The team at Pfizer was deep into preclinical research of a new drug when something strange began to happen that soon became the talk of the lab. The novel molecule in development had passed rodent testing and was now being trialed in beagles. At first, the dogs seemed to tolerate the drug. Then, while sitting in their cages, their jaws suddenly started to click together at random intervals.
Called “canine jaw snapping,” the unexplained side effect spelled doom for the treatment. The development team couldn’t figure out why it was happening and thus, wouldn’t have a sufficient explanation for the U.S. Food and Drug Administration — the gatekeepers for further trials. Ultimately, the molecule was scrapped.
“In truth, this might have been a good compound. It was a franchise I had hoped to continue working on at Pfizer,” explains David Gortler, a former FDA senior executive official who launched his career at Pfizer about 20 years ago as an investigational medicine research scientist. “But, this is what can go wrong when you use animal studies to predict drug safety and efficacy.”
Animal testing has been an integral phase of preclinical trials for every major drug on the market today. The FDA in fact requires that drugs pass through various animal tests before moving into human trials. But as the Pfizer story illustrates, a lot can go wrong with this approach. Aside from the ethical concerns related to animal testing, biological differences can make it an inefficient way of modeling the effect of drugs in humans.
“It’s a heart-wrenching and outdated technology,” Gortler says. “And it’s rarely — if ever — predictive in humans.”
To put it another way: Common estimates show that close to 90 percent of drugs fail in human trials despite showing positive responses in animal tests.
However, the use of animal testing in pharma has persisted — partly because there is no widespread infrastructure for alternatives. But now, with a rise of disruptive technologies, a major shift is underway and new drug development tests are evolving to take their place.
According to analysis released by Research and Markets last year, the global market for animal testing was valued at close to $10.74 billion in 2019 and is expected to grow at a compound annual growth rate (CAGR) of 4.27 percent between now and 2025 — but that growth will slow to 2.46 percent between 2025-2035.
And although the global market for animal testing alternatives was valued at just $1.11 billion in 2019, it is expected to swell by 10.4 percent over the next four years, with a CAGR of nearly 12 percent in the U.S. alone.
Within the field of animal testing alternatives in pharma is a swath of emerging technologies including computational modeling, organ-on-a-chip and advanced cell culture assays. And in addition to adding cost-savings and efficiency to early drug development, these technologies could shake up manufacturing processes in the coming years as well.
Have we reached a point where pharma can finally put an end to animal testing?
The dark side of development
Animals have long been behind the scenes of major breakthroughs in science.
In 1957, a dog named Laika helped pave the way to human space travel after becoming the first living creature to orbit the earth in a Soviet rocket called Sputnik 2.* And in 1996, a sheep named Dolly advanced our understanding of genetics and stem cell biology by becoming the first mammal cloned from another adult sheep’s cell.
The use of animals in medical research dates back to the ancient Greeks, who pioneered animal dissections for anatomical studies. And in pharma, the roots of documented animal testing traces back to the late 19th century when the practice helped pave the way to the creation of critical medicines such as vaccines, anesthetics and antibiotics.
But animal testing didn’t become mandatory until the early 20th century after a series of tragic drug and cosmetic launches. In 1937, a U.S. company released a new cough syrup called Elixir Sulfanilamide, which was dissolved with diethylene glycol — a human toxin. The product, which was not tested in animals before being commercialized, killed over 100 people.
A year later, Congress passed the Food, Drug, and Cosmetic Act, which sought to strengthen the FDA’s grip on the safety of many consumer products and mandated animal testing in new drugs.
Now, new medicines that survive lab testing must then graduate to a series of preclinical trials first focused on safety and dosing in rodents and small animals, then on safety and efficacy in larger mammals, such as primates.
But as the use of animals has grown throughout science, so has the backlash from animal rights activists. Today, animal testing has been phased out of most industries — yet, pharma is one of the last holdouts of required animal modeling. And although animal modeling has long been defended for its role in ushering safe new drugs onto the market, it’s also a moral quandary of drug development that the industry would prefer to avoid.
During our interview, Mike Clements, vice president of marketing at Axion Biosystems, a company that develops microelectrode array (MEA) assay systems, even side-stepped saying the words “animal testing” by referring to it as “traditional modeling” instead.
“I don’t want to be critical. I used animal research as a PhD student,” he admitted. “I think it’s something that is unfortunately necessary, but people don’t want to do it.”
Going forward, the focus has shifted onto how drug developers can use a more reliable model — the human body.
Chipping away at animal testing
In a 2019 TEDx Talk entitled “Why our grandchildren won’t know animal testing,” Manfredi San Germano, a PhD candidate at Imperial College in London, showed the audience the tiny technology making major changes in drug development.
Holding up a chip about the size of a AA battery, San Germano said, “Biochips like these…may seem like pieces of plastic to you…[but] they actually contain entire communities of human cells living together, talking to each other, moving, changing shape, growing…and we can induce dynamic environments.”
Packed with tubes less than a millimeter in diameter that are lined with human cells, organ-on-a-chip (OOC) technologies are able to model how organs function — a lung breathing in and out, or a heart contracting and relaxing — and then show the impact of a new variable, such as a compound.
Research that led to OOC started in the ‘90s, and the first successful model was presented in 2010 when Harvard’s Wyss Institute produced a lung model of the technology. Many have been holding up OOC models as a potential savior from animal testing ever since.
Since then, market researchers have estimated that about 60 organ-on-a-chip companies have sprung up around the world — many focused on modeling single organs. TARA Biosystems, for example, has zeroed in on the heart, where Misti Ushio, the company’s CEO, says safety flags in new molecules are most likely to be raised.
“One of the main reasons drugs fail is because of cardiac safety,” she says.
After developing its heart-on-a-chip model, the company commercialized its tech and now performs safety and efficacy screening tests on molecules in-house for its clients. Along the way, TARA has formed partnerships with about 30 different companies in drug discovery, and has published scientific papers in collaboration with GlaxoSmithKline and Amgen.
Although companies are not currently able to completely replace animals with TARA’s heart-on-a-chip tests, Ushio says the technology is “animal sparing.” It also boasts a range of other benefits, including its ability to generate disease models of heart disease and pump out large quantities of human relevant data on a continuous basis.
“Organ-on-a-chip is not just one workflow point in time,” Ushio says. “You can use it along an entire drug discovery and development continuum.”
OOC technologies like TARA’s can also be customized to model disease biology or different organ functions, offering users the ability to discover new medicines for heart disease, which remains the No. 1 cause of death worldwide.
“We’re looking at the effect of compounds on safety and efficacy by measuring a variety of functional parameters that model the diverse functions of the heart,” Ushio says.
Like others in the OOC game, an important part of TARA’s work with pharma companies involves validating that their tech can produce data as reliable (or better) as data from animal models. For example, Ushio says that TARA has been able to show that, if used instead of animal models, their model would have predicted safety concerns that would not have been apparent until human clinical trials. Having this information would have mitigated safety risks and saved pharma companies on development costs.
“We have seen drugs exhibit safety issues in clinical trials — an outcome our tech would have predicted,” Ushio explains.
Ushio also notes that in the growing landscape of OOC technologies, there are some single-organ systems with minimal biology that can run more compounds at a lower cost per compound. TARA, however, has strived to engineer heart models with the most integrated human biology to offer more “richness” in content, increasing the likelihood of predicting clinical results.
“I think that’s something to keep in mind,” Ushio says of pharma companies looking to implement OOC options. “It is about matching the right technology to the challenge at hand, balancing throughput, costs and translational relevance.”
For companies looking for even more integrated modeling options, multi-organ human-on-a-chip technologies have also emerged to offer assays that measure the impact of compounds on various systems in the body.
Dubbing itself “The Original Human-on-a-Chip Company,” Florida-based Hesperos has sought to create ever-more complex models that can mimic the functions of two to more than five organs.
“Our thinking is that while a drug could show efficacy in a single-organ, isolated system, it is not going to predict side effects. So you could take that drug into clinical trials and it could fail due to off-target toxicity,” Mike Shuler, CEO of Hesperos, says. “We want to offer a system that helps you make better decisions about which drugs will be both safe and effective in clinical trials.”
Like TARA, Hesperos works like a contract research organization and runs its tests in-house on models that churn out continuous data on the effects of molecules on critical organs, such as the heart and lungs.
“We monitor the functional changes to each organ as treatment enters and metabolizes in the system,” says Nate Post, director of business operations at Hesperos. “For example, we track the acute and chronic changes to key heart functions such as: beat frequency, contractile force, and conduction velocity. These types of changes to organ functions are indicators of a drug’s toxicity — and we are tracking it in real time.”
As the industry discovers more applications for OOC, Shuler says that the technology will be especially insightful for certain indications.
“There are diseases for which the current animal models are irrelevant or simply don’t exist, such as rare diseases and the immune system,” Shuler argues. “Chips will play a bigger role in that development.”
So far, Shuler says that at least one company has submitted an approval application to the FDA using data from Hesperos. Although the company — Bioverativ, which was bought by Sanofi in 2018 — wasn’t able to completely replace animal models in the development of its drug, it supplemented its testing with a Hesperos model that recreated a functional neuro-muscular junction, then damaged it and treated it with the new therapy.
These kinds of milestones in pharma and regulatory acceptance are going to be critical to wider adoption of the nascent OOC industry. Shuler argues that in many cases OOC offers more accurate data than animal trials, but acknowledges that the company is still “developing market trust in the technology” in pharma.
And although OOC tests can still take months to develop and complete, both TARA and Hesperos say that it is a cheaper option — especially if it helps pharma companies avoid ill-fated development investments.
“If you can have a higher success rate and it is now a 1 in 4 chance of success, versus 1 in 10 — that’s potentially a multibillion-dollar improvement,” Shuler says.
Neither Hesperos nor TARA publicly disclose sales figures, but both said that sales have been at least doubling year over year — a clear indication that the shift to these new models is truly underway.
“It’s the beginning of the next generation of what drug development and discovery is going to look like,” Post says.
Advancing cell analysis
Multielectrode array (MEA) assays, which measure electrical signals from live cells, have been around for decades. But in recent years, MEA system companies such as Axion Biosystems have been innovating testing devices that allow companies to perform increasingly complex cell cultures for disease modeling and drug screening.
Rather than traditional systems, which use lights or dyes, Clements says that Axion embeds “microscopic electrodes that act like antennas that can sense changes in the cell.” Axion’s multiwell models are also high throughput systems that can track the real-time effects of more cell samples in less time. Importantly, Clements says that Axion’s systems are also easier to learn and use.
“What makes us unique is that we have created a product where the end-user doesn’t need to be an expert in electrophysiology,” he explains.
Clements says that these kinds of assays, which can use human-induced pluripotent stem cells (iPSCs) with disease mutations to measure the impact of new molecules, are going to be fundamental models in the efforts to replace animal tests.
“[Our tests] have weeded out drugs that didn’t need to go into animal testing,” he says. “So if we can improve the success of the compounds that go into full animal models, that will ultimately mean fewer animals used in research.”
After Axion commercialized its first multiwell system in 2011, most of its early adopters were in safety and pharmacology at agencies such as the FDA. Now, the company is seeing more use among academic labs and pharma companies. And critically for drug development, the design of the system gives users the flexibility to ask targeted questions.
“That’s very important because a lot of pharmacology isn’t just interested in seeing if a compound is going to destroy the brain tissue itself — they want to know if it’s going to induce a seizure,” Clements says. “Our users are coming up with ways to make our ever-more intricate systems answer more complex questions.”
Upgraded humans
It was really just a matter of time before some company figured out how to conduct a clinical trial on a computer with just the click of a button. As it turns out, that company was Israel-based CytoReason.
Using the massive deluge of clinical trial data in the public domain, CytoReason has become a pioneer in the field of machine learning platforms that can simulate human systems on a cellular level with a computational model. Tissue by tissue, disease by disease, the company creates AI-driven discovery and development models that churn out biological data on illnesses in a matter of minutes.
According to David Harel, the company’s co-founder and CEO, the platform helps overcome one of the most limiting factors of animal tests.
“Animal testing in preclinical work is limiting the types of drugs that can be developed,” he says. “Our models are built on human data that can simulate clinical trials in significant and more accurate ways, and that opens up the ability to use and manufacture molecules that were considered too risky because they could not be validated in animals.”
Harel says that typically, CytoReason’s computational models, which have been implemented by several Big Pharma companies, are used in early discovery to better understand disease biology for targeted drug development.
Although the system is focused on predicting efficacy outcomes, Harel says it can provide some safety data. So far, the company’s core platform includes models for over 180 diseases and it can be customized for specific use cases, accounting for patient stratification, indication selections or genetic variables.
The cost to implement CytoReason’s platform varies widely and depends on factors such as how much of the body the company wants to model. But Harel says speed is a prime selling point.
“Every animal trial could cost $80-100,00 and take 18 months,” he says. “If you do it with human data, once you have the computational disease model set up, the marginal cost will be zero and it’ll take you 20 minutes.”
While there are some limitations — because the system is based on available data from previous clinical trials, it works best when measuring outcomes observed in live patients — ultimately, Harel says that platforms like CytoReason’s are playing an increasingly important role in reshaping what the landscape of early discovery and development will look like in the future.
“All of these assays are important,” Harel says. “They allow you to look at [drug development] from every angle and then you can aggregate all of it to get one full picture.”
Manufacturing implications
Because the industry for many of the emerging drug development technologies is so new, their impact on manufacturing has yet to be realized. But companies in the space are envisioning a number of ways their tests could one day be used in downstream processes.
Precision medicine
One common prediction about the application of OOC technologies in manufacturing is the impact they could have on developing drugs for subpopulations.
“One of the main reasons drugs fail in clinical trials is because they work fine for 95 percent of the population but for 5 percent, they don’t,” Shuler says. “But you could build these chips to represent different parts of the population to test.”
OOC could also aid in creating personalized drugs — a growing area of pharma investment.
“Patients could supply you with their cells and you could build a model off of them,” Shuler says, noting that having a piece of liver tissue, for example, could show how fast an individual will metabolize a drug — something that varies widely across the population and has a direct effect on efficacy for that patient.
According to Clements, MEA assays could also be used to inform development for drugs based on genetic variables.
“For example, we’ve seen research on pediatric epilepsy…which is not caused by one gene,” Clements explains. “So the researcher was looking at cells on a dish from different patients to study phenotypes and predict which compound could act as a potential therapeutic.”
An example of organ-on-a-chip technology.
Shuler admits, however, that because of the time it would currently take to build personalized OOC models, its application would be limited to long-term care scenarios.
“The idea of personalized medicine is great but it is probably going to take a while to be effective because of the time between taking the cells and putting them into the device,” he explains. “So it wouldn’t work well for a drug you need in a few days.”
Supply chain and scale-up considerations
According to Harel, companies are leveraging CytoReason’s platform during critical decision-making points in phase 2 trials that could have ramifications for commercialization.
“A company may try a drug for certain types of cancer, for example, but then think about extending the label to other indications,” Harel says. “Those decisions change the commercial dynamics.”
In particular, Harel says that the platform helps companies make these decisions quickly so that they can assess supply chain considerations related to commercialization.
Shuler says that OOC testing could also have a major impact on product testing during scale-up when process changes could alter the drug.
“The product is defined by the process by which it’s made,” he argues. “In scale-up, there is always the danger that a product could become toxic because of side effects that were introduced along the way.”
By continually testing the product during scale-up, Shuler says that companies can ensure that the drug is still behaving the same way it did during clinical trials.
“I do see a way to combine these technologies with the manufacturing process to assess quality control and ensure that the product is still going to be effective and not cause toxicities,” he says.
Release criteria
In the burgeoning field of cell-based therapies, Clements says that cell testing could play a role in determining how well manufactured therapies might work.
“Companies are generating CAR-T cell therapies, for example, and they might want to test these cells to see if they’ve produced enough of them with the right viability…to make sure they are working before they’re given to the patient,” he explains.
Currently, Clements says that as far as he knows, regulators don’t mandate this kind of release criteria. But as cell testing and manufacturing becomes more sophisticated, that could change.
“It makes sense that if you’ve made millions of cells in a bioreactor that you would want to know if you’ve made them with the properties you want,” he says. “That is something the field is very interested in.”
Waiting on regulators
While serving as a senior advisor to FDA Commissioner Stephen Hahn between 2019-2021, Gortler says he was spearheading an effort to pivot the agency towards the acceptance of non-animal models.
“Trump’s whole thing was, ‘Let’s speed up drug development,’” Gortler explains. “And I thought, ‘Let’s do that and get rid of or reduce the torturing of animals.’ It’s a no-brainer.”
With the White House’s blessing, Gortler says he took the initiative to research alternative technologies, such as OOC, and was pushing the agency towards policies that would reward companies for using them. Then after the Biden administration took over in January, Gortler — a Trump appointee — was forced to resign, leaving his efforts to create incentives for non-animal models unfinished.
But there are signs the agency is opening its arms to alternatives. In October, the FDA signed a collaborative agreement with Emulate, a maker of OCC technologies, to test the safety, efficacy and mechanisms of drugs regulated by the agency on their models.
And in December, the FDA launched a pilot program called the Innovative Science and Technology Approaches for New Drugs (ISTAND), which lays out a three-step process for biotech companies to qualify new testing technologies.
Meanwhile, companies are continuing to submit data from new models alongside animal data to help validate them as equals.
“There is an enormous push to create and validate these models,” Michael Graziano, chief scientific officer of TARA, says. “What it’s going to take is drug companies bringing parallel data sets forward that show these technologies predict human outcomes as well or better than animal models. This is the path that will help replace one technology with another.”
Are we there yet?
Although there is little debate that the process of drug discovery and development is undergoing a transformation that will make it look very different decades from now, there is still some disagreement over how much animals will factor into that future.
OOC technologies have been heralded by many as a replacement for animal models and to a certain extent, are already reducing that need. Yet, Shuler estimates that OOC technologies are still five to 10 years away from being developed enough to replace most animal models and being widely adopted by pharma.
“Animal models are going to be around for a little while,” he says. “But today, nine out of 10 drugs are failing to translate from animals to humans. If we can improve that rate, as we expect using our models, we’re better informing what’s going to work at the end of the rainbow.”
Will animals ever be completely eradicated from pharma drug testing? Unfortunately not, Shuler predicts.
“I think we get rid of 99.9 percent of animal testing,” he says, “While this technology will eventually eliminate the need for animal testing in virtually every use case, there are certain complexities of a living being that won’t fully translate to in vitro models.”
For now, Ushio says that any reduction in animal tests is a step in the right direction.
“You don’t have to go to animals to generate reliable data,” Ushio says. “Having a human relevant model increases your probability of success. That’s what matters at the end of the day.”
