Pharma is starting to use real-world evidence to get a more accurate picture of how drugs fare outside a clinical setting – but it needs to stop being a new concept and start being an integral part of operations
It may seem obvious that market access stakeholders would want to use real-world evidence (RWE) to truly understand how a drug works outside the artificial setting of a clinical trial, but it is only in recent years that people across the industry have realised just how beneficial it could be. In fact, FDA commissioner Robert Califf recently called real-world evidence a "top programmatic priority" for the future of regulation.
"No matter how good our decision-making, we function most effectively when we have access to high quality, reliable evidence," he said in a speech at the Food and Drug Law Institute Annual Meeting in May. "We now have the amazing situation in which nearly every American has an electronic health record (EHR), national repositories of curated claims data are available, and clinical registries are growing at a rapid rate. The consolidation of healthcare delivery into large integrated health systems means that research clinics are being folded into clinical care systems with increasingly robust governance systems and highly curated enterprise data warehouses.
"When these rich and diverse sources of digital data are coupled with increasingly sophisticated analytical tools, we are at a tipping point for generating the scientific evidence needed to make good decisions."
John Doyle, senior vice president, QuintilesIMS, has also seen a shift in the attitudes of market access stakeholders. "A lot of payers have come to appreciate that the data package that's submitted to a regulator is typically focused on a rather narrow type of patient population, often one that is younger and healthier than the population who ultimately receive the product," he says.
"That's a challenge, and real-world data (RWD) provides the means for the payer, the provider and the pharma company to work together to generate evidence of the effectiveness and safety of the product as it's used in a broader, more heterogeneous population, as well as begin to investigate the economics of using the drug. RWE begins to bridge the gap between the needs of regulators and other system stakeholders that control market access."
Alberto Grignolo, corporate vice president of PAREXEL, says that RWE helps address these gaps, where some data is good enough for regulatory approval, but not for reimbursement, by taking into account what really matters to patients.
"It relates to patient-centricity – what matters to patients, what endpoints do they want to see in clinical trials?" he says. "Those endpoints will also matter to payers, who can tell a patient that this drug will improve the quality of their life.
"Patients have been saying for some time that the endpoints in clinical trials aren't always relevant to them, so RWE should be brought into the conversation in late-stage and post-approval drug development to build evidence that matters to patients and payers so that a drug that gets approved also rapidly becomes relevant in the real-world, and you don't just get regulatory approval based on endpoints that matter to
Doyle says that this kind of information can then be fed back into drug development, meaning that the proliferation of RWE could even change the kinds of drugs that are developed in the future. "Generating RWE is in many cases still a late-stage effort in order to engage with market access stakeholders, or to follow up on a product to see how it is faring in the real-world. But what you could also do is create a feedback loop based on the information you're gathering in the real-world about clinical practice patterns, patient subpopulations, real-world outcomes, and even economics.
"Those types of data can be input into future clinical development planning so that you can better understand patient epidemiology, unmet medical need, and real-world practice patterns. You can also better understand the competition. You can use that input to actually begin to prioritise your portfolio of potential developmental assets based on your read of real-world need."
RWE in clinical trials
Because of this, the industry is increasingly considering using RWE evidence earlier on in development, in clinical trials.
"In certain diseases we're finding that drugs are more targeted, but that it makes it harder to recruit," says Rick Sax, senior vice president at QuintilesIMS Consulting Services. "If you wanted to do a phase 2 or phase 3 study, and you have a very specific patient population in mind, you could design the study in a way that could recruit patients broadly, but then you might not get the scientific precision to get a good outcome. Or you can recruit patients very narrowly, which might allow you to prove that the drug is beneficial in that targeted population, but narrows the population that will ultimately take the drug and narrows your label.
"You need to know how many of the right patients are out there, and one way to do that is to use real-world data to check your inclusion/exclusion criteria against the patient population. We can pull that directly from electronic healthcare records and understand the demographics of the population, what medications they're on, and what concomitant diseases they have. We also have tools that can analyse the data and allow it to be manipulated visually, for example to see what happens if you treat the population that's 18-65 versus 18-75 or what happens if you only have patients who are on certain medications."
But, generally speaking, pharma is still finding its way with RWE, and is often more focused on what it can do with the data in the present than how it can be integrated into future projects early in development. "In pharma timescales this is still a relatively new development," says Doyle. "Because of that they're often focused on the front burner issues such as the products that are about to launch, and how they differentiate them and secure optimal market access by using real-world evidence."
Marla Kessler, senior VP of strategy, marketing and communications at QuintilesIMS says that this narrow focus has been an issue for pharma companies so far, but they are beginning to look at RWD more holistically.
"In the past the industry has approached it as a series of one-offs," she says. "You have a problem, so you fix it by generating the evidence for it. It becomes a little chaotic because the scientific people and commercial people may be looking at different data sets and everyone is developing a different point of view about the same patient population.
"The industry needs to stop thinking about it that way and start to look at it as capability that really supports the entire business, perhaps creating a common foundation of RWD, a data network where anyone can access near-real-time data on demand. That way you can make more efficient data-purchase decisions and also figure out where the gaps are in the data that you need."
She continues: "You need to get as many different people together as possible and decide what issues you really care about in the market, what kind of patients you need to understand, what kind of outcomes you'd like to track, and the costs associated with these things. Then you build that fact-base to support those business, commercial and scientific needs and have an alignment about what is it you're investing in. That gives you the opportunity to work together and have a common sense of what's happening with all these other stakeholders. Instead of thinking about your real-world data as a set of databases, you think about it more as a network. It really gives you the ability to be more flexible."
Kessler says that using the right technology is one step towards this. "You need technology that's really built to deal with big data; then you can have machine learning, predictive analytics, and the ability to extract the data you want and figure out what insights you're generating. And then through the right governance and the right thinking about strategic business priorities you can give those insights to a different part of your business."
She says that one example is Novartis' branded and externally-promoted VERO (Value of Evidence in the Real-wOrld) platform. "It's a great example of a platform that robustly supports an entire franchise's understanding of what's happening in treatment so they can use that to talk to payers and understand what's happening in the market versus what they think is happening.
"The companies that have really gone big into this not only make better decisions for themselves, but also more effectively engage with the stakeholders around them."
There are additional challenges with the data itself. "The nature of RWD is that it's typically messy data," says Doyle. "One major difference between RWE research programmes and clinical trial programmes, is that when you don't randomise you risk confounding and bias. You need to deploy epidemiological and biostatistical methods to account for that. You could do it statistically, you could do it at your design stage, you could be thoughtful about clinical pathways and really try to hypothesise why a particular patient received a drug. Typically there's a host of risk factors that predispose a patient to receive one drug versus another. Confounding by indication can lead to erroneous conclusions if you don't account for it in your statistical analysis plan. You need to be very mindful of the threat to validity when you're doing RWE generation."
Sax adds that using electronic health records comes with its own problems: "You have to recognise that the purpose of electronic healthcare records is to capture things that are related to the broader healthcare delivery system. They are not yet really amenable for clinical trials.
"Therefore the records are really dirty – they're incomplete, they're not necessarily precise, and there are data gaps. It's also very difficult in the way that the records are structured to get any more than a snapshot of what's happening now. Following what happens to a patient over time is really difficult to do in the healthcare records system when the information is de-identified. Unique patient identifiers could be a way to solve this while still preserving absolute patient privacy.
"The incompleteness doesn't matter if you have a lot of records and you're looking broadly, leveraging big data techniques, but the level of precision isn't there for the regulatory purposes we need."
Kessler adds that there's a risk that people don't know enough before they make decisions. "For example, sometimes I hear people say 'I want the EHR data that has the most patient count'. Patient counts are one variable that help you understand how broad an EHR data set could be but is that data set representative? Or are you going to be skewed to a certain type of patient population or geography? Is EMR data going to give you the ability to answer the cost questions you have? Probably not, you probably need claims data for that. I think sometimes people are making decisions without having enough of the expertise as counsel to them to make sure they're making the right decisions."
Looking to the future, Doyle thinks that the balance between clinical data and RWE in market access decisions will be quite complementary. "It's almost like concentric circles, where you start with the tightest circle still being the clinical trial and the initial evidence generation around whether or not this drug has an effect. Then we begin to look at safety and begin to expand out to further concentric circles to evaluate how the product might work in regard to the standard of care, what its place is in therapy, and look at various patient subpopulations that we might not be able to see in the clinical trial setting. As you expand it will begin to move away from a trial setting and more towards an observational setting."
Likewise, in his speech, the FDA's Califf noted that RWD and randomisation are "different dimensions" rather than "polar opposites". He added: "When these two dimensions – source of data and research method – are optimised, a randomised trial using real-world data will be a highly valuable and efficient approach to generating evidence. This is not a new concept, and many important advances in medicine already have come from this type of study, often called the pragmatic clinical trial."
"Hopefully we will be making much better decisions for healthcare with RWD," says Kessler, "because instead of making a decision based on history, doing something just because that's the way you've been doing it for the last five or ten years, we can make decisions based on what's truly happening to patients."
Collecting real-world data
The rise of RWD has been facilitated by advances in digital technology that are able to accurately track patients and their health. Electronic health records are one prominent technology that HCPs and researchers can use, but others have been putting the power directly in the hands of patients.
"As a trend, people are increasingly taking their own well-being into their hands," says Robert McFarlane, head of labs at digital agency Head. "There are plenty of lifestyle products out there that foster this – Apple Watch, FitBit, Nike Fuelband, smart scales like the Withings Smart Body Analyser – adding to a quantified-self landscape. For clinicians and trial designers, this is an opportunity to use existing wearables, smart devices and smart environments to conduct trials in a safe and unobtrusive way, making the experience much more appealing to potential participants."
He adds that wearable tech and smart home devices don't have to be invasive or so confounding that it puts people off. Some examples of unobtrusive technology that can be used in clinical trials include:
- Micro needles (dermo patches)
- Infrared light used for non-invasive blood glucose monitoring
- Wearables that look fashionable
- Small cameras fitted around the home can track presence and alert cases of unusual behaviour – falls, seizures
- Mobile phones that can track location, give reminders, and be the means of delivering regimes
"When data collection can be automated, you can expand the variables you gather from just one or two points of interest to an entire lifestyle," says McFarlane. "If you're trialling a drug that affected blood pressure, it's easy to just gather blood pressure data. But adding more data points like exercise, walking distance, diet, sleep patterns, water consumption, browser history, air quality, temperature, even what they watch on Netflix (all of which can be digitally monitored) creates a richer profile of the subject and more points of interest that might have an effect on someone's well-being."