Gastric organoids are self-assembling, multicellular structures that can be classified into two categories based on the cell sources required for their generation: adult stem cell (AdSC)- and iPSC-derived organoids^[12-14] [Table 1]. AdSC-derived organoids originate from stem or progenitor cells isolated from normal gastric tissue or patient biopsies and can maintain region-specific adult epithelial programs during long-term expansion^[19,41,42]. They are therefore well suited to postnatal disease modeling, epithelial host-pathogen studies, and patient-specific drug testing. By contrast, iPSC-derived gastric organoids are generated through the staged specification of pluripotent cells toward definitive endoderm, foregut, and gastric lineages by manipulating defined signaling pathways^[20,43-46]. iPSC-derived models can incorporate mesenchymal components and are especially valuable for developmental biology and congenital disease modeling, but their epithelial maturation generally remains more fetal-like than that of adult tissue.

GC is classified into four molecular subtypes by the Cancer Genome Atlas (TCGA), including Epstein-Barr virus (EBV)-positive, genomically stable (GS), microsatellite instability (MSI), and chromosomal instability (CIN)^[54]. GC patient-derived organoids (PDOs) of these four subtypes have been generated^[55]. Importantly, mutational spectrum overlap was observed between organoid cultures and parental tissue, indicating that PDOs could recapitulate the molecular patterns of the GC subtypes^[55,56]. In these systems, PDOs have been used to analyze responses to targeted drugs against identified molecular alterations.

The most direct personalized-medicine workflow begins with an endoscopic biopsy or surgical specimen, followed by the rapid establishment of a patient-derived GC organoid, histologic and molecular quality control, and functional testing against clinically relevant drugs^[57]. Molecular profiling can then be integrated with the phenotypic response to prioritize candidate therapies or identify resistance mechanisms. For instance, aberrant tumor cell proliferation induced by loss of tumor suppressor CDKN2A could be inhibited by the cyclin-dependent kinase 4/6 (CDK4/6) inhibitor^[21]. As another example, trastuzumab has been demonstrated to be effective against PDOs with ERBB2 amplification, consistent with the response of patients harboring the same genetic alteration^[55]. Intratumoral heterogeneity plays a key role in therapy response. GC PDOs established from normal, dysplastic, cancer, and lymph node metastases showed regional heterogeneity and different subclonal architectures, which resulted in sensitivity to different drugs^[21,58].

Considering that drug resistance is a major clinical challenge affecting the prognosis and survival of patients with GC, organoids have been explored to clarify the underlying mechanisms of drug resistance. It was found that 5-fluorouracil (5-FU), a first-line drug for GC, led to tumor relapse upon therapy^[59]. Ukai et al. developed GC PDOs resistant to 5-FU and identified that KHDRBS3 induced 5-FU resistance by modulating CD44 variant expression and promoting PDOs to acquire stem-like properties^[60]. Because the commonly used post-surgery chemotherapy [the combination of 5-FU and cisplatin (CDDP)] frequently resulted in acquired resistance, Wong et al. generated 5-FU + CDDP-resistant GC subtype PDOs and revealed that adenosine deaminase acting on RNA 1 (ADAR1) induced chemoresistance by augmenting stearoyl-CoA desaturase (SCD1) mRNA stability, resulting in reduced ER stress and enhanced self-renewal [Figure 3A]^[61]. Zhao et al. established 57 GC PDOs from 73 patients (78% establishment rate) and reported concordance between the PDO drug-response results and actual clinical response in 91.7% (11/12) of evaluable patients^[62]. As expected, the genes/pathways involved in proliferation and invasion are enriched in chemotherapy-resistant PDOs^[62]. McDonald et al. further demonstrated the feasibility of rapidly generating organoids from endoscopic gastric cancer biopsies and performing drug testing within approximately two weeks in a proof-of-concept workflow^[63]. These results are encouraging because they directly connect ex vivo testing to patient treatment response, but the sample sizes remain too small to establish routine clinical utility. While these models provide mechanistic and predictive information, the inadequate efficiency, assay turnaround time, tumor sampling, and absence of nonepithelial tumor microenvironment components are important constraints that must be overcome for clinical deployment.

Considering that different regions of the human stomach show distinct morphological and functional properties, Jones et al. generated multiregional assembloids via the self-assembly of region-specific gastric organoids from pediatric patients with unusual antral foveolar hyperplasia [Figure 3B]^[64]. The assembloids efficiently recapitulate hyperplastic-like antral regions characterized by impaired gastric acid secretion due to decreased mucin secretion and glycosylated H⁺/K⁺ ATPase β^[64]. Overall, these studies highlight PDOs as a potential tool for investigating individual treatment responses, as well as disease development and drug resistance mechanisms.

The gastric mucosa, the innermost layer of the stomach wall, serves as a protective barrier and contributes to acid, enzyme, and hormone production, immune surveillance, and epithelial renewal^[75,76]. Several stomach chips have been developed to recapitulate the structure and functions of the gastric mucosa in vitro. To study how the mucosa protects the gastric epithelial cells against damage by the acidity in the lumen, a microfluidic device producing a continuous stream of mucus against incoming acid was developed to mimic the gastric mucus acid barrier, using a fluorescent pH reporter to monitor the dynamics of H⁺ penetration into the mucin layer [Figure 4A]^[68]. The study demonstrated proton sequestration and retarded diffusion within mucus, illustrating how a reductionist chip can provide a quantitative barrier readout that is difficult to obtain from closed 3D organoid lumens.

To model gastric epithelium with lumen and rhythmical peristalsis, a stomach-on-a-chip (SoC) model was established by Lee et al. in which gastric organoids were encased within a microfluidic device and underwent cyclic expansion and contraction induced by a peristaltic pump [Figure 4B]^[22]. The robust 3D growth of gastric organoids in the presence of luminal flow could be recorded, providing a platform for the long-term delivery of agents into the gastric lumen for disease modeling and drug screening. To more authentically recapitulate the gastric mucosa, a SoC comprised of the epithelial cell monolayer, basement membrane, and lamina propria (gastric fibroblast-embedding collagen I gel) was developed [Figure 4C]^[71]. In the bioinspired SoC, a 2 μL min^-1 perfusion rate in a 0.20 cm wide and 0.05 cm high channel corresponded to an estimated epithelial shear stress of approximately 0.003 dyne per cm², which is deliberately within the physiological shear stress range experienced by gut cells in vivo (0.002-0.080 dyne per cm²). Peristalsis-like motion was realized by applying a vacuum to the microactuator at cyclical intervals. Sinusoidal actuation at 0.15 Hz produced surface stretching of approximately 10%. Next, the growth and integrity of the epithelial barrier were assessed by measuring the transepithelial electrical resistance (TEER). The results showed that the TEER values were above 100 Ω cm² at two days post-seeding and reached an average of approximately 157 Ω cm² after five days of culturing, characterizing a fully formed barrier. These studies identified that the mechanical actuation of the SoC plays a key role in the development of epithelial polarization and differentiation traits that are characteristic of the gastric mucosa, illustrating how device architecture can directly affect model performance rather than simply provide a culture container.

As the organoid-derived gastric epithelium shows apical-basal polarization, a bilaterally accessible patterned epithelial chip was designed to ensure the manipulation of both its apical and basal sides^[70]. Basal stem cell niche growth factor delivery could be mimicked to enable proliferative and differentiation patterns of gastric epithelium organoids consistent with the stomach owing to the improved accessibility of the system. Compared with conventional enclosed 3D organoids, such open-access designs facilitate longitudinal imaging and cell-resolved perturbation, although some 3D gland geometry is sacrificed for experimental observability.

Pathogen-mediated homeostatic disruption of the gastric mucosal barrier leads to various types of stomach disease pathogenesis, among which H. pylori is a representative pathogen^[77]. The replication of host-pathogen interactions in vitro could improve the study of the gastric mucosal defense mechanisms against pathogens and the development of therapeutic strategies. A human stomach microphysiological system (hsMPS) was built by cultivating human antral organoids and gastric mesenchymal stromal cells in the luminal and abluminal channels of a compartmentalized microfluidic device^[23]. H. pylori infection was modeled with the hsMPS, and trefoil factor peptide 1 (TFF1) was revealed as a critical element in gastroprotection against H. pylori [Figure 5]. Moreover, the introduction of peripheral blood mononuclear cells (PBMCs) demonstrated the immune cell recruitment and inflammation induction properties of H. pylori [Figure 5B], indicating the potential of hsMPS as a tool for uncovering the immune pathogenic mechanisms of H. pylori infection^[23].

To explore cell-type-specific host-pathogen interactions, Hofer et al. proposed the generation of a patterned homeostatic human gastric organoid-on-a-chip model with bilateral access and an apical environment maintaining a pH of approximately 3. This can model H. pylori niche establishment and the persistent colonization of the gastric epithelium^[24]. By incorporating a spatially confined culture and a dynamic acidity regulation system, this model overcomes the limitation of unipolar differentiation observed in conventional gastric organoids. The study demonstrates that under the physiological acidic luminal environment, the maturation of gastric pit cells on the chip is significantly enhanced. H. pylori colonization showed a graded response among different gastric epithelial cell subsets: basal progenitor cells predominantly secrete inflammatory mediators, while apical pit cells specifically activate tight junction remodeling and the lectin stress pathway. For the first time, this model enables the in situ observation of the dynamic formation of bacterial microcolonies and their authentic three-dimensional adhesion behavior, indicating that the nonuniform immune response of the host is the molecular basis for sustaining chronic infection. Overall, this study demonstrates why acid control and spatial epithelial patterning can qualitatively change the infection phenotype observed in vitro.

Clinical profiling indicates that dysbiotic communities enriched in taxa such as Fusobacterium and Neisseria can be associated with GC risk after H. pylori eradication, and mechanistic experiments have shown that selected gastric bacteria can induce distinct inflammatory and pro-oncogenic responses in gastric epithelial systems^[78]. However, the non-H. pylori gastric microbiome remains substantially less developed as an in vitro model. Therefore, future gastric chips should introduce a defined microbial community rather than a single pathogen. In addition, apical access should be combined with controlled oxygen tension, low-pH gradients, mucus production, and defined microbial consortia while quantifying microbial abundance, epithelial functionality, and metabolite exchange to extend from single-pathogen infection models toward experimentally controlled gastric ecosystem models.

Although established cell lines are suboptimal for cancer studies, investigators have developed microfluidic chips to characterize GC growth and drug treatments in vitro. Bakhchova et al. proposed the establishment of a polarized gastric epithelial monolayer-on-a-chip with continuous and stable fluid flow for the imaging-based analysis of chemotherapeutic drug responses^[69]. Considering that the tumor microenvironment fundamentally shapes drug response, a microfluidic tumor-on-chip platform encompassing endothelial cells and 3D-cultured GC cells was designed to mimic a vascularized tumor in vivo^[72]. The chip contains the chamber layer [for the cultivation of human umbilical vein endothelial cells (HUVECs) and GC cells embedded in extracellular matrix (ECM) gel] and the gas layer (for oxygen concentration modulation) separated by a permeable membrane^[72]. Upon hypoxic microenvironment modeling followed by CD8⁺ T cell perfusion, the system identified that hypoxia-induced forkhead box O3 (FOXO3a) upregulation mediates PD-L1 elevation, thereby contributing to immunotherapy resistance^[72]. This model is particularly suitable for investigating the impact of oxygen gradients, immune-cell trafficking, and perfused drug exposure on tumors that are not captured by epithelial organoids alone.

As patient-derived organoids provide a more faithful, personalized, and physiologically relevant model of human cancer compared with cell lines, gastric tumor chips based on patient-derived organoids have been established for personalized drug screening and treatment. In particular, Meng et al. proposed a microcapsule-based OoC that incorporates gastric organoid-loaded microcapsules into a microfluidic chip, enabling the dynamic perfusion of various chemotherapeutic agents^[79]. Heterogeneity in drug sensitivity was identified among organoids derived from different patients using this platform, which was in accordance with the responses of the corresponding patients. To further model tumor vasculature and the tissue-tissue interface of tumor and blood vessels, a vascularized tumor chip was built by integrating patient-derived gastric organoids with perfusable endothelium and tumor-blood vessel structures in interstitial tissues formed by stomach-derived decellularized ECM [Figure 6]^[73]. The clinical responses of patients treated with a Vascular Endothelial Growth Factor Receptor 2 (VEGFR2) inhibitor were reproduced using this platform. These studies highlight microfluidic chips as valuable platforms for the prediction of individual therapeutic responses and drug discovery, but they currently require a large quantity of tissue, long setup times, specialized fabrication, and assay-specific validation. Consequently, most current gastric tumor chips should be regarded as advanced preclinical or proof-of-concept platforms rather than clinically validated companion diagnostics. Translation will require predefined turnaround times, standardized drug screening metrics, blinded prospective comparison with clinical outcomes, and demonstration that chip-guided decisions improve outcomes beyond standard pathology and genomic biomarkers.

Modeling the body as a system of interconnected organs is a powerful strategy for gaining a holistic understanding of human physiology. Despite the ability of organoids to model individual organs, utilizing them to recapitulate organ-organ interactions remains challenging. With the advent of human-on-a-chip technology, various types of organoids have been integrated into a chip to simulate multiorgan interactions^[65,80,81]. For example, Jin et al. developed a microfluidic array-based multiorgan chip model through the coculturing of stem cell-derived liver, intestinal, and stomach organoids [Figure 7]^[74]. The organoids cultivated in distinct chambers were allowed to communicate via rocker-induced media flow. This work aimed to simulate in vivo bile acid homeostasis, which is regulated by the interaction between the liver and intestine. The platform showed reduced expression of CYP7A1, a bile acid synthesis enzyme, upon exposure to bile acids in the liver organoids due to the paracrine effect of the intestinal organoid, demonstrating the physiological crosstalk between different organs^[74].

Despite the progress that has been made, several challenges must be overcome to broaden the application of gastric organoid/OoC models. First, future models must maintain a stable apical pH gradient without nonspecific epithelial injury as the stomach combines a highly acidic lumen with a protected basolateral tissue compartment. Moreover, the glandular mucosa contains strong apical-basal and pit-gland spatial organization that is difficult to reproduce in flat monolayers, while closed spherical organoids restrict luminal access. Patterning, bilateral access, tubular/gland-like geometries, and controlled niche-factor gradients might provide solutions to some extent^[24,70]. In addition, gastric mechanics are multiscale, considering slow-wave activity, local wall strain, luminal mixing, and antral contractions, and they should therefore not be reduced to a single “peristalsis” frequency. In terms of cell sources, although primary gastric organoids overcome the drawbacks of established cell lines by recapitulating gastric region-specific structure and function, their use is hindered by limited availability and a lack of standard quality control metrics and procedures. While iPSC-derived gastric organoids provide an alternative, they fail to become fully mature and precisely model adult tissues in vivo^[82]. For stomach-on-a-chip systems, while polydimethylsiloxane (PDMS) is the most frequently used material for chip fabrication, it is not suitable for scalable production. In addition, its absorption of hydrophobic molecules might impair the accuracy of efficacy evaluation of drugs with high hydrophobicity^[83,84]. The exploration of other materials or nonfouling coatings could resolve these challenges^[84]. Furthermore, the experimental throughput of recent stomach chips must be increased to enable the high-throughput assessment of epithelial function in parallel and drug screening.

While the incorporation of cells in addition to patient-derived epithelial cells, such as endothelium, stromal, and immune cells, improves the physiological relevance of gastric organoid and OoC models, these cells are commonly not obtained from the same patient. Therefore, the establishment of stomach-on-a-chip models containing all cells from the same patient could improve personalized medicine. These fully personalized stomach chips could be used to more authentically recapitulate individual disease phenotypes and drug responses, which might help to reveal the impact of sex, age, and race on interpatient variability^[79]. Moreover, controlled microbial consortia should be incorporated to study how non-H. pylori communities interact with acid, mucus, and host tissue^[78]. The integration of real-time monitoring tools and additional downstream analytical strategies should be further investigated. Gastric organoid/OoC technology would benefit from the introduction of biosensors that allow in situ and live-cell analysis of the dynamics of complicated biological processes and treatment responses without disrupting gastric architecture and cell viability. In addition to single-cell transcriptomics, other omics techniques (e.g., metabolomic, proteomic, and epigenomic approaches) should be included to identify the mechanisms involved in gene regulation, protein expression, epigenetic variation, and host-pathogen interactions^[85]. Furthermore, the integration of an artificial-intelligence-based drug discovery platform into a human stomach-on-a-chip might accelerate therapeutics discovery for specific gastric disorders and aid in the development of drug delivery strategies with high gastric epithelium selectivity for minimized systemic toxicity^[86,87]. In parallel, microfluidic manufacturing and delivery strategies may be developed to improve the reproducible control of materials, gradients, and dosing^[88].

By contrast, standardization is required to convert technically sophisticated gastric models into reliable tools for clinical application. Recent human gastric organoid standards emphasize traceable cell sourcing, morphology, sterility, genomic stability, lineage-marker characterization, functional assays (such as acid or pepsinogen secretion where appropriate), and defined storage and quality-control procedures^[89]. For microphysiological systems, reproducibility frameworks additionally recommend prespecified acceptance criteria and the reporting of device, biological, and analytical variables^[90]. For stomach chips, a minimum reporting set should include the cell source and passage, ECM composition, channel dimensions, material identity, calculated shear stress, actuation frequency and strain, apical and basal pH, barrier baseline (e.g., TEER or permeability), assay timing, and interdevice variability. These metrics would allow laboratories to determine whether a model is reproducible and suitable for its intended regulatory or translational use.