Basic and Applied Biomedical Research

A recent meta-analysis showed that cancer experiments on animals have smaller effect sizes and are less likely to replicate than non-animal cancer experiments.⁷ Oncologists have noted that “crucial genetic, molecular, immunologic and cellular differences between humans and mice prevent animal models from serving as effective means to seek for a cancer cure.”⁸ Former director of the National Cancer Institute Dr. Richard Klausner stated, “The history of cancer research has been a history of curing cancer in the mouse. We have cured mice of cancer for decades—and it simply didn’t work in humans.”⁹ In addition, the enormous pain and suffering experienced by animals raises ethical and welfare concerns.¹⁰¹¹

There are several methods by which rodents—predominantly mice—are used in cancer experimentation. These methods are categorized based on the tumor development mechanism: xenografting, genetic engineering, or, less frequently, spontaneous induction through exposure to carcinogenic agents.¹²¹³

To create xenografted animals, immortalized or patient-derived human cancer cells are transplanted either under the skin or into an organ of immunocompromised rodents, who may then be subjected to a range of experiments, such as treatment with a drug candidate or a substance of interest. Although xenografting is the most common approach to generate tumors in rodents, an analysis of 1,110 mouse xenograft tumor models concluded that these models face fundamental challenges that hinder their ability to predict therapy outcomes in humans.¹⁴ Transplantation of human cells alters the genetic landscape of mice in ways that are unlikely to happen in humans, and these changes alter responses to drug treatment.

Genetically modified (transgenic) mice are created by inserting or deleting human genes into a mouse’s DNA to induce the expression of oncogenes or inactivate tumor-suppressing genes, respectively. Since these modifications happen randomly, researchers cannot control gene expression, and off-target alterations are common.¹⁵ Transgenic mouse cancer models fail to mimic the sporadic nature of tumor development, resulting in unexpected outcomes that would not be present in human patients. Moreover, these models are time-consuming and costly since they require many animals to obtain the desired and stable genotype, and the “surplus animals” are euthanized.¹⁶

In August 2021, the European Commission’s Joint Research Centre published a report on immuno-oncology. It highlighted promising human-based, non-animal methods for developing new therapies, studying cancer biology and immunomodulation, identifying specific molecular biomarkers, and more.¹⁷ Some examples of these human-relevant models for cancer research include three-dimensional platforms, such as bioprinted tumors using patient samples,^18192021 organs–on-a-chip models for precision medicine using different cancer cell lines,^2223242526 and patient-derived organoids.²⁷²⁸²⁹ In addition, cancer genomic datasets^3031323334 and machine learning tools^35363738 are available to improve diagnosis and predict responses to therapies in real-time. 

Scientists using non-animal methods for cancer research face a smaller translational hurdle since they can use patients’ own cancer cells and because these human-relevant methods are grounded in human, not rodent, biology.³⁹ These new tools and approaches will advance cancer research, produce human-relevant results, and accelerate the field toward precision medicine, but only if funding for them is increased and allocated away from cancer experiments on animals.

Experimenters use a variety of animal species, from frogs to rats to cows, in an effort to study human CVD. However, the etiology and pathology of CVD in these animals often differ significantly from those of humans.⁴³⁴⁴ Most species have distinct cardiovascular functional and structural parameters, including resting heart rate, action potentials, protein isoforms, contraction, and force-frequency response.⁴⁵⁴⁶⁴⁷ They also exhibit species-specific genetic mechanisms that affect their susceptibility to CVD and responses to drugs intended for human treatment.⁴⁸⁴⁹⁵⁰ For example, rodents are resistant to atherosclerosis,⁵¹ a key component of CVD. Coronary artery disease, which leads to atherosclerosis, rarely occurs in animals and is difficult to induce, often requiring surgical or pharmaceutical interventions that are not relevant to the human context.⁵²

Additionally, behavioral and environmental risk factors, such as diet, physical inactivity, smoking, and air pollution¹ are complex and not reliably reproducible in animals. These factors contribute to the limited relevance and poor clinical translation of CVD experiments on animals. A recent study’s authors noted that “profound understanding of disease progression is limited. The lack of biologically relevant and robust preclinical disease models that truly grasp the molecular underpinnings of cardiac disease and its pathophysiology attributes to this stagnation.”⁵³

Human-relevant in vitro and in silico methods are more suitable for cardiovascular research, as they reflect human biology better than animal models. Researchers have generated heart organoids using human induced pluripotent stem cells (hiPSCs) that mimic the cellular composition of the heart and self-organize to create chamber-like structures. These heart organoids can recapitulate functional impairments seen in conditions such as cardiac fibrosis and hypertrophic cardiomyopathy.⁵⁴⁵⁵⁵⁶ A team of engineers in Taiwan has developed a microfluidic chip system to rapidly quantify four CVD biomarkers aimed at improving early intervention.⁵⁷ A recent study demonstrated that heart-on-a-chip technology can be used to model cardiac arrhythmias.⁵⁸⁵⁹ Additionally, machine learning techniques, in combination with patient data, can create models to predict CVD risk, enabling earlier identification of diseases and more effective treatment outcomes.⁶⁰⁶¹⁶² Scientists and clinicians have collaborated to develop an algorithm that predicts 10-year disease progression in hypertrophic cardiomyopathy using clinical data.⁶³ Finally, in silico modeling and simulation can be employed to assess the mechanistic understanding of cardiac pathophysiology.⁶⁴ These methods are valuable platforms for studying the human heart, identifying and screening drugs for CVD treatment, and application in regenerative and personalized medicine.

Considering that “[t]here is no ideal animal model available for cardiac research,”⁶⁵ CVD research must evolve toward modern methods that rely on human cells and patient-derived data. These new experimental models are more cost-effective and better recapitulate human physiology.⁶⁶ Non-animal research methods provide more accurate biological insights into cardiac function, enhancing the translation of preclinical findings into human benefits compared to animal models.⁶⁷⁶⁸⁶⁹

Cell therapy research is often conducted using animals, primarily genetically engineered mice, and faces significant limitations. Experiments on animals typically use young, healthy animals who do not reflect the complex etiology of human diseases that are often influenced by age and other co-morbidities. Additionally, experiments on animals lack the long-term analysis and follow-up needed to assess efficacy in humans, posing a challenge in predicting outcomes.⁷⁷ Additionally, immune and physiological differences between species lead to poor translation of results.

Although some cell therapies have been approved for use, these treatments still face challenges, especially for solid cancers, due to tumor heterogeneity and the scarcity of tumor-specific antigens.⁷⁸ Engineered chimeric antigen receptor (CAR) T-cell therapies have shown antitumor activity in experiments on mice but failed to work in human clinical trials for ovarian and metastatic renal cell cancers.⁷⁹⁸⁰ One cause for these failures is that preclinical studies are often conducted using immunocompromised mice with xenografted human tumors, whereas in clinical practice, these cells operate within a patient’s complex and intact immune system.⁸¹ For more on the problems with xenograft mouse models, see the section on cancer.

Because animals do not accurately replicate human biology, they may also fail to reliably predict adverse effects of cell therapies, such as cytokine release syndrome and immune effector cell-associated neurotoxicity. Additionally, variability in cell preparation and characterization during preclinical experiments on animals can result in inconsistent and irreproducible findings.⁸²

Non-animal preclinical methods for studying and testing cell therapies include in vitro models, such as organoids and those using hiPSCs. These models replicate human physiology more accurately, allowing for high-throughput drug screening, identification of human-specific mechanisms, and personalized medicine approaches.⁸³⁸⁴ Maulana et al. introduced a patient-derived breast cancer-on-chip model that enables real-time monitoring of CAR T-cell activity and prevention of cytokine release syndrome with an FDA-approved drug.⁸⁵ In another study, researchers using patient samples and clinical data identified CD22 as a potential marker for CAR T-cell therapy development in triple-negative breast cancer, which, despite ongoing cell therapy clinical trials, is currently without targeted therapy.⁸⁶⁸⁷

Interest in adoptive cell therapies has surged in the past decade and continues to expand to various cancers and other diseases. Recent advances in engineering technologies, human in vitro models, and combination therapies are enhancing cell therapy development, providing robust platforms for studying disease mechanisms and therapeutic interventions, and yielding more applicable results.

Beyond technical limitations, using animals to study diabetes also poses significant biological limitations regarding anatomy, physiology, and exposure.⁹¹⁹² For instance, mice rely principally on the liver for glucose homeostasis, while, for humans, skeletal muscle is also critical in glucose metabolism.⁹³ In addition, some transgenic mice models of type 2 diabetes are based on leptin deficiency, which is not an essential contributor to diabetes in humans.⁹⁴ Because of a low rate of spontaneous diabetes (only 2%), the LEW-iddm rat model for type 1 diabetes requires compensatory alterations in the rat’s immune cell repertoire in order to develop a diabetic profile but still does not entirely mimic the human condition.⁹⁵⁹⁶ In the same way, the human pancreas differs from that of rodents in its tissue architecture, cellular composition, and insulin regulation.⁹⁷

Many drugs developed to treat diabetes have adverse side effects, such as edema, cardiac risk, and weight gain, with some drugs being withdrawn from the market.⁹⁸⁹⁹ Recent findings reveal significant human singularities in pathology, environment, ethnicity, and treatment responses among type 2 diabetes patients,^100101102103 highlighting why the heterogeneity of diabetes cannot be replicated using animals. As a result, experiments on animals have not led to transferable findings for humans.¹⁰⁴¹⁰⁵

As interspecies differences continue to emerge, there is a clear need for human-based methodologies to advance diabetes research to bridge the gap between pre-clinical and clinical trials and discover new ways to prevent disease progression.^106107108

Numerous organ-on-a-chip models for studying insulin resistance and glomerular function for diabetic nephropathy have been developed to uncover biological mechanisms and provide insights into effective therapeutic opportunities. For example,a glomerulus-on-a-chip using human cells allows researchers to assess high glucose-induced kidney damage.¹⁰⁹ In another study, the glomerulus-on-a-chip mimicked the human in vivo kidney response to injury in patients exposed to serum and toxic agents, providing a valuable tool to investigate renal damage.¹¹⁰ Another 3D model used cadaveric pancreas islets for continuous insulin measurements, offering a scalable model to study diabetes and perform drug screening.¹¹¹ In silico modeling using diabetic patient data is also showing promising results.^112113114 For example, a model designed to quantify endogenous and inhaled plasma insulin after a meal was tested in a clinical study with healthy patients and can help estimate the bioavailability and pharmacokinetics of inhaled insulin in humans.¹¹⁵

Many other human 3D models are being explored for drug development and considered for future organ transplantation in diabetic patients,¹¹⁶¹¹⁷ including stem cells¹¹⁸¹¹⁹ and pancreatic islets.^120121122 These innovative approaches, based on patient-derived cells, have the potential to accelerate research on diabetes as they permit investigation into the underlying biological mechanisms of human diabetes-induced complications, which are impossible to replicate in experiments on animals.¹²³¹²⁴

Because humans are the only primates who contract HIV and develop AIDS, experimenters instead infect monkeys with simian immunodeficiency virus (SIV), a virus unique to African primates. The genetic homology between HIV and SIV is only 55%, and SIV is less genetically diverse than HIV.¹²⁵¹²⁶ Owing to differences in surface proteins and other molecular markers, antibodies that neutralize SIV have no effect on HIV and vice versa.¹²⁷ Importantly, the dose of SIV administered to a nonhuman primate in an experiment is often much higher than the typical amount of HIV-1 to which a human is exposed during sexual transmission.¹²⁸ Sometimes, experimenters use an engineered SIV/HIV concoction. AIDS researcher Mark Girard has stressed, “One should realize that we still do not know how the SIV or SHIV model compares to HIV infection in humans. Extrapolating from vaccine protection results in nonhuman primate studies to efficacy in man may be misleading.”¹²⁹

Even those who use nonhuman primates as models of HIV have admitted that they “do not allow direct testing of HIV vaccines” and that “because of the complexity and limitations of the NHP [nonhuman primate] models, it remains difficult to extrapolate data from these models to inform the development of HIV vaccines.”¹³⁰ Experimenters have developed dozens of vaccines candidates using monkeys, but all have failed in human trials.¹³¹ At least two clinical trials resulted in an increased likelihood of HIV infection in humans.¹³²¹³³ After one of the failed vaccine trials, Anthony Fauci, former director of the U.S. National Institute of Allergy and Infectious Diseases, acknowledged that the original positive results of a macaque study “might be a fluke.”¹³⁴

Scientists have noted that “[e]xisting animal models predicting clinical translations are simplistic, highly reductionist and, therefore, not fit for purpose.”¹³⁵ They reported that clinical attrition data “focusses the attention back on to early target selection/lead generation, but it also questions the suitability of current animal models concerning congruency with and extrapolation of findings for human hosts.”

Because of broad failures in nonhuman primate HIV/AIDS research, some experimenters have shifted their focus to mice—a species even more genetically removed from humans. The “humanized” mouse model for HIV/AIDS research is a mouse who has been partially repopulated with human immune cells, allowing for the animal to be infected with HIV-1. However, humanized mice are limited in their longevity with the disease and retain parts of their murine immune systems, “complicating immune response interpretations.”¹³⁶ Not surprisingly, the use of humanized mice has also failed to generate valuable results for clinical HIV/AIDS treatment.

Considering the differences between a laboratory environment and human society, experiments on animals will never capture the complexity of this human disease. Mice and rats used in experiments are kept in conditions where the primary pathogens are those found in their feces, and cofactors that may be present in human patients, such as other microbial infections, are absent. This lack of cofactors significantly alters the acquisition and progression of the virus.¹³⁷ Nonhuman primates used in HIV research, on the other hand, have been found to harbor confounding infections like valley fever, which compromises the findings of HIV studies.¹³⁸

Scientists acknowledge that even after costly and unreliable experiments on animals, human data are still needed to determine whether a drug is fit for the clinical setting. Researchers with the U.S. Military HIV Research Program noted that “human clinical trials still appear to be the only reliable way to determine whether an HIV vaccine candidate will have activity or efficacy in humans,”¹³⁹ adding to this 2007 comment from the associate editor of The BMJ: “When it comes to testing HIV vaccines, only humans will do.¹⁴⁰ Researchers recognize that human in vitro models are needed to replicate this human disease and develop treatments.¹⁴¹

Recent non-animal HIV research includes interactive molecular dynamics simulations to predict how drug molecules will bind to HIV proteins,^142143144145 novel imaging techniques revealing previously unknown aspects of HIV structure that open up the potential for new therapies,¹⁴⁶ and bioinformatics analysis of specimens from individuals with viremia and in vitro–infected cells from healthy donors to construct an atlas of HIV-susceptible cell phenotypes.¹⁴⁷ Additionally, single-cell multi-omic analyses of samples from healthy and HIV-infected donors have uncovered differences in T cell populations, protein expression, and glycan expression, which could be instrumental in developing novel immune-targeted therapeutic strategies.^148149150

Scientists around the world have been studying the immune cells of individuals called “HIV controllers,” who can become infected with HIV but can control the spread of the virus without any therapeutic intervention.^{151152153154155} The hope is that immune cells from HIV controllers can be transferred to HIV-infected patients to help them fight the virus. This promising research is human-specific and requires human-specific testing methods.¹⁵⁶

Key physiological and cellular differences between the tissues of mice and humans reveal their inadequacy as human experimental stand-ins and should disqualify the use of mice in experiments.¹⁵⁹¹⁶⁰ Specifically for immunological research, mice have unique dendritic epidermal T cells with sensory functions nonexistent in humans.¹⁶¹ Similarly, the composition of immune cells in human blood (55–70% neutrophils, 20–40% lymphocytes)¹⁶² is different than that of mice used in experiments (20–30% neutrophils, 70–80% lymphocytes),¹⁶³ which affects species-specific immune defense mechanisms.¹⁶⁴¹⁶⁵ Logically, these differences make sense, given that we humans have longer life spans and “do not live with our heads a half-inch off the ground.”¹⁶⁶

Mice have a unique genetic makeup that contributes to their phenotypic dissimilarities with humans, such as the lack of class II human leukocyte antigen expression on T lymphocytes and differences in the activation of these cells during immune response.¹⁶⁷ These immunological specificities, along with epigenetic modifications unique to mice, hinder the data translation and make comparisons between mice and humans unrealistic and risky.¹⁶⁸¹⁶⁹ For example, a deficiency of CD28 molecules results in severe immune dysfunction in mice, while humans with this deficiency remain healthy.¹⁷⁰ Due, in part, to differences in CD28 expression between species, clinical trials with Fialuridine resulted in organ failure in humans who took only 1/500^th of the dose that had been deemed safe in preclinical tests using animals.¹⁷¹

A mouse’s immune system is also altered by the barren, controlled housing conditions in which they are kept in laboratories. Consequently, mice develop a gut microbiome adapted to these conditions,¹⁷² which is distinct from that of wild mice and even more divergent from that of humans.¹⁷³ In a study that analyzed over 1900 mouse genomes, researchers revealed that humans and mice have only 2% of gut bacteria species in common.¹⁷⁴ The breeding process used to generate specific mouse strains with genetic variations also makes them more susceptible to human pathogens than humans are, adding another point of discrepancy.¹⁷⁵¹⁷⁶ Mice in laboratories fail to represent the genetic variability found among humans or their own species’ wild counterparts.¹⁷⁷¹⁷⁸ Despite these many glaring disadvantages, mice continue to be used for immunological research.

Human immunological research is slowly but surely bringing the “human” back into its focus. “Big data” and computational biology—proteomics, metabolomics, and clinical data—integrated with novel 3D models can bridge the gap in translational science and leverage personalized approaches.^179180181182 Human samples, such as bone marrow,¹⁸³ lymph nodes,¹⁸⁴ tonsils,¹⁸⁵ and liver,¹⁸⁶ are being used to generate patient-derived organoids to address mechanistic and hypothesis-driven immunological studies in different contexts.

A review summarizing the progress of immune-competent human skin disease models recognizes that the failures of experiments on animals to translate into effective treatments for diseases such as fibrosis, psoriasis, cancer, contact allergy, and autoimmune diseases is due, in part, to the immunological nature of these conditions. The authors go on to describe how co-culture, three-dimensional organotype systems, and organ-on-a-chip technology will “enable human models of well-controlled complexity, yielding detailed, reliable data, providing a fitting solution for the drug development process.”¹⁸⁷

Mice are the animals most commonly used in sepsis research—not because they make good models of human sepsis but because they’re cheap, plentiful, small, and docile.¹⁹¹ The difficulty in reliably translating results from mice to humans is considered a primary cause of the failure of nearly all human trials of sepsis therapies.

In 2013, Proceedings of the National Academy of Sciences of the United States of America published a landmark study that took 10 years to complete and involved the collaboration of 39 researchers from institutions across North America, including Stanford University and Harvard Medical School. Dr. Junhee Seok and his colleagues compared data from hundreds of human clinical patients with results from experiments on animals to demonstrate that humans and mice are dissimilar in their genetic responses to severe inflammatory conditions such as sepsis, burns, and trauma.¹⁹²

Former NIH Director Dr. Francis Collins authored an article about these results, lamenting the time and resources spent developing 150 drugs that had successfully treated sepsis in mice but failed in human clinical trials. He called this disaster “a heartbreaking loss of decades of research and billions of dollars.”¹⁹³ The paper reveals that in humans, many of the same genes are involved in recovery from sepsis, burns, and trauma but that it was “close to random” which mouse genes might match these profiles. Collins explains it as follows:

Mice, however, apparently use distinct sets of genes to tackle trauma, burns, and bacterial toxins—when the authors compared the activity of the human sepsis-trauma-burn genes with that of the equivalent mouse genes, there was very little overlap. No wonder drugs designed for the mice failed in humans: they were, in fact, treating different conditions!¹⁹⁴

Even before this landmark study, the criticism of mouse models had been documented in more than 20 peer-reviewed scientific papers. The mice used in sepsis experiments are young, inbred, and of the same age and weight, and they live in primarily germ-free settings. In contrast, it is mostly infant and elderly humans who live in a variety of unsterilized, unpredictable environments who develop sepsis.¹⁹⁵¹⁹⁶ When experimenters induce the condition in mice, the onset of symptoms occurs within hours to days, whereas in humans it takes day to weeks. Mice are not typically provided with the supportive therapy that human patients receive, such as fluids, vasopressors, and ventilators.¹⁹⁷ Unlike humans, mice are rarely given pain relief,¹⁹⁸ another difference that undermines data of already questionable value, as pain affects other physiological processes.

The “gold standard” method of inducing sepsis in mice is through cecal ligation and puncture, a procedure in which experimenters cut open a mouse’s abdomen and puncture their intestines with a needle before sewing the animal back up. However, mice’s responses to this procedure vary depending on age, sex, strain, laboratory, the size of the needle used, and the size of the incision, which makes results incomparable between laboratories.¹⁹⁹²⁰⁰ In addition, the procedure causes the formation of an abscess, whose effects may disguise or be disguised by the effects of the sepsis itself.²⁰¹ This means that an intervention that appears beneficial for sepsis may only appear beneficial because of its effects on the abscess.

Rats, dogs, cats, pigs, sheep, rabbits, horses, and nonhuman primates, including baboons and macaques, have also been used in sepsis experimentation. None of these species reproduce all the physiologic features of human sepsis. The pulmonary artery pressure responses of pigs and sheep differ from those of humans, so this aspect of sepsis cannot be compared between these species.²⁰² Furthermore, baboons and mice are less sensitive to a species of bacteria commonly used to induce sepsis in experimental settings.²⁰³ A recent study found that rhesus macaques and baboons differ markedly in their innate immune response to pathogens compared to humans.²⁰⁴

A 2019 report from the National Advisory General Medical Sciences Council (NAGMSC) Working Group on Sepsis states, “Despite decades of intensive study of the underlying mechanisms of this condition, no new drug or significantly new diagnostic technology has emerged. Dozens of prospective trials of agents or strategies targeting the inflammatory basis of sepsis have failed.” In its report, the NAGMSC Working Group on Sepsis recommended that the National Institute of General Medical Sciences (NIGMS), under NIH, “rebalance” its sepsis research–funding portfolio to “include a more clinical focus.”²⁰⁵ In a “Notice of Information” issued by NIGMS following the NAGMSC report, the institute expressed its intention to support sepsis research that “uses new and emerging approaches, such as clinical informatics, computational analyses, and predictive modeling in patients, and new applications of high-resolution and high-throughput bioanalytical techniques to materials obtained from septic patients” and called the support of “[s]tudies using rodent models of sepsis” a “low priority.”²⁰⁶ More recently, at the 2024 Shock Society Annual Conference, NIGMS announced that it was “unwilling” to fund projects proposing mouse models of human sepsis and encouraged the use of animal-free research methods moving forward.²⁰⁷ In other words, NIGMS intends to prioritize funding human-relevant sepsis research over sepsis experiments on animals. However, other NIH institutes and funders have yet to follow NIGMS’ lead.

In 2015, an expert working group consisting of veterinarians, animal technologists, and scientists issued a report on implementing the 3Rs (the replacement, reduction, and refinement of animal use) in sepsis research.²⁰⁸ The group identified several methods that could be used instead of animal models, including in vitro cell culture models for studying sepsis mechanisms, systems and computation biology for revealing the inflammatory processes occurring during sepsis, three-dimensional cell culture models to explore human disease progression and infectious mechanisms, synthetic human models to recreate disease-related cell types and tissues, and human genomic data to understand how sepsis affects individuals differently and which groups may be more at risk. The authors state that genomic information “will complement or even replace the need for mouse models in disease discovery and drug development.”²⁰⁹

The following are examples of recent developments in human-relevant sepsis research: 

• Scientists in Tokyo, Japan, used hiPSC-derived liver organoids to model the pathological events of septic-associated liver dysfunction and recovery following infection.²¹⁰
• A team of engineers, doctors, and researchers at Temple University identified an association between neutrophil types and the severity of sepsis using a human lung-on-chip model, which can be used to determine the appropriate therapeutic intervention based on sepsis severity.²¹¹
• Researchers in Hefei, China collaborated with physicians at First Affiliated Hospital to create a six-unit microfluidic device that comprehensively analyzes a sepsis patient’s white blood cell activity to monitor disease progression and severity.²¹²
• Massachusetts General Hospital scientists and physicians created a microfluidic device to accurately detect a biomarker of sepsis pathophysiology using a drop of blood, aiming to improve disease monitoring.²¹³
• Because early detection of sepsis is likely the most critical factor in reducing mortality from this condition,²¹⁴ researchers around the globe are exploring various artificial intelligence and machine learning tools to aid in the early prediction and diagnosis of sepsis.^{215216217218219220221222223}

Key differences in nonhuman animals render them inappropriate models for studying human GI diseases. The two species most often used in these experiments are rats and pigs.²²⁷ Both have GI tracts that are anatomically dissimilar to those of humans. For example, the jejunum constitutes 90% of the rat’s small intestine but only 38% of the human small intestine.²²⁸ Rats lack a sigmoid colon, gallbladder, and cystic ducts, while pig colons are larger than those of humans.^229230231 Beyond anatomical differences, behavioral disparities impact the relevance of these animal models. Rats typically consume small, frequent meals, whereas humans eat larger, less frequent meals.²³² Pigs, on the other hand, consume more food relative to their body weight than humans do.²³³

Laboratory conditions can further influence the study of GI diseases. In a 2024 study, researchers found that the temperature at which mice are housed within a laboratory can significantly affect their gut motility and microbiota.²³⁴ The source of the animals can also lead to variations in gut microbiomes due to differing environmental factors.²³⁵ Species-specific microbiome differences play another role: Pigs have little Bifidobacterium, a major species in the human gut.²³⁶ Given the role of gut microbiota in immune response, these differences may significantly impact study outcomes.²³⁷

Animal models of human GI conditions are criticized for their poor predictive value regarding disease outcomes and clinical efficacy in humans, especially for conditions like irritable bowel syndrome (IBS) and irritable bowel diseases (IBD), the pathogenesis of which remains not fully understood.²³⁸

IBS is a chronic condition affecting the lower GI tract. Fifteen percent of adults in the U.S. experience IBS symptoms, which include abdominal pain accompanied by diarrhea, constipation, or both. While the exact cause of IBS remains unclear, it is believed to involve a combination of physical and psychological factors, particularly stress and anxiety,²³⁹ which cannot be faithfully simulated in nonhuman models.

Animal models of IBS are typically created by subjecting animals to stress during early development.²⁴⁰ These models have significant limitations, such as their inability to replicate the constipation or mixed bowel responses of human patients. Additionally, human IBS patients often present with overlapping disorders, such as bladder pain syndrome, chronic pelvic pain, anxiety, and depression—none of which are modeled in experiments on animals. Behavioral changes, such as anxiety or depression, are difficult, if not impossible, to measure in animals. Most experiments use male animals, even though IBS is more commonly diagnosed in females. Additionally, abdominal pain, the primary symptom of IBS, cannot be accurately assessed in animals, as there is no measurable phenotype specific to the visceral pain experienced by humans. These shortcomings make IBS experiments on animals inappropriate for understanding IBS pathophysiology and developing effective treatments.²⁴¹

IBDs, which include ulcerative colitis and Crohn’s disease, are chronic inflammatory conditions often affecting the large and small intestines. IBDs impact 2 to 3 million people in the U.S.²⁴²²⁴³ IBD patients suffer from rectal bleeding, severe diarrhea, abdominal pain, fever, and weight loss. The causes of IBDs are believed to involve a combination of genetic, immune, microbial, and environmental factors, though the precise mechanisms are not fully understood.²⁴⁴

In IBD research, scientists induce colitis by administering irritating substances or using genetically engineered mice. However, reproducibility remains a significant issue. Different mice strains exhibit varying susceptibilities to chemically induced colitis, and microbiome differences across strains or vendors can also influence disease development in genetically engineered mice. Given that both genetic and environmental factors contribute to IBD, an animal model that lacks these human-specific characteristics cannot effectively replicate these diseases. For example, genetically engineered mice are often created by mutating a single gene, but human IBDs are polygenic.²⁴⁵ Furthermore, chemically induced colitis in mice typically results in acute injury over a few days, whereas IBDs in humans develop over years.²⁴⁶

A key example of the limitations of animal models is IL-17 inhibition, which effectively treats colitis in mice but has failed in Crohn’s disease patients, sometimes even worsening the condition.²⁴⁷²⁴⁸ A 2019 review noted that “while there are many in vivo models of IBD, none adequately predicts response to therapeutics.”²⁴⁹ The disappointment of IL-17 inhibition in clinical trials illustrates how a treatment that works in animal models can fail in humans. Conversely, some therapeutics that show promise for treating IBDs in patients have failed in mouse models.²⁵⁰²⁵¹

Given these limitations, it is clear that no animal model can accurately replicate human GI disorders. These conditions are influenced by a complex interplay of environmental, genetic, and microbial factors that cannot be fully captured in artificially induced animal models. Therefore, prioritizing human-relevant research methods, such as organoids, microfluidics, and organ-on-a-chip technologies, is crucial. Recent developments in this area include the following:

• Biological engineers at MIT created a human multi-organ model of ulcerative colitis to study its impact on the gut-liver-immune axis.²⁵²
• Scientists at the Francis Crick Institute, in collaboration with UCL and Imperial College London, used a multi-omics approach to identify a new biological pathway related to IBDs, finding the gene ETS2, which is linked to higher IBD risk.²⁵³
• A group of researchers and physicians in Missouri and North Carolina created a neonatal-intestine-on-a-chip to study necrotizing enterocolitis, a deadly GI disease seen in premature infants. They successfully showed that this model can recapitulate disease pathology and plan to use this method for therapeutic testing.²⁵⁴
• Physicians and scientists in Boston obtained biopsies, blood, and stool samples from patients at Cincinnati Children’s Hospital, Massachusetts General Hospital, Emory University Hospital, and Cedars-Sinai Medical Center to create a longitudinal molecular profile of their microbiomes. Using a multi-omics approach, they were able to identify microbial, biochemical, and host factors involved in IBD-induced dysregulation.²⁵⁵
• Researchers and physicians in Houston used patient-derived intestinal organoids to explore the link between telomere dysfunction and IBDs, suggesting that addressing telomeric dysfunction could be a therapeutic strategy.²⁵⁶

The anatomical and physiological differences between nonhuman and human GI systems, coupled with the artificial induction of GI diseases in animals, hinder reliable study outcomes. Furthermore, many of these induction methods involve invasive and painful procedures, leaving the animals in distress until they are killed.^{257258259260261} Given that animal models of GI diseases do not reliably reflect human pathology and contribute to animal suffering, it is essential to transition toward the numerous non-animal methods using human tissues or consenting patients.

For example, methylprednisolone, a routinely used treatment for acute SCI, has generated inconsistent results in animal models. A systematic review examining 62 studies of the drug on a wide variety of species, from rodents to monkeys, found that 34% reported beneficial results, 58% reported no effect, and 8% had mixed findings.²⁶⁴

Among species, rats are particularly unsuitable for nerve repair or regeneration research. Experts have pointed out three major problems with rat models in this field: 

(1) The majority of nerve regeneration data is now being generated in the rat, which is likely to skew treatment outcomes and lead to inappropriate evaluation of risks and benefits. (2) The rat is a particularly poor model for the repair of human critical gap defects due to both its small size and its species-specific neurobiological regenerative profile. (3) Translation from rat to human has proven unreliable for nerve regeneration, as for many other applications.²⁶⁵

More specifically, the inconsistencies between animal models and the clinical situation are significant²⁶⁶ and include the following:  

(1) healthy animals versus sick patients; (2) short versus long gap lengths (the clinical need for large gap repairs, while 90% of in vivo studies are in rats and rabbits where gap lengths are usually ≤3 cm); (3) animal models that almost always employ mixed sensory-motor autografts for repairing mixed defects, versus clinical repairs that almost always involve sensory autografts (usually sural nerve) for repairing mixed defects; (4) protected anatomical sites in animal models, versus repairs that must often cross articulating joints in humans; and (5) inbred, highly homogeneous animal strains and ages, versus diverse patient populations and ages: It is well recognized that animal models fail to mimic the human condition in terms of the uniformity of animal subjects used.²⁶⁷

To induce a spinal cord injury in animal models, experimenters use physical force to damage the spinal cord. There are many different methods, such as contusion, which involves displacing the spinal cord by dropping a weight, or distraction, which applies a traction force to stretch the spinal cord. Regardless of the method used, achieving consistency and reproducibility is challenging due to the inability to replicate the same spinal cord injury every time they perform the procedure. For example, in contusion-induced injuries, variability can arise from the rod bouncing after it hits the spinal cord, potentially causing multiple impacts.²⁶⁸

In addition to consistency issues, many of these models do not accurately reflect the mechanisms of SCI in humans. A compression model created using forceps does not replicate the acute impact seen in most human SCI, and the devices used for the distraction model often induce injury too slowly to emulate human injury. Chemically induced SCI is employed to study secondary injuries associated with SCI, usually involving the injection or application of a toxic chemical to the area of interest. However, challenges with chemically induced SCI include ensuring accurate delivery of the chemical to the correct region of the spine.²⁶⁹

Biomedical engineers have noted that researchers “are incapable of truly mimicking human neural injures in animal models because of the extensive anatomical, functional, molecular, immunological, and pathological differences between humans and frequently studied animals.”²⁷⁰ Human-relevant methods can bypass these limitations and should be the focus.  

Human-relevant methods for studying nerve injury and regeneration have been reviewed by a number of research groups and include human organoids, microfluidics, engineered human tissue scaffold molds, bioprinting, and other in vitro uses of human cells. Ex vivo models, such as those using three-dimensional engineered scaffolds, bioreactors, neurospheres, and organoids, allow for more controlled studies on specific parameters than animal experiments.²⁷¹ Bioprinting can use bioinks containing human cells and materials to construct heterogeneous tissue models in a single step and with remarkable consistency,²⁷² an aspect of nerve regeneration research that has been notably lacking in animal models.²⁷³

Engineers and researchers at the University of Pittsburgh Medical Center and Carnegie Mellon University have emulated mild and moderate traumatic brain injury (TBI) using human cerebral organoids. Their study identified important genetic repercussions of TBI on the brain that can be used to diagnose the condition and create personalized treatments for patients.²⁷⁴ Neuroscientists have engineered human spinal cord organoids that display functional neuronal activity and hold promise for investigating SCI therapies.²⁷⁵

Microfluidic devices are “adaptable for modeling a wide range of injuries” and provide advantages over traditional in vivo and in vitro experiments by “allowing researchers to (1) examine the effect of injury on specific neural components, (2) fluidically isolate neuronal regions to examine specific effects on subcellular components, and (3) reproducibly create a variety of injuries to model TBI and SCI.”²⁷⁶ For example, brain-on-chip platforms offer a promising avenue for personalized medicine, as a patient’s own cells can be used to create a custom device to investigate treatment options.²⁷⁷ Axons-on-a-chip can model diffuse axonal injury, allowing researchers to track the intracellular changes immediately following injury and offering a platform for screening treatments.²⁷⁸ These systems offer advantages in precision, scalability, and cost-effectiveness when compared to traditional cell culture or experiments on animals, and are currently on the market and available for neural regenerative medicine research.²⁷⁹

All these diseases are human-specific, meaning they do not occur naturally in other animals. No animal model has been developed that recapitulates all aspects of a particular neurodegenerative disease.²⁸⁰ For AD research, the clinical failure rate for new drugs was last estimated to be 99.6%,²⁸¹²⁸² and recent monoclonal drugs approved for AD have been controversial due to adverse effects and questionable efficacy.²⁸³²⁸⁴

A bioinformatics analysis comparing the transcriptional signatures of AD, PD, HD, and ALS with mouse models of these diseases produced the following findings: 

[M]ost available mouse models of neurodegenerative disease fail to recapitulate the salient transcriptional alterations of human neurodegeneration and … even the best available models show significant and reproducible differences compared to human neurodegeneration. Although the reasons for the poor transcriptional performance of mouse models varied, the unifying theme was the failure of mouse models to exhibit the variety and severity of diverse defects observed in human neurodegeneration.²⁸⁵

These molecular discrepancies underscore the artificial methods used to create such models. Physical and chemical lesioning or systemic administration of toxins are commonly used. These are acute stressors, not long-term degenerative processes, and as such, they initiate events in animal models that are not present in human patients. The acute and immediate nature of disease models, such as the 6-OHDA and MPTP animal models of PD and the 3-NP animal model of HD, fail to capture the progressive nature of the disorders they aim to mimic. In addition, scientists often use young animals to “model” diseases associated with aging,²⁸⁶ further reducing their relevance. For example, “[c]ommonly used AD mouse models, like the 5xFAD, display amyloid deposits starting at 2–4 months of age … this early accumulation can be translated to Aβ deposits occurring in 4–8 year-old humans, a scenario not found even in the most aggressive cases”²⁸⁷ of AD.

Genetically modified mouse models exhibit inconsistent pathological and behavioral phenotypes, partly due to variations in transgenes used, inconsistencies in transgene insertion and expression, and differences in mouse background strains.²⁸⁸ As of 2024, 210 transgenic rodent AD models have been developed.²⁸⁹ In a review on the relevance and translational validity of these mouse models, researchers described their shortcomings:

Some transgenic models can present a very aggressive disease phenotype compared to the human form of the disease … while others fail to demonstrate aspects of neuronal loss and dysfunction. … Of additional concern is the fact that mouse models often fail to show a substantive neuronal loss even in the presence of amyloid deposits and generate amyloid peptides different from those found in human brain. … In some instances, the failures encountered with animal transgene models reflect the fact that they are based on intrinsically flawed hypotheses and the constructs used to interrogate these; in other instances, they reflect a lack of diligence on the part of investigators to ensure best practices in the husbandry and use of these models. Despite their limitations, these flawed models become widely utilized, with their relevance being overstated because of the lack of any viable alternatives, while only lip-service is paid to their validity as they become de rigor and self-perpetuating—driving the field down a blind alley.²⁹⁰

Fundamental genetic differences further hinder translation. For example, “knock-in models require the presence of multiple APP [amyloid precursor protein] mutations not found in humans,” murine tau differs structurally from human tau, and “key amino acid substitutions make murine Aβ less prone to aggregation when compared to its human counterpart.”²⁹¹ These differences make animal models of neurodegenerative disease misleading and waste precious time: A genetic target for AD research previously identified as upregulated in mouse models was, unsurprisingly, not found to be upregulated in humans in a recent postmortem study.²⁹²  For PD, nonhuman primate studies do not “constitute a valid scientific modality for the complete understanding of PD and for the development of future neuromodulation therapeutic strategies.”²⁹³

As in much of biomedical research, animals suffer greatly when used to mimic neurodegenerative diseases. In an analysis of published research on animal models of HD, 51 studies referenced experiments “in which animals were expected to develop motor deficits so severe that they would have difficulty eating and drinking normally.”²⁹⁴ However, only three out of 51 reported making adaptations to the animals’ housing to facilitate food and water intake. The authors of this analysis concluded that experimenters are not adhering to the 3Rs principles and compromising not only animal welfare but also the relevance of their studies to HD.²⁹⁵

As animal studies fall short, scientists and policymakers are increasingly recognizing the need for human-relevant research strategies. Following a review of AD research, an interdisciplinary panel recommended reallocating funding away from animal studies and toward more promising techniques, such as patient-derived hiPSC models, “omic” technology (genomics, proteomics, etc.), in silico models, neuroimaging, and epidemiological studies.²⁹⁶

The following are highlights in recent cutting-edge, human-relevant neurodegenerative disease research. 

• At Brigham and Women’s Hospital, researchers differentiated hiPSCs into neurons that quickly develop protein inclusions mimicking those found in the brains of individuals who had died with inclusionopathies. Using this method, the team created more than 60 human cellular models that other laboratories can use to study human neurodegenerative diseases.²⁹⁷
• A team of scientists at Washington University in St. Louis used cells from patients with AD to develop a relevant, 3D human cellular model for late-onset AD (which accounts for 95% of cases). This model allows for the study of age-associated neurodegeneration.²⁹⁸ Another team conducted a proteomic study on the cerebral spinal fluid of patients with AD to identify biomarkers that can be detected decades before symptoms arise.²⁹⁹
• Researchers at the Barcelona Institute of Science and Technology developed an organ-on-a-chip to evaluate the brain permeability of nanotherapeutics and facilitate personalized research and therapy for AD.³⁰⁰
• At the Vienna BioCenter, scientists created an in vitro model of the human dopaminergic system with ventral midbrain–striatum–cortex assembloids to improve the study of PD cell therapies.³⁰¹
• Researchers at the University of Luxembourg used human organoids and assembloids—including those developed with patients’ own cells—to understand the early stages of PD and factors influencing susceptibility.³⁰²³⁰³
• Boston-based Emulate Inc engineered a human brain-on-a-chip that represents areas affected by PD, reproduced features of the disease, and can be used to identify and test new therapeutic targets.³⁰⁴
• Scientists in Germany used human brain organoids to identify a gene implicated in HD that may damage the brain before symptoms arise and could serve as a focus for drug development. Restoring the function of this gene reversed the HD phenotype.³⁰⁵
• University of Central Florida scientists used cells from patients with ALS to develop a disease-specific neuromuscular junction-on-a-chip and tested the effects of a compound on clinically relevant functional measures of ALS.³⁰⁶
• In another patient-specific study, a team at Utrecht University used human brain organoids to improve the understanding of synaptic changes in ALS patients before the onset of symptoms.³⁰⁷

For decades, experimenters have tormented monkeys, mice, dogs, and other animals in an attempt to model these devastating diseases. However, since other animals don’t develop these human neurodegenerative diseases naturally, experimenters have manipulated their genomes to force discrete symptoms. The results, after decades of tests, include more than 100 failed drugs, an untold number of animal deaths, and the continued suffering of humans living with these conditions. For these patients, a shift to human-relevant methods is long overdue.

No single animal model replicates all aspects of a human neuropsychiatric condition, and features of human behaviors that represent hallmarks of these disorders cannot be accurately produced or assessed adequately in animals.

For example, human depressive disorders are characterized in part by feelings of sadness, hopelessness, and despair. In an effort to measure “despair” in rodents, the most commonly used behavioral test is the forced swim test, in which an experimenter places a rat or mouse in a container of water with no way to escape or rest. Experimenters falsely interpret the amount of time the animal spends swimming or struggling to escape as a measure of the animal’s lack of despair. This misguided notion originated from the observation that swimming and struggling time could be extended by giving the animal some types of human antidepressants (even though this assumption ignores the many false positives and false negatives that the test produces). As has been widely discussed in the scientific literature, an animal’s behavior in the forced swim test may represent an evolutionary adaptation to the stressful situation and should not be used to determine their mood.³⁰⁹ The results can be influenced by an animal’s strain and many experimental variances, including water depth, container dimensions, and temperature.^310311312313

A PETA neuroscientist and collaborators have published papers discrediting the use of the forced swim test as a valid method for screening antidepressant drugs. Their findings revealed that the use of this test by the world’s top 15 pharmaceutical companies did not produce any drugs currently approved for treating depression in humans.³¹⁴ They also highlighted actionable steps regulatory authorities could take to eliminate the use of the forced swim test (and the similar tail suspension test) in the pharmaceutical industry.³¹⁵

Other animal behavioral tests—such as the sucrose preference test (for anhedonia),^316317318 open field test and elevated mazes (for anxiety),³¹⁹³²⁰ marble burying (for compulsion),³²¹ chronic unpredictable stress (to induce psychopathologies)³²² —have similar flaws. These concerns have led to the awareness that “some of these assays must be discontinued, and placed in the past; while we seek improved, innovative strategies for outcome measures.”³²³

A series of citation analyses demonstrated that researchers studying major depressive disorder in humans rarely cite results from experiments on rats or monkeys, two of the most commonly used species in this field. Instead, they more frequently relied on research results using human cells and human biological data.^324325326 A similar failure of animal studies to contribute to clinical knowledge has been noted in bipolar depression research,³²⁷ and animal studies have been cited as the primary source of attrition (failure of drugs) in neurobehavioral clinical trials.³²⁸ Despite these warnings, thousands of researchers have continued to use flawed assays like the forced swim test to draw erroneous conclusions about an animal’s mood³²⁹ or the potential effects of compounds on human depressive disorders.³³⁰

Significant physiological differences between humans and other animals contribute to the low translation rate. For example, the gene encoding tyrosine hydroxylase, the enzyme involved in dopamine formation, is regulated differently in humans than in mice.³³¹ Misregulation of tyrosine hydroxylase has been implicated in several psychiatric illnesses, such as bipolar disorder and schizophrenia. In a 2019 study published in Nature, 64 researchers analyzed the brains of mice and humans and found substantial species differences in types of brain cells and how they produce proteins critical to neuropsychiatric function. The authors noted numerous “failures in the use of [the] mouse for preclinical studies” because of “so many [species] differences in the cellular patterning of genes.”³³² Rodents and humans also diverge in other critical areas for neuropsychiatric research, including the diversity, organization, and volume of neuronal cell types; relevant neural circuitry; volume of neurotransmitters available in specific cell types; and neurotransmitter receptor availability and kinetics.³³³

Beyond the lack of applicability, animal neuropsychiatric models cause immense suffering. To induce “depression,” experimenters subject animals to uncontrollable pain through electric shocks or chronic stressors, such as restraining them for extended periods, starving them or denying them water, tilting their cages, forcing them to live in wet bedding, shaking them, or disrupting their circadian rhythms. Animals are often made to live in complete isolation from other members of their species, bullied and physically assaulted by other animals, deprived of parental care, and subjected to genetic or surgical manipulations in an effort to induce a depressed-like or altered mental state. In this field, in particular, “animals are likely undergoing experimental procedures that do not provide the epistemic benefit we are sacrificing them for.”³³⁴

Funds should be redirected from the use of animals toward relevant, human-based experimental methods, including the following:

• Human brain organoids: Advanced, 3D in vitro cultures of human brain cells that replicate the cellular organization and signaling of human brain tissue. These have been used to study mood disorders, psychoses, and neurodivergence.^335336337338 Organoids can be combined to form self-organizing assembloids that mimic complex interactions between different parts of the brain,³³⁹³⁴⁰ such as the cortico-striatal-thalamic-cortical circuit and thalamocortical assembloids recently developed by a team at Stanford University to study human neurodevelopmental conditions like autism, Tourette’s syndrome, and schizophrenia.³⁴¹³⁴² Researchers at the University of California-San Diego and the University of Massachusetts-Amherst are developing disease-specific brain organoids using cells from patients with genetic mutations linked to neuropsychiatric disorders for therapeutic applications.^343344345
• Omics research: This is being applied to better understand the underpinnings of human neuropsychiatric conditions. The PsychENCODE Consortium, a collaboration of multidisciplinary teams, uses state-of-the-art methods to create large datasets from human postmortem brain samples.³⁴⁶ Some teams are analyzing existing data to characterize gene variants related to these disorders.³⁴⁷
• Brain imaging: Techniques including magnetoencephalography, high-density electroencephalography, magnetic resonance spectroscopy, transport-based morphometry, and functional magnetic resonance imaging—often combined with machine learning and genomics—are being used to study human psychiatric conditions and neurodivergence directly in individuals with lived experience.^{348349350351352}
• Longitudinal studies: Tracking individuals over extended periods provides insights into the effects of environmental stimuli, medical history, and life events on the incidence and progression of neurodevelopmental conditions.³⁵³³⁵⁴
• In silico clinical trials: Virtual patient models have been used to evaluate the potential of drugs for conditions like attention-deficit/hyperactivity disorder and schizophrenia.³⁵⁵³⁵⁶

Given the psychological distress inflicted on animals and the inapplicability of the results to humans, the use of animals in human neuropsychiatric and neurodivergence experiments should end. Resources must be diverted to human biology-based research like the examples listed above.

Although time constraint was an obvious factor in this decision, it is essential to note that many species do not respond to SARS-CoV-2 infection in the same way humans do. When The New York Times asked about seemingly promising experimental results in rhesus macaques, Dr. Malcolm Martin, a virologist at NIH, “cautioned that monkeys are different from humans in important ways.”³⁵⁸ The interviewer noted that “[t]he unvaccinated monkeys in [the vaccine experiment] didn’t develop any of the severe symptoms that some people get following a coronavirus infection” and quoted Martin as saying, “It looks like they got a cold.”³⁵⁹ Even genetically engineered mice, who are made susceptible to the disease, only show mild symptoms. “Humanized” mice (those who are engineered to express human immune factors) do not solve this problem, as “many human factors cross-react with murine cells, which may lead to unexpected phenotypic changes.”³⁶⁰

Amid the COVID-19 pandemic and outbreaks of other infectious diseases like H5N1, it has become increasingly clear that infectious disease research and pandemic preparedness should be prioritized. Human-relevant research can lead the way.

Many scientists are using innovative non-animal methods to study existing pathogens and those with pandemic potential. These methods include human lung and intestinal organoids, three-dimensional reconstructed human respiratory tissue models, human oral tissue samples from healthy volunteers, advanced computer simulation and supercomputers, human genetic analyses, human challenge studies, human-derived antibodies, and human organs–on-chips modeling human lungs, mouths, eyes, noses, and intestines. Complex in vitro human models, such as organoids and organs-on-chips, are expected to be particularly valuable for infectious disease research and developing vaccines and antiviral drugs.^{361362363364365366} Here are a few recent examples: 

• Human lung and brain organoids are being used to study SARS-CoV-2 infection mechanisms, test potential therapies, and investigate the effects of the virus on the brains of healthy individuals and those with comorbidities.^{367368369370371}
• Researchers in Japan created patient-specific livers-on-chips to explore SARS-CoV-2-induced liver dysfunction and to evaluate drugs to treat it.³⁷²
• Using cells isolated from human lung tissue, researchers engineered human lung organoids to study H5N1 virus replication, host cell survival, and lung immune responses to different viral strains.³⁷³
• According to a recent review, “microphysiological systems and organoids are already used in the pharmaceutical R&D pipeline because they are prefigured to overcome the translational gap between model systems and clinical studies.”³⁷⁴ The authors explain that complex, human-derived systems like organoids and microphysiological systems will be essential for research on filovirus and bornavirus infection in humans, for which “animal models cannot capture the respective pathogenesis and disease in full.”³⁷⁵
• Respiratory syncytial virus is being studied using ex vivo samples from patients to determine why some have a more severe reaction to the infection³⁷⁶ and with human airway organoids to develop and test antibody therapies.³⁷⁷
• Individuals with post-infectious disease syndromes like long COVID and myalgic encephalomyelitis/chronic fatigue syndrome have been studied using brain imaging; analyses of skin biopsies, blood, and cerebrospinal fluid; monitoring of diet, sleep, and cardiac measures; and more to phenotype these conditions, understand how they occur, and guide potential therapies.³⁷⁸
• In silico tools have been used in drug repurposing studies to identify existing therapies that could treat COVID-19.³⁷⁹

In addition to adopting non-animal methods to study and develop treatments, it’s even more critical to take measures to prevent the spread of emerging pathogens. Ending the importation of wild species into laboratories for experimentation is a key step. Long-tailed and rhesus macaques are the most commonly used nonhuman primates in experimentation, the most commonly traded primate species, and the species that harbors the highest volume of potential zoonotic disease.³⁸⁰³⁸¹ While primate suppliers and buyers claim to support efforts to reduce the use of wild-caught macaques in research, investigations have revealed that international suppliers have falsely labeled wild-caught macaques as captive-bred and sold them to laboratories.³⁸² This practice risks disease spillover and compromises the results of experiments conducted on these animals, whose health histories are unknown.

Macaques are often captured and imported from regions endemic for melioidosis, a life-threatening illness caused by Burkholderia pseudomallei. Although the Centers for Disease Control and Prevention (CDC) requires that monkeys imported from these regions undergo a mandatory quarantine, Burkholderia pseudomallei can remain dormant for long periods, and animals have been released into laboratories while still infected.³⁸³ Macaques have also been imported while harboring tuberculosis-causing mycobacteria.³⁸⁴³⁸⁵ According to the CDC, “In the United States, there is no centralized system for reporting TB in NHP that are not in CDC-mandated quarantine (minimum of 31 days after importation). Therefore, it is unknown how common TB is in NHP in the United States.”³⁸⁶

Ending the global trade of monkeys for experimentation would eliminate a major risk factor in zoonotic disease spillover, reduce the dissemination of unreliable data collected from animals of unknown origin, and stimulate the move toward human-relevant research methods. This is a critical step in protecting public health and preventing the next pandemic.

Despite over a thousand neuroprotective drugs showing promise in animal models, none have translated into effective human therapies for stroke.³⁹⁰ Our understanding of the biological processes driving human stroke recovery remains limited,³⁹¹ and developing accurate models of the central nervous system is challenging due to the complexity of the human brain. Current animal models, which primarily use rats, lack essential human characteristics, differ in stroke recovery compared to humans, and raise ethical concerns.³⁹²³⁹³ For example, ischemic stroke typically occurs in elderly patients with comorbidities, whereas experiments are predominantly carried out on young, healthy animals who often exhibit spontaneous recovery.³⁹⁴

Significant differences in brain composition—such as white matter making up 60% of the human brain but only 10% of the mouse brain³⁹⁵ —and variations in blood-brain barrier physiology³⁹⁶³⁹⁷ play crucial roles in stroke pathology. Additionally, differences in clot composition, neuronal function, and inflammatory processes among species further contribute to the poor translatability of animal models in stroke research.^398399400

A 2010 analysis of 16 systematic reviews (including 525 different studies) on human stroke interventions tested in animal models revealed that the efficacy of these experiments on animals was overstated by approximately one-third due to publication bias (the propensity of researchers and journals to publish results showing positive outcomes and omit studies with negative or null data).¹³ The authors noted that “participants in clinical trials may be put at unnecessary risk if efficacy in animals has been overstated.”⁴⁰¹

In silico modeling shows potential to replace animal experimentation in stroke research. Projects like INSIST “IN-Silico trials for treatment of acute Ischemic Stroke” use virtual patients to simulate stroke treatments, replicating clinical characteristics, such as clot properties, vessel geometries, and patient medical records.⁴⁰² These models, which allow for virtual drug testing and the detailed study of thrombosis and brain perfusion in humans,“ have the potential to lead to a more effective human clinical trial design, reduce animal testing, lower development costs, and shorten time to market for new medical products.”⁴⁰³ A groundbreaking in silico trial published in 2021 predicted aneurysm treatment responses using 164 virtual patients with 82 unique anatomies.⁴⁰⁴ This model outperformed experiments on animals, identifying new risk factors for treatment failure in days instead of decades. Virtual modeling can also assist patient-tailored clinical decisions for strokes and other neurological conditions. However, regulatory reform for in silico trials is urgently needed to advance the field.⁴⁰⁵

Researchers are also exploring new technologies and cell-based methods to enhance recovery by replacing damaged brain tissue with stem cells.⁴⁰⁶ Recently, stem cell therapy using patients’ bone marrow or allogeneic umbilical cord blood has shown improved neurological outcomes in clinical trials.^407408409410 In preclinical research, the isolation of human stem cells and hiPSCs has advanced the development of scalable human models in neurobiology.⁴¹¹⁴¹² Innovative 3D systems, like organs-on-chips and brain organoids,⁴¹³⁴¹⁴ may mimic complex cell interactions and in vivo physiology better than animal models, while 3D printing⁴¹⁵ enables the creation of detailed nervous system models for preclinical drug testing and clinical applications.

Accurately modeling ischemic responses requires understanding cellular interactions that influence blood-brain barrier permeability, cerebral edema, and neurovascular responses under pathological conditions. Because these interactions ultimately affect stroke outcomes, it is essential to create realistic models. Combining hiPSCs with advanced cell culture technologies has allowed researchers to replicate specific human nervous system features. For example, Kook and colleagues developed a vascularized model by coculturing vascular and cerebral spheroids generated by hiPSCs.⁴¹⁶ In another brain organoid study, Xu et al. observed morphological and synaptic changes in microglia cells after viral exposure.⁴¹⁷ Additionally, microfluidic models enable the use of patient cells and real-time monitoring of human brain dynamics, such as blood-brain barrier permeability and shear stress, which are not feasible in experiments using other species. Ex vivo brain slices are another valuable method for studying human brain tissue, as they preserve in vivo properties, spatial organization, and complex networks of various cell types.⁴¹⁸

In recent years, in vitro systems for studying strokes and the human nervous system have advanced significantly, becoming sought-after tools for studying human brain function and improving stroke treatment strategies.⁴¹⁹ Now that these tools are available, researchers must adopt them, and funders must support their uptake.

[A]nimal models cannot capture many key aspects of human brain disorders that may be caused by an SUD, which often involve the interplay of genetic, developmental, and environmental factors. … In addition, studying the brain in live animals involves invasive techniques that can affect the health and behavior of the subjects, potentially confounding results. … Consequently, it’s hard to translate research outcomes from animal models into effective clinical treatments for SUDs due to the inter-species differences in neuro systems between human and animal models.⁴²²

Several diagnostic criteria for SUD are impossible to model in animals since they require an individual to self-report. These include “(i) subjective craving, (ii) taking the substance in larger amounts or for longer than intended and (iii) wanting to cease or reduce substance use but being unable to.”⁴²³

Second, the pharmacokinetic actions of drugs differ among species. For example, “the rate of metabolism of MDMA and its major metabolites is slower in humans than rats or monkeys, potentially allowing endogenous neuroprotective mechanisms to function in a species-specific manner.”⁴²⁴ Pharmacokinetic differences between humans and “model” animals likely explain why the neurotoxicity seen in rodents after MDMA administration has not been observed in a clinical setting.⁴²⁵ Since MDMA is being explored not only because of its use as a recreational drug but also for its potential therapeutic use, accurate knowledge regarding its safety in humans is paramount.

Third, serious flaws in the experimental design of substance use experiments on animals skew the interpretation of their results. Unlike humans, whose experience with SUD is primarily shaped by individual choice to consume an addictive substance—often over other rewarding alternatives—animals in laboratories are typically not given this option. When they are, the majority will choose an alternative reward, such as sugar, over the drug.⁴²⁶ This holds for primates as well as mice and rats. Even among animals with a history of heavy drug use, only about 10% continue to self-administer the drug when presented with another rewarding choice.⁴²⁷ In a review on the “validation crisis” in animal models of drug addiction, it has been said that the lack of choice offered to animals in these experiments raises “serious doubt” about “the interpretation of drug use in experimental animals.”⁴²⁸

The nonhuman animal has been called a “most reluctant collaborator” in studying alcohol use disorder and exhibits a “determined sobriety,” which the experimenter must fight against to overcome “their consistent failure to replicate the volitional consumption of ethanol to the point of physical dependency.”⁴²⁹ National Institute of Mental Health researchers reason that “it is difficult to argue that [drug self-administration by rodents] truly models compulsion, when the alternative to self-administration is solitude in a shoebox cage.”⁴³⁰

Despite the epidemic of drug dependence and overdose in the U.S. and the prevalence of SUD research conducted on animals, there are only limited treatment options available for individuals addicted to opioids, nicotine, and alcohol, and no approved treatments for marijuana, stimulant, or polysubstance users.⁴³¹ Leadership at the National Institute on Drug Abuse has noted that pharmaceutical companies show little interest in investing in treatments for SUD due to the stigma and complexities of the disease.⁴³²⁴³³ While data from animal studies were once hailed as promising in certain drug classes and relapse prevention, most have either failed to be effective in human trials or were not tolerated well by humans.⁴³⁴⁴³⁵ Some researchers argue that “these failures illustrate the inability of animal models to capture the complex nature of addiction and its treatment” and that “findings from animal models of addiction have generated a misleading picture of the nature of addictive behavior in humans.”⁴³⁶

Non-invasive human and human biology–based research methods are now providing answers to questions that the use of other animals is fundamentally unable to solve. Rutgers University Robert Wood Johnson Medical School researchers authored a review article describing how hiPSCs can provide a “unique opportunity to model neuropsychiatric disorders like [alcohol use disorders] in a manner that … maintains fidelity with complex human genetic contexts. Patient-specific neuronal cells derived from [induced pluripotent stem] cells can then be used for drug discovery and precision medicine.”⁴³⁷

Forward-thinking scientists around the world are carrying out human-relevant, non-animal research on SUD:

• Researchers are using postmortem human samples to model changes in the brain and brain cells induced by SUD. For example, at the University of Texas Health Science Center and the Baylor College of Medicine, researchers engineered a novel hiPSC model of neural progenitor cells and neurons from postmortem human skin cells, directly comparing the new models to brain tissue from the same donors to model opioid-induced brain changes.⁴³⁸ Heidelberg University scientists conducted an epigenomics study on postmortem brain tissue from individuals with cocaine use disorder to understand how the disorder alters synaptic signaling and neuroplasticity.⁴³⁹
• A recent University of Pennsylvania study used 3D genomic datasets to sequence more than 50 diverse human cell types to identify genetic and cell targets that underlie SUD.⁴⁴⁰
• A multi-omics study conducted by a team of researchers across the U.S. as part of the Million Veteran Program used systems biology to reveal key genetic targets for new drugs to treat opioid use disorder.⁴⁴¹
• University of Central Florida researchers have developed a hiPSC model for studying opioid use disorder and opioid-induced respiratory depression to combat the opioid overdose crisis.⁴⁴²
• At North Carolina State University, scientists co-cultured human neurons to form assembloids used to understand single-cell human molecular responses to cocaine and morphine.⁴⁴³ Human-derived assembloids and organoids “show unique potential in recapitulating the response of a developing human brain to substances”⁴⁴⁴ and will also be helpful in studying in utero exposure to drugs of abuse.
• Research on better ways to treat human pain is crucial for reducing opioid use disorder incidence and relapse. Researchers at Queen’s University Belfast used in vitro and in vivo human neuronal models to study a molecular basis for the modulation of nociception in human peripheral nerves.⁴⁴⁵ Biotechnology companies like AxoSim, NETRI, and others have developed human neuronal in vitro models that can be used for human pain research.

In addition, the funds currently supporting ineffective and wasteful SUD studies in animals could redirected to support effective drug prevention, rehabilitation, and mental health programs.

A significant obstacle to using other species to study women’s health is the anatomy of the reproductive tract. For example, mice have a closed reproductive system with tightly coiled oviducts opening into the bursal space. In contrast, the human reproductive system is open to the peritoneal cavity. This allows endometrial cells, shed during menstruation, to flow backward (retrograde menstruation) into the peritoneal cavity. This retrograde menstruation is linked to the development and symptoms of endometriosis. “[F]rom a morphogenetic perspective Müllerian duct development differs considerably in mice and humans,”⁴⁴⁸ resulting in the development of fallopian tubes in humans and the Müllerian vagina in mice.

Endometriosis and adenomyosis are closely related gynecological conditions that cause pelvic pain, miscarriage, and infertilityand affect around 10% of women.^449450451 Despite being first described centuries ago, significant gaps in the diagnosis and treatment of these conditions are due to the incomplete understanding of underlying mechanisms⁵ that have been repeatedly investigated using failed animal models.

Human endometriotic lesions, which are not yet fully characterized, vary significantly in location, size, color, and depth.⁴⁵² Additionally, endometriotic lesions have distinct etiologies that are impossible to fully replicate in animal models, requiring invasive methods such as surgical engraftment, intraperitoneal injection, or direct tissue injection into the endometrium.⁴⁵³⁴⁵⁴ These artificial approaches often result in cellular contamination with non-uterine tissue and local inflammation in animals.⁴⁵⁵ Transgenic de novo mouse models rarely succeed in replicating endometriosis due to the lethal phenotypes often associated with knocking out essential genes.⁸ In addition, the long latency period required for endometriosis to develop—something unachievable with short-lived species like mice—underscores the fundamental limitations of animal models.

The process of menopause and its symptoms vary widely among women, primarily influenced by factors such as the remaining number of eggs in the ovaries, lifestyle, diet, and ethnicity.^456457458 During the menopause transition, fluctuations in estradiol levels in the perimenopausal phase can cause specific, complex, and protracted physiological, behavioral, and neurological changes⁴⁵⁹ that experiments on animals fundamentally fail to replicate.

The estrous cycle of other primates and rodents differs considerably from that of humans.⁴⁶⁰ The vast majority of nonhuman animals do not experience menopause, and their fertility patterns differ significantly from those of humans. Fertility decline can occur in mice as early as 8 months,⁴⁶¹ or about one-sixth of their potential lifespan. The menstrual cycle of other primates and rodents differs in length, hormone fluctuation, and the ways in which these hormones regulate the hypothalamic-pituitary-gonadal axis compared to humans.^462463464

Given the many biological challenges described above, researchers attempt to replicate menopause and uterine lesions in animals using unnatural methods. Ovariectomy—the surgical removal of ovaries—is considered the “gold standard” for creating these symptoms in animals, but the procedure is an invasive and clinically irrelevant method for inducing menopause. Menopause is a gradual transition—not an abrupt event—and animals do not experience the same symptoms as humans, such as brain fog or the continued release of androgens by the ovaries.⁴⁶⁵ Other animal models created by the chemical induction of premature ovarian failure are prone to experimental confounds, such as discrepancies related to the dose and duration of the treatment, the development of unrelated neurological issues,⁴⁶⁶ and the inability to model responses to drugs that may reverse premature ovarian failure in humans.⁴⁶⁷

Most experiments use young animals, such as young marmosets, whose physiology drastically differs from that of the aging humans they are supposed to mimic. Genetic patterns in the brains of these animals don’t align with those of humans in the menopausal transition, meaning cognitive decline caused by estrogen fluctuation and loss during this period cannot be replicated.⁴⁶⁸

To design more effective interventions, it is essential to deepen the understanding of human-specific biological mechanisms that affect women’s health and fund the tools necessary for this critical yet often overlooked research.

Collective efforts for phenotypic characterization and biobanking of human endometrial lesions,⁴⁶⁹⁴⁷⁰ combined with machine learning tools that analyze patient data and wearable devices to identify potential risk factors, can produce data that has been historically difficult to replicate using simpler in vitro models. In one study, researchers developed a unified predictive model for the diagnosis of endometriosis using a dataset of over 5,000 women.⁴⁷¹ The model analyzed more than 1,000 variables, including lifestyle, genetic variants, and medical history.

The limitations of experiments on animals and traditional in vitro models have driven the development of advanced microfluidics platforms that accurately recapitulate the human reproductive system.⁴⁷² These include the human placenta-on-a-chip, which allows for the study of maternal-fetal interface and pregnancy-related conditions,^473474475 and standardized hiPSC protocols.⁴⁷⁶ Another vascularized multicellular model effectively mimics the hormonal fluctuations of the human menstrual cycle,⁴⁷⁷ enabling the study of endometrial permeability to contraceptives and serving as a proof-of-concept for studying human embryo implantation, which is impossible to replicate using animal models. Ultrasonographic data has been used to build a 3D bioprinted endometrium for diagnosing congenital uterine anomalies.⁴⁷⁸ Recently, the Human Endometrial Cell Atlas was published as a new reference for studying endometrial transcriptomics and guiding the development of human in vitro systems.⁴⁷⁹

Shifting resources away from inaccurate animal models and toward improvement in patient care would also profoundly affect outcomes. A recent study highlighted that misinterpreted symptoms are a major contributor to delayed endometriosis diagnoses.⁴⁸⁰ To tackle this issue, the authors proposed a comprehensive approach that includes educating physicians, offering specialized courses for medical students, and integrating other healthcare professionals into the diagnostic and care processes.

The human menstrual cycle and endometrium are dynamic and unique to every individual,highlightingthe need to prioritize personalized approaches using patient-derived models. Non-animal methods can revolutionize women’s health research, offering more accurate models for disease study, drug testing, and precision medicine.

According to the United Network for Organ Sharing (UNOS), as of February 2025, over 108,000 people in the U.S. are waiting for organ transplants.⁴⁸¹ Despite this monumental and urgent need, the current system for managing, harvesting, and transporting human organs is highly inefficient. Human organs remain the most compatible and effective option for transplantation, yet inefficiencies in the system lead to the waste of many viable organs. Rather than resorting to genetically engineering, breeding, and killing pigs for organ harvesting, the focus should be on refining the Organ Procurement and Transplantation Network (OPTN), the current U.S. human organ donation system. Creating a separate xenotransplantation network would demand substantial government oversight and funding, adding complexity and potential inefficiency to an already challenging system. Instead, the most responsible and effective solution is to strengthen the current human organ donation process, ensuring patients receive the best possible transplant options.

Until recently, UNOS was the sole organization managing the OPTN in the U.S., but it has faced decades of criticism for poor management. A 2022 Senate Committee on Finance investigation revealed that organs procured by UNOS were often lost, damaged, delayed, or never collected.⁴⁸² A 2022 report by the National Academies of Sciences, Engineering, and Medicine concluded that the U.S. organ transplant system is inefficient, inequitable, and inconsistent and needs significant improvement.⁴⁸³ Human organ transplantation is a critical and, by nature, scarce lifesaving resource. Yet one in five donor kidneys and one in 10 donor livers were procured but never transplanted, primarily due to the systemic problems described above.⁴⁸⁴

Moreover, the current system often wastes already available organs. A study of kidney transplants from 2000 to 2015 found that in nearly 8,000 cases, one kidney was used while the donor’s other kidney was discarded, often due to minor differences from ideal kidney organ donation criteria.⁴⁸⁵ These discarded kidneys would likely function well, especially compared to long-term dialysis.⁴⁸⁶ According to Dr. Dalvin Roth, a Stanford professor and Nobel Prize recipient for his work on kidney exchange programs, transplant centers are pressured to reject kidneys because they are penalized for unsuccessful transplants.⁴⁸⁷ However, transplant centers are not penalized for rejecting kidneys.⁴⁸⁸ This system perpetuates the organ shortage as rejected kidneys may not meet an unrealistic threshold; considering the significant morbidity and mortality of long-term dialysis, transplants offer far greater benefits to patients.⁴⁸⁹ Reforming these criteria could significantly increase the number of available kidneys among other organs. 

In response, President Joe Biden signed the bipartisan Securing the U.S. Organ Procurement and Transplantation Network Act in 2023 to modernize the national transplant system.⁴⁹⁰ This legislation aims to ensure that patients receive high-quality human organs,⁴⁹¹ in contrast to animal organs, which harbor risks of rejection and zoonotic infections and raise ethical concerns. In August 2024, the Health Resources and Services Administration announced that the OPTN Board of Directors, which governs national organ allocation policy, would be separately incorporated and independent from UNOS.⁴⁹² This is a critical step toward improving efficiency, but additional efforts to expand and improve the OPTN are needed as human organs remain the best option for transplant patients.

Xenotransplantation introduces additional risks, including transmitting pathogens from animals to humans, a phenomenon known as xenozoonosis. The FDA has recognized this as a significant risk, particularly for transplant patients who are inherently and medically immunosuppressed.⁴⁹³ These infections could potentially spread to close contacts and the broader community, raising an ethical dilemma by pitting the duty to protect public health against the need to provide organ transplants for patients with end-stage organ failure.¹⁰ Despite genetically engineering animals, raising them in pathogen-free facilities, and undergoing pathogen screening, viruses such as porcine cytomegalovirus or porcine roseolovirus have been reported even after pre-transplant screening.⁴⁹⁴ In May 2022, a pig heart transplant recipient died two months after his operation.⁴⁹⁵ The autopsy revealed that the pig’s heart carried undetected porcine cytomegalovirus and may have contributed to an unforeseen and untimely death in an immunocompromised individual.⁴⁹⁶

Historically, studies involving xenotransplanted organs in live human participants were approved through the Expanded Access Protocol, colloquially known as compassionate use, and even this approach has drawn criticism. Scholars have argued that the FDA misused expanded access in this context because the program was never intended to generate evidence of safety or efficacy to support the initiation of a clinical trial.⁴⁹⁷ However, since February 2025, the FDA has allowed the use of xenotransplanted organs in clinical studies under an investigational new drug application.⁴⁹⁸ This pathway permits formal clinical testing beyond compassionate use and signals a shift toward structured research, designed to support potential regulatory approval of xenotransplanted organs as alternatives to human organs. As a result, more clinical trials are expected moving forward.

Rather than investing in clinical trials exploring xenotransplantation, greater resources should be directed towards other approaches to increase the pool of human organs viable for allotransplantation. In 2025, the Health Resources and Services Administration directed the OPTN to propose policies to strengthen Donation after Circulatory Death (DCD) operations.⁴⁹⁹ Future research on DCD should examine how to implement practices that promote transparency and build public trust, ultimately fostering broader acceptance of DCD as an ethical and effective strategy to increase the availability of human organs. Additionally, increased resources should support the evaluation of advanced organ preservation technologies, such as ex vivo normothermic perfusion, to reduce discard rates and improve post-transplant graft function.

In January 2026, the U.S. Department of Health and Human Services announced new awards to research teams using bioprinting technology and regenerative medicine to create personalized, on-demand human organs that do not require immunosuppressive drugs.⁵⁰⁰ The goal of this technology is to use either a patient’s own cells or cells from a biobank to rapidly produce immune-matched replacement organs. There is significant hope that technologies like this can help the medical field move beyond considering xenotransplantation and could even reduce reliance on the current allotransplantation system.

In addition to funding research aimed at identifying ways to increase the availability of human organs, the U.S. should make systematic reforms to strengthen the current allotransplantation system. For example, experts suggest adopting a “presumed consent” policy, recommended by a 2019 University of Michigan study.⁵⁰¹ In this system, organ donation is the “default” unless individuals opt out, a practice that has already increased donation rates in other countries.⁵⁰² Furthermore, the U.S. could implement approaches similar to those of European countries that prioritize broad access to human organs and maximize the efficiency of their organ donation and transplantation systems.⁵⁰³ Their success is driven by government commitment, an opt-out donation process, fostering a culture of trust and confidence in the system and establishing dedicated institutions at multiple levels. In addition, proper hospital reimbursement ensures that financial barriers will not impede participation.⁵⁰⁴ These measures expand access to human organs and improve the efficiency of the transplantation system. By committing to improving the current U.S. organ donation system, policymakers could increase access to lifesaving human organs without resorting to the ethically fraught, risky, and unnecessary practice of xenotransplantation.