Article | February 25, 2019

Real-World Evidence: Explaining The Current State Of The FDA's RWE Program

Source: Ciox Health - Life Sciences

By Paul Roma, CEO, Ciox Health

Research scientist receiving drug approval virtual

Like most instances of adopting new ideas and government programs, few people in drug development have a firm grasp upon and truly understand the FDA’s real-world evidence (RWE) program and the impact it will have on clinical research and trials. This article will guide you through the program and what it means for drug developers and explain the potential that real-world data (RWD) offers in developing RWE to significantly and safely accelerate the drug approval process.

RWD vs. Trial Data and Converting RWD into RWE
To fully understand the RWE program, it’s essential to understand the relevance of structured and unstructured data, and the processes used to collect it, in the drug development process. While all data has some type of structure, the difference between structured and unstructured data is determined by whether it has a predetermined data model and if it’s organized in a predefined manner. Over the last decade or so, the gold standard in the drug development process has been randomized trials. In the trial process, data is collected, but it is not necessarily done in a real-world clinical setting in therapy; it’s more curated from a processing perspective. Data captured outside of the randomized trial process — produced at the point of care, in real-world clinical scenarios — is considered RWD. 

RWD has been used for about a decade in various phases of the drug development process. However, the distinction between RWD and RWE has been made within the last three years. In that time, there's been a big differentiation between data just captured in real-world scenarios and data that has proper support and evidence, so it can be used in clinical responses, as well as for use in the decision-making process while engaged with regulators regarding labeling and drug usage.

The conversion of RWD to RWE typically comes from a clear understanding of how RWD was gathered. This includes the method through which the data was accrued, the integrity of the data, and a knowledge that it is in the proper format. From there, the data can be matched to journaled evidence and can begin to be shaped into rudimentary RWE applicable to some decision-making. Decision-making using RWE is more easily understood and contributes to drug safety. Further, post-trial, RWE offers insight into adverse and/or rare side effects from long-term use of a drug that would not be seen in a randomized trial.

Accelerating Development and Innovation

Back in late 2016, The 21st Century Cures Act was signed into law. The law’s intention was to accelerate medical product development and bring new innovations faster and more efficiently to patients. The RWE program stems from a portion of the Cures Act, Section 505f of the Federal Food, Drug, and Cosmetic Act, that created a basis for evaluating the potential use of RWE to support the approval of a new indication for an already-approved drug or to support and/or satisfy post-approval study agreements. Then, in December 2018, the FDA published its guidance, Framework for a Real-World Evidence Program. The guidance was designed to enable drug developers that use RWD to develop RWE to support regulatory requirements.

The Standards (or Lack of) for RWE
Currently, no regulating body clearly defines the standard for RWE. Because of this, there isn’t a clear process to convert RWD to RWE. Despite the FDA offering help in the conversion process, there is still much confusion.

RWD, while it has many advantages, comes with limitations. The most significant disadvantage is the quality of the data. Let’s examine a hypothetical example. A study consists of 1,000 congestive heart failure patients using beta blockers and ACE inhibitors. Of those patients, only half of them are properly dosed. Then, as RWD is examined, it becomes clear there is no differentiation from the point of care on whether a dose was given for an acute episode or a larger dose in the loading phase. Caregivers do not necessarily dictate all thoughts regarding dosage in a medical record. Instead, they usually transcribe the order sets and prescriptions. This is a major limitation in the fitness of data captured at the point of care and the ability to analyze that data as part of the scientific process.

Still, this raises many questions. What does fitness mean? What is statistically rigorous enough within RWD to convert into RWE? What forms of journaling RWD should be published? What forms of peer review need to be performed? How much data is statistically relevant? What forms of data are statistically relevant?  Unfortunately, there are not many concrete answers to these questions. As a result, drug developers are beginning to answer those questions and write their own recipe. The FDA has provided guidance indicating that it wants to work with the industry to create this formula in hopes of creating a network effect — interactions between researchers, regulators, caregivers, clinicians, and others. This wide-ranging interaction should create more evidence over time, leading to exponential growth in data for evidence.

While there isn’t a literal, direct data format standard for converting RWD into RWE, the education and collaboration processes and the technology to shape a standard are currently happening. We are at the beginning of what will ultimately be an explosion of data yielding positive results for both the industry and patients.

Further, the industry is just beginning to break down data into syntactical and semantic standards. Right now, the industry uses syntactical standards or plain format. Diagnosis codes are sent for billing purposes, and they come with a great deal of context, but there is no standard on reporting. For instance, in pathology reports, there are no standards for tumor measurement or description. A rich data set is achieved, and, ultimately, this is beneficial, but the industry needs to work on definitions for complex observations that can be codified into both semantical (context) and syntactical standards. While syntactical standards can be interesting and a better read for humans, the goal is for a computer to analyze data with high confidence, convert relevant pieces into RWE, and apply it to trials and decision-making processes. Imagine that a computer with context and semantics adds 100 IQ points to the read of the chart and creates the best answer.

The FDA is interested in creating and promoting standards, and the industry is at the beginning of that process. However, first it must work on formats. Ideally, code formats will become code assignments, which develop further into code sets and compound from there into a way to be gathered, incorporated, and used. This will be a long, arduous process, but it can happen by breaking the problem into smaller parts. One example is by formalizing codes for capturing vital signs. Currently, we have the syntax for what vital signs are, but we don’t have a formality as to when they were captured, how they were captured, and when they were tested. We can begin codifying and publishing how they’re captured, laying a foundation everyone can use. This will improve the clinical process and the accuracy of vital signs measured and digested. From there, other areas, such as codifying oncology, could be tackled.

Ideally, both syntactical and semantic standards will be fulfilled for all clinical areas, leaving a data set for both computer analysis and human intake and communication. The human version should be a visualization of what the computer will interpret, but the computer doesn't need all the visuals to incorporate and process data, as it relies on code to create RWE.

Industry Perceptions of RWE
In addition to confusion about what constitutes RWE, the industry is delineated based on the perception of RWE. These perceptions generally break down into two camps. The first camp believes randomized trials are good enough and the industry should adopt a trial design with an interventionalist mind-set and improve on that. This camp believes the scientific approach has fully curated endpoints with precise targets. However, this is a minimalist view of the landscape.

The second camp, primarily supported in the industry, believes that much of the quality data available to both scientists and clinicians is ignored. This camp wants to stop the omission of this data and become educated on how to use it through the trial process and in post-trial and approval.

While the scientific process is very pure, at times it puts up roadblocks — timing, money, and lack of a patient study group, to name a few — on the path of innovations intended to help patients. The innovation process does not involve, or even care about, economics or the timing of situations. Instead, it cares only about efficacy and helping patients. For instance, Microsoft funded a RWD program on pancreatic cancer. That analysis was able to identify, with more than 90 percent confidence, cases of pancreatic cancer six to 12 months in advance in 5 to 15 percent of patients while having false positive rates as low as one in 100,000. This wasn’t RWE, but it was a real data process offering prognostic indicators that could be used in conjunction with other RWD to form RWE and fight a deadly disease up to a year prior to diagnosis. 

The ability of RWD to quickly establish hypotheses that have real statistic evidence is changing the face of R&D. A trial for oncology will pass a specific tumor type with flying colors with 95 percent apoptosis and no toxicity. Once approved and used on patients that have multiple tumor types, you immediately have RWD in bulk. This RWD is almost immediately adapted in practice by doctors, which is why so many prescriptions are off-label and still widely accepted.

The RWE Program and the Hybrid Design
Most countries already have RWE programs, including:

  • Japan’s Rational Medicine initiative
  • France has built a large national database spanning across millions of patients.
  • The U.K. has the National Institute for Health and Care Excellence (NICE), considered a leader in the way RWE can inform decision-making in a healthcare system.
  • Canada’s Canada Network for Observational Drug Effect Studies
  • Australia’s Victoria Comprehensive Cancer Center

The U.S. only became formal with RWE with the introduction of the previously mentioned Cures Act, embracing Big Data and realizing that more data is being produced than ever before, and it’s better, more accurate data than has ever been created. This has quickly started to change the way trials are completed.

In the legacy trial process, drug development only uses randomized trials. The FDA strongly desires that drug developers change their trial designs. In fact, the agency’s RWE guidance not only promotes the use of RWE but also pleads with drug developers to embrace it using a hybrid design for trials — the combination of the randomized trial process with professionally curated research data along with outside data (drug adherence, compliance, multiple co-morbidities, and potential social, economic, and lifestyle impacts). The hybrid design, by nature, allows drug developers to narrow the number of participants in the trial and what they are trying to get approved. In turn, this increases accuracy and hastens the approval process. Further, the hybrid design creates a narrowed data set for the approval process, as well as data regarding long-term efficacy and safety concerns.

In a hybrid design, data is available from both patients in a specific trial and all users of the drug from outside of a specific trial. This RWE captured in a post-approval setting paves a path of continuous improvement from start to finish and creates a lengthy scientific dialogue, ensuring approved drugs are closely monitored and any future complications or contradictions are understood. Essentially, it creates the long-term engagement model the FDA is hoping for.

RWE and Drug Safety
The benefits of RWE regarding safety have already been demonstrated in not only the previously mentioned countries’ programs but domestically as well. In the U.S., as recently as 2011, most health plans had contraindication of medications built into normal script processing. If the contraindication caused safety concerns, healthcare professionals, as well as the patient, were notified to create follow-ups and corrective actions. This is perhaps the first proven case study of how RWE, information outside of a randomized trial, is used to predict safety events and intervene to prevent issues.

Expanding upon that, the FDA is using data from the Sentinel Program to factor how data can be used to improve drug safety. The thought here is to embrace Big Data; an understanding of contraindications can be used to weed out morbidities and co-morbidities and, on a survey level, understand mortalities and potentially long-term effects. This creates an excellent data set that drug developers can factor into making hypotheses about safety concerns and intervention programs. From those hypotheses, drug developers can reach out to agencies heavily involved in the Sentinel Program, such as Duke Medical Center, and explore the hypotheses with small, purpose-built data.

Those hypotheses can be held against RWD to begin understanding co-morbidities of the drug and if it is achieving desired outcomes. For instance, using the beta blockers and ACE inhibitors example again, these drugs are developed for congestive heart failure. However, in a patient with congestive heart failure along with diabetes and obesity — multiple co-morbidities — the therapy will not have the desired outcome. Instead, there is no cure rate and the therapy has other side effects. In a randomized trial alone, there wouldn’t be a comparative data set and treatment patterns wouldn’t be fully understood to create therapies related to the real world.

The RWE program is designed, from a safety perspective, to give clear comparisons to the real world — what happens at the point of care and what happens with co-morbidities and their treatments. It also provides insight into social, economic, lifestyle, and behavioral aspects of therapies. These all factor into the overall safety of a drug in normal and adverse situations.

Capturing Disparate Digital Data and Making it Usable
Drug developers talk about clinical data being connected and digital, and while the industry is advancing in terms of digital data, it isn’t as connected as it should be. In fact, nearly 75 percent of electronic data in the U.S. is called electronic because it travels through a fax machine. Further, the average U.S. hospital currently uses 16 disparate electronic medical record (EHR) systems, leaving only about 2 percent of hospitals conforming to two or fewer systems. This creates a very broken and siloed process in collecting data and has been a problem for nearly two decades.

We are now able to bring disparate data sets together using AI to structure data and make it quickly usable, without human intervention. AI is giving researchers and clinicians the ability to take multiple EHRs, combine them, and create a structured output for analysis in real-world applications. Essentially, technology is building a bridge that allows for analog, siloed, unstructured data to be connected to digital data and used in meaningful, real-life ways. The real breakthrough will happen when the data sets link directly to the point of care, in large quantities, for a better standard of how data is captured and written. Fortunately, the industry is not far from reaching this. Much like Moore’s law says the number of transistors in a densely integrated circuit doubles every two years — exponentially growing the power of a computer — technology and software are on a path of similar exponential growth. AI will soon be able to seamlessly bridge disconnected, unstructured, siloed data with data from the point of care. When this happens, real and meaningful breakthroughs with RWE will take place.

Next Steps
Pharmaceutical companies with a heavy hand in drug development would love to see more cross-agency collaboration, particularly among the FDA, NIH, U.S. Department of Health and Human Services (HHS), and CMS. CMS and HHS have the most extensive collection of RWD in the world, and the data sets of CMS are the most curated. Cross-agency programs could refine the quality of data and jump-start the creation of RWE standards. For instance, CMS receives 40 million full EMRs, with both pharma and medical claims, annually. This is a tremendous amount of data, but the primary concern of CMS is reimbursements, so all that data is only used internally. An inter-agency collaboration with the FDA would be just the tip of the iceberg in uncovering new, meaningful ways to use the data and create RWE and the standards it’s currently lacking.

Converting RWD into RWE and the standards associated with that conversion are still in their infancy. The decisions that will be made from RWE and the outcomes from those decisions are yet to be seen. All of these factors, coupled with the current state of siloed data and medical records, have created a great deal of ambiguity and confusion for drug developers.

To align with the FDA, all stakeholders must come together and act in a more integrated, coordinated fashion. First, we should recognize that data is not solely for the purpose of billing and consider critical clinical implications as well. Second, we should form a leadership consortium between government and industry leaders to discuss standards for converting RWD to RWE in a repeatable, secure way to produce data we can trust and act upon.

While the RWD and RWE landscape in the U.S. is currently difficult to navigate, the FDA has deviated from the norm and given the industry a unique and incredibly valuable opportunity to create a road map for it. Ultimately, the efforts of clinical researchers and technology companies — traditionally, an unlikely partnership — will form the foundation of design requirements that should drive fast-tracked drug development, accelerate FDA acceptance and approval and, most importantly, improve patient care.