ESMO 2025: Highlights and analysis

Assoc Prof Bishal Gyawali - Queen's University, Kingston, Canada

Hello everyone. This is Dr Bishal Gyawali from Queen’s University, Kingston, Canada, and I’m very pleased to continue our annual tradition of bringing you the major highlights from ASCO and ESMO annual meetings. So here I am back with you again to discuss about the major takeaways from ESMO 2025 Annual Congress.

Let me start from some major trials from breast cancer. As usual, after every ESMO annual meeting we have lots and lots of new clinical trial information and sometimes it’s very difficult to tease apart what is actually meaningful and what is not as meaningful, because everything seems to get so super-hyped during the annual meeting that it’s difficult to tease what is important from what is not. So I’m trying to do an unbiased critical appraisal of the trials that caught my interest from different tumour types at this annual meeting. We’ll start from breast and then we’ll move on to GU, GI, lung and other tumour sites.

So first let’s talk about a very interesting study called the POSITIVE study in breast cancer. I like this study because this is a very pragmatic question that the study is trying to answer. This is about whether or not women with breast cancer who are on hormonal therapy can interrupt that hormonal treatment to become pregnant, and whether that affects their breast cancer outcomes. And it was a very reassuring study, which provided us data that interruption of endocrine therapy for up to two years to have pregnancy did not lead to detrimental outcomes. More follow-up is needed and that’s ongoing but this is a very reassuring study. So let’s start on a positive note.

Now let’s move on to clinical trials. First let’s look at the early breast cancer setting; there were some important updates here. First the updates from the monarchE and the NATALEE trials, which is, as you all know, the trial of adjuvant abemaciclib and adjuvant ribociclib respectively. We know that adjuvant palbociclib, two of those trials were negative, and adjuvant ribociclib (NATALEE) and adjuvant abemaciclib (monarchE) trials were considered positive based on the primary endpoint of disease free survival improvement. Now at this ESMO meeting we got details of overall survival as well and overall survival in the case of adjuvant abemaciclib has statistically significantly improved, and therefore this led to lots of buzz. Having said that, and in the case of ribociclib, the NATALEE trial, overall survival also showed improvement in absolute numbers but that did not reach statistical significance, so it has not still been significantly improved. So this led to lots of discussion about whether this ribociclib data will also turn out to be significantly better with maturity, and whose patients we should select for these adjuvant CDK4/6 inhibitors and so on? But I think some critical thinking is required before making these decisions, because although people were quoting the hazard ratio and saying that the risk of death has reduced by 16% and so on, the absolute risk reduction is not as great. The absolute risk reduction if you look at 5 years is just 1%. The 7-year OS has improved by a little less than 2%, 1.8%, so this is not a huge OS again. For instance, in GI cancers we look at OS improvement of 10%, 8%, so this over seven years 1.8%, I’d say that’s very marginal benefit. Yes, it is statistically significant, but it looks not as exciting in terms of its absolute number.

In addition to that, we should also look at what percentage of patient in the control arm received CDK4/6 inhibitors when they relapsed, and that percentage is not super-high. In the monarchE trial actually it’s less than 50%, so less than half of the patients, when they relapsed, received adjuvant CDK4/6 inhibitors, and that is not the standard of care. The standard of care is patients in a control arm, they should get CDK4/6 inhibitors when they relapse. And so if we are not comparing it with the standard of care, then how can we claim that this is practice changing, based on 1%, 1.6%, 1.8% improvement in survival over seven, eight years, against a control arm that did not get appropriate therapy when they relapsed? And some people are talking about why NATALEE has not shown OS improvement yet. I think there is not much of a difference to see between monarchE and NATALEE. If you look at the hazard ratio, it looks similar. If you look at the margin of benefit, 1%, 1.6%, they all are in the same ballpark. It’s not like we are just chasing for that p-value to go less than 0.05 over the next few years ,and we’re just waiting for that incident to happen – ‘OK, now the p value is less than 0.05 and we can celebrate.’ No, we need to look at the overall picture. There is a huge amount of informative censoring here; we have published about that before as well. The control arm patients are not getting CDK4/6 inhibitors when they are relapsing and the margin of benefit is small.

More than these trials, I was more excited, actually, about the low-dose pembrolizumab trial in the neoadjuvant setting, coming from India where they tested the addition of low-dose pembrolizumab, and by low dose here we mean 50 mg every 6 weeks, just three doses. And that seemed to improve pathological complete response rate comparable to … not a statistical comparison but the outcomes looked similar to what we would expect with full-dose pembrolizumab in this setting. So I have been saying this often and often, that with immunotherapy we are probably over treating our patients because a lower dose at a lower frequency or at a lower duration might be enough. So this was a very good study that showed one more evidence of proof of concept. But we already have enough evidence in several settings now that probably we are over treating our patients with immunotherapy at the full dose, especially the fixed dose.

And the other two trials in this curative setting that made a big splash were the DESTINY-Breast11 and DESTINY-Breast05. So DESTINY-Breast11 was in the neoadjuvant setting after trastuzumab deruxtecan, and as usual I find there is always an error where a certain class of drugs are super-super-hyped. It used to be targeted drugs in the past, then it became immunotherapy, now it’s the era of antibody drug conjugates. So trastuzumab deruxtecan is an antibody-drug conjugate which has already been proven to be effective by improving overall survival, both in second line and first line of metastatic HER2-positive and even HER2-low breast cancer. Now this is a trial in the neoadjuvant setting. It was a three-arm trial: trastuzumab deruxtecan alone versus chemo versus trastuzumab deruxtecan plus chemo, and that alone arm was closed for futility earlier, which actually surprised me. The primary endpoint was pathological complete response, which was improved and it was a statistically significant improvement, but I have several concerns about DESTINY-Breast11. The first is that pathological complete response is just a prognostic marker, it’s not a predictive marker, so it cannot be considered a surrogate for improvement in not just overall survival, even EFS, it’s not a surrogate at all. And EFS is not significant in this trial. So it’s very premature to claim success based on DESTINY-Breast11 results. But even beyond the debate about surrogate endpoints, there are also issues about patients – did they have access to trastuzumab deruxtecan when they relapsed? And there is also the issue of if you look at the Kaplan–Meier graphs, there is a huge amount of censoring here. There is a significant percentage of patients in the control arm who are dropping out of the trial, so that can easily inflate your pathological complete response results because if those patients were dropping out, who are being censored, are the ones who had pathological complete response, then it easily shows you a difference where there is none to begin with. And in fact this trial is overpowered. When they initially designed the trial, they were trying to detect a difference of at least 15% in the pathological complete response rate but that was then changed and the trial was overpowered and the ultimate difference of 11% was also shown to be statistically significant because of higher sample size. So I’m not convinced that DESTINY-Breast11 should change anything at all.

DESTINY-Breast05 looks more promising. This is for patients with inoperable or lymph-node-positive patients with residual disease who would normally get T-DM1. So the control arm was T-DM1 in this trial, and the experimental arm was trastuzumab deruxtecan and it improved DFS, 3-year DFS rate. A little more impressive in this case as compared to the previous CDK4/6 inhibitor trials that we discussed here. It was 3-year DFS rates, so they improved by 8.7%, which is a good amount of improvement. In the HER2-positive case actually DFS is a better surrogate of OS than in other cases. But what concerns me here is two things. One, the rate of serious and fatal adverse events, it’s not negligible. There is 10% rate of interstitial lung disease; there were three or four deaths with the drug, so this is not something that can be ignored. I hate it when people call it well tolerated despite all of these things. It’s not a well-tolerated drug. And especially losing patients due to side effects, it’s very, very troublesome in a curative setting, because many of these patients will have been cured even without getting this drug. If you look at the absolute DFS rate, it’s 84% versus 92%. That means 84% of the patients would actually be cured by surgery and chemo and T-DM1 alone. They would not need trastuzumab deruxtecan. Getting trastuzumab deruxtecan improves that from 84% to 92% which is not nothing, but when you have a high chance of getting cured without the drug, then any fatal adverse event, even one death due to drug toxicity becomes a little too much. The other point about this trial is again, as usual, did these patients in the control arm get trastuzumab deruxtecan when they relapsed?

Moving on from the curative setting to the metastatic setting. In metastatic setting there were a couple of trials in the hormone receptor positive subgroup: the evERA trial and the VIKTORIA trial. The evERA trial compared giredestrant plus everolimus versus endocrine therapy plus everolimus, and both trials are in the post-CDK4/6 inhibitor setting. The VIKTORIA trial tested with gedatolisib plus fulvestrant plus palbociclib against fulvestrant. But in either of these cases OS has not been shown, it is just an improvement in PFS. I find it problematic to rely on PFS, especially when we are combining drugs. When we are combining drugs we are always adding therapeutic burden; when we are combining drugs we are always adding toxicities, and that can only be justified if overall survival has improved. We have written about contribution of component. The standard of contribution of component is that each component in the treatment should be contributing to survival gains, should be contributing to clinical benefit, and that is valid for these perioperative trials as well as these combination therapy trials. So in combination therapy trials, PFS is never enough.

Which brings me to the triple-negative breast cancer setting, the ASCENT-03 and the TROPION-Breast02 trials, both of these in triple negative breast cancer. The ASCENT-03 trial is of sacituzumab govitecan vs chemo; the TROPION-Breast02, the same setting but datopotamab deruxtecan vs chemo. TROPION has overall survival data and there is an improvement in overall survival of 5 months, which is fantastic. ASCENT does not have overall survival data, so on the face of it you might think that one is better than the other, but there is some nuance here that I want to highlight. In the TROPION- Breast02 trial, what happened to the patients who progressed on the control arm? Crossover was not allowed but ASCENT-03, and I want to give props to ASCENT-03, this trial allowed crossover and it went beyond that. Most of the trials, when they let crossover happen, they just say we allowed crossover, but ASCENT-03 provided the drugs to the patients in the control arm when they relapsed, so it’s not just saying, ‘We’ll allow crossover,’ it’s also offering the drug to the patients, and this is exactly what we have advocated for in our Common Sense Oncology Checklist, for designing of clinical trials. We have been asking for this to happen for a long time, that when crossover is appropriate, it should be included into the design and the sponsor should provide the drug at the time of relapse, which is exactly what happened in ASCENT-03. So I very much want to applaud them for doing the right thing. And this trial has shown a 3-month improvement in PFS, OS is still immature, but I would have higher confidence in the OS data of ASCENT-03 than of TRIPION- Breast02, because if I see good overall survival benefit despite crossover, then that means the drug is really good, as opposed to showing a false positive overall survival benefit by not letting the patients in the control arm cross over. So ASCENT-03, good job by making crossover happen.

Speaking of contribution of component, I talked about contribution of component when we are combining drugs, but there is also contribution of component, and we wrote a paper about this in Nature Reviews two years ago, about perioperative treatments. Because in perioperative treatment we are giving the drugs in the neoadjuvant setting plus the adjuvant setting and we need to know if both components are contributing to overall survival benefit. Otherwise it would just be over treating the patients.

Which brings me to one of the most important trials that was discussed in the GU space, the bladder cancer space, in muscle-invasive bladder cancer. This is the KEYNOTE-905 trial, perioperative treatment of EV plus pembrolizumab versus placebo for patients who would not be eligible for cisplatin or who decline cisplatin. On the face of it the results look so impressive. EFS the hazard ratio is 0.4, overall survival hazard ratio is 0.5, 80% versus 63%. What’s not to like about it? And these are impressive results, but there are some nuances again that we should not be forgetting. I’m not very comfortable putting cisplatin ineligible patients and cisplatin decliners into the same group because patients who decline cisplatin, especially when they know that this trial is out there, could actually be very fit patients, good socioeconomic status, good educational background patients who declined cisplatin, and their outcomes will be better than patients who are actually cisplatin ineligible. And that is also reflected in the median OS of the control arm, it’s 42 months, which is remarkable for a cisplatin ineligible population. So I worry about informative censoring in this trial: that patients who are dropping out of the control arm, they are probably not dropping out for random reasons; they are probably dropping out for some reasons, leading to informative censoring.

The other issue here that I wanted to talk about was, as I hinted earlier, the contribution of component, as in when we’re talking about this perioperative treatment, is the adjuvant component actually needed or is neoadjuvant alone enough? But if we had not done this trial we would never know. And this trial did not have any comparison with the neoadjuvant component alone; this was just a perioperative versus placebo trial, so we’ll never know whether that adjuvant component is even needed.

And the other thing here is about, I have been beating the same drum again and again, the subsequent treatment. What did patients in the placebo arm get when they relapsed, when they progressed? Did they get immunotherapy at that time? And we should also put this trial in context with the CheckMate 274 and the NIAGARA trial of nivolumab and durvalumab respectively, because they are also effective options in this space. So it will depend on patient selection and patient preference and so on, but at this point I don’t think that there is any good quality evidence to say that one is favourable over the other. I appreciate these results as very impressive, but I do have my concerns about the control arm patients, the informative censoring and what they got when they relapsed. So we are now talking about bladder cancer, so we are now in the GU space, and in GU space there was also this trial called the ALBAN trial in the non-muscle-invasive bladder cancer space, of intravesical BCG alone versus BCG plus intravenous atezolizumab, which is another PD-L1 inhibitor. In this case there was no improvement in EFS, there was no improvement in OS, so this was a negative trial. Then there was the POTOMAC trial, in a similar non-muscle-invasive space, but with durvalumab. But in this case the DFS was significant.  

But let’s ask a couple of questions. Non-muscle-invasive bladder cancer where BCG alone is the standard of care and the DFS rates with BCG alone at 24 months, these are like 82-85%. OS is not significant in any of these trials; the POTOMAC trial was also not significant for overall survival. So is DFS alone enough in this context? And I would say no. You already have pretty good DFS rates for the control arm and these immunotherapies can have some important prominent long-term toxicities, and if you progress you can get immunotherapy at that point. So I don’t think treating everyone up front with immunotherapy at this early stage without survival data … if there is overall survival data then I’ll be convinced, but without overall survival data I don’t think that’s a very prudent kind of decision.

The other trial in GU space that caught my attention was the PSMAddition trial. This was a completely negative trial in that this trial tested whether adding lutetium-PSMA to ADT plus androgen receptor inhibitor if it improved outcomes, if you have to add lutetium and PSMA to that. It is positive for PFS but it is not positive for OS, it is negative for OS, and in fact there is a detriment in quality of life. So when we keep saying that OS and quality of life are the only clinical endpoints that are relevant for patients, and if OS does not improve and quality of life in fact decreases, then absolutely I don’t think this should be practice changing at all; this should not even be considered.

In prostate cancer, the EMBARK trial was positive, in that this trial looked at patients with high risk of biochemical recurrence and tested whether adding enzalutamide  to leuprolide improved outcomes. Indeed, 8-year overall survival was improved by 9% and this is what I’m talking about. In breast cancer we were talking about 8-year DFS improvement by 1.6% and claiming a big victory. In the prostate cancer space, just for comparison, this is a 9% improvement in 8-year overall survival, not DFS, with a hazard ratio of 0.6. The interesting thing is that the curves start to separate only after four years but, anyway, this is a practice changing trial. Which reminds me of a paper that we published in JAMA Oncology in January of this year, where we looked at this drug, enzalutamide, and how it was discovered. Because this is turning out to be one of the most important drugs in the prostate cancer space, it’s now indicated in almost every context: hormone sensitive, hormone resistant, biochemical recurrent, as this one, so many trials out there, positive trials of enzalutamide. But this drug was developed and the initial stages of this drug’s development were funded by public funds, by government grants, and then it was handed over to the industry who is now making billions of dollars. So in a sense the public is paying twice for this drug, so I just wanted to put this thought out there that we need to ask as a society if we have a drug that is being developed from public funds, public money, how can we make sure that we are not paying for the same drug again by handing the products on to the industry?

Anyways one of the most important trials in the GU cancer space was the IMvigor011 trial, which I’ll talk shortly but before that I wanted to go to lung cancer because in lung cancer there were a couple of trials that caught my attention. One was the PAULIEN trial. This was a Dutch trial that tested pembrolizumab alone versus pembrolizumab plus chemo in PD-L1 more than 50% population. I like these types of pragmatic trials and usually we get very good pragmatic de-escalation trials from The Netherlands. Very grateful for Netherlands as a whole, the oncology community there. For instance, the SONIA trial, the mature overall survival results in breast cancer were presented at this ESMO too and they still continue to show no benefit by giving CDK4/6 inhibitors first line as opposed to second line. We are talking about contribution of components for any combination therapy. This in fact should be the standard way to test treatments, and PAULIEN is doing the same thing for patients with lung cancer, PD-L1 more than 50%. We had a trial of pembrolizumab monotherapy and then there was also a trial of pembrolizumab plus chemo that lumped all PD-L1 more than 1% together, so included patients with PD-L1 more than 50%.

So there was a debate as to whether or not this additional chemo is even needed for these patients, and thankfully the Dutch group designed this trial and started running this trial, and the goal was to show superiority of pembrolizumab plus chemo versus pembrolizumab. It was not to show non-inferiority of pembrolizumab alone, and that is a very crucial point. There is a paper that I have written which is coming out soon about this idea that de-escalation trials do not always need to be in a non-inferiority design. Anyway, so in this trial, before they completed the enrolment, futility was shown as in there was very little chance that pembrolizumab plus chemo was going to show that it was better than pembrolizumab alone. So pembrolizumab alone should be the standard of care for these patients, and in fact I think that it might be worth doing this trial even for PD-L1 more than 1%.

So to complete the lung cancer space, there is just one more trial I wanted to talk about on lung cancer, which is the sacituzumab tirumotecan in EGFR mutation positive non-small cell lung cancer from China. In the post EGFR TKI progression space, overall survival has improved with a hazard ratio of 0.6. I understand that the paper has just been published as well, in New England Journal of Medicine, so I look forward to reading the results in detail. But this looks very intriguing, in the first reading, and if this overall survival is indeed positive and there are no other concerns, then I think this should change practice.

Speaking of trials from China, there was one in bladder cancer as well, in HER2 positive urothelial cancer – disitamab vedotin plus toripalimab. Toripalimab is an immunotherapy that was previously tested in, for example, nasopharyngeal cancers and oesophageal cancers as well, and with very good data there, so now this is being tested  in urothelial cancer in HER2 positive. This is an ADC plus immunotherapy combination. And overall survival has doubled, so that looks very, very remarkable. So if these results, and again I understand the paper has been published in the New England Journal, and I’ll take a deeper dive into it later, but if these results hold and if there are no other issues with these trials, I think we should start changing practice, like we should not forever relegate these trials, thinking that oh, these are coming from China. There is so much opportunity here for the free market and price competition that we always talk about but never happens in oncology.

Speaking of urothelial cancer, and I hinted earlier, IMvigor011 is one of the most important trials, I think, because this is a trial which, to my knowledge, proved for the first time that ctDNA can be a predictive biomarker, in addition to being prognostic. We have seen ctDNA being prognostic everywhere, but this is to my knowledge the first trial that showed that it can be used as a predictive marker as well. So what happened here was, after surgery for patients who were supposed to be in surveillance, the investigators followed them with the surveillance protocol for one year, but they also measured the ctDNA every six weeks and if the ctDNA turned out to be positive during any time point, they randomised these patients 2:1 to get either atezolizumab or placebo. It improved DFS, it improved overall survival, overall survival improved by almost 11 months, which is excellent. In this space I’m convinced that ctDNA can be used to tailor treatment and personalise treatment based on this.

Speaking of the role of ctDNA, let’s move into the GI cancer space, because the trial DYNAMIC-III was also presented which is also a trial of the utility of ctDNA. We saw the escalation component of this trial during ASCO and we saw that escalating treatment, this is adjuvant chemotherapy for colon cancer that we are talking about, we saw at ASCO that escalating treatment based on ctDNA positivity did not improve outcomes, in fact there were hints of harm. At ESMO they presented about de-escalation, as in if ctDNA is negative, can you de-escalate treatment? By de-escalation it means you could reduce your treatment plan, you could de-escalate a treatment plan from two drugs to one drug or three drugs to two drugs or from six months to three months. They could not prove that de-escalation based on ctDNA negativity was non-inferior. So non-inferiority could not be proven. In fact, if you look at the graphs for the high-risk population, de-escalation does look inferior. So yes, this is in colon cancer space as of now, this is just a prognostic marker, it’s not a predictive marker. But this was a super-important study and I am thankful to the investigators for having done this trial.

The only other thing in GI cancer that I wanted to talk about is the overall survival details from the MATTERHORN trial. I was very happy to see that overall survival results have panned out to be positive, a three-year overall survival difference of 7%. So, yes, now I am convinced that this can be practice changing. Speaking of overall survival results, in fact in colorectal cancer there was one more trial, STELLAR-303 of zanzalintinib plus atezolizumab versus regorafenib in the last-line therapy space. Overall survival did improve by 1.5 months but this is very miniscule improvement and we already have other drugs in this space that have better outcomes, like TAS-102 plus bevacizumab, or even fruquintinib. So based on this 1.5 delta OS and this combination had a more than 60% rate of grade 3 side effects, so this is not at all practice changing. I don’t think this should be considered.

So those were the major trials from this ESMO meeting that caught my attention and that I think are worth highlighting and discussing about. A couple more that piqued my interest were in melanoma there was the CheckMate 238 trial’s mature results published and the 9-year RFS is now 44%, the median overall survival is now more than 9 years. This is really, really remarkable. Very happy to see such an improvement in median survival from where we were 15 years ago to where we are now in melanoma. A median OS of more than 9 years, this is really remarkable and something worth celebrating.

But I have been talking about contribution of component, and the results of the NADINA trial were also presented again. This was a trial that compared neoadjuvant nivolumab plus ipilimumab versus adjuvant nivolumab, so even one more reason for us to consider that maybe neoadjuvant alone is enough. Not only here but in every tumour site, we should be doing these trials that compare each component. We should be doing the trials that compare if we are planning a perioperative trial, any future perioperative trials, ideally should be drawn as a three-arm trial where you’ll have one control arm, the other will be the perioperative strategy and the other would be neoadjuvant only. Because I think if we are treating with the adjuvant component without offering an additional benefit from the adjuvant component, then I think even one extra cycle is too much, but most of our adjuvant treatments go on for more than one year. So we should really think about that. These would be the most patient-centred trials.

In sarcoma, there was a trial called RAR-IMMUNE, which tested nivolumab plus ipilimumab versus pazopanib in rare sarcoma types, and this was a negative trial, negative for PFS, negative for OS. But I just wanted to highlight it here, that not all trials will be positive, which is OK, because if we are doing randomised trials we should expect half of our trials to be negative. So people think as if any negative trial is, ‘Oh my god, why did this negative trial happen?’ But if we already knew the results that it would be positive, we don’t even need to do a trial. We do randomised trials because there is clinical equipoise, so there is always a chance that the control arm might be better, the experimental arm might not work, (a), and (b) especially when we are running so many trials, immunotherapy trials. We have had now thousands and thousands of trials and there are always several false positives and that is what I worry about, because people think that negative trials are false negatives but they usually don’t think that the positive trials might be false positives. So in this universe of thousands and thousands of immunotherapy trials, some of the trials that we have been basing our clinical decisions on could even be false positives. And I fear that the same thing will happen with antibody-drug conjugates in the future, because now there are hundreds of antibody-drug conjugate trials, which will swell to be thousands in the near future. So bottom line, I think just repeating that the hazard ratio is significant, the p-value is significant, and therefore it should be the standard of care, just repeating those statements doesn’t solve anything of course. You don’t need any critical thinking to just say that, OK, the p-value is less than 0.05 so we can start celebrating. There will always be these nuances that we need to look into, to correctly understand the evidence and thereby implement it for our patients for their benefit.

I hope you have found this highlights video from ESMO 2025 useful and, as usual, I’ll see you at ASCO 2026.  Thank you.