Introduction The environment surrounding medical device regulation in the United States has always been rigorous, but recent events including well-publicized quality issues associated with implantable medical devices have raised the scrutiny to which devices are examined prior to product approval and monitored for performance post approval. European requirements are undergoing a similar change in rigor as evidenced by the change from Directive to Regulation. The safety of medical products should be paramount in any determination regarding whether a product should be marketed. The US FDA is charged with evaluating safety and effectiveness as part of its mission to ensure the safety of public health as well as fostering access to new and innovative therapies. The current FDA climate requires significant data and scientific evidence to support approval for commercialization of new device on the US market. This article is intended to describe the types of data required by FDA and provide recent examples of how FDA is interpreting the regulations differently than in the past. Clinical Data to Support FDA Submissions FDA is requiring supportive clinical evidence for many more submissions, including both PMAs and 510(k)s. The PMA (premarket approval) route to market is typically reserved for the highest risk medical devices and requires a product be proven safe and effective on its own merit, whereas the 510(k) route to market allows moderate risk products to leverage experience from a product that has been on the US market. A 510(k) submission must demonstrate that the new product is substantially equivalent to another product that has been on the market since before 1976, when the medical device regulations were implemented. The need for supportive clinical evidence is a relatively new trend for 510(k)s that appears to have been triggered by heightened public scrutiny of this route to market. As a result, FDA reviewers are less willing to accept bench testing or logic to demonstrate substantial equivalence of the new device to the predicate. In addition, the amount of clinical evidence required to support US FDA submissions is increasing. FDA expects well-controlled clinical trials that minimize typical sources of bias and compliance with good clinical practice (GCP) standards, including a formal protocol defining the study specifics, ensuring adequate protection of human subjects, and establishing standard operating procedures governing processes to control all aspects of study performance, data integrity, and data reporting. Hierarchy of Clinical Data When considering the need for clinical data to support FDA submissions, the randomized controlled trial is the gold standard. This type of trial offers the most credible data because the design ensures that most sources of bias are minimized. In this type of design, all subjects meet the same inclusion and exclusion criteria, are treated and followed at the same centers by the same health care professionals, during the same time period. Therefore, the only real difference between the standard of care group and the investigational therapy group is likely the treatment itself or statistical chance. 1
While the randomized controlled trial results in the most credible data, there are times when it might not be feasible, or ethical to randomize subjects. For example, when companies were developing endocardial leads for use with implantable cardioverter defibrillators (ICDs), there were physicians who felt it was not ethical to perform a thoracotomy to implant subjects with epicardial leads (which are placed on the outside of the heart through an open heart procedure) when those subjects could otherwise be treated with the endocardial lead (which is placed via an intravascular approach). Although the performance of the endocardial lead was unproven at the time of the study, there was sufficient experience with implantation of this type of lead with pacemaker devices and furthermore, testing of the lead was performed with each ICD procedure to ensure its effectiveness for its intended use. In this situation, credible data could be collected at clinical sites using the investigational endocardial leads with comparison to clinical sites using the standard of care epicardial leads. All subjects would meet the same inclusion and exclusion criteria as defined by the protocol, the same type of data would be collected from all subjects, and all subjects would be followed in the same manner. This type of study design is considered a prospective controlled trial. The sources of bias introduced in this type of study include the potential for site treatment or follow-up differences, statistical chance and, of course, differences due to the treatment. In the hierarchy of clinical evidence, the next most credible data would come from prospectively collected data from a single arm study compared to retrospectively collected data from the same subject, or matched control data selected from an available historical data set. In this type of study design, the control group data would be available in line listing form so FDA could reproduce the data analyses. The sources of bias introduced in this type of study design include differences in time and treatment standards, differences in site treatment or follow-up, statistical chance and of course differences due to the treatment. Moving lower on the clinical data hierarchy, then, would be single arm studies compared to historical data from literature reports. Literature represents some difficulties for comparison purposes because non-devicerelated and non-procedure-related safety issues are typically not reported in a manner similar to current clinical practices. Therefore, it is likely that differences will be seen in safety measures. Also, bias could potentially be introduced from the same sources identified previously. Note that although these data have the potential for bias, the data still may represent valid scientific evidence, and may be very useful for publication or for the advancement of scientific knowledge. Similar clinical evidence can be collected from data registries and meta-analyses from literature reports. These sources of clinical data again offer scientific evidence, but because sources of bias have not proactively been controlled, the level of credibility is less than what is typically required by FDA. To address FDA s need for higher levels of valid scientific evidence, we recommend that Sponsors design a clinical study to provide the highest level of credible clinical data as possible, balancing the need to be practical for clinical investigators. At study completion, highly credible data is a very valuable tool for regulatory purposes, for reimbursement purposes as well as for marketing purposes. 2
Statistical Data to Support FDA Submissions Another emerging trend in FDA review of submissions is the need to demonstrate statistical significance to prove clinical effectiveness of a new treatment. The reviewing clinicians at FDA rely on the statisticians to interpret the data and justify statistically whether the stated claims and intended use can be supported by the endpoints and analyses of the resulting data. During the clinical study design phase, objectives, performance endpoints, and the statistical sample size are determined a priori based on expected assumptions. If after the study is completed a statistical endpoint is not met, it may be because there was not enough subjects studied, or the assumptions used to develop the statistics might have been flawed or the product might not perform better than the standard of care. In these cases if a statistical justification cannot be supported the FDA and their medical reviewers will try to assess if the data support a clinical justification. The current trend in FDA reviews is that clinical justifications are generally insufficient in the absence of the statistical support. FDA and FDA s Advisory Committees have requested additional clinical studies be conducted, often with a new design or new assumptions in cases such as these. Even when FDA is involved in providing input into the clinical study design before the study is initiated, there is, of course, no guarantee that the study will yield statistically significant results upon its completion. There have also been cases where a study did meet its defined endpoints, but in the final analysis FDA and its medical reviewers felt that the defined endpoint was not sufficiently clinically relevant and required further confirmatory studies. Therefore, it is important to work with knowledgeable clinicians to understand the expected performance of the new therapy, determine conservative statistical assumptions, and ensure FDA is supportive of the study design. A pilot study is often helpful in providing early evidence upon which to base performance and statistical assumptions. These efforts will be helpful in determining the best possibility of success for the study for all parties. The Importance of Intended Use The key element when developing the FDA strategy is the Intended Use, or the statements made in the labeling regarding how the product is intended to be used by the healthcare professional. It is this language that will drive the evidence FDA will need to support an approval/clearance decision for commercialization of the product in the US. For example, a device filter on a cardiovascular catheter intended to prevent thrombus from embolizing to prevent stroke would require a significant clinical trial to demonstrate a difference in stroke events in a population of subjects with the filter versus a population of subjects treated without the filter. The same device filter intended for use in preventing embolization to the extremities and limbs would require a different study. The same device filter used with a cardiovascular catheter for general infusion use might only require bench data. The Intended Use statement is also a critical element in determining the regulatory route to market, particularly whether a product can be authorized for market via a PMA or a 510(k) route. Where a similar product is already on the market and is classified as Class II (falling into a moderate risk category), a new product can follow the 510(k) path by using an identical Intended Use statement as that of the predicate device. New Intended Use statements will automatically eliminate the option to pursue a 510(k), requiring a PMA path for high risk devices or a new De Novo process for moderate and low risk devices. 3
De Novo Process Until earlier this year, FDA law required that any new devices that did not fall into the 510(k) category were automatically classified as Class III requiring a PMA. US regulations, however, acknowledged that some of these devices would be low to moderate risk and as such did not need to be subject to the high level of scrutiny and resource requirements for a PMA submission. FDA issued a guidance document that allowed an alternative to the automatic Class III designation which was called a de novo application. In these cases, sponsors submitted a 510(k), received a not substantially equivalent decision, and within 30 days submitted a de novo application providing information to demonstrate that the product presents only a low to moderate risk to public health. The de novo process also required FDA to issue a guidance document for future devices that will use the subject device as a predicate. When FDA granted a clearance decision, the product was placed in Class II. Earlier this year, US Congress passed a new law that codified the de novo process and made it a single step submission no longer requiring the first step of a standard 510(k) application. It remains to be seen what FDA will expect to see as part of the submission to incorporate the information typically included in a 510(k) as well as the risk/benefit assessment. But it is clear that FDA is being very strict about 510(k) predicates and intended use language, thus sending more devices through the de novo process. FDA does prefer that companies who plan to pursue the de novo pathway request a pre-submission meeting to discuss the available data and allow FDA an opportunity to comment on information they would expect to see in a de novo application. Companies should be prepared to present their plans rather than expecting FDA to make recommendations. This pre-submission process can require 2-3 months to schedule and conduct the meeting, but the feedback can be valuable in understanding FDA s expectations when topics are controversial or it is unknown whether FDA will accept a proposed plan. FDA will often expect clinical data in support of de novo applications, particularly in situations where the intended use has been expanded beyond that of a predicate product. For example, a product that has a general indication, but for which a specific indication is being sought, would need clinical data support for the specific indication. Also, if a product has significant differences in technological characteristics from a predicate product, clinical data may also be required to demonstrate the safety profile of the product with those differences. The de novo submission process, although new, offers a faster alternative route to the US market compared to the PMA submission. Companies need to be aware of the requirements and solicit feedback from FDA to meet their expectations, in order to find the shortest path through the regulatory hurdles. Leveraging Other Sources of Data In the product development process overall, many types of data are generated including bench and mechanical testing, biocompatibility testing, pre-clinical animal testing, pilot clinical data, clinical data that is collected in specific countries to meet their regulatory or reimbursement needs, health economics data, post-market clinical follow-up and post-approval studies, etc. When developing a US regulatory plan, all types of data should be considered as to what will be useful in demonstrating safety and effectiveness, substantial equivalence or a positive risk/benefit analysis. For a typical 510(k), mechanical and bench data often are sufficient. For new and 4
novel products, FDA typically requires a small pilot study before allowing progression to a large scale pivotal trial. In many cases, companies can leverage clinical experience from outside the United States in lieu of a pilot study. The EU and FDA are now requiring post-market surveillance monitoring which usually consist of passive data collection on reported adverse events (e.g., MDR and vigilance reporting) and active data collection in the form of post-market studies and clinical follow-up. Additionally, third-party payors are requesting post-approval data to demonstrate that the product continues to perform as expected when used in all patients, in all healthcare settings. Companies should assess whether a single study design can be used to collect data to serve all the post-market clinical needs for US, EU and reimbursement purposes. Significant cost savings can be realized by setting up and conducting one study meeting many needs instead of managing multiple studies. Conclusions FDA s interpretation of the regulations is changing and has resulted in heightened requirements for clinical data and statistical relevance. Where FDA previously exercised some flexibility in evaluating Intended Use language and determining the regulatory route to market, they are now taking a strict view and holding firm to a narrow interpretation of the regulations. As such, companies wishing to enter the market in the US should develop a regulatory plan that includes sufficient data support and specifically, clinical evidence to support stated claims for the product. These requirements are not just new for the US market, similar requirements are being mandated for European regulatory needs, but are also useful for reimbursement and marketing as well. A well-designed, well-conducted clinical study can be leveraged for multiple purposes, but will particularly facilitate the submission process in the US when the data are supportive. Investment in clinical evidence to support regulatory applications in general and US applications in particular is funding well-spent, especially in light of FDA s current environment. 5