Scientists, pharmacists, and engineers may encounter several classes of challenges while developing and seeking regulatory approval for a novel drug. One challenge is understanding the uncertainty surrounding the development and evidence for regulatory compliance of a new drug. As no two patients are the same, scientists and engineers need to know how the drug functions under multiple operating conditions with an acceptable level of confidence. Other uncertainties can come from variations in dosage or changes to the original chemical formulation. These variations lead to uncertain performance and may result in compromised parts or decreased service life. Other challenges include the expense of running clinical trials or meeting stringent government regulations.
When faced with these challenges, simulation has become the great enabler. Simulation allows scientists and engineers to determine the optimal dosage for a given population and better understand sensitivities and uncertainties without extensive and costly clinical trials. However, simulation comes with its own risks and challenges.
In addition, modeling and simulation have been at the forefront of every major news story during the COVID-19 pandemic. For public health organizations, modeling and simulation are critical for predicting the spread of COVID-19. For pharmaceutical companies, modeling and simulation are critical for accelerating the development of vaccines.
The Food and Drug Administration (FDA) recommends the use of modeling and simulation for the development and regulatory evaluation of pharmaceuticals and medical devices. Examples of where the FDA advises the use of simulation and modeling include:
The FDA continues to work with the industry to further develop new modeling and simulation methodologies that will provide a path towards innovation and rapid introduction of new life-saving technology to the public.
Though simulation and modeling come with many advantages, MIDD has its fair share of challenges, including the quality of the data used in developing the models, lack of consistent application to critical drug development and regulatory decisions, and how to handle data from disparate sources like in vivo and clinical trials. Two major challenges are the computational cost of accounting for the many variations in a given population and establishing simulation model validity. SmartUQ’s predictive analytics and Uncertainty Quantification toolset can reduce the cost of performing in silico trials and increase the value of the information gained.
Statistical calibration is the process of tuning calibration parameters in a simulation to better fit reality while also quantifying the uncertainty in the model and suggesting areas of improvement to reduce the overall uncertainty. Using SmartUQ’s toolset, this process starts with building an efficient human clinical design for calibration and validation. SmartUQ offers specialized DOEs such as simulation to clinical trials that will take the DOE of the simulation and build a complementary clinical trial DOE.
Model validation is the process of determining the degree to which a model is an accurate representation of the real world from the perspective of the intended uses of the model. The ideal validation is predictive validation which can be difficult to do under uncertainty. This involves determining the best method to interpret all the information about uncertainties coming from input uncertainty distributions, discrepancy maps, and uncertainty in the emulator variance. For example, a PBPK simulation may be used to predict the absorption rate of an encapsulated drug. The in vivo data from experimentation can be used in conjunction with the simulation to determine where the simulation can reliably predict data, as well as to quantify the uncertainty of the predictions. This can help determine whether the new drug configuration will meet IVIVC standards, even under uncertain conditions. Additionally, sensitivity analysis can be performed as validation “sanity checks” throughout the validation process to see how local and global sensitivities affect the output(s) of interest.
