New Technologies Spread Precision Medicine to the Community

Mark G. Kris, MD; Christian Rolfo, MD, PhD, MBA; Pasi A. Jänne, MD, PhD


January 18, 2019

Mark G. Kris, MD: Hello. I'm Mark Kris, professor of oncology at Memorial Sloan Kettering Cancer Center (MSKCC). Welcome to Medscape Oncology Insights. Today we are going to talk about the growing role of technology and artificial intelligence in cancer care, with a particular focus on clinical trials and how these advances can hasten the emergence of new drugs that we so desperately need and have enjoyed joining our practice this year. Joining me today in this discussion is Christian Rolfo, director of thoracic medical oncology and director of early clinical trials at the University of Maryland Greenebaum Comprehensive Cancer Center in Baltimore. Also joining us is Pasi Jänne, director of the Lowe Center for Thoracic Oncology at the Dana-Farber Cancer Center.

Both of you spend your life conducting clinical trials. All of the advances we are putting into practice today would not have come about without clinical trials. We also know that trials need to be done more efficiently and faster to get these breakthroughs to our patients. Christian, you chaired a group of doctors today in talking about different types of trials in this new age and how to improve them. What are some of your thoughts?

Trials Being 'Trialed' in Cancer Care

Christian Rolfo, MD, PhD, MBA: Trials are very important and crucial for the final goal of getting access of new drugs to our patients. In order to have more efficient trials and an accelerated system for approval, we need to design new kinds of trials, such as master protocols and umbrella trials.

The ALCHEMIST trial is an example of an umbrella trial. In ALCHEMIST, patients with lung cancer are tested for EGFR and ALK translocations. Patients who are positive receive tyrosine kinase inhibitors (TKIs) that specifically target these two alterations in the adjuvant setting. There was an amendment in the trial to include a third arm for patients who were negative for these biomarkers. And with the new data we have on immunotherapy, patients can receive nivolumab prior to entering this trial. This is a very nice approach because we need an adjuvant setting for these populations.

Another kind of trial is the basket trial. The hypothesis is to test only one drug for different tumor types; that is the idea of agnostic tumors. Advantages of this kind of trial include fast inclusion and an opportunity to test very rare populations, like tropomyosin receptor kinase (TRK), whose frequencies in different tumors are very low. In these trials you can cancel or add another arm very quickly without spending money or time. For early stage, we have trials like Bayesian adaptive designs to quickly identify patients when we know that the mechanism of action is proven. That is obviously an advantage for us because regulatory agencies are looking for this kind of trial. Actually, the US Food and Drug Administration sent a message last week saying that there will be a statement on new trials in order to accelerate them in the United States.

Success When Drugs Are 'On Target'

Kris: The data published this year on larotrectinib are just amazing. When you have an effective drug and a target you can reliably identify, it does not matter where the cancer came from; you get equal results if you are young or old.

Rolfo: This is a success. There is not a lot of medication for patients with tumors like medullary thyroid cancer or children with infantile fibrosarcoma. Now they have a good opportunity to receive these kinds of drugs that have an amazing rate of success. Pasi, you were working in this trial and you know a little more.

Pasi A. Jänne, MD, PhD: It's amazing to be able to cut across different diseases. It does not always work, and I do think that we have to keep that in mind. Sometimes the mutation acts differently depending on the tissue of origin. The TRK example didn't make a case. We also learn from the clinic. The clinic teaches us a little bit about biology, even in the context of the mutation.

Kris: For example, BRAF mutations in lung cancer, melanoma, and colorectal cancer are different.

Rolfo: It's important to include patients in these trials so that we will have the opportunity to learn, and that is essential for advance.

Finding the 'Needle in a Haystack'

Kris: The other amazing thing is that these new targets that lead to very potent therapies are already in our next-generation sequencing (NGS) panels or can be added into the current generation of NGS panels very quickly. So whatever advance that comes can be pushed out to everybody who has access to NGS testing, and that is pretty much everybody. Clinical trials are difficult and take a long time. Sometimes they are difficult in cases where you can have a very specific drug for a very specific genetic abnormality. Pasi, do you have things in place at Dana-Farber to help you find that "needle in a haystack"?

Jänne: We have a system called MatchMiner. Let's say you are doing a trial in BRAF V600E mutant cancers. You can set up a system so that whenever a BRAF V600E mutation comes off the sequencer for any Dana-Farber patient who is being sequenced, it will alert you. Then you, as the investigator, can contact the treating physician and alert them about a trial for that specific alteration for their patients. That is a good process to identify patients, especially for these rare ones. A second strategy is also ongoing whereby you can then input the patient's mutation and ask what the potential trials are. You can not only identify the mutants but then the potential trials. All of these will help facilitate trial enrollment, especially when you have very effective therapies for patients with unique alterations. We don't want to lose that opportunity for our patients.

Kris: In the session that we attended today, a number of people were concerned that these kinds of strategies are not available routinely. I assume that the program you are talking about, like a similar one we have at MSKCC, is internal and something developed by your scientists.

Jänne: Right. It was developed by our scientists. We ultimately want these to be available more widely. But I think we want to make sure that they are working and doing the appropriate things, and we want to think about all of the other things that go into it and refine it. But ultimately these kinds of tools will be useful for the greater community.

Kris: We have a very similar story at our institution. But I discovered that these people are already there. They are serving many roles in an institution and they can take on these kinds of extra projects. So, it's not something that can't happen. Christian, do you have a similar system to try to find people for your clinical research?

Rolfo: We are not having the luck that you have with these two magnificent tools. Some companies are doing these services for free. A Hungarian company called Precision Oncology creates something like an atlas of mutations. You can chat with doctors who have had similar experiences in different subpopulations and you can customize the area where you live to find trials in your area.

Molecular Tumor Boards

Rolfo: I think the essential point here is the interpretation of the genetic data that we receive. We discussed this problem this morning. Nowadays, several companies' platforms are producing genomic determinations that are not matching the right information with the right mutations. Sometimes they are matching some mutations with overexpression, so we need to be very careful. It is crucial that institutions have molecular tumor boards. In our institution, our molecular tumor board helps us interpret different alterations and try to find the correct treatment.

Kris: Molecular tumor boards are becoming more of a standard around the country. I was approached by the Jackson Laboratory in Maine, which develops a lot of the mice used in laboratory research. They have taken on a project for the state of Maine to be an information source for these molecular tumor boards and take the citizens within the state of Maine who have these molecular results. But we lack many things: understanding of the significance of these findings, ability to find the abnormalities, and precision. This exceeds our knowledge of exactly what to do with it. What advice would you give to a real-world oncologist when they get back a mutation they have never heard of that may not have an obvious drug target to go along with it?

Rolfo: My advice is to contact the cancer center that they have nearby in their community. We are starting a strategy of visiting catchment areas and explaining to the oncologists that it's important that we determine and interpret the alterations together. So you may have molecular tumor boards near home. You also have the opportunity to use some tools. The American Society of Clinical Oncology has some molecular tumor boards by Web so you can present your cases there. It's very important that oncologists don't feel like they are alone. We are here to help them with the interpretation. Besides clinical trials, sometimes we have access to compassionate use or to expanded access of some drugs.

Kris: As a community we need to pay more attention to this. Pasi, would you like to comment?

Jänne: I agree with Christian. I get emails all the time from people asking about specific EGFR mutations and I'm always happy to answer them. I think we need to get the information out to our colleagues who may not be at a cancer center. We should help people out because we don't want to miss the opportunity that there could be a very effective therapy for patients based on their tumor genotypes.

Decision Support Tools

Kris: One way of doing this is to couple this kind of information into other ways that doctors gather information and make it more part of their day. I've been involved in the development of a decision support tool. One huge issue is that it needs to be integrated into the daily care of the patients and there needs to be a way to bring in more information without having to seek out that information. I think this is very easy for genetic data. As we are looking at different kinds of information and using it for decision support, mutation is probably about as easy as it gets. It's kind of yes-or-no. There is a finite amount of knowledge out there, particularly for a specific mutation in a specific disease, and martialing that is a lot easier than many other things where you have to make tough decisions. Is it time to change targeted therapy? We all stay up at night trying to figure that one out. But the genetic stuff is helpful, and more and more information comes with genetic testing, be it from blood or serum. Do either of you use a decision support tool in your departments?

Jänne: No, not at the moment.

Kris: Have they considered doing this? Have they looked into vendors that would provide these sorts of things?

Jänne: I think they are being discussed but nothing has been decided.

Kris: I've been involved with IBM Watson, which is part of a collaboration with Watson Oncology and MSKCC. It's a way of bringing in a lot of complicated information and placing it in the doc's hands at the time that they have to make a decision. A huge number of obstacles have to be overcome to make this happen. But it's happening more and more. I think that, in time, all docs will want this. They want easy access to information; they don't want to go to 10 places. They want authoritative information. We have the technology to make that happen, but it's not easy.

For example, simple things like the natural language processing piece [are not easy]. I don't think we realize all of the medical jargon that we use. Let's say that you are reading a physician note electronically and you see 'T2.' What could that mean? It could mean a certain tumor size and staging of the cancer. It could mean the second thoracic vertebral body. And it could also mean the T2-weighted images of MRI, which I've recently learned are very critical in making diagnoses of autism. So a computer sees T2 and it has to be taught that this has nothing to do with autism, it may have nothing to do with damage to bone, and it has everything to do with the size of the tumor and the proper staging of the cancer. This is just one little tiny example of what has to be done. But these are all solvable problems. We all flip to our iPhones and Androids constantly, and these technologies are able to move forward.

There is also a tremendous need for people who are not at a big medical center. The IBM-Watson product has found resonance worldwide in places that are geographically spread out—places where you don't have a tumor board. Some are part of a 20-hospital system and they can't all get together, and because of where they are, they have to treat all different kinds of cancer. They may be treating people who are going to be with them for a short time and then go back to their home country. These clinicians need to have a very precise, defined plan of care, and the computer helps them do that. It has the drugs, schedule, and doses. It can have printouts of the side effects that can quickly be put in the hands of the patient and in the hands of the doctor who is ultimately going to be caring for the patient. There is a lot of potential. But again, we have a way to go.

We have talked about advances in technology that exist, but they do need to be "democratized." The kinds of resources that we have at our centers that really facilitate research need to be put wherever they can be, and the decision support tool is a way of doing that. Commercial companies can help you find clinical trials. is always a reasonable place to start. But all of these things can be done. Just as we have talked about advances in the care of lung cancer patients this year, we have a really bright future in this area as technology is placed into the hands of oncologists, becomes part of routine care, and leads to more benefit for the patient than our current overfocus on completing electronic medical records.

Thank you both for joining me today in this discussion. And thanks to our audience for joining us here on Medscape.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: