Now we did mention duration of nivolumab, one year versus continuous treatment in lung. Do you want to pick up where we left off with that?

Yes, this is making a lot of noise and people are claiming that you should give nivolumab as long as the patient can take, so continuous therapy should be the standard. But I don’t think we are yet there to make this claim because, number one, this study was not designed to look at the duration of therapy, this study was actually designed to look at safety. Number two, another caveat is overall survival is not different, the difference is only in progression free survival. Number three, even that progression free survival is actually a bit intriguing. The progression free survival is not from the time of start of treatment, it’s after one year.

So they’ve got a trial for duration where they’ve not been powering it to look at the success of duration on overall survival but for PFS up to a year, or starting at one year after treatment when they’re looking at the success of continuous versus… That sounds like a very muddled way of going about things.

Yes, you are trying to make a comparison between continuous versus one year but you are starting to measure the duration only after the end of one year. So it seems very obvious that if you keep on giving a drug versus not giving the drug then it will…

The drug will keep drugging.

Yes, it will halt your progression to some extent. So finding a difference in progression free survival by giving the drug versus not giving anything at all might not be that much of a big deal. We have seen that there is no difference in survival so this is a bit of a tricky situation. You have a trial that is not powered to see the difference in duration, this is an exploratory analysis, this is not the primary analysis of the trial. So you are measuring progression free survival after one year and you are finding a benefit in PFS so we can’t make the claim that one year is inferior based on this data alone. There was a very nice discussion in this trial by the discussant and he made a very nice point that this question is a very important question, whether one year is enough or you need to keep giving the drug for a long time. So he argued that this should be tested in a proper, prospective randomised trial which is designed to answer this question. So we should not rely on exploratory analysis for this type of important questions. But the question is who is going to fund the trial and that’s a different story.

I’m reminded of the discussion raised by Dr Ben Goldstein, I think it was at ASCO, of the dosing according to recommended dose. Dosing according to recommended duration seems like an equally important measure for, first of all, like you say, controlling disease but also for cost. An unending future of possibly paying for one of the more expensive drugs available seems not very encouraging, that’s for sure.

Yes, it’s very prudent to ask whether a limited duration of therapy can give the same efficacy. It’s very important for PD-1 inhibitors, CTLA4 inhibitors, the immunotherapy class of drugs, because we have accepted for these drugs work for a longer duration and sometimes the patients receive benefit even after stopping the drug. Even in this case you take the drug for one year and then you progress and you can give the drug again. So we need to test this in a properly powered randomised trial. This is too important a question to leave just for exploratory analysis.

Moving on to some of the other lung data, we have had a good run of success with lung cancer in the last few years. Do you think that PACIFIC data will be hailed among one of the big breakthroughs?

It is one of the most important trials in recent years for stage 3 lung cancer. But whether it will be a game changer, whether it will be ground-breaking, will depend on the overall survival details. So right now we don’t have data for OS and this is a trial comparing maintenance durvalumab for one year versus placebo after chemoradiotherapy for more than two cycles with a platinum based agent. So in this setting the trials that we have had until now all had a primary endpoint of overall survival but this is the trial where we have co-primary endpoints of progression free survival and overall survival and overall survival data are not reported yet. We have data for PFS. But we have to accept that the data are a very substantial improvement; the data show that there is quite a substantial improvement in PFS with a very excellent hazard ratio. So all of us assume that this will, obviously, translate to OS benefit and that’s why we’re excited about it. But the fact of the matter is that we don’t know that yet, we hope it will. It looks very likely but, as we discussed from my study, the correlation between PFS and OS for immunotherapy is very low. So there is a chance that it might not as well. There is a high chance that it will lead to an OS benefit but there is a small chance that it might not and if it does not then what does it mean? It means that we are giving a drug as expensive as durvalumab for one year as maintenance. So until now what we do is we don’t give any maintenance therapy for these patients, so basically they don’t take anything after the chemoradiotherapy. Now we are asking these patients to take instead a drug for one year in the hope that they will live longer. So we need that data. So of course it is an exciting trial and most likely it will lead to an overall survival change; whether it will get approval or not, of course it will based on the current patterns with the FDA and EMA, they approve virtually most of the drugs that have p less than 0.05. So it will get approved. If it gets approved will it be used in clinic? Of course it will. If you have the drug you use it, usually. If the patient comes to you and you counsel the patient that we have a drug called durvalumab and would he like to use it then if the patient is informed and he can make a very good decision about OK, but what happens if I don’t use it and when the cancer comes back I can use immunotherapy again, right? So will my survival change in that case? We don’t have an answer to that. If the patient asks me if I use durvalumab now versus I don’t use it now but I use pembrolizumab when I get a relapse, will it affect my survival? I don’t have the answer to that question yet and that is a very important question that informed patients will ask and we need to answer the patients. So I’m very excited about it, of course, because such a huge margin of PFS benefit has not been seen for this patient group before but I’m also maintaining my cautious optimism. So I’m optimistic but I’m also cautious whether it will or will not lead to a survival benefit. Seeing the margin of benefit in PFS I hope it will lead to OS but I’m not sure yet, I’m not sure yet.

Cautious optimism.

Yes. I really need to see the OS data. It is important to report the OS data; I actually can’t understand why they did an interim analysis with PFS alone.

Their results were, like I say, very convincing at this interim.

For this patient population it’s very obvious that OS is the endpoint that should have been reported, that should have been, because all the trials until now for this patient population have reported overall survival. In the case of lung cancer you usually expect overall survival data because these are the patients who don’t have very long survival. But here we are, they have reported PFS data which looks very, very good and it will likely get approved and then there is confusion in the clinic.

Speaking of confusion in clinics, that takes us onto the French study IFCT-0302, I think it was, looking at the utility of CT scans in lung cancer. The findings being maybe, kind of unsure, it seemed.

For me it is a practice changing study because they showed that the frequent use of CT scans for surveillance after the resection of non-small cell lung cancer did not improve survival. That was very clear cut. So the only question that remains is whether the patient is OK with that because the patient will have anxiety, he wants to know whether his tumour has come back or not. So it’s a very delicate matter. But when I saw the presentation of this study I was reminded of one important study in ovarian cancer by Dustin et al a number of years ago in which they checked CA125 labels for ovarian cancer patients. They saw that whether screening for recurrence with CA125 actually improved survival or not and it did not, similar to this trial in which they checked CT scan screening frequently whether it improved survival or not and it did not. But a couple of years ago there was one study published in JAMA Oncology in which the investigators tested whether any practice pattern changed before and after the publication of that study showing CA125 measurement for surveillance of recurrence is useless. They found that it absolutely did not. Even if we had a study that showed there is no necessity to measure CA125 or there is no necessity to start immediate treatment, even if CA125 is high, people were still doing CA125.

We, in fact, spoke with Dr Craig Barrington from a Welsh hospital board talking about the unnecessary or excessive repeated tumour screenings that he was tracking and at potential cost of thousands upon thousands upon thousands of pounds but also that there are patients for whom having the scan increased anxiety, other patient having the scan were offered the sense of surety. It seems like that is going to be a very important balance in terms of their quality of life of do you live in fear or do you live with the certainty either way?

That study was also very important but the difference is that was in a primary care setting so this is the study that looked after using tumour markers for diagnosis of cancer. But right now we are talking about people who had cancer and who had the tumour resected and we are following them up for recurrence of tumour. So this is a bit different population but still. I am very interested to see maybe ten years after now we do a similar study to that of ovarian cancer and we see whether people change their practice with regard to CT scans or not. It would be a very interesting thing to observe. Now we have data that doing frequent CT scans is not necessary but you never know. It’s very difficult to change practice in medicine once a practice is ingrained. There was a very good editorial by Goodwin in JAMA Oncology which accompanied that ovarian cancer paper and the title of that editorial was The Fatal Attraction of Testing. So I think that is a very appropriate summary of what we see in oncology. We have been fatally attracted to testing and we have conveyed that message directly or indirectly to patients so patients are also always worried about whether they should get tested often. But patient education, physician education is a must. We need to teach them that no matter if we scan you at three month intervals or at six month intervals or yearly it doesn’t make much difference. If you are really worried then we should maybe do at least one per year rather than doing it three monthly or six monthly. That could be one reasonable trade-off. But the data are clear that at least for the first two years surveillance with the usual CT scan absolutely did not improve any survival.

I’m just reminded now, I think it was shared by yourself and Vinay Prasad on Twitter of that section looking at guidelines where patients are now becoming aware of guidelines saying you hit a certain age do you start screening for prostate, for PSA? Do you start conducting mammograms now? And rather than just rolling them out wholesale you have that conversation to say the guidelines recommend that we start going forwards with this. Maybe that is a way of embracing that change to say here is an option.

I completely agree but, as I told, The Fatal Attraction of Testing, the ovarian cancer guidelines don’t recommend to do CA125 measurement for screening for relapse in ovarian cancer patients. But still here we are, we are continuing to do that.

A note saying this is against guidelines, I’m doing this not following any regular recommendation but I’m still going to do it would make patients ask why.

I think we always try to play it safe even though we have the evidence that it’s not actually safer compared to not testing. Also maybe it’s also because the patients want to be sure that they are cancer free so you always try to play it on the safe side and let’s test, let’s be doubly sure. But the thing is, this is science and science has shown that doing that doesn’t actually matter.

One last lung trial to mention, the FLAURA trial for osimertinib.

Yes, it was a very important trial testing osimertinib versus gefitinib or erlotinib for EGFR mutated lung cancer patients. This was a trial that many of us looked forward to and it showed improvement in progression free survival with an excellent hazard ratio and nearly 8 months more of benefit in progression free survival. Again we are coming to the tricky thing – they saw benefit in overall survival which was less than 0.05 but this was an interim look so the p value criteria for saying this is significant was not met. So in simple words the PFS was significantly improved but OS is immature at the moment and it was at the moment the OS has not significantly improved. Now the question is whether this is a practice changing trial or not. So the simple answer is we need to wait for the overall survival data, then only will we know.

Another tricky thing about this trial is we need to also know how many patients who received erlotinib or gefitinib received osimertinib in the second line because right now patients who receive erlotinib or gefitinib when they progress, most of them progress because of the T790M mutation. Osimertinib is a drug that actually targets that mutation and so the current standard of care is you give erlotinib or gefitinib as the first generation TKI in the first line and when they progress you check for the 790M mutation and if they have the mutation you give osimertinib. So we need to know whether that strategy versus osimertinib up front as done in FLAURA makes any difference. Because if it does not then you are essentially depriving these patients from gefitinib or erlotinib because once you progress on osimertinib the patient doesn’t respond to erlotinib or gefitinib. So you need to be actually sure by using osimertinib first line you are not actually depriving the patient of erlotinib or gefitinib. Tony Mok made very good points about this trial in his discussion and I agree with what Mok said. In this FLAURA trial there were maybe out of 230 patients there were still 64 patients who were still responding to erlotinib or gefitinib. So you have a substantial number of patients who are still responding to the first generation TKI. Then they progress and then you can give osimertinib. So they will have quite a long duration of progression free survival. So this is a case of A B, combine testing A B, drug A B, versus A followed by B. So we can say that if the PFS of A is 7 and the PFS of A B is 10 then you say, ‘Wow, this is an improvement, 10 versus 7.’

It’s bigger.

But you need to see A followed by B. So A gives 7 and B gives, let’s say, 5 then actually when you follow A with B you are getting 7 plus 5, 12, which is greater than 10. So this is what we argued in our Nature Reviews Clinical Oncology paper. So this is a typical case example for such a scenario and it would be very interesting to see overall survival data as well as to see how many patients actually received osimertinib in the second line.

There were some breast trials which we haven’t mentioned yet which are something to talk about.

Yes, in fact there were a lot of breast trials at this ESMO. If you ask me in simple words were there any practice changing trials for me there were not. One interesting trial was the NeoPAL trial in which they tested palbociclib as a neoadjuvant therapy and it was a negative trial, it failed. It was important to share that palbociclib actually failed in the adjuvant setting which I had not expected, to be honest, because it has improved outcomes in the metastatic setting. So I was questioning whether it would improve overall survival in the metastatic setting too because it has already been approved in the metastatic setting and people are taking it to be standard of care because it has improved progression free survival by a substantial amount. But we don’t have OS data yet and we need actually improved OS.

We seem to keep coming back to this same point of ‘but do you live?’ With NeoPAL at least we have that question raised a lot more clearly of this negative trial. It’s not going to change practice but it does at least guide further practice, further research into what will be leading some way of improving survival overall.

Yes, and talking about ciclibs, CDK inhibitors, there was another study, MONARCH-3, of abemaciclib. People were discussing on Twitter about we have many ciclibs now, we have palbo, ribo and this new kid on the block, abemaciclib, so which to use. I had a simple answer – use the one that improves OS but the thing is that none of them have data to show that they improve OS yet. So we have options and people can argue that actually it’s very difficult to improve OS in the case of hormone receptor positive breast cancer because you have too many lines of therapy.

This is where the conversation regarding PFS2 first came about was, again, that relapse after relapse with so many confounding lines, perhaps. Like you mentioned about osimertinib closing off certain avenues for subsequent treatment.

Yes, and we had actually the same argument with everolimus, it was also approved for hormone receptor positive, HER2 negative, similar to the ciclibs and it also did not improve overall survival. We had similar arguments that time too. But if your drug is really good it will improve survival and we have an example from breast cancer itself – pertuzumab. Pertuzumab plus Herceptin plus a taxane in HER2 positive breast cancer has drastically improved OS. But in the case of hormone receptor positive breast cancer we have not seen a drug that actually improves OS for a long, long time. Everolimus did not improve OS and the ciclibs we don’t have data on OS yet. We actually wrote a paper in the JCO about this which was titled Same Data, Different Interpretations. In triple negative breast cancer we had a history of bevacizumab getting FDA approval on the basis of PFS but failing to improve OS and then its approval was revoked. But for other drugs which do exactly the same thing, improve PFS but don’t improve OS, but they stay on the market. So this is a bit of a double standard, either you should approve bevacizumab again because we don’t care about OS anymore or if you revoke bevacizumab then you revoke all other drugs that don’t improve OS.

I’m glad I’m not the kind of person who has to make that decision. Other breast news that came out was the LORELEI trial, I’m unclear on the pronunciation for that, with data which I believe is now already being fed forwards into the phase III SANDPIPER trial where, again, lots of good signs early on. The phase III trial at least I think it might still be recruiting or at least still ongoing and we’ll, again, have to wait and see how this turns out.

Yes, this was a trial of PI3K inhibitor with letrozole in the neoadjuvant setting.

Yes, taselisib.

They did not find any difference in pathological complete response. They had a marginal improvement in objective response rate. This was seen at 16 weeks and people are telling it as a very important study, positive trial, because this is the first positive trial in PI3K inhibitors for breast cancer. But these are hormone receptor positive patients and for these patients we usually give adjuvant hormone agent for five years, ten years and they do pretty well. We have data with 16 weeks of therapy and your PCR is negative and objective response rate is positive by a small margin. Then people say pathological complete response is maybe not a good surrogate endpoint for these types of drugs. Until now they have approved so many drugs based on pathological complete response, pathCR rate, for so many drugs in breast cancer and people have always said pathological complete response rate is a very good surrogate endpoint so we should approve based on PCR. Now this drug does not improve PCR and people are arguing that PCR might not be a good surrogate endpoint. So I found it very absurd.

There is definitely room for maybe a satellite symposium on endpoints. Probably a full conference on endpoints. Just who gets to use which endpoint for when depending on if you do or do not want it to agree with you.

Yes, I think that would make for a very good discussion or debate session.

A few quick final notes, then on a few of the diseases we haven’t got to looking at - cancers in the head and neck, the KEYNOTE-040 data for pembrolizumab.

It was intriguing because it did not improve overall survival but people were expecting it to improve overall survival and it has already got an approval based on PFS data. But we have nivolumab approved for a similar setting, second line head and neck; nivolumab has clearly improved overall survival, nivolumab has shown better quality of life. So why is this debate about pembrolizumab? It has not improved overall survival, it is a negative trial and you have a similar drug which has positive data with better quality of life. So I don’t understand why we’re having this debate.

Maybe another one for the endpoints conference, I believe. Well, it missed out on significance by 1% and 1% on the other way, 20% to 21% that’s significance and there would be a very different discussion coming from it. So significance but for who, I guess.

If we did not have any other option then maybe this debate might have some meaning to it. But we already have, we already have nivolumab which has improved survival clearly with a better quality of life.

If there is, for example, any subgroup analysis for PD-L1 expression, PD-1 expression, then that might give some stratification as to who this most suitable for, if there were any subgroups way down the significance that are just not quite reaching the thresholds they’re after then there’s room for subsequent investigation at least.

Yes, I need to look at that trial in detail but we need to be cautious with the subgroup analysis whether pre-specified or just exploratory looking back.

Finally onto the JACOB trial.

This was an interesting trial because this is a trial of HER2 positive gastric cancer. We have HER2 positive breast cancer and in HER2 positive breast cancer we have pertuzumab as a very good drug so it was a reasonable thing to test it in gastric cancer, HER2 positive gastric cancer, in which we have trastuzumab already approved which is also approved in breast cancer. So we tend to think about precision medicine and we say that we need to match the name of a drug to the name of a mutation. So this was a very nice example because pertuzumab is a HER2 targeting agent, very good results, overall survival results, in breast cancer. They tested it in second line gastric cancer and it failed, it failed to improve outcomes. So all HER2 positives are not the same so HER2 positive breast and HER2 positive gastric are different.

It’s hallmarks but certainly not the same disease, not by a long stretch.

Yes. So I have never been comfortable with this strategy of matching the name of a drug to the name of a mutation so there is more to that than just matching the names. So, HER2 positive colorectal, HER2 positive gastric, HER2 positive breast, they are all HER2 positive but they might behave differently. So they might behave the same but they might behave differently so we need to have trials like this to confirm the benefit.

Yes, like you say with your poster, just going on guesswork and expectation that if it improves PFS it will probably improve OS, it works for one HER2, probably works for another. That’s unscientific.

Yes.

Is there any conclusion we can draw apart from that we look forward to overall survival data for a lot of trials when it becomes available?

Yes, we really do look forward to a lot of trials for their mature overall survival data. I do hope that we get those data. This conference has been phenomenal in every sense of the word because there have been a lot of these important trials that we discussed. When we go back home there are a lot of things that we can change for the sake of our patients. That is the only thing that matters – the benefit of patients. I do hope that this discussion of ours will lead to some insights to some oncologists somewhere in the world.

Somewhere. Until next time, thank you very much for joining us.

Thank you very much. It has always been a great pleasure.

Until next time.