Innovative organ-on-a-chip platforms for exploring tumorigenesis and therapy in head and neck cancer

Abstract

Background

Head and neck cancer (HNC) presents significant research challenges due to the complexity of its tumor microenvironment (TME) and the heterogeneity across different cancer subtypes. Recent advancements in three-dimensional (3D) culture models and organ-on-a-chip (OOC) technology offer new avenues for mimicking the TME and enhancing the study of tumor biology, drug responses, and personalized treatment strategies. This study aims to summarize the current state of these models in HNC research and their potential in bridging the gap between preclinical models and clinical applications.

Methods

This review synthesizes findings from recent literature on the use of 3D models such as tumor spheroids, organoids, and co-culture systems in HNC research. A focus is placed on their applications in different cancer types, including laryngeal, oral, and nasopharyngeal cancers. Additionally, the integration of OOC technology in studying cancer metastasis, immunotherapy, and radiotherapy is discussed. Relevant studies on the role of AI and robotics in improving model efficiency and scalability are also examined.

Results

The review identifies key developments in 3D model systems and OOC technologies, highlighting their ability to replicate patient-specific tumor behaviors and predict therapeutic responses. While these models have advanced the understanding of HNC pathophysiology, challenges remain in terms of technical limitations, validation, and physiological relevance. The integration of AI and robotics has shown promise in enhancing the scalability and data analysis capabilities of these models.

Conclusions

Advancements in 3D and OOC technologies are essential for overcoming the current limitations in HNC research. These models offer valuable insights into tumor biology and therapeutic efficacy, and their integration with artificial intelligence (AI) can further accelerate the development of personalized treatment strategies. However, further validation and refinement are needed before these models can be widely translated into clinical practice, offering a more effective and individualized approach to treating HNC.

1. Head and neck cancer: an overview

Head and neck cancer (HNC), the sixth most common malignancy globally, encompasses various types, including oral cavity cancer, laryngeal cancer, and nasopharyngeal cancer. These tumors typically originate from mucosal epithelial tissue, with squamous cell carcinoma being the most frequently encountered subtype [1]. In head and neck squamous cell carcinoma (HNSCC), most are located in areas exposed to airflow, increasing their exposure to carcinogens like tobacco and alcohol, and the tumor microenvironment frequently involves chronic inflammatory responses [2]. Specifically, oral cavity and laryngeal cancers are generally associated with tobacco use, alcohol abuse, or a combination of both. Additionally, an increasing number of studies indicate a significant correlation between pharyngeal tumors and infection with human papillomavirus (HPV), particularly with the HPV-16 strain, which has a higher oncogenic potential [3]. However, regardless of whether HNSCC is caused by environmental factors or viral etiologies, males have a significantly higher risk than females. The classic clinical symptoms of HNSCC depend on the anatomical location of the primary tumor and the etiology of the tumor.

In histopathology, HNSCC is characterized by the extent of cellular atypia and squamous differentiation. Well-differentiated tumors closely resemble stratified epithelium, featuring mature cells arranged in layers with irregular keratinization, often appearing as “keratin pearl“ [3]. Conversely, poorly differentiated tumors exhibit immature cells, nuclear pleomorphism, and atypical mitoses, with little to no organized stratification or keratinization. Notably, HPV-positive HNSCC tissues tend to be poorly differentiated or exhibit basaloid morphology in histopathological analysis, while HPV-negative HNSCC is usually moderately or highly differentiated, retaining stratification and keratinization [4]. Clinically, HPV-positive patients are also more likely to present with positive lymph node metastases. However, due to the significantly greater radiosensitivity and chemotherapy sensitivity of HPV-positive tumors compared to HPV-negative tumors, HPV-positive HNSCC generally has a better prognosis, with HPV-positive patients showing higher survival rates [5, 6].

Additionally, the tumor microenvironment (TME) of HNSCC is a highly heterogeneous and complex mixture of various cells, matrix components, and factors (Fig. 1). The primary cellular composition includes cancer cells, immune cells, and stromal cells, specifically T cells, B cells, macrophages, cancer-associated fibroblasts (CAFs), and endothelial cells [7]. Understanding the crosstalk between cancer cells and the TME is crucial for inhibiting tumor progression and overcoming treatment resistance. For instance, CAFs play a key role in the pathogenesis of HNSCC as they remain in a persistently activated state and secrete various chemokines, cytokines (such as IL-6), and growth factors (such as VEGF and HGF). This activity promotes the formation of tumor-adaptive extracellular matrix structures, angiogenesis, and immune and metabolic reprogramming within the TME, all of which are associated with metastasis and resistance to radiotherapy and chemotherapy [8]. In HPV-positive patients, HPV can evade the innate immune system by directly altering the interferon signaling cascade, a critical antiviral and immunostimulatory pathway [9]. Moreover, HNSCC frequently exhibits inactivating mutations in tumor necrosis factor receptor-associated factor 3 (TRAF3), which further weakens the antiviral immune response [10]. Similarly, the TME of non-HPV-related HNSCC patients may be influenced by tobacco and alcohol use, leading to immune suppression and the recruitment of various immune cell types, including monocytes, macrophages, dendritic cells, and neutrophils [11] (Table 1)。.

Fig. 1.

Head and Neck Cancer and Its Complex Microenvironment A. The most frequent sites of head and neck tumors include: (1) oral cavity, (2) pharynx, (3) larynx, (4) nasal cavity, and (5) thyroid; B. The HNC tumor microenvironment (TME) consists of cancer cells, stromal cells (e.g., cancer-associated fibroblasts), immune cells (e.g., B cells, T cells, Natural Killer cells), and extracellular matrix (ECM) components (e.g., collagen, fibronectin), which collectively support tumor progression and therapy resistance. Abbreviations: CAF, cancer-associated fibroblast; VEGF, vascular endothelial growth factor; TAM, tumor-associated macrophage; TAN, tumor-associated neutrophil

Table 1.

Two subtypes of HNSCC, different tumor microenvironments and clinical features

HNSCC subtype	Carcinogenic factor	TME	Lymphatic metastasis(LM)	Radiosensitivity
HPV-negative HNSCC	Alcohol, Tobacco	Decreased immune infiltration	Less susceptible to LM	Less sensitive to radiotherapy
HPV-positive HNSCC	HPV	Increased immune infiltration	More susceptible to LM	More sensitive to radiotherapy

Clinically, due to the subtlety of early symptoms, most patients diagnosed with HNSCC present with locally advanced disease, requiring multimodal treatment [12]. Standard treatments for HNC patients typically involve surgery, radiotherapy, chemotherapy, or a combination of these approaches. However, nearly 50% of patients do not experience significant benefits, instead enduring severe toxic side effects that impact both individuals and society [13]. Despite decades of efforts to improve HNSCC treatment outcomes, progress has been limited. A substantial portion of these patients may experience local failure and/or distant metastasis, leading to poor prognosis. The advent of immune checkpoint inhibitors (ICIs) and immunotherapy (IT) has transformed the systemic treatment landscape for recurrent/metastatic (R/M) HNSCC demonstrating efficacy, activity, and safety in both platinum-resistant and platinum-naïve patients [14]. PD-1-targeting ICIs, nivolumab and pembrolizumab, have been approved for the treatment of R/M HNSCC. Unfortunately, despite the expression of PD-1 and PD-L1 in tumors being 2 to 3 times higher, only 15–20% of HNSCC patients exhibit a durable response to these drugs [15]. Alternatively, molecularly targeted agents, such as the epidermal growth factor receptor (EGFR) inhibitor cetuximab, have demonstrated modest success in treating locally advanced disease [16]. Moreover, anti-CD47 drug Ligufalimab, EGFR-targeting MRG003, and receptor tyrosine kinase (RTK) inhibitor Lenvatinib are currently undergoing various stages of clinical trials for targeted and combination therapies in HNSCC [17–20]. It is crucial to develop new therapeutic approaches and conduct preclinical validation.

This review was based on a comprehensive literature search of peer-reviewed articles published between 2011 and 2025 using databases including PubMed and Web of Science. Both preclinical and clinical studies were included. Studies were selected based on their relevance to HNC research involving two-dimensional (2D)/ three-dimensional (3D) in vitro models, organoids, organ-on-a-chip (OOC) platforms, tumor microenvironment modeling, drug testing, and metastasis. Articles not directly related to OOC or advanced 3D models in HNC were excluded to maintain thematic focus.

Traditional 2D cell models and animal experiments in head and neck cancer research and their limitations

Researchers must undergo extensive and lengthy in vitro and in vivo model testing and validation for the development and screening of every head and neck cancer drug. This process is not only time-consuming but also involves various complex techniques and methodologies (Table 2). Conventionally, cell culture experiments for HNC involve growing tumor-derived cell lines obtained from patients in a 2D format on culture dishes or flasks [21]. However, as a complex dynamic system, a tumor develops within a region composed of cells and intricate extracellular components that are synthesized or rearranged within this space. This process creates essential microstructures throughout the system. Meanwhile, tumor growth occurs with specific dynamics, influenced by the microenvironment and various endogenous bioactive substances [22]. Consequently, simple 2D cultures retain only a fraction of the tumor cell characteristics and cannot represent the intricate cell-matrix interactions or cell-to-cell crosstalk found within cancerous tissues [23]. Additionally, culturing cells in a 2D environment can significantly alter their gene expression profiles [24].

Table 2.

Conventional 2D-cell and animal models in cancer research

Model	Strengths	Limitations	Comparison with OOC (Throughput, Cost, Clinical Relevance)
2D cell culture (cell lines)	• Cost-effective and straightforward • Enables high-throughput drug screening and toxicity evaluations • Time-efficient with consistent reproducibility	• Overly simplistic representation of tumors • Low success rate in developing reliable tumor models • Limited ability to capture tumor heterogeneity and the complexity of the TME	Higher throughput and lower cost than OOC, but lower clinical predictive value due to lack of 3D structure and TME complexity.
Animals	• Recognized as the gold standard in cancer research • Partially replicates the TME for in vivo assessment of tumor growth and drug response • Facilitates the study of tumors within a living biological system	• Expensive, low-throughput, time-intensive engraftment process, with ethical concerns • Species-specific differences that may result in inaccurate drug testing outcomes • Limited success in transplanting certain human cancers to create patient-derived mouse models	Higher clinical predictive value than 2D models and some OOC, but lower throughput and higher cost compared to OOC.

Patient-derived xenograft (PDX) tumor models can partially compensate for the limitations of 2D models by directly transplanting tumor tissues or cells surgically removed from cancer patients into immunodeficient mice [25]. However, after multiple passages, the human-derived matrix in PDXs is almost entirely replaced by the mouse-derived matrix, leading to a decrease in model accuracy [26]. The long experimdental duration, limited engraftment success rate, and inability to replicate the immune microenvironment further restrict the application of PDX models [27]. Of note, recent systematic studies examining the correlation between animal data and human outcomes have demonstrated a weak predictive ability of animal models [28]. The clinical translatability of drug efficacy tests performed on these models remains highly controversial [29]. Moreover, the use of animal models in scientific research raises numerous ethical issues, which are similarly faced in head and neck cancer studies. These include the potential suffering of animals from pain, anxiety, and other discomforts during research, as well as concerns regarding public awareness and acceptance. The development of alternative methods has emerged as a preferred new direction.

From 3D spheroids to organoids and organ-on-a-chip technology

3D tumor-derived spheroid

The tumor-derived spheroid culture is a notable 3D culture technique for cancer cells, where primary cancer cells with stem cell-like characteristics are grown in vitro as floating spheres [30]. Based on their source, they can be categorized into cell line-derived multicellular tumor spheroids (MCTS) and cancer tissue-derived spheroids.

MCTS can be considered an extension of two-dimensional cultures, as representing a typical type of tumor spheroids derived from cancer cell lines. The culture environment consists of conventional media supplemented with fetal bovine serum (FBS). However, unlike two-dimensional cultures, cells in MCTS grow as spheres in suspension culture. Although MCTS show limited histological similarity to primary tumors, they effectively mimic the metabolic and proliferative gradients of in vivo tumors and exhibit multicellular chemoresistance. MCTS are often regarded as an effective tool for high-throughput drug testing, with advantages such as cell clonality, ease of maintenance, and simple manipulation [31, 32].

Cancer tissue-derived spheroids retain some histological features of the original tumors and can proliferate following mechanical dissociation [33]. Cancer spheroids are generally obtained by gently dissociating cancer tissues using mild enzymes. Cells cultivated by this method are often surrounded by non-tumor cells and stromal components that normally exist in the tumor microenvironment, thereby retaining many histological features and the cellular heterogeneity of the primary cancer. During subsequent proliferation cultures, cancer tissue-derived spheroids can still largely preserve their original histological characteristics even in the absence of non-tumor cells. The main characteristic of cancer tissue-derived spheroids is their enrichment of cancer stem cells (CSCs) or cells with stem cell-related properties, which are often used to evaluate CSC-related features in vitro [34, 35].

Generally, cancer tissue-derived spheroids are similar to MCTS in their formation of free-floating spheres. At times, serum-free culture conditions for tumor-derived spheroids are used in MCTS cultures to simulate stem cell-like states.

Tumor organoids

Tumor organoids were first developed over a decade ago by Sato et al. and have rapidly gained popularity as a representative ex vivo model for tumorigenesis [36]. By establishing tumor-specific growth conditions, including basement membrane matrix (Matrigel), tyrosine kinase receptor agonists, bone morphogenetic protein/transforming growth factor-β inhibitors, Wnt agonists, and various growth factors, researchers can simulate the microenvironment components of non-tumor cells associated with various tumor tissues in vitro [37, 38]. Additionally, researchers can develop a multistep carcinogenesis model by introducing sequential oncogene/tumor suppressor gene mutations into non-tumor organoids, providing an effective alternative method for studying cancer development [39, 40].

Besides, tumor organoids are generally produced from two primary sources: freshly removed tumor tissue and induced pluripotent stem cells (iPSCs) [38, 41]. Most cell materials are currently obtained by dissociating the original tumor tissues into a mixture of cell clusters that contain tumor stem cells. The processes of self-assembly and differentiation in organoid formation rely on specific cell signaling pathways influenced by intrinsic components and the surrounding extracellular environment, such as the extracellular matrix (ECM) and culture media [42]. Recently, a range of cancer organoids has been developed for various applications, including patient-derived prostate cancer, breast cancer, colorectal cancer, non-small cell lung cancer, and hepatocellular cancer, for purposes such as genome editing, drug screening, transplantation, radiotherapy evaluation, and oncogene identification [43–47].

Tumor OOC

OOC technology is a physiological microsystem that integrates biomimetic biology with microengineering, enabling the simulation of essential structures and functions of human organs on microfluidic chips [48]. OOC are enabled by the integration of cutting-edge knowledge in cell co-cultures, stem cells, genome editing, sensors, 3D printing, and microfluidics. Typically, the design of an organ-on-a-chip system is informed by reductionist principles that analyze the characteristics of the target organ [49]. By combining living human cells with synthetically created, physiologically relevant microenvironments, organs-on-a-chip can replicate the integrated functions of organs essential for maintaining physiological homeostasis and the intricate processes of diseases [50]. Notably, in addition to the cellular composition of the tissue microenvironment, perfusable vascular networks can be generated on OOC platforms. These networks can exist as standalone structures [51] or be incorporated with other tissues to develop functional vascularized chips [52].

Due to the biological complexity of human systems, researchers have begun to integrate multiple organs onto a single chip as chip technology advances, exploring microengineering and co-culturing various types of organoids on in vitro platforms to simulate multi-organ interactions [53]. Bovard et al. integrated liver and lung on a single chip to study the toxicity of inhaled substances. They constructed the chip framework using fluidic plates and a small molecular non-absorbable material, polyether ether ketone. In the lung chamber, normal human bronchial epithelial cells were cultured at the air-liquid interface, while HepaRG liver spheroids were cultured in the liver chamber. Compared to a single lung chip, the toxicity of the toxic substances significantly decreased after enzymatic metabolism through this integrated chip [54]. These systems replicate the functions of tissues and organs by constructing 3D arrangements of different tissues, where biological and physical forces are applied to simulate in vivo conditions. Additionally, multiple organs-on-a-chip can be combined to form a body-on-a-chip, allowing for the assessment of organ interactions that reflect the complexity of the human body [55], which is also a powerful tool for evaluating the absorption, distribution, metabolism, and elimination (ADME) of test drugs. de Haan et al. simulated the digestive chain representing the mouth, stomach, and intestines by serially connecting three Y-shaped micromixers to create a three-stage chamber chip. In this model, they introduced saliva, gastric juice, and intestinal fluid along with pre-mixed drugs into the three Y-shaped mixers. A fourth micromixer provided the necessary cell culture matrices and factors for the corresponding co-culture cell lines in a flow-through transwell chamber, allowing them to test the bioavailability of orally administered small molecules [56].

While OOC systems have shown broad applicability across various tumor types, their unique advantages in HNC research merit special emphasis. HNC exhibits marked heterogeneity driven by HPV status, with HPV-positive and HPV-negative tumors differing significantly in immune infiltration, molecular mechanisms, and treatment responsiveness [57, 58]. Traditional 2D cultures lack the structural and cellular complexity to model the distinct immune and stromal landscapes of HPV-positive and HPV-negative HNC. Similarly, animal models often fail to reflect the human-specific immune contexture and viral oncogene interactions that are central to HPV-related tumor biology. OOC platforms, however, enable precise control over cell composition, flow dynamics, and stromal architecture, making it feasible to construct TME models tailored to HPV status. For instance, chips can co-culture HPV + or HPV − tumor cells with immune cells, fibroblasts, and ECM components to mimic immunosuppressive or inflammatory milieus, allowing comparison of differential responses to immune checkpoint inhibitors or chemoradiotherapy. Moreover, the complexity of head and neck anatomy, reflected in its diverse tissues and vascular networks, can be captured through the use of regionally customized chip designs. This enables modeling of site-specific dynamics such as local invasion and perfusion relevant to oral, pharyngeal, or laryngeal cancers. As such, OOC platforms offer a powerful tool for dissecting HPV-driven heterogeneity and guiding personalized therapeutic strategies in HNC.

Traditionally, the development of personalized treatment strategies using patient-derived tumor spheroids has been limited by the low number of available cells, resulting in low-throughput assays and significantly increased costs for drug screening and biomarker identification. In contrast, OOC platforms utilize microfluidic technology to precisely manipulate limited tumor samples and enable high-throughput functional analysis [59]. For example, Schuster et al. developed an automated high-throughput microfluidic platform featuring a 200-chamber array integrated with 3D culture and fluidic control systems, allowing simultaneous testing of multiple standardized and dynamic chemotherapy regimens for clinical cancer patients [60]. These features make OOC a more cost-effective and faster approach to simulate human tumor pathophysiology and the tumor microenvironment, thereby providing a more accurate representation of cancer characteristics (Table 3).

Table 3.

Typical examples of OOC technology used for tumor modelling

Tumor type	Devices & Purposes	References
Pancreatic ductal adenocarcinoma (PDAC)	A 3D PDAC model cultured under flow for investigation of tumor-stroma-immune cell interactions.	[61]
Breast cancer	A breast cancer-on-chip model integrates an endothelial barrier, immune cell transmigration, and cytokine monitoring to study CAR-T cell efficacy, immune response control, and patient-specific therapy.	[62]
Prostate cancer	A Prostate-Cancer-on-Chip (PCoC) model recapitulates tumor-stroma interactions, enabling analysis of CAF conversion, tumor invasion, and androgen receptor modulation, for studying prostate cancer mechanisms and therapies.	[63]
Cervical Cancer	Cervical cancer-on-chip (CCoC) model, enables the generation and long-term cultivation of SiHa spheroids co-cultured with cervical fibroblasts, emulating tumor-stroma interactions, with potential for integrating immunocompetent components for developing novel (immuno)therapies.	[64]
Lung and Intestine	Human Organs-on-Chips platforms evaluate T-cell bispecific antibodies’ safety by mimicking target-dependent toxicity in lung and intestine tissues, offering insights into the safety and mechanisms of engineered therapeutic antibodies.	[65]
Breast cancer	Two organ-on-chip platforms, IC-chip and EX-chip, were developed to quantitatively assess breast cancer cell invasion and extravasation towards tissue-specific microenvironments, revealing that invasive MDA-MB-231 cells preferentially invade and extravasate into lung and liver tissues.	[66]
Non-small-cell lung cancer	A microfluidic OOC model of NSCLC revealed that lung cancer cell growth, invasion, and TKI responses are influenced by physical cues from b reathing motions, mediated by EGFR and MET signaling.	[67]
Breast cancer bone metastasis	A OOC model revealed that paracrine signaling between sympathetic neurons and osteoclasts boosts breast cancer aggressiveness in bone metastasis, offering insights into metastatic niche mechanisms.	[68]
Triple-negative breast cancer liver metastasis	Liver-on-a-chip model effectively simulating the 3D liver microenvironment to assess reprogrammed triple-negative breast cancer cells into hepatocyte-like cells using six transcription factors inhibits liver metastasis.	[69]

Tumor microenvironment in microfluidic organ-on-a-chip

Microfluidic technology is a versatile tool widely applied across various fields, including fluid dynamics, synthetic and analytical chemistry, biology, and medicine. It is used to evaluate drug toxicity, develop and study drug delivery systems, advance regenerative medicine, and conduct single-cell analyses [70]. Its advantages extend beyond efficient fluid control to exploring and leveraging fluid properties that are absent in macroscale systems. Typically, microfluidic technology employs chip modules made from materials such as glass or polymers, featuring micron-scale channels etched or molded into them. These channels enable the simulation of biomechanical forces, such as electrical signals, shear stress, and tension-valuable stimuli for cell maturation and development [48].

Beyond physical stimulation, microfluidic flows facilitate the creation of perfusion culture systems, allowing precise control of nutrient and factor concentrations in the medium while continuously supplying fresh culture media to cells. By incorporating multiple cell types, vascular network structures, and maintaining extracellular matrix factor concentrations, Organ-on-a-Chip systems can closely mimic the tumor microenvironment. These systems are also customizable to reflect individual heterogeneity. Currently, researchers have designed various organ-on-a-chip models tailored to different research goals and needs, including single-chamber, high-throughput multi-array, parallel chamber, and sequentially connected configurations [71–75] (Fig. 2A-D).

Fig. 2.

Different organ-on-chip designs are tailored for various research purposes (A) Single-chamber chip design. Reproduced with permission [71].Copyright 2023, Proceedings of the National Academy of Sciences of the United States of America; (B) High-throughput multi-array chip. Reproduced with permission [72]. Copyright 2021, Toxicology; (C) Parallel chamber chip. Reproduced with permission [73].Copyright 2021, Nature biomedical engineering; (D) Dual-organ and multi-organ interconnected chip platforms. Reproduced with permission [74, 75]. Copyright 2018, Scientific reports; Copyright 2020, Nature biomedical engineering

Chen et al. developed a vascularized tumor spheroid model that recapitulates the pathological features of solid tumors [76]. In this chip-based model, endothelial cells and tumor spheroids were co-cultured in a central channel, while stromal fibroblasts were introduced into peripheral channels. They found that tumor-endothelial cell hybrid spheroids exhibited significantly improved uniformity and peritumoral angiogenesis capacity compared to spheroids composed solely of cancer cells. Furthermore, RNA sequencing revealed that this model displayed invasive gene expression profiles associated with cancer progression, invasion, and metastasis. The study also demonstrated the dose- and time-dependent inhibitory effects of the anti-cancer drug axitinib on tumor growth and angiogenesis. Haque et al. developed a tumor-on-chip device that combined patient-derived organoids and stromal cells to simulate the TME of pancreatic cancer [77]. Using this complex organoid model, they validated that stromal cell-targeting drugs significantly enhanced the efficacy of chemotherapy when tested with stromal cell depletion agents and chemotherapeutic drugs.

Additionally, immune cells, including T cells, macrophages, and natural killer cells, can be integrated into chip models to mimic immune infiltration in the TME and study their interactions with tumor cells [78–81]. Advances in biomaterials and biosynthesis technologies have enabled the development of ECM-mimicking scaffolds with tunable mechanical and biochemical properties, allowing for more accurate TME simulations [82].

Tumor metastasis modeling on chip

The metastasis-on-a-chip platform typically integrates three key elements of the TME: the primary tumor site, the circulatory environment, and secondary metastatic sites [83]. It is designed to study critical steps in metastasis, such as regulating cell migration parameters and real-time monitoring of the invasion process. Even simplified in vitro models provide new insights into the mechanisms of cancer metastasis, enhancing our understanding and control over cancer progression [84].

Erdogan et al. developed a dual-channel co-culture chip, co-culturing tumor organoids with normal fibroblasts or CAFs. They found that CAFs guide the directional migration of cancer cells through tissue fibronectin (Fn) matrices, thus promoting tumor progression [85]. CAFs enhance the contractility and traction force of Fn fibers through myosin II, PDGFRα, and α5β1 integrins, allowing cancer cells to efficiently migrate via αv integrins. Wang et al. cultured renal carcinoma Caki-1 cells in a 3D biomimetic liver microenvironment to simulate metastatic kidney cancer progression in a chip-based model [86] (Fig. 3A). This metastatic tumor model, built using rat liver to construct the liver microenvironment and co-cultured with kidney cancer cell lines, was used to predict therapeutic effects and assess dose-response at different stages of tumor progression. Utilizing patient-derived metastatic tumor tissues could provide strong evidence for developing personalized precision therapies.

Fig. 3.

Application of Organ-on-a-chip in tumor metastasis and drug testing. (A) Co-culture of renal carcinoma Caki-1 cells and CAFs in a chip-model biomimetic liver microenvironment. Reproduced with permission [86]. Copyright 2021, Theranostics; (B) Multi-organ metastasis-on-a-chip model. Reproduced with permission [87]. Copyright 2016, ACS Applied Materials & Interfaces; (C) Lung cancer organoid drug testing model with CGG for screening patient-derived organoids and identifying optimal drug concentrations. Reproduced with permission [89]. Copyright 2013, Biomaterials; (D) a high-throughput chip model to test anti-PD-1 effects on breast cancer spheroids and T cells, using microcolumn arrays to monitor tumor-immune interactions. Reproduced with permission [91]. Copyright 2021, Small

Additionally, Xu et al. developed a multi-organ chip to study lung cancer metastasis to the brain, bones, and liver, analyzing cellular physiology and interactions under more physiologically relevant conditions [87] (Fig. 3B). In the “lung” chamber, they cultured bronchial epithelial cells, lung cancer cells, microvascular endothelial cells, monocytes, and fibroblasts, while astrocytes, osteocytes, and hepatocytes were cultured in distal chambers to simulate the metastatic process. The study showed that lung cancer cells formed “tumor masses” in this system, accompanied by epithelial-to-mesenchymal transition (EMT), characterized by changes in the expression of E-cadherin, N-cadherin, Snail1, and Snail2, demonstrating enhanced invasive capabilities.

Anticancer drug testing within organ-on-a-chip model

Cancer drug development is a lengthy and labor-intensive process, and large-scale screening of drug candidates often requires significant manpower. Microfluidic technology offers an efficient solution by enabling drug screening with limited materials, such as primary cells or patient biopsy samples. In microfluidic organ-on-a-chip systems, drug gradient generators can be integrated to record tumor cell responses to different drug concentrations in real-time [88].

For instance, in the lung cancer organoid drug testing model developed by Xu et al., an upstream cell culture chamber was equipped with a concentration gradient generator (CGG) [89]. By regulating drug concentrations with the CGG, they were able to generate a range of concentrations and inoculate patient-derived lung cancer organoids into downstream chambers to screen for suitable drugs and determine the optimal concentrations (Fig. 3C). This approach is particularly valuable for drug screening in clinical settings, where personalized treatment plans are crucial. Similarly, Zhu et al. designed a high-throughput aortic smooth muscle chip that can be directly placed under a microplate reader for post-drug treatment measurements, akin to a 96-well plate [90]. This model allows for rapid, simultaneous screening of multiple drugs using different cell lines, providing a powerful tool for high-throughput drug testing. This design can also be applied to other drug screening chip models.

Moreover, Jiang et al. developed a high-throughput immunotherapy observation chamber chip model to test the effect of anti-PD-1 on the interaction between breast cancer tumor spheroids and T cells [91]. By using microcolumn arrays to measure IL-2 concentration, this model enables convenient monitoring of T-cell inhibition and reactivation, capturing tumor-immune interactions in real-time. After coating the array surface with antibodies, the system allows for sensitive, quantitative detection, which is essential for evaluating the efficacy of immunotherapies (Fig. 3D). These examples demonstrate the growing potential of organ-on-a-chip models for personalized drug testing and treatment optimization.

3D models and Organ-on-a-Chip in HNC research

Representative 3D models in HNC

The treatment of HNC typically involves multimodal strategies. However, the substantial heterogeneity within head and neck tumors limits the efficacy of these treatments. To better mimic the tumor microenvironment, the use of technologies such as tumor spheroids, organoids, and co-culture systems models has been increasing in HNSCC research (Table 4). Below, we summarize some key 3D models and their applications across different types of head and neck cancers.

Table 4.

Representative studies on the construction of head and neck tumor 3D-culture models

Tissue type	Cell/Tissue Resource	Materials	3D Culture	Applications	Reference
HNSCC	Cell lines originated from squamous cell carcinomas (SCC) of the larynx (11B), the hypopharynx (22B), the tongue (74) and the oropharynx (UD01).	96-well Nunclon Sphera-Treated U-shaped-bottom Microplates	The cells were cultured into spheroids, incubating for seven days prior to treatment, with fresh mediums replaced every 48 h.	Construct 3D tumor spheroid models for HNSCC and compare the drug sensitivity and mechanisms of cancer cells in 3D culture versus 2D culture.	[92]
Hypopharyngeal squamous cell carcinoma	cell lines: HNSCC cells FaDu, human fibroblasts (HF), human monocytes (THP-1), and human endothelial cells (HUVEC)	Transwell inserts	Cultured in 6-well SPL3D cell floater plate (3D-4-culture) as mixed cells.	Establishing a 3D-4-culture model to better simulate the in vivo immune microenvironment of HNSCC	[93]
HNSCC (Hypopharynx)	HNSCC cell line FaDu; CAFs: patient tissue fragments	Tissue Roll for the Analysis of Cellular Environment and Response (TRACER) platform	The TRACERs co-culture three-layer structure consists of the first layer containing CAFs (GFP), the third layer containing FaDu (mCherry), and the middle layer containing only collagen.	Create a 3D in vitro co-culture model with CAFs and HNSCC tumor cells to study their organizational patterns and the impact of CAFs on HNSCC tumor cell phenotype.	[94]
Oral squamous cell carcinoma	Tumor specimens and adjacent normal tissue from patients.	Matrigel	Solidified matrigel-cells complex incubated with organoid culture medium.	Investigate that co-culture with paralleled CAFs promotes stem-like properties of OSCC	[95]
Primary Nasal Epithelial	human donor; 3T3-J2 cell line	collagen IV	Cultured conditionally reprogrammed cells into the apical side of a 24 well plate-sized transwell.	Studying latent and lytic infections can help explore and complement the mechanisms of EBV-associated nasopharyngeal carcinoma initiation.	[100]
NPC tissues and normal mucosa	patients	Matrigel	Cultured in the matrigel supplementary with organoid culture medium at 37 ℃.	Established a living biobank using the model, offering significant potential for both basic and clinical research on NPC.	[103]
HNC (Tongue)	Patient samples	Myogel and Matrigel; 3D microfluidic chip	Mix cancer cells with Myogel/fibrin to form a gel cell suspension, and load 3 µL of the suspension into each microchannel of the chip. Perfuse with culture medium and incubate in a cell culture incubator.	Construct a humanized in vitro microfluidic chip assay to test immunotherapeutic drugs against HNSCC patient samples.	[110]
HNSCC and normal epithelium-derived mucosa	Patient material and mouse-derived	Cultrex growth factor–reduced BME type 2	Cultured in 70% BME in organoid medium.	In vitro drug screening of tumoroids uncovers responses to both current HNSCC therapies and experimental treatments not yet utilized in clinical practice.	[123]
Nasopharyngeal carcinoma	nasopharyngeal epithelial cell line (NP460) and nasopharyngeal carci noma cell line (NPC43)	PDMS (Dow Corning Sylgard 184 kit) mixture	In 3D biomimetic platform with medium.	Cell migration in a complex 3D environment to model nasopharyngeal carcinoma metastasis.	[124]
Papillary Thyroid Cancer	Human Specimens	Matrigel	Solidify the 45-µL Matrigel-cell mixture in a 6-well plate, then nourish with organoid medium at 37℃.	Presenting a promising preclinical model for personalized anticancer studies, potentially aiding ERα-specific thyroid cancer strategies.	[125]

For laryngeal and hypopharyngeal cancer, Heid et al. compared 2D and 3D culture models using HNSCC cell lines from different anatomical sites [92]. They found that while EGFR expression remained high in both conditions after TKI treatment, EGFR phosphorylation was significantly stronger in the 3D culture. This highlights that 3D cultures more accurately reflect the in vivo response. For example, UM-SCC-11B cells from laryngeal cancer showed significant drug resistance in 3D cultures, whereas they responded well in 2D cultures. In a study by Xiao et al., the immune microenvironment in 2D and 3D models of hypopharyngeal HNSCC was compared. They found that the 3D model better simulated the in vivo immune environment, promoting the conversion of fibroblasts into CAFs, monocytes into tumor-associated macrophages (TAMs), and enhancing endothelial cell migration. This model provides an important platform for studying immune escape mechanisms in HNSCC [93]. Similarly, Young et al. used a 3D TRACERs co-culture model of CAFs and HNSCC cells (Fig. 4A), revealing that CAFs increased cancer cell proliferation, invasiveness, and migration, further validating their role in the tumor microenvironment [94].

Fig. 4.

Representative HNC 3D model applications (A) 3D TRACERs co-culture model of cancer-associated fibroblasts and HNSCC cells. Reproduced with permission [90]. Copyright 2021, Biomaterials; (B) 3D oral cancer model simulating tumor progression and testing nanoparticle-based drug delivery systems. Reproduced with permission [93]. Copyright 2023, In Vivo; (C) Air-liquid interface 3D culture model of nasopharyngeal carcinoma to study disease progression and EBV infection. Reproduced with permission [95]. Copyright 2018, mSphere

In oral squamous cell carcinoma (OSCC), Zhao et al. explored CAF interactions with OSCC stem cells using 3D organoids. They found that co-culture with CAFs enhanced organoid formation in CD44 + cells, with lactate playing a crucial role. This suggests that lactate has a significant impact on OSCC stem cell properties [95]. Flashner et al. also used 3D organoids derived from mouse models of oral-esophageal cancer to study disease progression and precursor cells, enabling drug screening [96]. Mendoza-Martinez et al. evaluated the efficacy of ZnO and MgO nanoparticles combined with chemotherapeutic drugs in 2D and 3D models (Fig. 4B), demonstrating that the 3D model is crucial for assessing the effects of nanoparticle-based drug delivery systems [97].

In nasopharyngeal carcinoma (NPC) research, EBV infection is a key factor in the disease’s development [98]. Traditional 2D culture systems have low infection rates of EBV in nasopharyngeal epithelial cells. Therefore, Caves et al. used the air-liquid interface (ALI) culture method to study EBV infection in nasopharyngeal epithelial cells (Fig. 4C), concluding that this method is suitable for reactivating established EBV-infected cell lines [99]. Later, Ziegler et al. improved the 3D culture system with the ALI method, significantly increasing the infection rate and providing a better platform for studying EBV infection and nasopharyngeal carcinoma [100]. Lucky’s team developed NPC patient-derived xenograft organoids, which showed good correlation with patient tumors, supporting personalized treatment strategies [101]. Ding et al. also constructed NPC organoids for personalized treatment through combined radiotherapy and chemotherapy [102]. Additionally, Wang et al. created a biobank of primary and recurrent NPC 3D models for drug screening and preclinical studies, using a novel two-step enzymatic digestion method [103].

HNC-related organ-on-a-chip model

In preclinical research on HNC disease models, Byun et al. developed a 3D respiratory mucosal chip model, which incorporated human nasal epithelial cells, fibroblasts, and endothelial cells, to study the effects of urban particulate matter (UPM) on the respiratory system [104]. Their study showed that UPM exposure altered the gene expression related to inflammation and adhesion in nasal epithelial cells, disrupting the integrity of the respiratory mucosa, as indicated by decreased expression of tight junction protein ZO-1 and endothelial cadherin. This work highlights the potential of 3D respiratory mucosal models as valuable platforms for assessing multicellular responses (Fig. 5A).

Fig. 5.

Representative organ-on-a-chip studies related to head and neck cancer (A) Respiratory chip derived from nasal cavity, consisting of three layers: epithelial, fibroblast, and endothelial layers. Reproduced with permission [104]; (B) Microfluidic chip design for tongue cancer. Immune cells are loaded in channel A, tongue cancer cells in channel B, and channel C is used for hydration. Reproduced with permission [110]; (C) Oral cancer organ-on-a-chip simulating the tumor microenvironment, schematic and in vitro characterization. Reproduced with permission [112]

Expanding this application to tobacco-related HNC risk, airway-on-a-chip platforms have been employed to investigate the pathophysiological impact of cigarette smoke exposure. These microfluidic systems typically co-culture primary human airway epithelial and endothelial cells at an air-liquid interface to mimic the in vivo airway barrier [105, 106]. Notably, Benam et al. constructed an airway chip using chronic obstructive pulmonary disease patient-derived epithelial cells and demonstrated that smoke exposure recapitulated hallmark features such as goblet cell hyperplasia, ciliary dysfunction, and oxidative stress, alongside excessive secretion of pro-inflammatory cytokines [107]. Dynamic exposure was achieved via a custom smoking device, allowing real-time monitoring of smoke-induced perturbations under physiologically relevant conditions. Further incorporation of fibroblasts and immune cells into tri-culture airway chips has enabled exploration of epithelial-mesenchymal-immune interactions that drive airway remodeling and inflammation, offering valuable insights into tobacco-induced carcinogenesis mechanisms relevant to HNC [108]. Gao et al. developed a compliant collagen-based airway-on-a-chip that enables long-term ALI culture and dynamic ventilation, faithfully mimicking the tubular geometry and biomechanical environment of small airways. This model allows precise investigation of ventilation-induced epithelial injury, including collapse-reopening cycles relevant to tobacco-related airway damage [109]. These advancements underscore the value of airway-on-a-chip models as platforms for dissecting the molecular underpinnings of smoking-related diseases.

In the realm of cancer immunotherapy, Al-Samadi et al. developed a 3D microfluidic chip model to test the effects of immunotherapy using the HSC-3 tongue cancer cell line and immune cells from healthy donors [110]. Their findings revealed that the migration of immune cells towards cancer cells was density-dependent, with IDO1 inhibitors promoting immune cell migration. Interestingly, the efficacy of PD-L1 antibodies and IDO1 inhibitors varied among patients (Fig. 5B). This research introduces a humanized microfluidic chip method that could predict individual patient responses to immunotherapy.

Further advancing cancer metastasis research, Nairon et al. created a thyroid-lung metastasis chip model, where thyroid cancer cells circulated through microfluidic channels parallel to a lung hydrogel structure, simulating single-cell-level invasion [111]. Their study demonstrated that knockdown of RCAN1.4 increased the expression of NFE2L3, which promoted invasion and metastasis of thyroid cancer cells, with varying metastatic capabilities observed across different cell lines.

In oral cancer research, Yada et al. developed an oral cancer micro physiological system by co-culturing patient-derived tumor-associated fibroblasts, HNC oral tumor-derived spheroids, and lymphatic microvascular cells to investigate metastasis mechanisms [112]. The inclusion of tumor-associated fibroblasts enhanced migration compared to normal oral fibroblasts and blank controls. Additionally, the study found that lymphatic endothelial cells in the tumor microenvironment secreted macrophage migration inhibitory factor (MIF), which enhanced the glycolytic dependency of migratory HNC cells, as demonstrated by single-cell metabolic imaging (Fig. 5C). This multi-cellular microfluidic platform offers high-resolution tools for visualizing and quantifying patient heterogeneity.

Complementing these efforts, Hachey et al. introduced a vascularized microtumor (VMT) platform that recapitulates the dynamic tumor-stromal-vascular interactions within a microfluidic device under physiologic flow conditions. This platform supports the co-culture of tumor cells with perfused human vasculature, enabling the study of immune cell trafficking, drug delivery, and tumor heterogeneity in a setting closely resembling in vivo environments. The VMT retains patient-derived tumor characteristics, including gene expression signatures and therapy responses, and has been validated across multiple cancer types, including head and neck cancers, thereby providing a robust organotypic platform for personalized oncology research [113].

Additionally, a well-known side effect of radiotherapy is tasting dysfunction, with approximately 96% of HNC patients experiencing acute taste disorders. While many patients experience partial recovery, 25% suffer from long-term dysfunction [114]. Taste progenitor/stem cells expressing Lgr5 or Lgr6 form spheroid structures in vitro, differentiating into taste bud cells [115]. Wu et al. developed a taste organoids-on-a-chip system that incorporates taste progenitor cells and microelectrode arrays (MEA) to act as taste axons, enabling the detection of taste information via real-time extracellular potential recordings [116]. By analyzing differences in the extracted features from these recordings, the system can biologically identify varying levels of sour, sweet, bitter, and salty, effectively reconstructing mammalian taste in vitro.

To capture the complex anatomical diversity of HNC, which spans soft connective tissues, glandular epithelium, and mineralized structures such as bone, recent OOC advances have emphasized tissue-specific microenvironment modeling. For soft tissue reconstruction, polydimethylsiloxane (PDMS) remains a widely used substrate due to its tunable stiffness (down to ~ 100 kPa) and optical properties. Mechanical modulation of PDMS, along with temperature-sensitive crosslinking of gelatin-based hydrogels, enables construction of dynamic ECM-mimetic scaffolds that emulate stromal heterogeneity and guide tumor cell behavior [117]. To replicate mucosal or glandular tissues, decellularized extracellular matrix (dECM) derived from native tracheal mucosa has been integrated into asthma disease model [118]. This tissue-specific matrix preserves native biochemical cues and structural organization, supporting epithelial-endothelial-stromal co-culture and offering a physiological basis for salivary or oral gland tumor modeling [119].

In the context of mineralized tissue modeling, PDMS alone is insufficient to simulate the high stiffness of osseous structures (~ GPa range). Therefore, hydroxyapatite (HA)-coated scaffolds have been incorporated into bone marrow-on-a-chip systems to provide bone-like rigidity and topographical features suitable for osteogenic or hematopoietic applications [120]. Sieber et al. demonstrated that HA-coated zirconia scaffolds enabled long-term maintenance of hematopoietic stem cells under perfused conditions [121], suggesting strong applicability in modeling jawbone invasion in HNC. Furthermore, multi-tissue OOC systems interconnecting organs such as the heart, liver, bone, and skin via microvascular channels have been used to simulate pharmacokinetics and identify circulating microRNA biomarkers during chemotherapy [122], as shown by Ronaldson-Bouchard et al. Collectively, these innovations enhance the relevance of OOC platforms for modeling spatially distinct tissue interactions and disease processes in HNC.

Current challenges, and future directions

Models of head and neck tumors across different subtypes

HPV-positive and HPV-negative HNC patients are often driven by different carcinogenic factors, resulting in distinct tumor microenvironments and immune responses, which also lead to varied treatment sensitivities. In research comparing viral-driven versus carcinogen-driven HNC, Cillo et al. conducted single-cell transcriptomic analysis on immune cells from peripheral blood and tumors of both HPV-negative and HPV-positive HNSCC patients, as well as healthy donors. They found distinct transcriptional features in the immune cells of these two patient groups, particularly in helper CD4 + T cells and B cells, which exhibited notable differences [126].

Currently, researchers are using cell lines such as HPV-positive SCC090, SCC152, SCC154, and HPV-negative SCC072 to study the mechanisms of HNC progression [127]. Additionally, they use Cas9 (pCas9) and gRNA plasmids targeting HPV16 E6/E7 and HPV18 E7/E6 to simulate HPV infection [128, 129]. However, cell-line-based studies have significant limitations in accurately modeling the HNC tumor microenvironment (TME) and immune infiltration characteristics.

Furthermore, due to the unique nature of HPV, which cannot be cultured in vitro, the pathogenesis of HPV-related diseases and their interactions with the host remain unclear [129]. Previously, Hu et al. studied HPV-driven cervical precancerous lesions and tumorigenesis by constructing long-term expandable organoids in a natural HPV infection state, replicating different HPV infection statuses, such as free and integrated HPV infection [130]. This culture strategy, when combined with organ-on-a-chip technology that incorporates interactions between stromal cells, immune cells, and microvascular flow, offers a more accurate reconstruction of the HPV-infected tumor microenvironment. Researchers can better observe how viral infection alters immune responses, the composition of the tumor microenvironment, and immune evasion mechanisms of tumor cells. This approach has significant implications for studying the pathogenesis and disease progression of HPV-driven HNC.

Additionally, HPV-positive patients are more likely to experience lymph node metastasis. Lymph node organ-on-a-chip models have increasingly been used in studies of cancer metastasis and immune evasion [131]. German et al. developed a reusable microfluidic device containing a 3D breast tumor spheroid, simulating secondary tumors following lymphatic metastasis in collagen-based sponge micropores [132]. Shim et al. designed a dual-organ chip to simulate communication between tumors and lymph nodes [133]. They cultured fresh tissue slices in two separate chambers and found that the tissue survived well. Notably, the protein communication between tumors and lymph nodes exhibited significant immune suppression compared to the normal tissue-lymph node communication. These lymph node organ-on-a-chip models provide valuable insights into simulating lymph node metastasis in HPV-positive patients, as well as understanding molecular mechanisms and screening for precision therapeutic drugs.

Future perspectives and challenges in the clinical translation of head and neck Tumor-on-a-Chip platforms

Different HNC subtypes exhibit distinct disease manifestations. To better advance the clinical application of HNC tumor-on-a-chip platforms, following strategies are proposed here: (A) Explore the mechanisms of different carcinogen-driven HNC types to uncover the key biological processes involved in tumorigenesis; (B) Conduct anticancer drug testing, which includes preclinical validation of new drugs, as well as the creation of personalized patient models to identify the most suitable treatment options through drug screening experiments; (C) Simulate lymphatic and distant metastasis in HNC to provide new insights and methods for studying recurrent/metastatic HNC (Fig. 6).

Fig. 6.

Construction strategy of HNC Tumor-on-a-chip

Organ-on-a-chip technology faces multiple challenges in its translational journey. Compared with traditional in vitro models, constructing and operating OOC systems require specialized expertise, fine manual skills, and comprehensive training in chip fabrication and device handling, which limits widespread adoption. Additionally, commonly used soft lithography techniques are labor-intensive and time-consuming, and rising demand will increase production costs and turnaround times. Therefore, there is an urgent need to develop more user-friendly designs, streamlined protocols, and efficient data acquisition and analysis tools to improve overall accessibility and efficiency.

In addition to engineering challenges, several systemic barriers continue to hinder the clinical translation of OOC technologies. Reproducibility and inter-laboratory variability remain unresolved issues, as most platforms still rely on internally developed validation protocols without harmonized performance standards, making data comparison and regulatory acceptance difficult [134]. Furthermore, current OOC systems often overlook key biological factors such as patient-specific stromal heterogeneity and immune cell recruitment—both of which are critical determinants of tumor progression and therapeutic response. While reviews such as that by Van Os et al. have proposed strategies including ECM tuning and chemokine gradient generation these approaches remain inconsistently implemented and lack standardized workflows across studies [135].

In response to these challenges, increasing attention has been directed toward the standardization of OOC systems. Current platforms often use lab-specific materials and protocols, hindering reproducibility and clinical scalability. Several international consortia—including the European Organ-on-Chip Society and the UK OOC Network—are actively working to establish standardized fabrication materials, operational workflows, and biological validation criteria to facilitate cross-platform comparability and regulatory acceptance [136, 137]. Moreover, efforts to replicate patient-specific biological complexity, such as stromal heterogeneity and immune cell dynamics, are ongoing. Strategies including ECM customization, stromal co-culture, and immune-tumor interface modeling have been proposed [135, 138], though their implementation remains inconsistent and rarely supported by unified technical guidance [139]. With advances in recapitulating tissue microenvironments, OOC platforms are increasingly recognized as promising alternatives to animal models in preclinical drug screening and toxicology assessment [88]. However, their application in guiding personalized clinical decision-making remains limited due to challenges such as system complexity, lack of standardized protocols, and poor integration with clinical workflows like biopsy processing and rapid turnaround requirement [140].

Furthermore, to tackle scalability, fabrication efficiency, and result analysis challenges, integrating robotics and artificial intelligence (AI) for automating tasks [141]—such as chip operation and data collection—becomes crucial [142]. Machine learning (ML) can significantly advance data analysis processes. With progress in deep learning, traditional methods of analyzing tumor-related changes have been revolutionized. For instance, identifying abnormal vascular morphology in tumors traditionally required cell fixation, immunostaining, confocal microscopy, and manual interpretation. New virtual staining techniques now enable the automatic identification of 3D vascular structures from bright-field images, bypassing laborious steps and providing rapid, reliable quantification of morphological changes. Deep learning has demonstrated clear advantages in analyzing complex structures like blood vessels and can similarly be applied to study immune cell behaviors and functions. By integrating high-throughput 3D cell culture in microfluidics with machine learning, researchers have successfully demonstrated the synergy between multi-parameter systems (MPS) and advanced data analytics. This approach not only offers rich 3D insights into the cancer immunity cycle but also lays the foundation for data-driven discoveries [143].

Lastly, as an emerging in vitro model, tumor-on-a-chip technology would effectively replicate precancerous lesions and tumor microenvironments while preserving patient heterogeneity, e.g. maintaining tissue characteristics, like ECM, cell population, viability, and enabling prolonged and synchronized drug response testing [144]. These would ultimately support precision oncology and improving therapeutic outcomes for head and neck cancer patients. Notably, the use of patient-derived OOC models brings forth important ethical considerations that must not be overlooked in translational cancer research. Prior to tissue collection, it is essential to obtain comprehensive informed consent, ensuring that participants are fully aware of how their biological materials will be utilized in experimental contexts. In parallel, rigorous data governance is necessary to protect patient confidentiality and comply with privacy regulations. As OOC platforms evolve toward more individualized and digitally integrated systems, the establishment of ongoing ethical oversight will be crucial in aligning scientific progress with the protection of patient autonomy and data integrity.

Conclusion

The design of tumor-on-chip platforms has evolved in recent years, trying to reconstruct and recapitulate more accurately the microphysiology of the tumor microenvironment. In the future, more precisely designed head and neck tumor chip models would be constructed to enhance the chips’ fidelity and complexity, reproduction of individual patient’s tumor microenvironment, providing stronger support for a deeper understanding of the pathophysiology of head and neck tumors, and exploring and developing new treatment methods and drugs.

Author contributions

Chen Lin, Zilin Zhang, and Feili Yang contributed equally to this research by drafting the manuscript, reviewing the literature, and writing the main sections of the article. Shanshan Gu and Jiyang Zuo contributed to the literature review and discussion sections. Yi Wu and Jing Zhang assisted with the structural design and revisions of the manuscript. Chen Lin, Tiantian Zhou and Yuna Zhang contributed to the design and preparation of figures and tables. Zaozao Chen, Zhongze Gu, and Zhisen Shen supervised the project, provided critical revisions, and coordinated the overall research efforts. All authors have read and approved the final manuscript.

Funding

This research was supported by the Ningbo Top Medical and Health Research Program (No. 2023030514), Ningbo Clinical Research Center for Otolaryngology Head and Neck Disease (No. 2022L005), the Medical and Health Research Project of Zhejiang Province (2025KY221, 2025KY222), the National Natural Science Foundation of China (Grant No.82341032), Science and Technology Project of Jiangsu Province (GrantNo.BK2024001144), Key R&D Program of Guangxi Province (Grant No.2023AB08121), the Educational Science Planning Project of Ningbo City (2025YGH007), and the Ningbo Public Welfare Research Program Project(2023S078).

Data availability

Not applicable.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

All authors agreed to publish this review.

Competing interests

The authors declare no conflict of interest.