IBCD 2016 - Innovation and biomarkers in cancer drug development
Genomic characterisation and risk stratification of AML
Dr Moritz Gerstung - European Bioinformatics Institute, Hinxton, UK
We’ve been briefly talking about research presented here. So I’ve been analysing a large cohort of AML patients where we had complete genetic information and detailed follow-up in the clinic for more than seven years. So that was really one of the first datasets that allowed us to make these detailed associations between mutations that those patients had acquired and their later course in the clinic, whether they were able to go into remission, so their leukaemia vanishes, but also whether it comes back and relapses at later stages. That helped us to train the model, a very detailed statistical model, that associates more than 100 variables which are genetic but also clinical, so comprising blood counts but also patient age, to make very detailed predictions about what the fate of those patients were which were about twice as accurate as current risk stratification procedures.
How was the discussion that followed your talk?
I guess one of the outcomes of our analysis is that it is really worthwhile first building those large patient cohorts and then doing detailed genetic, but potentially also other molecular, profiling on them in order to be able to understand how patients have been progressing clinically. So one of the implications is probably that we need many more patient data in order to be able to make even more precise statements about their fate. So far 1,500 patients, that was a highly selective group from the German/Austrian AML Study Group, if you really want to be sure that those predictions also hold true in other populations around the world then of course we also need to get data from those countries.
Another implication is if we want to apply the same logic to different tumour types then we also need cohorts from other cancers, maybe starting with leukaemias but also solid tumours, in order to be able to apply a similar logic and also see how genetics and outcome are associated.
Do you see patient recruitment as an issue?
Yes, that was what we were discussing outside. There was a notion of a clinical anarchy which is currently ongoing and I think one could actually turn that positive by saying that this could essentially be a new paradigm of a grass-roots large scale randomised clinical trial. But there’s the risk if everyone attempts in administering new treatments that this information is being lost. But I think we should not lose that momentum and the fact that there is the willingness of patients also to participate in these more experimental treatments is a great opportunity but if this data is not being recorded or if it’s just being recorded in a very fragmented way such that we will never be able to analyse which patient benefited and which patient potentially did not benefit then this may all be a wasted effort.
Do you think data sharing is the way forward?
I think this is absolutely what we needed and this is also in line with the recent Blue Ribbon panel report which clearly called in the US for installing a national ecosystem for cancer data sharing with additionally also the potential for patient involvement because really utilising that wish that many patients and also patient advocacy groups express that they can be re-contacted and provide their information into these databases is a great thing and we should really surf that wave rather than trying to duck under it.
Is there anything else you would like to add?
As an additional thing this should not be a national effort alone but I think one need is clearly also to make this international because that won’t just maximise our strengths in re-leveraging the information that comes from those interesting exercises.
