At ASH I presented a study called validation of measurable residual disease as a surrogate endpoint in acute myeloid leukaemia. This was a HARMONY Alliance project, meaning that we conducted it under the HARMONY umbrella. The background is that we know that for AML we still achieve a pretty poor long-term survival and although we have seen some clinical trials that show progress in overall survival, these studies are very slow and therefore very costly because they rely on overall survival as the primary outcome.

So, as we see with other haematological malignancies like ALL and multiple myeloma, who have a surrogate endpoint, they are able to look earlier if a drug is good for better overall survival, yes or no. They do that with measurable residual disease, or with MRD. So our main question is if this can also be the case for AML.

What was the study design?

We are very fortunate with the FDA that put out criteria for us to rely on to make sure that we produce the data that they want to see in order to regard MRD as a surrogate endpoint. This requires us to do a pooled analysis of several prospective, randomised European trials and we need to harmonise patient-level data for this. So it means that the trial groups had to get the data to us in order for us to do this analysis. So we ended up with almost 2,000 intensively-treated AML patients who, of course, had an MRD assessment at the timepoint of after two cycles of intensive chemotherapy. The MRD could have been assessed by flow cytometry or qPCR for NPM1.

The FDA requires surrogacy on two different levels which we both assessed. First is called the individual level correlation, so this looks at the association on an individual level between MRD and overall survival, meaning if a patient is MRD positive does that relate to a poor outcome in the long term? Then second is the trial of correlation and this determines whether a treatment effect on MRD correlates with a treatment effect on overall survival across the included trials. So if a trial achieves a higher MRD negativity in the treatment arm compared to the placebo arm does that then also correlate to a better overall survival difference in the long term?

What were the key results?

The key results were also on these two levels. First the individual level where MRD negativity was strongly associated with overall survival, with a superior overall survival, of course. This has been shown in other studies as well but the main difference is because we have this patient-level data we are able to do also multivariable analysis, stratifying for or correcting for variables such as age at diagnosis, other mutations or going to transplant. We could see that MRD was an independent predictor of poor survival even when we stratified for all these factors. So that’s that first level.

Then the second level is the trial level which was only focussing on flow cytometry MRD because we had enough data to do that analysis for this modality. We observed that there is a strong relationship between MRD differences and overall survival differences. So if the study drug was able to lower MRD, so more MRD negativity, then that would also correlate to a longer overall survival at the end of the study. There was some caution with this data because the confidence interval was wider than the FDA framework allows the data to be. So the lower bound was below the 0.6 cut-off that they typically use. We were able to do this analysis only for the non-transplanted AML patients, then we could see that the correlation increased and the confidence interval became very tight, meaning that there’s an even stronger relationship. That makes sense because probably the transplant that occurs after MRD assessment can mitigate the negative effect of MRD positivity.

What is the clinical significance of these results?

Also the results can be interpreted on these two different levels. The first level, the patient level, is more for the clinicians who use MRD on a daily basis. It shows that MRD is clearly associated with a worse survival outcome, even when correcting for all these different variables. So that’s really hopeful. So I guess you can use MRD for prognostication of your AML patients.

Then at the second level, the trial level, is more of interest for the regulatory bodies such as the FDA or EMA. Of course they need to validate this and see if they regard MRD as a surrogate endpoint. But the results are promising, so we’ll probably need to discuss with them the results as well and we hope that if they approve MRD as a surrogate endpoint that will allow us to have trials with MRD as an endpoint which will ultimately lead to faster drug approval for AML patients.

This study was only possible because of a lot of people that came together – four European major trial groups with, of course, data that they already had but still they needed to trust HARMONY, trust the Alliance and put this data together. It shows that we can have very good results and beneficial things for our patients but only if we truly come together and work together. That’s one of the powers of Europe and one of the things we should always keep in mind and continue doing.