Whether human, animal or engineered, cells that are grown in the laboratory and maintained (described as cultured) outside a living organism, have been used within science for generations. The first human cell line was obtained in 1951, during the treatment of a patient with cancer, Henrietta Lacks. These cells, called HeLa, have contributed to many research advances. Cells of interest can be extracted from living tissue to be examined more closely under carefully controlled conditions, allowing researchers to understand how tissue grows, study specific mechanisms within the cells, test the effect of substances on cells, generate pharmaceuticals based on living cells and organisms, and much more. 
The fact that cells are grown in a very controlled environment offers advantages in terms of reproducibility and consistency, which allow for precise quantification and validation of scientific findings. Cells are also cost-effective and a way to study human or animal health without ethical constraints. 
When it comes to regulatory testing, lab-based (in vitro) methods are already well-established and validated in assessing certain types of toxicity, such as for the skin, to test sensitivity and irritation. Here, different compounds that are found in common products, that may trigger an allergic reaction for instance, can be used to cause such a reaction in human tissue models, to understand the effect it will have. 
However, there are some key limitations of using in vitro, the major one being tied to the very nature of these techniques, as cells in a dish are unlikely to behave in exactly the same way as in a living organism or their natural environment. Cells are also grown flat in a dish, whereas in the body they are surrounded by other cells in a 3D arrangement, and they interact with their environment, engaging in complex communication with their surroundings and other cells and tissues. 

Organoids are small groups of cells grown in a dish that arrange themselves into a 3D system and mimic some features of an organ, offering valuable models for studying human biology and disease. As they are often derived from human stem cells, organoids allow researchers to observe certain biological processes directly in human tissue. Using specialised cell-culturing techniques, scientists can now grow organoids of almost any organ in the body, including the brain, heart, liver, kidney and lungs. 
However, the term “organoid” may be misleading. These structures do not replicate entire organs. It is very challenging to mimic the environment where these cells are found in the human body. Additionally, most consist of just a few of the overall types of cells existing in an organ and lack features such as communication with other organs, a working immune system, or fully functioning blood vessels. The latter has limited their growth, with cells at the centre of organoids dying due to a lack of oxygen and nutrients. 
In 2025, a major breakthrough in organoid research began to address this major limitation. Researchers from Stanford Medicine, US, combined blood vessel-forming cells, smooth muscle and human stem cells and were able to generate heart and liver organoids that were capable of forming their own blood vessels and, consequently, allow organoids to grow further before cell death began. This technology could be applied to improve the development of organoids, enhancing their contribution to disease modelling, therapeutic testing and organ replacement strategies.  
While organoids lack the structure of a proper organ, some advanced organoids can already partially mimic the complexity of certain organs. In 2025, researchers at the University of Southern California,US, developed mouse kidney organoids that could mimic complex structures and produce urine – a key function of the kidney.   
The team used mouse stem cells grown in a mixture of compounds that supported kidney organoid development and complex maturation. They also generated human kidney organoids, albeit not mature enough to produce urine.   

Organs-on-chips are small devices of a clear flexible polymer with channels and chambers designed to culture cells and closely mimic the human physiology of an organ. Organs-on-chips can be developed using material derived from a range of organs, including the heart, kidneys, liver and lungs, and have broad applications in drug development and toxicology. 
There have also been advances to better understand the different interactions between the organs of the human body. Organs-on-chips can be combined to create Multi-Organ-Chips (MOCs) to better study these interconnections and the exchange of materials – such as substances that break down drugs (metabolites).  
Even though MOCs are still far from reflecting a complete organism, the number of organs connected in MOCs is constantly increasing. Specific versions of MOCs allow for a rapid screening of different compounds as either toxic, or are promising for use in drug development. They have also been developed to study the liver in a way that provides a more cost-effective alternative. Drug development is often lengthy and expensive. By culturing liver cells in a chip together with blood vessels and immune cells, researchers could mimic certain aspects of liver toxicity in an automated system that allowed testing several compounds at the same time.  
Although organs-on-chips mirror aspects of both healthy and diseased tissue and organ function, there remains the challenge of replicating the physical characteristics of organs and the interactionsbetween the many different cells that make up a certain organ, as well as the interactions between different organs and how they respond to stimuli. 

Although computational models don’t generate biological data, they are crucial to take full advantage of the evidence obtained from non-animal and animal studies. As computers have developed, they have become a significant way to reduce reliance on animals in research in certain areas. Artificial intelligence (AI) and machine learning are helping to drive improvements in how researchers design their experiments, analyse data and predict outcomes. 
Data on toxicity can be obtained using different computational models – for example, models that can work out the potential toxicity for one species based on what is already known about another.Researchers at Chalmers University of Technology, Sweden, developed an AI method that improves the identification of toxic chemicals, based on their chemical structure, at an early stage, which holds the potential to reduce the reliance on animal tests.  
AI can also integrate data from different studies to help identify which procedures or experiments, that would otherwise involve animals, are unnecessary and can be skipped. Training AI for the best results can be done using both animal data, that incorporates findings from previous research, and human data, from results in the clinic or clinical trials. 
However, because AI’s predictive power relies solely on existing data and predefined algorithms determined by humans, it can miss unforeseen effects that would otherwise be identified in animal toxicology testing, as it cannot account for the biological variability that takes place in organisms. For AI to accurately predict the effect of a drug, science would have had to already arrive at a place where all human processes are fully described and detailed, and science is still very far from that. AI can therefore supplement the approval of new drugs and medicines, and contribute to the reduction of animals used, but does not meet all the requirements for safety and ethical approval that would bypass the need for animals.  

With the development of non-animal techniques, animal studies to assess the safety of cosmetics have increasingly been replaced by these alternative methods. By 2004, animal tests were no longer performed for cosmetic products and a ban has been in place since 2004. The ban was extended for ingredients of cosmetics in 2009. Since 2013, a full EU ban on animal-tested cosmetic products, as part of EU Regulation 1223/2009, applies also to imported cosmetics. Before these bans were implemented, safety assessments involved the use of animals, mainly rodents and rabbits.  
However, chemicals used in cosmetics may also be used in other products with a different type of exposure or higher concentration. Industrial workers in factories producing cosmetics are exposed to the ingredients in cosmetics for longer and at larger quantities than consumers. This is why chemicals used in cosmetics can also be subject to the REACH Regulation, which requires companies who make and supply chemicals to provide information about the properties of these substances, to control the risks they may pose to humans and environmental health. This means that, under REACH, certain cosmetics ingredients still need to be tested on animals to assess the hazards for both the production workers and the potential effects on the natural environment. 

NAMs offer great promise in toxicology testing since their introduction can potentially mean obtaining data more predictively for effects on humans, compared with animal studies, and cheaper and faster results.  
Examples of successful replacement of animals by NAMs include the rabbit pyrogen test, which is used to detect particles derived from microbes or chemicals (pyrogens) that can cause a fever (an inflammatory immune response), and is currently being phased out. This is largely because alternatives to the RPT, such as the cell-based monocyte activation test (MAT), have demonstrated that they can be used successfully instead of rabbits.  
However, there are only a few instances where NAMs exist as a direct replacement for animals in regulatory testing. These specifically are in areas that assess fast-acting and short-term effects only, such as eye irritation or skin irritation, corrosion and dermal absorption. Here, NAMs are also showing their value in predicting how toxic a chemical will be, based on the chemical’s structure itself – certain chemical structures are known to cause negative reactions in people – by using AI to assess thousands of chemicals at a time. 
These methods are still far from being routine and it is noticeable that NAMs have been adopted more often for external aspects of the body, such as eyes and skin, rather than for methods that assess the internal organs. For these more complex measurements such as phototoxicity (light-triggered tissue damage), genetic toxicology (how chemicals damage the DNA) and endocrine disruption (how substances interfere with the effect of hormones), both NAMs and animal tests are part of safety testing. While certain NAMs may be able to replace and reduce the number of animals used in testing, there will also be cases where some findings obtained using NAMs will still need to be verified in animals. For example, if there is limited information available on the mechanism by which a particular NAM works to give the intended effect, comparisons will need to be drawn from a model where this data is available – often animals. 

NAMs have great promise in personalised medicine, since clinical scientists can collect cells from a patient and develop a NAM with their own cells, which can recapitulate some of the individual responses and predict how a specific patient will respond to different therapies.  
For instance, patient-derived cell cultures – created from a patient's tissues – retain the genetic and molecular characteristics unique to that individual and are used to develop so-called personalisedmedicines. It is hoped that these cultures will model a patient’s response to targeted therapies, such as for cancer treatment, to sufficiently help predict the most effective drugs, improving treatment success and minimising side effects.  
There has also been progress in creating patient-derived organoids (PDOs) and organs-on-chips, allowing researchers to test a myriad of drug responses and interactions, for example on patient-derived tumour tissues, in a laboratory setting, helping doctors choose the most efficient treatment and increasing their safety when used in that specific patient.  
Although NAMs can’t fully account for the systemic interactions and the complexity of being integrated in the body, examples show their value in fields such as personalised oncology. In 2025, researchers at EARA member Research Institute of the McGill University Health Centre, Canada, and Harvard University, US, created an organoid from biopsy sample cells taken from patients with an adenocarcinoma with a high mortality rate. Then, the researchers circulated chemotherapy drugs in the concentration and exposure times that would be given to the patient through the organ chips, aiding their clinical decisions. 

NAMs are also used for greater scientific understanding (basic research) and translating basic research to develop new drugs and treatments. Through a combination of techniques and information, gleaned from both animal and non-animal studies, researchers can build the best possible understanding. 
NAMs are limited in the complexity of information they can provide about the living body, but often they can model more faithfully certain aspects of some isolated human biological systems, closed mechanisms and predict interactions, helping to improve the translation of research findings from the lab to the clinic.  
Artificial intelligence can simulate some biological processes or even how certain diseases progress, leading to the identification of possible drug compounds, as well as predictions about how they might behave once in the body. A study led by the Spanish National Research Council trained an AI system to detect brainwaves in monkeys that are disrupted in neurological disorders such as Alzheimer’s, based on brain recordings from mice. Without artificial intelligence these brainwaves could otherwise be missed by standard imaging techniques, contributing to the improvement of the diagnosis of brain diseases.  
Meanwhile, cells grown in the lab – from human cells to more complex organoids and organs-on-chips – have been developed for a range of different diseases and conditions, including neurodegenerative diseases and cancer, and are showing promise for studying disease mechanisms, drug responses and developing personalised medicines – and in some cases reducing or replacing the use of animals.  
In a study at the University of Manchester, UK, researchers used lung organoids, produced from human stem cells, to investigate the possible effects of carbon-based nanomaterials on human health, with the results mirroring the negative effects on lung health shown in animal studies. Alternatives have also been used in basic research, such as in work at the University of Birmingham, also in the UK, that developed an organ-on-a-chip which replicated the blood vessels in the human liver, allowing scientists to understand how immune cells reach liver cells. 
NAMs, while they do not reach the full maturation of an organ inside a living organism, provide an important window into early developmental stages. By trying to replicate the development of an organ in a controlled, simpler system, scientists can pinpoint the steps involved in organ development and they can understand how certain substances affect it. Because these models are often human-based, they can capture mechanisms that are specific to humans. In 2025, researchers from IBEC, Spain, using a platform that mimicked the wall of an uterus, were able to capture the first video of a human embryo implantation, which could help improve understanding of the natural process and improve fertility and assisted reproduction techniques.