So, I always enjoy giving a lecture on the non-role of surgery to surgeons. It's always a daunting topic, but I thought this might be a good opportunity to review some of the data that we have seen emerge and continue to emerge with respect to clinical trials that have tried to define this topic. These are my disclosures. So, kind of the first thing to note, I think we all would agree, is that there is an intrinsic bias to cut out cancer. These are just some books that you can, if you go on Amazon and you want to look for a book on surgery and its role in medicine, you find all of these different kind of books out there. And obviously, this one on the right-hand side that says, a chance to cut is a chance to cure. You hear lots of people saying that all the time, and patients realize this as well. So, when we think about surgery, we have to think about what it is. We have to look at the totality of the treatment plan. And I'll get to this in a little bit, but we aren't looking at whether or not surgery is better or equivalent or worse than chemotherapy. We're actually looking at it as a tool in a modality for treatment, and that's important to recall. The other piece of this that's important to recall as we go through this data is to understand what are the analytical endpoints that are being assessed in the trials themselves. So I refer you to these statistical definitions, because these are very important to understand when you're interpreting trials. The first thing you need to understand, and it should be stated in the paper, is what is the hypothesis or the null hypothesis that's being assessed. These errors that we talk about, alpha and beta, represent the amount of probability that we'll be willing to accept based on our testing of this null hypothesis. So the alpha error is this probability of rejecting this null hypothesis. So let's say there's no difference between surgery and non-surgery, when in fact it's actually true. And by convention, not by any kind of any rule, but by convention, we allow ourselves to be wrong in that decision one out of 20 times. That's where alpha of O5 comes into play. Just as important is the beta error, which is the probability of accepting that the null hypothesis is true, but it's actually wrong. So this would be to say, hey, listen, there is no difference between surgery and non-surgery, but in fact, actually it is. So here we allow a little more error. We allow to be wrong maybe one in 10, or maybe one in 55, if you want, in that probability assessment. And those two things essentially help us to define the sample size that we need to enroll to address these questions against the null hypothesis. The confidence that we have in making that decision is based on power, and that's one minus this beta error. So it's important to really understand these because frequently we'll look at a trial like GOG213 and SOC1 that says that we were unable to reject the null hypothesis and then conclude that surgery is the same as non-surgery, and that would be wrong. So it's really important to understand these definitions as we think about going forward. And it gets me to this point about superiority and non-inferiority. So again, these are, again, statistical definitions that help you to inform how you're going to interpret the study that you're reading. So if we power a trial for superiority, so we expect that A is going to be better than B, but we can't reject the null hypothesis that they're not different, then that is a negative superiority trial. But it doesn't mean that they're equivalent. Similarly, if we have a non-inferiority trial that we have designed to basically say that the two are within the statistical assumption of being similar in outcome, it doesn't mean that if we reject that null hypothesis, that they actually is superior. So both of these are really super important in understanding how you interpret the studies going forward. So as I mentioned, what we're not actually arguing here is surgery versus chemotherapy. What we are testing is whether or not in the context of chemotherapy, does surgery add anything over chemotherapy alone? So in both of these situations, we are looking at a cohort of patients defined as platinum sensitive who are going to get chemotherapy anyway, and then looking as to whether or not surgery is going to impact that. Now, I like to show this slide because now more than 30 years ago, we were debating this topic. And here we are today debating this topic. And at this point, this was actually done at a GOG meeting, which we have twice a year. And you can see that there were patients, there were two sides to this discussion. And back in this debate, if you read these two papers, they did define some very important concepts. One of the concepts was the idea of doing surgery in the platinum resistant setting. And that was basically determined that that was of no value. So chemosensitivity is an important component of surgical success. That was one of the major things. The other kind of important definition that came out of this discussion is that if you go to PubMed, for instance, and type in secondary cytoreduction, you get a lot of different types of definitions of what secondary cytoreduction is. In this concept, what we're talking about today is doing a secondary cytoreduction in the setting of recurrent cancer. But secondary cytoreduction has actually been defined as also doing surgery in the frontline setting, and then another surgery in the frontline setting, again, called a secondary cytoreduction. It's also been used in terms of patients who had surgery in the frontline setting, but then had what Dr. Slomovitz mentioned, a palliative surgical procedure for, let's say, bowel obstruction. So it's very important that when we talk about surgical secondary cytoreduction in the context of recurrent platinum-sensitive ovarian cancers, that those parameters are met. And I can tell you that at the conclusion of this debate, there was a call for randomized clinical trials to decide the course of action. Now, I keep updating this slide too, because more and more meta-analyses come out that basically look across the entire experience of trials that have been reported. There have been hundreds, actually, of retrospective studies that have been done to try to show the value of surgery versus no surgery. And if you look at this on the right-hand side here at the bottom, you can see that the hazard ratio for the role of surgery is almost lower than 0.5. So it shows very strong treatment effect for the success of surgery. In fact, if I put the unity line, it's not even on here. Here's the unity line. But one of the things you'll notice is that down here at this I squared here is that there's substantial heterogeneity in these trials, and they vary on a number of different things. I've listed them here, but these are all, they all vary with respect to what's considered an optimal surgical outcome. They vary on what line of therapy the patient actually had the surgery performed in. They vary on how patients were selected for surgery. They also vary on adjuvant therapy, so what they got after surgery, and they vary substantially on what is the primary endpoint. So for many of us, the primary, ultimate primary endpoint for ovarian cancer and intervention is overall survival. But maybe that's not the right endpoint, and these trials all have different outcomes. But one thing that is very clear from all of these trials is that there is substantial selection bias. If you think about it, if you have a patient like the one we discussed today who has isolated disease, has evidence of chemosensitivity, that patient is very, we would feel very strongly about wanting to operate on that patient like we discussed. But would you enroll that patient on a trial that had no surgery as an option? Some of you would, and some of you wouldn't. And that's one of the problems that we have with retrospective studies. So as I mentioned, there are hundreds of studies that have been published, and so we have found some very common ideas here that do flow in the results of these papers. But the major thing that came out of this was that we really needed to do randomized trials. And we have, as I mentioned, five of these trials that have been performed, have attempted to be performed. And we actually do have another one that's actually ongoing right now, which I'll share at the very end, which so we do have still continued effort here. But as Pedro mentioned, he said, what about patients who have enrolled to bulking surgery and then go on to surgery? And so as I mentioned, ERTC 55963 was a trial that was launched in 2000 and closed two years later because of lack of accrual. The three trials that have been reported are listed here, and we'll talk about them. The last one that I won't mention is the SOCR trial. It also closed for futility in 2007. All of the major trials that have been completed have as their primary endpoint, overall survival. The SOC 1 trial did have a co-primary endpoint of progression-free survival. So I say that because when we share the data, you will see PFS presented as an outcome. It wasn't a primary outcome, so there's no analytic alpha allocated to it. But it is of interest for hypothesis testing. So let me go through these trials real quickly. You know them well. GOTY 213 actually was a trial that started many years ago, actually more than two decades ago, as actually a chemotherapy study. And during the time of developing the chemotherapy questions that we were trying to address, much of that centered around the use of bevacizumab. It was amended to allow for surgery, and it was felt that surgery should be done in a randomized way. So when this trial emerged in 2007, it's hard to believe it's been that long, but 2007, we invoked a double randomization. So patients were felt, if they were felt to be surgical candidates from the doctor who was seeing the patient. Those patients were then rolled into a surgery versus no surgery, and then were allowed to go on to chemotherapy. The other part of this trial before the amendment in 2011 was to then do a second randomization to platinum paclitaxel versus platinum paclitaxel bevacizumab. Because that trial was positive, for both PFS and overall survival, this became a physician choice component to the surgical question, which was continued on. In the AGO-OVAR OP4 trial, which was desktop three, this trial used a selection criteria, which was defined as the AGO score, and you can see what it was here. Those patients that were eligible for that were then randomized to surgery versus no surgery, and the chemotherapy they used were standard platinum doublets without bevacizumab. And the SOC 1 trial, which was published a couple of years ago, and then the updated analysis just published in Nature this year, was basically the same study design. It used a selection criteria called the AGO-I model, which was an amendment of the AGO trial criteria, but you can see it included platinum-sensitive patients, surgery versus no surgery, and then platinum-based doublet as the outcome. These trials have a lot of similarities. I listed them here on one slide just for context. You can see that they were relatively younger patients, so this goes along with selection for surgery. They were good-shaped patients eligible to go on to surgical cytoreduction. Most of these patients were advanced stage. All of them had some form of selection criteria. The least formal was in GOG-213, which was individualized for the outcome of a complete gross resection. So in this particular setting, it was left up to the individual surgeon's discretion and their surgical skillset to be able to render a patient complete gross resection and then randomize them. Most of these were high-grade serous. Most of them were highly platinum-sensitive, as you can see here, and that the crossover to surgery in the initial randomization was very low. So again, only 2% of patients in 213 and as high as 6% of patients in SOC 1 who were essentially randomized to no surgery, but the doctor did surgery anyway. The complete gross resection rates, lowest in GOG-213, highest in SOC 1, but they're all basically about the same, so around 70%, and the mortality rates were very low. Now, one of the key differences between these trials is listed in this gray box, and this is gonna become more of an issue as time goes on. In the GOG-213, we don't have the number yet. I'm still trying to get that number, but it's far less than 10% of patients who went on to surgery after they were randomized. The reason that's the case is that in GOG-213, more than almost 85% of the patients went on to Bevacizumab maintenance until progression. So there were very few patients that went on to a surgical procedure after that progression. However, in the SOC 1 trial, there was overall about a 37% subsequent surgery in the control arm, and as much as 50% in at least a couple of the major centers that were participating in the trial. The other major difference here is the use of these adjuvant maintenance strategies, either Bev or PARP. PARP, at the time that these trials was done, was not really available, and so these are low numbers, but Bevacizumab was allowed as an option for GOG-213, and it was four times as high as any other trial that was done, so that becomes an important differentiation. So if you look at the primary endpoint for overall survival, I have them listed here. These are the three outcomes. In the SOC 1, it looks fuzzy here because I've overlaid the two results that we have for SOC 1, so the initial and then the final overall survival. You can see that in GOG-213 that there was no benefit. The hazard ratio is actually above one. In SOC 1, it's below one, but the upper limit of the confidence interval is above one, and DESTOP-3 is the only one which is positive. Now, what's really interesting about these studies is that the surgical arms actually perform very much the same, even though you can see that the trials have different hazard ratios relative to control. If you look at the median outcome for surgery, it's 53.6, 53.7, 58.1, and if you overlay these, what you find is that there is a clear difference in the performance of the control arm of GOG-213, and that's been the major question. It's a clear outlier. You can see that in SOC 1 and DESTOP-3, the control arms are lower, at least at the median, but in GOG-213, it's higher, and so this represents some inconsistency in heterogeneity of the patient populations. Now, one of the things that I mentioned was that crossover to surgery could have confounded the study in the outcome of the control arm, meaning that if a patient went on to get chemotherapy followed by surgery, then they might have done better than expected, so the control arm would have performed better, and so in SOC 1, because they had almost 37% of patients who crossed over to surgery after randomized to control, they did an analysis where they censored those patients and looked at, or adjusted those patients, and looked at the patients who did not cross over. The green line here are the patients who did not cross over, the 70% of patients that did not cross over, and you can see that in that setting, then the hazard ratio looks more like desktop three. Now, remember, in desktop three, about 11 to 10% of those patients crossed over as well to surgery. As I mentioned in the discussion for the case, there has been interest in looking at patients having experienced time off of treatment, and so they have an analysis that's called this treatment-free survival that's been adjusted for that completion of that second-line treatment, and so what you're seeing here are these longer durations of time that patients were alive without having to be on treatment, and as mentioned, this has been a endpoint that they, the SOC 1 investigators, feel is important for patients. Now, these are non-analytical endpoints with the exception of SOC 1, but I just wanted to show you also the PFS curves and how these can be different. Once again, notice that the PVS, the PFS and the surgery arms are very similar, but what is very different is the PFS and the control arm, again, in G02 to 13, which is higher. So if I overlay these curves, you can see that there is a lot of overlap with respect to surgery, but one of the key differences has to do with the chemotherapy alone arm in G02 to 13. So it goes back to what I mentioned. What was the difference? Well, the difference here was that patients on this trial received Bevacizumab in nearly 85% of the patients until progression. So this was part of the experiment that showed the value of using Bevacizumab, whereas in SOC 1 and in desktop 3, adjuvant maintenance therapy was not used. So there's a lot of kind of different ways to assess patients for surgery. As mentioned, G02 to 13 did not use a prospective defined score, but the AGO desktop 3 and the SOC 1 used the TIN score, the eye model, essentially showed that these studies have very high positive predictive value, but relatively low negative predictive value. So they are directional in how well they can help predict patients to who would have a good outcome from surgery. So they're not bad at picking them prospectively, but in picking the patients that did not get good surgery, they did kind of fall apart. So one of the key messages that has come up here in why selection is so important is that if we get it wrong, we actually don't help our patients at all. These are the three trials that have looked at patients who were taken to surgery, but did not get complete gross resection. And you can see that even with the help of Bevacizumab, that there is a huge difference in outcomes in patients who did not get a complete gross resection. So this could be displayed another way. I've listed here that if you look at the gain in patients who got a complete gross resection versus the negative impact in patients taken to surgery, but didn't get a gross resection, you can see how drastic these curves flip. So it just drills home the point that the consideration of complete gross resection and the consideration of adjuvant therapy is very important. I know I'm almost out of time here, so I'm gonna skip ahead a little bit. You have all the slides. I do wanna mention one thing that I believe is probably important to future consideration of surgery in this setting, and that is the interaction between surgery and the novel treatment. So it might be that the type of treatment that we pick is actually more or influenced by whether or not patients get surgery or not. Now, we tried to assess this question in GOG213, but only about 100 patients were able to go through this double randomization. And in doing so, what we saw was that there probably is an interaction between surgery and certain types of chemotherapy, including novel therapy like PARP inhibitors. And this was actually nicely shown in a retrospective study that looked at the role of surgery and chemotherapy with Olaparib in patients that carried a BRCA mutation. So this is kind of like the patient we're talking about in the case example today. And this, again, retrospective, lots of problems with that, but you can see that surgery seemed to have a bigger impact here in this cohort of patients who were selected for secondary cytoreduction. Fortunately, we are not done with this question. There is a trial going on right now called the SUNY trial, which is gonna look at secondary cytoreduction patients who are all gonna get PARP maintenance therapy. It is randomized. It will be 400 patients, and we should know something in about 18 months about this trial. So I wanna just provide some takeaways. I think that as we've discussed today, the case is nicely examined, is that these patients who have good portfolios for surgery, tumor biology is very important to outcomes. So we try to pick those patients who are most likely to have complete gross resection, or tend to be highly chemosensitive, and that can be related to their genomic status. And with that, I'll go ahead and close. Thank you so much.

surgery

cancer treatment

clinical trials

chemotherapy

statistical endpoints

tumor biology