Pharmacometabolomics in Early‐Phase Clinical Development

Pharmacometabolomics is an emerging field that uses the body's complement of metabolites to identify individuals likely to experience treatment or adverse effects. Nevertheless, review of clinicaltrials.gov reveals that <1% of trials used metabolomic principles and only 1.5% of 469 metabolomic studies were of new molecular entities.We review the history, current usage, and potential for future use of pharmacometabolomics in early−phase drug development, and conclude with recommendations for applications in clinical trials. Metabolomics represent the downstream end-products of cellular reactions, the "foot soldiers" of the genomic-transcriptomic-proteomic-metabolomic process, and the components that are most closely associated with the phenotype.1 Indeed, the organism's metabolic composition, the "metabotype," is a phenotype in its own right, a convergence of genetic, environmental, and pathophysiological effects.2 One of the fields that evolved from metabolomics is "pharmacometabolomics," initially termed "pharmaco-metabonomics" and defined as "the prediction of the outcome (for example, efficacy or toxicity) of a drug or xenobiotic intervention in an individual based on a mathematical model of preintervention metabolite signatures".3, 4 Pharmacometabolomics complements genomic, transcriptomic, proteomic, and epigenomic "systems biology" approaches to drug development and contributes to a comprehensive and holistic understanding of drug effects by taking into account both intrinsic and extrinsic contributions to interindividual variation in drug response.5, 6 Most importantly, there is the potential to understand and manage nonresponders and partial responders to conventional treatments, phenotypes that are likely the product of our incomplete understanding of pathophysiology and incorrect nosology, and grouping of diseases, For example, conditions such as coronary artery disease and schizophrenia are likely syndromes composed of many entities, with distinct etiologies and management requirements, as suggested by the wide variation in response to treatment and high percentage of nonresponders, partial responders, and remitters, and those suffering from adverse drug reactions.7 In that capacity, pharmacometabolomics hold the promise not only of delivering personalized drug treatment, but also improving the efficiency of drug development. This review describes the role and utilization of pharmacometabolomics as a tool in early-phase clinical development (i.e., the first human testing of new drugs), and its potential to facilitate translational effectiveness in drug development. We include assessment of utilization of pharmacometabolomics in clinical drug development using an analysis of clinicaltrials.gov records, discuss the related challenges and opportunities, and conclude with recommendations for future development of the field. The concept of metabolomics, as manifested in the use of bodily products to infer the state of health of the individual, dates back to antiquity. Examples include references in ancient Chinese and Ayurvedic medical literature to insects attracted to patients with sweet-tasting urine as markers of diabetes, and the use, albeit erroneous, of "black bile" and "phlegm" as markers of mood and alertness, respectively.5 However, because of the complexity of interactions between the multitude of metabolites and respective physiological and pathological states, and consequent dependence on sophisticated bioinformatics and powerful analytical and computational tools, the field has made significant progress only in last few decades (Table 1).1, 4, 5, 8, 10, 14-28 The term "metabonomics" (later converging with the parallel coined "metabolomics") was coined in 1999 as "the quantitative measurement of the dynamic multiparametric metabolic response of living systems to pathophysiological stimuli or genetic modification."8 It captures a dynamic process of metabolic changes over time in response to internal and external influences. One such influence is pharmacotherapy, and the term "pharmacometabolomics" was introduced in 2006 to describe the use of the metabolome to study drug effects, first applied to an animal model of liver damage associated with paracetamol metabolism.4 The analysis revealed that a certain metabolic profile was associated with increased liver damage after paracetamol treatment. Later studies provided further insight into the application of the metabolic profile as an early indicator of drug-related metabolism and toxicity in humans.9, 10 Using a range of complementary platforms for comprehensive chemical analyses, metabolomic approaches enable identification and quantification of physiologic-, pathologic-, and treatment-specific metabolites in cell extracts, tissue, and biological fluids (e.g., serum, plasma, urine, and cerebrospinal fluid). The product is a biochemical fingerprint of the organism at a specific timepoint containing information that may be relevant for diagnostic and therapeutic considerations and may be used to identify causal factors (i.e., biomarkers) most strongly affecting the organism's steady state. The most commonly used analytical platforms are nuclear magnetic resonance spectroscopy, noted for its capability for the comprehensive, simultaneous, "unbiased" quantification of a wide range of compounds, and the highly sensitive mass spectrometry (MS), including liquid chromatography MS (LC-MS), tandem MS (MS-MS), and gas chromatography MS methods, and, more recently, the more powerful ultraperformance liquid chromatography.8, 11-13 The complex multivariate nature of the data obtained with metabolomics requires sophisticated statistical, visualization, chemometric, and bioinformatics methods for analysis and interpretation.13 The vision of the "personalized medicine" initiative is the prospect of selecting treatments according to individual patient's unique characteristics, and, in particular, those characteristics that are relevant to treatment safety and efficacy.28-30 Metabolomics is potentially a useful prognostic indicator to complement other personalized biomarker domains (genomics, transcriptomics, and proteomics) because endogenous biochemical are ontologically closer and interact directly with the elements affecting the organism (e.g., pathological factors, environmental modifiers, treatment interventions, and the genome as well), thus, a more complete and authentic representation of disease effects and intervention outcomes. In contrast, genomic, transcriptomic, and proteomic information is controlling in nature and has yet to be "translated" and "actualized" downstream before exerting its effects and does not take into account the dynamic status of the entire organism or external effects.8, 25 Indeed, variation in response to pharmacotherapy is determined by both genes and the environment. Pharmacometabolomics identify characteristics of response to pharmacological interventions based on individuals' metabotype.1, 4, 27, 31, 32 The "metabotype" is the totality of person's characteristics associated with metabolic health and that which dictate disease heterogeneity and drug response.7 The "metabotype" reflects not only the constitution of the individual (e.g., the genetic makeup, gender, age, and ethnicity), and the impact of the disease (including any genetic components), but also the product of environmental exposure (e.g., diet, climate, environmental xenobiotics, gut microbiota, and circadian rhythms) as well as any effects of past and concomitant treatments (e.g., polypharmacy) that have impacted the organism during its lifetime and therefore provide a unique and comprehensive profiling of its constitution.2 Beyond the general conceptual argument it is evident that some metabolic products are more sensitive indicators of health states than others.8, 25, 33 In addition, drugs may affect gene expression and protein synthesis and may also have direct pharmacological interactions with metabolic products not directly affected by the genome or proteome that lead to therapeutic and toxicological effects.8 Finally, heterogeneous populations that appear phenotypically similar could display variability in molecular, metabolic, and other biological factors, which are important in determining drug response, allowing the use of metabolomics to decipher "behind the scenes" heterogeneity. In all these cases, choice of the optimal metabolomic biomarker out of a complex interconnected biological environment is of essence to the accomplishment of healthcare objectives, including successful drug development programs and pharmacotherapy. Several studies have demonstrated the use of pharmacometabolomics to guide the selection of the right drug for the right metabotype9, 34, 35 (Table 2).9, 10, 35-45 Early-phase development is defined as the first-in-human safety, and proof of concept efficacy clinical trials, typically conducted in healthy volunteers and patients, respectively. They are usually small (10–80 research participants) and short in duration (days to weeks) and aimed at obtaining initial information about drug effects in humans before the definitive large, long, late-phase approval clinical trials. Pharmacometabolomics can contribute to drug discovery and development at multiple points along the translational and clinical process (Figure 1, Table 346). The specific benefits are outlined in Table 2. The contribution is particularly relevant and valuable in early-phase human clinical development where little is known about drug toxicity and efficacy, and where reliably identifying "true positives" and "true negatives" can spare the expensive late-phase studies or loss of effective therapeutics, respectively, and reduce the costs and delays of developing innovative treatments. About half of all new chemical entities fail at phase III stage of clinical development, meaning they are the "false positives" of earlier trials; the "false negatives" may never be known as they do not get a "second chance" at retesting in large, adequately powered trials.47, 48 Because early-phase clinical development studies (i.e., phase 0, I, and II) are usually small, short, and underpowered, the value of traditional outcomes is limited. Any improvement in this expensive and lengthy outcome of early-phase inefficiency, such as the availability of reliable and powerful surrogate pharmacometabolomic biomarkers, can increase the predictive validity of early-phase trials and overall effectiveness of clinical development.49 Identifying new drug targets relevant to the drug's efficacy, safety, PKs Mechanistic insight into disease pathophysiology Insight into the impact of genotype and phenotype variability on pharmacotherapy outcomes Study design: Outcomes: Participant selection – by establishing more meaningful inclusion/exclusion criteria Dose selection – influenced by existing population and subpopulation information on dose-response and concentration-response relationships relevant to the drug or disease under study Validation of biomarkers identified in preclinical work and thus: Increasing the efficiency of later-phase trials Pharmacometabolomics used to inform the design of pharmacogenomic studies Sample collection: can be collected noninvasively, in most cases, with multiple samples easily collected over any required time course Ethics: adhering to pharmacometabolomics principles would enable more ethical study designs by limiting the testing of new medications to those most likely to benefit and least likely to experience adverse outcomes: Identifying at-risk population Identifying those most likely to experience beneficial response to the drug Limiting duration of exposure to ineffective drugs Early identification of toxic potential Increasing the efficiency of drug development with quicker delivery of new therapeutics Drug "rescue" and "repurposing": using newly validated metabolomic biomarkers to identify new value in existing drugs or previously unseen value in drugs that had their development terminated (Collins51) Vulnerable populations, disease subpopulations, and rare disease drug development: pharmacometabolomics could increase the efficiency of identifying subpopulations, and reduce the duration of exposure, leading to accelerated development for these conditions Increasing translational effectiveness: by lowering risk, duration, and, ultimately, cost of drug development Pharmacometabolomics, as other "omics" platforms, holds the promise of reliably predicting pharmacotherapy outcomes in a quicker and more efficient way than traditional approaches. This can be accomplished by identifying and utilizing metabolomic components as "surrogate" or "intermediate" biomarkers of longer-term clinical outcomes of interest to drug developers (e.g., toxicity, remission, mortality, and wellbeing). In addition, pharmacometabolomics can help account for the "non-genetic" components of human heterogeneity (e.g., lifestyle, diet, and environmental exposures). Such heterogeneity could account for important efficacy and toxicity variability in humans.50, 51 Pharmacometabolomic studies can then be done with limited exposure (dose and duration) to the novel drug, allowing fewer risks of adverse effects, minimal delays in delivery of standard treatment to research participants, quicker arrival at "go-no-go" developmental decisions, and reduced developmental timelines. Several unique features of pharmacometabolomic approaches need to be considered when incorporating into clinical development programs. The ethical aspects (e.g., confidentiality and inclusiveness) of sample collection and use should be taken into account in study design, storage, processing, and dissemination of results. Samples collected for the pharmacometabolomic evaluation are generally minimally invasive and could be easily collected over study time points or therapeutic time course. However, to maximize their "informatics" potential, sophisticated banking infrastructure has to be established and maintained so that the complex and large amount of information could be analyzed, processed, and compared with intra- and interindividual samples over long periods of time. ANALYSIS OF CLINICALTRIALS.GOV DATABASE Purpose The purpose of this study was to assess the type and scope of metabolomic applications in clinical trials as reflected in trials registered in clinicaltrials.gov. Methodology Clinicaltrials.gov database was accessed on 4 July 2015 using the key word "metabolomics." Each study entry was independently reviewed and categorized by the two authors (T.B. and S.D.) by phase, sponsor, therapeutic area, objectives, study start date, and outcome data (see Supplementary Material). Studies were categorized as "discovery" if the clinical trial was used to identify, study, or validate metabolomic biomarkers, and were identified as "clinical development" (phase 0 through phase IV) if the biomarkers were used as study outcome of pharmacotherapy interventions in clinical trials. We defined "early-phase development" as phase 0, I, or II studies, of developmental programs of new molecular entities, or new indications of known drugs. Studies evaluating only drug metabolite profile (e.g., mass balance studies) were not included. Any discrepancies between the authors' assessments were reconciled in a consensus discussion. Results Over the 18 years (1997–2015) available in the clinicaltrials.gov database, a total of 469 studies were identified in which metabolomic biomarkers were used as primary (51.8%) or secondary (48.2%) outcomes. One hundred sixty-six (35.4%) were drug development studies, 270 (57.6%) discovery studies, and 72 (15.4%) other (e.g., diet, exercise, and acupuncture) studies, with some overlap (see Figure 2). Study objectives were efficacy (57.4%), pathophysiology/pathogenesis (20.3%), diagnosis (19.2%), safety (16.0%), and prognosis (15.4%), with some overlap. Of the drug development studies, 92 (19.6% of the total) were "early-phase development" studies, however, only 7 (1.5%) of all metabolomic studies were used in development of new molecular entities. There has been a gradual increase over the past 14 years in trials utilizing metabolomic outcomes as one of the end points, especially after 2006. The trend appears to plateau after 2011 with another increase in 2014 (see Figure 3). Nevertheless, even the highest utilization frequency (92 studies) in 2014 constituted <0.5% of reported clinical trials (0.39% of 23,286 trials). The majority of the studies (438; 93.4%) were conducted by or in collaboration with academic institutes, 66 (14.1%) were conducted by industry, and 35 (7.5%) were industry/academia collaborations. Endocrinology, oncology, central nervous system, cardiovascular, and gastroenterology were the most represented therapeutic areas with endocrinology, at 215 studies, comprising almost half (45.8%) of all studies followed by oncology at 12.4% (see Supplementary Figure S1). A search using the near-synonym term "metabonomics" yielded 42 studies, of which 19 included drug intervention (Table 4,52 Supplementary Material). Of these, two were early-phase and two were late-phase clinical development studies. Sixteen (84.2%) were done by academia and four (21.1%) by industry (one study was done in collaboration). Results are similar to those from the "metabolomics" search. Limitations Studies before phase II (i.e., phase I and phase 0, or exploratory clinical trials) are not required to be registered in the public domain and may have not been included in the clinicaltrials.gov database. This may have exposed our analysis to reporting bias. Our search strategy was dependent on the use of the term "metabolomics." It is possible that studies used metabolomic biomarkers but have not identified them as such. Conclusions Over the 18‑year period of the clinicaltrial.gov database, a total of 469 studies included metabolomics applications in clinical trials, most (57.6%) in discovery phase (i.e., clinical trials used to discover/validate biomarkers), 19.6% in early phase drug development but only seven studies (1.5%) used metabolomics in development of new molecular entities. Almost half (45.8%) of the applications were in endocrinology, followed by oncology (12.4%). The large majority of metabolomic trials (93.4%) are conducted by academia rather than by drug developers and even with recent growth in utilization metabolomics are used in <0.5% of all reported clinical trials. The limited application may be due to the complex nature of metabolomics, the limited availability of qualified metabolomic biomarkers, and with sophisticated combinations of statistical, analytical, and scientific capabilities necessary for interpretation of yet to be The application of pharmacometabolomics multiple potential challenges in of study design, bioinformatics infrastructure and and ethical and (Table and may increase the complexity of clinical trials and associated early developmental A pharmacometabolomic in early-phase clinical development may need to with the that metabolomic markers are not yet of molecular entities may challenges because of limited with the The metabolome may have complex and relationships not only with the drug under development but also with disease states and environmental Such variability may be especially in the of the typically and early-phase clinical development trials. The of use and incomplete with the application of metabolomic principles in clinical are available in Table factors may initially be associated with high trial but costs are to as of into A recent of on biomarkers in drug development that before utilization as a clinical trial end a should be defined and validated using a developmental process discovery and evaluation of clinical and In the discovery the and computational are and are then and validated in a clinical population and to Nevertheless, in a recent to a the and Drug the to work with drug developers to maximize the use of novel biomarkers in drug development, even in in which the biomarkers are not yet validated or In in which the use of biomarkers for is not standard of potential patients for the trial may be confidentiality of information in bioinformatics systems and the of delivery of optimal healthcare because of using metabolomic markers that are not validated or not the standard of The lengthy of could also Sample collection and should be to variability the Drug development programs should include a for the identification and development of pharmacometabolomic biomarkers that may be useful in drug testing (Figure Such should be as early as before This time to and validate the biomarkers allowing application during the clinical development should at a to with metabolomics (i.e., the of a response to drug treatment with aimed at identifying and or a (i.e., the of a of based on validated and aimed at developing or utilizing an existing validated biomarkers they should be as outcomes in the design of early-phase clinical trials. they are primary or secondary outcomes may on existing experience with the biomarker and its predictive validity with to clinical outcomes. novel metabolomic biomarkers are identified as valuable and and should be as early as and potential benefits are then development and clinical of the should Early-phase programs would then a role in biomarker allowing utilization of the biomarkers in development. should be in and the of the process and clinical trials from the early such early (e.g., new drug and data more efficient clinical development can with the choice and (see of biomarkers and provide on the approval process and the role the biomarkers can in of study and of the of metabolomic data to the development process should be considered and In the clinical and process, should be made to biomarkers with the and and, predictive It is also in the and process to use a study that is from the analytical and clinical studies in which the diagnostic was initially the analytical (e.g., of a diagnostic should be based on a that is from the samples with which it is to be In and to maximize it is important to capabilities in of understanding related and complementary "omics" (e.g., markers of disease and drug effects, of the bioinformatics identifying platforms, and related to the use of the biomarkers, and The of metabolomic biomarkers with drug development applications but potentially with important applications in translational may collaboration and industry, and should include about future use of and potential risks because of or clinical The methods and for sample amount of sample processing, storage, out and related should be established and in and well in of the clinical trials. The time for biomarker especially novel and used in clinical should be taken into during the design of clinical trials. and central should be and methods, bioinformatics and data and related and should be established in in with research selection in clinical trials should only take with biomarkers validated and qualified in respective However, the of and amount of data required should be determined in with the on a and may well be influenced by the healthcare benefit of the drug under development (e.g., A for criteria used to the of data to guide selection has been by the Nevertheless, clinical development programs may be and could an important role in the of metabolomic In these cases, however, the drug development programs should use other for selection and of primary outcomes. The should include the relevant not only the nature of the drug under study but also the nature of the biomarker used to assess drug It should also include a of confidentiality including those of and if for biomarker development collection of for future studies should be pathophysiological and disease with and drug response, and powerful surrogate end In metabolomic with of response or of with traditional biomarkers could provide important into disease For example, the profile with response to treatment in which traditional an to previously disease and treatment and heterogeneous disease populations for in clinical trials. including disease response to treatments, and may be by their including with a may reduce study variability and increase its and to meaningful treatment effects. biomarkers could also help out research at for adverse and clinical trial populations by identifying variation in drug of healthy volunteers or patients with the under study may to the drug, even after the This clinical trial to of disease and treatment heterogeneity. For example, a metabolite or metabolomic profile may be in the assessment – help identify the with and including considerations of with drug – help the of research likely to experience and those likely to not experience adverse – help the of patients likely to to therapeutic intervention or likely to not pharmacogenomic trial – help provide (i.e., of in disease and treatment response, and identify biomarker combinations that are more powerful as surrogate end points than biomarker Pharmacometabolomics is an emerging "omics" biomarker field that has potential to drug development by early in the clinical development process, patients most likely to experience beneficial treatment effects and least likely to experience adverse outcomes. information the of genomic, proteomic, and environmental on the organism and can provide information on drug response not by the other The potential value is in early-phase clinical development, in which studies are small, short, and underpowered, and where pharmacometabolomics can help reduce variability of study populations and as a powerful surrogate of drug Nevertheless, analysis of clinicaltrials.gov in 2015 identified only limited application of pharmacometabolomics in drug development clinical trials. We for and of pharmacometabolomics principles in clinical development. include early and identification of potential biomarker to and sample The most is to start early in the discovery phase, with by and relevant pharmacometabolomic biomarkers so that they can be used at the of human the required in novel and pharmacometabolomics has the potential to clinical development overall developmental and lead to in overall translational effectiveness and delivery of healthcare The authors of Supplementary have been but not The is not for the or of any information by the Any than should be to the for the