The MS neurologist at Cleveland Clinic’s Lou Ruvo Center for Brain Health discussed the impact that propensity score has had on real-world data analysis, the use of additional outcome measures in trials, and the increasing understanding of progressive disease.
Carrie Hersh, DO, MSc, MS neurologist, Cleveland Clinic Lou Ruvo Center for Brain Health
Carrie Hersh, DO, MSc
Even with dozens of treatment options available for multiple sclerosis (MS), as research has plunged further into progressive disease and the optimization of therapy, the treatment landscape in MS has shown no signs of slowing.
As it continues to expand, the process of making treatment decisions, comparing therapies, and the focus on more personalized patient care has become far more complex. However, with the addition of new analytical methodologies and the further involvement of patients in the decision-making process, there has been some relief from these challenges. At the 2019 Annual Meeting of the Consortium of Multiple Sclerosis Centers (CMSC), held May 28-June 1, in Seattle, Washington, this very topic was the theme for a presentation aimed at helping clinicians and ancillary staff better understand how to interpret and analyze clinical trial data which has increasingly become less reliable as the only source of information about disease-modifying therapies.
One of the physicians leading that session was Carrie Hersh, DO, MSc, an MS neurologist at the Cleveland Clinic Lou Ruvo Center for Brain Health. Afterward, Hersh sat with NeurologyLive® to share her insight into the use of propensity score analysis, the incorporation of patient-reported outcomes, and the impact this will have as treatment for progressive MS advances.
Carrie Hersh, DO: Today's course [with Kavita Nair, PhD and Enrique Alvarez, MD, PhD] overall was a synopsis of the emerging research landscape in multiple sclerosis. Specifically, we wanted to focus in on how to understand analyze and interpret the different study designs that are currently in existence in the MS research landscape, but we also wanted to focus in on different research methodologies that are becoming more popularized now that we have different sorts of outcome measures that are being developed and also looking at how real-world effectiveness can complement randomized control clinical trials when we are trying to make conclusions on comparative effectiveness between existing disease-modifying therapies.
The reason why we find especially the latter is so important is because there are new drugs being developed very consistently and there are new drugs that are being pioneered, and because of the plethora of drug development that we have in MS, we understand that these drugs are going to become increasingly more challenging to compare head-to-head. Randomized, controlled clinical trials are not going to be able to satisfy that unmet need, and where we have some benefit is that we can start employing some unique statistical methods to turn very messy, retrospective observational data into a usable format so that we can make outcomes analysis. We can look at treatment effect differences between 2 different interventions. We’re essentially comparing apples to apples as opposed to apples to oranges, and part of the discussion of the course today was learning about a statistical method called propensity score adjustment that we can actually employ to essentially extrapolate usable data from very messy, observational real-world data.
Real-world studies are growing exponentially an interest and, again, this has to do with the emerging research landscape, the drug development landscape, where we have a very quickly growing armamentarium of available disease-modifying therapies and we're no longer going to be able to just rely on our clinical trials because we're just not going to have those comparisons. Being able to leverage what we're seeing in real-world practice using all sorts of different datasets from different institutions is going to play an increasingly important role.
Where we really need to be conscientious is using statistical methods to make sure that we are actually comparing our patients adequately, meaning that we want to make sure that our populations are similar enough so that way we are able to compare apples to apples as opposed to apples to oranges. Something called indication bias becomes a potential problem when we're trying to make conclusions on treatment effect differences between 2 different therapies. We have to be very careful about that, and that's where the propensity score methodologies certainly have an important use in our real-world effectiveness retrospective observational studies.
There are actually a whole lot of things that come into play when we're trying to make a decision between the existing therapies when we have a patient who is sitting in front of us. It's becoming increasingly more important to use something called shared decision-making or personalized treatment choices with our patients because they're just so many different therapies that are available. When we are trying to decide on 1 disease-modifying therapy versus another, we're looking at very important things, like making sure that it's probably going to work the best for that particular individual. It can be very difficult to predict if someone's going to respond to a particular therapy, but we have some indicators that might be able to give us an idea whether or not they should be on a more highly-effective therapy versus a lower-effective therapy.
It can depend upon the number of negative prognosticators they have, but we also want to look at tolerability. We want to make sure that our patients are going to be able to persist on a certain medication. We also want to make sure that we're looking at affordability and accessibility; we want to take into account if our patients are younger, that family planning might actually be something that's on the horizon, so is this particular disease-modifying therapy going to be appropriate if this patient wants to start family planning in the next year or so? There are different facets that we want to make sure that we're taking into account when we are having these very important conversations with our patients at the bedside in terms of clinical effectiveness and what might be the best outcomes. Those are some of the most important questions that we have.
Historically, we've used annualized relapse rates as our primary outcome measures for our pivotal phase 3 clinical trials that are looking at the efficacy of a certain drug versus either placebo or an active comparator, but now you know we're starting to use more drugs that have a place for our progressive patients. Is an annualized relapse rate the best overall primary outcome measure for that particular patient population if they're no longer as inflammatory? Looking at other kinds of surrogate measures of disability or the expanded disability status scale score, and looking at clinical definite progression or clinical definite improvement, and whether or not we can start using composite measures—so not just having 1 outcome but having 3 different outcomes that are all part of a single measure—is that something that we should start employing into our randomized control trials?
In terms of clinical effectiveness, those are the some of the things that we can look at, but then there's increasing interest in MRI radiographic disease activity, not necessarily just T2 lesion volume, which is essentially how much inflammation we have in the brain and the spinal cord. But looking at brain atrophy rates, especially when we're looking at a more progressive population of patients where brain atrophy is starting to become more prominent and is a measure of neurodegeneration. Patient-reported outcomes are also becoming crucial, especially when we are personalizing care at the bedside, so being able to implement these patient-reported outcomes, these questionnaires, into not only our randomized control trials but our observational studies as well. It's not really just necessarily 1 measure, but all of these things together that ultimately are going to be able to give us the best estimation of how a particular drug is working for a particular patient. It's going to be very difficult to take generalized, population-based data into an individualized case, but we do the best that we can, based on making sure that we are representing the patient populations well.
Transcript edited for clarity.
Hersh C, Nair K, Alvarez E. The evolving research landscape: understanding, designing, and analyzing studies in multiple sclerosis. Presented at: CMSC 2019; May 28 to June 1, 2019; Seattle, Washington.