How to Use Modeling and Simulation to Optimize a 505(b)(2) Application

A 505(b)(2) application is a hybrid of a full new drug application and an abbreviated new drug application. The terms 505(b)(2), 505(b)(1), and 505(j) all refer to sections of the federal 505 code that governs the applications for new and generic drugs. Modeling and simulation is a broad field that uses mathematical models to describe and predict the actions of drugs and can be utilized across almost all types of regulatory applications. Modeling and simulation can be particularly useful for efficiently filling in some of the requirement gaps to support and optimize a 505(b)(2) application which uses new and previously approved information.

505(b)(2), 505(b)(1), and 505(j)

A 505(b)(1) is a traditional full new drug application (NDA), which requires all the mandatory safety and efficacy data. All of the data must be original and there is no option to use previously approved safety and efficacy data for this type of application because it is being evaluated for the approval of a new drug. A 505(j) is an abbreviated new drug application (ANDA), where the sponsor only needs to demonstrate that the drug candidate is the same as the previously approved innovator drug and has sufficiently similar pharmacokinetics (PK). An ANDA is used for the approval of generic drugs where all the safety and efficacy data are inferred from the innovator drug, also known as the reference listed drug (RLD). In this type of application, no new safety and efficacy data has to be created. A 505(b)(2) is considered a hybrid of an NDA and an ANDA because new safety and efficacy data and data from the original RLD can both be referenced. The new drug candidate is not intended to match the PK of the RLD exactly. Thus, some new safety and efficacy data must typically be generated, but not the complete safety and efficacy package that is required for a full NDA. Some reasons to submit a 505(b)(2) application include pursuing a new indication or new route of administration. The degree to which existing safety and efficacy data can be inferred from the RLD and how much additional data must be generated is a matter for detailed discussions with the agency.

Role of Modeling and Simulation for 505(b)(2)

Modeling and simulation can be beneficial for many different types of 505(b)(2) programs. However, it is difficult to be very specific about the role that modeling and simulation can play in a 505(b)(2) as the requirements for having new safety and efficacy data versus inferring from existing data varies greatly for each program. The most straightforward example of how modeling and simulation can benefit a 505(b)(2) application is when developing a modified release formulation of an already approved drug. If a detailed exposure-response relationship is known for the RLD, it can be used to estimate safety and efficacy for a new formulation. However, the exposure-response relationship needs to implicitly include all the relevant physiological processes. Commonly, exposure-response relationships connecting AUC and drug response are available for RLDs. One may (falsely) assume that for a drug with linear PK and if the driver of the drug’s exposure-response relationship is AUC, then giving twice the dose of a drug once a day would be expected to result in the same exposure (as measured by AUC) and efficacy as taking a single dose twice a day (if the response is dependent solely on AUC). However, this is rarely true. Taking this example to the extreme, using an AUC-based exposure-response relationship, one could conclude that a patient will have the same efficacious response when taking a drug twice a day for 7 days as taking 14 times the dose once a week. In both of these scenarios, the AUC-response is not physiologically accurate and cannot capture the impact of other biological processes that drive the efficacy of the drug. A more physiologically “correct” model will be needed to leverage existing efficacy data in a 505(b)(2) application and the type of model used will vary for each program.

Benefits of Modeling & Simulation to Optimize 505(b)(2) Applications

Modeling and simulation can provide a great deal of support for a 505(b)(2) application by leveraging existing data. This can greatly reduce the costs and time to market for a 505(b)(2) development program. The two main modeling approaches used to support a 505(b)(2) application are a full pharmacokinetic/pharmacodynamic (PK/PD) model or a physiologically based PK/PD (PBPK/PD) model. A full PK/PD model links plasma concentration to some receptor occupancy, which is then linked to some relevant clinical endpoint. These models are usually relatively simple, such as a group of indirect response models, but can be more complex. These models are data-driven and typically require a large amount of subject-level source data to develop. However, once such a model is established and qualified, it can be used to predict safety and efficacy for new dosing regimens or routes of administration. A similar approach can be used to support 505(b)(2) applications with PBPK models. PBPK models can be connected to PD models, like indirect response models or quantitative system pharmacology (QSP) models to determine response. Once a PBPK/PD model is established and qualified for the RLD, it can then be used to extrapolate the safety or efficacy for different dosing regimens, release profiles, or routes of administration. PBPK/PD models describe biology and pharmacology in much greater detail than a full PK/PD model approach and are able to predict drug responses over a wider set of scenarios. PBPK/PD models cannot be constructed by simply using numerical optimization methods and need to be developed and fit to the specific drug being tested. This process is more resource-intensive than a typical full PK/PD model.

Obstacles to 505(b)(2) Application Approval

As is common, the biggest obstacle to applying modeling and simulation to a 505(b)(2) program is the amount of available subject-level data. Ideally, the sponsor has access to the data from the reference application. However, a small company typically would not have access to the source data (often from a larger company) that is needed to establish a full PK/PD model. Frequently, the literature or a summary basis for approval (SBA) will describe a model in enough detail to reproduce it and creating a model from source data may not be necessary. This is particularly true for PBPK/PD models, as the physiology is frequently well described in the published literature.

Conclusions

Modeling and simulation is a key tool in leveraging existing data to support a 505(b)(2) application. By establishing a link between concentration and efficacy or safety (or both), different release profiles, formulations, or routes of administration can be explored. The extent to which these extrapolations can be used in lieu of generating new safety and efficacy data is to be discussed with regulators. Oftentimes, if the model is well supported by data and is biologically plausible, costly efficacy trials can be avoided for a 505(b)(2) application with the help of modeling and simulation approaches. Allucent has a wealth of experience supporting 505(b)(2) applications as well as experience with the regulatory processes and discussions with regulators on how to best leverage existing data with modeling and simulation. Contact us today for modeling and simulation support for your 505(b)(2) application.