After devices enter the market, it is typical for manufacturers to monitor incidents and complaints and to determine if there are performance issues that are occurring at a greater than expected rate— that expected rate having been previously judged to be below an acceptable threshold. Or, an acceptable rate judgment might be newly made, especially for unanticipated problems. What, then, must follow is a sequence of steps related to investigation, determining a fix if indicated, and rolling out that fix. There must also be an ongoing consideration about what, if anything, to do about product already in use or in the pipeline. All of these technical and business decisions have a time element associated with them that must also be managed, with the general objective to move as quickly as reasonably possible.
Product issues are typically identified either from external reports or internal quality controls. For external reports, there will often be some number of incidents that occur that are noted to be more than was anticipated or more than has historically occurred.
The first step in addressing problems is to have an effective process to record, note, and trend issues, whether they are direct complaints, MDRs, published papers, other field feedback, or information from other sources. What number of events triggers attention, and then action, is a local decision, which is however subject to being second-guessed.
Complaint handling is a regulated process (21 CFR 820.198) but it must be remembered that regulations set a minimum standard and that additional or parallel activities can be undertaken at a manufacturer’s discretion. Among other things, 820.198 requires that
each manufacturer shall review and evaluate all complaints to determine whether an investigation is necessary. When no investigation is made, the manufacturer shall maintain a record that includes the reason no investigation was made and the name of the individual responsible for the decision not to investigate."
When there is an investigation, there must be documentation that includes the nature and details of the complaint; the dates and results of the investigation; and any corrective action taken.
Notable here is that each complaint requires an active decision, and that not doing anything about the complaint also requires documentation. In addition, according to 820.198:
Any complaint involving the possible failure of a device, labeling, or packaging to meet any of its specifications shall be reviewed, evaluated, and investigated, unless such investigation has already been performed for a similar complaint and another investigation is not necessary.”
This clause seems to limit the ability to decide not to investigate in the case of a deviation from specifications. That there has been a prior investigation is an interesting consideration in that it establishes that there is an already known problem. In this regard, I once saw a commonly reported failure mode lead to the inclusion of a new checkbox on an evaluation form to save the investigator the time of writing down the reason for the failure. One might argue that deciding to preprint a failure mode suggests that more effort is needed in prevention rather than simply making the recording easier. Complaint handling must also be an input in reporting to the FDA under MDR. Complaint handling is a frequent citation in inspection 483’s, partly driven be complaint handling being an inspection focus. In this regard, it has been observed that collecting and filing complaints, along with trending, is not the same as doing something effective with them.
Other than the use of the term “promptly” in the MDR clause, no timelines are given in the regulation, yet time can be a significant factor when an investigation should or does lead to later action, such as ceasing distribution and undertaking a recall or other field action. People injured by a device that was later recalled will want to know where their injury is on the timeline from trigger point to action and if their injury was contributed to by unnecessary and/or deliberate delays. This is the "what did you know and when did you know it" question, which may be supplemented with what should you have known and when should you have done something.
Investigations of the cause of a problem may fall under Corrective and Preventive Action (CAPA). The CAPA regulation (21 CFR 820.100) states that there must be a process for
analyzing processes, work operations, concessions, quality audit reports, quality records, service records, complaints, returned product, and other sources of quality data to identify existing and potential causes of nonconforming product, or other quality problems.”
There must then be the identification of “action(s) needed to correct and prevent recurrence of nonconforming product and other quality problems”. The FDA Inspection Guide for CAPA notes that the rationale for undertaking a CAPA should be examined along with the investigational procedures including the depth of the investigation and prior history of identifying and acting on root causes.
As with complaint handling, the CAPA regulation sets a minimum, if vague, level of activity, but any other appropriate action at appropriate speed remains a local decision. The answer to the question “Why didn’t you undertake an investigation or a CAPA?” should never be “Because we weren’t required to.”
An investigation timeline can be challenging in that there are a number of factors that influence how long an investigation might take once it has begun. One factor is how technically difficult it is to determine, first, what is the cause of a problem, and second, how to correct it. In this regard, problems might arise from the original design or manufacturing, or changes that might have been made in either. Such changes might be intentional or inadvertent, and if intentional, they may not have been adequately validated. Manufacturing changes that design doesn’t know about, and which have not been validated, is a classic form of lack of communication between the two. Incoming materials and components might also have changed, with or without the manufacturer knowing. Manufacturing issues are linked to inspection and test. If there was an adverse manufacturing change why didn’t inspection and test detect it?
The use of the product might also have changed. For example, a wider range of applications might be occurring including off-label uses. Some of those applications may be more problematic than earlier uses, and they may also not be reflected in prior design and risk management considerations. Known and problematical off-label use might have to be addressed with additional instructions and warnings. In some causes there might also have to be a reduction in off-label promotion.
There might also be a wider range of users, especially when early adopters are joined by others with perhaps lower skill or training. However, a simple, and repetitive conclusion of “user error” ignores the well-established role of design, instructions, and training in mitigating errors.
Solving a problem also requires the application of adequate investigatory resources, and insufficient resources is likely to delay a resolution. Are the people involved working many problems, too many problems, or only a few? In either case, how hard are they pursuing each? Is there a sense of urgency or one of leisure? This is also a management issue of how closely anyone is watching the number and resolution of open problems (whether CAPAs or not). It also requires the establishment of acceptable effort and pace. On the other hand, the objective is not just to “close” CAPAs or open fewer of them as suggested by some advice, but to actually solve problems.
Existing Product and Production
What to do with products in the pipeline, and when, is a technical, business, and sometimes ethical question. The pipeline may include products currently being manufactured, products on the shelf or in the distribution chain, products already in customer’s hands, and, in some cases, products already implanted in people. Those in customer’s hands might be more or less permanent equipment, or they may be single use. If there is a problem with a permanent product, that problem will persist until a solution is found and implemented. For single-use devices the problem will disappear over time provided no new devices with the same problem are produced and distributed, including when the old problem appears in redesigned products. This might invite model changes whose purpose is more making old problems appear to go away rather than actually fixing them.
Technically, the pipeline decision will depend on the discovery of the cause of the problem and the identification of a proposed fix, with some fixes subject to regulatory review. Business decisions, perhaps influenced by risk assessment, might be driven by potential loss of income, loss of market share that might not be recovered later, negative spill-over into other product lines, and other effects of adverse publicity. Here you must remember that risk assessment is highly subjective and otherwise imperfect. For example, allowing low probability to offset high severity inherently suggests that it is “ok” if only a small number of people are dying. In this regard, a decision based on avoiding bad publicity now might result in even worse publicity later, and such a decision might also be addressed during litigation. In some cases, an argument for keeping a product on the market might be based on a claim that the public would suffer if a product became unavailable. However, in some cases the public might be well-served by a product becoming unavailable. Moreover, there are relatively few devices that are so unique that lack of the device being a public health problem would actually be true. Such an argument for a Class II device might be particularly complicated by the past assertions that it was substantially equivalent to other legally marketed devices. This raises the more basic question of whether or not a substantially equivalent device can be claimed to be significantly better than its predicates, and, if so, better in what verified and validated ways.
Pipeline decisions might include halting production, halting distribution, and/or undertaking a recall (or other similar field action, even if calling it a recall is avoided). Halting production or distribution but not undertaking a recall can be particularly problematic, especially if the apparent reason is to avoid getting stuck with unsellable inventory. Even more problematic is when a solution is implemented such that newly produced product is (presumably) free of the problem, but the old product continues to be sold off. A recall that seems to illustrate this situation occurred in August 2017 when intra-aortic balloon pumps manufactured before June 2013 were recalled. It was disclosed that a design change had been implemented in June 2013 such that those pumps manufactured after that date did not have the problem that led to the recall. Therefore, it appears that the manufacturer knew about a problem, fixed the problem for new units, but did nothing about old units for four years. A similar case arose for an orthopedic device for which it was noted that the recalled product was manufactured before September 2011 and sold between October 2008 and May 2015. Thus, sales continued for 3.5 years after manufacturing ended. Information made public also showed that devices manufactured after September 2011 had design enhancements that increased the strength of the device. Therefore, beginning in September 2011, the manufacturer began making stronger devices but was still selling the already manufactured less-strong devices. This might be hard to explain to someone who suffered from a failure of the old weaker device when that device was implanted in them after the new stronger device was already being manufactured.
The identification of design and/or manufacturing issues in currently marketed medical devices, as well as investigating these issues and correcting them, is an ongoing challenge. There are no universal formulas that tell you what to do and how fast to do it. Whether you are forthrightly dealing with these issues might benefit from the “smell test,” or the do-you-want-this-publicly-reported test. What customers and users of medical devices would prefer—if not necessarily expect, given some prior bad examples—is that these device issues be quickly identified and aggressively pursued, and that they be notified in a timely manner. On the other hand, at least some companies have on occasion followed a “don’t tell” strategy coupled with “sell off the bad stuff.”