Are mouse "avatars" useful for predicting human response to chemotherapy and targeted agents in cancer?

From time to time, I tussle with various animal rights activists online. Over the summer, unfortunately the radical animal rights types, those who think that, at the very least, vandalism is perfectly acceptable in the name of their cause, some of whom think that action up to and including murder of scientists could potentially be morally justified, came to my campus. They also threatened and harassed an excellent scientist whom I know personally and with whom I've served on a committee. That's a bit too close.

Then there are animal rights apologists whom, although they make a point of eschewing violence—and admirably so, as far as they go—I suspect of being much more pals with some of animal rights thugs such as Steven Best and Jerry Vlasak, the latter of whom advocates killing "vivisectionists." My suspicions aside, one thing they do try to do is very much like what antivaccinationists try to do with vaccines, anthropogenic global warming denialists try to do with climate science, and quacks try to do with science-based medicine; i.e., try to spread fear, uncertainty, and doubt (FUD) about it. This generally involves cherry picking data and studies to paint as dim a view of animal research and its usefulness as possible, usually through a combination of exaggerating how bad the animals are treated and claiming that the research using animals doesn't produce any useful results. One variant of this latter ploy is to claim that animal models aren't "predictive" of human responses. Again, any animal models that are predictive are dismissed, and situations where animal models fall short are hyped to the max.

The funny thing is, those of us who actually do animal research acknowledge shortcomings of the various animal models. We have to, because we work with them, and our ultimate goal is to develop treatments that help people. To do that, we use all the tools at our disposal, including biochemical assays, cell culture models, and, of course, animals when appropriate. These were the thoughts running through my mind as I came across this article by Andrew Pollack in the New York Times entitled Seeking Cures, Patients Enlist Mice Stand-Ins. Basically, it's about a trend in science and among patients to use custom, "personalized' mouse xenograft models in order to do "personalized" therapy:

Megan Sykes, a medical researcher, has a mouse with a human immune system — her own. She calls it “Mini-Me.”

There are also mice containing a part of 9-year-old Michael Feeney — a cancerous tumor extracted from his lungs. Researchers have tested various drugs on the mice, hoping to find the treatment that would work best for Michael.

In what could be the ultimate in personalized medicine, animals bearing your disease, or part of your anatomy, can serve as your personal guinea pig, so to speak. Some researchers call them avatars, like the virtual characters in movies and online games.

“The mice allow you the opportunity to test drugs to find out which ones will be efficacious without exposing the patient to toxicity,” said Colin Collins, a professor at the University of British Columbia.

Well, yes and no.

My first reaction to this story was a massive yawn. Scientists have been removing bits of patient's tumors and implanting them into immunosuppressed mice for decades. Such models certainly have their uses and can be moderately predictive of at least initial human response (namely, tumor shrinkage), but they have been less successful at being predictive of what therapies will actually prolong human life or cure disease. It's been fairly hit or miss. So what's different about these mouse models in this story? Two things. First, instead of using a cancer cell line, these tumors are derived from a patient's own tumor. In essence, they are chunks of the patient's tumor implanted under the skin of the appropriate mouse strain; i.e., the strain with the "human" immune system.

As "un-novel" as this all is, what was novel (and to me somewhat alarming) is that companies are apparently already marketing these sorts of animal models to patients. I've written a lot about "personalized medicine" before (just type "personalized medicine" or "targeted therapy" into the search box of this blog), but the vast majority of this blogging has been about the revolutions in genomics that have allowed us to probe the nature of tumors in a way that was never before possible. Moreover, it might very well be that combining the use of genomics to identify specific genetic abnormalities in cancer cells, matching them with targeted drugs, and then testing these drugs in mouse models of the patient's very own tumor could revolutionize cancer therapy. After all, one problem with using just genomics is that the first test of whether the prediction is right is in patients, and using cell culture models has been quite unreliable to predict whether a given patient's tumor will be sensitive to a given drug. In other words, avatars are an interesting research tool that could become very useful over time when combined with the genomics revolution. One example is the Breast Cancer Genome Guided Therapy Study (BEAUTY) project (which has the most god-awful contrived acronym I've seen in a very long time), which is being carried out at the Mayo Clinic's Center for Individualized Medicine.

There's more on this study here:

And here. It's a fascinating, systematic study that is likely to provide very useful information. Basically, the investigators will not only do next generation sequencing (NGS) on patient's tumors before and after chemotherapy, but they will keep the patient's tumor cells alive and grow them in mice to test the predictions that these NGS techniques lead to. The whole thing is likely to be hideously expensive; although the cost of NGS sequencing techniques is plummeting, the cost of keeping a large colony of mice is skyrocketing. That's why, right now, this sort of combination of techniques is (and should probably only be) limited to cutting edge research laboratories. It's not something that just anyone can do, nor has it been validated yet in clinical trials.

Out of curiosity, I wandered over to the website of the company mentioned in the NYT article, Champions Oncology, which markets its test under the name Personalized Champions TumorGrafts™ as "empowering patients and physicians using an in vivo mouse avatar-based diagnostic model that has shown to be predictive of a patients' clinical response to anticancer therapies." TumorGrafts are further described thusly:

A piece of the patient's living tumor is removed during surgery or biopsy and is implanted in mice. By implanting the tumor together with it’s microenvironment, TumorGrafts continue to very closely resemble the patient’s tumor with 94% genetic correlation to the tumor in the patient. Our TumorGrafts successfully grow over 80% of all tumors implanted.

And:

Our scientists work closely with treating physicians to determine which drugs to test on the patient’s TumorGrafts. Our labs will administer these treatments to the TumorGrafted mice and measure the living tumor’s response to each drug regimen. Physicians receive a robust report on the effectiveness of each tested therapy on the TumorGrafts. With this information in hand, physicians can personalize each patient’s cancer treatment.

It sound suspiciously to me like making educated guesses and throwing them at the mouse tumor models. In the article, a boy named Michael Feeney is described. Michael has Ewing's sarcoma, and his parents paid over $25,000 to have a bit of his tumor sent to Champions Oncology for its TumorGraft test. The results came back suggesting gemcitabine, docetaxel, Avastin and Afinitor as a combination of drugs to use, which is described by Michael's oncologist as "not something oncologists would typically choose." It's probably for the same reason that the combinations that Stanislaw Burzynski chooses based on a gene test that he uses are not combinations that oncologists would typically choose because of the potential for synergistic toxicity. In any case, I have to wonder how Champions Oncology came up with the idea of testing that combination in the first place. At least with NGS, there is the guidance of specific mutations uncovered through sequencing. In the case of the BEAUTY study, those mutations will guide testing in mouse avatars. In the case of the TumorGraft test, there are a practically uncountable number of potential drug combinations that can be tested, each taking a certain number of mice. How does one prioritize? I'm surprised it only cost the Feeneys $25,000!

This made me wonder what the evidence base for this test was. I was disturbed by what I found. What I would normally expect to find would be well-designed basic science, animal studies, and clinical trials supporting the hypothesis that using TumorGraft-guided therapy improves response rates or, better yet, survival rates. That is the minimum that I would expect. What I found were preliminary studies, some of which weren't even based on the company's test, and then, under a section called Our Experience was nothing of the sort. What I found was a single case report of a man with pancreatic cancer who survived over five years with pancreatic cancer. Unfortunatley, one case study does not adequate evidence make, and I note that this particular patient also had his tumor genome sequenced, as described in this case report. In other words, this case report is much like the BEAUTY trial than it is like what is described on the company website for what it does. In other words, it's not exactly a fair comparison. Champions narrowed down its choices of therapies to try based on sequencing of the patient's cancer cells, not the way that the test is described. If the company is routinely sequencing patient tumors and then testing combinations in its TumorGraft test, that would be less questionable to me, although I would still consider it way premature to offer such a test to patients. Way premature.

Nor do testimonials, which are the other main type of evidence on the Champions Oncology website. I can't help but notice in these testimonials that there doesn't seem to be any testimonials stating that the results of the TumorGraft test has actually saved a life that wouldn't have been saved using conventional methods. There are, however, testimonials like this:

A combination of world-class advice and the data from the mice is ensuring that my oncologist has far more information about the potential effectiveness of various treatments than would normally be the case. I feel more in control and not a passive victim of my condition.

Or

Working with Champions’ experts has provided my physician, my family and me with a hopeful path to treat my disease... using Champions TumorGrafts™, [they] identified a novel drug combination that has provided promising results and an improved quality of life.

Oh, sure, there are scientific studies listed. They're all fairly small and definitely preliminary. One was carried out by Bayer Schering Pharma AG, which didn't really demonstrate that TumorGraft or TumorGraft-like tests improve patient outcomes. There was, as I pointed out, the aforementioned case report of the man with pancreatic cancer who had a prolonged remission. I'm not saying that the technique of sequencing a patient's tumor and then testing targeted therapies identified by that sequencing doesn't have promise; I am saying that it's not ready for prime time and that offering it to patients before it's been validated (and charging patients for it) is the same thing as charging patients for an experimental test or therapy, which is what TumorGraft is. Even if it turns out to be the be-all and end-all of personalized cancer therapy, charging patients for it right now is in my opinion at best highly dubious and at worst completely wrong. In other words, count me as one of the critics:

Critics are numerous and have a lot to say. The total length of the procedure is long and patients have died waiting for their mice to be done growing their tumors. Sometimes the tumors do not grow at all. Mice can die during the transplant or when undergoing treatment. The tumor may not behave the same in mice. Treatments that are effective in mice may not have the same effect in humans. Critics are unsure whether it will prolong patients' lives and the process is very financially draining; it can cost upwards of tens of thousands of dollars, which insurance does not cover. Skeptics say that a randomized trial would need to occur in order to prove that people with avatars would do better than more traditional methods.

Exactly. Although I now count myself as a critic, I'm a hopeful critic, though. There are studies, for instance, that suggest that the use of these animal models in which patients' own tumors are grown in mice do correlate pretty closely with patient response to the chemotherapy tested. For instance, this study, which is the main study touted by Champions Oncology, involved testing 63 drugs in 232 treatment regimens, and only a few of the tumors were subjected to gene expression profiling. Interestingly, in the conclusion the authors write something that I completely agree with:

The limitations to this approach certainly challenge the broad clinical application of the process and will need to be resolved before this can be first tested in a randomized clinical trial. The process requires large amounts of fresh tumor material and intense resources to generate the tumorgraft. Even in the best conditions, 25 to 30% of implants fail, and those that engraft require 6 to 8 months of additional propagation to be useful for treatment.

This paper was published in 2011 and was done in collaboration with scientists associated with Champions Biotechnology. One wonders why a year later Champions thinks that it's OK to charge patients for this test. Believe it or not, I do think that sequencing cancer genomes and doing expression profiling, then using mouse models like this, could hold considerable promise for predicting individual patient response to different regimens. I just don't think the contention that this approach does has been documented. I also view this as a test that will not be helpful to a lot of cancer patients because it just takes too long and is way more expensive than most insurance plans and governments would be willing to pay for without slam dunk evidence that it does a lot better than what we're doing now. In the NYT article, Dr. Ronnie Morris, the president of Champions, noted that the company has had about 160 patients so far and has tested drugs on mice for 60 of them. However, the other patients "either died too soon, or the tumor did not grow in the mice, or the patients are too new to have reached the drug testing stage."

So in the end, what we have here is an animal model that very well might be predictive of human response in a way that is more direct and individualized. Unfortunately, it's not ready for prime time, and might never be. Personally, I tend to view this sort of test as a better research tool than actual diagnostic and predictive test, given its expense, how labor intensive it is, and how long it takes.