Courtesy of
froggmister
Dec. 5 Al Musella interview Dr. Linda Liau transcript
Yesterday when Al tweeted out that he had reposted this video he noted that he was hoping to get the Q and A part up soon. If he does I'll add to the transcript and repost.
One thing that sticks out to me is that when LL repeats something multiple times it's probably best to listen, and she continues to come back to "the activation of T-cells into the tumors is a necessary, although not always sufficient, first step in terms of activating an immune response to glioblastoma. It's a necessary step." Take that with some of the amazing preliminary results she has shown in the UCLA trials with DCVax and PD-1 and Poly-ICLC and it becomes really hard, in my opinion, to continue to keep that door closed.
Here you go.
---
Al Mussela: Welcome to the Musella Foundation’s webinar series. Tonight’s topic is the DCVax vaccine, and our very very very special speaker is Dr. Linda Liau, who is the chairman of the Dept Neurosurgery at UCLA and professor and Director of the UCLA Brain Tumor program, and is the former editor-in-chief of the Journal of Neuro-oncology. Take it away Dr Liau.
LL: Thank you Al, thanks for having me. I’m going to start my screenshare and get started with my presentation, and I’ll take some questions at the end. So can you all see my screen OK?
AM: Yes, perfect.
LL: (slide 1, 0:50) OK, so today I’m going to talk about the paper we recently published on the Phase 3 trial results and kind of go into a little more detail, about the details of that paper.
(Slide 2, 1:04) Here are my disclosures. (Research grants/funding to institution: NIH, NWBO, Merck. Advisory Boards: Insightec, Inc., ImmPact Bio. Stock Shareholder ClearPoint Neuro).
(Slide 3, 1:11) So, dendritic cells vaccines are really based on the concept of the dendritic cell; the dendritic cell is a professional antigen presenting cell. They are a normal cell in the body and they were discovered in the 1970s by Dr. Steinman, but it wasn’t until the late 90’s when people learned how to grow these cells in large numbers outside of the body. We were one of the groups that started exploring the use of dendritic cells as a vehicle for vaccines. So the concept is really taking the antigen presenting cells and loading them with tumor antigens; in this particular case, the antigens from autologous patient tumors. The cells are then injected into patients and thereby activate T-cells, and these resting anti-cancer T-cells can then grow and proliferate. They get activated and then they can divide and proliferate then go on to t tumor site to attack the tumor.
(Slide 4, 2:33) So, we and others have over the years done lots of pre-clinical studies. We were the first to do the early phase clinical trials using this vaccine, and one thing that we do know is that dendritic cell vaccination does get T-cells into the tumors. The infiltration activation of T-cells into the tumors is a necessary, although not always sufficient, first step in terms of activating an immune response to glioblastoma. We have done several early studies with this vaccine in patients, many of whom are still alive today, and one thing that is consistent among long term survivors is that they do have T-cell infiltration into their tumors.
(Slide 5, 3:27) This is the schematic for the Phase 3 trial. It was a Phase 3 multicenter randomized clinical trial of autologous DC vaccination. 331 patients. The patients underwent screening and surgery. Following surgery they had leukapheresis in order to collect the blood cells to make dendritic cells. The dendritic cell vaccine is essentially the combination of the dendritic cell and the patient’s tumor tissues taken at the time of surgery. Then patients underwent standard chemo-radiation following surgery during which time the dendritic cell vaccine was made. After chemo-radiation patients were randomized to placebo vs. DCVax and they were treated on Day 0, 10, and 20 with the vaccine, and had booster injections every 2 months for the first year and every 6 months for the two subsequent years.
(Slide 6, 4:31): This study was conducted at 94 sites in four different countries including the US, Canada, and the UK and Germany. Just completing this trial itself was quite a feat as you can imagine. There was a lot of logistical coordination that went into getting this trial started and getting treatments to these various sites. One thing it did show is that it is feasible and can be done in a wide variety of clinical settings which I think is a strong point because some clinical trials can only be done in very specialized tertiary and quaternary academic centers whereas this is something that can be more widely accessible.
(Slide 7, 5:23) So this was the patient enrollment. 331 patients. This was initially designed as a randomized controlled trial with 2 to 1 randomization. 233 in this arm, 99 in this arm who got placebo. Of note, even though the trial was initially started in 2007, recruiting was paused for economic reasons following the recession in 2008, so the majority of patients were actually enrolled between 2012 and 2015, so that’s something to keep in mind when considering the comparator trials that this trial was compared to. The reason we compared this to external controls was because of this crossover arm. Because of the crossover arm most of the placebo patients eventually crossed over to get DCVax, and the few that did not crossover, many of those had actually either died or were lost to follow up, so there were very few control patients left for analysis. Basically the control arm was depleted, and that’s why we couldn’t do a comparison to that cohort.
(Slide 8, 6:53) One issue – and we’ve been very transparent about this – is the issue of Progression Free Survival (PFS). When this trial was initially started and the protocol was written in 2007 the endpoint of progression was to be determined by using something called the McDonald criteria. But then, as many people may know, as the trial was underway the field began to realize the problems with using strictly the McDonald criteria for progression, so it went on to the RANO criteria, and RANO was not sufficient so subsequently that has been changed to iRano, and there have been problems using iRANO as well, so now there’s been talk of using modified RANO (mRANO). And the problem is essentially illustrated here. This is a patient who received vaccination, received surgery and got vaccination, and then is noted to have progression based on strict McDonald criteria but this patient was clinically doing very well; didn’t have any symptoms, didn’t have any other issues, and then over time this area regressed on its own without any further treatment. So in this particular case this would have been deemed as a pseudo-progresser but at the time we wouldn’t have been able to know that. During this trial the radiology review was done centrally by two radiologists that were unaware of the treatment cohorts that the patients were on, and in over 50% of the cases the radiologists themselves did not agree. So that of course made determination of PFS a very difficult endpoint. When you have an endpoint that could not be reliably determined that made that endpoint essentially very difficult to assess as a valid endpoint. Even when iRANO came on board – iRANO stands for immunotherapy Response Assessment in Neuro-Oncology – one thing that iRANO required was a follow up scan in 3 months, but the reason we moved from iRANO to mRANO is because, if you can imagine if you are a patient and they saw progression and the response is “well let’s go 3 more months and get another scan” a lot of patients on clinical trials that used iRANO had to be censored because there was a lot of dropout from those trials. So I think the field is still trying to figure out how to determine PFS in an immunotherapy setting in glioblastoma.
(Slide 9, 10:11) With all that being said, while the trial was underway, but before data lock and before unblinding, the Statistical Analysis Plan (SAP) that was submitted to the regulators was designed to focus on Overall Survival (OS) as opposed to PFS. This was the design of the SAP, and essentially because of the depletion of the control and because of the crossover the primary endpoint was then written to include survival compared to external controls. Then as you can imagine, this was a trial done in the newly diagnosed setting, so when patients actually crossed over they were deemed to be patients who had first recurrence. So that group that crossed over were then compared to overall survival in clinical trials of rGBM. This was the primary endpoint and this was the secondary endpoint.
(Slide 10, 11:23) So talking a little bit more about the SAP, the first thing that needed to be done was to match the trials to find the comparator trials on which to compare these patients. This was the selection of the comparators was done by an independent statistics firm based on four pre-determined criteria used to match these trials, which included the contemporaneous time period from which the patients were enrolled into the trial, the similarity with eligibility criteria, and treatment protocols. Using these 14 criteria there were 5 trials that were done around that same time period that were used as comparators. This is a graph showing the control arm of those five trials. As you can see they overlap quite a bit, actually they were consistent in terms of the Kaplan-Meier survival curves for the control arms of these trials. This comprised more than 1300 that essentially received radiation and temozolomide and served as the control population.
(Slide 11, 12:49) In addition to taking these patients in control arm of these trials, one bias that could be introduced if there are different characteristics of these control arms. So we used an analysis called a MAIC analysis, Matching Adjusted Indirect Comparison, this is used quite frequently in health care economics, in population health analysis. What it does is actually tries to compare individual level patient characteristics with weighted characteristics in a population of patients - in this case, the control arms of these trials. This is a way to do as close of matching as we can of these patients when individual level patient level data is not available for propensity score matching. One thing I would advocate for and hope for in the future is that when we do these large trials that the patient level data can be made openly available for comparisons for subsequent trials. I think that would be very beneficial to the field. But essentially these characteristics were a match for each patient in our trial were matched to characteristics of patients in the comparator trials and they were matched for things like age, extent of resection, MGMT methylation and several other factors. In addition to the matching we also did sensitivity analysis to check for comparator differences. There were five different sensitivity analyses that were performed doing each of these comparisons, leaving out one comparator at a time, and even with each of these analyses the statistical differences between our treatment arm and the external control population did hold out to be true. We also did a 6th sensitivity analysis whereby we took out two comparators where it was unclear whether the early progressors were excluded from their trials and only did the comparison to the 3 other other trials where they did, as in our trial, take out early progressors, and those data also showed a statistically significant difference. So basically with the sensitivity analysis the outcomes came out to be the same.
(Slide 12, 15:54) The was also specificity analyses that were done for validation of external controls because as you could imagine one potential bias could be that if using this approach you could erroneously have a negative trial turn out to be positive. So in a way to kind of control for that each of the 15 studies that were used as comparators, the 10 newly diagnosed studies and the 10 rGBM studies, were taken and we took the treatment arms of these trials and compared them to the external control population using the same methodology that we used for this trial. What that showed was that all the trials that were negative in the randomized control setting, you know where these randomized trials are done, trials that were negative were still negative when compared to these external population of patients, and all the trials that were positive, which was only one, the TTF trial, actually did turn out to be positive. So at least with this level of analysis there was some validation that the negative trials, if a randomized controlled trial were done, were still negative and the positive trials were still positive.
(Slide 13, 17:21) This is the baseline demographic and clinical characteristics of the comparator trials, compared to our treatment group, and this is the pool of external control of 1366 patients. Frankly I don’t think we should be subjecting another thousands of patients to randomized controlled trials when the data from these arms are very consistent. The KM curve is very consistent among these various trials. MGMT methylation was very similar, as well as residual disease for the trials that we had those data for.
(Slide 14, 18:27) These were the baseline characteristics for the rGBM patients. In this group there were 10 different trials amounting to 640 patients and the 64 patients that crossed over constituted our treatment arm for the rGBM group.
(Slide 15, 18:44) Here are some additional baseline characteristics for these trials and of our external control patients as well as treatment patients in our nGBM DCVax group as well as the rGBM DCVax group. I won’t go through all the details expect to say that the characteristics were very similar. Of note, we also looked at IDH mutation, because there was the thought that if there were a lot of IDH mutated patients that could be why there was increased survival. The percentage of IDH mutated patients is only 3% in our treated patients which is similar to the external controls.
(Slide 16, 19:33) So this is the data. There was a significantly significant difference in the median OS in the DCVax treated patients compared to the external controls; granted this was not a randomized controlled trial and I realize the limitations because of that, but given the circumstances probably as close a match as we could perform to validate an external control population. One thing I thought was particularly interesting was in the different subgroup analyses which showed even more robust hazard ratios, when we looked at patients for instance who were over 65 or who had significant residual disease or MGMT methylation.
(Slide 17, 20:37) This is the KM curve for the crossover arm, so placebo crossed over meant that the patients crossed over and received DCVax at recurrence, that group compared to the external controls from the external group. One thing I would like to note (went back a slide) is that these tails are actually KM estimates. A lot of these external comparator trials did not actually follow patients all the way out to five years. They ended in less time than that. So these are not actual data on the external controls but the KM estimates based on the statistics. (Back to rGBM slide) With that being said the difference is still significant.
(slide 18, 21:31) This is the landmark survival rates based on the KM estimates for the external control population. Our DCVax group had a 13% 5 year survival in the nGBM setting and 11% at 30 month survival (in the rGBM group). This may actually be higher as many patients are living out past 5 years but we stopped the analysis at a specific time point so this is the KM estimates at 13%.
(slide 19, 22:18) As far as the subgroup analyses, and I think there were some very interesting hypotheses that came out of the subgroup analyses, were somewhat unexpected, and could lead to further study in these areas. One is that there seemed to be a significant survival advantage in patients over 65 who had the dendritic vaccine vs the controls, and that bodes well for this having some effect in these older age group patients. Note this doesn’t mean that the vaccine worked better in patients older than 65, just that it suggests that worked relatively better than the control patients who were over 65. If you look at the median survival here at 15.6% whereas here in the patients under 65 it was 19.6%, so it still works better in younger patients, but relatively better in older patients compared to the external controls.
(slide 20, 23:35) This one was very surprising to us. There is actually a greater survival advantage in patients with significant residual disease (SRD) compared to minimal residual disease (MRD). As a surgeon we always taught and we thought that taking out as much of the tumor as possible leads to a better prognosis, and that still is the case; you see in the MRD group that survival is still longer than the survival in the SRD cohort, but the relative survival advantage is greater in this group. What this kind of suggests is that perhaps…one thing that dendritic cell vaccination that we know it does is that it gets T-cells into the tumors, and perhaps in order to have a more diverse repertoire of antigen presentation in epitope spreading, there may need to be some residual tumor still there so T-cells that are still there, once they get into the tumor and get activated there may need to be residual tumor there to enhance the immune response and promote epitope spreading. This hypothesis perhaps needs further validation but I found that to be very interesting in this case.
(slide 21, 25:09) Another subgroup analysis that showed a significant survival advantage were the MGMT methylated patients. In this subgroup the median survival was 30 months. What I thought was particularly interesting was that when these patients got to about 3 years the majority of these patients continued to live on, not just continued to survive, did not have progression. It could be that in this subgroup of patients there is a % of these patients, roughly 20%, that do have a significant long-term survival advantage with DC vaccination. And why it’s more effective in MGMT methylated patients we don’t know; MGMT methylation might be a surrogate marker for something we have yet to discover. There have been some reports that suggests that methylated tumors are more hyper-mutated so they may have mutations to induce an immune response.
(Slide 22, 26:27) These are the relative percentages of long-term survivors in the overall DC vaccinated patients. At the end of the day it turned out there was not a clear cut prognostic advantage to some of these prognostic indicators particularly as related to age or IDH mutation. With MGMT methylation you can see there is a greater extent of long term survivors in the MGMT methylated group.
(slide 23, 27:07) So, in conclusion, personalized autologous tumor lysate-pulsed cellular vaccines such as DCVax appears to be safe with minimal toxicity, and was feasible to administer in >90 sites internationally. In nGBM there was a statistically significant difference in median survival in both the nGBM and in the rGBM patients. Again, this was compared to external control populations, this was not a randomized controlled trial although the external control populations were matched as well as we could in this situation. It’s probably the best we could have done given the data that we had.
(slide 24, 27:59) Long term survival, I think this is even more interesting, it was significantly increased in our GBM patients who received DCVax; and it wasn’t just 13% long term survival, it was actually longer term survival without progression, which actually was quite interesting. Again there seems to be a significant long term survival tail, more than 5 years without recurrence in the MGMT methylated patients.
(slide 25, 28:34) So, I think that data, it is what it is. It is level 2 data, it is compared to an external control cohort, and in terms of where it will go next, that’s still yet to be determined. But I do think that it is certainly a first step to really, hopefully getting to more significant longer term survival in GBM patients. This was actually a review written by these authors a year or so ago, and it shows the potential role of dendritic cell vaccination in combination with various different modalities. This is really where they have the most power in terms of future place in the treatment of GBM patients. Because one thing that we know that dendritic cell vaccination does is that it gets T-cells in the tumor, and although that’s sometimes not sufficient, it’s necessary. So that’s a necessary first step. Once the T-cells are in, there is, and I won’t go into that data today, but there is there is a micro-environment within glioblastomas that could actually deter an effective immune response, and that may have to do with checkpoint inhibition, some has to do with this population of immunosuppressive myeloid cells that come in, so there are different ways that we can modulate that with other agents, and that’s something we are looking at at UCLA, we’re looking at dendritic cell vaccination in combination with co-stimulatory molecules and checkpoint inhibitors. We are also looking at collaborations to look at CAR T cells in conjunction with DC vaccination and other types of protocols. This is just some preliminary unpublished data on our trial at UCLA whereby we are combining dendritic cell vaccination with PD-1 inhibition, and what we are showing is that…interestingly, when we give PD-1 inhibition neoadjuvantly – before vaccination – you actually get a survival curve that is better than just giving PD-1 alone, but if you give PD-1 inhibitor after dendritic cell vaccination you can boost that survival rate up to greater than 50%. Again this data is very preliminary, we are only half way through this trial so we don’t have final data yet but I think it’s encouraging in the fact that we probably can combine vaccine therapies with immune modulators to hopefully get a positive response for our GBM patients.
(Slide 26, 32:03) So with that I just wanted to thank all of the investigators that were involved in the Phase 3 clinical trial and all of the co-authors on this JAMA-Oncology paper that was published a couple weeks ago. It was really a wonderful group of people that I’ve worked with, and with that I will stop and take any questions.
franknez.com
