Why Real-World Evidence Will Challenge Biosimilars

By Anna Rose Welch, Editorial & Community Director, Advancing RNA

In a recent article, I discussed some of the questions stakeholders need to consider when approaching real-world evidence (RWE). But there are a number of challenges the industry will face when attempting to answer these questions.

There are several ways the industry can create or find viable post-marketing evidence to determine the long-term safety and comparability of biosimilars. For one, companies can choose the observational cohort study route. There are also the large claims databases established by the immense U.S. healthcare system.

But despite the existence of an already vast amount of available biosimilar-related RWE (or Big Data), there are pitfalls in securing and interpreting this data.

As Brian Bradbury, executive director and head of the Data and Analytics Center within Amgen’s Center for Observational Research, stated at the DIA Biosimilars 2016 conference, “A lot of us in the epidemiology community are talking about better data, not just bigger data. With bigger numbers, you can get a precise, but still wrong, answer. A critical issue we have to think through is how to understand the factors impacting drug performance and design studies to minimize error in obtaining reliable evidence.”

Why Is It So Difficult To Interpret Real World Evidence?

A number of factors can influence a patient’s outcome, including age, exposure to other medicines, and inconsistencies in reporting outcomes. There are also issues with adherence. In clinical trials, medication is part of an established protocol. But in the real world, doses are missed. Patients aren’t always capable of making it back to the hospital for their treatment, nor can they always afford the copays (or the medicine as a whole).

It may seem, on the surface level, that discontinuation of a medication would eliminate some valuable RWE on the long-term performance of a specific drug. But Bradbury emphasized the importance of looking further into those patients who don’t continue on their biologic. “Discontinuation is a very important construct when approaching medicines that are given chronically,” he said. “It’s very important to identify those who stay on the medicine and those who discontinue. Is there a factor driving adherence that may be related to the patient’s underlying clinical status?”

Another big issue hindering interpretation of effectiveness is the different dosing regimens of biologics. Take for instance, erythropoietin-stimulating agents (ESAs), which are often administered to dialysis patients three times weekly during dialysis sessions. The ESA can also be given to patients once a week or once a month. Similarly, if we look at the biologics in the anti-TNF space (i.e. Enbrel and Humira), there may be different initiation doses, or they may be administered on different schedules.

 “You have lots of different dosing paradigms,” Bradbury explained. “If one’s given monthly and the other quarterly, how do you differentiate between effectiveness and safety in the real world?”

We also can’t forget the impact of age. For instance, a dialysis patient, on average, is in their mid-60s. Diabetes and high blood pressure are common conditions affecting this demographic. As such, these patients are likely on several therapies to control these other illnesses. These factors make it difficult to find a comparable group of patients in the real world.

The Ins-And-Outs Of Current Databases

Bradbury homed in on Medicare and private insurance claims databases, because they provide beneficial insights into what goes on in inpatient and outpatient settings. There are also electronic health records (EHRs) systems. However, all of these systems currently have limitations.

For one, though claims sources provide a thorough look at in-patient and outpatient settings, they often lack detailed laboratory data and diagnostic testing information. Similarly, while some EHRs may be more integrated — take, for instance, the partnership between Kaiser Permanente and Veterans Affairs — others tend to be electronic medical record platforms that decouple outpatient from inpatient care.

The ultimate goal is to find a way “to combine those systems to get rich, granular, clinical data, coupled with complete capture of the necessary outcomes,” Bradbury described.

A big question circulating through the biosimilar space is how we will distinguish biosimilars from each other in claims information, since all biosimilars to a reference product will share the same J-code.

A possible solution is to rely on the drug’s national drug code (NDC) when notating what medication a patient has been given in the hospital. But, after seeing some of the data Bradbury provided from administration claims, it’s clear we are not yet at a place where we can distinguish drugs using the NDC method.

Take, for instance, the data Bradbury shared from UnitedHealthcare’s Optum dataset. Upwards of 50 percent of the G-CSF claims (i.e. filgrastim and pegfilgrastim) did not report NDC codes. He also showed the crowd a slide illustrating the use of ESA NDCs per state on J-code claims. The inclusion of an NDC in addition to a J-code varied widely from state to state. (For those unfamiliar with J-code or administration claims forms, there is a field for recording the NDC.) 

It’s also important to consider the fact that key effectiveness indicators for medicines are not routinely captured for a variety of indications, including psoriasis, rheumatoid arthritis (RA), diabetes, and neutropenia. These effectiveness measures include the Psoriasis Area Severity Index (PASI); DAS28 scores (RA); HbA1c (diabetes); and absolute neutrophil count (ANC in neutropenia). In other words, the information captured in these data systems “may not be exactly what we need to establish comparative effectiveness,” Bradbury argued.

There is a potential beacon of hope for biosimilars in the dialysis space, considering Medicare captures hemoglobin levels — the dialysis effectiveness measure — on a monthly basis. This opens the door to evaluate effectiveness of biosimilar ESAs. “But I would posit to you that effectiveness studies may be quite challenging until we get this better data,” he offered. “I think we’re moving toward better data. However, with the systems we have in place right now, we have to be very artful in how we design studies so we get the answers right.”

How To Obtain Reliable Real-World Evidence

Bradbury ended his presentation with a great list of questions we need to consider when setting out to evaluate long-term use of biosimilars. Overall, the answers (or lack of answers) we come up with will determine how resources are allocated.

For starters, the industry needs to determine whether it is primarily more concerned with the safety or effectiveness of biosimilars. (We can certainly be concerned about both, and we are. But we need to determine “a first order of questions.”) For instance, was efficacy essentially determined through the biosimilar’s development program? Are we more concerned about safety, and, if so, which components of safety specifically? Which questions can or cannot be answered currently?

Then, it’s time to consider how to design studies to determine the answers to these questions. Bradbury’s presentation did not go into the nitty-gritty aspects of establishing the right design — that was the topic of another presentation on the same panel. However, he did ask the million- dollar question the industry is facing now as more biosimilar data continues to be released: “Are we willing to do prospective data collection and wait that long, or can we use available data systems?”