Transcript: Marla Dubinsky, MD, on Proactive Therapeutic Monitoring
Rebecca Mashaw: Hello, everyone, and welcome to Podcasts360, your go‑to resource for medical news and clinical updates. I'm your moderator, Rebecca Mashaw.
I'm here today with Dr Marla Dubinsky, division chief of pediatric gastroenterology and nutrition and codirector of the Susan and Leonard Feinstein IBD Clinical Center at the Icahn School of Medicine at Mount Sinai in New York.
Dr Dubinsky and colleagues recently published a review in the journal Clinical Gastroenterology and Hepatology, entitled, "How, When, and for Whom Should We Perform Therapeutic Drug Monitoring?" which she's going to discuss with us today, particularly in regard to the treatment of inflammatory bowel disease.
Thanks for being with us today Dr Dubinsky. In your article, you mentioned that reactive therapeutic drug monitoring or TDM, particularly for anti‑TNF antibodies, is widely accepted but that the data are more scarce for other classes of monoclonal antibodies. Why is this the case?
Marla Dubinsky: A good question is, why do we have so much more data on understanding the pharmacology and pharmacokinetics of anti-TNFs? First, with some of our newer biologics, it's really due to time. We have been using, for example, infliximab over 20 years now and adalimumab heading into 13 years. So we just have more data, we understand how these therapies work, we now have luxurious time.
People have been investigating the role of drug concentration dating back to 2006, or maybe even before that. I think that we just know more because we've had more experience with them. The idea of understanding the pharmacokinetics of the newer biologics, such as the IL-12/23 target, or the integrins with vedolizumab, there's a couple of important differences.
One of them is, we don't really have a lot of dose escalation or dose adjustment opportunity. We can change the frequency of vedolizumab to every 4 weeks, for example. We can, in theory, change the frequency of injections for ustekinumab more frequently. We just don't have as much flexibility in dosing like we do with infliximab, because we still don’t know that we can play around a lot with the dosing as well as the frequency.
When fixed dosing, which is what we're seeing with our newer biologics but also applies to adalimumab, where you're only able to go from one pen every other week, for example, to one pen every week. It's not as flexible a dosing. For us, choosing infliximab, working with experts in the field of pharmacokinetics, and using dosing calculators to help us improve the precision of our dosing, infliximab is a great place to start.
Rebecca Mashaw: What challenges do practitioners face in implementing proactive TDM to help prevent loss of response in patients with IBD? How difficult is it to maintain the tight control that is the goal of IBD care?
Marla Dubinsky: If you go back to the history of therapeutic drug monitoring, it really started in a reactive approach. Meaning, we understood that when a patient wasn't feeling well and we wanted to see if the reason for why they weren't doing well on infliximab was driven by, perhaps, just they didn't have enough drugs in the tank, so to say, that led us to say when a patient is not well, let's reactively check a drug concentration, confirm that the patient (1) has enough drug and should be responding. Therefore, we would classify them as failing, and maybe switch to something else.
Or (2) do they have antibodies already developed against the drug and that's why they've lost response? We call that secondary loss of response, for example. Or (3) probably the most common is the fact that the patient just doesn't have enough drugs.
We've learned quickly that dose adjustment, based on clinical intuition alone, when a patient just has symptoms may not always be the right option, because people may be active because they've already developed antibodies to infliximab because they've had a prolonged course of underdosing or undertreatment or low drug concentration.
The body finally just said, "I needed more of the drug or I needed it more frequently, and this every 8‑week regiment of the starting dose of 5 mg/kg doesn't work for me and my body."
We were really missing an opportunity to personalize that approach. We were waiting for people to reactively fail. Well, that doesn't work in IBD. Our whole point is to keep people always feeling well and being ahead of the flare.
The idea that we started getting data to say that if you get a drug concentration of X and you change the dose or the frequency in order to get a drug concentration of Y, and we know that when a patient has that higher drug concentration they do better, wouldn't it be amazing if you get a low drug concentration while the patient's not expressing any symptoms? To actively proactively optimize the dose based on the target concentration, so that you avoid a flare.
That's that proactive mindset that there's no doubt where we've moved into, and we should be all residing in a place where we definitely should be proactively finding things out in advance of a patient getting sick. I mean, that's what they come to us for. They want to make sure that we have all the pieces of information we can to keep them well durably.
We know that if you could proactively obtain a stool marker, or like a triple tracer teptin, or a CRP for disease activity, and a drug concentration in advance of any patient developing symptoms, and proactively doing something about it, you're going to have better outcomes.
The premise for our study was not only checking drug levels early—which I should point out using this pharmacokinetic dashboard and a dose calculator—it wasn't just checking patients early. It was checking patients really early, meaning during the induction phase. This study has a number of key components of our hypotheses, or our key components of clinical practice, that we needed to address.
One is, should we be checking drug levels earlier than by the time they're in maintenance? Even proactively, meaning you can proactively check the drug levels of someone who's been on the drug for 5 years, sure. Maybe make a dose change, but imagine being able to actually make these dose changes when a patient's just starting on the drug, when their disease is most active, and they need the most drug. I mean, that's sort of addressing proactive, but we're going to add the term early proactive.
We added that layer of early, meaning an induction, starting at infusion II. Then we said, "It's not good enough just to get a drug concentration," because drug concentrations alone don't necessarily tell you the whole story. What these dashboards allow you to do is integrate not only the drug concentrations of that individual patient, but all the factors that influence the clearings of these drugs, meaning the influence of drug concentration.
If you could precisely dial in a dose and frequency specific to that individual patient's clearance or the way that their body manages these therapies specifically anti-TNF , you're not only proactively optimizing, but you're proactively optimizing based on the needs of the individual patient, which has not been done before.
I think the study allows us to get into multiple layers of where we need to change our paradigms when it comes to biologics, particularly infliximab.
Rebecca Mashaw: Can you give us an overview of the different types of assays to monitor drug concentrations and anti‑drug antibodies that you reviewed in your research?
Marla Dubinsky: We only used one type of assay, because at Mount Sinai, we use the Prometheus Mobility shift assay for both our standard of care and in the research setting. We were dealing with the assay that allows you to accurately detect anti‑drug antibodies even when drug is present.
The bigger issues of some of these non‑, shall I say, drug tolerant assays is that you can't really tell if it's drug or antibodies that the assay’s picking up. If there's measurable drug, they don't even comment on the level of anti‑drug antibody titer level.
What this allowed us to do was that, even though there was plenty of drug around, we did have some patients who started to have low levels of anti‑drug antibodies that actually goes into the model. The model, based on the drug concentration alone, may not have access to escalate drugs, but because there was antibody even low level, it allowed us to overcome those anti‑drug antibodies.
That's the unique feature as well is that, you know, it's usually one or the other. Does the patient have antibodies, or does the patient have enough drugs, or neither, or both are missing, or both are there? “Both there” are a small group of our patients, but I think in the induction phase that is an important subgroup because normally, if we would just have drug level information, we may have left the patient without any adjustment.
But having a knowledge that you could input the presence or absence of an anti‑drug antibody early on in the dashboard allows us to make some dosing changes that we otherwise would have not done.
I think the secret sauce to this is that, if you take a table or a list of anything that impacts drug concentrations and anti‑drug antibody, they are accounted for in the model. If we hadn't used in this particular, I can only speak from my research and my study here, is had we not used the mobility shift assay we would have missed the presence of even low level of the anti‑drug antibody, which did play a role in driving a dosing.
Rebecca Mashaw: You also looked at algorithmic approaches to TDM. What did you discover in that area of study?
Marla Dubinsky: In overall, just to highlight the design of the study, any patient who entered standard of care infusions for infliximab to our infusion unit were eligible for enrollment. This was part of standard of care. We offer this precision-dosing dashboard that will help direct the physician and you in knowing when your next dose to be. That is what it would be if we integrate into clinical practice tomorrow.
What is was, was that the starting dose was decided by the physician himself. The study team was not deciding what dose a patient should start on, that is entirely up to the treating physician, which is how it would be in clinical practice as well. We try to make this as real world as possible.
What happens is the patient gets the first dose. Then at the time of second infusion, which is usually 2 weeks after the first infusion, which is the patients we've studied. They have to have a Week 0 and a Week 2 to be eligible. If they receive their second infusion any time before Week 2 that sort of bias the model, patients who were not included. There were plenty of patients who got the second dose earlier, but they were excluded from this particular analysis.
At the second infusion before the patient gets their drugs, they actually get a drug concentration and anti‑drug antibody collected, along with standard of care labs such as the CRP and albumin and a weight and a disease activity index. That all went into the machine, sort of the model, which is the software. That software told us when the patient should come back for their third infusion. Not surprisingly, that a large proportion of patients who started on the 5 mg/kg dosing, which is the standard of care approved dose, that many of them needed to come in earlier than the standard 4-week interval between the second and third infusion, and had to actually come in, instead of 28 days, they had to come in at 17 days.
Essentially, 2 weeks after the second infusion, which is almost half of the length of time that is in the label. That standard of care dictates that you come in 4 weeks after your second. The idea that we had a whole bunch of patients that needed to come in for the third infusion earlier, so they were forecasted. We picked up the phone, we called the patient, we organized with the infusion unit, that the patient needs their third infusion earlier than what the standard of care on label dosing shows.
They came in for their third. Again at the third infusion they had a drug concentration, anti‑drug antibody, CRP, albumin, weight, disease activity, dosing, goes into the model and a software we calculate, based on the drug concentration that they achieved at their third infusion, when they should come in for their fourth.
Now, what I'll tell you is that our target concentrations, which were used predict when patients should come in, is based on lots of data in the literature. It has been shown that we should try for a goal of a level of at least 17 at the third infusion. If the idea that most patients aren't going to get a level of 17 at the third infusion, that's what drove the model to say come in earlier.
Now similarly, for the fourth infusion it's usually an 8-week period between the third and the fourth infusion. What we targeted is, we wanted a drug concentration of 10. The beauty of the software and the model is you could toggle your way and choose any trough that you want to. Based on what your target trough concentration, meaning predose concentration you want for that patient, the model will thereafter tell you, if you want a concentration of 10, based on literature that is a very valid literature on why a level of 10, for example, is that it'll tell you when the patient will hit a level of 10. It shows you when the patient will go below a level 10. You want to catch the patient when their drug concentration, before their fourth infusion, would be 10.
The fact, and it shouldn't shock anyone who are listening to our discussion, that 80% of patients who started on 5 mg/kg, meaning the on‑label dose of infliximab, 80%, 8 out of 10 patients are forecasted to need a regimen that is different than what the label says.
Moreover, even patients who we doubled the dose on the first infusion, like we started at a 10 mg/kg, 6 out of 10 needed a different frequency by the time their fourth infusion hit. The fact that we underdosed 70 or so percent of our patients in induction, when the patients most need the right drug dose, because that is when their burden of inflammation is higher, it's backwards.
This idea of this step‑up vs top‑down, and a lot of people use these terms when it came to which class you choose, I continue to recoin the term to say top‑down is where you start with the right dose for the patient's individual burden of disease in their own personal characteristics and then you de‑escalate. You can then step‑down to dosing that is lower and less often, because the patient is now in remission. What we were reactively doing is dose escalating once the patient was flaring.
The idea of proactive optimization in doing this higher dosing when the patient most needs it will prevent loss of response, will prevent the development of anti‑drug antibodies, and will increase the durability of these drugs well beyond what any clinical trial has ever shown, because you're precisely dosing anti-TNFs, specifically infliximab, for the individual patient's own needs.
Rebecca Mashaw: What additional research needs to be done to further explore this topic?
Marla Dubinsky: Well, you know I'm biased. I don't think you need much else! We have 180 patients along with previous studies that have shown that my clinic. I thought I was pretty good at managing patients historically and how it all started is I sent my clinical decision‑making to Diane Mold, who is actually the inventor of this software program called IDOS, is actually the name, and I said to her, "Tell me, in these patients that got anti‑drug antibodies, if I had used your software would your software have told me the dose-adjust these patients before they got the anti‑drug antibodies?"
The answer is, "Of course, it would have," because it knew that the patients drug concentrations were already too low. So, I waited for the patient to flare and look what happened, the patient flared, but they already have anti‑drug antibodies had I just used the software. Again, I'm pretty good at understanding how to dose infliximab. This totally made my clinical acuity look subpar. After that, it was like, for me, I almost feel guilty not using the software.
If I have a really sick patient, I sort of in my head have become my own software. I'm like, "I know if they have this lab, this concentration, they should come in at this timeframe.” I do my calculator in my brain now when people ask me, "Hey, I saw you present that data. What should be the regimen for this patient? They have this CRP, this albumin.” I’ve become my own calculator in my brain, because when you've seen, I don't know, 10 predictions for 180 patients, you start to see thousands plus of predictions, you start to understand the needs of patients.
If someone would have to be a hard scientist, I believe in the importance of confirming and validating any finding than a one‑arm open‑labeled type fashion to feel the cause and on the idea of getting rid of reactive would be to randomize patients to what we're doing reactive standard of care, dosing, and TDM vs a proactive optimization early adoption with the software, so go against what we just found in 180 patients.
I will tell you that those of us who have been proactive and who know this data inside and out, it's almost like, although I respect the idea of needing to prove it, once you’ve lived it and feel it and understand it, you almost feel like why do I need to do that again? In that sense, I understand the importance of the rigor and I'm very conscious of the need to always validate findings.
One thing that's interesting is the group in Amsterdam looked at this using this model in maintenance so didn't optimize it in induction. That's why we focused on the induction piece here. They did it in maintenance, and they showed very similar findings that the ability to stay on drugs is far better when you use personalized dosing as opposed to clinical acuity. In addition, I think a very important point, which can't get lost here, is we've been struggling with infliximab and adalimumab for a long time, especially infliximab. Do we need to use a second drug to make infliximab work better? Meaning, do we need to add a thiopurine or methotrexate, which is an immunomodulator, which increases the risk of infliximab substantially by adding another drug, just to make the infliximab drug concentrations stay higher or to lower the likelihood of anti‑drug antibodies.
My answer to that is, we can get rid of these second drugs. We don't need them if we would just dose them right from the beginning. For those high‑risk groups such as those that are under, let's say, the age of 21, 25, groups of patients where thiopurines proves to be increased risk in young patients who've had a splenic T‑cell lymphoma by using combination or even just non‑hepatic T‑cell lymphoma, we know is higher, particularly in males, for example, who are on thiopurines. Also, our advancing age population above the age of 60, 65.
If we could take high‑risk patients away from needing a second drug to make the first drug work better, we would not only improve the outcomes of the biologic and keep people on these drugs longer, we would also improve the safety profile of these therapies.
Probably some of the biggest debate I have, as I said, I know there's a lot of patients between the age of 25 and 60. If I was one of them, which I do fall into that age category, I would want the approach where I don't need a second drug and just optimize the drug I'm on. I think that is going to be really the most important, to me, piece of information.
Speak with listeners about, and for them to hear that, the research is not just about the at‑risk population. To me, it's about giving everybody the right dose at the right time and making sure they're on the right drug.
Rebecca Mashaw: Thank you very much, Dr Dubinsky, for a very interesting discussion on therapeutic drug monitoring.
Marla Dubinsky: Thank you for having me. It's always a great opportunity to bring what's happening in the field of therapeutic drug monitoring to all your listeners. I think this is a way of the future. It makes me happy, and I'm appreciative of you asking me to comment.