Use of Computer and In Vitro Modeling Techniques during the Development of Pediatric Circulatory Support Devices

Pediatric mechanical circulatory devices are currently being developed by several programs. Along with the engineering and testing challenges to be considered is the regulatory process which will be used to review applications to the Food and Drug Administration (FDA) to initiate clinical trials with these devices. This paper considers the appropriate use of computer and in vitro modeling data as a part of an FDA Investigational Device Exemption submission for a pediatric circulatory support device. Initially, the types and value of modeling techniques that may be used are discussed. How the unique physiology, flow requirements, and anatomy of pediatric patients may affect modeling efforts is then reviewed. Modeling assumptions, justification, and validation are key aspects of modeling data to be submitted as a part of an FDA submission. Finally, key recommendations are made for the appropriate use of modeling data for an FDA clinical trial submission. Introduction The Pediatric Circulatory Support Program of the National Heart, Lung, and Blood Institute (NHLBI) is currently sponsoring five contracts for the development of pediatric ventricular assist systems.1 In February, 2008, the NHLBI convened a meeting of the five funded contractors. At this meeting, the attendees updated the field on their progress, and attendees met to discuss the critical, unmet needs in the area of pediatric cardiovascular device development. In addition to a discussion of clinical trial design challenges due to the limited number of pediatric patients available for participation in studies, a separate group also discussed engineering challenges and barriers for development of these devices. Aspects of how clinical trial submissions for these devices will be reviewed by the Food and Drug Administration (FDA) have been given on-going consideration.2,3,4 Although computational and bench-top modeling techniques have been extensively used in the development process by each of the contract program participants, their importance in a regulatory submission remains unclear. As these devices progress toward a clinical trial, the appropriate use of information generated by the modeling techniques for an FDA submission needs to be defined for the contract participants and other developers of pediatric blood pumping systems. Computer and in vitro modeling techniques are valuable tools in the process of medical device development. These techniques can be used to guide the device design process and make initial predictions of device performance. Furthermore, these techniques can be used to evaluate the responsiveness of proposed control algorithms, predict in vivo performance, and provide guidance prior to conducting pre-clinical animal experiments. While these tools can generate useful information for the device developer, due to multiple inputs and variability in programming of modeling techniques, the data provided by such tools may or may not be useful to assess reasonable safety. Why Model? As a developmental process, in vitro or computational modeling can provide verification of intended functions of device operation that may not be possible to determine by other methods. As a part of the modeling process, by trying to create or recreate an accurate representation of the device, one gains a more complete understanding of flow-through regions, recirculation areas, and/or areas of stasis. Furthermore, modeling may allow for visualization or understanding of device function, when bench-top testing does not suffice. This can lead to valuable predictive capability. Before a model of a medical device is created, all of the components that make up the model and how they relate to each other physically and functionally need to be understood. As the model is constructed from its parts, a better understanding of the function of the fully assembled device can be gained. Computational modeling is used when measuring a difficult parameter, such as determining the blood flow path and the gas transport between oxygenator fibers or the immeasurably small flow field between a blood pump’s impeller blade tips and the housing. Modeling can help to identify the critical regions of interest that affect the redesign or safety evaluation of the device. Many types of models may be used in the development of a pediatric circulatory support system. Computer-based mathematical models, such as computational fluid dynamics (CFD), are useful in defining and optimizing the design and blood flow path through the device. Computer-based models of cardiovascular system function are useful to test the responsiveness of blood pump control strategies. Computer-based visual models, such as 2-D and 3-D graphical reconstruction of anatomy, are useful in guiding the external configuration design and placement of the device so that it will be able to fit appropriately into the body and still function as needed. Computer-based statistical models can be useful in establishing sub-system and system reliability and for performing a Fault Tree Analysis, a Failure Mode Effects Analysis, and making a Mean Time between Failures prediction. Physical models, such as bench-top in vitro flow loops and mock circulations, provide an initial evaluation of the device pumping performance with clinically relevant cannulae and the consequence of inter-action with blood, such as hemolysis and leukocyte and platelet activation and depletion. If the results of the in vitro tests meet the performance specifications, then it is reasonable to proceed to animal model testing. If the results are insufficient, it may indicate that there are aspects of the blood pump development process that were inadequate, including any modeling steps along the development path. The finding of insufficient blood pump performance may serve as the basis for an iterative re-design process. Matching the Model to the Patient Device developers and designers recognize the challenges of establishing appropriate design parameters for the model when representing a pediatric patient. Understanding the intended patient population when modeling the device establishes that it is distinctly different from a device that may be intended for use in adults. Yet, further clarification of performance requirements is necessary as pediatrics covers a broad range of patient ages and sizes. Consequently, modeling efforts need to use parameters appropriate for the intended pediatric population. For example, the NHLBI is interested in funding programs that are designing devices that specified a patient size of 2 to 25 Kg, sizes that range from newborn to approximately 8 years old. The flow requirement needed to support infants and small children is different than adults. Typical flow requirements for pediatric patients are: neonates 200–250 ml/kg/min, infants (7kg to 15 kg) 150ml/kg/min and children (>15 kg) 100 ml/kg/min. Adults typically require between 60–80 ml/kg/min. Given the patient size range specified for the pediatric contract program, these flow requirements translate into a flow range of 0.5 L/min to 2.5 L/min. It is not likely that any one device and clinically appropriate cannulae configuration could cover that entire flow range, so it is important to specify the desired flow range over which the modeling is to be conducted. The anatomy relevant to the device, such as the size of the left ventricle, aorta and intrathoracic space (which will influence the size and shape of the pump and cannulae) is also going to vary greatly in this range of patient sizes. For example, the diameter of the aortic arch in a newborn with normal anatomy is approximately 7 mm, but in an 8 year old, it can be 16 mm, whereas in an adult, it can be 25 mm or greater. Again, it is important to use dimensions appropriate for the intended pediatric population when conducting a modeling study. Another consideration is that rheological properties of neonatal and infant blood are different from those of adults and may vary greatly depending on the disease state. Neonates have larger red blood cell volumes and surface areas while the plasma viscosity tends to be lower than adults. However, pediatric patients with cyanotic congenital diseases (e.g., hypoplastic left heart syndrome) requiring Norwood, Glenn, or Fontan procedures tend to have higher hematocrits and viscosities than other disease states, resulting in a wide range of whole blood viscosity that needs to be taken into account when modeling pump performance. A final consideration for modeling is that the heart rate of pediatric patients is greater than that in adults and that the blood pressures may be quite different than adults, especially if congenital heart defects are present and depending on the specific nature of the cardiac dysfunction that necessitates circulatory support. These differences in pressures and heart rates may influence the cannulae and pump inflow and outflow conditions used in a device performance modeling effort. Considerations for the Use of Modeling Data in an FDA Investigational Device Exemption (IDE) Submission It is important to recognize that any modeling data presented should be directly applicable to the version of the device being considered for clinical trial. Data from earlier modeling efforts may be useful in the description of the developmental design process to the final version of the device. However, this data may not be directly applicable to the consideration of whether the final device is reasonably safe, unless it can be shown that there has not been any change in that aspect of the device design or performance since the modeling effort was conducted on a previous version of the device. Any computer modeling presented to the FDA should be linked to device risk assessment and have a clear and obvious goal. The results should be presented in the context of in vitro and in vivo preclinical testing. Computer modeling performed for the demonstration of reasonable safety may have different (but overlapping) goals than that performed for device design, development, and virtual prototyping. Whenever modeling is used, it is imperative to identify the advantages and limitations of each modeling technique. The following questions are important for consideration: (1) What are the assumptions that are made in the development of the model, and what are the boundary conditions applied? In submissions to the FDA, the model assumptions need to be provided and supported with scientific justification. (2) How has the natural variability of key parameters, such as blood viscosity, been treated in the model? In other words, how closely do the model inputs and results match clinical operation of the circulatory support device? (3) How does the modeling data support reasonable assurance of device safety? The device should also be modeled over its full range of performance capabilities (e.g., low, normal, and high flow rates under appropriate backpressure conditions). (4) Will residual blood flow through the heart affect the device’s performance? What happens if the device stops or reverse blood flow occurs? (5) What is the variability in the important physiological parameters that need to be examined? (6) How sensitive are the model’s results to changes in the input parameters (i.e., perform parameter sensitivity testing)? (7) How relevant is the turbulence modeling used to device and blood characteristics? (8) In all situations, model verification (of the quality of the software) and validation (how well the model replicates the experiment) is required and critical to assess the applicability of the computational modeling. The validation of CFD results may require more careful scrutiny when the conclusions go beyond the mechanical aspects of a device. For example, the prediction of hemolysis rates from wall shear rates or the prediction of oxygen uptake from flow distribution may not be straight-forward relationships, but rather geometry, situation, and fluid-composition specific. While each case must be made on its own merits, interpretation of the CFD results in terms of non-mechanical outcomes will require greater corroboration and more caution. A major limitation of current CFD computations that applies to most cardiovascular devices is the difficulty of inferring blood cell trajectories, particularly movement across streamlines and toward and away from boundaries, in both laminar and turbulent flows. Conclusions based on the assumption that cells follow predictions of how homo-genous fluid elements move require especially careful corroboration. One situation where modeling may be useful is examining the effect of a proposed change in device design before committing to the manufacturing and assembly of the modified device for verification testing. The modeling may help the FDA to assess the rationale and the justification for the design change, as well as to verify that the change accomplished the desired result without introducing other undesirable results as a consequence. In that regard, the accepted, revised computer model can also serve as a quality assurance tool for adherence to the indicated design change. Key Recommendations (1)Any modeling effort for pediatric blood pump devices needs to accurately represent the relevant anatomic structures and dimensions that will be involved with the function of the pump and cannulae. Additionally, the desired pump flow range and physiologic blood pressure, blood viscosity, heart rate, and other physiologic factors unique to pediatric patients, that will affect the device performance, need to be considered. (2) Any modeling data should be directly applicable to the version of the device being considered for the clinical trial. (3) Any modeling data should be linked to device risk assessment/analysis. (4) The model used to generate data should be verified and validated for the specific intended use. (5) The advantages and limitations of a specific modeling technique used to generate data should be identified. Acknowledgment The efforts of this working group were supported by contracts from the Pediatric Circulatory Support Program of the National Heart, Lung, and Blood Institute. Working Group Members and Report Authors: George Pantalos,1 Zhongjun Jon Wu,2 Guruprasad Giridharan,1 Sonna Patel,3 Jean Rinaldi,3 Sandy Stewart,3 Prasanna Hariharan,3, Qijin Lu,3 Michael Berman,3 Branka Lukic,4 Steve Deutsch,4 Patrick Cahalan,5 Edward Leonard,6 Keefe Manning,4 James Antaki,7 Dave Paden,8 Rich Malinauskas,3 William Smith,9 Lyle Mockros,10 Trevor Snyder,2 Greg Johnson,5 J. Timothy Baldwin11 1University of Louisville, 2 University of Maryland, 3 U.S. Food and Drug Administration, 4 Pennsylvania State University, 5 Ension, Inc., 6 Columbia University, 7 Carnegie Mellon University, 8 Launchpoint Technologies, Inc., 9 Cleveland Clinic Foundation, 10 North-western University, 11 National Heart, Lung, and Blood Institute