Friday, May 31, 2013

Personalized Medicine - turning drug development upside down!


American Society of Clinical Oncology is meeting in Chicago this weekend.  That means that you will be hearing about a bunch of important developments in cancer medicine. 

I am very fortunate to serve as the medical director for the US Oncology Hematology Research Committee.  Leading the largest network of community practice based research sites in the world gets me access to senior leadership of many of the major pharma companies.  Spending a day with the decision makers at Novartis, Genentech, Celgene, Janssen, and Gilead makes it pretty clear what will be happening in the next several years for patients with CLL/NHL (see posts about why it takes so long and costs so much).

So what is making me so excited?  I think I can now say that drug development is getting turned upside down.  We are entering a new era where therapies are rationally designed and deployed.  This creates hope for treatments that are far more effective and far less toxic.  That "double benefit" is part of the enthusiasm behind drugs like ibrutinib, idelalisib, and the antibody drug conjugates.

If you are not aware of the book, The Emperor of All Maladies I highly recommend you purchase a copy of it and read it cover to cover.  It was written by one of my classmates from residency and was awarded the Pulitzer Prize.  The book details the history of cancer medicine in a way that keeps you awake and burning through the pages.  I gave a 4 part you tube lecture series that is a broad overview of the book that doesn't do justice to Sid's book, but can hit the highlights pretty quickly.  Furthermore, my version is probably a good antidote to the insomnia you are suffering currently that has you reading my blog in the first place.

One theme that comes out clearly through the book and is forcefully validated in all of my pharma meetings is that the old way of developing drugs is running out of steam and a new era is starting to take shape.

So what is the old way?

In the old way, you might take a "compound library" which was merely an inventory of random chemicals.  Sometimes these came from sources like the pacific yew tree or the indian sea hare (not kidding).  You then grew some cancer cells in a laboratory dish and looked to see if you would wipe them out with your chemical of interest.  If that worked, you gave it to mice.  If the mice died from cancer - you abandoned the effort.  If they died from the horrible toxic brew that you just gave them, that was a bad sign too.  If the mice somehow survived the treatment and you saw ANY improvement in the mouse's survival - you took it to a phase I clinical trial.  In the old model, the likelihood of deriving clinical benefit in a phase I trial was less than 5% and many ethicists debated if such trials were fair to the patient.

Provided you could demonstrate that you could get the drug into patients, you then tested it on just about every type of cancer there was.  If you got lucky, most patients survived the treatment and you saw a few cancers shrink.  If that was "successful," you then did a randomized phase three study in a population of patients with a specific type of disease (perhaps you had seen patients with breast cancer respond in earlier studies).  You crossed your finger that the patients treated with drug x did a little better than patients with drug y.  If that difference was big enough, or you proved it in a large enough sample of patients.... you had a drug.

I am sure I have just thoroughly offended a bunch of medicinal chemists and drug developers with my grossly oversimplified version of the story - but that was the general theme.  There is a reason why people dread cancer so much.  For years, we made a bunch of drugs toxic enough to wipe out the cancer cells but often wiped out the patient in the process.  Provided the patient was able to get through the process, they might have been marginally better off for going through it.... feeling miserable the whole time.

The whole process was very empiric.  It wasn't based so much on good science as much as it was the remarkable strength of so many patients who wanted to live a day longer no matter what the cost.  I call it a "drug led process" meaning we had drugs - we didn't necessarily know why they should work, or who they should work in - but give it to enough people and you could often figure it out.

What is the new model?  I think we are entering a "diagnostic led" era.  Some have called this "precision oncology" others have called it personalized medicine.  What it means is that we are leaving the "drug led" process into a "diagnostic led" period.  We are several years into a remarkable transformation in cancer medicine that is only picking up speed.  We are matching drugs the appropriate mutations and doing a vastly better job helping patients live longer - WHILE FEELING BETTER!

The human genome project has unleashed an unbelievable set of discovery tools in cancer medicine.  This gave way to the cancer genome atlas which has systematically categorized the fundamental causes of a bunch of different cancers.  Now you can go to a single webpage like COSMIC (which stands for catalog of somatic mutations in cancer) and look up a single type of cancer and figure out what the important mutated genes are.

What you find is that there are really only a small handful of really bad mutations that cause a bunch of different types of cancer and the same mutations crop up in a bunch of different diseases.  Mutations in RAS, B-RAF, P53, Notch and others keep popping up just about everywhere you look.  In total there are about 100-150 really important targets (out of a genome of 4 billion base pairs).

If you take a disease like lung cancer - just a few years ago we looked into a microscope and said it was either small cell lung cancer or non-small cell lung cancer.  By the time I was in medical school, it was helpful to figure out if the non-small cell lung cancer was adenocarcinoma (gland like) or squamous cell (like the skin).  By todays standards that would be grossly insufficient in many cases.

Today you have to know their RAS mutation status.  If that is negative, you might like to know if their EGFR is normal or not, their ALK translocation status, and maybe even their ROS status.  If they are EGFR mutated, you need to know where the mutation is because exon 19 mutations might be predict sensitivity to erlotinib but T790M mutation mean erlotinib won't work yet afatinib will

This story is evolving right in front of our CLL focused eyes with a new understanding of ibrutinib resistance and a bunch of other unique prognostic markers / mutations.

At the same time a bunch of the new drugs may inhibit several of these important enzymes with a single drug.  Take a drug like gleevec (imatinib).  This drug inhibits the key protein in chronic myelogenous leukemia, but it also inhibits different enzymes that drive the growth of some intestinal cancers, skin cancers, and even rare melanomas.  In fact, the drug is approved in 10 different cancer settings where it is truly revolutionary in most of them.   

So we have two parallel developments going on at the same time.  We can find out what makes a patients cancer actually tick and in many cases we can give a drug unique for their profile.  That is personalized medicine and it is happening now.  Today my organization announced a joint effort with a company called Foundation Medicine to do tumor profiling on thousands of patients in order to get them into appropriate clinical trials.

So why is drug development "upside down?"

The traditional model to get a drug approved was to test it in a bunch of patients with a single disease and hope it worked.  Now we can find patients with specific mutations (regardless of what sort of cancer they actually have), give them specific drugs and the results can be dynamite.... but there is a problem.... a bunch of these mutations are uncommon in a single disease (2-10% of cases) which makes it REALLY HARD to do the sort of clinical trials to prove a point.

Let me give an example.

I was taking care of a patient with multiple myeloma.  He had some of the truly most awful myeloma I've ever seen.  His disease was way off the charts in terms of severity.  He didn't respond to two different types of powerful treatment.  Finally a super potent regimen got him into remission.  He got a stem cell transplant and within two months his disease was exploding once again.  In the process, he fractured several bones, became virtually bedbound and slept 16-18 hours per day. 

I was able to get him tested with one of these fancy diagnostic tests and found he had a specific mutation that is only seen in 2% of all myeloma cases (not going to say which one because the next part of the story gets a little more adventurous than I would typically recommend).  Amazingly there is a drug approved by the FDA for a totally different cancer that targeted his particular mutation.  Even more amazingly, I was able to get his insurance to cover the drug. 

His response was unbelievable.  His blood markers rapidly improved, his transfusion dependence ended, he felt better, he was able to get back to doing home repairs, etc. He survived several extra months, feeling MUCH BETTER (until the cancer came back in his brain where the drug couldn't get to).

So here is the challenge.  An individual patients cancer may have a small and unique number of critical mutations that make the disease grow.  Another patient with the exact same type of disease has a different set of mutations.  Take 100 patients with the same disease and you begin to see patterns to suggest that there may only be 10-20 important mutated genes for a single type of cancer  and 2-4 important ones for a single patient.

The challenge is that they can be in various combinations for different patients - but fortunately a drug from an unrelated type of cancer might work provided the right mutation is there.  Instead of studying a single disease to get a drug approved - we need to study a single mutation in a bunch of different types of cancer.  For a bunch of reasons that is really hard to do but companies like Novartis, Seattle Genetics, and Genentech are starting to figure it out and that makes me really excited.

Now we have tests that can tell us what an individuals profile is and in many cases we have research drugs that might work in their unique profile.  We are no longer in a "drug led" era but we are in a "diagnostics led" era.  We can make a pretty informed decision about what sort of therapy might work in a particular patient - even if we have to borrow a drug from melanoma or kidney cancer to treat a patient with leukemia.

I hope I've made this complicated subject actually make sense.  If I've failed, I encourage you to pick up a copy of "The Emperor of All Maladies" - Sid is far more clear about the subject than I am.

Thanks for reading.
Jeff