Real World Evidence: Supporting Decision-Making
How can pharma use observational data to support discovery, research & early development?
With all the excitement around RWE these days, it can be easy to forget that epidemiologists in industry have been using it for many years to generate insights alongside other forms of evidence. “What’s changed is the availability of it,” said Alex Thompson, Head Strategic Epidemiology, UCB. “The scale and penetrance of RWE is greater than it’s ever been as more and more healthcare providers move towards linked computerized records. Moreover, demand for RWE is increasing as decision-makers within and outside pharmaceutical companies become increasingly aware of what it can offer. The landscape is changing.”
Thompson is setting up a new epidemiology function at UCB, which will support discovery, research, and early development up to proof-of-concept. Their goal is to use observational evidence to better characterize unmet need; to exploit genetic and biomarker evidence to increase understanding of pathophysiological pathways; and to generate valuable real-world patient and disease insights. “Our intention is that epidemiology will complement preclinical and clinical studies so that we can better prioritize research areas; inform target product profiles and clinical development plans; and kill bad ideas as soon as possible,” Thompson explained.
Raison d’être
One of the main ways we use RWE is to test the assumptions we make about disease natural history and patient characteristics. In addition, “RWE can be used to provide quantitative information to support forecasting; to better understand treatment pathways; and to inform the specifics of clinical trial design”.
Despite fears among some that RWE can be used for “fishing expeditions,” we aren’t using RWE for hypothesis generation. Rather, it is a tool to verify the generalizability of observations from key opinion leaders or that are found in the literature. “This could be particularly helpful when considering really novel indications, as this kind of data would allow you to confirm whether a given phenomenon is just one individual’s or study’s experience, or whether it is happening across the board.”
Looking into the future
With increasing accessibility of RWE, its potential uses are also growing. “There are a lot of exciting directions in which it could go,” Thompson suggested, naming the use of RWE as historical controls for single-arm trials. Another potential application is in case ascertainment, which could allow study sponsors to conduct trials that Thompson described as “larger trials, with a lighter touch.”
A key benefit of large-scale datasets that can be interrogated in real time is that they can generate greater insights than is possible using summary data from publications alone. Such datasets allow you to answer the 2nd, 3rd, and 4th-line questions, which inevitably emerge when you answer the 1st line query. Moreover, as we move towards greater sub-classification of disease and increased acceptance that not all patients will respond to a given therapy, RWE allows us to better understand what patients in these potentially under-studied subsets look like. However, while “the promise is there, the delivery is far from systematic. There are ad-hoc examples of where this has been useful, but we’ve still a long way to go,” Thompson pointed out.
More than one source
Amid the interest in RWE, it’s important to remember that it’s just one source of information. For example, in early development, other forms of useful non-experimental data include patient and physician surveys, market and outcomes research, the published literature and competitive intelligence. But, when evidence from multiple sources aligns, decision-makers can gain confidence; when they diverge, they have the opportunity to challenge their assumptions and seek additional information prior to decision-making.
While recognizing the need to use multiple sources of data is crucial, it is just as important to keep in mind their strengths and limitations. “One of the key differences between observational and experimental data is that RWE often provides answers that have greater variability and more associated uncertainty,” Thompson explained. In experimental settings, we carefully remove many of the potential sources of noise and uncertainty that exist in the real world. As a result, the estimates you get from RWE can be less precise than experimental data. “Imprecision is not a bad thing if it appropriately reflects what’s happening out there. Still, it is a real challenge to encourage people to embrace that uncertainty.” Alternatively, because of the very large numbers of participants found in many real-world datasets, it’s also possible to generate estimates that are perhaps spuriously precise. Moreover, “given the timescales of early development, a real challenge for us is how to get real-world insights quickly enough,” Thompson said.
Managing expectations
Culturally, we work in organizations that are used to dealing with precise, unbiased, or randomized evidence. It requires a shift to embrace the nuances in observational data".
Despite providing large patient numbers, increasing detail, and highly valuable contemporary real world context, RWE isn’t a panacea. In addition to technical pitfalls that await anyone working with a large dataset, pharma executives need to manage their expectations about what insights RWE can provide. The interest in real world evidence is growing, and will not subside, as there is an increasing number of impressively sized and outwardly detailed datasets becoming available. “One of the challenges for epidemiologists is to explain what this data can and cannot do, and how to interpret it. Culturally, we work in organizations that are used to dealing with precise, unbiased, or randomized evidence. It requires a shift to embrace the nuances in observational data,” Thompson concluded.
RWE is an old resource with new opportunities. With growing availability of multi-million-patient databases and linked electronic health records, among others, the possibilities to generate information for decision-making are unprecedented. But limitations apply, and RWE is as much about careful interpretation as it is about size, detail and reach.
Alex Thompson will be speaking at the Real World Evidence Conference, June 3-4th, 2015 in Amsterdam. For more information, click here.