Are Animal Models Necessary? Exploring (Dis)advantages and Alternatives

Animal models have been crucial for scientific development, allowing researchers to understand the underlying mechanisms of various human conditions, and are far from becoming obsolete in scientific research. However, the ethics of animal experimentation has been a prevalent question between both experts and nonexperts. This essay tackles the advantages and disadvantages of the usage of animal models while delving into new alternatives that have emerged in light of contemporary science. Animal models have been pivotal in several biomedical research fields (including neuroscience and neurobiology, pharmacology and toxicology, and many others), helping develop therapeutic approaches to several human conditions, such as vaccines and new drugs. Given their unique characteristics and, for most models, their close connection to human genomics and physiology, several species have been used: nonhuman primates, rodents, dogs, birds, ruminants, pigs, among others (Doke and Dhawale 2015; Domínguez-Oliva et al. 2023). Although animal models have been crucial for the research of prominent and increasing diseases that affect global human health (Domínguez-Oliva et al. 2023), many ethical questions still arise when considering studies and experimental techniques involving animals. On the other hand, animal models present methodological challenges that could be avoided with strictly controlled standardized alternatives (e.g., diverse animal species and disease induction models with varying degrees of resemblance to human conditions; variations in drug dosage, efficacy and toxicity) (Robinson et al. 2019). This creates a necessity for innovative approaches to animal experimentation in modern-day research. This review explores the necessity of animal experiments and their implications on ethical regulation, as researchers begin to consider other rising alternatives to animal models. Animal models have been instrumental in the research of human and/or animal diseases, by mimicking these pathologies and, thus, providing novel therapeutic approaches, diagnostics techniques and surgical procedures (before advancing to clinical trials) (Doke and Dhawale 2015; Domínguez-Oliva et al. 2023). The discovery of vaccines and, more recently, the significant advances in worldwide diseases (such as COVID-19 or cancer) were made possible thanks to the usage of animal species (Domínguez-Oliva et al. 2023; Mukherjee et al. 2022). In neuroscience, several studies have been conducted on animal models focusing on specific pathologies of the nervous system, namely, epilepsy, Alzheimer's and Parkinson's disease (Domínguez-Oliva et al. 2023). Alongside these advances, the improvement and testing of new therapeutic options, which are also suited for animal models, are another emerging field (Domínguez-Oliva et al. 2023). This highlights animal models as essential tools for preclinical drug screening, serving as primary in vivo models for assessing pharmacokinetic parameters such as drug efficacy, safety and toxicity (Mukherjee et al. 2022). Although animal models have become indispensable tools to allow research of disease instalment and progression, they present some limitations, as they may not be able to accurately capture the complexity and nature of human conditions (Neurath 2012; Rieder et al. 2012). Albeit this, there remains an ongoing debate regarding the significance of animal models, which gives rise to moral and ethical questions and to the necessity of further evaluating the scientific necessity of animal experimentation. Given their remarkable contribution, it is unlikely to preview the fallout of animal experiments in modern science. However, along with the worldwide exponential increase in scientific research and publication output (National Science Board 2023), it is important to note that the number of animals used in research has gone up significantly in recent years, raising questions about the ethical aspect of animal experimentation (Doke and Dhawale 2015). For example, the European Commission reported that around 9.58 million animals were used for ‘research and testing’ in the year 2017: 45% of these animals were used for basic research (4.3 million uses), of which almost 1 million uses were directed to nervous system research (Directorate-General for Environment (European Commission) 2020). These rules keep being brought up to researchers, in hopes of gradually making them a worldwide reality. More recently, in 2010, Directive 2010/63/EU was approved by the European Parliament, establishing welfare criteria for animal usage across all testing stages (Directive 2010/63/EU 2010; Kanďárová and Pôbiš 2024). Despite all efforts to implement and integrate clear standards and procedures, there are still unanswered questions and a rising call for the development of new alternatives to animal experimentation. As previously mentioned, new approaches like genetic engineering, artificial intelligence, organs-on-chips, mathematical models and advanced machine learning technologies are being driven by ethical pressure, as well as the implementation of initiatives to decrease and replace the number of animals used in experimental protocols. This multimodal strategy is thought to be the most effective way to handle the moral conundrums that come with employing laboratory animals while highlighting their important contributions to both human and animal medicine (Domínguez-Oliva et al. 2023). In vitro models have been widely used to study disease pathogenesis, as they create the possibility to target specific mechanisms and investigate possible treatments (le Feber 2019). These systems may consist of two-dimensional (2D) cultures of immortalized cell lines or primary cells, which involve their isolation, maintenance and expansion (growing as a monolayer) in a laboratory environment (Doke and Dhawale 2015; Gibbons and Dragunow 2010). In brain research, in vitro culture of adult human brain cells has been fundamental to studying the (patho)physiology of the brain. Conversely, organs, slices or multiple cell types can be cultivated to high density in a three-dimensional (3D) matrix, constructing replicates of a portion of the tissue in study (Gibbons and Dragunow 2010). In the last few years, organoids have been developed based on the expansion potential of stem cells in culture, particularly embryonic stem cells and their synthetic induced pluripotent stem cell counterparts. In the case of brain organoids, these can reach up to a few millimetres and include several brain regions (Clevers 2016; Qian, Song, and Ming 2019). Furthermore, microphysiologic systems (MPS), which include organotypic cultures, organ-on-chip (OOC) technologies, scaffolds and other 3D cultures, have been helpful in the research of organ function and diseases (Blutt et al. 2017), as they allow the study of interactions among various cell types within a 3D structure that more closely mimics the actual organ environment (Barreras et al. 2023). By combining these efforts, neuroscientists have tried to understand the development and pathogenesis of human conditions (e.g., Alzheimer's disease; Blanchard, Victor, and Tsai 2022) by creating in vitro models of the brain, as they are becoming more complex and increasingly able to provide insight into the molecular pathways and components of disease conditions. As with in vivo models, there are many benefits to employing in vitro models, but it is important to acknowledge their constraints as well, particularly the fact that not all clinical symptoms of the illness can be replicated in in vitro models (le Feber 2019), which not only complicates the translation of findings into the clinic but also hinders the identification of underlying mechanisms for in-depth research. Nevertheless, complementary information offered by these models can help reduce and improve the usage of animal models in neuroscientific research, as well as many other fields (Kaplan et al. 2024). Invertebrates, such as Caenorhabditis elegans (worm) and Drosophila melanogaster (fruit fly), are commonly used as substitutes for mammals due to their brief life cycle, small size, simple anatomy and mouldable genetics, making them cost-effective and suitable for mass culture, with reduced ethical concerns (Austad 2009; Doke and Dhawale 2015; Søvik and Barron 2013). Given their simplistic nervous system, with minimal number of neurons connected through defined networks, invertebrates have become more relevant in neuroscience research: There are over 15,000 articles in PubMed connecting the usage of invertebrate models and nervous system research, from the last 10 years alone (Romanova and Sweedler 2018). However, these models are insufficient for translating studies into human diseases because invertebrates have undeveloped organ systems, lack adaptive immune systems (Doke and Dhawale 2015) and do not have homologous human counterparts (Søvik and Barron 2013). Despite these limitations, these simpler models have been key to many new discoveries and, in neuronal diseases, in particular, have been crucial in understanding conditions like Parkinson's (Surguchov 2021; Zhang et al. 2024) and Alzheimer's disease (Link 2005; Wu et al. 2023). As such, realizing their potential could offer significant benefits to comprehend the molecular pathways behind human diseases and disorders (Søvik and Barron 2013). In silico models have been used in a variety of industries, with a wide range of tools that allow the rapid increase of structural, chemical and biological data available. Recently, with the rapid advancements in computer technology, there have been more initiatives in the development and adoption of in silico methods, more transparent and driven by clear mechanistic insights. These methods have also provided easier access to an increasing amount of data, catalysing the emergence of innovative technologies and tools (Madden et al. 2020; Pinzi and Rastelli 2019). By quicky analysing information (including from ongoing experiments), in silico models are able to provide predictions, while learning and adapting its methods, based on pattern recognition from input data (Vu et al. 2018). Some examples are physiologically based pharmacokinetic (PBPK)—comprised of multiple compartments representing various body tissues, interconnected by blood and lymph flow, that can predict drug concentration-time profiles and exposure in blood and tissues (Kuepfer et al. 2016; Li et al. 2017; Zhuang and Lu 2016); quantitative structure–activity relationship (QSAR)—a ligand-based method that utilizes mathematical statistics to develop models, establishing a relationship between the structure and the activity of a compound (Neves et al. 2018; Wang et al. 2021); and molecular docking—originally designed to study molecular recognition between a small molecule (ligand) and a larger molecule (protein) and now widely used to assist different tasks of drug discovery programmes (Pinzi and Rastelli 2019; Stanzione, Giangreco, and Cole 2021). These models are already well implemented in pharmaceutical research and development, namely, in bioactivity profiling, in silico dosimetry and predictive toxicology (Kaplan et al. 2024). In recent years, many initiatives have been implemented towards incorporating in silico technologies in neuroscientific research, ranging from brain network modelling (Pronold et al. 2024) to genetic screening in neurodegenerative diseases (Libonati et al. 2024). Together, given their time-effectiveness, low costs and absence of ethical considerations, these advancements are pivotal in pushing a shift away from animal testing towards scientifically robust alternatives, which can be mechanistically interpreted, and are relevant across different species (Madden et al. 2020). However, the applications of these models are still limited, partially due to the complexity of in vivo transport mechanisms, which are not adequately represented (Li et al. 2017). Additionally, there are problems with the in vitro systems' ability to replicate the in vivo conditions accurately (Neves et al. 2018). Therefore, these models should be viewed as a first step or as a component of a workflow, rather than as the end product. Thus, by addressing data gaps, assisting in more efficient chemical screening (finding candidate chemicals with superior safety profiles during new product development) or forecasting toxicokinetic parameters (avoids animal testing for bioactivity) (Kaplan et al. 2024), computational models can reduce the necessity for animal model experimentation. The recent advances in biometric data, algorithms and computational pathology allow us to anticipate the implementation of artificial intelligence (AI) approaches in the near future. As technology rapidly develops, scientists are now able to obtain great amounts of data across all kinds of research fields. Particularly in neuroscience, there has been a growing interest to gain more insight into brain function by extrapolating this vast volume of new information data into new measurement techniques and experimental paradigms (Vu et al. 2018). Furthermore, as datasets are growing rapidly and becoming more complex, machine learning techniques become more accurate (Glaser et al. 2019). In the United States, the National Center for Toxicological Research (NCTR, FDA) is developing initiatives based on AI in hopes of diminishing animal usage (Kaplan et al. 2024). For example, ToxGAN aims to predict experimental results of new compounds based on existing animal toxicogenomic data without actual animal testing (Chen et al. 2022). Another approach, SafetAI, strives to identify safety issues, based on publicly available data from in vivo testing using mice and rats (Li et al. 2021). With contemporary technological advancements, AI's learning algorithms have the potential to enhance disease monitoring, refine diagnostic accuracy, discover new treatments (Fogel and Kvedar 2018) and aid in a variety of tasks, including precise measurements of immunohistochemical biomarkers, cell counting and analysing tissue properties and architecture (Sandeep et al. 2024). However, as these models gain more recognitions, there is a growing concern that AI-powered technologies may outperform humans (Fogel and Kvedar 2018). Ultimately, the multifaceted and broad applicability of AI shows great potential to fulfil the demands of modern science (Sandeep et al. 2024) and pave the way for truly personalized medicine (Fogel and Kvedar 2018). Animal models are far from being replaced because they can mimic human conditions and pathologies and, as such, have been responsible for many discoveries in the biomedical field. Animal experimentation is conducted in compliance with government legislation, which reinforce the regard for animal welfare and the correct training of professionals (Barré-Sinoussi and Montagutelli 2015). Nonetheless, there are still rising ethical concerns relating to the usage of animals in research. For this reason, there has been an increasing effort to follow the 3R's rule to reduce animal harm to the lowest possible amount. Alternative models have been created, which can be used under controlled conditions and allow the implementation of the ‘replacement’ rule. However, although they are rapidly becoming more sophisticated and have supported major contemporary scientific discoveries, these models have not yet been perfected, as they do not accurately represent the complexity of human conditions (see Table 1). Particularly in neurosciences, the intricate nature of the human brain hinders the creation of an effective in vitro model (Benam et al. 2015). Many methods overlook the intricate interactions of the CNS, failing to take into account several biological and physical inputs. Furthermore, there is unfortunately little in vitro translation due to our still lacking understanding of how different cell types interact within the CNS in vivo (Walczak et al. 2021). This further underscores the need for whole-organism experimentation, in order to characterize physiological functions and systemic interactions (Barré-Sinoussi and Montagutelli 2015). As such, nonanimal models should be used as a starting point in workflow protocols, as long as researchers are well informed and educated on the correct methodology and interpretation of in vitro and in silico results. Although alternatives are still being assessed and slowly implemented, it is important to highlight that animals used for research should always be free from unnecessary pain or distress and placed in proper housing. Therefore, animal experimentation can be continuously improved and carefully planned (Barré-Sinoussi and Montagutelli 2015), using other alternative models to optimize experimental design (Glaser et al. 2019). Ana Isabel Guimarães: writing–original draft. The author declares no conflicts of interest.