Pharmacometrics: A Drug Development Paradigm
By Pierre-Olivier Termblay
While the term pharmacometrics has become fashionable since the start of the 21st century, the methods it encompasses have evolved from early pharmacokinetics (PK) work (circa 1920) to full-blown clinical trial simulation, thanks largely to enhanced computational capability and the work of many insightful minds.
Whatever the computational intensity or method, the purpose of pharmacometrics is best expressed in the International Conference on Harmonisation (ICH)-E4 Guideline titled, “Dose-Response Information to Support Drug Registration”:
“Knowledge of the relationships among dose, drug concentration in blood, and clinical response (effectiveness and undesirable effects) is important for the safe and effective use of drugs in individual patients. This information can help identify an appropriate starting dose, the best way to adjust dosage to the needs of a particular patient, and a dose beyond which increases would be unlikely to provide added benefit or would produce unacceptable side effects.”
In today’s fast-paced drug development environment, data is produced at an exponential rate. Efficient extraction of knowledge from the data to support development and decision making is becoming a cornerstone of success. Pharmacometrics specifically aims at extracting this knowledge and identifying the degree of uncertainty associated with it through the use of quantitative methods.
Armed with such knowledge, pharmaceutical companies can make enlightened decisions regarding the next development steps. At the regulatory level, this knowledge is helping to render decisions on approval, make dosing recommendations on the product label, and evaluate the need for further studies.
Although there is no one universal definition of pharmacometrics, all definitions currently in use emphasize the application of quantitative methods to characterize and predict a system’s behavior, whether at the individual, disease, or clinical trial level. Perhaps a good description of pharmacometrics, if not a formal definition, is one offered by the FDA: “Pharmacometrics is an emerging science defined as the science that quantifies drug, disease, and trial information to aid efficient drug development and/or regulatory decisions.”
The FDA’s description has the advantage of clearly emphasizing pharmacometrics’ role in drug development and is not limiting in scope. Nevertheless, pharmacometrics is generally understood as a collection of model-based approaches used to (1) extract from data and organize our understanding of a system’s behavior in a concise manner and (2) do so in a language (i.e. mathematics) that allows simulation of the system output. These models can be divided into three broad classes: (1) exposure-response models that specifically describe the relationships among dose, drug concentration in blood (or another matrix), and clinical response (effectiveness and undesirable effects), as stated in the ICH-E4 guidance document; (2)disease models that, as the name implies, aim to describe disease progression; (3) clinical trial models that describe patient demographics, adherence, dropout rates, trial structure, and so on.
While exposure-response models are by far the most commonly used pharmacometrics applications in the pharmaceutical industry today, the future should see greater use of clinical trial simulation applications.
Pharmacometrics In The Regulated Industry
Although from the above discussion it may sound as if pharmacometrics is something brooding deep within universities’ walls, it is, in fact, a lively discipline both at the regulatory level and within pharmaceutical companies. Several regulatory documents address the use and need of pharmacometrics in drug development. The ICH, followed by the FDA, produced scientific guidance on exposure-response studies emphasizing the use of concentration-effect relationship modeling in individualizing therapy, in preparing dosage instructions, and in providing primary evidence of effectiveness. The FDA and European Medicines Agency (EMA) guidances on population pharmacokinetics outline specific provisions on the conduct, context, and use of nonlinear mixed-effect modeling in drug development and regulatory submissions. One pertinent document is the FDA guidance titled, “Providing Clinical Evidence of Effectiveness for Human Drug and Biological Products,” published after the Food and Drug Administration Modernization Act (FDAMA) of 1997 to clarify what constitutes “confirmatory evidence” of effectiveness. The guidance makes provisions for the use of extrapolation from existing studies to support effectiveness. In this regard, pharmacometrics is an undeniably potent tool.
Finally, the 2009 FDA guidance on the end-of-Phase 2A (EOP2A) meetings for sponsors of investigational new drug applications (INDAs) closes the loop by unequivocally advocating the use of pharmacometrics in drug development and regulatory evaluation: “The purpose of an EOP2A meeting is to facilitate interaction between FDA and sponsors who seek guidance related to clinical trial design employing clinical trial simulation and quantitative modeling of prior knowledge (e.g. drug, placebo group responses, disease), designing trials for better dose response estimation and dose selection, and other related issues.”
Pharmacometrics In The Development Phases
While it is not uncommon for late-phase trials to incorporate pharmacometrics analyses, the full potential of pharmacometrics is achieved when data is modeled from the start of development at the nonclinical level, up to the confirmatory phase 3 and postmarketing trials. During the nonclinical development phase, a major concern is the prediction of human exposure-response behavior from animal data. Two modeling strategies often applied to circumvent this problem are allometric scaling and physiologically based modeling. The former approach is based on the relationship that exists between animal size and metabolic rate and is generally used to extrapolate appropriate doses to humans from other species data. The second strategy, physiologically based modeling, works by dividing the system into organized anatomical compartments and using physiologically oriented parameterization; this approach has intrinsic, between-species, scaling proprieties that allow for exposure-response modeling in various species.
In early clinical trials (e.g. Phases 1 and 2A), several doses will generally be tested, and safety data will be collected as well as preliminary efficacy data. This stage often involves the quantification of dose dependence in PK, along with the effects of food, concomitant drugs, and exposure in certain populations. Exposure-response modeling should be performed to relate blood concentration to safety outcomes of interest such as QT interval prolongation and relevant biomarkers, if available. The resulting model, along with appropriate trial model components, can be used prospectively to simulate later-phase trial designs and predict drug exposure, adverse events expectancy, and efficacy. In fact, the FDA had this very intention in mind when it issued the EOP2A meeting guidance.
Late-stage trials also benefit from pharmacometrics integration, not only to extract knowledge from observational data (as is often the case for late-stage PK sampling, for example), but also to confront the early-phase models to the more diverse patient population encountered in those trials. Only with a thorough knowledge of the dose-response behavior in specific patient populations can adequate dosing recommendations be made. Having an incomplete characterization of the shape and location of the exposure-response profile can lead to significant dosing errors. For example, in a 2002 paper by Cross et al., it was found that of 499 new molecular entities approved by the FDA from 1980 to 1999, one in five had a postmarketing dosage change, and for four out of five, the change resulted in a decrease in dose.
A Role For CROs
Until recently, pharmacometrics was seldom heard of as a service provided by CROs. The reasons for this are twofold: CROs do not usually have access to the entire body of data from sponsors, and typically only “big pharmas” had the means to assemble pharmacometrics teams within their organizations. However, the situation is changing rapidly. The pressure to reduce development costs within pharmaceutical companies, the increasing interest from the FDA in predictive (as opposed to descriptive) analysis methods, and the existence of commercially available analysis tools are making the offering of pharmacometrics services attractive to CROs.
CROs now offer a full range of pharmacometrics services, from nonclinical to postmarketing applications. Being relatively new, current pharmacometrics service providers are best used for circumscribed analyses that require handling data from one or a few selected trials. When deciding to outsource pharmacometrics services to a CRO, some aspects should be taken into consideration. First is the CRO’s ability to handle data. Pharmacometrics typically involves analysis data sets that will be built from one or several study databases. It represents a challenge because study databases are typically class-oriented (e.g. interventions, events, findings), while pharmacometrics analysis data sets are subject- and chronology-oriented. Second is the scope of work. The modeling tasks to be undertaken by the CRO are to be clearly established from the start and ideally should not be modified too much in the course of work. This is because in the modeling process, timelines tend to elongate significantly as the tasks at hand become more complex or diverse. Of course, additional tasks or changes in orientation may arise, but to the extent possible, they should be covered in the initial contract by allowing for changes in timelines and scope. Third, the type and level of quality control for the data and the model(s) should be clearly stated in the contract. Because pharmacometrics analysis is typically not a linear process, the quality of inputs and outputs should be reviewed regularly in the process. The level of quality control will depend on the task (descriptive vs. predictive modeling) and the intended use (internal vs. submission).
As contractual pharmacometrics services mature, it will also become possible to enter into functional service partnerships. While a few CROs can offer pharmacometrics partnerships, it is not yet a common practice. For it to be a viable alternative to in-house pharmacometrics groups, service providers need to have appropriate data management capabilities, a sufficient number and diversity of personnel (e.g., data management, biostatistics, clinical pharmacology, programming), and the ability to provide independent quality assurance. The requirement for clear-cut deliverables, although always important, may be relaxed to accommodate stakeholders’ priorities, since a drug development program is not, incidentally, a very linear process, either. Indeed, pharmacometrics often deals with, and evolves within, largely nonlinear processes.
The Future of Pharmacometrics
Even if some may argue that clinical trial simulation (CTS) is a well-implemented practice, it remains a challenging and resource-intensive endeavor at all levels. CTS requires computational capabilities, demonstrated expertise, and diversity in personnel backgrounds, making it an activity essentially reserved to a few well-organized pharmaceutical heavyweights. In addition, access to the critical data required to simulate some components of clinical trials (i.e. data on the natural progression of diseases) are often not easily accessible.
However, with the advent of concerted efforts aimed at making disease data and models publicly available, it now becomes possible to envision a broad use of CTS in drug development. One has only to think about the Alzheimer’s disease model from the Coalition Against Major Diseases (CAMD), the OpenDiseaseModels.org initiative of the Metrum Institute, or the FDA’s Specific Disease Model Library. With the continued development and sharing of such models amongst the industry and regulatory stakeholders, it will become easier for scientists to build and for regulatory bodies to review larger, more complex trial simulations that may eventually replace some of today’s live clinical trials.