OneMedRadio: Aptiv Solutions and a Novel Approach to Trial Design

How can one optimally design a clinical trial program for a medical device for a FDA clearance/approval submission? Typically, a clinical trial program involves a pilot study that aims to establish basic safety information, and possibly a preliminary estimate of effectiveness, followed by a similar pivotal trial intended to comprehensively address safety and effectiveness of the device in the intended use population. Aptiv Solutions (formerly MDCI) is a leading medical device CRO and consulting firm, focused on delivering strategic regulatory planning and support, FDA and international regulatory submissions, global clinical trial design and execution, and unmatched quality assurance and compliance services to address FDA QSR and ISO 13485. Aptiv Solutions takes an integrated approach and uses in-house medical device professionals to help clients respond to the increased rigor demanded by worldwide regulatory agencies for medical device development, approvals and ongoing compliance.

OneMedRadio spoke with Dr. Vladimir Dragalin and Dr. Michael Feldstein of Aptiv Solutions to discuss the changing tides of medical device regulatory and the advantages of using an adaptive design trial. For the transcript, see below.


Matthew Margolis: Greetings from OneMedRadio, I’m Matt Margolis. Today, I’m with Vladimir Dragalin and Michael Feldstein of Aptiv Solutions. Dr. Dragalin and Dr. Feldstein are Senior Vice Presidents for the full service CRO. Thank you for joining us gentlemen.

Vladimir Dragalin: Thank you for inviting us.

Matthew Margolis: So firstly, Vlad, let’s start with you. What is an adaptive design?

VD: An adaptive design is defined as a multistage study design that uses accumulating data to decide how to modify aspects of the study without undermining the validity and integrity of the trial. So this means that an adaptive design learns from accumulating trial data in real time and applies this knowledge to optimize subsequent study execution. So these adaptations are usually prospectively defined prior to the start of the trial and can include different aspects. Stopping early either for futility or success, expanding the sample size to great than what’s expected because of data variability, allocating patients perhaps preferentially to treatment regimen that show a better therapeutic index. So virtually any aspect of the trial may be a potential target for design modifications, but what is important…this modifications should be a design feature aimed to enhance the trial not a remedy for inadequate planning or poor planning. They are not ad hoc study corrections made via protocol amendment, which are used in the industry.

MM: So how does the Aptiv Solutions technology make clinical trials more efficient?

VD: So we do have integrated platform for executing clinical trials. That’s IT or information technology platform for running clinical trials. But we also do have a full service coverage for support of our sponsors in implementing adaptive clinical trials. So we do have expertise in our innovation center, which constitutes experts in methodology of adaptive clinical trials. We do have experience in running these clinical trials. We have so far more than 100 adaptive clinical trials already run for different clients starting with the very big pharma and going up to small biotech companies. And we do have the processes and SOPs in place, which were specifically developed in order to cope with the challenges and complexity of adaptive clinical trials.

MM: Now, Michael, let’s shift gears a little bit. What is the state of medical device regulatory?

Michael Feldstein: Well two words come to mind, Matt. One is tentative and the other is cautious right now at least if we’re talking about the FDA. It’s taking longer. It’s taking longer for the FDA to review and to make decisions. This has had an effect on sponsors who wish to go through the FDA process. We are hearing from time to time, more often than we used to, that sponsors are first now going to Europe for CE mark rather than deal with the FDA in its present state of mind.

MM: And so how can the technology from Aptiv actually be an advantage in the medical device field, and you know, how can it kind of combat this current difficult state of FDA regulatory?

VD: I think that definitely there are challenges. There are challenges with our regulators that Michael just mentioned, but also I think that developers of drugs recognize that the conventional way of drug development is not sustainable. So this is why they are looking towards adaptive design as innovation in drug development in order to improve the situation. I think that it’s recognized that adaptive designs in medical devices can provide an opportunity for example to calibrate initial assumptions of the trial design based on this partial information. You may get the let’s say pilot studies or even pivotal study faster, perhaps less expensive sometimes, increase likelihood of success through either early determination of the trial because you already have evidence that device is efficient, or maybe to stop the development of that device because you have evidence that is not going to be efficacious or effective.

MM: Now, Michael, how have the appropriate study endpoints shifted in recent years and what additional clinical data has the FDA sort of required for intended use for medical devices in clinical trials?

MF: Well the first thing to keep in mind is roughly speaking, about 10% of 510k submissions, premarket notification submissions, require clinical data and what we’re seeing recently is a tightening up of definition of what will be an allowable predicate, which is a device to which you claim that you are substantially equivalent. In the past, the FDA has allowed rather innovative approaches to defining predicate devices. We’re seeing some evidence on their part that they’re going to be looking more strictly at the definition of predicate so that one can’t be as innovative in putting together four or five devices, which taken together perhaps constitute a predicate. That’s certainly one area where we are seeing a kind of tightening up if you will.

Another area has to do with endpoints, which are subjective. The FDA is looking for measurement of objective endpoints through the use of validated instruments. So they are not very predisposed to the use of instruments, which are developed specifically for a device and which haven’t been tested rigorously in the field and there are lots of studies where subjective endpoints are important especially for example in the field of cosmetology.

I think finally the last thing I would say is that the FDA is certainly going to be looking to the 510k process to tightening it up generally speaking and I think we may see that 10% figure rise going forward.

MM: It’s very interesting. So, Vlad, I want to dive a little bit deeper into your technology and what are Bayesian methods and when should they be used?

VD: Okay. Bayesian methods is just one methodology, statistical methodology, which allows this use of adaptive clinical trial. So you can say Bayesian designs are a subclass of adaptive designs. Why they are so important and why they are used? Because they put a formalism for learning under uncertainty. So it’s a good tool in order to recognize that whenever you start a clinical trial for a device, let’s say a pilot study, you start the study under some assumptions. Now, the Bayesian formalism allows you to put on that assumption some kind of it’s called a prior. So your knowledge is not just one number about the effective of your device, but you recognize that you perhaps have some interval of that effect. It’s not just one number.

And then that methodology of Bayesian updating of the information taking into consideration new data is exactly suited for this online learning, which is nothing but adaptive methodology. So you use what you started with as a prior, you collect the data, and now you update your knowledge, your prior and that becomes now your so-called posterior. So you do this repeatedly in different stages and from stage to stage, you accumulate and update your knowledge. So that’s what is in the cornerstone of these Bayesian methods and why they are so appropriate if you like for being used in adaptive clinical trials.

So through the modeling, mathematical modeling, you can use the information on the same patient from previous time points. Through the hierarchical modeling, you can use the historical data perhaps on the device, which is the previous version of the device you are investigating now. So this methodology allows you to implement or to use the knowledge you have of the previous version of the device in designing your trial for the new version of the device.

MM: And so can you talk a little bit of the process that biostatisticians use in designing clinical trials and analyzing the clinical data?

VD: Yes, that’s a very interesting question. So this is exactly how statisticians as a profession in drug development can contribute to enhancing this drug development. So usually, you go through several steps in developing let’s say a design for a clinical trial with a device. Start with defining study objectives, what will be the endpoints, what will be the measurements you will be measuring on the patients in the study in order to estimate the treatment effect. What is the target for adaptation? Like I said, adaptive design may have different components and you have to specify and agree with the clinical team what targets for adaption will be there.

What is the study population? So after you kind of set up these objectives, then you look into what are the adaptive options, what are the options for adaptation, what design options you have. You can do a conventional design or you can do a simple adaptation or you can do a more complex adaptation. In order to compare and contrast these design options, usually a statistician will evaluate those in a simulation study. So simulate kind of in silico the performance of different patients, which you later will enroll in your trial, but you do it in silico. Then on different simulation scenario, you run all these different design options and then you collect the performance metrics and compare and contrast different options. Taking your decision then which option is the best, you have to go now in the final round where you can do a kind of sensitivity analysis of that design when you simulate data from some deviation of the initial assumptions to see if design will be appropriate for real situations. So that is how a statistician then kind of fine-tunes the design, which will then be taken in the implementation of a clinical trial.

MM: So Michael, I actually want to circle back to how the regulatory agencies are supporting adaptive trials, so what guidance documents have these regulatory agencies been publishing for adaptive trials?

MF: Well there are really three relevant guidance documents. There could be others, but the three most important are one that’s emerged in August of 2010, I believe, in final form design consideration for pivotal clinical investigations for medical devices. This is a general guidance document and actually it doesn’t speak to the notion of whether something should be Bayesian or not and it doesn’t speak to whether something should be adaptive or fixed design. But it is this foundational guidance document that anybody who’s going to embark upon a clinical investigation really needs to be familiar with.

Then there are two other guidance documents that are most relevant to your question, Matt. The first is one that is a guidance for the use of Bayesian statistical methods in medical device trials. That came out in February of 2010. I suppose you might consider a companion one, but it’s not really in devices, it’s a draft. It’s in draft form. It also came out in February of 2010 and it’s a draft for adaptive design clinical trials for drugs and biologics.

I think those three taken together would give a pretty good framework of ideas for a sponsor as to how they would proceed. It’s important to keep in mind first of all and the reason I say to take together is that we are seeing more and more combination products where we have device, drug, biologic device, that kind of thing. So these guidance documents in some sense are companions to one another. What I was going to say was that it’s important to keep in mind that today, roughly speaking, the Center For Diseases And Radiological Health, which sees these submissions for devices, only about 5% to maybe 8% of all their review time is spent on Bayesian submissions. So we’re looking at something, which is a very small portion of all of the time that is spent by CDRH in reviewing submissions. Nevertheless, there is a very concerted effort on the part of CDRH and other agencies at the FDA to use adaptive methods to use Bayesian methods when they are appropriate.

MM: That was a discussion of adaptive designs specifically as an advantage in the medical device field with Vladimir Dragalin and Michael Feldstein, senior vice presidents of Aptiv Solutions. With OneMed Radio, I’m Matt Margolis signing off.

The comments are closed.