Pharmacogenomic Biomarkers of Susceptibility to Adverse Drug Reactions: Just Around the Corner or Pie in the Sky?

Mark I. Avigan


Personalized Medicine. 2009;6(1):67-78. 

In This Article

Conclusion & Future Perspective

A Blueprint for the Future

Such an array of potentially difficult challenges should engender an 'open-eyed' response. The implementation of a number of scientific strategies and logistical measures to address them will be crucial in order to move meaningfully towards success. What are some of these?

Development of Consistent Study Designs in Pivotal Clinical Trials with Uniform Monitoring & Ascertainment Practices for Both Clinically Mild & Severe Adverse Reactions to Drugs and Biologics

In 2005, the US FDA released a Guidance for Industry on good premarketing risk assessment practices during the development of prescription drugs and biological products.[105] Based on principles described in this document, clinical study protocols should specify sufficiently frequent testing to enable reliable detection of systemic or organ-specific toxicities. Consistent monitoring practices of all study subjects in conjunction with specified protocols for diagnostic workups of individuals with serious ADRs will promote adequate ascertainment and clinical characterization of drug safety signals when they occur in clinical trials. In addition, it would enable the accrual of reliable clinical information surrounding all study subjects to justify pharmacogenomic measurements in those individuals with ADRs versus those who tolerate drug exposure without a drug reaction in a drug development program and across drug development programs.

Development of Uniform Clinical & Diagnostic Testing Criteria that Define Phenotypes Within Each of the Major Categories of ADRs

It is important that these be utilized consistently by academic and pharmaceutical industry investigators during performance of clinical trials. Adherence to agreed-upon criteria would greatly facilitate cumulative analyses of drug-related safety data across multiple studies in a drug-development program and across different programs. ADR phenotyping using such a set of specifications implies consistent appraisal of the physiological site and severity of toxicity, and assessment of causal linkage to drug exposure, using clinical and diagnostic test criteria. Although there can be different mechanisms of toxicity associated with different drugs, even with the same target organ of injury, unifying case identification criteria will be critical for performing effective cross-study analyses to identify pharmacogenomic markers of ADR susceptibility.

It needs to be emphasized that phenotype criteria should be developed for both severe and mild forms of each significant category of drug-induced toxicity. Randomized clinical trials provide an optimal framework for consistent ascertainment, adequate diagnostic workups and retrievable documentation of ADRs. Generally, when a drug-related safety signal appears in a clinical trial, study subjects in the subset of individuals treated with the test drug who develop toxicity present with a range of clinical severity. Typically, across this severity scale there is an 'iceberg' effect, such that cases with mild, self-limited drug-induced injuries are more commonly observed than life-threatening severe forms.[28] Although individuals with the milder forms of drug-induced toxicity may have critically important adaptive mechanisms that prevent acceleration of injury, they may also have genetic markers that determine susceptibility to initiation of a toxic response which overlap with the rarer patients who develop the most severe forms of toxicity. Inclusion of phenotypes with milder clinical severity into the pool of patients with toxicity in clinical trials analyzed for a genomic marker association with ADR susceptibility will enhance the likelihood that sufficient powering would be achievable. A recent example illustrates this point. SLCO1B1 is a gene that encodes the organic anion transporting polypeptide OATP1B1. The polypeptide plays an important role in the uptake of statins into hepatocytes, target cells in which these agents block cholesterol biosynthesis by inhibition of 3-hydroxy-3-methyl-glutaryl-CoA reductase. In a previous study, healthy volunteers with an allelic nonsynonymous polymorphism of SLCO1B1 in exon 6 (Val174Ala) showed reduced serum clearance of the active simvastatin acid metabolite, suggesting that standard drug dosing might be associated with reduced cholesterol-lowering efficacy and also increased risk for ADRs.[29] Recently, in an analysis of DNA samples systematically obtained in a large randomized study, two SNP variants that are in nearly complete linkage disequilibrium and located within SLCO1B1 were found to have a strong association with simvastatin-induced myopathy.[28,30] One of these variants is the Val174Ala polymorphism. The recently completed Study of the Effectiveness of Additional Reductions in Cholesterol and Homocysteine (SEARCH) performed in the UK enrolled 12,064 post-MI patients who were randomized to receive simvastatin in doses of either 20 or 80 mg. Susceptibility to idiosyncratic statin-induced myotoxicity is somewhat related to drug dosing levels. Thus, treatment with higher doses of simvastatin has been associated with a greater, albeit still low, risk for this ADR. Not surprisingly, because of its rarity, cases of statin-induced rhabdomyolysis were not observed. On the other hand, based on criteria of onset of muscle symptoms and prespecified elevations of creatine kinase levels, 98 cases of either reversible myopathy or incipient myopathy associated with 80-mg doses of simvastatin occurred. DNA samples from 96 of these in conjunction with 96 control study subjects who were matched for background risk factors were analyzed by genome-wide scanning of over 316,000 SNPs dispersed throughout the genome. Although the background population prevalence of one of the SNP variants that was further analyzed in the study is only 15%, more than 60% of the myopathy cases were associated with this SNP. Moreover, the odds ratio point estimates for association of the DNA polymorphism with the ADR were 4.5 in heterozygotes and 16.9 in homozygotes. The association of this SNP with simvastatin-induced myopathy has been reported to be replicated by UK investigators in the analysis of DNA samples obtained from statin-induced myopathy cases in a separate placebo-controlled study of the protective benefit of 40-mg doses of simvastatin (Heart Protection Study).

In vivo, simvastatin is a substrate for CYP3A4. It is not surprising that in SEARCH the relative risk for myopathy was 6.4 in study subjects who were concomitantly treated with amiodarone, a CYP3A4 inhibitor, compared with subjects not treated with this anti-arrhythmic agent. The increase in risk was observed despite an adjustment of the study protocol that limited maximal simvastatin daily dosing to 20 mg in users of amiodarone. Similarly, simvastatin-treated subjects who were concomitant users of calcium channel antagonists showed a modestly increased risk for myopathy (both verapamil and diltiazem are CYP3A4 substrates). In addition to drug-drug interactions, independent demographic and clinical risk factors for myopathy that were identified in SEARCH included older age, female gender, reduced glomerular filtration and diabetes mellitus. The finding of an association of the SLCO1B1 SNP variants with simvastatin-induced myopathy does not imply that these variants alone have enough predictive power to be used as markers to determine patient susceptibility to drug-induced rhabdomyolysis in a clinical setting. However, if testing for these SNPs will be combined in the future with other genetic and/or nongenetic markers to gain sufficient specificity for the ADRs of interest, a clinical application could quickly follow.

Despite an imperative to identify complementary markers in order to develop a clinically useful predictive algorithm, the SEARCH result demonstrates that, with appropriate protocols in place, large randomized clinical trials can provide important information on ADR susceptibility associations with genomic markers, when their prevalence in the population is not rare. Because of the 'iceberg effect' in which clinically milder forms of specific idiosyncratic ADRs are often more common compared with the most severe forms, during clinical trials ADR case definitions used as a guide in the collection of pharmacogenomic samples should be all-inclusive to achieve sufficient powering. Although this approach would not reveal concomitant genetic determinants that underlie transitions from mild to life-threatening forms of drug-induced toxicity it could have a number of important benefits. First, because of randomization and prospective monitoring, ascertainment of virtually all ADRs in the treatment groups is achievable and, in contrast to case-control study designs, hidden biases between comparator groups are minimized. Second, it would provide a basis for study and elucidation of basic genetic mechanisms underlying inception of idiosyncratic ADRs, serving as a prelude for a more complete understanding of the full range of pathological mechanisms that come into play. Third, even if the specificities of pharmacogenomic markers for susceptibility to life-threatening ADRs identified in clinical trials are unlikely to be high, in some instances they may still serve as important clinical decision making tools, if their sensitivities in defined demographic populations are robust. For this purpose, linking biomarkers for milder forms of ADRs uncovered in clinical trials with the most severe 'tip of the iceberg' forms will be a critical step. This could be accomplished by the performance of complementary case-control studies of genomic samples obtained from patients with ADRs collected in postmarketing outcome registries.

Development of Best Practices for Prospective Systematic Collection & Storage of DNA for Genotyping of All Study Subjects Enrolled in Clinical Trials During Drug Development

To enable analysis of association of genomic variants with risk, the collection should include not only samples from the patients who develop ADRs, but also the controls - study subjects who do not develop a toxic response when exposed to the test drug as well as those treated with a comparator agent or placebo. Often, drug-related serious idiosyncratic adverse events are rare. Owing to powering constraints for the identification of genomic ADR susceptibility biomarkers that I have described, optimizing biospecimen collection is critical. Given that idiosyncratic ADRs cannot be predicted prior to treatment with a test drug, systematic sample collection should be complemented by characterization of patient baseline medical histories at the time of study enrollment and adequate clinical and laboratory monitoring and follow-up practices during the course of the clinical trials.

Best practices should also be developed for uniform phenotyping and documenting nonstudy postmarketing patients with ADRs entered into a registry, when biospecimens are obtained for future case-control studies.

Development of 'User-friendly' Clinical Study Pharmacogenomic Databases

These would be web-based and contain an index of all clinical studies and investigators or sponsors that have systematically stored DNA and clinical outcome information. They would also serve as repositories of accumulating pharmacogenomic datasets that have been made available in the public domain from successively performed clinical trials and case-control studies. Critically important would be the uniform incorporation of information (e.g., ADR monitoring practices, ADR phenotyping and case definition criteria, and study subject demographic characteristics) from each clinical trial into the databases. This would permit analyses across trials. Datasets should be easily retrievable and amenable to integration for cumulative analyses and/or replication studies for genomic marker associations with ADRs. In the USA, development of shared resources for pharmacogenomic data-banking has begun. The NIH Pharmacogenetics Research Network formed in 2000 has funded a number of interactive research groups, some with a focus on an identified drug toxicity problem of interest.[106] Results of genomic studies are made available by the funded investigators and posted in the web-based Pharmacogenetics and Pharmacogenomics Knowledge Base (PharmGKB).[106] This web posting aims to stimulate collaborations and interactions among academic, pharmaceutical and governmental regulatory scientists. Separately, the National Library of Medicine has initiated a web-based database for genotypes and phenotypes (dbGaP).[107] Many of the current postings contain systematically collected genomic information obtained from clinical trials or case-control studies. So far, the phenotypes that have been entered into the database reflect different categories of human diseases. In the future, it may be possible to also include datasets with ADR phenotypes into this evolving repository.

Development of Focused Research Programs to Identify Human Genomic ADR-susceptibility Biomarkers for Specific Drugs or Drug Classes that Have Significant Public Health Impact

There are a number of drugs with well-known risk profiles for serious idiosyncratic ADRs that nonetheless play a central role in the treatment of a common disease(s). One example is isoniazid (INH), which continues to be a mainstay of both treatment as well as prophylaxis of TB worldwide. This agent causes clinically significant hepatotoxicity and asymptomatic elevations of serum liver transaminases in 1-2%, and in as many as 20% of treated individuals, respectively.[31,108] Although risk for INH-induced liver injury is modified by nongenetic factors such as age, gender, concomitant exposure to other anti-TB drugs and underlying liver disease, a number of studies have demonstrated that patients who are 'slow acetylators' due to reduced activity of N-acetyltransferase-2 (NAT2) are at increased risk.[32,33] Impaired acetylation may lead to an increase in hydrolysis of INH, which produces hydrazine, a toxic compound. Another proposed mechanism is that with slowing of conversion of acetylhydrazine to diacetylhyadrazine, due to reduced NAT2 activity, there is increased microsomal oxidation of the mono-acetylated form by CYP2E1 into a reactive metabolite which is toxic in hepatocytes.[33,34] This is supported by the finding that patients who express an allelic variant of CYP2E1 with reduced activity are relatively protected from the risk of INH-induced hepatotoxicity.[34] Much more pharmacogenomic research remains to be done to fully understand the interplay with genetically determined defects in metabolism and possible defects in adaptive liver responses that are protective. INH remains an important treatment in the worldwide struggle to eradicate TB, a disease with high prevalence in some parts of the world, and thus will continue to have significant usage. Because the phenotype and rate of INH-induced hepatotoxicity can be predicted in certain treatment populations, it could serve as a useful model to learn more about genetic susceptibility to idiosyncratic drug-induced liver injury (DILI). Accurate prediction of DILI susceptibility may eventually require the development of a multilayered algorithm based both on measurements of both polygenic biomarkers and nongenetic patient characteristics.

Development of a Comprehensive Strategy to Assess the Utility of Potential Pharmacogenomic Biomarkers & Clinical Predictors of ADR Risk for Healthcare Providers & Patients

An overall evaluation would incorporate information regarding the predictive power, sensitivity and specificity of testing for the ADR(s) of interest as well as analyze the projected utilization of pertinent healthcare system resources and test cost-effectiveness to achieve the systematic screening of patients. In addition, it would determine whether testing would significantly enhance the overall effectiveness of clinical care and consider the comparative risk/benefit calculus of alternative drugs that might be used. Such an evaluation would be iterative with the availability of new information.