Analytical Challenges and Opportunities Associated with the Analysis and Interpretation of Clinical Studies in Precision Medicine
Recent advances in technology have led to a proliferation of data and in genomic testing alternatives for complex diseases as more treatment options are developed to target genomic and/or proteomic alterations in subsets of the overall population. Due to these recent advances in the capturing of genomic variables (Kalf et al., 2015; Marzuillo et al., 2014), there has been an increased clinical, analytical and regulatory need to identify genomic and proteomic biomarkers relevant to drug response. The challenge precision medicine studies face is that patients are sampled from a population that (due to practical constrains) are not usually fully representative of the target population that will ultimately receive a drug, biologic or vaccine.
Big data provides another tremendous resource for hypothesis generation as well as increasing the sampling population from which variables are being derived (Huang et al., 2016). Even so, disparities within and between these data and sampling bias that are inherent to the data pooled into big data further convolutes its potential utility. These burdens are small in comparison to the isolation problem that biostatisticians, bioinformaticists, geneticists and clinicians face within and across industry, governmental organizations, academia and clinical practice.
The Population Approach to Precision Medicine
The unique challenge to precision medicine studies is that they are sensitive to genetic as well as environmental influences. These influences could vary in their impact over time and the interaction between genetic by genetic, genetic by environmental and environmental by environmental factors could contribute to large sources of variation that are analytically complex to elucidate. Another factor of complexity lies in the added interaction between drug response and disease, which creates a complex phenotypic response measure. Because of the complexity of interactions and how these interactions could be influenced at different time points in a patient’s lifetime exposure, it is important to consider how, when and where endpoints and covariates are captured for use in precision medicine studies. This is of particular importance in RNA and expression studies where the time of collection is essentially the capturing of a moment in time for each patient. The timing of drugs, pretreatment and PK/PD characteristics of single and combination compounds are of particular importance in the consideration of analysis and interpretation of results in precision medicine studies.
The Unique Characteristics of Every Patient in Precision Medicine
Based on the unique combination of genetic and environmental factors each patient encounters during their lifetime, and even prior to birth, it is reasonable to conclude that the unique genotypic, environmental and epidemiologic characteristics within and shared between patients offer distinctive advantages in the field of precision medicine. This is beneficial if and only if care is taken in collecting, analyzing and interpreting these variables, considering the different sources of variability that could be introduced including time of collection and/or timing of treatment. The challenge in the analyses of precision medicine data is that we can generate more data than we have the biologic understanding for.
The Hardy-Weinberg Example
Population geneticists, biostatisticians and those who brave the world of precision analytics make use of the Hardy-Weinberg principle in the analysis of genetic data. The Hardy-Weinberg principle states that allele and genotypic frequencies in a population will remain constant from generation to generation in the absence of evolutionary influences namely, mutation, migration, selection and mate choice. Because one or more of these principles are typically present in studies being evaluated for drug response, the Hardy-Weinberg principle sets the stage for the ideal population against which these influences could be investigated. The challenge however is that deviations from Hardy-Weinberg are frequently interrelated and in modern society, with increased travel, migration and a more accepting culture for interracial relationships, the deviations are intertwined and complex. The Hardy-Weinberg principle is frequently used as a go-to in elucidating genotypic quality control, but maybe a better understanding and use of this principle in the design and analyses of precision medicine data could lead to an explanation of why we often fail in our predictive capabilities of precision medicine studies.
Choice of Endpoint – Phenotype
According to www.vocabulary.com, the description of phenotype is contrasted against genotype: “whereas the ‘genotype’ is the genetic makeup of an organism, the phenotype is how genetic and environmental influences come together to create an organism’s physical appearance and behavior”. Robinson (2012) describes phenotypes in the context of precision medicine as some deviation from normal morphology, physiology, or behavior. The concern is that phenotypic descriptions in clinical notes and medical publications are often imprecise, and more so phenotypic measurement of drug response does not always take the end result in mind that drives the success of a drug post approval. This is the moment when the patient and their clinician or team of care providers decides what drug is best for that patient, at that stage of their disease, given the underlying factors (other phenotypes) we know about their disease. Unfortunately the first hurdle to overcome is drug approval and clinical trials are mostly designed with success of approval in mind. We “hinge our bets” on choice of primary, co-primary, secondary and exploratory outcomes or phenotypes, sometimes knowing that underlying processes are little understood, while facing a tremendous burden to achieve success.
Deep phenotyping can be defined as the precise and comprehensive analysis of phenotypic abnormalities in which the individual components of the phenotype are observed and described (Robinson, 2012). The emerging field of precision medicine aims to provide the best available care for each patient based on stratification into disease subclasses with a common biological basis of disease. Unfortunately the common biological base of disease and more challenging drug response to disease at any given time is frequently not well understood. This requires assumptions, which in turn can lead to faulty conclusions of success and ultimately poor replication of findings. The comprehensive discovery of such subclasses, as well as the translation of this knowledge into clinical care, will depend critically upon computational resources to capture, store, and exchange phenotypic data, and upon sophisticated algorithms to integrate it with genomic variation, omics profiles, and other clinical information.
Translation of Complex Biomarkers to Physicians and Patients
Traditionally bioinformaticists utilized big data and resources to generate hypothesis, while biostatisticians play a role towards hypothesis confirmation. Unfortunately these two groups may compete instead of complementing their skill sets. A complimentary approach towards using biostatisticians and bioinformaticists to synchronize efforts as a team could largely benefit prediction of response in precision medicine. Bioinformaticists, biostatisticians, clinical geneticists, epidemiologists, clinicians and pharmaco-economists (to mention a few) could accomplish so much more if they would not work in silos, but rather create an environment where knowledge could be more freely shared.
Regulatory Landscape in Precision Medicine
Advances in the ability to generate new biomarkers are taking place within a complex regulatory environment, with oversight split between the Food and Drug Administration (FDA) and the Centers for Medicare & Medicaid Services (CMS) (Horn & Terry, 2012). According to the FDA, the first regulatory challenge that must be overcome for precision medicine to advance is scientific (Nicol et al., 2016). Therefore proof of analytic and clinical validity and clinical utility is required. However, the FDA also recognizes that regulatory policy and management challenges are posed by these technological advances (Food and Drug Administration, 2015). There will need to be significant changes in the way the FDA and equivalent agencies in other countries oversee application of new, complex biomarkers, along with cooperation across multiple centers and departments to better coordinate regulatory requirements and avoid (Greenbaum, 2012). Such significant changes will likely require corresponding legislative reform (Riley, 2015).
Precision medicine provides unique opportunities to identify subgroups of patients that are more likely to respond to treatments geared towards the unique phenotypic make-up of the unique person. This provides challenges in how we sample patients from the population at large into clinical trials, identify appropriate endpoints, choose optimal genomic and non-genomic biomarkers, quantify the environmental impact on response and use the gained knowledge to predict response. It is therefore essential for the different experts available in the fields to find ways to interact better when designing, collecting, analyzing and interpreting data in precision medicine.
Huang, Bevan E., Widya Mulyasasmita, and Gunaretnam Rajagopal. “The path from big data to precision medicine.” Expert Review of Precision Medicine and Drug Development 1.2 (2016): 129-143.
Marzuillo, Carolina, et al. “Predictive genetic testing for complex diseases: a public health perspective.” QJM 107.2 (2014): 93-97.
Kalf, Rachel RJ, et al. “Variations in predicted risks in personal genome testing for common complex diseases.” Genetics in Medicine 16.1 (2013): 85-91.
Robinson, Peter N. “Deep phenotyping for precision medicine.” Human mutation 33.5 (2012): 777-780.
Horn, Elizabeth J., and Sharon F. Terry. “Regulating genetic tests: issues that guide policy decisions.” Genetic testing and molecular biomarkers 16.1 (2012): 1-2.
Nicol, Dianne, et al. “Precision medicine: drowning in a regulatory soup?.” Journal of Law and the Biosciences (2016): lsw018.
Food and Drug Administration. “FDA notification and medical device reporting for laboratory developed tests (LDTs): draft guidance. October 3, 2014.” (2015).
Greenbaum, Dov. “Regulation and the fate of personalized medicine.” The virtual mentor: VM 14.8 (2012): 645.
Margaret F. Riley, An Unfulfilled Promise: Changes Needed to the Drug Approval Process to Make Personalized Medicine a Reality, 70 FOOD & DRUG L. J. 289 (2015)