Epilepsy: Predicting Outcomes of Temporal Lobe Epilepsy Surgery

Podcast Transcript

Introduction: Neuro Pathways, a Cleveland Clinic podcast exploring the latest research discoveries and clinical advances in the fields of neurology, neurosurgery, neuro rehab, and psychiatry.

Glen Stevens, DO, PhD:

Temporal lobe resection is known to be highly effective in reducing or eliminating seizures in patients with medication resistant epilepsy. However, some patients experience postoperative complications such as depression or cognitive decline. In today's episode of Neuro Pathways, we are discussing the development of prediction models that can assist clinicians in identifying patients who may be at increased risk for these complications and ultimately aid in clinical decision making for both physicians and their patients. I'm your host, Glen Stevens, neurologist/neurooncologist in Cleveland Clinic's Neurological Institute. I'm very pleased to have Dr. Robyn Busch. Join me for today's conversation. Dr. Busch is head of the section of Neuropsychology in the Center for General Neurology and a staff neuropsychologist in the Epilepsy Center within Cleveland Clinic's Neurological Institute. Robyn, welcome to Neuro Pathways.

Robyn Busch, PhD:

Good afternoon, and thank you so much for having me.

Glen Stevens, DO, PhD:

So Robin, just for our audience that's out there so that we can all know you a little bit better, just give us a little introduction to your background, how you ended up coming to the Cleveland Clinic and your interest in neuropsychology.

Robyn Busch, PhD:

Well, I'm a clinical neuropsychologist. I did my graduate work at the University of Cincinnati and then did an internship down at the James Haley VA in Tampa, Florida. I originally hailed from Los Angeles and had planned to go back home, but came to the Cleveland Clinic to do a postdoc, fell in love with neuropsychology and working with epilepsy patients, and then joined the Epilepsy Center after I completed my postdoc.

Glen Stevens, DO, PhD:

Well, great. Well, we're glad to have you here.

Robyn Busch, PhD:

Thank you so much.

Glen Stevens, DO, PhD:

And all the great work you're doing. When I trained terminology changes, when I trained, we always used to call the medications that we would use to treat seizures and epilepsy, antiepileptic drugs. And of course, even though I'm getting old, I'm okay to change. And now I know that we should call an antiseizure medications and in general, anti-seizure medications are good for controlling seizures. And boy, there's been a lot developed in the 30 years that I've practiced. I mean, when I read reviews it says greater than 20, but I'm pretty sure it's getting closer to 30 medications. But sadly, despite that, my understanding is that about 30% of patients have drug resistant epilepsy, which as you know, we sort of define as failure of two adequate medications. So, we need to look at other options, which is where you come in and your area. And that is surgery in at the Cleveland Clinic we do a lot of epilepsy based surgery. So, tell us about epilepsy based surgery and some of the issues and potential complications and why a neuropsychologists needs to be involved in patients that have epilepsy based surgeries?

Robyn Busch, PhD:

Yeah, so neuropsychologists are a part of the epilepsy surgery team. Really our role is to characterize patients cognitive strengths and weaknesses prior to surgery and to help predict what types of issues a patient may have after they undergo epilepsy surgery, so that a patient can be counseled appropriately and it can inform clinical decision making in terms of determining the best treatment option for the patient.

Glen Stevens, DO, PhD:

So, talk to us a little bit about your predictive models. What kinds of things do you look for in your predictive models?

Robyn Busch, PhD:

So, the models that we've developed to date have addressed the most common issues that patients experience following temporal lobe resection. And so to date, we've developed models for verbal memory outcome, for naming outcome, and also for mood outcome or an increased depressive symptoms. And so, all of the models have taken information gleaned from the literature regarding the risk factors for these different types of outcomes, and really tried to consolidate them into multi-variable models that can take numerous risk factors into account simultaneously, and help the clinician to predict the likelihood of a decline in an individual patient given their unique clinical and demographic characteristics.

Glen Stevens, DO, PhD:

So, I take it you could then put into a computer various parameters and then it would tell you the likelihood that somebody will have a specific abnormality, or what do you glean from it?

Robyn Busch, PhD:

Yes, that's exactly right. So, what we've done is taken large cohorts of patients who have undergone temporal lobe surgery in the past, who have completed neuropsychological evaluations before and after their surgery, and then we look at a host of different risk factors and put them into logistic regression models to try to determine the best combination of variables that will give you the most accurate prediction in identifying someone who declined versus someone who did not.

Glen Stevens, DO, PhD:

Yeah, I thought when I go see my financial guy and they put my stuff through the Monte Carlo, it gets complicated but this probably gets fairly complicated, right?

Robyn Busch, PhD:

Yeah, it can be. Now, what I will say is for the initial models that we developed, we actually tried to keep them quite simple. And so, we wanted to use variables that could be very easily accessed, so that the models could be widely used regardless of what epilepsy center you're at. So, in these initial models, we've included things like age and sex and preoperative cognitive ability, age at onset, duration, these types of variables that should be easily accessible. What we hope to do in the future is include more complicated and sophisticated variables that might help increase our predictive accuracy.

Glen Stevens, DO, PhD:

And have we learned anything that we didn't expect, or not necessarily, we kind of had a pretty good idea of what would happen?

Robyn Busch, PhD:

No, I don't think there was anything that we didn't expect. I think the variables that we knew were associated with postoperative outcome were important. What I think this has really done is help clinicians consolidate risk factors when there's conflicting information. So for example, we know that preoperative cognitive performance is one of the strongest predictors of outcome with individuals who have higher baseline functioning, more likely at risk for a decline after surgery. We also have known, for example, that individuals whose seizures start later in life or who are older at time of surgery are more likely to demonstrate cognitive declines. The difficulty as a clinician is to consolidate these factors when they're contradictory. So, what do you do if some of the risk factors, the patient would be at high risk and some they would be at low risk, can be very difficult to determine how to weight those different factors. And so, what these nomograms or prediction models will do is consolidate all of that information to tell the clinician the probability that their patient is likely to have a decline on whatever specific cognitive measure of interest.

Glen Stevens, DO, PhD:

And do you have cases where patients are having so many seizures that it's affecting their cognitive function, they're on so much medication that when you do predict that they're going to have a certain outcome, yet when their seizures stop and they're on less medication, maybe there's a benefit there? So, I can see how it can get a little complicated.

Robyn Busch, PhD:

Yes, exactly. You've hit the nail on the head that there are certainly a number of factors that can contribute to cognitive function in individuals with epilepsy, not only the seizures themselves but the medication side effects. Oftentimes there's mood or attention difficulties that come with that as well that can impact cognitive function. At this point, the models are pretty simplistic and sort of use some of those basic variables, but you are correct that if patients are rendered seizure free, able to come down on some of their medications, oftentimes their attention may be better, which can help with other cognitive aspects as well.

Glen Stevens, DO, PhD:

And are these models specific for temporal lobe epilepsy? What about extra-temporal epilepsy, different models, things you're working on?

Robyn Busch, PhD:

Yeah. So, all of the models we've developed so far are in temporal lobe epilepsy because that is the population of patients that is most likely to demonstrate a change in cognition following surgery, and that's the most common type of epilepsy surgery. So, we have the largest data sets in patients who have undergone temporal lobe resections, but we do hope to extend that to other types of epilepsy surgery

as well. As well as so far we've mostly focused on adults, but we do also plan to develop models for children.

Glen Stevens, DO, PhD:

And I'm curious, are there other models that are out there, other centers are using or national guideline models available or no?

Robyn Busch, PhD:

So, in the literature historically, different groups have developed regression models, but the vast majority of those have not been externally validated. And oftentimes they'll include variables that are specific to that center. So, maybe a particular FMRI protocol or a particular test that may not be used more widely. So, the goal here was to develop models that as I mentioned previously, included variables that were accessible at any center. And then we also externally validated them in a large cohort of patients from six other epilepsy centers in the United States and Canada to try to increase the generalizability of the model and the utility, so that other centers could also use it.

Glen Stevens, DO, PhD:

Well, I'm Canadian, so I'm very happy to hear that you're validating that in Canada as well. We're a socialized system up there, so we can't afford to do all those things ourselves. So, thank you. Are these prediction models available then to other, I'm sitting at a hospital, a state away and I'm thinking, boy, this sounds good. Can I get access to this?

Robyn Busch, PhD:

Yes. So, we've developed nomograms, which are sort of a paper tool that you can use to use these, and those are published and individuals can access those by going to the publications. But we also have an online risk calculator that's available at riskcalc.org, and anybody, any clinician can go there and feel free to use that software. If you enter in the different risk factors, it will ask for those individual variables. The clinician inputs them into the system, clicks a button that says "Run," and it will give you the predicted probability for decline on each of the different cognitive measures that we have developed tools for so far.

Glen Stevens, DO, PhD:

So, the outcome is really sort of a number, is that right?

Robyn Busch, PhD:

Yes, exactly. So, what it does is it gives you a number that tells you the probability that this patient will decline. So for example, let's say it came back as 70%. Then basically what that tells you is that if you had a hundred individuals that were similar in their disease and demographic characteristics to this patient, then 70 would decline.

Glen Stevens, DO, PhD:

And then ultimately the clinician with the patient would have to define the risk for each individual and what they're willing to accept or not accept.

Robyn Busch, PhD:

Yes, exactly.

Glen Stevens, DO, PhD:

So, let's say I'm going to use one of these models nomograms, anything specific I should know about it? Are there any nuances to it or fairly straightforward?

Robyn Busch, PhD:

The most important thing to know is that you should only apply them to patients that are similar to those that were included in the studies that we've published. So for example, most of these models are developed for older adolescents and adult patients who underwent a temporal lobe resection. So, you wouldn't want to apply it to a patient who had a different type of surgery such as an ablation, and you wouldn't want to apply it to a patient that was very demographically dissimilar to the patients in the study. The other thing to note is that we excluded individuals who were left-handed and who did not have a language lateralization procedure. So, you wouldn't want to apply it in those individuals either.

Glen Stevens, DO, PhD:

Very helpful. So, what are your long-term goals for these prediction models?

Robyn Busch, PhD:

Well, we'd like to continue to update and refine the models over time in order to increase the predictive accuracy. The predictive accuracy is pretty good and most of them, and certainly much better than chance. But I think we can continue to do better, particularly if we use more complex predictors such as preoperative workup, evaluation, information such as EEG, MRI, more complicated and sophisticated models, and we're working on that now. We'd also like to develop models not only to estimate the presence of postoperative decline, but also the magnitude of that decline and how it might impact functioning for the patient. And then ultimately what we would like to do is develop and validate a suite of prediction models on the same patient cohorts in order to simultaneously predict all outcomes that are relevant and meaningful to patients, such as the chance of seizure freedom, along with the potential risks to cognitive and mood changes.

Glen Stevens, DO, PhD:

So Robyn, what do these prediction models mean for the future of epilepsy surgery and clinical decision making for physicians and patients?

Robyn Busch, PhD:

Well, the goal of this research really is to help clinicians consolidate identified risk factors for cognitive and mood outcomes, which are often contradictory. And using these models, the goal is to have high predictive accuracy in order to assist clinicians in accurately counseling their patients regarding the potential risks and benefits of epilepsy surgery to aid clinical and surgical decision making, and ideally to permit early identification of those small subset of patients that may require closer postoperative monitoring. Ultimately, our hope is that these types of studies will pave the way toward precision medicine approaches in the treatment of epilepsy.

Glen Stevens, DO, PhD:

And can your work be leveraged for other disorders?

Robyn Busch, PhD:

Absolutely. These methods are really not novel. They've been used in a number of other disorders like cancer for many years. And so, the methods can easily be employed to predict relevant treatment outcomes for almost any disease treatment or outcome of interest.

Glen Stevens, DO, PhD:

And then Robyn lastly, there's always a concern with these models, is it going to inhibit patients from having epilepsy surgery? They're going to think oh, if I have a certain score and it's below 50% or something, I shouldn't do surgery. What's your advice to people out there that are concerned about these models that maybe inhibiting people from having surgery?

Robyn Busch, PhD:

Yeah, really the goal of them is just to provide more information prior to surgery, so that patients can be appropriately counseled. Certainly there's a risk for cognitive decline with continued seizures as well. And so, really the goal is to provide more information, so that the patients can consider the risks and the benefits of epilepsy surgery and have realistic expectations regarding their outcome. It also can be beneficial to clinicians because if we can identify those patients who may be at higher risk for some of these changes in cognition or mood following surgery, that they can follow them more closely during the postoperative period, so that if they do experience those changes, they can intervene early.

Glen Stevens, DO, PhD:

Well, Robyn, this has been a very insightful conversation. I'd like to thank you very much for your time today and look forward to trying your nomograms one day.

Robyn Busch, PhD:

Great. Thank you so much for having me.

Conclusion: This concludes this episode of Neuro Pathways. You can find additional podcast episodes on our website, clevelandclinic.org/neuropodcast, or subscribe to the podcast on iTunes, Google Play, Spotify, or wherever you get your podcasts. And don't forget, you can access real-time updates from experts in Cleveland Clinic's Neurological Institute on our Consult QD website. That's consultqd.clevelandclinic.org/neuro, or follow us on Twitter @CleClinicMD, all one word. And thank you for listening.