCF studies
- the potential to cover the known under-reporting of adverse events
The challenges are:
- conceptual: the value of the data; the level, quality and credibility of the information; the risk of misinformation
- environmental: compliance; the regulatory framework
- technical: data mining and analysis; large amounts of data, with a significant amount of noise expected; duplicate reports from parallel posting; text classification, for example of colloquial language and misspellings
Scholars agree that social media data could be a tool to augment post-market surveillance, but that much work must still be done to overcome the challenges. Listening may not suit every device, since its utility depends on the device's nature and market share. Reluctance to explore it in the device industry seems to come from the overwhelming number of adverse events, but its many potential benefits make it worth exploring.
Listening monitors patients' opinions on a device as they state them publicly on social media or other online channels, and patients may, for example, express their views, good and bad, after having a prosthesis implanted. The data are supposedly unbiased because the manufacturer did not request them, but negative results are likely to be reported more often than positive ones. Listening needs continuous monitoring, unsolicited information can be difficult to collate and analyse, and a device that is low risk and also well established may draw very few comments.
Some minimum criteria are needed before social media mentions can be imported into surveillance databases with confidence. These primary considerations could be recommended for establishing listening as a surveillance activity:
- A clear purpose should be established for the listening.
- Queries should be clearly defined and fine-tuned, including the number of queries and the keywords for each query.
- A systematic process should be established, covering, for example, the social media searched, data crawling, analysis of mentions, alerts, automated reports and team roles.
- The elements required for a valid adverse event should be defined: an identifiable reporter (user name, handle or email), an identifiable patient (the reporter, or someone the reporter knows of who experienced the event), an identifiable brand and an identifiable adverse event.
- Reportability requirements must be established, specifically on how to proceed when information is incomplete.
- Rules for following up with posters should be defined, and follow-up should not be a requirement where social media is used as a complementary surveillance activity.
- Adequate documentation should be created to outline the process for recording and archiving primary source data from social media.
- Data protection laws, ethical standards and regulatory compliance should be considered.
Implant cards could make listening easier, because more accurate information on patients' knowledge and awareness will be gathered. Device manufacturers and regulatory authorities should expect to work more closely to reach agreements and to educate the sector on the value of listening in post-market surveillance.
Social media listening could eventually become a standard source of adverse event reporting, and one editor of device writing expected that some writers may soon be integrating its outcomes into clinical evaluation reports and PMCF plans and reports.
Off-label use
Identifying possible systematic misuse or off-label use is one of the aims of the PMCF plan. Case reports may reveal misuse or off-label use of a device, and reviewing them is a good way of identifying it.
Devices can seem designed so specifically for one condition that off-label use is hard to imagine, but one writer who worked as a physician confirms from that experience that there is a lot of off-label use in hospitals. Stone baskets, for example, are designed to retrieve stones from the urological or biliary system, and they are often used to retrieve foreign bodies: when a guide wire fractures during a vascular intervention and part of it stays in the patient's body, many physicians use such a device to retrieve it. Most stents and stent grafts are labelled only for treating, for example, stenoses or occlusions in a specific vasculature, but many physicians also use them for a tear or an aneurysm in a vessel; this is feasible, but the devices are usually not labelled for that indication.