Data Sharing, Clinical Trials, and Biomarkers in Precision Oncology: Challenges, Opportunities, and Programs at the Department of Veterans Affairs

Abstract

Cancer genomic research reveals that a similar cancer clinical phenotype (e.g., non‐small cell lung cancer) can arise from various mutations in tumor DNA. Thus, organ of origin is not a definitive classification. Further, targeted therapy for cancer patients (precision oncology) capitalizes on knowledge of individual patient mutational status to deliver treatment directed against the protein products of these mutations with the goal of reducing toxicity and enhancing efficacy relative to traditional nontargeted chemotherapy.

THE NEED FOR COLLABORATIVE DATA SHARING

Subclassification and study of cancer patients based on mutational status presents opportunities to learn the significance of genomic alterations (and their combinations) and to develop additional therapies (and combinations of therapies). However, the large number of mutations known to be important in cancer development and the presence of multiple mutations in any individual patient combines to create a great diversity in populations of patients with a specific tumor type. Shrager and Tenenbaum note that cancer is “in effect, a large number of rare diseases occupying a very high dimensional space with very few opportunities for action and observation in each subtype. To efficiently search a space of this nature, one needs to capture the learnings from as many patients and treatment experiments as possible in a continuously updated knowledge base.”1

The stratification of cancer patients by mutational status and resultant decrease in the proportion of patients available for study enrollment presents serious problems for observational and interventional research. The low prevalence (1–2%) of many driver mutations in solid tumors precludes recruitment of sufficient numbers of subjects from traditionally sized research consortia and has led to new models of clinical research based on interinstitution collaboration, community outreach, and broad data sharing, with the goal of learning from every treatment encounter. This article reviews the challenges and opportunities of such collaborations from the perspective of the Department of Veterans Affairs Healthcare System, the largest integrated healthcare system in the United States.

Creating generalizable knowledge as well as informing current individual patient care is therefore enabled by “learning” from all available previous treatment experiences of every relevant patient in the entire system, aggregating data across many medical centers. Familiar challenges to this approach include technical issues, such as data element provenance, data quality, and database variability across institutions, ensuring patient protections in data sharing related to informed consent, privacy/confidentiality, and Health Insurance Portability and Accountability Act (HIPAA) authorization, scalability and sustainability of aggregated databases, and cultural and financial barriers to data sharing in a research community.

The ability to commoditize healthcare data has created opportunities for new consortium models that enable data sharing and patient access (for clinical trials). Data sharing is an important element of the collaborations exemplified by ORIEN,2 MED‐C,3 TAPUR,4 and APOLLO5 (see Table 1). The National Cancer Institute (NCI) has created the Genomic Data Commons, a unified data repository6 that enables data sharing across these and other cancer genomic studies, in support of precision medicine. The repository houses clinical health record and genomic data (FASTQ file format) and complements the Cancer Imaging Archive,7 another NCI‐sponsored repository that contains radiographic and pathology images for cancer patients.

Table 1.

Sharing with the research community

Group	Leading institutions	Consortium activities (beyond data sharing)
ORIEN2	The Moffitt and The Ohio State University Comprehensive Cancer Centers	Biobank and investigational drugs procurement
MED‐C3	Molecular Evidence Development Consortium	Reimbursement for tumor sequencing
TAPUR4	The American Society of Clinical Oncology	Availability of targeted therapy off label
APOLLO5	National Cancer Institute, Department of Veterans Affairs and the Department of Defense	Tissue and data for biomarker development (proteogenomics)

The Department of Veterans Affairs (VA) has begun to move consented and HIPAA‐authorized patient data from the VA electronic medical records to these NCI data repositories for subsequent sharing with the research community (see Table 1). This approach replaces the need for cancer patients to sign multiple forms consenting to data sharing with a single broad consent that satisfies the requirements of the various project Institutional Review Boards (IRBs). It also provides data collection and curation that meets requirements of the individual project aims. The “Big Data Scientist Training and Education Program” is a joint effort by the VA and the NCI to provide these data to early career scientists to develop analytics and other tools in support of clinical and research objectives.8

CLINICAL TRIALS

Clinical trialists confront obstacles unique to precision oncology. As discussed above, traditional recruitment approaches from single or limited groups of institutions do not provide sufficient numbers of eligible study subjects to fulfill inclusion criteria with specific tumor–mutation combination requirements. Furthermore, third‐party payers seldom reimburse for mutational analysis required for patient screening prior to entry into research, thus shifting to the research enterprise the cost of screening large numbers of patients, of whom only a small fraction will be found eligible. The lack of data on the mutational status of patients is the major bottleneck for clinical trial execution and slows progress in precision oncology. In the era of precision oncology treatment, the new standard of care requires clinical reimbursement for expanded panel testing in cancer patients, with subsequent recruitment in clinical trials when appropriate for the individual patient. That standard then leads to sharing and reuse of patient data for clinical trials and observational research.

Major efforts underway to solve these and related problems are exemplified by the MED‐C and ORIEN initiatives. The MED‐C Program offers insurance coverage of testing for institutions and patients who agree to contribute clinical data to the N1 registry for subsequent analysis. The Oncology Research Information Exchange Network (ORIEN) is a research collaboration founded by the Moffitt Cancer Center in Tampa and the Ohio State University Comprehensive Cancer Center to match patients to targeted treatments and promote data‐sharing activities. The group engages industry partners on sponsored projects across the clinical trials continuum and serves as a broker between research and healthcare communities. These and other programs (particularly NCI‐sponsored clinical trial consortia) that foster collaboration between clinical care and research communities represent new models to advance precision oncology.

Membership of VA sites in NCI Consortia and programs such as MED‐C, ORIEN, and TAPUR makes data sharing and clinical trial participation opportunities available for Veterans at participating VA Medical Centers but leaves behind patients at facilities that lack oncology research programs and infrastructure. This structural problem is not unique to the VA, as opportunities for patients to participate in cancer clinical trials are similarly limited in communities not located near cancer centers.

If participation in clinical trials is considered a new standard of care, then this phenomenon exposes a new and important access disparity in a healthcare system. The Distributed Enrollment Program under development at the VA Cooperative Studies Program presents a model to “move clinical trials to patients” remote from VA cancer centers by seeking pre‐approval of cancer protocols by a central IRB, reduction of research‐specific training requirements imposed on participating clinicians, and centralized trial management (such as data collection and submission). These enhancements are engineered to deliver “just in time” research opportunities for patients while preserving patient safety and data integrity. Indeed, much of the administrative overhead built into the clinical research apparatus in the name of “quality assurance,” while highly appropriate for research designed primarily to benefit the broader community (such as registration studies for drugs where effective alternatives exist), may have reduced relevance in a precision oncology setting, where the patient's motivation to participate is first and foremost to obtain study drug, whether or not alternatives exist.

BIOMARKER DISCOVERY AND VALIDATION

By its nature, precision oncology is a biomarker‐driven field critically dependent on acquisition of clinical biosamples and electronic medical record data for discovery and validation with lack of access to both resources as a central limitation to the pace of discovery. Collaborations to cost‐share for data and tissue procurement between biotech, pharma, and healthcare systems, in a “precompetitive fashion,” are emerging.4 To reach full potential, such collaborations require more complete integration within healthcare systems, as exemplified in the APOLLO Program, a partnership between the NCI, DoD, and VA. In APOLLO, tumor tissue is made available for biomarker analysis (proteomics in this case) and the results transmitted back to healthcare providers if they are determined to offer incremental value to patient care, beyond mutational analysis. While observational data is useful in this regard, randomization accelerates learning, and results in more certain knowledge. For example, demonstration that patients randomized to genomic augmented with proteomic analysis had superior treatment outcomes to those randomized to genomic analysis alone supports a compelling argument to adopt this new biomarker of response to targeted therapy. Problematically, the introduction of randomization (to biomarkers) into clinical care using traditional clinical trial methods is cost‐prohibitive. The Department of Veterans Affairs continues to make progress in this area through the Point‐of‐Care research9 and Precision Oncology10 Programs whereby patients are randomized to minimal risk alternatives with relaxed regulatory requirements appropriate with the degree of risk (risk‐based monitoring). Data generated from these embedded studies are derived exclusively from the EHR (real‐world evidence) and require FDA acceptance if used for registration of a new companion diagnostic.

DISCLAIMER

The contents do not represent the views of the U.S. Department of Veterans Affairs or the United States Government