Materials and Surface Chemistry of Organ-on-Chip Microdevices: Polymer Networks, Interfacial Functionalization, and Small-Molecule Partitioning in Pharmaceutical Microphysiological Systems

1. Introduction

Pharmaceutical research and development remains dominated by a translational failure that has proven remarkably resistant to decades of methodological refinement: the majority of drug candidates that succeed in preclinical animal studies still fail when tested in humans, most often on grounds of unanticipated toxicity or lack of efficacy that the preceding animal and 2D cell-culture models did not predict. The financial consequence of this gap is substantial — the fully loaded cost of bringing a single new therapeutic to market was estimated at approximately US$2.23 billion in 2024, a figure that reflects not the cost of successful programs alone but the amortized cost of the roughly nine in ten candidates that enter human trials and do not reach approval.[1] Animal testing, meanwhile, carries its own well-documented limitations as a predictor of human response, together with ethical and, increasingly, legislative pressure to reduce its use under the long-standing 3Rs framework (replacement, reduction, and refinement of animal use), first articulated by Russell and Burch in 1959 and since embedded in regulatory guidance on both sides of the Atlantic.[2,3]

Organ-on-chip (OoC) technology — also termed microphysiological systems (MPS) or tissue chips — emerged from bioengineering laboratories over the 2010s as a candidate response to this problem. Rather than culturing cells in a dish, OoC devices place living human cells within microfabricated channels that reconstitute key physical and biochemical features of a functional organ unit: polarized tissue-tissue interfaces, physiological fluid flow and shear stress, and, in several designs, cyclic mechanical strain analogous to breathing or peristalsis. Landmark early work demonstrated that a microengineered ‘lung-on-a-chip,’ built around a porous, stretchable membrane separating alveolar epithelium from pulmonary microvascular endothelium, could reproduce organ-level physiology — including cytokine- and bacteria-induced inflammatory responses and drug-toxicity-induced pulmonary edema — that static culture systems could not.[4,5] Comparable designs have since been extended to liver, kidney, gut, heart, skin, and blood-brain-barrier models, and to multi-chip ‘body-on-a-chip’ configurations intended to capture cross-organ drug distribution and metabolism.[6] The U.S. National Institutes of Health formalized federal investment in this trajectory in 2012 through the NIH/NCATS Tissue Chip for Drug Screening program, funded through the Cures Acceleration Network, whose stated purpose is to develop bioengineered devices that better predict human drug safety and efficacy than existing preclinical models and, ultimately, to combine single-organ chips into an integrated body-on-a-chip system.[7] The program remains active in 2026, most recently through the Tissue Chips in Space 2.0 initiative, announced in February 2026, to study age-related disease biology aboard the International Space Station, and through the newer Complement-ARIE (Complementing Animal Research in Experimentation) Common Fund program, which targets non-animal methods more broadly.[7,8]

A decade and a half on from the earliest lung- and gut-chip demonstrations, the relevant question for the pharmaceutical industry is no longer whether OoC devices can reproduce interesting biology in a laboratory setting, but whether — and under what conditions, for which organ systems, and against which regulatory pathway — they can be relied upon as evidence within a regulated drug development pipeline. This perspective addresses that question in depth, beginning from the chemistry of the devices themselves. Section 2 details the materials and surface chemistry that differentiate OoC platforms from conventional in vitro assays — polymer network formation, surface activation and bonding chemistry, bioconjugation chemistry for extracellular-matrix immobilization, and, in particular, the small-molecule sorption chemistry that governs whether the measured drug concentrations on a given platform can be trusted. Section 3 reviews organ-system-specific pharmaceutical evidence in detail: liver, kidney, heart, blood-brain barrier, gut, and immune-competent chips. Section 4 surveys the commercial landscape of organ-chip developers. Section 5 details the U.S. regulatory pathway and specific submissions that have tested it. Section 6 covers the European and international regulatory picture. Section 7 summarizes market and program-level adoption trends. Section 8 situates organ chips in relation to the competing and complementary organoid platform. Section 9 catalogs the technical and institutional barriers identified in the most rigorous independent assessments available. Section 10 considers the growing role of artificial intelligence and digital-twin approaches, and Section 11 offers a set of near-term, organ-system-specific recommendations.

2. Materials and Surface Chemistry of Organ-on-Chip Systems

The physiological fidelity and, critically, the pharmacological trustworthiness of an organ-chip platform are governed at root by chemistry: the polymer network chemistry of the device substrate, the surface chemistry used to bond layers and immobilize biomolecules, and the interfacial chemistry that governs how a dissolved small-molecule drug partitions between the aqueous culture medium and the solid device material. This section addresses each in turn, because — as later sections make clear — the pharmaceutical use case a given chip can credibly support is, in a real sense, dictated by these chemical properties rather than by cell biology alone.

2.1. Polymer Network Chemistry of the Device Substrate

The dominant organ-chip substrate remains poly(dimethylsiloxane) (PDMS), a silicone elastomer formed by platinum-catalyzed hydrosilylation cross-linking between vinyl-terminated poly(dimethylsiloxane) prepolymer chains and a hydride-functional cross-linker, typically supplied as a two-part kit (commercially, Sylgard 184) mixed at a base-to-curing-agent ratio — most commonly 10:1 by weight — that directly determines cross-link density and, downstream, the material's elastic modulus, gas permeability, and optical clarity.[9] PDMS is favored specifically because this cross-linking chemistry is mild (curing at 60–80 °C, well below temperatures that would denature biological molecules present during fabrication), because the cured network is highly permeable to oxygen and carbon dioxide (supporting long-term cell viability without active gas exchange hardware), and because it is optically transparent from the near-UV through the near-IR, permitting real-time bright-field and fluorescence microscopy directly through the device wall.[9,37] Device molds are produced by photolithographic patterning of SU-8, a multifunctional bisphenol-A-type epoxy novolac photoresist that cross-links via cationic ring-opening photopolymerization initiated by a triarylsulfonium photoacid generator upon ultraviolet exposure; the resulting highly cross-linked, chemically resistant epoxy network permits the same silicon master to be reused for repeated PDMS replica molding, the fabrication route through which the microchannel geometries described in Section 3 are physically defined.[9]

2.2. Surface Activation and Covalent Siloxane Bonding Chemistry

Two cured PDMS layers do not bond spontaneously; irreversible device assembly relies on a specific, well-characterized surface chemistry. Oxygen plasma or corona discharge treatment generates high-kinetic-energy electrons that ionize the process gas, and the resulting reactive oxygen species bombard the PDMS surface, cleaving nonpolar surface methyl (Si–CH₃) groups and oxidizing the exposed silicon centers to generate polar silanol (Si–OH) groups — a reaction that can also be driven by UV/ozone exposure or piranha (H₂SO₄/H₂O₂) etching, each generating silanol groups through a distinct oxidative pathway.[37] This silanol-rich surface is transiently hydrophilic and, critically, reactive: when two freshly activated PDMS surfaces are brought into contact, surface silanol groups undergo a condensation reaction to form covalent siloxane (Si–O–Si) bridges across the interface, and subsequent mild heating (60–80 °C) drives this condensation toward completion, producing irreversible interlayer bonds with reported strengths on the order of several hundred kilopascals.[37] This bonding chemistry is time-sensitive: surface silanol groups are thermodynamically unstable relative to the native hydrophobic surface, and PDMS's very low glass transition temperature (below approximately −120 °C) permits rapid diffusion of low-molecular-weight, non-cross-linked oligomeric chains from the bulk to the activated surface, re-covering it and restoring hydrophobicity — a phenomenon termed hydrophobic recovery that occurs over a window of minutes to a few hours and that in practice dictates how quickly two plasma-treated layers must be brought into contact during device assembly.[37]

2.3. Silanization and Bioconjugation Chemistry for Extracellular-Matrix Immobilization

The same silanol-terminated PDMS surface chemistry that enables interlayer bonding is exploited, via a different reaction pathway, to immobilize the extracellular-matrix (ECM) proteins that render a bare polymer channel receptive to cell attachment. Alkoxy- or chlorosilane coupling agents — most commonly aminosilanes such as (3-aminopropyl)triethoxysilane (APTES) — react with surface silanol groups to form covalent Si–O–Si linkages, and subsequent thermal curing promotes lateral cross-linking into a self-assembled monolayer that displays a defined density of primary amine groups at the outward-facing surface.[37] These amine groups serve as anchor points for covalent immobilization of ECM proteins (fibronectin, collagen I, laminin) via carbodiimide (EDC/NHS) coupling chemistry, which activates protein carboxyl groups for amide bond formation with the surface amines — a route that yields a more stable, defined protein layer than the alternative and more commonly used approach in the specific protocols reviewed in this article, simple physical (non-covalent) adsorption, in which the hydrophobic domains of a globular ECM protein partition onto the native hydrophobic PDMS surface through hydrophobic interaction, a process that is operationally simpler but carries a documented risk of partial protein unfolding and denaturation at the polymer interface, with consequences for the fidelity of the integrin-mediated cell adhesion the coating is meant to support.[37]

2.4. Small-Molecule Sorption Chemistry: The Central Pharmacological Confound

The chemical property of PDMS most consequential to its use specifically in pharmaceutical (as opposed to purely biological) applications is small-molecule sorption. PDMS's cross-linked siloxane network has substantial free volume and a low glass transition temperature, properties that together allow small, sufficiently lipophilic solutes to dissolve into and diffuse through the bulk polymer rather than merely adsorbing at its surface — a bulk-absorption mechanism, chemically and quantitatively distinct from the surface-adsorption behavior of more rigid, glassy thermoplastics.[38] Quantitative partitioning studies across a panel of pharmaceutically active compounds spanning a wide lipophilicity range (logP from approximately −0.07 to 5.13) demonstrate the practical severity of this effect: for imipramine (logP ≈ 4.80), aqueous-phase concentration in a PDMS microchannel fell from a nominal 100 µM to as little as 0.0384 µM after 24 hours of incubation — a greater than 2,500-fold depletion — while the same compound in a cyclic olefin copolymer (COC) channel retained a substantially higher 31.5 µM under identical conditions.[38] For loperamide (logP ≈ 5.13), only 37.8% of the compound washed out of a PDMS channel within the first 5 hours of a release experiment, compared with 71.5% washout from COC over the same interval, reflecting PDMS's bulk-absorption-driven, diffusion-limited release kinetics as against COC's faster, surface-adsorption-dominated desorption.[38] Regression analysis across the compound panel identified lipophilicity and rotatable bond count as significant predictors of sorption for both materials, with PDMS sorption additionally correlated with topological polar surface area and COC sorption more strongly correlated with hydrogen-bond-acceptor count and molecular weight — indicating that the two materials' sorption behaviors are governed by measurably different physicochemical mechanisms rather than simply differing in magnitude.[38] Because this sorption is concentration- and time-dependent and varies systematically with a compound's physicochemical descriptors, it introduces a device-material-dependent, compound-specific bias into any measured dose-response or toxicity threshold generated on an unaddressed PDMS chip — a bias that computational models have begun to explicitly predict and correct for, treating the device itself as a sink term within the pharmacokinetic model of the experiment rather than an inert vessel.[39]

2.5. Alternative Substrate Chemistries for Pharmacological Applications

This sorption chemistry has directly motivated substrate materials engineering aimed specifically at pharmaceutical use cases. Cyclic olefin copolymer (COC), a glassy, largely amorphous thermoplastic produced by copolymerization of cyclic olefins (such as norbornene) with ethylene, lacks PDMS's open, rubbery free volume and correspondingly exhibits predominantly surface-limited, more readily reversible small-molecule adsorption rather than bulk absorption, as demonstrated by the partitioning data in Section 2.4.[38] Styrene-ethylene-butylene-styrene (SEBS) triblock copolymer elastomers, produced via hydrogenation of styrene-butadiene-styrene block copolymers, have been engineered specifically as injection-molded or extruded microfluidic substrates with substantially reduced drug absorption relative to PDMS while retaining useful elastomeric flexibility, positioning them as a candidate substrate where mechanical actuation (Section 3.5) must be combined with lower sorption than PDMS provides.[40] Castable polyurethane elastomers offer an additional alternative, combining PDMS-like optical clarity and rapid, low-temperature curing with a substantially more polar polymer backbone chemistry that reduces the partitioning of lipophilic small molecules relative to PDMS.[13] A comprehensive review of materials beyond PDMS for organ-chip and microphysiological system fabrication catalogs this broader shift toward thermoplastics, alternative elastomers, and hybrid material strategies explicitly in response to the sorption and chemical-inertness limitations documented in Section 2.4 and elsewhere in this section, alongside considerations of manufacturability at scale.[41,42] No alternative substrate chemistry yet matches PDMS's combined optical clarity, gas permeability, low-temperature soft-lithographic processability, and ease of surface functionalization (Section 2.2 and Section 2.3) in a single material, so substrate selection in practice remains a compound-class-specific and context-of-use-specific engineering trade-off rather than a solved problem — a conclusion with direct consequences for how organ-chip pharmacological data should be interpreted, discussed further in Section 9.8.

2.6. Mechanical Actuation and in-Line Biosensing Chemistry

A defining structural feature across most organ-chip designs is a thin, porous, often flexible membrane that separates two (or more) microchannels, allowing an epithelial or parenchymal cell layer on one side to be juxtaposed with a vascular endothelial layer on the other — reconstituting the tissue-tissue interface that conventional monolayer or Transwell culture cannot fully capture, since the latter formats typically omit organ-specific endothelium entirely. Where mechanical activity is physiologically relevant — breathing motion in the lung, peristalsis in the gut, contraction in the heart — this membrane or substrate can additionally be actuated to apply cyclic strain, and multiple published studies show that mechanically dependent phenotypes (villus morphogenesis and CYP450 activity in gut models; nanoparticle transport and cytokine-induced edema in lung models) are only observed under dynamic, not static, culture conditions.[4,5]

Because organ chips are intended to generate quantitative, regulator-facing data rather than qualitative observations alone, in-line biosensing has become a central design consideration. Common continuous readouts include transepithelial/transendothelial electrical resistance (TEER) as a barrier-integrity metric — an electrochemical impedance measurement across the ion-permeable cell layer, used across gut, lung, kidney, and blood-brain-barrier chips alike — dissolved oxygen and pH for metabolic activity, multi-electrode arrays for cardiac and neuronal electrophysiology, and outlet sampling for secreted biomarkers and drug metabolites.[9] A decade-spanning European regulatory review of organ-chip use in medicines safety assessment (2010–2020) found that most published OoC toxicity studies over that period measured only a single endpoint, typically viability or cell death, rather than the multi-parametric readouts (barrier function, metabolic activity, electrophysiology, and secreted biomarkers together) that a regulator would need to distinguish mechanisms of toxicity or to build confidence in a specific context of use — a gap the review identifies as a central obstacle to regulatory qualification, independent of any given platform's underlying biological fidelity.[14] The move toward standardized, continuously logged, multi-parametric sensor data — rather than single-endpoint microscopy alone — is accordingly a direct response to this regulatory expectation, as discussed further in Section 6 and Section 7.

3. Organ-Specific Pharmaceutical Evidence

Not all organ-chip platforms carry equal evidentiary weight, and the pharmaceutical relevance of the technology is best assessed organ system by organ system rather than as a single undifferentiated category. This section reviews the six systems for which the most substantial pharmaceutical-relevant evidence base currently exists.

3.1. Liver: Drug-Induced Liver Injury (DILI)

Drug-induced liver injury is among the most common causes of clinical-stage and post-market drug withdrawal, and it is the pharmaceutical application in which organ-chip technology has advanced furthest along the U.S. regulatory pathway. A decade-spanning European regulatory review identified liver as one of the three most frequently modeled organs in the OoC literature, alongside kidney and heart, with hepatotoxicity and metabolism-induced toxicity dominating the published safety-assessment use cases.[14] Emulate's Liver-Chip, evaluated in blinded validation testing, has been reported to achieve 87% sensitivity and 100% specificity for detecting known hepatotoxic compounds — performance that directly led to its designation as the first organ-chip platform admitted to the FDA's ISTAND qualification program (detailed in Section 5).[15,16]

3.2. Kidney: Drug-Induced Kidney Injury (DIKI) and Transporter-Mediated Clearance

Drug-induced kidney injury is estimated to account for roughly 10% of failures across both the preclinical and clinical phases of drug development, making the kidney a high-value target for improved in vitro prediction.[17] Kidney-on-chip platforms incorporating proximal tubule epithelium under physiological flow have demonstrated nephrotoxicity detection for a defined reference-compound panel including cisplatin, gentamicin, adriamycin, tenofovir, tobramycin, and cyclosporine A, and have shown improved prediction of transporter-mediated drug clearance relative to conventional static in vitro transporter assays — directly relevant to renal drug-drug interaction and nephrotoxicity assessment during preclinical development.[17,18] Beyond the proximal tubule, glomerulus-on-a-chip designs have been used to assess albumin filtration and podocyte injury, and multi-organ liver-kidney chip configurations have been used to evaluate how hepatic metabolism of a parent compound alters its downstream renal toxicity — a mechanistic question that single-organ models cannot address.[17] Recognized limitations specific to kidney chips include difficulty obtaining mature, fully differentiated podocytes, incomplete epithelial polarization relative to native tubule, and incomplete vascularization of more complex multi-compartment designs.[17]

3.3. Heart: Torsadogenic and Structural Cardiotoxicity

Cardiac organ-chip platforms, built around human induced-pluripotent-stem-cell-derived cardiomyocytes (iPSC-CMs) under physiologically relevant electromechanical loading, have reported area-under-the-receiver-operating-characteristic (AUROC) values of 0.85 or higher for classifying torsadogenic (pro-arrhythmic) risk — a performance benchmark directly relevant to the cardiac safety pharmacology studies required for most small-molecule and several biologic drug candidates.[9] This work builds on a broader and more mature evidence base for iPSC-CM-based cardiac safety testing, including the multi-site Comprehensive in vitro Proarrhythmia Assay (CiPA) initiative, which has validated iPSC-CM electrophysiology as a component of a regulator-recognized paradigm for detecting drug-induced arrhythmia risk independent of the specific culture format used.[19,20] Heart-on-chip designs that integrate contractile-force and multi-electrode-array biosensing directly into the microfluidic device extend this iPSC-CM evidence base toward continuous, higher-throughput phenotypic drug screening beyond electrophysiology alone, including detection of structural (as opposed to purely electrophysiological) cardiotoxicity.[21]

3.4. Blood-Brain Barrier: CNS Drug Permeability

An estimated 98% of small-molecule pharmaceutical compounds and essentially all large-molecule biologics fail to cross the intact blood-brain barrier (BBB) at pharmacologically useful concentrations, making poor BBB penetration one of the most persistent bottlenecks in CNS drug development.[22] BBB-on-chip platforms incorporate brain microvascular endothelial cells — increasingly iPSC-derived — together with astrocytes and pericytes, and use TEER as the primary barrier-tightness metric; iPSC-derived brain microvascular endothelial cells cultured under physiological shear stress have been reported to reach TEER values considered physiologically meaningful for an intact barrier, an improvement over earlier static Transwell-based BBB models.[22] Reported applications include modeling impaired barrier function in Alzheimer's disease and Parkinson's disease and evaluating brain-tumor drug penetration, though the same review identifies PDMS's incompatibility with hydrophobic CNS-active compounds, incomplete replication of full BBB cellular and structural complexity, and cost/scale barriers to commercial-scale use as continuing limitations specific to this organ system.[22]

3.5. Gut: Absorption, Peristalsis-Dependent Differentiation, and Microbiome Coculture

Gut-on-chip platforms extend the same two-channel, membrane-separated architecture used in lung- and liver-chip designs to intestinal epithelium, typically incorporating trickling flow and cyclic peristalsis-like mechanical strain. Human Caco-2 intestinal epithelial cells cultured under these dynamic conditions differentiate into a polarized columnar epithelium and spontaneously form three-dimensional villus-like folds with basal proliferative crypts, elevated mucus production, and increased CYP450 enzyme activity — phenotypes not observed under static culture of the same cells, and directly relevant to first-pass intestinal drug metabolism and oral bioavailability prediction.[4] A further capability distinguishing gut chips from most other organ-chip systems is stable, extended coculture with living intestinal microbes on the luminal epithelial surface, supporting study of host-microbiome interactions relevant to oral drug absorption, metabolism, and microbiome-mediated toxicity.[4]

3.6. Immune-Competent Chips: Cytokine Release Syndrome and Immuno-Oncology

A more recent and rapidly developing category of organ-chip integrates immune cell populations — T cells, macrophages, dendritic cells, and others — into organ or tumor microenvironments using tunable extracellular-matrix hydrogels, multi-chamber microfluidic architectures that model vessel-to-tissue immune cell trafficking, and physiological interstitial flow and shear stress to drive immune cell recruitment and activation.[23] Named pharmaceutical applications include testing chimeric antigen receptor (CAR) T-cell efficacy, evaluating checkpoint-inhibitor resistance mechanisms, and modeling gut, lung, and blood-brain-barrier inflammatory and infectious processes.[23] Because unpredicted immune activation — most acutely cytokine release syndrome — has caused serious adverse events in first-in-human trials of immunomodulatory biologics, immune-competent chips are positioned as a potential preclinical safety-signal source specifically for this failure mode, though the same review identifies standardization across laboratory-specific designs, the continued absence of lymphatic drainage and systemic endocrine signaling in most designs, and the need for prospective validation against actual patient immunotherapy outcomes as unresolved barriers to regulatory use.[23]

4. The Commercial Organ-Chip Landscape

A number of specialized companies have emerged as the principal commercial developers of organ-chip platforms for pharmaceutical use, each generally associated with a particular organ-system focus, device architecture, or biosensing approach. Table 1 summarizes the platforms most frequently cited in the pharmaceutical and regulatory literature reviewed in this article; it is not exhaustive of the wider and still-growing commercial field.[24,25]

The commercial landscape illustrated in Table 1 reflects a broader pattern: platforms with a narrowly defined, single-organ context of use and a validated reference-compound panel (most visibly Emulate's Liver-Chip) have progressed furthest along the formal regulatory qualification pathway, while multi-organ and immune-competent platforms remain earlier-stage from a regulatory standpoint even where their underlying biology is well published, consistent with the GAO's finding (Section 9) that single-organ chips are considerably more developed than multi-organ body-on-a-chip systems.[26]

5. U.S. Regulatory Pathway and Status

The FDA Modernization Act 2.0, signed into law on December 29, 2022, amended the Federal Food, Drug, and Cosmetic Act to explicitly permit — though not require — the use of qualified non-animal test methods, including cell-based assays and microphysiological (organ-chip) systems, in place of animal testing to support an investigational new drug application.[2,27] The Act did not eliminate animal testing requirements outright; it removed a specific statutory reference to animal testing as a mandatory step, opening a regulatory door that sponsors and technology developers have since sought to walk through via existing FDA drug development tool (DDT) qualification programs.

The applicable qualification route for organ-chip platforms is the Innovative Science and Technology Approaches for New Drugs (ISTAND) pilot program, which explicitly covers the use of tissue chips (microphysiological systems) as a DDT category. ISTAND follows a mandatory three-stage process — a Letter of Intent (LOI), a Qualification Plan (QP), and a Full Qualification Package (FQP) — before a tool can be formally qualified for a defined context of use. As of the most recent public accounting, ISTAND had 16 projects in development across all technology types and zero fully qualified DDTs to date; across all DDT qualification programs combined, FDA reports 151 projects in progress against only 20 fully qualified tools, underscoring how demanding and lengthy the qualification process is even outside the organ-chip category specifically.[27]

Within that pipeline, Emulate's DILI-focused Liver-Chip became the first organ-chip platform admitted to ISTAND, with its Letter of Intent accepted on September 24, 2024 — a milestone, but only the first of three required stages, with full qualification realistically still years away.[15,27] Separately, an AI-based (non-chip) liver-injury prediction model received LOI acceptance around June 3, 2026, signaling that FDA's interest in improved DILI prediction extends across technology types and is not organ-chip-specific.[27]

FDA's broader posture toward non-animal methods has continued to shift in the same direction. On April 10, 2025, the agency announced a roadmap to phase out certain animal testing requirements for monoclonal antibody development, explicitly promoting organ-chip and other New Approach Methodologies (NAMs) as replacement evidence sources.[26,27] A draft NAM guidance document released March 18, 2026 names organ-chips directly and lays out validation principles for their regulatory use, including material-safety considerations such as the leaching of device materials into cultured tissue — a consideration directly connected to the PDMS absorption issues discussed in Section 2.[27] Legislatively, a further FDA Modernization Act 3.0 remains pending in the U.S. Congress as of mid-2026, having passed the Senate (S. 355) and the House (H.R. 2821) in July 2026 but not yet been reconciled or signed into law.[27]

A May 2025 U.S. Government Accountability Office (GAO) report — the most comprehensive independent assessment of the field to date — offers a more conservative reading of near-term regulatory reality: organ-chip technology will complement, but is not positioned to fully replace, animal testing in the near term, and single-organ chips (heart, liver, lung, kidney) remain considerably more developed than multi-organ ‘body-on-a-chip’ systems.[26] As of that report, FDA had received only limited organ-chip data within clinical trial applications, no organ-chip platform had been qualified through ISTAND, and the agency itself was reported to lack sufficient specialized staff to evaluate organ-chip submissions at scale — a capacity constraint as consequential to near-term adoption as any technical limitation of the devices themselves.[26]

5.1. Direct Use Within Individual IND Submissions

Independent of the formal DDT qualification process, several sponsors have already submitted organ-chip-derived data as supporting evidence within individual investigational new drug (IND) applications — a route that does not require prior FDA qualification of the platform itself, but allows chip data to inform a specific development program. Sanofi and True North Therapeutics, working with the tissue-chip company Hesperos, supported an IND for a rare neuromuscular disorder primarily with tissue-chip rather than animal data, with the associated Phase 2 trial beginning recruitment in April 2021. Dianthus Therapeutics, also working with Hesperos, cited organ-chip data in the IND submission for its complement-targeting antibody DNTH103, now in an active Phase II trial (NCT06282159). argenx, working with MIMETAS's OrganoPlate platform, announced an IND filing citing organ-chip data in July 2024.[27] These cases represent, at present, the clearest evidence that organ-chip data can carry regulatory weight in an active drug program, even in the absence of a formally qualified platform.

6. European and International Regulatory Landscape

The European Medicines Agency (EMA) situates organ-chip technology within its long-standing 3Rs framework and its more recent New Approach Methodologies (NAMs) initiative, and has published guidance on the principles of regulatory acceptance of 3Rs-oriented testing approaches more broadly.[3,28] A decade-spanning (2010–2020) review of the OoC literature from a European regulatory perspective — published in ALTEX, the principal peer-reviewed venue for alternatives-to-animal-testing research — found that liver, kidney, and heart were the most frequently modeled organs for safety assessment, consistent with the organ-specific evidence reviewed in Section 3, but identified single-endpoint assay design (Section 2), insufficient platform standardization, and unclear applicability across different therapeutic contexts as the principal obstacles to EMA qualification.[14] The review's core regulatory recommendation — that successful qualification requires a clearly defined context of use, an appropriately selected endpoint, and validation against a reference-compound set with known clinical outcomes — closely parallels the FDA ISTAND framework described in Section 5, suggesting convergence in what U.S. and EU regulators will require of any OoC platform regardless of jurisdiction.[14]

The EMA has taken further institutional steps in this direction, including a 2025 reflection paper addressing the reduction of non-human primate use in safety testing of human medicinal products through 3Rs implementation, and continued horizon-scanning of NAMs (including OoC platforms) for eventual incorporation into EU regulatory science.[3,29] The European Commission's Joint Research Centre — which hosts the EU Reference Laboratory for alternatives to animal testing (EURL ECVAM) — launched a review of models and methods used in cardiovascular research in September 2025, an initiative directly relevant to the cardiac-chip evidence summarized in Section 3.3.[30]

Beyond the United States and European Union, Japan has established a coordinated national microphysiological systems (MPS) initiative explicitly framed around innovation in pharmaceutical development and a defined path to regulatory acceptance, reflecting a broader international trend toward government-coordinated, rather than purely industry-driven, validation infrastructure for organ-chip technology.[31] Taken together, the U.S., EU, and Japanese regulatory tracks described in this section and Section 5 indicate that organ-chip qualification is increasingly a multi-jurisdictional undertaking, and that a platform validated principally against one regulator's context-of-use framework should not be assumed to transfer automatically to another's.

7. Market and Program-Level Adoption Trends

Commercial investment in organ-chip technology has grown alongside — and to some extent ahead of — its regulatory validation. Market analyses estimate the global organ-on-a-chip market at approximately US$186.1 million in 2025, rising to an estimated US$234.8 million in 2026, with one widely cited forecast projecting growth to roughly US$1.21 billion by 2033 at a compound annual growth rate near 26.4%.[32] Reported market composition offers a useful proxy for where the technology is currently being deployed: services (rather than device sales alone) accounted for the largest 2025 revenue share (51.8%), drug discovery represented the leading application (62.7% share) ahead of toxicology research, and pharmaceutical and biotechnology companies accounted for the large majority of end-user demand (72.4%), with academic institutions identified as the fastest-growing end-user segment. North America led regional market share (50.8% in 2025).[32]

This commercial growth sits alongside, and has been substantially catalyzed by, sustained public investment. The NIH/NCATS Tissue Chip for Drug Screening program, established in 2012 with flexible Cures Acceleration Network funding, has since supported a wide portfolio of academic tissue-chip development projects and, more recently, funded dedicated NIH Microphysiological Systems (TraCe MPS) centers explicitly tasked with making tissue chips ‘FDA-ready,’ alongside the 2024-launched Complement-ARIE Common Fund program targeting non-animal methods across NIH's portfolio more broadly.[7,8,26] Read together, these figures describe a technology whose commercial center of gravity remains internal pharmaceutical R&D decision-making — where organ-chip data can inform go/no-go calls without requiring regulatory sign-off — rather than regulatory submission itself, even as public funding programs explicitly target the regulatory-readiness gap.

8. Organ-on-Chip Versus Organoid Platforms

Organ chips are frequently discussed alongside — and sometimes conflated with — organoids, self-organizing three-dimensional cell aggregates typically derived from pluripotent or adult stem cells. The two technologies are complementary rather than substitutable, and the choice between them (or the increasingly common combination of the two, ‘organoids-on-chips’) should follow from the specific pharmaceutical question being asked rather than from a general judgment that one platform is superior.[33,34]

Organoids capture genetic and histological features of native human tissue effectively, and — because they can be derived from a specific patient's cells — are particularly well suited to personalized medicine applications and to oncology research requiring retention of tumor heterogeneity. Their principal limitations are structural and reproducibility-related: most organoids lack integrated immune, nervous, and circulatory systems; larger organoids commonly develop a necrotic core from inadequate nutrient and oxygen diffusion to their interior; and dependence on variable, batch-to-batch-inconsistent hydrogel matrices such as Matrigel constrains reproducibility across laboratories and production runs.[33]

Organ-chip systems, by contrast, provide a dynamic, continuously perfused environment with precise control over chemical gradients, mechanical stimuli, and shear stress, and — as reviewed in Section 3 — support the tissue-tissue interfaces and multi-organ linkage that organoids generally cannot replicate. Their principal limitations are practical rather than biological: fabrication and operation require specialized expertise and, in most designs, cleanroom access; multi-tubing fluidic connections introduce contamination risk and are susceptible to flow-obstructing air bubbles; and both platform standardization and manufacturing scalability remain significant unresolved challenges, echoing the standardization concerns raised for both regulatory pathways in Section 5 and Section 6.[33] Emerging hybrid ‘organoid-on-chip’ approaches — placing organoids within a perfused, mechanically dynamic microfluidic chip — aim explicitly to combine the genetic and histological fidelity of organoids with the physiological dynamics of chip culture, and represent an active area of platform convergence rather than competition.[33,34]

9. Barriers to Broader Adoption

The 2025 GAO assessment provides the most systematically documented account of what currently separates organ-chip technology from routine pharmaceutical use, and its findings are summarized here — alongside organ-system-specific technical limitations already noted in Section 3 — because together they define the practical (rather than purely scientific) agenda for the field over the next several years.[26]

9.1. Human Cell Access and Diversity

Only an estimated 10–20% of purchased primary human cells meet the quality standards required for reliable organ-chip experiments, and available cell sources lack demographic diversity, with more than 70% sourced from White donors — a limitation with direct consequences for a technology whose central promise is capturing human-specific and, potentially, patient-population-specific biology.[26]

9.2. Cost and Operational Complexity

Initial capital investment in organ-chip infrastructure exceeds US$150,000 before accounting for recurring per-experiment costs of chips, cells, and reagents, and running a single experiment can require ten or more days of specialized staff time for preparation and execution — a throughput and cost profile that constrains adoption outside well-resourced pharmaceutical and large academic laboratories.[26]

9.3. Single-Endpoint and Standardization Gaps

As detailed in Section 2, much of the published OoC toxicology literature has relied on a single endpoint — typically viability or cell death — rather than the multi-parametric readouts a regulator needs to establish mechanism and confidence for a defined context of use, and both the GAO and the ALTEX regulatory review identify the continued absence of organ-chip-specific standards for terminology, measurement, and data reporting as a barrier that compounds this problem by preventing meaningful comparison across platforms and laboratories.[14,26]

9.4. Incomplete Biological Complexity: Vascularization and Immune Integration

Even the more advanced organ-chip designs reviewed in Section 3 generally omit at least one physiologically consequential system: kidney chips face documented difficulty achieving mature podocyte differentiation and full vascularization of multi-compartment designs; immune-competent chips lack lymphatic drainage and systemic endocrine signaling in most current designs, limiting their ability to model whole-body immune phenomena such as cytokine release syndrome with full fidelity; and BBB chips still incompletely replicate the full cellular and structural complexity of the native barrier.[17,22,23] These are not failures of the platform concept but active, well-documented areas of ongoing engineering work, cataloged in a 2024 review of advances and challenges in organ-on-chip technology toward mimicking human physiology and disease in vitro.[35]

9.5. Validation and Benchmarking Gaps

Perhaps the most consequential barrier remains the continued absence of sufficient benchmarking and validation studies comparing organ-chip outputs against both animal-model and human clinical outcomes across a standardized set of contexts of use. Without such benchmarks, sponsors and regulators alike lack a shared basis for judging when a given chip's performance is adequate to support a specific decision — the same conclusion reached independently by the U.S. GAO and the European ALTEX review despite differing regulatory jurisdictions.[14,26]

9.6. Data-sharing Constraints

Companies developing and using organ-chip platforms have generally been reluctant to share performance and validation data publicly, citing competitive positioning and intellectual property concerns — a dynamic that slows the field-wide accumulation of the comparative evidence described in Section 9.5, and that stands in some tension with the precompetitive data-sharing norms that have accelerated validation of other preclinical technologies.[26]

9.7. Regulatory Capacity

As noted in Section 5, FDA's own capacity to evaluate organ-chip submissions — in terms of specialized reviewer expertise — has been identified as a limiting factor independent of the platforms' technical readiness, suggesting that accelerating qualification timelines will require investment in regulatory science capacity alongside continued platform development.[26]

9.8. Substrate Chemistry Mismatch

A barrier distinct from those identified in the GAO and ALTEX assessments, but directly implied by the materials chemistry reviewed in Section 2, is the risk of mismatch between a chip's substrate chemistry and the physicochemical class of compound under study. A PDMS-based platform validated for a hydrophilic, low-logP compound class provides comparatively little assurance for a lipophilic, high-logP candidate, given the more-than-1,000-fold sorption differences documented in Section 2.4; conversely, a program's decision to adopt a lower-sorption substrate such as COC, SEBS, or polyurethane (Section 2.5) is not chemistry-neutral, since substrate change can alter surface functionalization chemistry, gas permeability, optical properties, and fabrication cost simultaneously. At present, published validation and regulatory-qualification packages (Section 5 and Section 6) rarely report substrate absorption behavior for the specific compound class the qualification targets, and none of the review-level sources synthesized in this article describe a standardized, compound-class-stratified sorption-correction protocol in general regulatory use — indicating that substrate chemistry selection remains, in practice, an underspecified and under-scrutinized step in most organ-chip pharmaceutical validation workflows.[38,39,41]

10. Artificial Intelligence and Digital-Twin Integration

A growing share of recent organ-chip literature concerns not the devices themselves but the computational layer built around them. Continuous sensor streams from TEER, electrophysiology, and metabolic monitoring generate substantially more data per experiment than endpoint assays, and machine learning methods are increasingly used to automate signal analysis, harmonize datasets collected across different chip platforms and laboratories, and flag safety signals that might not be apparent from any single readout in isolation.[9] This trend has precedent in adjacent iPSC-based cardiac safety testing, where deep-learning approaches applied to high-content imaging of iPSC-derived cardiomyocytes have already been used to detect cardiotoxic phenotypes at a scale and consistency difficult to achieve through manual scoring — an approach directly transferable to the sensor-rich data streams generated by cardiac and other organ chips.[36]

A further and more ambitious direction couples organ-chip measurements to in silico pharmacokinetic/pharmacodynamic models in a ‘digital twin’ framework, in which a computational model of a patient or patient population is continuously updated with chip-derived data to simulate drug response — an approach positioned as a bridge between the mechanistic, cell-level resolution of organ chips and the population-level predictions that regulatory and clinical decision-making ultimately require.[9,27] These approaches remain earlier-stage than the chip platforms themselves, but they directly address two of the adoption barriers identified in Section 9: the need for standardized, comparable output across platforms (Section 9.3), and the need to translate chip-level findings into a form directly usable for dose- and population-level regulatory judgments.

11. Outlook and Organ-System-Specific Recommendations

Taken together, the evidence reviewed in this article supports a specific and fairly narrow characterization of where organ-chip technology now stands: it is a maturing, regulator-engaged evidentiary category that has already demonstrated the ability to inform individual drug-development decisions in specific organ systems, but it is not yet — and is not imminently going to be — a wholesale substitute for animal testing or a routinely available pharmaceutical R&D tool outside well-resourced organizations. The GAO's six policy recommendations, directed principally at U.S. federal agencies, capture the priorities that would most directly close this gap, and are reproduced in summary in Table 2 alongside a brief note on their relevance to pharmaceutical sponsors specifically.

Beyond these general recommendations, the organ-specific evidence reviewed in Section 3 supports three additional, more targeted conclusions. First, sponsors evaluating where to invest limited organ-chip budget should weight liver-DILI and cardiac-torsadogenic applications most heavily in the near term, given their comparatively mature validation record (Section 3.1 and Section 3.3) and liver-chip's lead position within ISTAND (Section 5). Second, kidney, BBB, and gut chips, while scientifically well characterized, currently carry a thinner formal regulatory validation record and should be positioned by sponsors principally as internal mechanistic and go/no-go decision tools rather than as anticipated regulatory-submission evidence, pending further benchmarking (Section 3.2, Section 3.4, Section 3.5 and Section 9.5). Third, immune-competent chips represent the least regulatorily mature but arguably highest-unmet-need category, given the clinical severity of unpredicted immune activation events in immunomodulatory biologic development; sponsors working in this space should expect to contribute directly to the field's still-developing standardization and validation base rather than drawing on an existing one (Section 3.6). Fourth, and irrespective of organ system, given FDA's own stated capacity constraints, sponsors planning to rely on chip data for a regulatory submission should engage FDA early and directly — through existing INTERACT or pre-IND meeting mechanisms — rather than assuming ISTAND qualification status of a given platform is a reliable proxy for submission readiness (Section 5).

12. Conclusion

Organ-on-chip technology has traveled a genuine distance from laboratory demonstration to regulatory engagement over the past decade and a half: a platform capable of reconstituting organ-level physiology in a microfabricated device now has, in at least one instance, entered the first formal stage of FDA drug-development-tool qualification, organ-chip data have already informed several active clinical development programs, and coordinated public investment — from the NIH Tissue Chip program in the United States to Japan's national MPS initiative — has grown alongside a commercial market now approaching a quarter-billion dollars in annual size. At the same time, the most rigorous independent assessments available — the U.S. GAO's 2025 review and the decade-spanning European ALTEX regulatory analysis — are unambiguous that no organ-chip platform is currently qualified in any major jurisdiction, that validation and benchmarking against clinical outcomes remain insufficient across most contexts of use and most organ systems, and that structural barriers in cell sourcing, cost, single-endpoint assay design, incomplete biological complexity, and data sharing will need deliberate policy and industry intervention to resolve. The most useful framing for pharmaceutical stakeholders over the next several years is therefore not ‘animal-testing replacement,’ but rather a new, still-maturing, and organ-system-differentiated category of mechanistic and safety-signal evidence — one whose regulatory and commercial value will accrue chip system by chip system, organ by organ, and context of use by context of use, as the validation record documented in Section 3 through 9 continues to build.