Personalized Medicine for Chronic, Complex Diseases: Chronic Obstructive Pulmonary Disease as an Example

Josiah E Radder; Steven D Shapiro; Annerose Berndt

Personalized Medicine. 2014;11(7):669-679.

Abstract and Introduction

Abstract

Chronic, complex diseases represent the majority of healthcare utilization and spending in the USA today. Despite this, therapeutics that account for the heterogeneity of these diseases are lacking, begging for more personalized approaches. Improving our understanding of disease phenotypes through retrospective trials of electronic health record data will enable us to better categorize patients. Increased usage of next-generation sequencing will further our understanding of the genetic variants involved in chronic disease. Utilization of data warehousing will be necessary in order to securely handle, integrate and analyze the large sets of data produced with these methods. Finally, increased use of clinical decision support will enable the return of clinically actionable results that physicians can use to apply these personalized approaches.

1 2 3 4 5 6 7 8

Next Section

Box 1. Overcoming problems with large data integration: a success story in genetic data quality control.
While there are certainly many challenges to integrating complex data from different sources, many creative solutions to these challenges are already being employed. One excellent example of this is the rigorous quality control of genetic data, led by the Genomics Working Group of the Electronic Medical Records and Genomics (eMERGE) Network. Although whole-genome sequencing offers the possibility of genome-wide information, the reality is that for most clinically utilized tests and even for many research studies, the collection of genotypes at only specific loci, such as in genotyping arrays or targeted sequencing, are likely to remain more common. Even in the analysis of whole-genome sequencing, certain genotypes cannot be identified with confidence, and the specific genotypes that are missing may differ from sample to sample. Thus, when integrating genetic data from different technologies, platforms or analysis pipelines, it will be necessary to confirm that the genotype calls that are being compared are both of equal quality and, where comparable, equally complete. The eMERGE Network initially demonstrated the feasibility of integrating a large collection of genome-wide genotypes from multiple institutions for the purposes of conducting large-scale genome-wide association studies [92]. In order to do this, the group combined nearly 17,000 genotypes from multiple institutions that had been typed on one of two technologies at either the Broad Institute or the Center for Inherited Disease Research. In order to ensure that these variable data sources were of similar quality, each sample was checked for specific characteristics, including sex inconsistencies and population stratification, while each marker was required to pass a call threshold, which was checked for Hardy–Weinberg equilibrium and minor allele frequency. Ultimately, the group also introduced further quality control measures in order to accommodate the large-scale integration of many genotypes, including checking for strand orientation differences between sites and batch issues such as population stratification. The group concluded that, "With careful merging of datasets, we enable an increase in sample size, and likewise power to detect genetic associations and improve our understanding of the genetic architecture of complex traits" [79]. This has been validated by a number of genome-wide association studies linked to eMERGE's electronic record resources [82,93]. This demonstration of controlling the quality and completeness of a large composite dataset is one of many ways that researchers are overcoming some of the data challenges confronting the combination of genotypic data with phenotypic data that will be necessary in order to personalize medicine for chronic diseases.

Sidebar

Executive Summary

The challenge of personalized medicine for chronic, complex disease

Personalized medicine offers the hope of more precise diagnostic classification, leading to more successful targeted therapeutics.
There have been a number of success stories in personalized medicine from targeted therapies for tumor subgroups to the US FDA labeling of pharmacogenomics.
Although chronic diseases represent the majority of healthcare utilization, there are almost no personalized medicines for these diseases.
The challenge to personalizing medicine for chronic disease is their complexity and, until recently, the lack of technologies to adequately handle this complexity.

The importance of phenotyping chronic disease for personalized medicine

Chronic diseases such as chronic obstructive pulmonary disease are due to a complex interaction of many genetic and environmental factors, and this complex pathogenesis leads to heterogeneous clinical presentations, making it more difficult to determine individual pathogenesis.
Phenotyping aims to determine any measurable attributes of a patient that help to better classify that patient into a subgroup of disease.
Improved phenotyping for chronic disease can be achieved both through prospective, longitudinal studies and through retrospective studies.
Retrospective studies can complement classically designed prospective studies through greater lengths of time.

Finding missing heritability using more personalized approaches in chronic disease

While classical genetic studies and genome-wide association studies have identified candidate genes, significant missing heritability information remains, and the biological relevance of many results has not yet been elucidated.
Next-generation sequencing offers the opportunity to look for variations, such as rare variants and structural variations, which have not been previously explored in these diseases and are likely to contribute to the missing heritability information.
Next-generation sequencing has recently rapidly decreased in cost, opening up its use to more researchers in order to ask more questions about chronic diseases.
Better phenotyping also improves the ability to use modern genetic techniques, as it reduces the number of subjects that are necessary for a study and more clearly identifies the role of potential causal variants.

Combining genotype & phenotype for personalized medicine

Integrating data in a reliable and structured format will be =key to using large sets of clinical data for retrospective trials and for combining phenotypic and genetic data.
Some of the challenges to combining clinical data from multiple sites include data incompleteness and improperly entered data.
An enterprise data warehouse that follows certain importation rules can improve this situation.
Genetic data also present challenges, including their large size and the consideration of how to update the results based on new knowledge.
The Electronic Medical Records and Genomics (eMERGE) Network and the Electronic Health Records for Clinical Research (EHR4CR) project are exploring some of these issues now.

Bringing personalized medicine to the bedside

Although the promise of better classification of complex phenotypes will aid in developing personalized therapies, it also complicates implementation of guidelines for clinical practice.
Clinical decision-support systems can support physicians, assisting them in adhering to more complex guidelines within the electronic health record.
Advanced clinical decision-support systems will need to be context dependent and frequently updated.

Conclusion

Although challenges exist, we now have the technology to personalize medicine for chronic, complex diseases, offering the promise of improved patient care.