Read the Full Video Transcript

Ashish Kamat: Hello, and welcome to UroToday's Bladder Cancer Center of Excellence. I'm Ashish Kamat, Professor of Urology at MD Anderson Cancer Center in Houston, Texas. And it is a distinct pleasure to have today, Dr. Steven Boorjian, who is a Professor of Urologic Oncology at the Mayo Clinic in Rochester, and Kevin Wymer, who is a fifth year resident at the Mayo clinic and who is going to stay in the Mayo system once he graduates.

Kevin and then Dr. Boorjian had a very interesting presentation on the cost-effectiveness analysis of pembrolizumab for BCG-unresponsive carcinoma in situ of the bladder, and this not only gained a lot of press at the meeting, which was virtual, of course but also on social media and various types of journal clubs that are ongoing in the social media world. So we thought it would be great to have both of you here today to go through the presentation and then have a little round table discussion. So Kevin, if you're ready, the stage is yours.

Kevin Wymer: Absolutely. Well, thank you so much, Dr. Kamat and UroToday.  We are excited to be able to talk about our work, as you mentioned, Evaluating the Cost-Effectiveness of Pembrolizumab and the BCG-Unresponsive CIS Setting. So as you are well aware, BCG-unresponsive carcinoma in situ of the bladder represents a particularly high risk and difficulty to manage the disease. Treatment options are relatively limited and cystectomy has remained the gold standard.

In this area, alternatives have traditionally included various salvage intravesical chemotherapy regimens with response rates that are much lower than primary BCG. In 2020, based on preliminary findings from the prospective single-arm KEYNOTE-057 trial, the FDA approved systemic pembrolizumab as an additional treatment option in this disease space. Given the previously limited options for these patients, these findings were quite significant with a three-month complete response rate of 41%. And of those patients, about three-quarters had sustained response at one year.

Despite this promising data, however, there were multiple concerns associated with pembrolizumab, and one of those is cost. So this is approximately $9,000 per treatment, and this comes out to over $100,000 per year. In addition, as a systemic therapy agent, the KEYNOTE trial reported a relatively high toxicity rate of pembrolizumab, with approximately 63% of patients suffering from adverse events. It is therefore unclear if the oncologic benefits of pembrolizumab outweigh the toxicity and costs.

In order to help address these questions, we conducted a cost-effectiveness analysis, comparing pembrolizumab, radical cystectomy, and salvage intravesical chemotherapy for the treatment of BCG-unresponsive CIS. We selected combination gemcitabine-docetaxel intravesical therapy as our prototypical salvage intravesical chemo regimen for the model because this is the only one with data published specifically for BCG-unresponsive carcinoma in situ.

In order to perform the analysis, we created a Markov model comparing these three treatment modalities. We had two separate models for two index patients, and you can see here, index patient one represented a patient who was willing and able to undergo radical cystectomy. Index patient two represented a patient who was unwilling or unable to undergo radical cystectomy. In general, when looking at Markov models, these consist of multiple health states, and each health state is associated with specific costs and quality of life metrics. For each treatment modality that we included in our model, the specific health states included surveillance, recurrence, toxicity, progression, metastatic disease, and mortality.

The likelihood of a patient being in each health state was based on published data from the literature.  Costs were based on Medicare reimbursement and quality of life data were derived from utility values reported in the literature. You can see some of the specific utility values associated with the different health states in our model here on the table on the right. As far as the model structure, we won't get too into specifics, but the cycle length was set at three months and the total time horizon or total time for our base analysis of the model was set to five years.

These represent a simplified diagram of the model structure for both index patients. The model begins at the top with the initial treatment choice as the first branch. And then you can see subsequent branches represent various outcomes, as mentioned previously, this included recurrence, progression, toxicity, and death, again, each with specific costs and quality of life metrics. As with any cost-effectiveness analysis, the primary outcomes included cost and effectiveness, where effectiveness was measured in quality-adjusted life years. We also calculated incremental cost-effectiveness ratios or ICERs, and these essentially just represent the additional cost per quality-adjusted life-year gained for a given intervention.

An intervention was considered cost-effective if it was the least expensive, and if another regimen that was more costly, had an incremental cost-effectiveness ratio that was below a willingness to pay threshold at $100,000 per quality-adjusted life. So essentially, for a treatment to be cost-effective, if the added cost per quality-adjusted life-year and that cost was less than $100,000, it met the requirements for cost-effectiveness. To test the robustness of the model, we performed multiple large-scale sensitivity analyses. This included one-way, two-way, and multi-variable analyses, as well as bearing the willingness to pay threshold, and the overall time horizon of the model.

Getting into some of the results of the model, this table kind of represents our base-case analysis or the baseline results for both index patients. For index patient one on the top, there are a few main takeaways I want to point out. So the first on the left side, there is the cost. Specifically, pembrolizumab had a five-year cost of approximately $190,000. And this, as you can see is more than fourfold higher than both gemcitabine-docetaxel combination therapy and radical cystectomy and cystectomy was the least costly. The second column is the effectiveness. Again, this is measured in quality-adjusted life years, and you can see that pembrolizumab did result in higher quality-adjusted life years, than both other treatment options.

However, this was a relatively modest increase, and putting these together with the incremental cost-effectiveness ratio or ICER, which as we discussed, represents the cost per additional quality-adjusted life-year, you can see that for pembrolizumab, given the cost and increase effectiveness for one additional quality-adjusted life year, it cost greater than $1.4 million. And this is obviously much higher than our baseline willingness to pay threshold of $100,000, and so pembrolizumab was not cost-effective. Similarly, although gemcitabine-docetaxel was closer to that willingness to pay threshold, it was over that threshold and therefore it was not cost-effective, either leaving cystectomy as the most cost-effective option.

The results were similar for index patient two. Obviously, the main difference here is radical cystectomy is not included, but again, you have pembrolizumab with a very high ICER of over a million dollars and therefore not cost-effective relative to the gemcitabine-docetaxel. I also want to point out briefly here that for a lot of these data and particularly for pembrolizumab, some of the long-term follow-up data was limited.

And so certainly for five-year outcomes, it required some assumptions and extrapolations for our model. For most of these extrapolations, whenever possible, we tried to skew these in favor of pembrolizumab. So for example, there is a 0% metastasis rate reported in KEYNOTE-057, and we assume that it remained at 0% all the way through five years as was the case for bladder cancer mortality, 0% acute mortality rate, and no chronic toxicity. And even with these assumptions, we did not find pembrolizumab to be cost-effective.

When we tested the robustness of the model by varying each input value one at a time, pembrolizumab only became cost-effective if the price was reduced by greater than 90%. And this remained true for both index patients and additionally remained true, even if we varied multiple parameters at once. Now, the difference between this and the gemcitabine-docetaxel, although gemcitabine-docetaxel was not cost-effective at baseline, it only required slight variations from baseline parameters to become cost-effective relative to radical cystectomy. So for example, the two-year occurrence had to decrease from 57.5 to 55%, or the two-year metastasis rate would have to decrease from 6.1 to 5.9% for it to become cost-effective relative to radical cystectomy.

For the probabilistic sensitivity analysis, for those who vary all the input values across pre-specified distributions, and you do this simultaneously, and then you run thousands of what we call micro-simulations through the model, and these graphs demonstrate the probability of each treatment modality being preferred among those thousands of micro-simulations at varying willingness to pay thresholds. So for index patient one on the left, the main takeaway is that at lower willingness to pay thresholds, cystectomy, as the least expensive option was preferred. However, for willingness to pay thresholds above approximately $120,000, a combination gemcitabine-docetaxel became preferred. For both index patients one and two, pembrolizumab represented by the red line there is unlikely to be preferred even at high willingness to pay thresholds up to $500,000 per quality-adjusted life.

So in conclusion, based on a Markov model for patients with BCG-unresponsive carcinoma in situ of the bladder, pembrolizumab is unlikely to be cost-effective without a greater than 90% price reduction. When comparing radical cystectomy and salvage intravesical chemo, gemcitabine-docetaxel are cost-effective if recurrence and metastasis thresholds are met in future studies.

And then I think perhaps the most significant takeaway from this study is the potential implications Markov modeling can have in this area, particularly in disease spaces, such as this, whereas the International Bladder Cancer Group pointed out, placebo control is not acceptable, there is no approved standard of care alternative to cystectomy, and significant unmet needs exist. These models could potentially offer a way to design and transparently compare outcomes across single-arm trials, moving forward. Thank you, and I am happy to discuss further.

Ashish Kamat: Great. Thank you so much, Kevin, and yeah, let me start off by first just clarifying one thing. The ICER that you are talking about is not the same as ICER the institute, right?

Kevin Wymer: Correct.

Ashish Kamat: Because we don't want our audience to get confused about that. So the institute, because they published recently their findings on comparing pembrolizumab with nadofaragene and (vicinium) oportuzumab monatox essentially, and I don't want our audience to mistake the two. So the ICER term that you came up with is terminology. And if you would just restate that, what it actually means, that would be great.

Kevin Wymer: Absolutely. So that is a great point. And the ICER terminology is a specific outcome, the standard for cost-effectiveness analyses. So it's an incremental cost-effectiveness ratio, and it is essentially the ratio of the cost difference over the effectiveness difference and simplistically, it's the price per quality-adjusted life-year gained. So for our model, the threshold was $100,000 per additional quality-adjusted life-year, which is commonly accepted for these models. And if a treatment results in higher quality-adjusted life years, if the cost of each one of those is above that cut point of $100,000, or the ICER is above $100,000 per quality adjusted life year, it is not considered cost-effective. Now, again, for our model, we did vary that ICER from $0 to $500,000 for all of our treatment modalities as well, but yes, to your point, it is a specific outcome from these cost-effectiveness models.

Ashish Kamat: Great, thanks. So when you did this evaluation, what was essentially the impetus or the hypothesis going in, was it to generate sort of a benchmark that you could assess any therapy that comes into view, or what was the... share with us some of those thoughts?

Kevin Wymer: Yeah, I think initially as you know, this is a very tricky disease space and one with limited options, and also where randomized controlled trials are very difficult. And so modeling, in general, can be useful in these scenarios. And also pembrolizumab was recently released onto the market, and so it was a very salient topic that we were starting to see more from a clinical standpoint, and we are also aware of some significant costs. And so I think together, that sort of drove the idea. The more that we've and Dr. Boorjian I'm sure can speak to this as well, the more that we have been involved in this disease space and this process, though, it's becoming clear that these types of models could play a huge role moving forward. So yes, this model was evaluating pembrolizumab, gemcitabine, and docetaxel, but if we have this type of model, we can use it for other agents as they become available. And if you have additional single-arm trials, you could use it to create effectiveness thresholds or cost-effectiveness thresholds for interventions moving forward.

Ashish Kamat: Right. And Stephen, you obviously have been very active in this space and you've led the SUO-CTC trial on nadofaragene, with the projected costs again of, for example, nadofaragene, some people are putting it in the $160,000 to $200,000 per year range, and then Vvcinium, people are calling it at $160,000 or thereabouts of $190,000 to $160,000 thereabouts. Is there a way you think to use this model then... again, it's not fair, right, because the trials were not done head to head with the different agents, but to sort of model one agent versus the other, given they are vastly different mechanisms of actions and potential target populations?

Stephen Boorjian: Yeah. So absolutely Ashish, I mean, that really crystallizes what I had in mind when we started to do this, which was the debate that has been generated as pembrolizumab came out, and then the nadofaragene trial was conducted and other ongoing trials are conducted in the single-arm space. At varied meetings, and in working groups of how do we take single-arm trials, which were approved for design by the FDA in this BCG-unresponsive disease state, and then arrive at some form of comparison, comparative-effectiveness? Because they are not truly randomized trials of agent to agent in one study.

And I think it's going to be very difficult to do that going forward. So, I absolutely agree that yes, our idea would be that as new agents are approved and then as price points are established and determined, this modeling can be utilized just as you mentioned to sequentially one at a time, sort of compare those new agents to other existing treatments. Now we have pembrolizumab, and as data on the gemcitabine-docetaxel matures, and we see other studies on it, we will have more on that. So yes, I would envision this as a very nice way, hopefully going forward, to be able to do comparative studies across agents that haven't been directly compared in prospective head-to-head trials.

Ashish Kamat: Great. And this would sort of almost be like a phase four evaluation, right? I mean, in the real world as well. So yeah, it's certainly great to have a tool like this that you could look at and evaluate the burden, essentially, the economic burden on the patients and society of all these new agents coming out. A quick question to either one of you, obviously when you look at the radical cystectomy cost, it varies based on centers of excellence, right? The radical cystectomy done at your place is going to cost a system much less than a radical cystectomy done somewhere where there are potentially a lot more complications. And that might be true of the agents too because if they are giving pembro somewhere where they don't know how to handle the toxicity, that could add a whole bunch of nuances. How do you factor those things when you do this sort of modeling?

Kevin Wymer: So I think costs are one of the trickier aspects of these models. And as you're probably aware, costs are not always publicly available. For us, we found that Medicare reimbursement represents the most uniform and accurate way to model this, but certainly, the Medicare reimbursement is based on DRG codes. So it's a lump sum for the procedure performed, in this case, cystectomy. So it doesn't account for the variations in hospital costs, or OR costs, or time in the OR, things like that, those subtle details that certainly can make an impact.

But I think on a population level, at least on a nationwide level, it is the most uniform comparison. And we do account for things like subsequent toxicities of the treatment agents and of cystectomy. For cystectomy, we did both acute and long-term chronic complications and included those costs and their impact on quality-adjusted life years. I do also think it's, the cost is one thing that we focus on, but these models can be used just from an effectiveness standpoint too. A lot of the thresholds for these BCG-unresponsive therapies are based on complete response rates at certain time points. But how do you balance that with the toxicity rates? And so this modeling gives you a way to kind of merge those things together, even if you put cost aside and say, "Is the toxicity level of pembrolizumab, worth the improved, complete response rate?"

Stephen Boorjian: And I have to say that this will take some behind the scenes work with Kevin and some of our research team, but I actually think you kind of opened the next type of study for us, which is to be able to go one level further and put into our models, location of treatment. Because my sense of where you are going with this is that these costs will vary based on center and treatment center. And I think that would be first of all, to be able to demonstrate that would be fascinating at a quantitative level, and then potentially have implications for policy, potentially have implications for reimbursement, but at the very least, have transparency in reporting. So when counseling patients and whatnot, you are able to present them with as granular data as possible. So yes, I absolutely think that is a potentially impactful thing, it was not captured here, but I think that is a really important thing to do going forward.

Ashish Kamat: Yeah, that is sort of what I was driving at. Just recently, I think maybe it was even just this morning or yesterday, there was a big article on the front page of a popular newspaper about how hospitals are not transparent. And how they are not putting, even though they've been asked to state prices, they are not putting it out. And of course, our patients are older, many of them come from far away places, and they can't really access healthcare. So cost is relative to what a patient, he or she can access per se, too, right? So it is very, very important to keep that in mind. I'm glad you guys are doing all this.

Again, in the interest of time. I mean, I could chat with you guys forever, but let me give you both, kind of a closing statement, 30, 40 seconds, take your time, but where do you see this going both from your next phase of the research standpoint, but also really looking at let's assume, pembro, nadofaragene, and vicinium are all three approved, radical cystectomy is something that we have to talk to our patients about anyways because everything else seems to look like it's falling in the 20%, 25% efficacy rate. As a counseling tool, how would you use this when you are talking to your patients?

Kevin Wymer: Yeah. So, as someone going into urology, obviously my interests are starting to move towards those specific disease states. But I think what you guys touched on in the last answer and what you kind of mentioned here, as far as the future of this modeling is really, and Dr. Boorjian and I have talked about this, transitioning it from this higher level, kind of population-based modeling to models that can be used to help actually guide patient by patient decision-making.

And Dr. Boorjian and I have talked about patient cost toxicity and how a treatment decision impacts them financially.  And so to be able to have models that are almost able to be individualized based on where the patient is, what disease type they have, what they value as far as quality of life and outcomes and specific complications, and really creating these into almost decision aids or decision tools to help patients, providers, and potentially institutions, make decisions going forward, whether that's in the BCG-unresponsive disease space, the nephrolithiasis disease space or other disease spaces. Again, I think this patient population represents a good opportunity for this because of the inherent limitations with these studies and as these additional therapy agents come out, we already have the infrastructure and backbone to help guide and increase our efficiency of incorporating them into a model like this.

Stephen Boorjian: So I'll offer if it's okay, just a little extension of that. And I will start by saying, we're not going to let you out of the oncology world yet, Kevin. So I'm glad Dr. Kamat brought up the future of this because you're going to be heavily invested. So I would say that I could see, that I do see our future plans here are two-fold. One, as you mentioned, new agents are reviewed and approved, assessing the cost-effectiveness in models like this. But I think more than that, what Kevin hinted at, was to apply this in sort of a mixed-methods approach, where we look at financial toxicity, we look at patient-reported outcomes, and we are beginning to do this in some other disease states as well, including advanced prostate cancer, where we take sort of patient interviews, talk about some of the implications of treatment toxicities, how they value that, put it together with the financial data.

And I think as you're well aware Ashish, the other issue with some of these new agents is motive delivery. So how would a patient view an IV medication versus an intravesical therapy? How would a patient view dosing schedules, more frequent touchpoints, less frequent touchpoints, together with just the numbers? So I view this study that was published by Kevin and his colleagues as a first stepping stone towards what I hope to be a much more comprehensive assessment of the financial implications and quality of life, as well as efficacy implications to do comparative effectiveness research.

Ashish Kamat: Yeah, absolutely. I mean, I applaud you again on starting on this journey because we need tools like this. And as you had said, putting everything together, because a patient's perception, mode of delivery, all of that is extremely important from not just our plan, but more importantly from the patient's perspective, right? So that's key.

This has been a great discussion and I want to thank both of you for taking the time out of your busy schedules. It's 2021, but I still have to say, stay safe and stay well. Hopefully, we will get to see each other soon in person. Again, thank you, thank you both for taking the time to chat with us today.

Stephen Boorjian: It was our pleasure. Thanks for highlighting our work, and it was fun speaking together. I enjoyed it.

Kevin Wymer: Yeah, really appreciate it. Thank you.