A lot of what follows is very technical but I wanted to get it all written in one place. I hope patients actually read and re-read this material several times. For people who are prone to sleeping every time they read one of my posts, here are two videos I did with Brian Koffman in Sept 2013 that goes over the same material in video format:
Part 1: New Prognostic Markers
Part 2: Another on New Prognostic Markers
I've been wanting to write this post for a while but a recent paper has really brought this to the forefront of management of our CLL patients. Unfortunately the names are strange and I worry this post may fall toward the technical side - sorry. I will create a separate post that specifically defines many of these terms.
Integrated mutational and cytogenetic analysis identifies new prognostic subgroups in chronic lymphocytic leukemia
For people who have read all my posts on FISH testing, you are probably aware that it is an antiquated technology that has served us well for 20 years but needs desperately to be replaced. Sequencing technology has advanced incredibly quickly and is now poised to refine our understanding of CLL risk groups with new molecular detail.
While most patients are aware of the incredible advances in CLL therapies (ibrutinib, CAL-101, GA-101, ABT-199), fewer are aware of the really important advances in molecular markers that have been recently discovered. Once these are rolled out to the general public we will be able to understand with much more precision how a patients disease will behave. Pretty soon, folks will not only be talking about 13q, 17p without also talking about BIRC3, SF3B1, and NOTCH.
In the last 24 months, genomic sequencing has been applied to cases of CLL with pretty remarkable results (see New England Journal of Medicine article or Journal of Experimental Medicine article)
Several key findings have emerged from these data sets.
1) CLL has a relatively simple genome. While some "smart cancers" (cancers that quickly gain resistance to our treatments and are far more aggressive) like small cell lung cancer may have 50,000 mutations per tumor, CLL (a comparatively dumb cancer - which is typically slow, responds well to most treatments, does not gain resistance all that fast) may have fewer than 100 mutations per case and only a small fraction of those (around 10-20) affect important proteins (the enzymes that make all things happen inside a cell).
2) Certain mutations seem to be observed fairly commonly in CLL and have some defined prognostic or predictive value. For instance BIRC3 turns out to be a really bad thing to have - it is the new 17p. NOTCH probably is one way to get to Richter's and helps sort out the trisomy 12 cases, SF3B1 makes you resistant to fludarabine chemotherapy.
3) Certain mutations are seen early in the disease, while others seem to accumulate with time. Furthermore, some of the ones present later on are actually present early but only emerge through "clonal selection."
4) Some cases of "familial CLL" (ie those cases that run in families) have an unifying genomic explanation that point toward things we already knew were important.
With all of this new information, it was only a matter of time before someone took on the herculean effort to figure out which of these were most important and what they all mean when you analyze them simultaneously in a large group of patients (1300 of them to make this model).
The old risk groups were:
High risk: 17p changes (home of the p53 protein)
Intermediate risk: 11q changes
Low risk; normal cytogenetics & trisomy 12
Very Low Risk: Isolated 13q changes
Unfortunately, there is a lot of biologic diversity that FISH testing misses since it only looks at large chunks of missing or added DNA. Using sequencing technology (think microscope compared to telescope) as an adjunct to FISH we can now help sort all of these out.
The new risk groups
Very high risk: 17p deletions, p53 mutations, or BIRC3 mutations (10 year survival 29%)
High risk: 11q deletions, SF3B1 mutations, NOTCH mutations (10 year survival 37%)
Low Risk: Normal cytogenetics, trisomy 12 (without NOTCH mutations) (10 year survival 57%)
Very low risk: Isolated 13q deletions (10 year survival same as age matched controls).
There are some really interesting observations contained within this.
1) It is not a surprise that 17p deletion and p53 mutation are both really bad - we've known that for a long time. They commonly run together (ie. most 17p deletions also have p53 mutations - but not all cases).
2) BIRC3 is a new kid on the block. It has only been recognized for about 18 months. Turns out it is really bad to have. It confers chemotherapy resistance and is often very discrete from p53 alterations (i.e., if you have one, your probably don't have the other). We've known for a while that p53 doesn't explain all cases of chemotherapy resistance - BIRC3 explains a lot of them.
3) We have known for a while that 11q deletions often associate with bulky lymph nodes, unmutated B-cell receptors, faster growth kinetics, requirement for alkylating drugs (cytoxan, bendamustine). It has often been considered a poor risk feature. SF3B1 and NOTCH are totally new though and we didn't know where these fit in terms of hierarchy. Turns out, they are about equal.
4) Last year the relationship between NOTCH and trisomy 12 was identified. About half of trisomy 12 cases carry a NOTCH mutation - particularly those with unmutated BCR (ie. cases with unmutated BCR and trisomy 12 have high frequency of NOTCH mutations - sorry if this gets confusing). We have been aware that trisomy 12 was a bit of a wild card - some did fine, some did poorly. Turns out that NOTCH mutations can sort the two apart. Those with mutations do worse, those without mutations are now considered "low risk." I am very eager to learn if the new NOTCH antibodies turn into personalized medicines for patients with the NOTCH (or even FBXW7 changes).
5) Our good old friend 13q is still "good risk." The surprise here is that 25% of 13q cases get put into higher risk categories when you do the mutation analysis. They might have an SF3B1 mutation or BIRC3 mutation you would have otherwise never known about. By carving out the bad players, it makes the good group even better. "Matching age controls" does have some limitations because the model is built upon typical CLL cases. There are probably not sufficient number of 42 year olds with 13q in the model to say that they necessarily match their peers.
6) This model holds true no matter when you evaluate a patient. In other words, if clonal evolution occurs and you go from very low risk to high risk by molecular definition - your clinical outcome changes too.
There are some important questions in all of this.
1) The most obvious is - how do I know what I am? Right now - you can't easily tell. There really are not commercial tests to sort this out - I'm trying to make one but seem to running into more walls than doors. If anyone out there wants to finance this idea, let me know!
2) What defines "positive" for mutation? For 17p by FISH we do not define a patient as positive until 20% of their cells are positive. With ultrasensitive testing you may find 0.07% of cells have a BIRC3 mutation. That patient isn't "positive" but I would be very concerned that clone may evolve in the future. Do you therefore do anything different when you choose to treat them?
3) This analysis may miss some of the subtlety of different FISH abnormalities. We already know there are type I and type II deletions on chromosome 13 with different prognostic value. We also know that the overall percent of cells with 11q or 13q makes a difference. This model does not capture that degree of subtlety.
4) Mutated vs unmutated is not included necessarily in this model - I would like to know if it "sub-stratifies" amongst the various different risk groups (although it is more common to see unmutated with 17p and 11q than the 13q cases so perhaps the model was just not big enough to take it all into account)
5) How do these markers hold up in the face of the new drugs. ABT-199, ibrutinib, CAL-101, GA-101 are so remarkable. Will traditional markers hold up in the "new era?" It is important to note that this model is based upon cases that have already been followed for quite a few years. Some didn't get rituxan with their first line of therapy. Presumably none were able to take advantage (since it is an Italian study) of the new drugs. By definition, this is a backwards looking model and does not capture what I see as a very optimistic future. For example, 29% 10 year survival for poor risk does not reflect the impressive durable control obtained in front line 17p patients treated with ibrutinib.
Though there are questions, the authors of this paper are to be thanked profusely for their remarkable effort to create a single predictive model of this magnitude. I would imagine that there were thousands of hours put into creating and analyzing the data. This paper will serve as a landmark for quite a few years and will help guide countless numbers of patients.