For many years, the assessment of chemical and pharmaceutical safety has largely relied on traditional animal models and conventional 2D in vitro cell cultures [13]. While these methods are essential, their inherent limitations have become more evident, prompting a thorough reevaluation of their ability to forecast human health [14].

Traditional animal models present several notable limitations, often considered the gold standard. The 2-year rodent bioassay exemplifies a study that requires significant resources and prolonged observation periods, making it well-known for being time-consuming and expensive [15]. The urgent need to minimize, improve, and replace their use is motivated by significant ethical issues related to animal welfare, as well as practical and financial challenges [16]. Significantly, the predictive accuracy often falls short because of the physiological and immunological differences that exist between animal species and humans [17]. Over 90% of drugs deemed safe and effective in animal studies ultimately do not succeed in human clinical trials, often due to unforeseen toxicity or lack of efficacy. The estimated average cost for developing monoclonal antibodies ranges from $650 to $750 million, highlighting the challenges posed by a high attrition rate, which presents both regulatory issues and substantial financial implications [18]. Specific historical examples that underscore the significant shortcomings in interspecies extrapolation involve the teratogenic effects of thalidomide, which were not foreseen despite extensive animal testing [19]. Furthermore, the immunogenicity observed in animals often fails to align with that in humans, which adds further complexity to safety assessments [20].

While conventional 2D in vitro cell cultures offer cost-effectiveness and high-throughput capabilities, they also possess inherent limitations [21]. Unlike their natural in vivo habitat, cells grown in a monolayer on flat, rigid plastic surfaces exist in an artificial environment. This simplified environment lacks the complex tissue structure, 3D cellular interactions, and crucial physiological gradients (including waste products, nutrients, and oxygen) that are necessary for cellular response and function in a living organism. Consequently, in comparison to in vivo findings [22], results may exaggerate or misrepresent toxic effects, as cells in 2D cultures often exhibit modified morphology, polarity, gene expression, and metabolic profiles [23]. The simplification of biological complexity in 2D in vitro systems, along with the challenges of inter-species extrapolation in animal models, consistently demonstrates that traditional models often fall short in predicting human responses [24]. This synthesis of various issues highlights the intricate nature of toxicology assessment, indicating the need for a fundamental shift in approach rather than minor improvements to existing frameworks [25].

The deficiencies of conventional approaches have directly driven the urgent advancement and regulatory acceptance of NAMs [26]. These novel methodologies include a wide range of non-animal-based techniques, such as advanced in vitro human-based systems [including cell assays, organoids, and microphysiological systems (MPS)] and in silico (computational) models [4]. The “3Rs principle” provides the fundamental motivation for the development of NAMs, which is to replace, reduce, and refine the use of animals in scientific research and testing [16]. This transformation is not merely a scientific preference; it is a strategic business and ethical imperative that has the potential to revolutionise the entire drug development landscape [27]. These sophisticated models have the potential to significantly reduce research and development costs, accelerate timelines, and enhance predictive accuracy [28]. The schematic representation of biomaterial-assisted models for immunotoxicology and carcinogenicity assessment is shown in Figure 2.

In order to successfully build physiologically appropriate 3D microenvironments that closely resemble the original extracellular matrix (ECM) and facilitate intricate cellular interactions, these sophisticated in vitro models must be developed [29]. Biomaterials are the basic components that offer structural support, biological signals, and mechanical qualities to these humanised systems [30]. The urgent development and regulatory push for NAMs is directly driven by the significant limits of conventional approaches, and biomaterials are emerging as a key enabling technology for designing the required physiological complexity and human relevance in these novel models [5]. Because of this causal link, human-relevant NAMs are required when standard models fail, and biomaterials are crucial for creating these sophisticated models. This shows how the use of biomaterial-enhanced NAMs is changing the whole drug development landscape and is not only a scientific preference but also a strategic commercial and ethical requirement.

The transition from 2D to 3D cell culture systems has been instrumental in the development of in vitro models that are more physiologically pertinent [22]. These cutting-edge systems are designed to replicate the intricate cellular and tissue architecture that is present in vivo, thereby providing a more precise substrate for the investigation of drug responses and disease mechanisms [22]. A comparative overview of conventional and advanced models used for toxicology and carcinogenicity assessment is provided in Table 1.

The relationship between physiological relevance and screening throughput across various 3D models is summarized in Figure 3.

These are microsized aggregates of closely-packed cells that comprise a fundamental 3D model in cancer research due to their simplicity [31].

Spheroids can be homotypic, which means they are made up of only one type of cell, or heterotypic, which means they are made up of more than one type of cell, including immune cells, fibroblasts, or endothelial cells [35]. The straightforward production and handling of spheroids make them suitable for high-throughput screening (HTS) when they come from established cell lines [36]. They effectively mimic or successfully recapitulate important in vivo features, such as creating nutrient and oxygen gradients and forming separate cell layers-a growing outer layer, a resting middle layer, and a hypoxic/necrotic core-which makes them more resistant to drugs than 2D cultures [37]. This layered structure is important for reliable investigations of medication penetration and effectiveness because it mimics the problems that drugs experience in solid tumours in living organisms [38].

Organoids are 3D cell cultures that self-assemble into miniaturized organs. They are derived from various stem cell sources, including embryonic stem cells, induced pluripotent stem cells (iPSCs), adult somatic stem cells, or primary tissues [32, 39]. They precisely replicate the structure, function, and cellular diversity of natural organs, encompassing the intricate variations observed in tumors [40]. Patient-derived organoids (PDOs) are effective instruments for personalized treatment, as they retain the functional, genetic, and intrinsic variations among tumors. This capability enables clinicians to predict the efficacy of therapies and the potential for resistance in individual patients. The capacity for personalization is a critical element in the development of advanced in vitro models. The study emphasizes the development of patient-derived systems capable of capturing and predicting individual responses to treatment and the progression of their conditions [41].

OoC platforms signify a significant advancement by integrating dynamic fluid flow and mechanical forces, such as cyclic strain in lung chips and shear stress in vascular channels, into 3D cultures [42]. These systems are generally constructed from polydimethylsiloxane (PDMS) or alternative polymers and feature microfluidic channels populated with living cells. OoCs provide a significant advantage by effectively modeling interactions between multiple tissues and their systemic effects. A liver-tumor-on-a-chip model can effectively simulate the first-pass metabolism of a prodrug and its subsequent impact on cancer cells [43]. Linking an immune component, such as a synthetic lymph node, to a tissue model facilitates the investigation of immune cell trafficking and systemic immunotoxicity, which static models are unable to achieve [44]. Although currently less suitable for HTS compared to spheroids, advancements in automation and multiplexing are swiftly mitigating this limitation.

Co-culture models are indispensable for recapitulating the complex interplay between different cell populations and the ECM [45]. These assays, which permit the concurrent cultivation of multiple cell types (e.g., immune cells and cancer cells, fibroblasts and epithelial cells), are critical for investigating intercellular signaling pathways [46]. Such systems are fundamental to understanding the TME, a dynamic ecosystem defined by reciprocal communication between neoplastic cells, stromal support cells, and immune infiltrates [47]. Consequently, they provide a more physiologically relevant context than monoculture approaches, where cells are grown in isolation.

The sophisticated in vitro models mentioned earlier have extensive uses in toxicology, especially in immunotoxicology and carcinogenicity evaluation [7].

3D cell cultures continue to face challenges in gaining widespread acceptance, despite their successes. The production of varying findings by different laboratories presents a challenge in terms of comparing work and obtaining regulatory approval, as there is no single best practice for the establishment and maintenance of these cultures [8]. Data collection and verification are difficult due to the extreme complexity of 3D models, which necessitate sophisticated tools and computational models, despite their usefulness for real-world relevance [56]. Furthermore, the utilization of these complex cultures may be restricted, particularly for large-scale initiatives, due to the time and expense necessary to acquire and maintain them. Organoids are highly appropriate, but their variability is a significant challenge for mass screening due to the complexity of their natural detail and the difficulty of generating uniform forms from a limited number of cells [57]. Many organoid settings still lack the comprehensive immune-defense component necessary to comprehensively test novel immune treatments and cancer-immune conflicts [58]. A significant judgment must be made regarding the transition from basic spheroids to more complex systems, such as multi-OoC. Spheroids are renowned for their simplicity and efficacy in mass screening [36], whereas organoids, with their complexity and diverse applications, are difficult to integrate into mass screening due to their fluctuations [57]. This illustrates that 3D models that are more realistic and complex are more costly, more challenging to disseminate, and less suitable for mass screening. Consequently, the decision-making process is complex and contingent upon the study’s requirements and stage [59]. A comparative summary of the advantages, limitations, and applicability of these advanced 3D in vitro models is provided in Table 2.

Robust filaments create a network in the ECM. It comprises adhesive proteins, substantial proteins, and collagens [62]. The body’s cells get crucial signals from this network [30]. Biomaterials are meticulously designed to emulate these complex components. High-level laboratory models offer the requisite shape and function as a framework [5].

The ingredients that are used to create the biomaterial are sourced from both the laboratory and natural sources.

Natural polymers: Collagen [64], fibrin, HA [65], and gelatin [66] are all substances that are well-absorbed by the body, do not induce numerous immunological issues, bind to cells [67], and are degraded with the help of enzymes. These fragments have the potential to transform into polymers that resemble the squishy, moist regions where cells reside in our bodies. This assists cells in maintaining their wrapping and acquiring nutrients.

The composition of biomaterials affects the behavior of cells [71].

Porosity: The pores in biomaterial grids are just the correct size to allow food to pass through, waste to be eliminated, and cells to adhere to one another to build 3D structures. Cell migration and development are determined by the hole diameters and their connections.

The ECM’s mechanical properties change throughout time and have a big impact on how cells develop and work. Biomaterials are made to copy these mechanical signals [73].

Stiffness: Hydrogels, a common type of biomaterial scaffold, can be tuned across a wide physiological range of stiffness, from extremely soft (mimicking brain tissue at ~0.1 kPa) to highly rigid (mimicking bone at ~10 MPa) [74]. This property is a critical regulator of cancer progression. Matrix stiffness directs cancer cell morphology, epithelial-to-mesenchymal transition (EMT), and the proliferation, invasion, and drug resistance of cancer stem cells. For instance, increasing matrix rigidity can induce irregular spheroid morphologies and promote the widespread dissemination of pancreatic cancer cells [74]. Stiffer microenvironments, particularly when combined with biochemical cues like growth factors, can also drive invasive phenotypes.

The precise regulation of porosity, topography, and stiffness is paramount for designing biomimetic microenvironments. These parameters are not intrinsic but are actively engineered through specific fabrication techniques and material choices. Control strategies include modulating polymer concentration, crosslinking density (chemical or physical), and employing advanced fabrication methods like electrospinning and 3D bioprinting. The following table summarizes key parameters and methods for controlling these critical properties. The principal parameters and techniques for modulating porosity, topography, and stiffness in biomaterials are outlined in Table 3.

In addition to providing physical support, biomaterials deliver specific biochemical signals.

Biomaterials can be designed to contain growth factors (e.g., BMPs, TGF-β, FGF-2, VEGF) and establish small chains (e.g., the RGD sequence) that alter cell function, assist in cell adhesion, growth, and the formation of new blood vessels [76]. The cells are provided with these clues in a specific manner, which replicates the endogenous signals present in the body. Motifs that are degradable proteases can be incorporated into synthetic biomaterials, enabling a cycle of ECM change and cell movement. This replicates the natural changes in body tissue and enables cells to modify their own environment [75].

Complex methods of producing biological materials, such as 3D bioprinting [77], are not merely instruments for producing objects. They are inextricably linked to novel biomaterial designs, such as specialized bioinks [78]. This link enables us to oversee the construction and composition of objects at unprecedented levels. Consequently, we can develop more accurate and realistic models for laboratory testing.

Moving from stationary frames to open systems that better adapt to the evolving behaviors of living tissues, the future of biomaterials in toxicological models seeks to create smart and dynamic microenvironments that can act upon and modify biological responses.

Biomaterials are being developed to establish human-specific immune microenvironments in vitro, as animal models are not reliable predictors of human immune responses due to inherent species differences, particularly in immunogenicity to human monoclonal antibodies [20]. This is an essential step towards the development of immunotoxicity assessments that are more human-relevant and reliable.

Biomaterials can be engineered to facilitate the delivery and support of immunomodulatory stromal cells, including fibroblastic reticular cells (FRCs). These cells can enhance immune regulation by influencing T cell activity and promoting tolerogenic phenotypes. For instance, 3D macroporous gelatine scaffolds have been demonstrated to promote the survival of FRCs in vitro, augment the expression of their phenotypic markers, and increase the secretion of anti-inflammatory cytokines such as TGF-β2. These scaffolds resulted in decreased cytotoxic T cell activity and the promotion of regulatory and anergic T cell phenotypes when co-cultured with diabetogenic T cells. These phenotypes are associated with immunological tolerance rather than autoimmune activation. This illustrates the ability of biomaterial architecture to influence the behaviour of FRC and the interactions between immune cells. Additionally, the behaviour of a variety of immune cells, such as macrophages [93], dendritic cells, and T cells, can be actively influenced by biomaterials. By precisely regulating surface properties, including charge, hydrophobicity, and roughness, and by regulating the release of immunomodulatory factors, including anti-inflammatory cytokines, this modulation is accomplished [94]. This represents a shift from the passive observation of immune responses to the engineering of them in a controlled in vitro environment, a significant innovation for the accurate assessment of immunotoxicology.

The interaction between biomaterials and the host immune system is intricate and multifarious, involving dynamic phenotypic shifts, receptor engagements, and intricate signalling pathways (e.g., M1/M2 macrophage polarisation) [93]. This requires the development of advanced biomaterials in order to obtain immunomodulatory results that are both predictable and specific [94].

B and T cell activation, differentiation, and cytokine production can be either stimulated or suppressed by biomaterials through their signalling pathways and transcription factors. Specific immune cell receptors and pathways are influenced by a variety of natural biomaterials, including collagen, gelatin, silk, fibrin, alginate, chitosan, and hyaluronic acid. For instance, collagen influences cell adhesion, proliferation, and inflammatory cytokine production by interacting with integrins, discoidin domain receptors (DDR), and OSCAR. Chitin-binding receptors and TLR-2 can recognise chitosan, which results in the activation of immune cells and the production of pro-inflammatory cytokines such as TNF-α and IL-6. Hyaluronic acid has the potential to influence the polarisation of macrophages towards an anti-inflammatory (M2) phenotype [65].

The surface chemistry (e.g., hydrophilicity, roughness, functional groups) and topography (e.g., pore architecture, size, shape) of biomaterials are essential in determining macrophage activation, polarisation (M1/M2 phenotypes), and overall immune cell migration and infiltration [94]. For example, macrophage activation markers can be upregulated by increased surface roughness, and anti-inflammatory macrophage polarisation can be improved by specific pore architectures. The modulation of immune reactions is also significantly influenced by the mechanical properties of biomaterials, such as stiffness and elasticity, which influence macrophage polarisation and neutrophil migration. Moreover, the immune system can be influenced by the degradation products released from biomaterials, which can affect the macrophage phenotype and the overall inflammatory response. This underscores the dynamic nature of these interactions over time. This level of detail underscores the complexity of the task of designing biomaterials for immunomodulation; it necessitates a profound comprehension of the intricate molecular and cellular interactions in order to accomplish predictable, desired immune responses, rather than merely general “biocompatibility” [83].

Biomaterial-enhanced 3D models are essential for the development of in vitro assays for previously intractable and complex immunotoxicology endpoints, such as T-cell dependent antibody responses (TDAR) and systemic immunotoxicity [48]. Traditional 2D in vitro methods were unable to capture these endpoints due to their lack of physiological complexity. Peripheral blood mononuclear cells (PBMCs) and other immune cell types are commonly employed in in vitro humanised models to evaluate immunotoxicity, including cytotoxicity, proliferation, and cytokine release, in co-culture with target cells. These assays are capable of detecting any monoclonal antibody that could potentially induce a hazardous cytokine surge, thereby offering a critical safety net that animal tests frequently failed to provide [49]. In order to evaluate vaccine formulations and adjuvants, dendritic cell activation assays are implemented to evaluate innate immune activation.

In order to evaluate intricate immune responses, including TDAR and immune enhancement, engineered lymph node models and other advanced 3D immune MPS are currently being developed [50]. These intricate endpoints necessitate appropriate matrices and spatial organisation of pertinent immunological compartments, which 2D in vitro systems are incapable of delivering. Within a single human microphysiological model, multi-OoC setups can simulate systemic pharmacodynamic effects by connecting a variety of tissue chips (e.g., liver, tumour) with an immune compartment to investigate the effects of immunotherapy, including on-target tumour killing and off-target organ toxicity [51, 95]. The co-culture of cancer spheroids (e.g., MCF7 breast cancer) with T-cells from human PBMCs within microfluidic systems offers a platform for the study of immune cell infiltration and modulation, as well as tumor-immune cell interactions and immunotherapeutic responses [52]. This suggests that biomaterials are directly addressing critical gaps in in vitro immunotoxicology by facilitating complex 3D structures and multi-cellular co-cultures, thereby enabling the study of functional immune responses that rely on tissue-like organisation and cell-cell crosstalk.

Biomaterial properties, such as their architecture (dimensionality, topography, porosity) and mechanics (stiffness, elasticity, plasticity), significantly influence critical cancer hallmarks [76]. These properties have the potential to substantially influence cell migration, invasion, and drug resistance, as well as to promote EMT, enrich cancer stem cell populations, and modulate proliferation. Additionally, they can direct cancer cell morphology. Mechanical and biochemical signals that are essential for tumour progression can be influenced by 3D scaffolds that are engineered from biomaterials, which can effectively replicate the hypoxic and nutrient-deprived conditions of the in vivo TME [95].

Biomaterials are a critical component of personalised oncology in vitro, as they facilitate the development of patient-derived tumour models that more accurately predict therapeutic responses and faithfully conserve individual tumour heterogeneity than conventional, non-patient-specific models [98]. Grown within biomaterial scaffolds, patient-derived tumour organoids (PDTOs) are becoming potent models for personalised cancer therapy [99]. The functional, genetic, and intrinsic heterogeneity of the original tumour is consistently preserved by these models, rendering them highly appropriate for the identification of resistance mechanisms, the discovery of prognostic biomarkers, and the assessment of patient-specific drugs [41]. Biomaterials facilitate the growth, differentiation, and maintenance of PDTOs by providing the essential scaffold and biochemical cues required for their physiological relevance and the ability to replicate the in vivo TME [5]. The precise construction of biomimetic models using patient-derived tumour cells and their associated ECM is made possible by advanced 3D bioprinting techniques [77]. This approach further advances the potential for truly personalised treatment strategies by enabling the arrangement of cells along complex 3D architectures. Biomaterials facilitate the accurate recapitulation of patient-specific tumour biology, which in turn enables the identification of resistance mechanisms and personalised drug screening. Consequently, a direct causal link is established.

Biomaterial-assisted 3D models are fundamentally changing the direction of cancer drug discovery from simple efficacy screening to a deeper, mechanism-based understanding of drug resistance, particularly those mediated by the TME, by accurately replicating the complex TME and its associated gradients [100].

Compared to 2D models, 3D cell cultures, particularly those supported by biomaterials, provide a more precise prediction of drug effects and toxicity, thereby bridging the divide between preclinical models and human physiology [22]. By emulating the ECM, hydrogel matrices facilitate the formation of multicellular tumour spheroids (MCTSs) [100]. The architectural structure of these MCTS is highly valuable for the study of drug permeation, efficacy, and discharge, as it provides significantly greater resistance to antineoplastic agents, similar to solid tumours in vivo [38]. This implies that biomaterials create a realistic TME that enables researchers to investigate the reasons why medications fail in vivo, such as diffusion barriers, hypoxic core, and ECM interactions, rather than merely determining whether they fail. As a critical stage in the development of more effective cancer therapies, this shifts drug discovery towards mechanism-based approaches for overcoming resistance.

The gene expression profiles of 3D MCTS, such as survival, proliferation, and resistance genes, are more similar to in vivo tumours than those observed in 2D cultures, as revealed by chemo-sensitivity analyses [101]. The role of the ECM provided by biomaterials in 3D systems in mediating chemo-resistance is significant, as it mirrors the physiological ECM’s influence during chemotherapy [95]. To further improve the predictive potential and physiological relevance of drug screening platforms, biomaterials can be engineered to include features such as biochemical gradients and vascularization. This facilitates a more thorough comprehension of the TME’s role in drug resistance, extending beyond basic cytotoxicity evaluations to investigate intricate molecular and cellular mechanisms.

By integrating genomics, transcriptomics, proteomics, and metabolomics, omics technologies offer an unparalleled opportunity to develop a thorough comprehension of the molecular mechanisms of toxicity. By capturing global changes in gene expression, protein abundance, and metabolite levels, these technologies can identify potential biomarkers of exposure or effect and enable personalised risk assessment [102].

Biomaterial-engineered 3D cell cultures are becoming more amenable to HTS, which enables the creation of large-scale omics datasets that are abundant in biological information [103]. The burgeoning discipline of “materiomics,” which employs omics methodologies to analyse biomaterials, has the potential to expedite the development of medical devices by elucidating the relationships between material properties and their impact on intricate biological systems [104]. The analysis of biological effects induced by material properties is significantly improved by computational modelling, which is essential for the correlation of high-throughput gene expression profiling with combinatorial material design strategies. This integration enables the construction of libraries with high-throughput data on both biological (transcriptome) and material (materiome) scales, thereby enabling potent correlation analyses to define the biological functionality of material properties and enhance material design.

Computational models, including Artificial Intelligence/Machine Learning (AI/ML) algorithms [105] and Quantitative Structure-Activity Relationship (QSAR) models [105], offer cost-effective and time-efficient alternatives to traditional animal testing for assessing chemical hazards. These models predict chemical activity, carcinogenicity, immunotoxicity, and ADMET (Absorption, Distribution, Metabolism, Excretion, Toxicity) properties by utilizing structural features and comprehensive datasets. AI/ML algorithms can discern patterns, correlations, and predictive relationships that remain obscured through conventional statistical methods by analyzing extensive datasets from sources like ToxCast, ChEMBL, and PubChem [106].

The design and optimization of novel biomaterials for in vitro models are increasingly informed by computational models, especially AI/ML and QSAR, which are advancing beyond simple toxicity prediction. This has led to the creation of a robust iterative feedback cycle that enhances the speed of material discovery and refinement. This represents a significant shift from the traditional trial-and-error methodology in biomaterial development. Computational models enable in silico parameter optimization for the rational design and development of novel biomaterials, thus streamlining the development process by allowing the variation of parameters that are difficult to modify in a laboratory setting. QSAR models are utilized to predict essential parameters for in vitro to in vivo extrapolation (IVIVE), thus connecting in vitro results with systemic effects [107]. These models are employed to enhance the physical, chemical, and biological characteristics of biomaterials for targeted in vitro applications. This process serves as the fundamental mechanism for rational design.

Despite significant potential, considerable challenges exist in integrating multi-omics data with biomaterial-based in vitro models. The challenges encompass the inherent complexity of data analysis and interpretation, the necessity of ensuring data quality and completeness, and the difficulties posed by the “black box” nature of certain AI/ML models, which complicates mechanistic understanding [108]. The properties of polymers are significantly affected by molecular weight, distribution, and processing, resulting in the absence of straightforward, universal correlations between biological activity and chemical structure.

This integration presents numerous opportunities. The comprehensive biological cascade of effects is elucidated through the synergistic integration of multi-omics data, offering a more systemic and holistic understanding of toxicity pathways, thereby exceeding the limitations of single-endpoint analyses [104]. The integration of individual variability in genomic, biological, and physiological parameters facilitates personalized risk assessment. This process accelerates the identification of biomarkers by detecting molecular signatures linked to toxicity and chemical exposure. The integration of advanced in vitro models allows AI/ML to enhance the predictability of human responses by considering complex factors, including environmental influences, lifestyle variations, genetic diversity, and drug metabolism in preclinical studies. The field of predictive toxicology is poised for a significant transformation due to the integration of biomaterial-based in vitro models with multi-omics data, providing a more comprehensive, personalized, and systemic understanding of chemical-biological interactions. This method goes beyond simple correlations to clarify complex biological networks and pathways, offering a more thorough basis for drug development and safety evaluation.