Epigraph Vol. 24 Issue 1, Winter 2022
Deep brain stimulation for epilepsy: Dr. Robert Fisher
Reported by Laurent Sheybani, PhD | Produced by Nancy Volkers, ILAE communications officer
Listen below or download the episode.
Find Sharp Waves episodes on Spotify, Apple Podcasts, Google Podcasts, or Amazon Music.
Neurostimulation, or neuromodulation, uses electrical pulses to inhibit seizure networks in the brain. The newest type, deep brain stimulation or DBS, was the focus of the SANTE trial. The trial included 110 people with epilepsy that had not responded to anti-seizure medications. Its results, published in 2010, showed promise for stimulating part of the brain called the anterior nucleus of thalamus.
In this episode, Dr. Laurent Sheybani talks with Dr. Robert Fisher about research in the field since the SANTE trial, as well as DBS’s possible mechanisms and its potential. Dr. Fisher is the director of the Epilepsy Center and EEG lab at Stanford University and was the lead principal investigator on the SANTE trial. He is a consultant for Medtronic, which manufactures DBS devices, though he receives no device-related compensation.
Laurent Sheybani, PhD: Maybe we can start with a general question – could you explain what DBS is?
Robert Fisher, MD, PhD: Deep brain stimulation or DBS is a method of modulating the excitability of brain in both the region of stimulation, where electrodes are implanted, and also in the connected network in the brain. The purpose is to inhibit seizures, at least in this case, although neuromodulation is used for other things, most notably movement disorders. We call it inhibition but it may not be so simple as just inhibiting neurons – it’s undoubtedly a matter of inhibiting activity in neuronal networks, disrupting synchrony, causing both short and long-term changes in systems.
Sheybani: So for refractory epilepsy, there is surgery, and there is also neuromodulation. Why would someone receive anterior thalamic stimulation rather than epilepsy surgery?
Fisher: Neuromodulation is a palliative procedure, not curative. Resective epilepsy surgery is unfortunately much of the time also palliative, but it has the possibility of eliminating seizures. So does neuromodulation, but in a relatively small percentage of cases. The numbers are in the range of 15% to 20% of people with either DBS or RNS, and we can talk about the differences later, would become seizure free for 6 months or more. But a greater percentage would become seizure free with epilepsy surgery.
Epilepsy surgery works better if you’re a surgical candidate, but you also pay a potentially bigger price. Once a piece of the brain has been cut out, suctioned out, or lasered out, it can’t be put back, so if you have memory or speech deficits or a surgical complication, recovery can be somewhat difficult. Whereas neuromodulation involves putting thin wires in or on the brain, and no explicit removal of tissue. It’s less destructive, less invasive, and it’s more adjustable. Radio paddle programming from the outside can turn it up, turn it down, change parameters, turn it off -it is removable and reversible..
Sheybani: Epilepsy surgery is not indicated in patients with generalized epilepsy, right?
Fisher: That’s another thing potentially in favor of DBS which is that it can address multifocal epilepsy, which surgery cannot, and for those who don’t know what resective means it means cut out, remove, destroy, a piece of the brain.
DBS is not officially indicated anywhere, including the USA, for generalized epilepsy, but it is used for that in some cases. There is published evidence for it. Not so much in the anterior nucleus of thalamus as the centromedian nucleus. That was pioneered by the Velascos in Mexico City. I did my first study of DBS back in 1990 with central median stimulation for generalized epilepsy. But it’s not yet approved for that. Focal epilepsy, with or without generalization, and treatment in the anterior nucleus , is the approved treatment.
Sheybani: We could think that even generalized epilepsy relies on this cortical thalamic network so thalamic network could be helpful in these cases. But maybe it depends on the region we stimulate in the thalamus?
Fisher: I think it does and you’re correct – the notion of “generalized-from-onset epilepsies” – what does that mean? It has to start somewhere. The debate still goes on, but my belief is that generalized-from-onset epilepsies very often start in a region of cortex but it’s not one region – you’re dealing with a hyperexcitable network, and prominent in that network are thalamo-cortical connections, so if something plays into the thalamus, it may set off a big broadcast to both regions of cortex so quickly that we can’t tell from EEG or behavior where it starts. That’s my notion of generalized epilepsy, rather than it starting in the thalamus or brain stem. Others would disagree with me.
It would make sense to do DBS of thalamus for generalized epilepsies but I can tell you, being a veteran of many clinical trial wars, that we tend to use the “Casablanca” approach – we round up the usual suspects. And the prevalent number of patients in epilepsy clinics that are amenable for a procedure have focal onset seizures, so that is the first group we chose to test. I look forward to definitive trials on generalized epilepsy.
Sheybani: So if we go back to the clinic, what should a patient expect who is a candidate for DBS, in terms of seizure control and side effects?
Fisher: Let’s go back to what is evidence based and approved in the US, Europe, many Asian countries, South America. Your ideal DBS candidate would be someone who has refractory and disabling seizures. Of course if the patient is in the two-thirds of people who respond to medication, then DBS or any neuromodulation or surgery would not seriously be considered.
Let’s assume they’re refractory, and this is something we’ve never been able to define precisely. Does it mean one seizure a month, four seizures a month? For some people one seizure a year may be disabling, if you’re a commercial airline pilot or something like that. But “refractory” has usually been defined in terms of seizures frequent and disabling enough to affect quality of life. But it’s a slippery issue.
It’s unusual for us to do DBS on someone who isn’t having a seizure or month per more, but it’s not a hard cutoff. Then you’d like to know they are not a good candidate for resective surgery and they would not prefer that as an option. So you’re probably talking about people who have multifocal epilepsy – could be two, could be bilateral temporal – or epilepsy documented, usually by video EEG, but from region unknown. This is an advantage of anterior thalamic DBS over surgery and over RNS – you don’t need to know precisely where the seizures are originating: the seizure onset zone, the seizure focus. Because playing into the thalamic network covers a lot of territory.
The stimulation doesn’t seem to have too much of a disruptive effect on function. It does have some. One of the things I regret not paying more attention to in the original trial of anterior nucleus, which is called SANTE – Stimulation of the anterior thalamus for Epilepsy –is sleep, because we’re stimulating structures that are involved in generating sleep potentials, and it can disrupt sleep. And yet very few people complained of being sleepy or having clinically disrupted sleep. In general I’ve been impressed that all modalities of neuromodulation are, with a few exceptions, remarkably well tolerated. Much better than cutting out a piece of the brain.
Sheybani: For DBS I think there has been a publication from someone and colleagues in Epilepsia in 2015 that showed an increased rate of nighttime arousal with DBS.
Fisher: That’s right. You certainly can measure changes in the sleep cycle, but they don’t complain much. But it’s something to think about. VNS [vagus nerve stimulation] also has effects on sleep and in rare cases it exacerbates sleep apnea in a way that might not even respond well to CPAP.
Sheybani: Could you indicate the rate of success of the therapy and the evolution of seizure control over time?
Fisher: What’s your definition of success?
Sheybani: That’s actually my question – what should a patient expect in terms of seizure control? Decrease of 50% of seizures per month, six months, what should they expect?
Fisher: That’s a question I can answer. People have different definitions of success, and that’s something to make clear to the patient from the beginning. If their expectation is that right after implanting the electrodes and turning on of the stimulator, they’ll be able to stop their medications and get a driver’s license, they need to be given a realistic description of what’s likely to happen.
The improvement with neuromodulation for seizures, as opposed to for tremor or Parkinson’s, which is immediate, the improvement is over months, really to become maximal in the 2- to 5-year range. So a certain amount of patience is required.
In clinical trials, both for RNS [responsive nerve stimulation] and DBS, by 3 months of stimulation the seizure reduction was in the 40% range. Which was significantly better than placebo in both cases. And that continued to improve to about the 70% maybe even 75% range, after a few years in open-label follow up.
That improvement has to take into account, always, when you’re talking about long term open label improvement – what happened to the people who didn’t do well and dropped out? Because if you’re only looking at the improvement in people who stayed in and for whom it worked, that’s not interesting. So that’s taken into account usually by the last observation carried forward method. You assume when they drop out because they’re doing poorly that they’re still in the study with the same bad seizure frequency that they had before they dropped out. That still leads to better than a two-thirds reduction in seizure count from baseline.
You also have reduction in the severity of seizures. Even in the blinded phase with DBS we saw that. Maybe because it’s blocking propagation through a network of bigger seizures that make you fall or have injuries. That is also a benefit. And then there may be other secondary benefits – you may change a focal impaired awareness seizure to a focal aware seizure, or a bilateral tonic-clonic seizures to a focal impaired awareness seizure, less severe, you may have quicker recover in the post-ictal state -- these things have all been seen.
But complete seizure control is elusive. Getting off medications entirely with neuromodulation – you can find a few shining cases but it’s pretty rare. Being able to reduce medications though is pretty common, and patients appreciate that. So those things are all success and degrees of success, in my definition.
Sheybani: You said the control of seizures is progressive over a long time. Do we know the mechanisms underlying this progressive improvement?
Fisher: That’s been a big question. It’s true for VNS, RNS that comes on when electrical seizure-like discharge occurs, and it’s true for DBS. It’s also true for some non-epilepsy conditions such as dystonia or other psychiatric issues. Sometimes pain control as well. We don’t really know what causes it yet, but when I give a talk on this, I show results from 5 or 6 animal studies that indicate synaptogenesis, gene induction with long-term changes, receptor changes, changes in BDNF, changes in autophagy (which is important for long-term neuronal function) and growth of neurons in, for example, the dentate gyrus of the hippocampus. So I think we’re changing genes, receptors, synaptic connections over long term, but there’s a lot more to be learned.
Sheybani: The changes that you are talking about – autophagy, changes in gene expression – are those at the site of stimulation, the epileptic focus, or elsewhere in the network?
Fisher: They’re certainly at the site of stimulation – the experiments were often rat experiments done in hippocampus. But you can also see changes throughout the network.
Sheybani: In the paper by Fasano and colleagues in 2021, we saw some variability in DBS stimulation parameters. Did you expect this variability and how do you explain it?
Fisher: I don’t object to the variability. It’s a matter of gaining more experience and learning better ways to do it. For young investigators who are planning to go into randomized clinical trials, I should tell them that there are a lot of guesses that have to be made in the beginning. And we made an enormous number of guesses. With drugs, there are not so many guesses. You have to guess the dose, mostly. You know the pharmacokinetics so you know how many times a day to give it. But with stimulation you have to guess where you’re going to stimulate, at what voltage or milliamperage, at what frequency. Sometimes stimulating at 2 Hz can have an opposite effect of stimulating at 100 Hz. You have to decide how long your pulse width is. You have to decide whether you’re giving continuous stimulation, as movement disorder people usually do, or intermittent, and if it’s intermittent, what interval on and what interval off? You have to decide if you’re giving it bipolar with adjacent electrodes, or what’s mistakenly called monopolar-- it should be called referential, with one electrode to a very distant point, which makes a bigger current spread. You have all of these choices and if you multiply all of the possibilities together there’s more than 1,000 things to choose from.
So we took what we knew from the literature, what we knew from VNS, which had been around for more than a decade, and we made our best guess. It was a good enough guess that we ended up with a significant difference between the placebo group and the control group, but there’s nothing that tells me it was the best guess. There were a few cases where I know the guess was wrong. For example we started with 5 volts, which would be about 5 milliamps with constant current. That was too much – some people felt too many paresthesias, and 2 of 100 patients had seizures that were actually triggered at that level. So now most of us start at 2 milliamps or 3 milliamps and work up to 5 over a few months. Experience and trying different things gives us some useful variance on how we do stimulation. Unfortunately you almost never can prove that it's better, because that would require another large, expensive, controlled trial.
Sheybani: All of these guesses and stimulation in protocols could be off putting for patients – how would you address these concerns?
Fisher: If they come with a question, I tell them that we start with what we know works. Since we’re talking about DBS and the SANTE trial, that’s 2 to 5 milliamp stimulation with the cathode at the top of the termination of the mammillothalamic tract, which experience has shown to be the best subregion. We do 1 minute on, 5 minutes off, 90 microseconds, 145 pulses per second. That’s what we did in the trial.
But if it’s not working well for them, then we may try some alterations in the amount of current, or the electrode we’re stimulating, or maybe longer times on versus off, what we call a bigger duty cycle. In my experience, they find it reassuring that if this doesn’t work for them, instead of having to take the whole system out, we can make painless adjustments that might make it better for them, without any other surgical procedures. I find it a plus rather than a minus that we have that flexibility.
Sheybani: So the variability means that if it doesn’t work there are other options.
Sheybani: You showed that anterior thalamic stimulation was effective in decreasing seizure frequency, and last year there were two publications in Brain showing that the stimulation of the medial septal nucleus was able to stop ongoing epileptic seizures. What is your clinical opinion on these publications?
Fisher: I don’t know of anyone who’s doing medial septal stimulation. Those two papers were not clinical trials in any sense – one was in rats and the other was mostly another review and a thought piece and a justification for why it should work. My friends at UC-Davis have been looking into this significantly, and I hope someone will make a trial for it. There’s no question that the anterior nucleus of thalamus, while being the only approved target, is not the only useful target and may not even be the best target. Different targets would be best for different syndromes and different individuals with epilepsy.
I would welcome different ways to map the seizure network so we would know – what are the control nodes? What might be the best places to do neuromodulation? Some cases might be septal, some might be centromedian, medial dorsal thalamus, or pulvinar, for posterior based seizures. One of the most influential thalamic gateway nuclei is the nucleus reticularis. Which would be hard to stimulate because it’s like a candy shell around the thalamus, but it’s very influential in relaying activity of nuclei, and in animal models at least it’s influential in absence seizures. Just direct stimulation of the seizure focus may be the most obvious way to go. If I had to imagine what would be the next approved indication for focal seizures, it would probably be direct hippocampal stimulation. There are several people making progress on that.
Sheybani: Going back to these two papers in Brain, they showed they could stop ongoing seizures that have already started. In your publication I think the aim was to prevent seizures. Do you think that depending on the region of the brain we target, or the parameters, could we prevent seizures or stop those that have already started?
Fisher: That’s a fascinating question. If you look at the VNS trials and goals and the DBS trials and goals, it was to prevent or reduce the overall number of seizures. If you look at the RNS trials, it was to shut a seizure off once it was detected. As we’re learning over time, it seems like all three of these modalities involve both mechanisms. Occasionally I think DBS can interrupt a seizure, although that was not the plan, and it doesn’t happen every time. But consider responsive neurostimulation, which I take no responsibility for in terms of the trials, but I do a lot of it with my patients. My colleague Martha Morrell led those studies. It may stimulate 2,000 times per day, and the patient is not having 2,000 seizures a day. But it’s detecting abnormal electricity according to the parameters you set, and in so doing we think it probably has a significant neuromodulatory effect or preventive effect, as well as the effect it can have when responding to an individual seizure. I’ve come to think of both sides of the coin being in play – prevention and treatment, for all the neuromodulatory therapies.
Sheybani: If we look at the original paper in 2010, the SANTE trial, we see both arms – the placebo group and the group with DBS, have decreased seizures at early time points. Is that a placebo effect, or is it something else? If it’s a placebo effect, what are your hypotheses on the mechanisms?
Fisher: It’s an interesting subject that we talk about a lot. There is about a 20% drop in median seizure frequency in the first month after surgery and before turning the stimulator on for DBS. In the case of RNS, they had a 2 -month period before turning it on and about the same 20% decrease. SO what’s going on there. Could be placebo effect. Could be a microlesion effect – just passing the electrode through the tissue takes out enough of the local network that seizures are hard to start. It could be an inflammatory effect of putting the electrodes in. I don’t think it’s an anesthesia effect because it goes on too long. Or it could be something much less interesting, which is regression to the mean.
What that’s about is that patients don’t go into an invasive surgical trial when they’re doing well. We might schedule them and then they’ll call me and say, “Hey doc, I haven’t had a seizure in two months, maybe we should wait a bit and not do this hole in my head thing.” So if patients go in only when they’re doing badly, the next chance observation will likely happen closer to their mean number of seizures. I actually think that’s the most compelling reason because we’ve seen this for lots of invasive procedures in the past. Pneumoencephalography, for example. In the old days you wouldn’t get a pneumoencephalogram if you were doing well.
Sheybani: So as you said before, when the DBS is turned on in Parkinson’s disease the symptoms alleviate spontaneously. This instantaneous efficacy could be interesting in epilepsy, in light of multi-day rhythms. But as we know, DBS is progressively effective in epilepsy. Is there a way to reconcile this?
Fisher: Epilepsy is very much a matter of rhythms – short-term, long-term rhythms. So you have these waves, different factors, and when they add together at the same time – menstrual cycle in women, missing medication, stress levels – the wave gets high enough to exceed a threshold for seizures. So if we have neuromodulation, we can make those waves more gradual, like a protected area of the sea. There still are these waves, and they still will account for a lot of random variation that we see with the numbers of seizures.
While you’re talking about biological rhythms, I should point out the big advantage and what we’re learning from the responsive neurostimulator. Aside from the invasive nature of RNS, it is by far the best seizure monitor that we have, because the long-interval abnormal electrical discharges turn out to be good surrogates for seizures. So you get counts, saying left hippocampus and right hippocampus, accurately, of seizure frequencies over time, for years. And analyzing these counts has shown that there are many rhythms in the brain that we did not expect. My friend at UCSF, Vikram Rao, has published several papers on this. It’s not just the menstrual cycle in women; men have cycles as well. Some people have cycles during the day, some have them over days and weeks, some have them over months. This is very interesting and seems like something we should be able to use to improve therapy, by adjusting amounts of stimulation according to the cycles, or adjusting medication doses according to the cycles.
Sheybani: It’s obviously quite an expensive procedure as well. How do you think it could be implemented in resource limited countries?
Fisher: It is expensive, but not as expensive as having uncontrolled epilepsy for several years, and it’s not as dangerous as having uncontrolled epilepsy for several years. All three neurostimulation modalities seem to reduce the SUDEP rate – sudden unexpected death in epilepsy. SUDEP is the main epilepsy-related killer of people with epilepsy. If DBS can reduce hospitalizations, emergency room visits, time off work, medication costs, then a number of pharmacoeconomic models show that it can pay for itself in the long term.
I’ve been going to epilepsy surgery conferences for more than 30 years. In past years it was common for us to say, “We can’t help this person” – medicines didn’t work, and they don’t have a resectable seizure focus – either we can’t find it, it’s too dangerous to take it out, or there are multiple foci.
Now, it’s very rare in our conferences where we say we don’t have anything to try or don’t think we can help someone. Usually if there isn’t a simple resectable answer, the debate comes, do we do RNS, DBS, or VNS? And with RNS and lately DBS as well, where do we put the stimulator? So we have more options to help people – maybe not cure them but help them in a significant way.
Sheybani: What are the next steps for DBS research? What is the potential of DBS?
I would like to have a noninvasive way of predicting who’s going to benefit. I’d also like a better way to know a person’s individual seizure network more precisely. Where is the focus, with what is it connected, what are the points of influence, so we would have better placement of our stimulation. I know that the anterior thalamus is not on the network of every focal seizure disorder and how it spreads. So perhaps some of the reason DBS doesn’t work in some people is that it’s not in the right place. So how do we know that? How do we know where someone’s individual seizure network is?
I think we also need to know more about the individual characteristics of seizures. There was a fascinating study in rodents a few years ago about how stimulating at the terminal frequency of a seizure -- maximal synchrony – was more effective than stimulating at any other frequency. So perhaps we shouldn’t be using the same 145 Hz, one-size-fits-all stimulation but we ought to look at the individual frequencies of seizures, the principal components, and tailor stimuli to that.
And then of course I’d like to know more about, as we talked about before, what are the best nuclei to stimulate, the best regions to stimulate for different syndromes. These are some of the things that I think might get long term efficacy from 70% to 90%. You’ll never get to 100%, seizure surgery is not 100% and medications are not 100%. And of course all of these are interim therapies until we find a way to really cure epilepsy.
Sheybani: To sum up, we need to better know the epileptic network in individuals, we need to be able to predict who will respond to the stimulation, and we need some kind of personalized stimulation protocols for each patient with epilepsy.
Fisher: Well said.
Sheybani: Well thank you very much for your time, it was great speaking with you. Do you have anything you’d like to add?
Fisher: Yeah, get into this field! It’s where it’s happening. There’s too much that’s empirical and unknown. Ideally you start with animal models and an understanding mechanisms, and then move logically to patient care. Historically, the field of neurostimulation has gone in the other direction – we started with uncontrolled observations in people from the 1950s and we said, “Hm, this seems interesting, why don’t we go into the laboratory and figure out how it works.” So please, help us figure out how this works.
Subscribe to the ILAE Newsletter
To subscribe, please click on the button below.
Please send me information about ILAE activities and other
information of interest to the epilepsy community