Read the Full Video Transcript

Ashish Kamat: Hello and welcome to UroToday's Bladder Cancer Center of Excellence. I'm Ashish Kamat, Professor of Urologic Oncology and Cancer Research, at MD Anderson Cancer Center. And it's a distinct pleasure to invite a friend, a colleague, Professor Pavlos Msaouel, who is an Assistant Professor, but we call him Professor, at the GU Medical Oncology Department here at MD Anderson, itself.

The topic that Dr. Msaouel is going to present to us today is not specifically related to bladder cancer, although a lot of the work that he does is in bladder cancer. But it is so relevant to all of our audience, that we really would like to hear him talk to us about his recent publication and the work that his group has done in the evaluation of technology-enabled monitoring of patient-reported outcomes for toxic effects linked to immune checkpoint inhibitors. This is clearly a topic that is not just of interest to our medical oncology colleagues, but also to our urology colleagues that have now had checkpoint inhibitors coming into earlier stages of the disease for non-muscle invasive bladder cancer, for example, and are now looking to incorporate this into their own practices. So it's critical for everybody to know how to detect and treat these toxic effects. So with that, Pavlos, let me hand the stage over to you.

Pavlos Msaouel: Thank you so much, Dr. Kamat, and UroToday, for the opportunity to present our electronically enabled strategy to monitor for immune toxicity.

These are my disclosures. None are pertinent to this presentation.

The goal of our platform is to reduce immune-related toxicities without compromising antitumor efficacy. And to achieve this, we had to accommodate the following considerations: monitor for a diverse set of remotely identified immune-related toxicities, develop dynamic electronic symptom alerting rules that efficiently account for sensitivity and specificity, and link our laboratory findings with our digital monitoring and clinical observations to enable specific interventions appropriate for each toxicity. Addressing these considerations will produce accurate and reliable recommendations that can be successfully transferred from centers of excellence to the broad oncology community.

Now to accomplish these goals, we needed first to account for several considerations. Cultural changes are a painful, but essential first steps as our clinic teams adapt to the new digital environment. Data security is key to protecting patient information and generating trust in our approach. And patients need to accept, and actually use, our platform. And at the same time, the platform needs to be flexible for use in living environments, and facilitate reliable and effective communication between patients and care teams.

We accordingly developed the illustrated trial design, focusing on patients treated with immune checkpoint inhibitors, alone or in combination. And this is an electronically enabled trial to screen immune-related adverse events, patient-reported outcomes, and biomarkers, to immediately alert the care team when meaningful symptoms are reported. Patients are monitored throughout the full course of therapy with immune checkpoint inhibitors. And we estimate that with 1,000 patients, we will be able to accommodate that heterogeneity in immune-related toxicities in patients with genital urinary malignancies across the Houston sites at MD Anderson. And while we focused on genital urinary cancers because this is our expertise, this paradigm is broadly applicable, and the collected blood, broad longitudinal observations, and data points will allow us to anticipate progression, respond to queries, and develop candidate predictive biomarkers.

Critical to the success of this project is developing a partnership with patients through effective communication. And to achieve this, we incorporated the expertise of Project Ronin to develop a smartphone-based patient interface, that can be easily embraced and be seen here. And when the patient-reported symptoms exceed a pre-specified alert threshold, their remote monitoring app informs patients to call their clinical care team, and simultaneously sends automated alerts to the clinical care team.

The interface with care teams needs to prioritize accurate data elements that assign risks for toxicity and inform specific interventions. That interface, illustrated here for a mock patient example, was developed with the help of our Prometheus team, with the goal, to be familiar with the care teams, and prioritize the critical data. Our platform and service data security via high encryption standards and all collected data resides at the Prometheus database at MD Anderson, for use by our clinicians and researchers.

The recent JAMA Network Open paper reported the results of our pre-planned analysis in the first 47 patients to be enrolled in our trial. The primary endpoint was feasibility, as determined by patient and care team adherence, and by the lack of increase in care staffing needed to monitor our patients. The median age of our patients was 65 years old. As expected, most of our patients had renal cell and urothelial carcinoma and were treated with regimens that included a PD1 inhibitor.

Now at the time of data lock, the patients had been followed for a median of 63 days. And patient adherence was generally consistent throughout the study period, with a median study adherence rate of 74%. The median response time from the care team to automatic alerts was 19 hours, and 73% of these alerts were reviewed within three days. And the implementation of our platform did not require an increase in care team staffing to help monitor or address these alerts.

As shown here, our patient adherence rates were stratified by gender, and by age group. Most of our patients were older than 60 years old, but still demonstrated acceptable compliance rates, that were generally higher than 60%. As shown on the right column of this table, the most frequently reported symptoms were fatigue, pain, and diarrhea. And on the other hand, as shown on the left column, the symptoms that most frequently generated alerts were arthralgia and myalgia, followed by pain, fatigue, and shortness of breath.

And a key feature of our approach is the ability to adapt the symptom alert thresholds by dynamically calculating their positive predictive value and their negative predictive value, and events defined as meriting alerts had to be at least grade two by CTCAE and associated with clinical intervention. And clinical interventions included pause, dose reduction, or termination of immunotherapy, or new interactions with healthcare providers due to toxicities that develop after treatment initiation, including emergency department visits or hospitalizations.

Using this strict definition of events meriting alerts, we prioritized increasing the negative predictive value of our alert thresholds, but without decreasing sensitivity. And for clinicians, this means that they will not receive excess alerts. And our patients are comforted by the knowledge that there is longitudinal oversight preventing excess delays in timely interventions. And the symptoms with the highest positive predictive value for adverse events requiring intervention were dizziness, nausea and vomiting, and shortness of breath. On the other hand, the symptoms most likely to result in unnecessary alerts were arthralgia and myalgia, fatigue, and cough.

Illustrated here, is the picture of a single patient's course monitored longitudinally. And for this patient, the event leading to intervention was grade two adrenal insufficiencies, which led to an ER visit. Notice how the alert, triggered by reporting of arthralgia and myalgia, was not related to any adverse event meriting clinical intervention. Conversely, alerts triggered by shortness of breath have a high positive predictive value for adverse events requiring intervention. And the platform provides the necessary infrastructure, which allows the collection of the many data elements, shown here, without disrupting patient care.

And shown here are the total number of alerts per patient collected over time for arthralgia and myalgia, as well as fatigue. And these symptoms are required alert threshold updatings, due to the high frequency of unnecessary alerts, relative to the missed alerts.

We are currently analyzing the original primary endpoint of improvement in immune toxicity outcomes in the first 100 patients enrolled in our trial. As mentioned before, we have now expanded to 1,000 patients, this protocol, and we are accruing now towards the final sample size, with the goal to assess for heterogeneity across our centers in Houston, at MD Anderson.

Our plan is for our platform to evolve in the following way: Integrate our IO toxicity data into the physician dashboard, that is allowing the comprehensive characterization of patient care, efficiently integrate into Epic care algorithms with order sets, as well as to adapt them to disease site-specific requirements, and to integrate with cross-discipline immunotherapy experience, by monitoring for interaction specific to the therapeutic environment, and by using our platform to look into the impact of immunotherapy interactions that may not be anticipated for ongoing and upcoming cellular therapy trials.

Thank you very much for your attention. This is a team effort supported by the many people acknowledged here, and I will be happy to answer any questions you might have.

Ashish Kamat: Thank you so much, Pavlos, for that very concise and precise talk, as is getting to be a trademark of how you present your data. That is very exciting information, obviously, that we have from your pilot study, and looking forward to the expansion.

You addressed a lot of the key points that people often wonder about when you are looking at app-based technology for our patients who are older, in the sense that you stratified it based on the age of patients. Is there also built-in, and maybe I missed this, a way to make sure that it is the patient, indeed, that's entering the symptoms, and not their spouse, or son, or daughter?

Pavlos Msaouel: That's a fantastic question. And while we emphasize the need for the patient, themselves, to put in those patient-reported outcomes, there is no way that we can know that it was indeed them, other than the fact that this all goes through the particular patient's specific phone. So we make sure, during onboarding, that we teach them, and we show them how to do this. So we double-check that they can indeed do this. It goes to their personal phone, but other than that, indeed, somebody could report for them. Now, this may not be a bad thing per se, because if, for example, somebody feels too fatigued, they might ask the person that cares for them, "Hey, can you input for me, that I feel severe fatigue?"

Ashish Kamat: Absolutely. No. And in fact, it might be a good thing, or it might not be a good thing. I guess, looking at the false alerts, maybe moving forward, if we could see if those are associated with somebody else inputting the data, that might actually be a pro or a con, it's just a thought. The next question I wanted to ask you is, clearly from your intent and everything that you presented, this looks like it's a supplement, and not a replacement, for normal monitoring. Is that correct?

Pavlos Msaouel: You're absolutely correct. It supplements normal monitoring, but in a way that does not add additional burden to clinicians. And that was a very key thing for us. That is why we needed first, to start this in the first 47 and 50 patients. And we were very careful initially, to make sure that we've got everything figured out, so that as not to add unnecessary alerts, et cetera. It is designed to make things easier for the patient and the clinicians, but it cannot, of course, replace real-life interactions with patients. But it did end up being particularly useful during the COVID-19 pandemic. And indeed, we showed an increase in enrollment as soon as the COVID-19 pandemic started, exactly because telehealth became particularly relevant at that point.

Ashish Kamat: Yeah. The timing was perfect. A follow-up question, and then this is not meant to be leading in any way, but I'll ask it anyway. There is a lot of interest amongst nonmedical oncologists, in the US, to administer IO therapy, partly because, of course, it is seen as something that doesn't have much toxicity. This is true, if you compare it to chemotherapy, but not true as an absolute. Right? And there's a mistaken belief that it is much easier to give. Which again, is true, relative to chemo, but not on an absolute.

On the other hand, there are a lot of patients that are seen, almost exclusively, by a nonmedical oncology colleague, for example, for non-muscle invasive bladder cancer, and almost would never make it to the medical oncology office unless the referral was put in. But we do hear from patients saying, "Well, I drive three hours anyway, to see my surgeon, do I need to go see a medical oncologist just to get a drug? Can you not give it to us?" Now, do you think that a platform such as this, eventually will help practices in smaller communities, places where patients do not have access to multidisciplinary care, to actually be able to get the IO therapy in a centralized or localized fashion?

Pavlos Msaouel: Ashish, you hit the nail on the head with this question. Absolutely. This is a major goal of this platform. And we're starting this with the idea that we can use this to produce accurate and reliable recommendations and data, that we can then transfer from, let's say, tertiary academic centers of excellence to the broad oncology community. That is exactly it. Because for example, the knowledge that shortness of breath is a very high yield, has high positive predictive value, as a symptom, is now something that we can communicate to the broader community of clinicians taking care of patients who are receiving a new checkpoint therapy.

The same goes with the symptoms that we're learning now may not have as high of a positive predictive value. So, that's exactly it. Our hope is, that by using this integrated platform, not just the app, but everything that comes with this platform, we are going to be able to detect earlier, those potential immune-related adverse events from afar, and mitigate them, so that they do not require ER visits and tertiary type of care.

Ashish Kamat: Dr. Msaouel, thank you so much for taking the time, and spending it with us this morning. In closing, are there any thoughts, or points that you want to highlight for our audience to take away?

Pavlos Msaouel: Yes. Thank you so much for having us. I would like to highlight that the goal exactly, of this app, is, at the same time, to both inform the clinicians, and our patients, about the potential for immune-related adverse events, and a big component of this is the dynamic symptom alert thresholds. It is that change in the dynamic symptom alert thresholds that make it more feasible for clinicians to take care of patients, without being overwhelmed by constantly receiving false-positive alerts. I think that is the key point that we are continuing to build upon.

Ashish Kamat: Thank you once again. Obviously, we will see each other, because we work in the same institution, but otherwise, I'm looking forward to when we can actually start meeting other experts in person and sharing data such as this in in-person conferences. Till then Pavlos, stay safe and stay well.

Pavlos Msaouel: Thank you, Ashish. Have a great day.