Pharmacogenomics in pregnancy

8.1

Pharmacogenomics

If it were not for the great variability among individuals, medicine might as well be a science and not an art. Sir William Osler (1892).

While much drug development and many clinical practice guidelines do not directly address variability in drug response, and in many cases assume that the effects of drugs on patients can generally be predicted, the evidence indicates otherwise. Significant numbers of patients do not respond to many medications, and adverse events that accompany drug therapy often compromise the quality of life of patients, limiting compliance with therapy, and can even be fatal in rare circumstances. The reasons for this variability in drug response often lie in easily accessible clinical factors including disease severity, age, weight, gender, ethnicity, or drug–drug interactions. Other factors may also be important, however, and in situations where readily available clinical predictors such as these are inadequate, alternative biomarkers of drug response can be used. In many situations the need for new biomarkers is urgent, perhaps most clearly in the case of diseases such as psychiatric disease or cancer, where considerable morbidity is incurred when therapy is ineffective or impossibly toxic for individual patients.

While improved efficacy is clearly a goal of the new era of “personalized medicine” heralded by the development of increasingly sophisticated new biomarkers of drug response, the occurrence of unanticipated adverse effects is also of great concern. It is clear that considerable damage is done to the public health by such adverse events. In one of the largest studies of in-hospital morbidity published to date, the incidence of serious adverse drug reactions (ADRs) was 6.7%, of fatal ADRs was 0.32%, and it was estimated that of 2 million patients, 216,000 experienced serious ADRs and over 100,000 had fatal ADRs in 1 year, making these reactions between the fourth and sixth leading cause of death [ ]. The cost was estimated at more than 100 billion dollars per year in 1994. Even with advances in alerts, electronic records, and other safety checks, it is estimated that ADRs cause an increase of over $5 billion to US health-care payers from injectable drugs alone [ ]. It follows that biomarkers that can predict , and also prevent adverse events, would also be of great potential value.

Biomarkers of drug response in clinical practice are far from new. Tests such as the international normalized ratio (INR) used to monitor warfarin response, the presence of estrogen or progesterone receptors on breast tumors used to guide antiestrogenic therapy, and the testing of patients with HIV or hepatitis C for viral loads are all a routine part of daily practice that health-care professionals have become comfortable with. We have learned that clinically useful biomarkers of drug response are of most value in situations where there is great variability in response, and a clear clinical decision, such as a change in drug, dose, or therapeutic approach, results from a test. It is equally clear that a test must have iterative value over existing easily available clinical predictors in order to be useful. For example, a test designed to predict the efficacy of an antihypertensive agent that had less predictive ability than routine measurement of blood pressure would be of little value.

The advent of genomics has brought a series of powerful new tools to this predictive science. While proteomics and metabolomics show great promise, it is with germline genomics, the study of the genetic sequence that we inherit from our parents, that we have the most experience. There are a number of reasons why the science of pharmacogenetics (or pharmacogenomics) appears valuable in this context. Not least among these are the simple facts that DNA is very stable and easy to amplify, and that there exists a map of the human genome and of the international hapmap ( http://www.genome.gov/10001688 ). In addition, the cost of DNA testing continues to drop dramatically.

While many definitions of the differences between the science of pharmacogenetics and that of pharmacogenomics have been put forward, a useful distinction appears to be simply that “pharmacogenetics” refers to the study of individual candidate genes, while “pharmacogenomics” refers to the study of whole pathways of genes, and indeed the entire genome.

8.2

Genetics and polymorphisms

Genetic variation in the sequence of about 3 billion nucleotide pairs that make up our DNA comes in many forms, but the most common differences between people are in the form of single-nucleotide polymorphisms (SNPs). These are single letter nucleotide changes and they are referred to as a “polymorphism” if they occur in 1% or more of the population. This is because variants that are that common tend to be stably present in a given population, whereas variants present at less than 1% tend to drift out. There are 12–15 million of such variants, and they have been meticulously cataloged by the Human Genome Project in the publicly available database called dbSNP ( http://www.ncbi.nlm.nih.gov/projects/SNP/ ). Since SNPs are the most common and easily accessible form of variability, they form the basis of the first genome-wide association studies (GWASs) testing that has been used to test for associations between common variants in the genome and nearly every form of human pathology ( http://www.genome.gov/gwastudies/ ).

Other important forms of variation include deletions and insertions of sequence, variable number tandem repeats of short sequences that are clustered together and oriented in the same direction [ ], and copy number variation: regions of the sequence that are copied with high fidelity within the genome itself. It has been estimated that such regions constitute up to 12% of the entire sequence in the genome [ ].

Since only about 1.5% of the human genome sequence is used for the ∼24,000 genes that code for proteins in humans, we presume that not all of it is relevant to therapeutic response, and that not all of this variability has functional or clinically meaningful consequence. That said, large numbers of variants that influence function via “nonsynonymous” changes in coding SNPs (cSNPs) have been found, and a growing number of functionally important variants in intronic and regulatory regions have also been identified [ ].

The use of GWASs to identify new genetic associations between SNPs and drug response has begun and already a significant number of important discoveries have resulted. These include the discovery of the SLC transporter with the muscle toxicity incurred by the use of the statin class of drugs [ ], and of a gene in the IL-17 pathway with the musculoskeletal toxicity associated with the use of aromatase inhibitors in patients with breast cancer [ ]. It is widely appreciated that a large number of new patterns of multiple genetic associations will result from this effort [ ], such that tests that involve large numbers of variants organized into a predictive pattern will become commonplace. The use of such predictive patterns is already commonplace in breast cancer, where arrays that test for 20–100 RNA species in a tumor at once are routinely used to predict the value of chemotherapy in individual patients [ ]. Additionally, there has been exciting work at implementing CPAC-recommended pharmacogenetic information related to treatments into health record problem lists to help guide clinicians [ ].

Within this large field of research, our understanding of genetic factors that affect drug disposition far exceeds our understanding of the factors affecting response. This is in part because pharmacokinetic changes are relatively easy to measure whereas the “phenotype” of overall drug response is more complex. In addition, cloning of most drug-metabolizing enzymes and drug transport proteins within the past 20 years combined with the genetic polymorphism information generated by the sequencing of the human genome and cataloged in dbSNP have allowed a comprehensive characterization of variability in drug metabolism and transport. As the practice of searching for, identifying, and then using determined genetic characteristics as predictors of drug effect becomes more common, it is clear that the entire community of health-care providers—physicians, pharmacists, and nurses—will have to play an increasing role as the value of carefully defined, valuable clinical phenotypes and their individual genetic and genomic associations increases.