Preclinical models and what they predict
A preclinical result is a result in a model, whether a cell culture, an animal or a computer simulation, and what it predicts for patients depends on how closely that model stands in for the human disease. Each kind of model shows some things and hides others. Animal results have often failed to predict human ones, whether animal testing is needed at all is disputed, and replacements for it are reaching regulators. A report of an animal result has to carry its limits.
Choosing a model against the target population
Obtaining preclinical results with a high degree of generalisability requires models as comparable to the target population as possible, and typically this involves a series of experiments in in vitro, in vivo and, more recently, in silico models. Research for a new drug begins in the laboratory with in vitro experiments, for example on commercially available cell lines, and with animal testing, to answer basic questions and understand diseases. The lack of good models of human disease is among the challenges pharmaceutical companies face in identifying new drugs, together with not knowing the cause and mechanism of many human disorders. Preclinical research is indispensable to protect human subjects in clinical trials, and adequate design of preclinical studies and careful choice of model systems are vital for results that translate to clinical settings.
In vitro, in vivo and in silico models
In vitro studies are a relatively fast, simple and cost-efficient way of preclinical testing. They use cell, tissue and organ cultures, or focus on particular cell components such as proteins and other biological macromolecules. They permit tight control and monitoring of the experimental setting and often provide mechanistic evidence for the compound's mode of action. But isolated cells may not behave in a petri dish as they would in the body, where they take part in crosstalk and interaction with millions of other cells, so more sophisticated models are required to establish the compound's safety profile before it moves to a clinical setting.
In vivo studies consider the complete organism, in various animal models. Animal testing is tightly regulated in most countries, and permission from local ethical review boards is required to ensure that no unnecessary harm is done to the animals. Animal experiments can only be conducted after a harm-benefit analysis and approval by the authorities, and European and American committees have created guides to ensure that an experiment is scientifically, technically and humanely appropriate. Non-invasive imaging, microsampling and telemetric monitoring have refined the use of animal models in drug development. Controlling the experimental setting is far more complicated in vivo, and because the living organism is complex, compounds may behave differently from what results in a test tube predicted. The choice of animal model depends on many criteria: species-specific physiology, similarity in the target organ and in metabolic pathways, and financial, regulatory and ethical considerations. Mice, rats and dogs are among the most frequently used animal models, while testing in primates such as monkeys and apes is done occasionally and typically for larger molecules.
In silico models are based on computer simulations and give information on how a compound might behave in later in vitro and in vivo experiments. Progress in bioinformatics has made them attractive enough that they often precede or complement in vitro and in vivo studies. Apart from the technology, they demand expert knowledge in biochemistry and molecular biology.
Mouse models
The mouse is one of the most popular animal models in pharmaceutical testing, and most research laboratories rely on mouse experiments. Mice have a genome similar to the human one, are relatively cheap, reproduce fast and have a short life span, so they are considered suitable for initial trials in drug discovery. Ninety-nine per cent of mouse genes overlap with human ones, and genomic manipulation in the mouse has become fairly simple. Nevertheless, species-specific differences in host immune response, drug metabolism and tumour heterogeneity affect therapeutic outcomes, and pharmacokinetic and pharmacodynamic differences between species are not negligible, so mouse models often have poor predictive power for clinical efficacy. For lack of superior alternatives, mouse models are the gold standard for testing cancer-targeting drugs.
Transplanting cells, tissue or organs from one species to another is called xenografting. Classical mouse cancer models were limited to transplanting cultured human tumour cells (cell lines) into immunodeficient mice such as nude or severe combined immunodeficiency (SCID) mice. In these cell line-derived xenograft (CDX) models the cancer cells are injected subcutaneously and tumour growth curves are built by measuring the tumour at regular intervals, so treating tumour-bearing mice with a candidate shows its potential to reduce tumour growth, which is its in vivo efficacy. But the cell lines have been passaged under artificial conditions that do not recapitulate the natural tumour microenvironment, so CDX models may lack similarity with human disease. A range of other mouse models has been developed to improve clinical relevance:
- Patient-derived xenograft (PDX) models: tissue from a patient's primary tumour is implanted directly into the animal, omitting in vitro adaptation of the tumour cells, so these models are more similar to human disease in stromal composition and tumour heterogeneity than CDX models. The approach is challenging, but advances in sample retrieval and transplantation technology have made it feasible, and in 2017 PDX models consisted almost exclusively of subcutaneous transplants.
- Orthotopic tumour models: tumours are implanted into their organ of origin to better mimic the microenvironment and recapitulate the metastasis pathways of human tumours, so they are more clinically relevant. Orthotopic PDX models are technically challenging and uncommon, while orthotopic transplantation is widely used for CDX models.
- Genetically engineered mice (GEM): mutations in oncogenes or tumour suppressor genes associated with human malignancies are introduced, so tumours develop orthotopically from initiation through progression in their native microenvironment, recapitulating human tumourigenesis. Preclinical studies in GEM therefore have the potential to provide more relevant data for later clinical trials. Genetic engineering has also given rise to humanised mouse models.
Human-derived cells circulate in humanised mice, but the mice's organs, central nervous system and muscles are not altered. A missing gene, protein or enzyme in an animal model could ruin drug discovery and innovation. Although such models are useful and have contributed to a better understanding of disease mechanisms, science needs further innovation to complement or even substitute animal models to advance drug discovery.
How far animal results predict human ones
Despite all efforts to identify animal models with translational value, drugs often show different pharmacodynamic characteristics in human subjects. That most anti-cancer drugs do not pass efficacy evaluation in phase II and III studies suggests that the preclinical models in use fail to mimic tumour heterogeneity, host factors and drug-resistance mechanisms.
In one 2015 assessment, basing drug development on animal models had produced many more failures than successes, and researchers and regulators have raised concerns about these methods for most major diseases, including cancer, heart failure, stroke, diabetes, and neurological disorders such as Alzheimer's disease and amyotrophic lateral sclerosis. The same assessment held that overreliance on animals ignores genetic differences between species, variability within species, such as age, sex and strain, and differences in how animals are housed and raised: "A mouse is not a mouse, is not a mouse — so how can we accurately extrapolate that data to humans?"
Almost every gene found in humans is found in a similar form in other mammals, which makes them models for studying disease and researching new drugs, and fish, flies and parasites also share some similarities with humans. But large variations in specific gene families separate humans from other animals. The genetic difference between individual humans is around 0.1% and between humans and chimpanzees 1%, yet non-human primates account for only 0.28% of laboratory animals used in research in the USA, while 90% are mice, rats and other rodents, from which the genetic difference can reach 2.5%. The Encyclopaedia of DNA Elements (ENCODE) Project lets scientists compare human and mouse genomes, pinpoint differences in metabolism or the immune system at the genome level, and decide whether a murine model is the best one for their research, which can help reduce animal use. More is understood about mouse biology than about human biology, so hidden threats and missed opportunities are hard to identify in research using animal models, and ex vivo experiments alone cannot be relied on either.
Animals also differ from humans in the absorption, distribution, metabolism and excretion of drugs. When a drug candidate does not progress into clinical studies, the usual reason is toxicity in animals, while some approved drugs are later identified as potentially hazardous to human health and are relabelled or removed from the market.
Grapes can cause severe reactions, kidney failure and death in dogs, for reasons still unknown, and chocolate is similarly dangerous to them. If chocolate were a new drug and were tested in dogs because their metabolism is close to the human one, the dogs would show panting, vomiting, tremor, hyperthermia, tachycardia, hypokalaemia and raised enzymes, and 98% of those affected would die even after decontamination. The report would conclude that chocolate could be harmful to humans and should not be sold, and its chemical structure would be flagged so that drugs of similar structure were excluded early in discovery. Animal studies alone can lead to the loss of valuable drugs and so to the waste of animals, which is one reason for regulators and private organisations to invest in non-animal alternatives.
Animals are directly affected by their surroundings, and the availability and type of food and habitat can affect their response to treatment. Animals bred at different research centres are fed and treated in different ways, which affects how they develop a disease and respond to treatment. Mice are more anxious when picked up by the tail, the most common method of handling them, and this can significantly influence experimental results. Carrying them in acrylic tunnels without direct contact, or letting them walk freely on the handler's open hand, leaves them less stressed, so a researcher who handles the animals carelessly on a particular day may obtain different results from one who handles them with care.
The argument over animal testing
Animal experiments date back to ancient Greek and Roman scientists, and criticism of them also emerged in earlier centuries. Queen Victoria's opposition to animal testing in England strengthened the anti-vivisection campaigns, which resulted in the first laws controlling animal experiments, and Great Britain's Cruelty to Animals Act took effect in 1876. The case for animal testing is that it contributes to medical progress and that alternative testing systems are inadequate; the case against is that it is cruel and inhumane and that alternative methods already in place could replace it. A common criticism of required animal testing is that its results can hardly be extrapolated to humans, so that it is an unnecessary cruelty.
Animals have played a vital role in many medical and scientific advances of the past century, among them the discovery of insulin, penicillin and the polio vaccine. Scientists can reproduce human disorders in animal models, mimic their pathophysiology and use drugs to cure the condition. Research on animals avoids risking human lives in the initial phases of drug discovery, whose experiments are invasive and could result in human death. A safety risk remains after animal testing, but scientists argue that it is significantly higher without it. The European Animal Research Association describes forty reasons why animals are needed for biomedical research, among them that many important findings, such as the development of vaccines and drugs, relied on animal data, and that animals and humans are physiologically very similar. An open letter from the Confederation of Spanish Scientific Societies argues that animals cannot be substituted in making antibodies, because animals yield antibodies of higher affinity and specificity than other methods, and that substituting traditional technologies requires further scientific validation.
Doctors Against Animal Experiments Germany (Ärzte gegen Tierversuche), an association founded in 1979, opposes animal experiments for ethical, medical and scientific reasons, provides researched information for doctors, scientists and the public, and works with international organisations such as the European Coalition to End Animal Experiments to be heard at EU legislative level. The PETA International Science Consortium addresses the replacement of animal experiments by alternative techniques and lists publications and posters on the applicability of animal study results to humans and on alternative testing procedures for pharmacological development and for food and cosmetic safety.
Alternatives to animal testing
When planning animal experiments, scientists should apply the 3Rs principle: replacement, methods that avoid or replace the use of animals; reduction, methods that minimise the number of animals used per experiment; and refinement, methods that minimise animal suffering and improve welfare. The European Partnership for Alternative Approaches to Animal Testing is a collaboration between the European Commission, European trade associations and companies from seven industries, including the pharmaceutical industry, that aims at the 3Rs in regulatory testing. In 2017 its project groups focused on eight topics, one of them the vaccines consistency approach: vaccine quality control includes batch safety testing in animals as legislation requires, and the approach strives to implement the 3Rs in vaccine manufacturing by strictly applying quality systems that ensure batch consistency.
Pharmaceutical companies constantly seek to reduce both the attrition rate and the need for animal studies in nonclinical development by introducing new in vitro screening methods, such as human organs-on-a-chip, so novel methods are continually tested and presented to the regulatory bodies. One way to generate robust data that reflect how the human body works is to combine in silico data with accurate in vitro data, while improving techniques such as dynamic culture, bioprinters, organoids and organ-on-a-chip and using advances in computing to upgrade in silico technology.
Technologies for producing antibodies without animals have proven a powerful substitute, generating antibodies that would be extremely difficult to obtain from animals. Antibodies that do not derive from mice have shown increased therapeutic efficacy and can avoid harms such as allergic reactions against mouse-generated antibodies. Phage display, an in vitro antibody selection method, has been used to isolate antibody candidates, and many antibodies developed with it have been approved by the FDA for diseases such as cancer; the PD-L1 inhibitor atezolizumab was approved in 2016 for bladder cancer and in 2022 was in clinical trials in other tumour types.
Thomas Hartung, a toxicologist, held in 2015 that research was ready to move beyond animal models, and saw promising advances toward more reproducible, human-relevant research in toxicology and drug development. To spread these advances to other preclinical fields, he suggested that researchers should work to:
- Increase agreement that animal tests are not sufficiently reproducible or accurate for translation to humans.
- Increase understanding that some animal tests lead to costly wrong decisions: "f (trash) = trash".
- Understand and overcome the limitations of alternative methods such as in vitro testing, many of which have as many limitations as animal testing.
- Improve the capabilities of human-relevant models such as organotypic cell culture and organs-on-a-chip.
- Develop approaches such as automated cell cultures, computer models, and combined cell and computer methods.
- Generate big data and mine it with bioinformatics to make "big sense" of results.
- Practise evidence-based medicine in preclinical research by systematically and objectively evaluating whether results are consistent among studies.
- Find pragmatic, early ways to test targets in human-relevant models, as green toxicology and read-across approaches do.
Hartung granted that many scientific discoveries have come from animal studies, but held that in many cases they are not the best method for the question: "All models are wrong" in some manner, and "some are useful".
How far the alternatives are accepted
By 2022 ethics in animal research had increased significantly over the preceding few years, yet approval of new technologies to substitute animals was growing slowly: many scientists familiar with traditional methods were still not convinced of the benefits of new technologies, and wanted further validation to justify replacing animals.
By December 2017 some alternatives to animal testing had gained approval, including tests for eye irritation, skin irritation and skin sensitisation. According to the European Pharmacopoeia, a monocyte activation test, which measures cytokine levels after human blood is exposed to the test substance, could then be used in place of the rabbit pyrogen test for assessing pyrogenicity. The US FDA had recently announced a collaborative research agreement with Emulate, the developer of organs-on-chips technology, which uses micro-engineered living human cells to simulate human organs and can be used instead of animals in drug testing; according to Emulate, the system predicts the human response with greater precision and control than cell culture or animal-based testing.
In September 2021 the European Medicines Agency began promoting the 3Rs through its Innovation Task Force, a free, multidisciplinary forum for early dialogue between regulators and developers of medicines, by opening it to discussion of new approach methodologies that comply with the 3Rs and with EU legislation on protecting animals used for scientific purposes. Any such methodology that can be used to fulfil testing requirements is eligible for consideration, and alternatives such as improved tests based on human and animal cells, organoids, organs-on-chips and in silico modelling offer the opportunity of better and more predictive tools to protect human and animal health and the environment. The step is expected to encourage prioritising and speeding up the integration of alternative methods into the regulatory framework, and it is in line with the agency's Regulatory Science Strategy to 2025.
Agencies such as the European Union Reference Laboratory for Alternatives to Animal Testing, and other laboratories and centres, are doing work that shows many options to substitute animals in research with better results than animal models give. More laboratories need to implement these techniques to show the specificity and reproducibility of their results, until the scientific field, and government agencies especially, fully accept them.
In 2022 many countries had established national centres to develop and validate alternative methods, while government agencies were investing heavily in legislation and strategic roadmaps to allow drug approvals on in vitro and in silico methods, and the large body of omics data and tools available could make research cheaper, faster and more relevant to human physiology. Completely stopping the use of animals in research was still a long way off.
Reporting an animal result
Many writers have struggled to report translational research: they have sensationalised findings, omitted the model or not put the study in context. Writers might also assume that drug response mechanisms and complications in animals mirror those in humans, and so overlook the experimental and biological limits of the research. In one researcher's view, responsibility for poor reporting is shared among researchers, regulatory agencies, writers and all parts of the system, but writers can illuminate the limits of the current system by asking whether the research has been, or could be, done with more reliable, human-relevant methods. She holds that medical writers can deeply inform stakeholders and move research standards toward human-relevant methods. Her general advice for writing about such research:
- Minimise hope and hype.
- Handle "hot" findings with care and place them in the context of the study's limitations.
- Do not mistake association for causation.
- If the science is unclear, avoid making conclusions.
- Do not regurgitate institutional news releases.
Her more specific tips for writers promoting human-relevant methods:
- Question the human relevance of animal or alternative methods.
- Watch for contradictory findings among animal models.
- Ask researchers about their plans for more human-relevant testing, such as with human cells or tissues.
- Use "triple dose caution" when discussing translation, and be clear about the inherent unreliability and extended timeframes of many methods.
- Explain any caveats readers need to interpret the findings properly.
- Do not distort the severity of the limitations of translational research.
She frames this advice in the context of the AMWA Code of Ethics, which calls on medical communicators to communicate objectively, accurately, rigorously, completely and with fair balance.