OR WAIT null SECS
Using real-world databases for evidence development has strengths and weaknesses; with electronic health records (EHRs), new approaches are evolving
Drug developers have long considered randomized controlled trials (RCTs) to be the gold standard in clinical research, but with the rise of electronic health records (EHRs), real-world studies that leverage existing medical and healthcare administrative data have become practical tools for getting actionable answers to clinical questions, including safety, outcomes, and many other health care endpoints. Availability of high quality real-world data is currently irregular but will be increasingly available globally over the next decade.
Database analysis has its own unique set of challenges, but when used appropriately, it is an excellent tool that supplements RCT and prospective observational study data to provide a more comprehensive picture of efficacy and effectiveness. Innovative hybrid study designs that leverage routine data sources are quickly capturing the interest of drug developers. Healthcare providers, and national and local payers are using real-world databases to provide quality and outcomes measurement to drive the most value for their patient populations.
The current landscape for real-world database studies
Access to existing high quality clinical data is increasingly in demand. However, there are a number of issues that confound the collection of this data. We have payers increasingly asking for more evidence of cost effectiveness that applies to the real world. Availability of this data is, however, contingent upon a product being reimbursed and prescribed.
The quality of real-world databases from country to country is also mixed. Those that exist are often incomplete across different health care sectors; for example, many are focused on general practitioners or the hospital sector, but rarely does one database cover all the different settings that play a role in medical treatment.
Still, even given these limitations, interest in real-world databases continues to grow, driven by a number of factors. First, there is a recognized need for more timely and cost-effective research approaches, and it is now widely understood that real-world data is an essential component of evidence-based medicine. Real-world databases contain answers to questions important to payers and providers such as quality of care and value of provision.
Further, technological advances and common data standards are enabling integration of disparate data sets. Scientific methods are growing in their sophistication to allow valid conclusions from real-world database studies. Adoption of EHR to support healthcare is accelerating. In fact, Global Data, a leading provider of business intelligence, projects worldwide EHR growth to surge a whopping 12 percent annually through 2016. Much of this growth will be driven by government financial incentives in the United States, Australia, China, and Canada, as well as by recent US legislation; namely, the Affordable Care Act of 2010, the Health Information Technology for Economic and Clinical Health (HITECH) Act of 2009, and the Medicare and Medicaid EHR Incentive Programs. New standards evolving out of these acts will allow for common data elements across EHR systems.
Why not only use RCTs?
Increased ease and availability of real-world data does not alone justify their use. The true reason for the increasing popularity of these databases is that RCTs are simply not well suited to answer all research questions. Although RCTs provide the strongest evidence for efficacy of a treatment, real-world database studies provide an ideal way to study effectiveness—that is, how a treatment works in actual practice.
An RCT protocol is set up with strict inclusion and exclusion criteria, making the study population vastly different from the population at large. These types of trials also do not give insights into why clinicians may use drugs off-label or in risky situations, and the RCT’s short treatment and follow-up duration limits our long-term understanding of the study subject.
Real-world databases, on the other hand, are far less expensive and time-consuming, and their limited inclusion/exclusion criteria mean patients are more representative of usual practice. They allow for access to broad information on how a drug works given a specific condition, so we can more easily assess comparative effectiveness. Also, the treatment population in a real-world database is more realistic and not limited to patients who may be predisposed to respond well to treatment. Real-world databases can add value in the pre-approval stage, as they allow for an exploration of the population that could benefit from a drug, facilitating better clinical trial design. When study results must be generalizable, insights on a large number of patients is quickly required, there is a need to identify trends over time, or recruitment of the target patient population is challenging, real-world databases can be an advantageous tool.
However, as previously noted, there are challenges associated with these databases that must be considered. A database may not be available for the specific research objectives of interest, or the existing data may be incomplete. A pilot study is recommended to ensure the question can be addressed with the data available. Another challenge is free-form text entry, which often makes key information difficult to extract from a database. EHR databases may show medicines that have been prescribed, but were not necessarily filled, or a database may include the start date of a medication, but not the end date.
Even if these challenges can be overcome, there are specific issues inherent to databases. In an observational database, assignment of treatment is not random. This means there may be varying severity of illness between patients who receive different treatments, which limits what we can infer in terms of treatment comparisons and outcomes. While randomization controls known and unknown differences between groups, databases do not. Statistical methods, such as propensity score matching and instrumental variable analysis can help provide balance and compare like patients to like patients, but there is still work to be done in this area.
Evolving hybrid study designs can strengthen EHR data
Hybrid study designs that examine both EHR data and survey data are an innovative way to maximize the potential of real-world databases. Such approaches incorporate not only passive data collection via EHR, but also active data collection via surveys. Such designs have broad application, such as the assessment of practice patterns, patient experiences, and outcomes in physician offices.
In the hybrid model, patients can be identified in the EHR database and invited to participate. As these patients go online to register and provide consent, baseline survey information can be collected from sites about characteristics of the patients and providers. Prospective information can be gathered via surveys from healthcare providers, and EHR data can be continually assessed to determine if new patients are qualifying.
One key advantage of this approach over registries is that it reduces the research burden at the site and provides cross-validation of what the patient reports and what the physician reports, therefore enhancing the strength of the EHR data. A “look back” period can be incorporated to shorten the overall study duration and facilitate time and cost efficiencies.
With the expected surge in EHR growth globally, we can expect to see increased interest in novel approaches to use real-world data to supplement RCT data, providing a more holistic perspective on a drug’s true efficacy and effectiveness. The landscape for real-world studies will be changing dramatically over the next decade as EHR records become more accessible, and it may be that we can execute most of our studies by recruiting patients via social media and online patient communities, obtaining informed consent online, and combining EHR data with patient surveys.
Complementary approaches to evidence development are necessary to gaining the bigger picture of a product profile and to address different stakeholder requirements. Hybrid designs bring new opportunities for robust, more rapid evidence development at lower cost. Each approach (RCTs, prospective observational and retrospective database studies) has unique strengths and weaknesses. By using a combination of these approaches and leveraging the ability to supplement any missing information with existing data from routine sources such as EHR, researchers can fill evidence gaps more efficiently and effectively.
ABOUT THE AUTHOR
Dr. Terry Cox is director of biostatistics at Quintiles Outcome. He is a board-certified ophthalmologist with fellowship training in neuro-ophthalmology and a PhD in biostatistics. Since joining Quintiles Outcome, he has worked in a variety of disease areas, including cardiology, oncology, pediatrics, and diabetes.