It was my privilege to present the results of IMpower010 at ASCO 2021. IMpower010 was a randomized phase III study looking at giving adjuvant atezolizumab to patients who had completed surgical resection for early stage lung cancer and chemotherapy, and then were randomized to either receive atezolizumab or be followed as per standard of care.
1,005 patients were randomized and we had designed the study to look at the endpoints in a hierarchical way, looking to see if PDL1 was important or not. The way things were set up was the first group we looked at, where patients who had PDL1 expression on their tumors and stage II to IIIA disease. I should mention that over half the patients had PDL1 expression, 55%, and the vast majority had stage II to IIIA. Only 12% of patients had stage IB.
So the first endpoint we looked at was this group with PDL1 expression, and the hazard ratio for disease-free survival was 0.66, so statistically significant. The next planned analysis was to look at all comers, regardless of PDL1 expression. There, the hazard ratio was 0.79, and with that group, we were able to do analysis by PDL1 expression.
In the study design, we had this preplanned greater than 1% PDL1 expression, and I'll note this was the SP263, so one of the tissue-based essays, tumor-based essays. And the other specified PDL1 expression point was greater than 50%.
So in this group of patients with all comers with stage II to IIIA disease, which, again, is almost 90% of patients enrolled on the trial, we were able to show that, for those with the highest PDL1 expression, which was greater than 50%, their disease-free survival hazard ratio was 0.43, so pretty striking.
For all comers with greater than 1%, It was 0.66. But when we looked at those who did not have PDL1 expression, it was 0.97 and not significant. So we were really focusing on the fact that we seem to have a biomarker with PDL1 in this adjuvant trial. And, of course, when we look at PDL1 as a biomarker across multiple studies in the metastatic setting, sometimes it's an important biomarker and sometimes it's less so, especially when we're looking at combinations with chemotherapy.
So we weren't really sure how things were going to pan out in this trial, and to me, it's really important to point out that the PDL1 did seem to correlate to help us select which patients are most likely to benefit. We're still waiting to see what the overall intention to treat endpoint will be when we bring in the stage IB patients, but we know that when you're talking about a disease-free survival hazard ratio, you have to look at what's the probability of an event happening anyway and what's the probability of the event being altered by the treatment.
If you have a group of patients who have 100% chance of being cured, it doesn't matter what you give them. You're not going to get any sort of a hazard ratio. So when we look at the 1B patients, we know it's going to take longer before we get outcomes for them because, fortunately, many of them have been cured with the surgery and chemotherapy alone.
But we are continuing to follow. We're also continuing to follow to look for the overall survival endpoint. I'm hopeful, obviously, we don't know yet, that what we're seeing with these profound disease-free survival benefit from IMpower010 is going to actually translate to overall survival.
But we always have to weigh benefit against the other sides, and the other sides of treatment are going to be toxicity and they're going to be costs. So when we look at the atezolizumab toxicity data, there were patients who had significant toxicity from atezolizumab. Not surprising when we know what the immune checkpoint inhibitors do in metastatic disease and what they've done in earlier stages of disease, such as after chemoradiation for stage III lung cancer and across other diseases.
So about 10% of patients had grade 3 or 4 adverse events with atezolizumab that were felt to be treatment-related. There were other toxicities seen on both arms with the best supportive care and atezolizumab, because these are patients who had had surgery and chemotherapy.
But when we look on to those immune therapy-related ones, many different organ systems could be involved, just as we've seen in metastatic disease. None of them at particularly high frequency, but for the patients who were impacted in that way, it was pretty significant.
So what we'd ideally like to get to is a point where we can meet with a patient who has early stage lung cancer and help them weigh, do they need this treatment or not. And that's where these newer biomarkers, I think, are going to be helpful. Because if we can say, you're done, you don't need any more treatment, that's great.
But for those where we are seeing it, we can say, look, there's a very high chance your cancer is going to come back. With immune checkpoint inhibitor, we know-- with atezolizumab now, that, by taking this treatment, you're going to significantly reduce the risk of this cancer coming back for some period of time. Maybe for good. We don't know that yet. And this is the possibility of having some toxicity. And then we can really weigh, for an individual patient, how meaningful that is.
The cost question is another one. I recently saw an analysis looking at what's the cost of adjuvant osimertinib, and it's pretty high per QALY, versus what the cost for chemotherapy. And that becomes more of a-- like, really a global health question. When you're having that individual patient decision, it's very difficult to bring cost in because how does one value the life of the person right there talking with you?
But we all need to be able to step back and think about that from a societal perspective because if, by treating someone with ajemian immune therapy, which is going to reduce their chance of their cancer coming back by 34%, then is that going to be enough to, say, that avoid giving vaccines to another large group of patients, where those health care dollars are being used here versus there? Those are the big global questions that I'm not going to be able to answer in this presentation, but things that we need to be thinking about from a societal standpoint as we do proceed with approving different treatments.
So to kind of step back from it, with IMpower010, this is the first global trial looking at adjuvant immune therapy in lung cancer. And we hit the disease-free survival endpoint. It seems to be biomarker-driven with the patients whose tumors express PDL1 being the ones who benefit and the patients whose tumors do not not benefiting.
We wait to see how approvals go, but very optimistic that this is going to become an option for our patients. And, of course, we need to wait to see what the overall survival benefit is to know the true benefit. But at this time, I think there's a lot of hope.
As we are awaiting the longer term outcomes from this study, the things that we need to really focus on are understanding the subsets a little bit better. So we know that in the stage II to IIIA with the highest PDL1 expression, that greater than 50%, that the hazard ratio there is 0.43.
We didn't put it in as a prespecified and point to look at what about other cut points on PDL1 expression, but I think that is going to be important as we then get to that decision making with the patient about, well, based on what we know about you and your tumor, what's the probability this treatment is going to help. Because the possibility of harm pretty much stays static regardless of the PDL1 expression. It's just that the ratio changes of the benefit to harm. So we'll need to look at that.
And we obviously need to continue to improve the biomarkers so we know which patients we're helping the most, and the biomarker being the PDL1, but also what about the driver mutations? We did look at that in the subset of patients with nonsquamous histology. And for those with ALK, didn't seem to benefit at all, regardless of whether they had PDL1 expression.
For those with EGFR, there's questions there that need to be further explored. Obviously, the disease-free survival hazard ratio with osimertinib is far better than what we saw with this study. But once we get into overall survival results from those two different studies, there might be subsets where we have to really discuss one versus the other as possibilities.
The other things we'll need to be continuing to look at are, are there differences in patterns of recurrence, are there differences based on the toxicity seen. So there are a lot of questions that still remain to be asked, surgical subsets as well, and whether there were any interactions based on which chemotherapy the patient received. So many questions still to be asked.