Abstract and Personalized Medicine: A Framework for the Development of Pharmacogenomic Predictors of Drug Treatment Responses
Two overarching goals tied to a futuristic vision of personalized medicine are to maximize individually derived benefits from available therapeutic drugs or biological agents and identify a reliable set of predictors of toxic outcomes to optimally manage a patient's treatment-related risk. This article will primarily address the second of these goals with a specific focus on challenges faced in the development of pharmacogenomic markers that would be used by practitioners to predict individual susceptibility to serious adverse drug reactions. The potential complexity of polygenic modifiers of risk for certain adverse drug reactions and demographic diversity in the prevalence of genomic variants pose significant challenges in the development of useful predictive biomarkers. Implementing rigorous scientific and logistical strategies to address these will be crucial in order to achieve meaningful success.
Personalized Medicine: A Framework for the Development of Pharmacogenomic Predictors of Drug Treatment Responses
The identification of accurate predictors of individual responses to treatment with a drug or biologic agent will be a central feature in the growth of personalized medicine during the 21st century. Personalized medicine has its roots in an understanding of disease that dates back to the writings of Hippocrates and Galen. This was predicated on a concept that specific mixtures of the four essential humors underlie an individual's personality, as well as inherent tendencies for the development of particular diseases. Within this framework, treatment strategies were directed to reverse idiosyncratic imbalances of these humors. Upon discovery and elucidation of the molecular basis of hereditary diseases beginning in the early 20th century, a modern version of this ancient tradition focusing on inherent characteristics of the individual that predict susceptibility to many different diseases has emerged. In a modern context, the central tenet of personalized medicine is to elucidate both the basis of individual risk for many common diseases, as well as identify the genetic and environmental bases for differences in benefit and/or risk linked to drug and biologic treatments. This approach has gained even further prominence with the recent advent of human genomics and epigenetics. With remarkable improvements in rapid genomic DNA sequencing methodologies, we are now in the midst of a large ongoing global effort to catalogue genetic variations or polymorphisms that are dispersed throughout the human genome and which form a fingerprint, reflecting the genetic determinants for an individual's risk of disease and responses to drug and biologic therapies.
Can pharmacogenomics in a framework of personalized medicine fundamentally transform how drugs and biologics will be developed and used by patients, or is this approach destined to provide, at best, added value for only a small proportion of therapeutic agents that will be used over the next century? The verdict on this question is still out, as there are some fundamental challenges in the design and execution of pre- and post-marketing clinical studies that must be faced, including the specter that in some instances, as discussed below, complex polygenic traits in concert with nongenetic factors may influence variability in individual responses to a drug. Nonetheless, if a comprehensive strategy to address these challenges will be adopted both by academia and the pharmaceutical industry, an optimistic perspective on the impact that pharmacogenomics will have in the future is highly justified. There is already a substantial number of pharmacogenomic markers that inform appropriate patient selection for drug-specific treatment or dosaging. This list is expected to expand over the next few years.
Plans for the development and implementation of pharmacogenomics as a vital tool to optimally treat individual patients are being adopted by governmental healthcare and research authorities on both sides of the Atlantic. For example, the US Department of Health & Human Services (HHS) May 2008 Report of the Secretary's Advisory Committee on Genetics, Health and Society (SACGHS) has recommended implementation of an extensive plan for HHS to promote basic, translational and clinical research in pharmacogenomics, facilitate co-development of pharmacogenomic biomarker tests with new drugs, collect pharmacogenomic data to predict individual differences, and promote pharmacogenomic technologies in the arena of clinical practice. In alignment with this plan, 4 years ago the US FDA launched the Critical Path initiative.[3,4] This program is committed to enhancing the development and use of medical products to promote better benefit/risk outcomes. In conjunction with the adoption of more efficient trial designs, Critical Path supports the discovery and qualification of biomarkers that will reliably predict individual patient responses to drug treatment. Use of such biomarkers in guiding treatment decisions could enable better matching of specific drugs with individual patients, more appropriate patient dosaging and significant improvements in management of drug-related risk.
With increasing costs of studies attached to drug development programs and simultaneous pressures by healthcare systems to contain increasing costs of new drugs and biologics, reliable algorithms that take into account factors that influence variability in individual responses to these agents would play an instrumental role in patient management. In this regard, there is a growing need to predict not only which patients are likely to be 'mainstream' responders to treatment with a specific drug, but those who would be nonresponders, as well as those who would require modification of dosaging or avoidance of exposure, in order to mitigate risk for serious drug-related adverse events. In the future, measurements that would identify individuals likely to develop 'outlier' responses to drug treatment could become an invaluable tool in the management of many diseases. It needs to be emphasized that the value of such algorithms in patient management would be to complement rather than replace information gathered from randomized, double-blind, controlled clinical trials. Protocols based on randomization remain the gold standard by which bias of results in a treatment population can be systematically eliminated. Results of randomized clinical trials remain the essential backbone used by regulatory agencies to determine efficacy and safety of drugs when approvability is being evaluated for new clinical indications. However, clinical studies that employ randomization are often limited in that they are not designed to investigate factors that predict efficacy and safety responses to test drugs in small subsets of 'outlier' patients. Even when mean attributable response effects of a drug within a test population subjected to the rigor of statistical evaluation are robust, it is inevitable that there will be a range of levels of individual response, including nonresponse, to test drugs. Likewise, risk for drug-related toxicity/injury may vary enormously among study subjects treated with the test agent. It follows that to optimize evaluation and use of drugs and biologics it would be ideal to overlay an analysis of population-level responses with elucidation of predictive factors that account for individual variability. If achievable, such a synthesis might permit an increased success rate in drug development programs with simultaneous improvement in overall treatment benefits and drug-related safety outcomes associated with approved marketed drugs.
There are many drugs for which benefit can be shown to outweigh risk in some, but not all, patients with the disease(s) for which the drug is indicated. The degree to which benefit outweighs risk in a set of clinical studies would then be directly determined by the proportions of study enrollees from each of the different patient subsets. Since clinical study enrollees could include a mix of responders and nonresponders, the proportions of these would determine the treatment effect size result. This point is exemplified both by the correlation of increased Her/neu-2 expression and the efficacy of trastuzumab (Herceptin®), and the association of biomarkers of CYP2D6 activity and the efficacy of tamoxifen in patients with breast carcinoma. Treatment with trastuzumab has a selective advantage in patients with amplification of the Her/neu-2 gene, compared with patients with nonamplified forms of the gene. This has formed the basis of the recommendations in Herceptin's FDA-approved labeling to test for gene amplification, originally with an indirect measure of immunohistochemical (IHC) staining with fluoresceinated anti-Her/neu-2 antibodies and, more recently, by direct analysis of gene copy number with fluorescence in situ hybridization (FISH). Tamoxifen is a prodrug that must undergo biotransformation by a member of the cytochrome P450 multigene family, CYP2D6, to the potent anti-estrogen endoxifen. It is not surprising that in a retrospective study of medical records of postmenopausal women enrolled in a trial designed to test tamoxifen as an adjuvant for the treatment of early breast cancer, poor metabolizer (PM) status of CYP2D6 (presence of the variant allele CYP2D6*4) was a predictor of earlier cancer relapse, compared with the absence of the PM metabolizer variant allele of this isoenzyme.
The views expressed are those of the author and do not necessarily represent the position of, nor imply endorsement from, the US FDA or the US Government.
Personalized Medicine. 2009;6(1):67-78. © 2009 Future Medicine Ltd.
Cite this: Pharmacogenomic Biomarkers of Susceptibility to Adverse Drug Reactions: Just Around the Corner or Pie in the Sky? - Medscape - Jan 01, 2009.