The use of animal research in drug toxicity testing doesn’t safeguard humans, argues Pandora Pound, research director at Safer Medicines Trust. So why does medicine still rely on it?
In the spring of 2006, Rob Oldfield was 31 and in perfect health. He had recently returned from an acting course in LA, and a friend of his suggested that participating in a medical trial would be an easy way to make some money.
An American company, Parexel, was offering volunteers $3,680 (£2,000) to test a drug intended to treat a form of leukemia, as well as multiple sclerosis and rheumatoid arthritis, by modulating the immune system. Rob was attracted by the possibility of making a scientific contribution and had gained the impression that the risks were minimal.
The experimental drug, TGN1412, had already undergone extensive tests in animals. In particular, it had been tested in long-tailed macaques and rhesus monkeys because of their relatively close relation to humans and had been found safe at doses 500 times higher than the dose given to the trial volunteers. This was the first time the drug would be tested in humans.
On the ward at Northwick Park Hospital, London, eight healthy men aged 19 to 34 lay on their beds and received their doses 10 minutes apart, watching as the medication slid down the clear tubes and into their veins.
David Oakley, newly engaged and in the middle of planning his wedding, was first. Within minutes of receiving the drug, he had a major headache, followed by severe pain in his lower back. He twisted and turned, trying to find a position that was less painful.
Then, almost immediately after receiving his dose, Rob realized something was terribly wrong. “My whole body went freezing cold and I started shaking,” he recalls. “This wasn’t something you could stop, it was so extreme. It was horrendous.”
Recent graduate Raste Khan, unaware that he was one of two men given a placebo, could only look on in horror at the scene unfolding around him. “It was all manic, everything was happening all at once. They were vomiting, they were screaming in pain, people were fainting, they couldn’t control their bowels . . . it was like a horror movie.”1
The doctor in charge of the trial, Daniel Bradford, recalls the chaos. “They tumbled like dominoes. One man tried to walk to the toilet and collapsed. The wards became chaotic, with blood, vomit, and staff and patients shouting. It was clear which two had been given the placebos.”2
All six men who took the drug were treated for multiple organ failure. All mercifully survived but were told they could suffer long-term disruption to their immune systems. Ryan Wilson, a trainee plumber who was taking part to fund some driving lessons, lost all his toes and the tips of several fingers to gangrene.3
A report on the disaster by the UK’s Medicines and Healthcare Products Regulatory Agency (MHRA) concluded that the serious adverse reactions experienced by the volunteers were the result of an “unpredicted biological action” of TGN1412.4 In other words, the animal studies had given no clue as to how the drug would behave in humans.
Detailed follow-up studies by the National Institute for Biological Standards and Control found there was a subtle but important difference in the way human and monkey blood cells process the drug, meaning the monkey studies would never have been able to predict the catastrophic reactions suffered by the human volunteers.5
Only 10 years later, a similar story unfolded in northwest France. Once more, six healthy men were taking part in a drug trial. Scientists hoped BIA 10-2474 would treat a range of conditions, including anxiety, chronic pain and neurodegenerative disorders such as Parkinson’s disease. So far, the drug had been tolerated in humans at doses of up to 20 mg; this trial was to assess its safety at a daily dose of 50 mg.
After five days on this dose, one man became ill and was hospitalized with symptoms similar to a stroke. On the sixth day, four of the five other men were hospitalized with similar neurological symptoms, including headache, memory impairment and altered consciousness.
Less than a week later, the first man to become ill lapsed into a coma and died. Two others were left with residual neurological impairment.6 BIA 10-2474 had been tested in extensive animal studies—on mice, rats, dogs and monkeys—and at doses up to 650 times stronger than those given to the human volunteers.
After the disaster, Dr Annelot van Esbroeck, then at Leiden University in the Netherlands, tested the drug on human cells. She and her colleagues discovered it deactivated multiple proteins, disrupting the metabolism of human nerve cells. These effects had not been identified in the animal tests.7
Both the TGN1412 and BA10-2474 disasters took place during phase 1 clinical trials. Also known as “first-in-human” trials, these are conducted with a cautiously small number of healthy human volunteers.
If all goes well in a phase 1 trial, the experimental drug will progress to phase 2 trials, which are conducted with small numbers of patients for whom the drug is intended. Phase 3 trials are conducted with larger patient populations.
If a drug successfully passes through all three phases, the pharmaceutical company can apply to the medicines regulator for the drug to be approved for use in the wider population. In the post-approval stage, sometimes known as phase 4, the safety of the new drug is monitored in the general population.
Prior to these human trials, however, a series of “preclinical studies” is conducted. These may include tests on human cells and tissues, as well as computer-based studies or simulations, but to satisfy the requirements of regulators, they always include animal studies.
Animals are used in this context to study drug toxicity, the argument being that whole living organisms are vital for this purpose, especially for new drugs about which nothing is known. Certainly, for a drug that has never been used before, it is useful to understand whether and how it interferes with a biological organism and at what concentrations.
Yet, while we may gain a perfect understanding of how a new drug behaves in the whole living organism that is a rat, mouse or monkey, when we try to translate this information to humans, we come up against the familiar problem of species differences. The information may translate to humans . . . or it may not. Species differences make it an unreliable and inherently risky process.
Animals are also used to identify which organs might be affected by toxicity and to work out a reasonably safe dose for testing the drug in humans. This is done by determining the dose at which no adverse effects are observed in animals and then converting it to a “human equivalent dose” through a process of scaling according to body surface area.
A safety factor is then applied, which involves dividing the human equivalent dose by a certain figure; the US regulator, the Food and Drug Administration (FDA), for example, uses a default safety factor of 10.8 This safety factor is the only protection against any unexpected toxicities that may arise due to interspecies differences.
“What people don’t realise,” says Kathy Archibald, founder of Safer Medicines Trust (safermedicines.org), “is that we have no idea what a safe starting dose is. We just test it in animals and then reduce it by a certain factor, it’s completely unscientific.”
Disasters in phase 1 trials tend to be newsworthy because they involve healthy volunteers, but things can also go wrong in the later phases, when drugs are tested in their target patients.
A combined phase 1 / phase 2 trial of the drug fialuridine, tested by the US National Institutes of Health in the 1990s as a potential treatment for hepatitis B, resulted in the deaths of five of the 15 patients involved. Two others were saved only by emergency liver transplants.9 Toxicity tests in animals, including a six-month trial in dogs, had given the drug the go-ahead for testing in humans.
Because phase 2 and 3 trials involve patients with health conditions, it’s sometimes difficult to unravel whether adverse reactions are due to the experimental drug or to underlying ill health. But since these phases involve more volunteers, if anything does go wrong, it tends to be on a larger scale. Phase 2 trials can involve up to 100 or so patients, while in phase 3, the drug may be given to hundreds or thousands.
The MHRA collects data on “suspected unexpected serious adverse reactions” (SUSARs) that occur during human trials of medicinal drugs, but these data are not easily accessible. Following a Freedom of Information Act request from the Daily Mirror newspaper in 2014, however, the MHRA revealed that 7,187 clinical trial participants had suffered SUSARs from 2010 to 2014, over 10 percent of whom had died, although it could not be proven that their deaths were directly caused by the experimental drugs.10
Adding to a lack of transparency in this area, researchers are not very good at reporting the adverse effects of drugs when publishing the results of their trials.11 Nevertheless, it is clear that patients receiving experimental drugs frequently suffer serious, and sometimes fatal, adverse reactions, despite the preclinical safety tests.
A brief look at drugs for two very common conditions, stroke and cardiovascular disease, illustrates this point. The experimental stroke drugs diaspirin, enlimomab, selfotel and tirilazad all increased the risk of adverse reactions and death for patients in phase 3 clinical trials. Each of these drugs improved outcomes in animals (e.g., reduced brain injury, improved neurological function) but led to a greater number of serious adverse events and deaths in stroke patients who took the drugs than in control groups.12
And between 1990 and 2012, 63 phase 3 trials of drugs for cardiovascular disease had to be halted. Almost a quarter of them were stopped for safety reasons, including seven that were associated with an increased risk of death.13
One of these was for torcetrapib, a highly anticipated drug that was intended to prevent heart disease and expected to be a major blockbuster. The trial recruited patients at high risk of heart disease and compared torcetrapib with atorvastatin, an existing medication for reducing cholesterol.
By the time the study was terminated, there were 34 more deaths in the torcetrapib group than in the atorvastatin group and a significant increase in major cardiovascular events such as heart attack and stroke.14 Animal studies had suggested the drug would benefit human cardiovascular health.15
There are many such examples across a wide range of different medical fields. Despite animal studies being conducted to safeguard humans, between 17 percent16 and 24 percent17 of the drugs that fail in clinical trials do so due to safety issues.
Even licensed drugs can result in adverse reactions and deaths. Approximately half the drugs withdrawn from the market in Europe and the US are withdrawn due to safety issues.18
Rofecoxib, for example, better known as Vioxx, was approved in the US in 1999 to treat arthritis and other painful conditions.
From 1999 until its withdrawal in 2004, there were an estimated 88,000–140,000 excess cases of serious coronary heart disease in the US, many of which were fatal.19 One study found that people taking the drug were 67 percent more likely to suffer a heart attack in the two weeks after getting their first Vioxx prescription than those who didn’t take it.20
Worldwide, twice as many people were exposed to Vioxx as in the US, meaning the scale of the disaster was enormous. Yet Vioxx had a protective effect on the hearts of mice and other animals.
Troglitazone, approved in 1997 in the US to treat diabetes, provides another example. The drug was withdrawn in 2000 after reports of deaths and severe liver failure that required transplantation. Animal studies hadn’t detected its potential to cause adverse effects in humans, but an international study led by the US Evidence-Based Toxicology Collaboration and the Norwegian Institute of Public Health found that tests on human cells and tissues pointed to an effect on the liver, which could have clearly revealed the hazard.21
Adverse drug reactions (ADRs)—excluding those caused by prescribing errors—are estimated to kill more than 10,000 people in the UK22 and 100,000 in the US each year.23 Indeed, a 1998 study calculated ADRs were between the fourth and sixth leading cause of death in the US.24
Studies conducted since then have not shown any decrease in the burden of ADRs in the US or elsewhere, while many smaller studies suggest the burden continues to grow. A large proportion of hospital admissions are also due to ADRs: 6.5 percent in the UK and 3.6 percent in the rest of Europe.25
In the general population, ADRs may be caused by interactions with other medicines as well as by toxicities not predicted by animal studies. These include rare adverse reactions that are difficult to detect until the drug is taken by a large number of people.
Even if an ADR is rare, when millions of people are taking the drug, significant numbers will be affected. A study of 93 serious human ADRs concluded that only 18 of them (19 percent) could have been detected on the basis of the preceding animal data.26
Frustrated at the absence of a robust, comprehensive evaluation of the use of animals in drug safety testing, Dr Jarrod Bailey, science director at Animal Free Research UK, and his colleagues conducted a series of analyses comparing animal and human toxicity data for over 2,000 drugs.27
They confirmed that the presence of toxicity in animal tests is indeed likely to correlate with the presence of toxicity in humans but that this correlation is neither reliable nor consistent. Importantly, however, they established that an absence of toxicity in animal tests was unable to predict an absence of toxicity in humans.
In other words, if a drug appears safe in animals (i.e., no toxicity is detected), it can nevertheless go on to be toxic in humans. Or as visiting professor in statistical science at Aston University Robert Matthews put it, “If Fido goes paws up, that’s bad news, but if he’s wagging his tail, it means nothing.”
Although some of Bailey’s findings were challenged by a 2017 study of 182 drugs,28 they were subsequently confirmed in analyses conducted by pharmaceutical industry scientists on data sets containing over 3,000 drugs.29
These large-scale analyses verify that the examples given above, of humans being harmed by drugs found safe in animals, are not isolated or unusual but illustrative of the general inability of animal tests to safeguard humans.30
An increasing number of scientists have become disenchanted with traditional, animal-based methods and have turned instead to more direct ways of investigating human disease and developing drugs, such as using human cells in various forms.
Cell lines are permanently established cell cultures that multiply indefinitely as long as they have the appropriate conditions. Because they can be maintained over days or months, they can be useful for studying the effects of drugs.
However, there are quality issues with cell lines. They may diverge genetically over time or lose specific functions when they are transferred from their original culture into fresh growth media for the purpose of propagating the line.
Primary cells are taken straight from a living organ or system, such as blood. While they don’t last as long in culture as cell lines—usually only days or, rarely, weeks—their strength is that they closely represent the organ or system from which they are derived. So, in pharmaceutical research, the expectation is that drugs will work in the same way in primary cells as they do in the whole organ.
Heart cells, then, stand in as a proxy for the heart, making them useful for understanding how drugs affect that organ. By testing drugs on primary human heart cells, scientists have been able to predict how drugs will affect actual human hearts and have shown that human heart cells are superior to canine heart cells in this respect.1
Organoids are more sophisticated 3D cell cultures—microscopic versions of parts of human organs that can self-assemble given the right conditions. They are formed of complex clusters of organ-specific cells, and because they often contain multiple cell types, they allow scientists to study inter-cell communication.
Organoids using “induced pluripotent stem cells” (iPSCs),2 cells that are able to renew themselves and differentiate into different cell types, have proven invaluable in preclinical drug development, enabling scientists to study the progress of disease and the actions of drugs, including toxicity.
A system using multiple organoids, for example, was used to test six drugs that had been recalled due to adverse effects in humans, and for almost all the compounds, it was able to demonstrate toxicity at human-relevant doses that animal studies had failed to detect.3
Organs-on-chips are clear, flexible polymers about the size of a computer memory stick that are lined with living human cells taken from an organ, through which blood, air and nutrients can be pumped. The organ-on-a-chip recreates the unique microenvironment that cells are exposed to within the human body, and, because each chip is crystal clear, researchers can observe what is happening at the cellular and molecular level and extract data for analysis.
Organ chips have been used to retrospectively identify drug toxicities that animal studies failed to detect and to throw light on why some drug trials failed. Liver-on-chip technology, for example, revealed that rezulin, a drug for type 2 diabetes that had caused unexplained liver damage in clinical trials, caused liver stress even at low concentrations and before any damage was visible.4
Computers are increasingly being used to test the safety of drugs. In the US, software has been developed to predict whether new drugs will cause liver injury and to understand the mechanisms that contribute to drug-induced liver injury, or DILI. The aim is to prevent human ADRs, decrease the demand for animal tests, reduce costs and speed up drug development.
The software, known as DILIsym®, predicted that the migraine drugs telcagepant and MK3207 would be toxic to the human liver, a prediction that led to their development being terminated even though animal studies had failed to raise any significant safety concerns. Had only animal studies been used, telcagepant and MK3207 may have gone on to harm humans.
Additionally, DILIsym® predicted that a related drug, ubrogepant, would be relatively safe for the liver.1 This was confirmed in human trials, and ubrogepant was subsequently approved by the FDA without any precautionary labeling regarding liver safety.2 This provides an excellent example of the ability of “in silico modeling,” as it’s known, to predict both the presence and the absence of toxicity, the latter being a particular challenge for animal studies.
Artificial intelligence (AI) is also being employed to predict which pharmaceutical drugs will work safely in humans. While in silico models use data that we provide, AI learns from the data we give it.
Israeli company Quris developed an automated platform to test new drugs on hundreds of organoids (see “The potential of a human cell” above), a platform they call “patients-on-a-chip.” Testing a drug on a single organoid gives only limited information, explains CEO Isaac Bentwich, because different people react differently to the same drug.
For this reason, his company tests thousands of drugs known to be safe or unsafe on “male” and “female” organoids with different genomic makeups. The data generated are used to feed into and continuously retrain the machine learning model.
The scientists at Quris believe that, for AI to learn whether a drug is safe, machine learning models need to routinely run thousands, and eventually millions, of patient-on-a-chip experiments. They anticipate that massive experiments will eventually be possible and at a fraction of the cost of traditional animal and clinical trials.
Adapted from Rat Trap: The Capture of Medicine by Animal Research—and How to Break Free by Pandora Pound, PhD (Troubador, 2023). See safermedicines.org
What do you think? Start a conversation over on the... WDDTY Community
