Since its introduction in clinical pharmacology and pharmacostatistical modeling (6), the idea of representing disease progression as a fundamental component of the quantitative representation of the PK and pharmacodynamics (PD) of a drug has made considerable strides. This paradigm shift has also made academic and industrial research and development entities closer than ever. The main reason is as follows. In this era of evidence-based medicine (7), where scientific findings are routinely used to make policy decisions, it is only natural that drug development makes aggressive use of the same body of scientific knowledge utilized by regulators and policymakers worldwide to make decisions that potentially affect the health and well-being of millions. It has now become quite clear that a computer model of disease progress, tightly integrated with a quantitative representation of the drug distribution and PD, has a previously untapped potential to streamline drug development at many levels. The cost of drug development has risen severalfold over the past decade, and the process is staggeringly expensive and far from efficient: only five out of 5000 candidate compounds reach the human experimentation stage, and only one out of these five will be approved by the FDA (8). There is a role for quantitative, testable, and queryable models of the drug-disease system that incorporate current knowledge, together with population variation and outcome mapping. The interest in these topics is starting to cross disciplines, from traditional mathematical biology to drug development, for example, as a recent mechanistic study about the effect of imatinib (Gleevec™) (9) testifies.
There seems to be a strong preference these days for biomarkers that are easy to measure, preferably univariate, and that have at least the potential to diagnose and predict therapy outcome for diseases that are at the worst multifactorial (10). In practice, this is going to be a lot trickier than it seems. We will neglect for now the role of environmental factors, but we will come back to them later. Straightforward diagnosis and obvious needs, even when coupled with a knowledge of the disease mechanism, such as for some monogenic diseases, do not necessarily translate into successful therapies. On the other hand, a lot of promise may lie in revisiting existing treatments or compounds, with the increased awareness that comes from a better understanding of the drug-disease system and its working mechanisms. An example is acute lymphoblastic leukemia, where significant increases in cure rates have been achieved largely through "the optimization of the use of existing drugs, rather than by the discovery of new agents" (11). Needs are multiple: while it is important to clearly discriminate the genetic basis of a disease, it is also important not to overlook disease aspects, which are more related to gross phenotypes, such as demographics, dosing extent, delivery, and timing of administration. Optimization of therapeutic conditions is also required, together with the development of novel genetic or molecular paradigms. The challenge with these systemic events is that they are best addressed from a holistic, or system-wide, point of view. The questions need to be posed in a specific language, and the grammar of the language is made of seemingly abstruse concepts, such as differential equations and probability densities, while its vocabulary consists of parameter values and regression techniques. Its end results are readily understood, since they are phrased in terms of dosing recommendations and effectiveness predictions.
Was this article helpful?