Will Mouse Research Cure Cancer? Uncovering Scientists' Failures & Faith
HHN: Past studies gave many of us hope that cancer could be cured, because scientists were able to cure cancer in mice. Ultimately, though, the interventions in mice didn’t cure human cancer. How reliable, then, are mouse studies in correlating with human patients? Should we be using them?
TVD: It’s true that traditional mouse models don’t correlate well with human cancer outcomes. The success rate with this type of model is only about 5%. That’s because in these models the mouse cancer doesn’t truly reflect the original cancer in the human.
The traditional models, which are still heavily used, developed from an artificial process: Scientists take a cancer from a human, put it in plastic,"The success rate with the traditional mouse models is only about 5%."grow whatever cells grow out of an artificial media, created a culture or group of cells, and then grow tumors under the skin of immune-compromised mice. Although these models were the only ones available for a long time, it turns out that those cancers did not mimic the process of human cancer. This is because significant changes occur when cells are cultured and the cancers don’t start or grow at the site of the original cancers. In short, this environment does not allow the cancers to replicate in all their complexity of function, gene expression, movement, metabolism, and proliferation.
A key problem was the matrix within the cancer. Cancer is comprised not just of cancer cells, but also of non-cancerous cells in the surrounding tissue matrix,"That's because in these models the mouse cancer doesn’t truly reflect the original cancer in the human."and there’s a complex interaction between the two. We now know that in order for the cancer cells to accurately respond, those surrounding non-cancer cells may also need to be targeted, and today’s developing therapies target the matrix as well. But when the cancer is grown in plastic, only the cancer cells grow, and new matrix is formed only after the cells are put into mice.
As a result, in the traditional mouse models, some cells might be similar to human cancer, but overall, the cancer is so far removed from human cancer that it doesn’t accurately predict efficacy in humans.
However, the good news is, today we have newer models that better replicate the development of a human cancer, as close as a mouse can replicate that.
New Mouse Models
HHN: What new mouse models better replicate the development of human cancer?
TVD: First, scientists are now working with genetically engineered mice—that is, mice in which the exact same mutations that are in human cancers have been engineered into mice to be expressed in the exact same locations."From start to finish, the tumors in genetically-engineered mice develop with the same complexity and the same genetic changes as human cancers."Basically, we can engineer DNA to create an altered genetic sequence that is similar to what happens in human cancers, and then replace the genetic sequence in the mouse by introducing the new one into the mouse embryo in its early stages. From start to finish, these tumors develop with the same complexity and the same genetic changes as human cancers. There are a few nuances that are different in mouse and human, such as in the immune system and drug metabolism, which scientists need to be aware of when they interpret their results, but so long as these differences are known, they can be compensated for easily.
A number of the studies using genetically engineered mice have been successful. The first major success leading to a change in clinical practice was in the study of"A study of a pancreatic endocrine tumor using genetically engineered mice led to a treatment which doubles the time it takes for the disease to progress."a pancreatic endocrine tumor. Treating genetically engineered mice bearing these tumors with the drug Sunitinib (Sutent) proved successful, subsequent human clinical trials based on these preclinical results were also successful, and this led to FDA approval. The treatment doubles the time it takes for the disease to progress, and is now routinely used to treat patients with pancreatic endocrine cancer.
Our second new model type is the patient-derived xenograft (PDX), in which scientists put a piece of a tumor from a human patient directly under the skin of mice in which the immune system has been compromised—the compromised system prevents the tumor’s rejection. PDX models are valuable because cancers grown in mice in this way bear the complex properties of the original patient’s cancer. Because of this, and because PDXs in mice are easier to work with than genetically engineered models, they are becoming widely used, particularly by pharmaceutical firms, in the early stages of therapeutic evaluation.
However, unlike genetically engineered mouse models, the PDXs do not have a functional immune system, and therefore cannot be used for evaluating therapies that trigger the immune system to fight cancers. In addition, the tumors don’t develop in the original location. Still, when we look at those therapies that can be tested using PDXs, often they do predict responses in patients. Since pharmaceutical data is not usually shared, we don’t yet have enough data of our own to say how well or how often.
From Mouse to Human
HHN: When an intervention is successful in mice genetically engineered to mimic human cancer, what percentage of these mice studies translates into success in human studies?
TVD: It’s too early to estimate exact percentages, because this is a new field. The genetically engineered mouse model has been used in earnest for direct evaluation of therapeutic treatments only in the last 5−7 years in the public sector.
What we know to date is based on three kinds of relevant studies:1Post-clinical studies: studies in genetically engineered mice that replicate conditions in which the clinical results are known. Since most therapies fail to show efficacy in patients, these mouse studies generally confirm negative outcomes. The good news is that the few positive outcomes are also generally reconfirmed in well-designed studies.2Parallel studies: pre-clinical studies on genetically engineered mice done in parallel with human clinical studies in order to validate the research model. Having recently completed a review of this published literature, I would estimate that 50% or higher show some correlation with the human outcome, again when the studies are well designed.3Bona fide pre-clinical studies: the interventions are evaluated first in the genetically-engineered mice, and if the results seem promising, a similar human clinical trial follows. Because this field is so new,"Of the few published, well-designed studies using genetically engineered mice, I would estimate the correlation success to human treatment is about 50%."these kinds of studies are just beginning, so there aren’t many cases where final results are known. To my knowledge, the endocrine tumor study we discussed is the only one where an intervention has made it all the way to FDA approval for human treatment. Several other studies, though, are now in Phase II clinical trials in humans, where researchers are studying their effectiveness, and the results look very promising. Of the few published, well-designed studies, I would estimate the correlation success to human treatment is about 50%. That said, I believe this percentage would be even higher if we were starting anew, because in some cases, analyzing the failures showed us why the preclinical work was faulty and led to improvement in the preclinical designs.
HHN: In estimating a 50% success rate in translating most study results from genetically engineered mice into humans, you’ve added the caveat: "well-designed studies." Do many genetically engineered mice studies fail to meet this criterion?
TVD: Yes, these numbers exclude many studies. Due to several constraints, most individual investigator studies are often not well powered and/or well controlled. In an evaluation of these studies published about three years ago, Amgen scientists attempted to replicate genetically engineered mice and other studies published from some of these individual academic community labs. Of the 53 studies they worked on, only six—11%—could be replicated.
HHN: How do you define "well powered" studies, and why don’t many of the mouse studies in the academic community measure up in reproducibility?
TVD: It’s partly a numbers game. In many of the academic studies, 3−5 mice in each treatment and placebo arm are used to test one drug or a combination of drugs. We know, however, that many more mice—a minimum of 10—per arm are needed to get meaningful findings, because, just like there’s variability in human response, there can be variability in mouse responses, and in addition there is experimental variation. That said, in some studies using fewer mice, every mouse responded similarly, and the subsequent human clinical trials had the same result, so it’s possible that fewer mice can dictate the right outcome. But it is impossible to know a priori that such a study will be successful.
HHN: If researchers are aware that a large number of mice are needed for a meaningful outcome, why are they using fewer mice in their studies?
It’s not because they have poor ideas or flawed thinking; the problem is in the execution for reasons not in their control. Doing sufficiently well-structured studies requires financial resources they typically don’t have. Whereas some labs in the academic community are funded well enough to conduct well-powered studies with a good chance for correlation in clinical trials, the majority of labs do not have access to the needed resources.
The Amgen study concluded that other issues affecting the non-reproducibility of results are the researchers not always being blind to the experimental vs. control groups, and their hypotheses not always being reflective of all their data.
The underlying problem is that, in biomedical research, to get grant money to do any studies, academic investigators have to publish hypothesis-driven findings, what’s called basic science discoveries, in a high quality science journal."Often, academic labs don’t have the financial support or infrastructure to enlist the expertise they need to conduct the kinds of studies with a good chance for correlation in human clinical trials."They can’t commit all their resources to a single study, because if that study doesn’t pan out, they won’t have any more money for research. In addition, getting published in high-quality journals is key to obtaining jobs, promotions, or tenure, As C. Glenn Begley (former vice president and global head of Hematology and Oncology Research at Amgen) and Lee M. Ellis (of the University of Texas M.D. Anderson Cancer Center) explain (Nature, March 2012), "Journal editors, medical reviewers, and grant review committees often look for a scientific finding that’s simple, clear, and complete—a ‘perfect' story. It is therefore tempting for investigators to submit selected data sets for publication or even to massage data to fit the underlying hypothesis."
Another issue concerns how technically demanding it is to optimally do this research. It calls upon many different types of expertise, and often the academic labs don’t have the financial support or infrastructure to enlist the expertise they need.
Enough evidence now exists that when these experiments are done well, very positive outcomes can result. People in the academic community realize there’s real power in this, and they’re trying to move the field forward, but right now they just don’t have the dollars to do it. If biomedical research had the funding it needs, many institutions could put in the technology and support to assist investigators, and investigators could execute more well-powered studies to move the field forward.
But unfortunately, the current reality is, given the National Institute of Health’s significantly reduced budget and many essential competing objectives, the NIH must support a broad range of research with the most certain promise to benefit people: keeping major new discoveries coming while at the same time translating existing discoveries into disease prevention and treatments. At a time of limited resources, major new and costly initiatives with uncertain impact simply do not attract funding, even though the potential for both human impact and cost savings is enormous.
HHN: How much NIH biomedical research funding has been lost?
TVD: The federal cutbacks have been huge. Funding has dropped by 27% since 1974 and about 12% over the last decade. And because there is so much more research going on and many more investigators to fund, the funds don’t go as far as they used to. Today we are at an all-time low in the percentage of grants that can be funded and in the resources available for NIH scientists and core missions.
HHN: What politics led to these drastic funding reductions?
TVD: I can give my own perspective, but the problem is likely more complex.
The more recent decline is in part due to the 2013 "sequestration." Two years earlier, a super committee of Democrats and Republicans had been charged with hashing out $1.2 trillion in federal budget cuts, in part to resolve the "fiscal cliff." If they failed, automatic cuts to federal programs important to both parties would go into effect. In the end, the super committee did not reach agreement, the sequestration starting date kept being postponed, and even though President Obama and congressional representatives spoke out against sequestration, the cuts across the political spectrum went into effect.
While there has been a continued push by the directors of the National Institutes of Health and the National Cancer Institute to reinstate the lost funding, no budget increase has been realized.
It’s unfortunate, because in the early years of the 21st century, when the NIH budget was considerably higher, among other advances, we were able to develop the genetically engineered cancer models that can help patients today.
HHN: How did additional funding in that more robust period enable you to develop the genetically engineered models?
TVD: Better funding enabled us to fully validate the value of these models for informing us about how cancer works and to develop technology to produce and study models that best emulate human cancers. Much of what we now know about the complexities of cancers resulted from research in these models during that time. Moreover, the first effective targeted therapies came out of basic studies followed by clinical research at that time. For example, the drug Gleevec©, which works by targeting a molecule that’s commonly mutated in that disease, is now used to treat patients with a leukemia called CML, with remarkable results. Nearly all CML patients respond to it, and there’s a dramatic increase in the time before relapse. This treatment is also now approved to treat additional cancer types.
In another targeted therapy, there’s a subset of breast cancers that are driven by a particular protein (Her2). The drug Herceptin©, which inhibits its activity, is helping patients whose cancers contain high levels of this protein. The drug improves overall survival and, in early stage cancers, reduces the rate of relapse.
In addition to these successes, seminal studies on lung cancer during this period led to a vital realization: During therapeutic treatment, not all cancers of a certain type respond the same way. This recognition solidified the need to screen patients to determine the required treatment, a practice referred to as "personalized" or "precision" medicine."Seminal studies on lung cancer led to a vital realization: During therapeutic treatment, not all cancers of a certain type respond the same way. This opened up the whole field of using targeted therapy for cancers based on molecular characteristics."The backstory is that basic research had shown that a particular protein (epidermal growth factor receptor or EGFR) was activated in some lung cancers and could fuel lung cancer in genetically engineered mice. The studies that reviewed the clinical trials of a drug (called Erlotinib© or Tarceva (c)) targeting that particular factor, which at first had not appeared to be successful, actually showed that the drug was effective, but only in about 10% of lung cancer patients—specifically in those cases where the lung cancers uniquely had the same activated factor. Now, screening of lung cancers for such mutations prior to therapy is routine practice. Moreover, as a result, scientists have begun to assess lung cancers, the most prevalent cancers in the U.S., not just in terms of what they look like, but also by what their larger molecular landscape looks like. This has really opened up the whole field of using targeted therapy for cancers based on molecular characteristics.
In short, without the ample funding of years past, the whole cancer field would be in the dark ages, so to speak.
What cutting-edge treatments is your lab helping to develop now using genetically engineered mice?
Working with a biotech company called Clovis Oncology, the center I direct, the Center for Advanced Preclinical Research, has contributed to the evaluation of a treatment for a specific kind of lung cancer: EGFR-positive lung cancers. The drug, CO-1686, or Rociletinib©, appears to be very promising. It’s doing very well now in clinical trials, showing manageable side effects, and tackling many of the cancers that came back resistant after the first round of treatment by the drugs we discussed earlier. My center did some of the preclinical work with genetically engineered mice, particularly showing that the compound could be used both as a first-line treatment and for treating these resistant cancers. We also have numerous ongoing projects in collaboration with companies, foundations, academic and NIH labs, and clinicians that hold promise to positively impact the field.
Perhaps the Pig
Earlier, you said that the genetically engineered cancers in mice "really do replicate the development of a human cancer, as close as a mouse can replicate that." Although the mouse field is still new and very promising, is it possible for us to get closer to replicating cancer in humans by genetically engineering a different animal?
Yes. We’re not going to be perfect in identifying the drugs that work best in humans by working with mice, so large animal models that are more similar to humans will be valuable for a final checkpoint before human research. For example,"Some research is now developing in pigs to create and evaluate human disease models."some research is now developing in pigs to create and evaluate human disease models, as well as to generate cancer models which will be available for the future evaluation of promising drug candidates. The pig’s biology is similar to humans in many respects, and you can engineer pigs like you can engineer mice. For example, genetically engineered pig models of diabetes and cystic fibrosis develop diseases that are far more similar to human diseases than the comparable mouse models.
That said, large animal models will not be useful for sifting through the huge number of potentially promising therapies. A beauty of using mice for research is their short gestation phase, about 20 days from pregnancy to birth, which provides scientists with many mouse subjects—10−12 in each litter when conditions are good—over a short period of time in order to do well-powered studies. In contrast, a pig’s gestation period is 112 — 115 days on average, about five times' longer than the mouse, and litter sizes are smaller. As a result, the mouse will always be used in early stages of drug development where many therapies need to be tested. Once a few promising candidates are identified in mice, however, they can be examined in swine models before they go into patients. Yet, it will take a considerable amount of time before we know whether this strategy will be fruitful, or even possible.
Let me add that the swine, or any large animal model, is unlikely to be used by pharmaceutical companies under current FDA requirements, since only conventional testing in mice is needed for cancer drugs before proceeding with human studies. Drug developers will want to move forward with their compound as quickly as possible from mouse to human trials. So, any real progress in routinely using any animal models to improve the success rate of cancer drug development will take changes in the FDA requirements
Why Conventional Testing Continues
HHN: Why does the FDA require conventional testing on mice when, as you noted earlier, there’s only about a 5% success rate between those studies and the ones on human patients?
TVD: The FDA currently requires conventional testing in mice solely for safety and not for effectiveness. However, as the newer mice models begin to consistently better predict outcomes in patients, the hope is that there will eventually be requirements for efficacy testing in these cancer models as well.
HHN: You also said that the traditional models are still widely used in cancer research. Beyond the FDA requirement, where and why are these models still prevalent when the success rate is so poor?
TVD: The traditional models, known as standard xenograft models, remain widely employed by public and private sector scientists alike. That’s because, first, these models are the easiest to work with. Second, significant expertise is required to effectively utilize the newer models,"The traditional mouse models remain widely employed by public and private sector scientists alike."and not all the labs used to working with the older models have the needed proficiency to change course. Third, accomplishing a change of the magnitude needed to transform the entire research process will probably require a successful and fully established replacement model, and as we discussed earlier, this is not yet the case with the new models, since the field of predictive preclinical science for cancer is still developing. Fourth, the older models can still be helpful in evaluating drug toxicity and in generating a "short list" of potentially effective drugs to be tested in GEM and/or PDX models. As such, these older models may continue to be used in drug development even after a future large-scale transition to the new models.
But, right now, once again, we come to the problem of limited resources. There needs to be a significant investment in team-based science in order to develop and incorporate new standard operating procedures for drug development.
That said, on the positive side, there is major movement to use the newer, more replicable models in cancer drug development, and I expect this to build steam as we refine and establish them.
Beyond Failures to Faith
HHN: Overall, how do you feel about the progress and promise of cancer treatment in the future utilizing all of the new therapies and paradigms?
TVD: There is a lot of good news.
Our understanding of cancer is so much better than it used to be. We now know that:1Cancers are far more complex than we ever imagined, including a complex interaction between the cancer cells and the surrounding matrix. As a result, new intervention studies target the matrix as well.2Cancers work differently in different people. Therefore, studies need to enlist larger numbers of subjects in order to account for variability and determine the right, targeted treatment for each patient. There is significant awareness and associated progress in this critical area.3Well-designed studies working with both genetically engineered mice and PDX models are showing promising reproducibility with humans. Early studies indicate that drug identification could reach about a 50% or more success rate, a significant improvement over the current success rate of 5% using conventional mouse models. Moreover, both of the new models are already aiding development of tests for early diagnosis of aggressive cancers, such as pancreatic, ovarian, and lung cancers, by identifying early-detection blood markers.
Now, a growing number of experts in the field—in modeling, pharmaceuticals, clinical research and treatment—are applying this new knowledge in their work.
As all this comes together, I believe that we will be able to cure some cancers."At the very least, we should be able to turn those cancers we cannot cure into chronic and manageable diseases, like diabetes."And, at the very least, we should be able to turn those cancers we cannot cure into chronic and manageable diseases, like diabetes—basically by keeping the cancers at bay using drug "cocktails" and by knocking the cancers back with a toolbox full of effective drugs if and when they reemerge.
There’s no saying whether there will or won’t be a magic bullet, but with all we have going for us, we can expect considerably more success in cancer treatment in the years ahead. The time it will take, of course, depends mostly on the availability of national resources.Curing cancer with mice, not humans. Behind scientists’ failures and faith. Click To Tweetby