Re-engineering insulin for oral delivery: structural modifications, advanced formulation strategies, and future directions

Abstract

Oral insulin delivery represents a transformative approach to diabetes management, offering improved patient compliance and physiological insulin delivery patterns compared to subcutaneous injection. However, multiple gastrointestinal barriers, including enzymatic degradation, mucus entrapment, epithelial impermeability, and first-pass metabolism, have limited oral bioavailability to below 1% for unmodified insulin. This review comprehensively examines contemporary strategies to overcome these barriers. We analyze structural modifications of insulin, including PEGylation, lipidation, cyclization, and glycoengineering, which enhance stability while maintaining biological activity. The analysis extends to sophisticated formulation technologies incorporating nanocarriers (polymer-based, lipid-based, inorganic nanocarriers, and metal organic frameworks), biomimetic systems, and stimuli-responsive mechanisms for protection and delivery. A central focus is on absorption-enhancing strategies, which range from chemical permeation enhancers to precise biological mechanisms like receptor-mediated transcytosis and other active transport pathways. Emerging tools such as microbiome-based carriers and smart devices are also discussed. Despite significant progress in preclinical models, challenges remain in manufacturing scalability, inter-patient variability, long-term safety, and regulatory approval. Future directions emphasize hybrid delivery systems, digital health integration, and personalized formulations. Realizing clinically viable oral insulin requires continued multidisciplinary collaboration addressing biological, technological, and translational barriers to transform diabetes care.

1. Introduction: the elusive goal of oral insulin

The discovery of insulin by Banting and Best in 1921 revolutionised diabetes treatment, transforming a once-fatal disease into a manageable condition (Mathieu et al. 2021). Since then, insulin therapy itself has undergone significant innovations (Vecchio et al. 2018). In 1922, insulin was first used in human treatment. In 1936, long-acting insulin (such as protamine insulin) was developed. In 1955, Sanger determined the amino acid sequence of insulin. In 1978, recombinant DNA technology successfully synthesised human insulin. In 1982, the first human insulin, Humulin, was marketed. In 1996, the first insulin analogue (Lispro) was approved. In 2006, inhaled insulin Exubera received approval. In recent years, personalised insulin therapy and artificial pancreas systems have been progressively developed. These milestones have not only advanced diabetes treatment but also laid the foundation for research on novel delivery systems, such as oral nanocarriers for insulin.

Despite long-standing challenges in clinical translation due to extremely low bioavailability and formidable gastrointestinal barriers, the field of oral insulin delivery has nevertheless achieved incremental progress. The continued evolution of nanodelivery technologies, such as polymer nanoparticles, liposomes, has provided key strategies for enhancing intestinal absorption and stability of insulin. Importantly, the successful market approval in 2019 of the first oral glucagon-like peptide-1 (GLP‑1) receptor agonist semaglutide tablet (Rybelsus®) demonstrated the feasibility of oral peptide drug delivery, greatly boosting confidence in the pursuit of oral insulin. Currently, several oral insulin candidates employing permeation enhancers or novel delivery systems have advanced into late‑stage clinical investigation, continuously exploring the potential pathway for this century‑old therapy toward an oral era (Brayden 2021).

Subcutaneous injection has remained the primary administration route despite significant advances in delivery systems. Early attempts at oral insulin delivery date back to the 1920s, with numerous approaches being tested over subsequent decades (Madhav 2011). These attempts faced consistent challenges related to insulin's biochemical properties, resulting in negligible bioavailability when administered orally. The evolution from simple enteric coatings to sophisticated molecular engineering approaches reflects the persistent quest to overcome these fundamental barriers. For example, I338, formulated in a tablet with the absorption enhancer sodium caprate developed by Novo Nordisk, was discontinued due to high production costs (Halberg et al. 2019). The development of ORMD-0801 by Oramed Pharmaceuticals represents an oral insulin capsule that combines enzyme inhibitors with permeation enhancers and an enteric coating (Eldor et al. 2013). Despite advancing to phase III clinical trials, the formulation ultimately failed to meet primary endpoints, highlighting persistent challenges in the field.

Oral insulin delivery represents a significant unmet clinical need. Injection-based regimens are associated with poor adherence due to pain, inconvenience, and psychological barriers. Insulin-dependent patients report injection-related anxiety, contributing to suboptimal glycemic control (Zambanini et al. 1999). Oral delivery would more closely mimic physiological insulin secretion patterns, with direct hepatic delivery through portal circulation, potentially reducing systemic hyperinsulinemia and its associated adverse effects (Arbit and Kidron 2017). The convenience of oral administration could improve treatment adherence, particularly in paediatric and geriatric populations, ultimately enhancing glycemic control and reducing long-term complications.

Multiple physiological barriers present formidable challenges to oral insulin delivery. The gastrointestinal tract's enzymatic environment rapidly degrades insulin through the action of pepsin, trypsin, and chymotrypsin, while the acidic gastric environment causes denaturation and aggregation (Gedawy et al. 2018). The negatively charged mucus layer is one of the main barriers faced by oral peptide/protein drugs, but the efficiency of drugs/carriers penetrating the barrier is highly dependent on their charge density, hydration, size, and shielding effect (Netsomboon and Bernkop-Schnurch 2016; Vedadghavami et al. 2020). The intestinal epithelium, with its tight junctions and efflux transporters, severely restricts the absorption of large hydrophilic peptides (Murakami and Takano 2008; Lundquist and Artursson 2016). First-pass hepatic metabolism further reduces the bioavailability of absorbed insulin (Titchenell et al. 2017). When operating together, these barriers result in oral bioavailability typically below 1% for unmodified insulin, necessitating multifaceted approaches to achieve clinically relevant delivery.

The pursuit of oral insulin has evolved from basic protective formulations to sophisticated engineering strategies. Early approaches focused primarily on enteric coatings and protease inhibitors, yielding limited success. Contemporary approaches integrate multiple complementary strategies, including molecular modifications of insulin itself, advanced carrier systems, bio-inspired delivery systems, stimuli-responsive mechanisms, cell-penetrating strategies and microbiome-based approaches (Scheme 1). The shift from single-mechanism approaches to integrated systems biology-based strategies reflects a growing recognition of the complexity of the challenge. Recent advances in understanding biological transport mechanisms, material science, and computational design have accelerated progress, bringing the long-sought goal of effective oral insulin within reach.

Scheme 1.

Schematic diagram of structural modification and advanced formulation design of oral insulin (generated by the authors using Adobe Illustrator 2021).

2. Structural modifications of insulin for enhanced stability and absorption

2.1. Chemical modifications for improved stability in the GI tract

Chemical modifications have significantly enhanced insulin stability under gastrointestinal conditions.

2.1.1. Amino acid conjugation

An insulin analogue with high thermal stability and pharmacological activity has been developed. Four insulin derivatives were successfully prepared by covalently linking the ε-amino group of the 29th lysine of insulin with acetic acid, phenylacetic acid, alanine, and phenylalanine (Sen et al. 2024). Experimental data showed that phenylalanine-conjugated insulin can tolerate high-temperature environments of 65 °C and high-salt stress of 25 mM sodium chloride, while maintaining stability between extreme acidic conditions (pH 1.6) and physiological neutral environments (pH 7.4) (Sen et al. 2024). In a streptozotocin-induced diabetic mouse model, phenylalanine-conjugated insulin demonstrated a blood glucose-lowering effect comparable to native insulin and retained its efficacy even after heat treatment.

2.1.2. PEGylation

PEGylation strategies, which involve introducing polyethylene glycol (PEG) at specific sites, have demonstrated increased stability and extended half-life by shielding susceptible regions from enzymatic attack. For example, insulin at the phenylalanine-B1 or lysine-B29 position can be coupled with low-molecular-weight monomethoxy poly (ethylene glycol) (mPEG) to enhance its anti-aggregation ability and prolong its residence time in systemic circulation (Hinds and Kim 2002). After subcutaneous administration, the relative bioavailability of PEGylated insulin (F750 and F2000) was approximately 1.5-fold and 1.8-fold higher, respectively, than that of unmodified insulin HumulinR® (0.3 IU/kg) in male beagle dogs.

Compared to mono-PEGylated insulin at position lysine-B29, di-PEGylated insulin formed by conjugating with PEG at lysine-B29 and glycine-A1 of insulin exhibited an extended half-life and a 10-fold increase in maximum tolerated dose. Di-PEGylated insulin could demonstrate more sustained blood glucose management efficacy with negligible hypoglycemia, and show potential to ameliorate diabetic nephropathy with long-term administration (Zeng et al. 2024). The oral insulin tablet, IN-105, conjugates PEG to the B29 position of insulin via an acyl chain, exhibiting protease-resistant properties, a prolonged half-life, and improved absorption (Heinemann and Jacques 2009).

2.1.3. Fatty acid conjugation

Lipidation approaches, particularly acylation with fatty acid chains, improve stability while facilitating interactions with intestinal membranes. Insulin at lysine-B29 position was coupled with saturated fatty acids containing 10-16 carbon atoms (Kurtzhals et al. 1995) or α, ω-dicarboxylic acids (octadecanedioic acid) (Ellmerer et al. 2003) to construct fatty acid-acylated insulin with affinity for albumin, which could be used to extend or prolong the action profile of insulin. Albumin-binding acylated insulin analogues such as insulin detemir demonstrate this principle in injectable formulations (Havelund et al. 2004), offering potential advantages for oral delivery.

Another study reported that introducing dual-site mutations of A14E and B25H, and connecting an albumin-binding side chain based on C18 or C20 to a lysine at position B29, can improve insulin protease stability and achieve ultra-long-acting pharmacokinetic properties of insulin (Kjeldsen et al. 2021). In male beagle dogs, the oral absolute bioavailability (3.0%) of C18-based analogues (A14E, B25H), defined as [(AUC/dose)_oral/(AUC/dose)_IV], was significantly higher than that of C20-based analogues (0.43%).

2.1.4. Cyclization and conformational stabilisation

Cyclization techniques that introduce disulphide bridges or other covalent bonds (Hayes et al. 2021) confer proteolytic resistance. For example, an engineered disulphide bridge between positions A10 and B4 has demonstrated enhanced stability while maintaining receptor binding (Vinther et al. 2015). Emerging techniques that combine computational modelling with high-throughput screening accelerate the identification of optimal cyclization strategies (Jin 2020; Hayes et al. 2021).

2.1.5. Glycoengineering

Glycosylation, the attachment of carbohydrate moieties to proteins, represents a biomimetic approach to enhancing stability (Moradi et al. 2016). While native insulin is not glycosylated, introducing glycosylation sites through chemical methods can enhance its stability and reduce its tendency to aggregation. Site-specific glycosylation at positions not critical for receptor binding, such as B1, has demonstrated enhanced stability (Baudys et al. 1995).

Chemical glycosylation can also effectively enhance peptide stability and inhibit molecular aggregation. Through investigation, twelve insulin glycoforms with different glycosylation sites and glycan structures, which comprise mannosylation modification, significantly optimise insulin properties. The C-terminal region of the B-chain was identified as the optimal site for glycan attachment (Guan et al. 2018). O-mannosylation at ThrB27 of the insulin B-chain reduces peptide susceptibility to proteases and self-aggregation, extending half-life to more than twice that of unglycosylated insulin while slightly enhancing bioactivity. Furthermore, insulin modified with a tri-mannose glycan at the B chain Thr27 site demonstrated a significantly longer half-life compared to mono- or di-mannose modifications (Guan et al. 2018). Selection of optimal glycan structures and attachment sites remains an active area of investigation. Additionally, certain glycan structures may enhance interactions with membrane transporters (Varamini et al. 2012; Moradi et al. 2013) or receptors, potentially improving epithelial permeability. The glycosylation of endorphin-1 could enhance its permeability in Caco-2 cell monolayers by 700-fold via a lactose-selective transporter (Varamini et al. 2012), while N-terminal glycosylation of luteinizing hormone-releasing hormone (LHRH) significantly improves both permeability and metabolic stability through the glucose transporter (GLUT2) and sodium–glucose linked transporter (SGLT1) transport mechanisms (Moradi et al. 2013; Moradi et al. 2014).

2.2. Site-specific mutations enhancing epithelial permeability

Strategic modifications to insulin's structure can enhance its ability to traverse intestinal epithelial barriers. Insulin, combined with cell-penetrating peptides known to traverse epithelial barriers, has shown promising results. Insulin can be conjugated with a penetrating peptide, low molecular weight fish sperm protein (LMWP), to form an insulin-LMWP conjugate, which could effectively penetrate intestinal mucosal epithelial cells and significantly improve bioavailability (He et al. 2013). The Insulin-PEG-LMWP conjugate, administered at a high dose of 50 IU/kg via in situ intestinal loop administration, significantly reduced blood glucose levels to 28.61% of the initial value in rats.

Targeting receptors (such as vitamin, folate, and transferrin receptors) and transporters (such as amino acid, bile acid, and nucleoside transporters) on the intestinal epithelial cell membrane can improve physicochemical properties and transmembrane transport capacity (Hamman et al. 2007). Insulin-transferrin conjugates have been developed to protect insulin from enzymatic degradation (Kavimandan et al. 2006). Additionally, the fusion protein formed by proinsulin and transferrin offers advantages such as oral absorption, liver-targeted activation, and prolonged hypoglycaemia effects (Chen et al. 2018b). Oral administration of ExpressTec-ProINS-Tf (800 nmol/kg) maintained blood glucose levels within the normal range (approximately 100 mg/dL) during the 6–12 hour period, producing a glucose-lowering effect comparable to that achieved by subcutaneous injection of 22.5 nmol/kg.

Covalent modification may affect the biological activity of insulin. Insulin at phenylalanine-B1 or lysine-B29 position was combined with bile acid to form two bile acid-insulin complexes (B1-Phe-cholyl-insulin, B29-Lys-cholyl-insulin). Experimental results showed that ileal infusions of B1-Phe-cholyl-insulin caused a long-lasting hypoglycaemia response, while B29-Lys-cholyl-insulin did not have a hypoglycaemia effect (McGinn and Morrison 2016). Recent advances in site-specific conjugation techniques enable the precise attachment of permeation-enhancing moieties (Lieser et al. 2020) while preserving the structural elements essential for biological activity. These approaches represent a significant advance over earlier, less selective modification strategies.

Critical assessment of the emerging paradigm

Structural modifications, while enhancing the efficiency and stability of insulin delivery, may also introduce a range of potential risks. Chemical modifications can alter the structure of insulin, potentially inducing the formation of aggregates with cytotoxic effects (Kamelnia et al. 2024). Moreover, such modifications may lead to uncontrolled changes in the pharmacological activity of insulin, triggering severe adverse reactions such as prolonged hypoglycemia. In particular, PEGylation carries potential immunogenicity risks, as it may create new antigenic epitopes and induce the production of anti-drug antibodies, thereby leading to drug inefficacy and allergic reactions (Fu et al. 2025). It is noteworthy that modifications alter the natural metabolic pathways and organ distribution of insulin, and the long-term metabolic impacts remain uncertain. These uncertainties necessitate rigorous safety evaluations and long-term monitoring to ensure medication safety.

2.3. Critical analysis of structure-function trade-offs in modified insulins

Structural modifications create tradeoffs between stability, permeability, and biological activity (Akbarian et al. 2018). Emerging evidence suggests that certain regions of insulin, particularly the C-terminal of the B-chain, offer greater tolerance for modification without compromising its activity (Hartmann et al. 1989). The development of standardised in vitro screening cascades (such as artificial intelligence prediction and high-throughput screening) has accelerated the evaluation of structure-function relationships (Cheng et al. 2025; Zhang et al. 2025). Artificial intelligence (AI) prediction models, particularly advanced algorithms based on deep learning, graph neural networks, and transformers, have become a pivotal driving force in accelerating drug discovery. These models enable the efficient screening of candidate molecules from ultra-large chemical libraries and accurately predict their biological activity, physicochemical properties, and ADMET (absorption, distribution, metabolism, excretion, and toxicity tolerance) characteristics, thereby significantly facilitating the rational evaluation and optimisation of structure-function relationships (Ferreira and Carneiro 2025). Concurrently, the exceptional performance of AI in predicting compound pharmacokinetic properties directly guides our design of oral delivery systems with enhanced stability and oral bioavailability (Wu et al. 2024a). Successful oral insulin candidates will likely require balanced modifications that sufficiently address stability and permeability challenges while maintaining adequate biological activity, potentially sacrificing some potency in exchange for dramatically improved bioavailability.

3. Advanced formulation technologies: beyond traditional approaches

3.1. Approaches utilising nanocarrier systems and their mechanisms

Nanocarrier technologies (polymer-based, lipid-based, inorganic nanoplatforms, and metal organic frameworks) have emerged as powerful tools for protecting insulin from degradation while enhancing epithelial transport.

3.1.1. Polymer-based nanocarrier

Polymer-based delivery systems stand out due to strong designability, high biocompatibility, and multifunctional potential. The polymer matrix used to construct insulin delivery nanocarriers primarily consists of two categories: natural sources and synthetic materials. Polymer materials can effectively encapsulate insulin, resist degradation by gastrointestinal enzymes, provide pH-responsive protection, enhance mucus permeability, promote cell bypass or cross-cell transport, and significantly improve the stability and absorption efficiency of drugs in the gastrointestinal tract (Pridgen et al. 2014).

Natural polymers, composed of polysaccharides (including chitosan, starch, alginate, pectin, and dextran) (Meneguin et al. 2021) and proteins (zein (Inchaurraga et al. 2020)), offer protective environments for insulin, with release profiles controlled through composition and manufacturing parameters.

The pectin-based drug delivery system exhibits good adhesion and can effectively resist degradation mediated by proteases and amylases, making it suitable for oral insulin delivery (Zhang et al. 2022). Polygalacturonic acid is a biodegradable polysaccharide prepared by demethylation of plant pectin, which can specifically recognise and bind to liver asialoglycoprotein receptors through galactose residues (Li et al. 2016), achieving active targeted delivery to liver parenchymal cells. The polygalacturonic acid-based biodegradable targeted polymer carrier has the potential to enhance oral insulin delivery (Zhang et al. 2020c).

Lysine-based poly(ester amide)s (Lys-aaPEAs) decorated with hyaluronic acid (HA) were developed for the oral delivery of insulin (Han et al. 2023). Within the easily degradable gastrointestinal tract, Lys-aaPEAs can maintain structural stability and effectively overcome the intestinal luminal-to-basolateral barrier, enabling efficient transport across the intestinal epithelium into systemic circulation to release insulin (Figure 1a) (Han et al. 2023). A HA-coated chitosan nanoparticle, HCP-INS, was developed for oral insulin delivery (Wu et al. 2022b). Subcutaneous insulin injection (2 IU/kg) demonstrated rapid yet transient hypoglycaemia effects, with a significant decrease in blood glucose concentration within 2 hours after administration and a return to baseline levels at 6 hours. In contrast, the oral formulation exhibited prolonged glucose-lowering characteristics: the 5 IU/kg dose group showed a kinetic pattern of rapid glucose reduction followed by gradual recovery over 8 h, while the high-dose 20 IU/kg group maintained significant hypoglycaemia effects for over 12 h, demonstrating stronger durability of drug efficacy (Wu et al. 2022b).

Figure 1.

Innovative nanocarrier systems for protecting insulin from degradation. (a) Schematic representation of polymer-based nano-platforms (Lys-aaPEA). Reproduced with permission from Han et al. (2023). Copyright 2023, Wiley. (b) Illustration of liposome-loaded cysteine-alginate hydrogel (Lip-Gel) and its morphology at different pH values. Reproduced with permission from Wu et al. (2023). Copyright 2022, Elsevier. (c) Schematic representation of the MOF framework for encapsulation of insulin and its release in stomach acid. Reproduced with permission from Chen et al. (2018a). Copyright 2018, American Chemical Society.

Artificially synthesised polymers mainly include polycaprolactone, poly (lactic co-glycolic acid, PLGA), and poly lactic acid (PLA). Various coated or modified PLGA nanoplatforms can be used to deliver insulin for diabetes management orally (Pang et al. 2023). Chitosan-coated porous PLGA particles were prepared for oral insulin delivery. Compared to uncoated particles, chitosan coating increases the drug loading capacity and promotes sustained drug release (Eilleia et al. 2018). Oral insulin suspension showed no reduction in blood sugar, while subcutaneous injection of insulin caused a rapid decrease in blood sugar levels in the first hour, but only continued to rise again after 2 hours. Compared with uncoated formulations, coated oral formulations exhibited a prolonged blood glucose-lowering effect, with a duration of action exceeding 8 hours (Eilleia et al. 2018).

3.1.2. Lipid-based nanocarrier

Lipid-based nanocarriers (such as solid lipid nanoparticles, nanostructured lipid carriers, liposomes, etc.) have become a research hotspot in drug delivery systems due to low toxicity, good biocompatibility, ability to encapsulate hydrophilic and hydrophobic drugs, and ability to regulate drug release (Plaza-Oliver et al. 2021). Currently, in addition to the US Food and Drug Administration (FDA) approval of Doxil and Onpattro, various other lipid-based nanoparticles are also undergoing clinical trials and being developed into commercial products (Thi et al. 2021). Lipid-based nanocarriers for oral peptide delivery exhibit dual advantages: resistance to protein hydrolysis and enhanced intestinal permeability. (Niu et al. 2016).

Lipid-based nanocarriers with surface coatings have been developed to enhance stability in the gastrointestinal tract and promote adhesion to intestinal epithelial cells, thereby enhancing the hypoglycaemia effect of insulin. Chitosan-coated solid lipid nanoparticles (Fonte et al. 2011), chitosan-coated anionic nanoliposomes (Zhang et al. 2021b), hydrophobic ion pairing and a self-microemulsifying drug delivery system (SMEDDS) (Goo et al. 2022) have been developed for efficient oral insulin delivery. Protein corona liposomes, formed by covering cationic liposomes with bovine serum albumin, were used to enhance the oral absorption of insulin (Wang et al., 2019a). In addition to chitosan, protein was also used to coat liposomes for oral administration. Protein corona liposomes can enhance the permeability of the intestinal mucosa and improve epithelial transport, significantly increasing the oral relative bioavailability of insulin by up to 11.9% via intrajejunal administration (at the insulin dose of 75 IU/kg) compare to subcutaneous injections of insulin solution (5 IU/kg) (Wang et al., 2019a).

To improve the adhesive ability of intestinal mucosa, cysteine-modified alginate hydrogel was used to encapsulate arginine-insulin complexes (AINS)-loaded liposomes (AINS Lip) to form liposome-in-alginate hydrogels (AINS-Lip-Gel) (Wu et al. 2023). In vitro, the intestinal permeability of arginine-insulin complexes reached 2.0 times that of free insulin, while AINS Lip exhibited a permeability enhancement effect of 6.0 times (Figure 1b) (Wu et al. 2023).

A novel formulation (ORLN-PHI) termed ‘oil-soluble’ reversed lipid nanoparticles was designed by encapsulating the hydrophilic peptide recombinant human insulin (PHI) in a phospholipid (PC) shell and then dissolving it in oil (Wang et al. 2020). Compared with free PHI, ORLN-PHI enhanced the oral hypoglycaemia effect in mice and the absorption of oral insulin in rats (Wang et al. 2020).

Although surfactant-based micellar systems can load hydrophilic molecules, their reliance on the critical micelle concentration (CMC) causes disintegration upon significant dilution in the gastrointestinal tract. This leads to drug leakage and burst release (Mirchandani and Patravale 2021), consequently failing to provide effective protection for insulin. In contrast, Solid lipid nanoparticles do not rely on CMC, as the physical stability of their solid matrix effectively restricts drug mobility, enabling a more stable and controlled release profile (Mirchandani and Patravale 2021; Upadhyay and Soni 2023). Furthermore, nanostructured lipid carriers, by blending solid lipids with liquid lipids, create a matrix with more structural imperfections. This structure not only enhances drug loading capacity but also minimises drug leakage during storage. As a result, nanostructured lipid carriers exhibit superior stability and higher entrapment efficiency compared to liposomes, while offering greater drug loading capacity and reduced drug leakage compared to solid lipid nanoparticles (Mall et al. 2025).

3.1.3. Inorganic nanoparticle platforms

In biomedical applications, commonly used inorganic nanoparticle materials include iron oxide nanoparticles, metal nanoparticles (such as gold and silver), solid or mesoporous silica, and quantum dots (Giner-Casares et al. 2016). Inorganic nanoparticles have great therapeutic potential for oral drug delivery. The physical and chemical properties of inorganic nanoparticles, including composition, size, shape, porosity, and surface chemistry, can significantly impact the solubility, adhesion, transport efficiency, and biocompatibility of particles, thereby influencing the oral drug delivery efficiency (Asad et al. 2022).

Small, negatively charged inorganic silica nanoparticles can open tight junctions and increase intestinal permeability by binding to integrins and activating myosin light chain kinase (MLCK), thereby promoting the oral delivery of insulin (Lamson et al. 2020). The effect of particles of different sizes (20–1200 nm) on intestinal barrier function was evaluated in vitro. All silica nanoparticles increased permeability, with the most significant effects observed in particles of 20 nm and 50 nm (Lamson et al. 2020).

Mesoporous silica nanoparticles (MSNs) with adjustable pore size can encapsulate different-sized biomolecules, allowing for further optimisation of drug loading and release characteristics through modification (Castillo et al. 2020). Dendritic mesoporous silica nanoparticles (DMSNs) have controllable mesoporous channels (diameter 6–20 nm) capable of carrying insulin, while 50–100 nm can effectively penetrate the intestinal epithelium. A thiol-functionality-anchored mesoporous DMSN could load insulin through electrostatic attraction and further bind to succinylated β-lactoglobulin, which could reduce release/degradation in the stomach (pH 1.2) and enhance insulin transport to the intestine (pH 7.4) (Juere et al. 2020).

Gold nanoparticles (AuNPs) have shown significant potential in the biomedical field due to their unique properties (Georgeous et al. 2024). The easily modifiable surface properties of AuNPs enable targeted drug delivery and enhanced drug loading efficiency. Additionally, the combination of optical properties (e.g. surface plasmon resonance and fluorescence quenching) with imaging techniques such as computed tomography can significantly enhance the sensitivity of disease diagnosis (Georgeous et al. 2024). The potential of AuNPs as carriers of insulin has been demonstrated. Chondroitin sulphate-capped AuNPs for oral delivery of insulin could significantly reduce blood glucose levels (up to 32.1%) and achieved a 6.61-fold higher plasma insulin concentration at 120 min compared to oral insulin solution (50 IU/kg) (Cho et al. 2014).

3.1.4. Metal organic framework

The recent awarding of the 2025 Nobel Prise in Chemistry to Prof. Susumu Kitagawa, Prof. Richard Robson, and Prof. Omar M. Yaghi for the development of metal-organic frameworks (MOFs) has cemented the profound impact of this class of materials (Wang 2025c). MOFs are a type of porous crystalline material formed by self-assembly of metal ions/clusters and organic ligands through coordination bonds (He et al. 2021), constructing three-dimensional architectures capable of hosting various molecules. In recent years, MOFs have garnered significant attention in the field of drug delivery due to their high specific surface area, tunable pore structures, exceptional drug-loading capacity, and potential for surface functionalization (He et al. 2021). Crucially, their porous structure enables the efficient encapsulation of insulin, not only to enhance its solubility but also to protect it from degradation by gastric acid and digestive enzymes. Furthermore, by rationally designing ligands or incorporating functional groups (e.g. pH-responsive units, mucus-penetrating agents), MOFs can promote intestinal absorption, thereby directly addressing the core challenges associated with oral insulin administration (Raza and Wu 2024).

The mesoporous MOF, NU-1000, (mesopores with size ~30 Å and micropores with size ~12 Å in diameter), could efficiently load insulin (13 × 13 Å), achieving a drug loading capacity of 40 wt% within 30 min, while effectively excluding pepsin (48 × 64 Å), shielding insulin from degradation of gastric acid and digestive enzymes, as well as gastric protease (Figure 1c) (Chen et al. 2018a). Experimental results showed that insulin@NU-1000 exhibited exceptional stability in simulated gastric fluid (pH 1.29), with only 10% drug release observed over 60 minutes. Conversely, under physiological conditions (PBS, pH 7.4), insulin@NU-1000 could trigger 91% of insulin release (Chen et al. 2018a).

Acid-resistant Zr-based MOF (PCN-777) nanoparticles (I@P) demonstrated efficient insulin loading (up to 121.6 wt% at a 3:1 mass ratio) and retained over 60% of insulin activity under trypsin digestion conditions for 3 hours. In a diabetic rat model, the oral relative bioavailability of I@P nanoparticles (50 IU/kg) was only 0.34% compared to subcutaneous injection of insulin (5 IU/kg, n = 5) (Zou et al. 2025).

Another study reported that efficient loading of insulin (65 wt%) through imine-linked covalent organic framework (nCOF) nanoparticles could protect insulin under harsh conditions while preserving its activity upon release (Benyettou et al. 2021). Additionally, in a type 1 diabetes (T1D) rat model, the system demonstrated stable blood glucose regulation within 2–4 hours after administration, with a prolonged hypoglycaemia effect lasting up to 10 hours in vivo (Benyettou et al. 2021).

Critical assessment of the emerging paradigm

Nanoformulations show great potential in the field of drug delivery, but their clinical application is severely limited by a series of safety concerns. The potential toxicity of nanoparticle formulations mainly stems from their physicochemical properties (size, shape, surface charge, hydrophobicity, and targeting ligands) and their nonspecific distribution and metabolic behaviour in vivo. Furthermore, the unclear processes of absorption, distribution, metabolism, and excretion of nanocarriers in vivo, as well as the drug release mechanisms, further restrict their clinical translation (Zhang et al. 2020a). Table 1 provides a comparative summary of key characteristics of major oral insulin nanocarrier systems.

Table 1.

Comparison of various nanotechnology-based formulations.

Property	Polymeric systems	Lipid systems	Inorganic platforms	Bio-inspired systems
Drug loading (w/w)	Moderate	Moderate to High	Moderate to High	Moderate
Protection in SGF/SIF	Good	Moderate	Excellent	Good
Mucus penetration	Moderate	Moderate	Moderate to High	Excellent
Epithelial transport mode	Endocytosis	Endocytosis	Tight junction modulation/Endocytosis	Active transcytosis
Variability	Moderate	Moderate	Low to High	High
Scale-up potential	Moderate to High	High	Moderate	Extremely difficult
Excipient status	Solid powder	Liquid/Solid	Solid powder	Complex

The extremely low relative bioavailability of oral formulations remains the primary bottleneck. Carriers designed to enhance oral absorption by exploiting the paracellular pathway may risk compromising the integrity of the tight junctions in the mucosal barrier, and the kinetics of their recovery remain to be fully elucidated (Saker et al. 2024). Furthermore, studies in animal models have indicated that intravenous administration of inorganic nanoparticles (such as titanium dioxide, silica, and gold nanoparticles) can induce endothelial leakage, which significantly promotes breast cancer metastasis (Peng et al. 2019).

The opsonization and subsequent phagocytic clearance mediated by the surface protein corona, along with the accelerated blood clearance phenomenon associated with PEGylated carriers (Zhang et al. 2020a; Saker et al. 2024), reveal the complex interactions between nanomaterials and the immune system. The most critical challenge stems from the resistance to degradation and the long-term accumulation of many inorganic carriers within the reticuloendothelial system (RES). Retention periods can extend for months or even years, and the chronic toxicological consequences of this prolonged in vivo retention represent a significant gap in current research (Saker et al. 2024). The exogenous metabolites produced by the degradation of nanocarriers in vivo or the metal ions released from inorganic materials and MOFs may have unknown pharmacological activity or toxicity (Zhang et al. 2020a; Saker et al. 2024).

Compounding these biological challenges are the formidable manufacturing obstacles. Organic carriers struggle to control the uniformity of particle size and drug loading during scale-up production, while inorganic carriers commonly face synthesis control issues, such as precise morphology regulation and the removal of toxic templates. Many MOFs that demonstrate exceptional performance in the laboratory rely on synthesis routes that often depend on expensive metal salts (such as zirconium and hafnium), organic ligands, high-boiling-point toxic organic solvents, as well as energy-intensive reaction conditions (e.g. high-temperature and high-pressure solvothermal methods) (Raza and Wu 2024), which hinder scalable and cost-effective production.

Therefore, future research urgently needs to systematically evaluate these risks in accordance with regulatory guidelines, establish critical safety thresholds and therapeutic indices, and thereby facilitate the safe and effective translation of nanomedicine into clinical practice. Additionally, the translation of nanoformulations from the laboratory to the market faces formidable challenges, including the scaling-up of manufacturing processes and ensuring batch-to-batch consistency, coupled with the dilemma of lacking clear regulatory guidelines. Furthermore, to aid the clinical translation of these technologies, it is necessary to advance the standardisation of preclinical parameters and procedures, integrate translational medicine considerations into the technological design, and establish knowledge-sharing mechanisms. Conducting preclinical in vitro and in vivo studies under unified conditions enables precise comparison among different technological approaches (Chen et al. 2022).

3.2. Bio-inspired delivery systems mimicking natural transport processes

Bio-inspired approaches that emulate natural biological transport mechanisms show particular promise for oral insulin delivery. Exosome-mimetic nanocarriers derived from cellular membrane components exhibit enhanced interactions with intestinal epithelial cells and reduced immunogenicity (Liu et al. 2024a). The insulin-loaded milk-derived exosomes demonstrated a superior hypoglycaemia effect compared to subcutaneously injected insulin (1 IU/kg) in a type I diabetic rat model, with a relative bioavailability of 4.03% at an oral dose of 30 IU/kg. This enhanced efficacy is attributed to the versatile capabilities of milk-derived exosomes, including active multi-targeting uptake, pH-adaptive membrane restructuring, and efficient mucus penetration, which collectively overcome multiple barriers in oral drug delivery (Wu et al. 2022a).

Virus-like particles offer highly defined structures that can protect insulin while leveraging viral entry mechanisms to enhance cellular uptake. For example, virus-mimicking nanoparticles consist of insulin-loaded poly(n-butylcyanoacrylate) core, coated with positively charged folic acid (FA) grafted chitosan copolymers and negatively charged HA to enhance mucus penetration (Figure 2a) (Cheng et al. 2021). PLGA nanoparticles with modification groups could extend from the surface of the intestine by mimicking spike proteins on the viral surface, enabling efficient intestinal barrier penetration (Figure 2b) (Yang et al. 2022). In another study, virus-mimicking mesoporous silica nanoparticles with an electrically neutral and hydrophilic surface were successfully engineered to effectively penetrate the mucus layer and pass through the intestinal epithelium (Figure 2c) (Zhang et al. 2021a).

Figure 2.

Bio-inspired delivery systems. (a) Folic acid decorated virus-mimicking nanoparticles for enhanced oral insulin delivery. Reproduced with permission from Cheng et al. (2021). Copyright 2021, Elsevier. (b) Schematic illustrations of the composition and mechanism of Pep/Gal-PNPs. Reproduced with permission from Yang et al. (2022). Copyright 2022, Springer Nature. (c) Virus-mimicking nanoparticles (MSN-NH₂@COOH/CPP5) for penetrating the mucus layer and the underlying epithelium. Reproduced with permission from Zhang et al. (2021a). Copyright 2021, American Chemical Society.

Biomimetic systems incorporating components of bacterial outer membrane vesicles (Sartorio et al. 2021; Gong et al. 2025) or milk fat globule membranes (MFGM) (Nie et al. 2024) exploit evolutionary-optimised transport pathways. Phospholipids in MFGM of camel milk exhibit anti-diabetic properties, and liposomes derived from MFGM could encapsulate insulin for targeted diabetes therapy (Shafiq et al. 2024). These bio-inspired approaches benefit from billions of years of evolutionary optimisation, potentially offering superior biocompatibility and efficacy compared to purely synthetic systems. The integration of specific receptors or transport proteins into these biomimetic structures further enhances the ability to navigate biological barriers through active transport mechanisms (Hamman et al. 2007).

Critical assessment of the emerging paradigm

The endotoxin present in natural outer membrane vesicles is the primary source of toxicity. However, a variety of effective detoxification strategies, such as detergent extraction and genetic engineering of lipid A, have been developed, providing a clear pathway to reduce the toxicity of biomimetic carriers (Sartorio et al. 2021). It is important to note that outer membrane vesicles derived from pathogens may carry other virulence factors, posing a potential risk of pathogenicity, which necessitates the use of non-pathogenic production strains. Outer membrane vesicles are potent immune activators, as the pathogen-associated molecular patterns (PAMPs) they carry can effectively stimulate both mucosal and systemic immunity (Sartorio et al. 2021; Gonçalves et al. 2025). While this property is advantageous for oral vaccine applications, it requires precise engineering when used for insulin delivery to avoid excessive immune responses. Furthermore, to prevent ecological imbalance, the impact of outer membrane vesicles on the gut microbiota must be considered, as their effect is beneficial when derived from commensals like Bacteroides but potentially disruptive when derived from exogenous species (Sartorio et al. 2021).

Regarding production, the heterogeneity and purity of outer membrane vesicles can vary between batches, a challenge underscored by the need for sophisticated isolation and characterisation techniques to ensure reproducibility (Gonçalves et al. 2025). Although long-term colonisation is less likely with non-replicating outer membrane vesicles, their profound immunomodulatory capacity could have indirect, sustained effects on the host environment. Various methods and yields for the separation and purification of outer membrane vesicles have been listed, providing technical references for scalable production and consistency control (Gonçalves et al. 2025). Meanwhile, efficient and safer production can be achieved using non-pathogenic engineered bacteria, such as Escherichia coli, thereby addressing safety and mass-production challenges associated with deriving outer membrane vesicles from pathogenic strains (Sartorio et al. 2021). Outer membrane vesicle-based vaccines, such as the MenB vaccine Bexsero, have already received approval from regulatory agencies like the FDA (Sartorio et al. 2021; Gonçalves et al. 2025). This establishes a critical regulatory precedent and builds confidence in the clinical translation of outer membrane vesicles as a delivery system, provided that the risks above are adequately managed and controlled.

3.3. Stimuli-responsive formulations for site-specific release

Nanocarriers can be designed to be responsive to either a single factor (i.e. pH, enzyme, glucose) or multiple elements to achieve controlled release of drugs. Stimuli-responsive systems enable precise control over insulin release in response to environmental conditions or external triggers (Zheng et al. 2020).

3.3.1. pH-responsive

Enteric coating technologies protect against the acidic gastric environment while enabling site-specific release in the intestine. The rational design of enteric systems requires consideration of regional gastrointestinal physiology, including pH gradients (stomach: pH 1.5–3.0; duodenum: pH 5.5–6.5; jejunum/ileum: pH 6.5–7.5; colon: pH 6.5–7.8) and transit times (stomach: 1–2 hours fasted, 2–4 hours fed; small intestine: ~3–5 hours; colon: ~20–35 hours) (Stillhart et al. 2020). Region-selective insulin release can be achieved using pH-responsive polymers that remain intact in the stomach but dissolve at specific intestinal pH values. These include hydroxypropyl methylcellulose phthalate (HPMCP; dissolving at pH > 5.5, targeting the duodenum/jejunum), polyacrylic acid derivatives such as Eudragit® L100 (pH > 6.0) and S100 (pH > 7.0), which target the jejunum/ileum and distal ileum/colon, respectively, and cellulose acetate phthalate (dissolving at pH > 6.0) (Wang and Zhang 2012).

The pH-dependent release profile of the formulation directs the drug to specific intestinal regions, thereby influencing the dominant absorption pathway. Formulations released in the duodenum and jejunum (pH 5.5–7.0) primarily exploit transcellular pathways via enterocyte uptake and receptor-mediated transcytosis, taking advantage of the large surface area and high metabolic activity of these regions. Conversely, colon-targeted systems leverage the extended transit time (~20–35 hours) and reduced proteolytic activity to enhance paracellular transport through the looser tight junctions characteristic of colonic epithelium (Stillhart et al. 2020; Mathur et al. 2025).

Eudragit®S100-coated chitosan nanoparticles (Chen et al. 2017) and colon-targeted nanocomposite system of organic clay/ethanol chitosan/Eudragit®S100 (Lee et al. 2020) were developed for enhanced oral delivery of insulin. Eudragit®S100, as an enteric coating material, is formed by copolymerisation of methacrylic acid and methyl methacrylate in a 1:2 molar ratio. Eudragit®S100 is a pH-sensitive enteric polymer designed for colonic delivery. It remains stable in acidic environments (pH < 7) due to the protonation of its carboxylic groups, protecting insulin from gastric release. Upon reaching the distal ileum and colon (pH > 7), the polymer dissolves due to carboxylate group ionisation. In the colon (pH 6.5–7.8), this system exploits the reduced proteolytic activity and extended residence time to facilitate insulin absorption through paracellular pathways, as the chitosan component further enhances tight junction opening (Lee et al. 2020).

A pH-triggered self-unpacking capsule encapsulating zwitterionic hydrogel-coated MOF nanoparticles has been developed for efficient oral drug delivery (Zhou et al. 2021). The capsule utilised a pH-sensitive poly(methacrylic acid-co-ethyl acrylate) coating, Eudragit L100-55, as the outer layer, which could protect the drug from acidic degradation in the gastric environment and dissolve under neutral intestinal conditions, enabling targeted release (Zhou et al. 2021). Besides, the capsule was loaded with a gas-generating mixture of sodium bicarbonate and citric acid. After the coating dissolved, the mixture rapidly produced gas bubbles, enhancing the transport of nanoparticles through the mucus layer in the intestinal tract (Zhou et al. 2021).

Another study reported that sodium alginate-coated metal organic framework (PCN-222) has been developed to achieve precise intestinal delivery and reduce gastric irritation (Figure 3a) (Zhu et al. 2025b). Sodium alginate decoration can protect contents by inducing structural contraction through the protonation of carboxyl groups in gastric acid, and release the payload through deprotonation dissolution in the alkaline intestinal environment (Zhu et al. 2025b).

Figure 3.

Schematic illustration of pH-sensitive and enteric coatings for protecting insulin from gastric acid degradation. (a) Design and delivery mechanism of the sodium alginate-coated metal organic framework. Reproduced with permission from Zhu et al. (2025b). Copyright 2024, Elsevier. (b) The synthesis of the layer-by-layer calcium phosphate nanoparticles for improved oral delivery of insulin. Reproduced with permission from Verma et al. (2016). Copyright 2015, Elsevier.

Advanced enteric systems utilise multiple polymer layers with distinct pH-triggered release thresholds, allowing for targeted release at specific intestinal regions optimised for insulin absorption. Layer-by-layer coating approaches, which deposit alternating layers of oppositely charged polyelectrolytes (such as chitosan, alginate, polyallylamine hydrochloride, and polyacrylic acid), can create films that undergo swelling or disintegration at specific pH values, thereby providing finely controlled, pH-responsive release profiles (Shukla et al. 2010). Vitamin B12 (VB12)-grafted chitosan and sodium alginate served as cationic and anionic polyelectrolytes, respectively, for layer-by-layer coating of insulin-loaded calcium phosphate nanoparticles (VitB12-Chi-CPNPs), which exhibited pH-responsive release and utilised multiple pathways to enhance overall bioavailability of insulin (Figure 3b) (Verma et al. 2016).

The integration of pH-responsive elements within nanoparticle systems, rather than as external coatings, represents an emerging approach in nanotechnology. pH-responsive polymers can form a core or shell of nanoparticles, triggering structural changes upon pH transition that alter release kinetics or expose functional groups, enhancing epithelial interaction.

PLGA-Hyd-PEG nanoparticles with a pH-sensitive core-shell structure, consisting of PEG and PLGA connected via a hydrazone bond, were developed for oral insulin delivery (Figure 4a) (Li et al. 2022). The system, as a freeze-dried powder, was encapsulated in enteric capsules, which can release drugs in the jejunum. Among them, the PEG shell exhibits excellent stability (pH 6.5–7.4) in jejunal fluid, which can effectively protect drugs and promote rapid penetration through the mucus layer (Li et al. 2022). However, in the acidic microenvironment of the jejunum epithelium (pH ~5.5), the PEG layer was shed by the hydrolysis of hydrazine bonds, and the surface hydrophobicity of the nanoparticles increased with a decrease in PEG shell density, significantly enhancing cell uptake efficiency (Li et al. 2022).

Figure 4.

pH stimuli-responsive formulations. (a) Schematic illustration of the delivery mechanism of the PLGA-Hyd-PEG nanoparticles with a pH-sensitive core-shell structure. Reproduced with permission from Li et al. (2022). Copyright 2021, Elsevier. (b) Schematic illustration of pH-triggered neutral nanoparticles-extruded microcapsules for oral delivery of Insulin. Reproduced with permission from Lu et al. (2023). Copyright 2023, American Chemical Society.

pH-responsive carboxylated cellulose microspheres loaded with insulin via electrostatic interactions were developed (Gong et al. 2021). The insulin-controlled release mechanism was regulated by the ionisation of carboxyl groups and the balance of protons. In an acidic environment, protonated carboxyl groups maintain a compact, collapsed state of microspheres through hydrogen bonding, inhibiting insulin release. However, in an alkaline environment, carboxyl deprotonation (COO^─) can generate electrostatic repulsion, expanding the pore size of the carrier and promoting the infiltration of water molecules (Gong et al. 2021). Although electrostatic repulsion between insulin (negatively charged) and the carrier could accelerate outward diffusion, a small amount of positively charged residues within the molecule could act synergistically with the swelling hydrogel barrier, resulting in partial insulin retention within the carrier (Gong et al. 2021).

Natural pollen shells were used to wrap net-neutral nanoparticles (γ-glutamic acid/chitosan, mass ratio: 1:2) and insulin to form pH-triggered nanoparticles-extruding microcapsules (Figure 4b) (Lu et al. 2023). The sturdy outer structure of pollen could protect the drug from degradation by stomach acid and enzymes. The electrostatic interaction between γ-glutamic acid and chitosan was also stable at low pH (Lu et al. 2023). In the small intestine, the amino group of chitosan could gradually deprotonate, triggering swelling and allowing nanoparticles to squeeze out pollen shells. Additionally, net-neutral nanoparticles could enable insulin to penetrate intestinal mucus and be effectively internalised by epithelial cells (Lu et al. 2023).

Another study utilised a hydrogel formed by crosslinking chitosan with glycidoxypropyltrimethoxysilane (GPTMS) to coat porous silica for pH-responsive insulin release (Wu and Sailor 2009). The gel in the top layer can block insulin release under neutral conditions (pH 7.4), whereas the hydrogel expands due to solvation and electrostatic repulsion effects in a weakly acidic environment (pH 6.0), promoting the stable release of drugs from the bottom porous silica reservoir (Wu and Sailor 2009).

3.3.2. Enzyme-responsive

Enzyme-responsive materials could also be developed to release insulin. Insulin-conjugated silver sulphide quantum dot coated with chitosan/glucose polymer to produce an enzyme-responsive oral insulin nanoformulation (Hunt et al. 2024). Chitosan/glucose polymer could trigger insulin release through glucosidase-mediated degradation in vivo. When co-incubated with β-glucosidase or cellulase, 50% insulin was released within 0.5 h (single-compartment system) (Hunt et al. 2024).

3.3.3. Glucose-responsive

Insulin embedded in a matrix containing glucose-responsive elements can regulate the release rate of insulin in response to environmental glucose levels through structural changes, such as swelling, contraction, degradation, or dissociation. Common glucose reaction elements include phenylboronic acid (PBA) (Yu et al. 2018; Li et al. 2018; Wang et al. 2019b) and derivatives (Xiao et al. 2021), polyboroxole (Kim et al. 2012), glucose binding proteins (Xu et al. 2022a; Zhou et al. 2024), and glucose oxidase (GOx) (Xia et al. 2018). A switchable polymeric complex was formed through electrostatic interactions between a positively charged polymer and negatively charged insulin (Wang et al. 2019b). Under hyperglycemic conditions, glucose binds to PBA in the polymer, reducing the positive charge density and weakening the polymer’s electrostatic attraction, which facilitates the rapid release of insulin (Wang et al. 2019b).

Phenylboronic acid (PBA)-modified hyaluronic acid (HA) can encapsulate a neonatal Fc receptor-targeted liposome core loaded with insulin through boronate ester bonds formed between PBA and catechol groups on the surface of liposomes (Yu et al. 2018). Under strong acidic conditions (pH 1.2–3.0), HA-PBA cannot complex with catechol groups on the surface of Fc-liposome, while protonation on the surface of Fc-liposome could lead to charge reversal (negative→positive) to maintain the integrity of the HA shell structure via electrostatic interactions, preventing insulin leakage and enzymatic hydrolysis in the stomach (Yu et al. 2018). However, boronate ester bonds in nanosystems could break due to competitive binding between glucose molecules and PBA, triggering the detachment of the HA shell and facilitating insulin release (Yu et al. 2018).

PBA derivatives, 4-aminophenylboronic acid pinacol ester (PBAPE) (Xiao et al. 2021), 4-carboy-3-fluorophenylboronic acid (FPBA) (Wang et al. 2021; Wang et al. 2023a; Ji et al. 2024), and (4-(2-acrylamidoethyl) carbamoyl)-3-fluorophenyl) boronic acid (Wang et al. 2024a), were used to construct glucose-responsive insulin delivery systems. PBAPE was grafted onto poly-_L-glutamate polymer to synthesise glucose-responsive copolymer P(GA-co-GAPBAPE) via amidation, which could self-assemble into nanoparticles with 1,2-dithioaryl-sn-glycero-3-phosphoethanolamine-N-[maleimide (polyethylene glycol)] (DSPE-PEG-Mal) for intelligent controlled release of insulin (Xiao et al. 2021). The nanoplatform demonstrated rapid insulin release in glucose solution (‘on’ state), while exhibiting slowed release in PBS (‘off’ state), successfully achieving three cycles of self-regulating on-off switching (Xiao et al. 2021). Nanomedicine was composed of amphiphilic diblock copolymers PPF (containing FPBA-modified PEA and polycarboxylicbetaine) and insulin, which self-assembled into worm-like micelles, establishing a glucose-responsive reservoir in the liver (Ji et al. 2024). Elevated blood glucose levels could enhance the binding of FPBA groups to glucose, thereby reducing the attraction between insulin and polymers and promoting insulin release (Ji et al. 2024).

Monosaccharide-responsive block copolymers consisted of Poly(styreneboroxole) (PBOx) and PEG that self-assembled to form polymersomes (PEG-b-PBOx) in water (Kim et al. 2012). PEG-b-PBOx exhibited sugar concentration-dependent dissociation behaviour at neutral pH, releasing encapsulated insulin in physiological environments (Kim et al. 2012). A glucose-responsive oral insulin nanosystem (VB12-FU-ConA) was developed through the crosslinking of VB12-Fucoidan (VB12-FU) with concanavalin A (ConA). ConA could specifically bind to free glucose, and glucose molecules compete with the original binding sites of ConA, causing material expansion and triggering insulin release, helping to maintain blood glucose stability (Figure 5a) (Zhou et al. 2024). Matrix systems incorporating GOx generate localised acidification in hyperglycemic conditions, promoting insulin release through degradation or protonation. In another study, biotinylated GOx (biotin-GOx)-modified insulin-loaded red blood cells were found to catalyse the conversion of glucose to gluconic acid and hydrogen peroxide (H₂O₂) through the action of GOx, leading to the rupture of the red blood cell membrane and the subsequent release of insulin (Figure 5b) (Xia et al. 2018).

Figure 5.

Glucose stimulus-responsive formulations. (a) The synthesis and mechanism of a responsive nanosystem containing Con A. Reproduced with permission from Zhou et al. (2024). Copyright 2024, Elsevier. (b) GOx-modified erythrocytes as a glucose-activating switch for insulin release. Reproduced with permission from Xia et al. (2018). Copyright 2018, Elsevier.

3.3.4. Dual-responsive

Dual-response controllable release systems, such as glucose/pH, glucose/hypoxia, and glucose/H₂O₂, have been developed to ensure precise drug release. GOx was entrapped or immobilised within a pH-sensitive matrix to achieve glucose- and pH-responsive insulin release. Glucose/pH dual-responsive systems were developed by co-loading insulin with GOx/catalase (CAT) onto chitosan oligosaccharide-vanillin nanoparticles (GRNs) and coupling them with erythrocytes (Xu et al. 2022b). GOx catalyses the conversion of glucose into gluconic acid and H₂O₂, leading to a decrease in the microenvironmental pH. Meanwhile, CAT decomposes excess H₂O₂ to reduce oxidative damage and prevent GOx deactivation. Subsequently, the cleavage of imine bonds and protonation of amino groups in GRNs lead to polymer rupture, triggering the release of insulin (Xu et al. 2022b).

Glucose/pH dual-responsive microgels containing chitosan, GOx/CAT nanocapsules, and insulin were developed to catalyse the conversion of glucose to gluconic acid, resulting in a decrease in the surrounding pH (Gu et al. 2013). Subsequently, the protonation of the amino-rich polymer chain increased the charge in the gel matrix, leading to the expansion and dissociation of the microgel and the release of insulin (Gu et al. 2013).

pH-responsive polymer, synthesised by reacting dextran with methoxypropene, was used to encapsulate insulin, GOx, and CAT (Volpatti et al. 2020). GOx could convert glucose into gluconic acid, thereby reducing the pH value of the microenvironment. Acetal groups of modified dextran were subsequently cleaved, dissolving the nanoparticles and triggering the release of insulin (Volpatti et al. 2020). In another study, a glucose/pH dual-responsive polymer–covalent COF composite material was developed (Zhang et al. 2020b). Thiol-terminated PEG was anchored onto the COF surface, while insulin and GOx bound to the boron group in the backbone of COF. Under elevated blood glucose conditions, low pH can be triggered by GOx to disrupt boronate ester bonds in the COF structure, leading to the disintegration of the nanostructure and subsequent insulin release (Figure 6a) (Zhang et al. 2020b).

Figure 6.

Multiple stimuli-responsive nanoformulations. (a) The synthesis and release mechanism of glucose/pH dual-responsive polymer–COFs composites. Reproduced with permission from Zhang et al. (2020b). Copyright 2020, Wiley. (b) The chemical structure and schematic of glucose/H₂O₂ dual-responsive vesicles for insulin delivery. Reproduced with permission from Hu et al. (2017). Copyright 2017, American Chemical Society. (c) Schematic illustration of glucose/hypoxia dual-responsive switch for blood glucose regulation. Reproduced with permission from Zhou et al. (2020b). Copyright 2020, Springer Nature. (d) Mechanistic diagram of glucose/hypoxia/H₂O₂-sensitive vesicles for enhanced insulin delivery. Reproduced with permission from Yu et al. (2017). Copyright 2017, American Chemical Society.

GOx was loaded into an H₂O₂-sensitive matrix to achieve glucose- and hypoxia-responsive release of insulin. Glucose/H₂O₂-responsive vesicles were formed by encapsulating GOx and insulin self-assembly in a block copolymer composed of PEG and phenylboronic ester (PBE) (Hu et al. 2017). GOx catalyses the generation of H₂O₂ from glucose, triggering dissociation of PBE side chains. Subsequently, the copolymer could become water-soluble, causing the vesicle structure to disintegrate and release insulin (Figure 6b) (Hu et al. 2017).

GOx was encapsulated in a hypoxia-sensitive matrix to achieve glucose/hypoxia-responsive release of insulin (Zhou et al. 2020b). An intelligent glucose/hypoxia dual-responsive switch was constructed by co-loading insulin and GOx using amphiphilic 2-nitroimidazole-_L-cysteine alginate (NI-CYS-ALG) conjugates (Zhou et al. 2020b). Glucose oxidation catalysed by GOx can create a hypoxic environment, triggering the substitution of hydrophobic nitroimidazole with hydrophilic aminoimidazole, which promotes the dissociation of the nanoparticle system and the release of insulin (Figure 6c) (Zhou et al. 2020b).

3.3.5. Multiple-stimulus responsive

Multiple stimuli-responsive release systems, such as glucose/hypoxia/H₂O₂ and glucose/pH/redox, have been developed. The integration of multiple stimuli-responsive elements creates hierarchical release systems capable of responding to sequential environments encountered during gastrointestinal transit, thereby significantly enhancing the site-specific delivery of functional insulin.

Glucose─, hypoxia─, and H₂O₂─responsive polymersome-based vesicles (d-GRPs) have been developed for the delivery of insulin. A diblock copolymer sensitive to both hypoxia and H₂O₂ was developed using PEG and polyserine modified with 2-nitroimidazole via a thioether moiety (Yu et al. 2017). The thioether group can be converted to hydrophilic sulphone triggered by H₂O₂, while the nitroimidazole group can be reduced to hydrophilic aminoimidazole under hypoxia conditions (Yu et al. 2017). Under hyperglycemic conditions, GOx-loaded d-GRPs can catalyse the consumption of oxygen by glucose, producing local hypoxia and H₂O₂, which facilitates the solubility switch of the polymer and the disintegration of polymersome-based vesicles, thereby achieving controlled release of insulin (Figure 6d) (Yu et al. 2017).

Glucose, pH, and redox-responsive amphiphilic copolymer containing poly(ethylene oxide) (PEO) and poly(3-acrylamidophenylboronic acid) (PAPBA) linked by a disulphide bond (PEO–SS–PAPBA) was developed for controlled release of insulin (Yuan et al. 2014). Copolymer-based micelles containing disulphide bonds could detach the hydrophilic PEO shell from the hydrophobic PAPBA core in the presence of glutathione (GSH), while the PAPBA core could be changed from hydrophobicity to hydrophilicity under pH and glucose stimulation, resulting in dissociation of micelles to trigger insulin release (Yuan et al. 2014).

Critical assessment of the emerging paradigm

The design of responsive nanoparticles aims to achieve site-specific drug release at disease locations (e.g. tumours, inflamed intestines) to enhance targeting efficacy and reduce side effects (Hu et al. 2024). However, upon entering the systemic circulation, most nanoparticles interact with various blood components, which can alter their stability and lead to premature release of the drug before reaching the target site. This off-target release not only diminishes therapeutic efficacy but may also cause tissue toxicity (Zhang et al. 2020a). Furthermore, inter-individual variations in disease states and physiological microenvironments significantly influence the responsive release behaviour and ultimate therapeutic outcome of nanoparticles (Zhang et al. 2020a; Hu et al. 2024). Accurate safety assessment of bio-sensitive nanoparticles necessitates real-time and precise in vivo tracking. Yet, current tracing technologies struggle to distinguish between intact nanoparticles and released signals or drug cargo within complex physiological environments, leading to misinterpretation of their true biodistribution and metabolic fate (Zhang et al. 2020a).

Consequently, design parameters obtained from in vitro studies may fail to accurately predict the dynamic in vivo performance in humans, introducing uncontrollable risks. Additionally, responsive nanoparticles often exhibit complex structures (e.g. multi-layer coatings, multifunctional ligand modifications), making their production processes far more complex, time-consuming, and costly than those for conventional formulations. This poses significant challenges for large-scale, low-cost, and high-consistency manufacturing, creating a substantial barrier to clinical translation and commercialisation. The previous regulatory science framework lacked comprehensive evaluation criteria for such complex nanomedicines, increasing the uncertainty in the approval process.

3.4. Synergistic Integration

Integration of structural modifications with advanced delivery systems represents a particularly promising approach, leveraging complementary mechanisms to overcome multiple barriers to oral insulin delivery. Modified insulin analogues with enhanced stability can be incorporated into protective nanocarriers, creating dual protection systems with improved robustness against the variable conditions encountered during gastrointestinal transit.

Fast-acting insulin peptide (Actrapid®) and slow-acting insulin detemir (Levemir®) encapsulated in a lipidic cubic phase were embedded within a poly-(methacrylic acid)-b-(methyl methacrylate) (PMMA) capsule shell (Strachan et al. 2023). Insulin detemir enhances hydrophobicity and alters pharmacokinetics by attaching a 14-carbon fatty acid (myristic acid) to the lysine at the B29 position and removing threonine at the B30 position, thereby prolonging the glucose-lowering effect to 14 hours. The PMAA capsule matrix could disintegrate in the intestinal tract, triggering enzymatic breakdown of the lipid carrier to dissociate, ultimately leading to the release of encapsulated insulin (Strachan et al. 2023). The thickness of the enteric coating could significantly influence the in vivo glucose-lowering efficacy of insulin, quantified as the percentage of the blood glucose level-time curve (AAC). Compared with subcutaneous injection, a thinner coating (~160 μm) enhanced the pharmacological availability (PA) of fast-acting and slow-acting insulin to 99% and 150%, respectively, while a thicker coating (~500 μm) achieved only 47% and 48% (Strachan et al. 2023).

PEGylated insulin can be combined with the poly(methacrylic acid-grafted ethylene glycol) glycol P(MAA-g-EG) gel to form a pH-sensitive anionic hydrogel, which could protect insulin from degradation after release from the carrier (Coolich et al. 2023). In an acidic gastric environment (pH < pKa), the hydrogel can form a hydrogen-bonded network to safeguard the drug. Conversely, the deprotonation of carboxylic acid groups can trigger charge repulsion and cause swelling in a neutral intestinal environment (pH > pKa), enabling the intelligent controlled release of insulin in the intestinal tract (Coolich et al. 2023).

In another study, mucoadhesive N-trimethyl chitosan chloride-coated PLGA nanoparticles can be loaded with an insulin-LMWP conjugate (insulin-LMWP MNPs) to prevent enzymatic hydrolysis and penetration through mucosal barriers (Sheng et al. 2016). Subcutaneous injection of insulin-LMWP conjugate showed long-lasting hypoglycaemia effects (12 h), while oral administration of the conjugate alone was ineffective due to protease degradation. After encapsulation with MNPs, the PA of oral insulin delivery with MNPs (at the insulin dose of 20 IU/kg) was significantly increased to 11.24% compared to subcutaneous injections of insulin solution (2 IU/kg, n = 6). More importantly, the PA of insulin-LMWP MNPs reached 17.98% (Sheng et al. 2016).

3.5. Critical comparison of competing technologies with analysis of clinical translation potential

Analysis of competing technologies reveals varying potential for clinical translation. Lipid-based systems generally demonstrate superior stability profiles and manufacturing scalability, but often have limited loading capacity for hydrophilic peptides, such as insulin (Niu et al. 2016). Polymeric nanoparticles offer excellent control over release kinetics but face challenges related to batch-to-batch reproducibility and the potential toxicity of certain polymers (Eltaib 2025). Due to their non-biodegradability and nonspecific distribution, inorganic nanoparticles require further research in non-human primates to evaluate their translatability across species and potential for clinical studies (Huang et al. 2020). Bio-inspired systems exhibit promising biological interactions but present greater manufacturing complexity and regulatory hurdles (Ming et al. 2024). The integrated approach recognises that no single technology is likely to overcome all barriers to oral insulin delivery simultaneously. A strategic combination of complementary technologies, each addressing specific aspects of the delivery challenge, presents the most promising path toward clinically viable oral insulin. Technologies incorporating well-established pharmaceutical excipients with proven safety profiles possess advantages in regulatory approval pathways. Economic considerations favour approaches compatible with existing pharmaceutical manufacturing infrastructure. Clinical translation will likely favour technologies that balance efficacy with practical considerations, such as manufacturing, stability, and regulatory compliance, rather than those demonstrating the highest theoretical performance in idealised laboratory conditions (Agrahari and Hiremath 2017).

4. Osmotic enhancer and biomimetic strategies

4.1. Osmotic enhancers and their application to insulin delivery

To address the barrier of intestinal epithelial cells, the introduction of osmotic enhancers can improve permeability of intestinal epithelium to promote drug absorption, such as sodium dodecyl sulphate (SDS) (Lin et al. 2017), sucrose laurate (McCartney et al. 2019), Labrasol® (McCartney et al. 2019), d-Polyarginine lipopeptides (Garcia et al. 2018), thiolated polymers (Zhang et al. 2018), and ionic liquid (IL) (Banerjee et al. 2018). Surfactant SDS-coated acid-resistant metal organic framework (PCN-222) has been developed to enhance the oral delivery of insulin (Zhu et al. 2025b). SDS regulates cell deformation by reducing the surface hydrophilicity of nanocarriers, enhancing the permeability of the epithelial barrier, and significantly improving the absorption of nanocarriers by intestinal epithelial cells (Zhu et al. 2025b). A study utilised surfactant polysorbate-80 (Tween-80) and zinc ions to prepare self-assembled insulin micelles (Figure 7a) (Xu et al. 2024). Tween-80 enhanced mucus penetration and opened tight junctions to improve cellular uptake and drug permeability, and inhibited functional activity of P-glycoprotein (P-gp), to reduce drug efflux and promote insulin transport (Xu et al. 2024). Additionally, pre-activated thiolated polymeric poly(acrylic acid)-cysteine-6-mercaptonicotinic acid (PAA−Cys−6MNA, PC6)-coated chitosan nanoparticles can also overcome both the mucus barrier and epithelial barrier (Figure 7b) (Zhou et al. 2020a). The formulation could dilute mucus to enhance permeability, open tight junctions to facilitate insulin paracellular transport, neutralise surface charge to reduce resistance, and expose cationic properties of chitosan via coating shedding to augment epithelial cell adhesion and uptake efficiency. These mechanisms collectively enhance oral relative bioavailability by up to 16.22% (Zhou et al. 2020a).

Figure 7.

Osmotic enhancer for enhanced oral delivery of insulin. (a) Schematic illustration of the formation and mechanism of insulin-loaded Tween-80 micelles. Reproduced with permission from Xu et al. (2024). Copyright 2024, American Chemical Society. (b) Schematic illustration of a thiolated-polymer-based nanodrug delivery system to overcome the mucus barrier and epithelial barrier. Reproduced with permission from Zhou et al. (2020a). Copyright 2019, American Chemical Society.

Cell-penetrating peptides (CPPs) have demonstrated significant potential for enhancing insulin delivery across intestinal epithelia. Cationic CPPs, such as TAT, and poly-arginine sequences, facilitate interactions with negatively charged cell membranes, thereby enhancing cellular uptake through various mechanisms, including direct translocation and endocytosis (Khafagy and Morishita 2012). Amphipathic CPPs like MAP and MPG induce temporary membrane destabilization, facilitating drug transport (Avci et al. 2018; Langel 2021). Hydrophobic CPPs, such as Pep-7, interact with membrane lipids, creating transient pathways for the passage of drugs (Avci et al. 2018). Strategic conjugation of CPPs to insulin requires careful consideration of attachment site and linker chemistry to maintain both penetration enhancement and insulin activity. Recent developments include stimuli-responsive CPPs that activate only under specific conditions, potentially reducing off-target effects and enhancing site-specific delivery in the intestinal environment.

N-(2-hydroxypropyl) methacrylamide copolymer (pHPMA) copolymer and CPP (penetratin)-modified nanoparticles could enhance epithelial tissue absorption of insulin by promoting mucus penetration and facilitating epithelial cell transport. Absorption of nanoparticles on mucous-secreting epithelial cells is 20 times higher than that of free insulin (Shan et al. 2015). Another study has also reported the development of colon-specific (CS-CPP) nanoparticles co-modified with CPP and amphiphilic chitosan derivatives to enhance oral colonic insulin absorption (Guo et al. 2016; Guo et al. 2019). Four amphiphilic chitosan derivatives and three CPPs (TAT, penetratin, and R8) were used to modify PLGA nanoparticles, investigating the penetration efficacy of different CPP-modified nanoparticles (Guo et al. 2019). A study revealed that CS-CPP nanoparticles demonstrated significantly enhanced cellular uptake and extracellular transport performance compared to N-trimethyl-N-dodecyl chitosan (TDCS) nanoparticles. Furthermore, in the Bama minipig model, insulin-loaded TDCS-TAT nanoparticles (10 IU/kg) exhibited more pronounced hypoglycaemia effects than TDCS nanoparticles, achieving a 40% reduction in blood glucose levels. Pharmacokinetic parameters, including AUC (the area under the curve of serum insulin concentration versus time) and maximum blood drug concentration C_max, increased by 1.45-fold and 1.82-fold, respectively (Guo et al. 2019).

Some peptides that enhance paracellular transport (such as zonula occludens toxin and phosphatase peptides) (Taverner et al. 2015; Lee et al. 2016; Almansour et al. 2018) can also enhance drug penetration across the epithelial barrier. Chitosan and zonula occludens toxin-derived peptide-functionalized nanocarriers can increase intestinal epithelial permeability by enhancing adhesion and opening tight junctions, thereby enhancing oral absorption of insulin and effectively reducing blood glucose levels (Lee et al. 2016).

Critical assessment of the emerging paradigm.

At specific doses (e.g. oligoarginine R6 ≤ 25 mg/kg, penetratin ≤ 0.5 mM), CPPs did not cause significant lactate dehydrogenase (LDH) leakage in the rat ileal model, and the integrity of the intestinal mucosa was preserved (Khafagy and Morishita 2012). This favourable safety profile may stem from their unique mechanism of action—primarily entering cells via endocytosis mediated by cell-surface proteoglycans, rather than directly disrupting the cell membrane or tight junction structures. This principle may confer CPPs a potentially wider therapeutic window; however, systematic long-term toxicological studies and precise safety thresholds remain to be established.

Permeation enhancers can achieve an acceptable therapeutic window through careful selection of the enhancer type, controlled concentration, and utilisation of formulation technologies. Mucosal toxicity is significantly model-dependent. For example, Labrasol® (8 mg/mL) caused tissue damage in ex vivo models, but in in vivo instillation studies, even at higher concentrations (40 mg/mL), it only resulted in mild and reversible cell sloughing (McCartney et al. 2019). This highlights the critical importance of in vivo models with intact blood flow and repair mechanisms for assessing the true risk. Besides, studies on Labrasol® (8 mg/mL) directly demonstrated that the reduction in transepithelial electrical resistance (TEER) it induced was reversibly restored by up to 70% after the enhancer was removed, indicating that its effect is transient and reversible (McCartney et al. 2019). Moreover, both SDS in bubble carriers and sucrose laurate in intestinal instillation models showed that their effective concentrations were considerably lower than those causing significant histological damage or endotoxin absorption (Lin et al. 2017; McCartney et al. 2019).

Based on existing research, the safety risks of tight junction modulators primarily encompass cytotoxicity, reversibility of action, tissue specificity, bystander absorption, and potential impacts on long-term pathological processes such as tumour metastasis. The immunogenicity of biologically derived molecules (e.g. bacterial toxin derivatives or recombinant proteins) must also be prospectively evaluated (Brunner et al. 2021). Among these, the reversibility of the mechanism of action is emphasised as a core indicator for assessing their safety. For example, the effect of AT1002 on Caco-2 cells is reversible within 48 hours and non-toxic, whereas the effect of 10 carbons is reversible after 24 hours but exhibits toxicity at a concentration of 20 mM (Brunner et al. 2021). An ideal reversible modulator should achieve transient, controllable opening of the barrier function, followed by rapid recovery after drug absorption, which is a key prerequisite for achieving an acceptable therapeutic index. The clinical advancement of Larazotide (Phase II) and 10 carbons-based Gastrointestinal Permeation Enhancement Technology (GIPET) technology (Phase III) depends on their ability to balance efficacy and safety for specific routes of administration and indications.

The synthesis of complex molecules, such as CPPs, particularly their D-enantiomers, presents significant challenges. At the formulation level, issues of compatibility and developability arise (Khafagy and Morishita 2012; Brunner et al. 2021). Tight junction modulators can readily interact with insulin and other formulation components, leading to instability or degradation, while their inherent physicochemical properties often restrict the development of conventional dosage forms. These factors together constitute the fundamental obstacles to its regulatory and commercialisation path.

4.2. Leveraging natural transport mechanisms across intestinal barriers

Exploitation of endogenous transport pathways represents a sophisticated approach to enhancing insulin absorption. Carrier-mediated transport systems, including peptide transporters (PEPT1) (Giacomini 2010), organic anion transporters and apical sodium-dependent bile acid transporters (ASBT) (Alrefai and Gill 2007), can be targeted through conjugation of insulin to substrates recognised by these transporters. These approaches harness evolutionarily conserved mechanisms for nutrient absorption, potentially offering more predictable and physiologically compatible absorption profiles compared to approaches that disrupt epithelial barrier function.

Zwitterionic structures, which possess both positive and negative charges, can readily absorb water to form a protective layer, reducing unnecessary protein adhesion and facilitating drug penetration through the intestinal mucus layer (Qian et al. 2022). A betaine-based zwitterionic micelle formulation utilises proton-coupled amino acid transporter 1 (PAT1)-mediated transcellular transport, allowing for enhanced insulin delivery across the intestinal epithelium without compromising tight junction integrity (Han et al. 2020). Polyzwitterion/protein nanocomplexes formed by mixing proteins with three monomers containing anionic, cationic, and zwitterionic groups through a polymerisation reaction can enable the effective oral delivery of proteins without opening tight junctions (Fang et al. 2023). The oral relative bioavailability of polyzwitterion/insulin (M₈-insulin) capsules reached up to 16.2% in a diabetic rat model. Moreover, the formulation demonstrated a significant hypoglycaemia effect across multiple animal species, including mice, rats, and pigs (Fang et al. 2023). Additionally, the functional modification of chitosan nanoparticles with γ-polyglutamic acid has also enhanced intestinal absorption through calcium-sensitive receptors and amino acid transporters in the intestinal epithelium (Urimi et al. 2019).

Bile acid transporters offer another potential pathway. In order to enhance intestinal absorption, bile acids such as cholic acid (Zhang et al. 2018), deoxycholic acid (DA) (Fan et al. 2018; Liang et al. 2025), and ursodeoxycholic acid (UDCA) (Ma et al. 2024) are also used to adorn the nanocarriers to achieve ASBT-mediated transmembrane transport. Liver-targeted delivery of insulin-loaded nanoparticles consisted of cholic acid, quaternary ammonium modified chitosan derivative and HPMCP via enterohepatic circulation of bile acids for treatment of diabetes mellitus (Zhang et al. 2018). DA-modified chitosan could promote intestinal epithelial transport and basolateral release of insulin via ASBT and cytosolic ileal bile acid-binding protein (IBABP)-mediated pathways. Orally administered freeze-dried DA-modified nanoparticles (DNPs)-loaded enteric-coated capsules produced a significant hypoglycaemia effect (Fan et al. 2018). Other literature has reported that DA-modified Heyndrickxia coagulans spores can effectively penetrate mucus and increase basolateral drug release through the bile acid pathway. In a rat model of type 1 diabetes, the drug achieved an oral relative bioavailability of 15.1% and provided significant hypoglycaemia effects (Liang et al. 2025). Another study has prepared UDCA-decorated zwitterionic nanoparticles for liver-targeted delivery of insulin (UC-CMs@ins) (Ma et al. 2024). UC-CMs@ins could prevent insulin degradation in the gastrointestinal tract by maintaining a crosslinked structure and releasing insulin in a closed-loop manner (Ma et al. 2024). The introduction of betaine and UDCA could overcome mucosal and intestinal barriers through the ASBT and PAT1 dual pathways, promoting basal insulin release and increasing insulin accumulation in the liver (Figure 8a) (Ma et al. 2024).

Figure 8.

Exploitation of endogenous transport pathways and receptor-mediated transcytosis approaches for enhanced oral delivery of insulin. (a) Schematic illustration of the formation and mechanism of UDCA-decorated zwitterionic nanoparticles. Reproduced with permission from Ma et al. (2024). Copyright 2024, Wiley. (b) The synthesis and release mechanism of FA-modified Zr-based MOF. Reproduced with permission from Zou et al. (2025). Copyright 2025, Elsevier.

Critical assessment of the emerging paradigm.

Despite the significant enhancement of oral relative bioavailability achieved by utilising endogenous transporters (such as ASBT and PEPT1) for insulin delivery, its clinical translation still faces multiple challenges. The surfactants and bile acid derivatives employed may pose local risks by disrupting membrane integrity (Wang et al. 2025b). Dynamic alterations in transporter expression under pathological conditions (e.g. upregulation of ASBT/proton-coupled folate transporter (PCFT) in diabetes, downregulation of ASBT/monocarboxylate transporter 1 (MCT1) in inflammatory bowel disease could compromise dosing consistency and competitively interfere with the physiological balance of endogenous substrates like bile acids (Cho et al. 2024). The complex synthesis process, which requires precise control over ligand density and orientation, together with the resultant batch-to-batch variability and high cost, presents a major obstacle to its translation from the laboratory to the market (Pangeni et al. 2021). Moreover, any strategy that modifies intestinal barrier function warrants rigorous regulatory scrutiny, necessitating comprehensive data on colloidal stability, functional reversibility, and batch-to-batch consistency (Cho et al. 2024).

4.3. Receptor-mediated transcytosis approaches

Receptor-mediated transcytosis offers a specific, energy-dependent mechanism for insulin transport across the intestinal epithelium. Intelligent delivery systems with biomimetic properties can be constructed by integrating functional ligands, such as Fc fragment (Azevedo et al. 2020), transferrin (Zou et al. 2022), and vitamin (Zhang and Wu 2014a), onto the surface of nanocarriers to target widely distributed receptors in intestinal epithelial cells. Claudin-4 targeting peptides interact with tight junction proteins, potentially offering a complementary paracellular enhancement mechanism to transcellular approaches (Erramilli et al. 2024). These receptor-targeting strategies benefit from the specificity of receptor-ligand interactions, potentially reducing off-target effects compared to nonspecific permeation enhancement approaches (Zheng et al. 2021). Expression of these receptors throughout the intestinal tract offers the possibility of absorption along extended segments of the gastrointestinal tract, potentially reducing variability in absorption (Hamman et al. 2007).

For example, transferrin-coated acid-resistant metal-organic framework nanoparticles (UiO-68-NH₂) also improved the oral delivery efficiency of insulin (Zou et al. 2022). The formulation could protect insulin from acid and enzymatic degradation, facilitate rapid absorption by intestinal cells, and achieve a relative bioavailability of 29.6% orally at a dose of 20 IU/kg compared to subcutaneous injection (5 IU/kg, n = 6) (Zou et al. 2022). FA-conjugated MOF nanoparticle (FA-PCN-777) was developed for FA receptor-mediated endocytosis, achieving specific intestinal transport and highly controllable insulin release (Figure 8b) (Zou et al. 2025). Importantly, oral insulin relative bioavailability can reach up to 35.5% in diabetic rat models (50 IU/kg, n = 5), and achieve a prolonged oral hypoglycaemia effect for 48 hours in diabetic New Zealand rabbit models (30 IU/kg, n = 3) compared to subcutaneous injection of insulin (5 IU/kg). The high relative bioavailability can be attributed to the selective uptake of the nanoparticles mediated by the upregulated PCFT in the gut of alloxan-induced diabetic rat models, and the delayed degradation of the MOF, which ensures sustained release. Furthermore, interspecies extrapolation must be interpreted with caution due to physiological variations. The system's translational potential depends on further validation of batch-to-batch reproducibility and long-term safety in higher species.

Biotinylated liposomes (Zhang et al. 2014b) and FA-modified liposomes (Yazdi et al. 2020) have been developed to enhance the oral delivery of insulin. FA-targeted PEGylated liposomes can enhance stability in the gastrointestinal tract and improve intestinal epithelial absorption, leading to improved hypoglycaemia effects and higher serum insulin levels (Yazdi et al. 2020). Additionally, liposomes modified with biomimetic thiamine and niacin can enhance the resistance of liposome carriers to acid- or enzyme-induced damage and exhibit significant hypoglycaemia effects lasting up to 12 h (He et al. 2018).

Critical assessment of the emerging paradigm

Receptor-mediated endocytosis (e.g. the VB12, folate, or transferrin receptor pathways) offers enhanced cellular specificity for the oral delivery of insulin. Although the use of natural ligands (such as folate or transferrin) offers advantages in biocompatibility, receptor saturation may competitively inhibit the absorption of endogenous nutrients (e.g. folate, iron), potentially leading to long-term metabolic deficiencies (Hamman et al. 2007; Zhang and Wu 2014a). The risk of immunogenicity cannot be overlooked, particularly when using heterologous or modified ligands (e.g. anti-transferrin receptor antibodies or plant lectins), which may trigger immune responses against the receptor or the carrier complex (Zhang and Wu 2014a). Furthermore, the complex chemical conjugation of ligands to nanocarriers and the large-scale manufacturing process required to ensure the structural integrity of the complexes are extremely complicated and costly (Hamman et al. 2007). Additionally, bioconjugates or biological products face stricter regulatory scrutiny regarding their pharmaceutical properties, stability, and bioequivalence compared to traditional small molecules. These factors collectively contribute to greater uncertainty in the regulatory approval pathway for such complex bioconjugates.

4.4. Binary mutual-assist

Functional nanoparticle (PG-FAPEP) with FA and charge-convertible tripeptide dual-modified PLGA was constructed to improve oral insulin delivery through receptor-mediated endocytosis as well as transporter-mediated transport (Figure 9a) (Xi et al. 2022). Neutrally charged PG-FAPEP could penetrate the mucus layer and target the apical sides of enterocytes via the FA receptor. Subsequently, the surface tripeptide undergoes protonation in acidic microenvironments, endowing PG-FAPEP with enhanced lysosomal escape. In the cytoplasm, it can become electrically neutral and recognise the proton-coupled oligopeptide transporter (PHT1), facilitating the basolateral release of insulin. The relative bioavailability of insulin-loaded PG-FAPEP in HPMCP-coated capsules reached 14.3%, and the hypoglycaemia effect lasted (Xi et al. 2022).

Figure 9.

Ternary mutual-assist nanoformulations for enhanced oral delivery of insulin. (a) Schematic representation of the composition and delivery mechanism of PG-FAPEP with dual modifications (FA and charge-convertible tripeptide). Reproduced with permission from Xi et al. (2022). Copyright 2021, Elsevier. (b) Schematic diagram of the synthesis and delivery mechanism of PLGA nanoparticles composed of IL and VB12-chitosan. Reproduced with permission from Zhou et al. (2023). Copyright 2023, Elsevier.

Ligand-switchable poly PLGA nanoparticles with dual modifications (a pH-responsive stretchable cell-penetrating peptide (Pep) and galactose) were constructed, which could achieve efficient oral delivery by adjusting the conformation of surface proteins and targeting the liver (Yang et al. 2022). The formulation could penetrate the intestinal tract via surface extension in acidic environments and target the liver by exposing galactose under physiological conditions. Results showed that Pep/Gal-PNPs could efficiently deliver insulin to the liver, increase liver glycogen production by 7.2 times, and significantly improve blood glucose control (Yang et al. 2022).

Additionally, multifunctional PLGA nanoparticles composed of IL and VB12-chitosan have been constructed for efficient oral insulin delivery via extracellular and paracellular transport (Figure 9b) (Zhou et al. 2023). IL, as an osmotic enhancer, can significantly improve the storage stability of insulin at room temperature and exhibit a synergistic effect with chitosan, increasing the space between intestinal epithelial cells and effectively promoting the paracellular transport of drugs. Then, VB12-chitosan could achieve mucous adhesion and prolong intestinal retention, while active transport was mediated by the vitamin B12 receptor. Experimental results demonstrated that the formulation could significantly enhance PA, up to 31.8%, thereby improving the hypoglycaemia effect of insulin (Zhou et al. 2023).

4.5. Analysis of physiological implications and safety considerations

While biomimetic approaches offer promising efficacy, physiological implications require careful consideration. Approaches that enhance paracellular transport through tight junction modulation may temporarily compromise barrier function and increase the prevalence of autoimmune diseases (Lerner and Matthias 2015; McCartney et al. 2016). Risk must be balanced against the magnitude and duration of permeation enhancement. Similarly, cell-penetrating peptides with nonspecific mechanisms may facilitate the uptake of co-present substances (Kristensen et al. 2016). Receptor-mediated approaches generally present fewer concerns regarding barrier disruption but may lead to receptor downregulation with repeated exposure (Myers et al. 2022). Safety evaluations must consider not only acute effects but also the consequences of chronic administration, as diabetes management requires lifelong therapy. The ideal approach would maintain epithelial barrier integrity while specifically enhancing insulin transport, with particular attention to limiting systemic exposure to absorption enhancers and preventing unintended absorption of intestinal contents.

5. The microbiome frontier: an untapped resource

5.1. Gut microbiota influence on insulin absorption and metabolism

The gut microbiome has a significant influence on intestinal physiology, with implications highly relevant to oral insulin delivery (Lee et al. 2021). The microbiota modulates intestinal pH, mucus production, and epithelial turnover, which are factors that directly affect drug absorption (Li et al. 2020). Evidence suggests that microbiota-derived short-chain fatty acids enhance tight junction integrity while promoting the secretion of glucagon-like peptide-1 (Cong et al. 2022), potentially creating complementary effects on insulin absorption and glucose homoeostasis. Significant interindividual variations in microbiome composition may partially explain the heterogeneous responses observed in clinical trials of oral insulin formulations (Zhao et al. 2023). These complex interactions between microbiota, host epithelium, and insulin delivery systems represent a largely unexplored dimension in optimising oral insulin therapy.

5.2. Microorganism-based carriers

Microorganisms can also serve as oral drug carriers, such as spores, yeast, and microalgae. Heyndrickxia coagulans spores as generators of autonomous bio-based nanoparticles can withstand the harsh conditions of the gastrointestinal tract and achieve efficient mucus penetration through germination, thereby improving oral insulin delivery and hypoglycaemia therapy (Figure 10a) (Liang et al. 2025). Yeast can open tight junctions between intestinal epithelial cells, thereby improving oral insulin delivery. Recombinant Rhodotorula glutinis strain GM4-DTS-PGK1-CCT, when used as a live cell-based liposome, can be employed for the oral delivery of insulin, H22-LP, and α-MSH (Fei et al. 2019). In the recombinant GM4-DTS-PGK1-CCT strain, the content of phosphatidylcholine increased from 40.8% to 60.7%, and accumulated lecithin was also higher than that in the wild-type strain (Fei et al. 2019). In type 2 diabetic mice, oral administration of recombinant insulin-loaded GM4-DTS-PGK1-CCT resulted in a significant downward trend in blood glucose levels within 1-4 hours after administration, and these levels remained relatively low for at least 6 hours (Fei et al. 2019). Notably, recombinant Rhodotorula glutinis lacking thymidylate synthase (TS) was unable to proliferate in vivo and carried a polypeptide drug for safe and controlled release (Fei et al. 2019). Some studies have utilised microcapsules derived from Saccharomyces cerevisiae for insulin loading (IYMC) (Sabu et al. 2019). The 1,3-β-glucan polysaccharide of glucan particles in the outer shell of baker's yeast facilitated efficient drug uptake through the endocytosis of M cells, and alginate-coated IYMC successfully reduced blood glucose levels in diabetic-induced Sprague-Dawley rats (Sabu et al. 2019). Another study reported a microalgae-based oral insulin delivery strategy with both hypoglycaemia and insulin-sensitising effects (Ren et al. 2023). Two common microalgae species, Chlorella vulgaris (CV) and Spirulina platensis (SP), were selected as drug carriers. Research found that spiral-shaped SP could only load 72.50% of insulin, whereas CV effectively achieved nearly 100% insulin loading at the same insulin concentration (2000 μg/mL) (Ren et al. 2023). An alginate surface-modified CV-based insulin delivery system could reduce blood glucose levels in type 1 diabetic mice, and improve insulin sensitivity in type 2 diabetic mice by modulating gut microbiota, thereby maintaining glycemic homoeostasis (Figure 10b) (Ren et al. 2023).

Figure 10.

Microorganism-based carriers for oral delivery of insulin. (a) The composition and mechanism of the spore-based insulin delivery system. Reproduced with permission from Liang et al. (2025). Copyright 2024, Elsevier. (b) Schematic Illustration of microalgal hydrogels in Type 1 diabetes mellitus and gut microbiota regulation in Type 2 Diabetes Mellitus. Reproduced with permission from Ren et al. (2023). Copyright 2023, American Chemical Society.

Critical assessment of the emerging paradigm.

When utilising microorganisms as oral insulin nanocarriers, it is imperative to evaluate their potential risks carefully. Live microorganisms may pose pathogenicity or trigger opportunistic infections. Their endotoxin load can induce inflammatory responses and potentially disrupt the host's gut microbiota balance (Shuwen et al. 2024). Furthermore, uncertainties remain regarding dose-dependent toxicity, immunogenicity, and the consequences of long-term colonisation or gene-level transfer (Shuwen et al. 2024; Pandey and Pandey 2024). The stability of large-scale fermentation, along with the challenges in purification and preservation of live biotherapeutic products, hinders their clinical translation. Moreover, their unique living nature introduces complex regulatory considerations. The ecological risks that may be triggered by genetically engineered bacteria not only pose complex ethical issues but also further compound regulatory complexity (Shuwen et al. 2024; Pandey and Pandey 2024). Additionally, the costs associated with genetic engineering, fermentation, purification, quality control, and cold chain logistics are significantly higher than those for traditional nano-formulations (Pandey and Pandey 2024). Therefore, advancing the development of such carriers necessitates the implementation of stringent process controls, rigorous preclinical safety assessments, and proactive communication with regulatory agencies to address these challenges systematically.

5.3. Microbiome engineering approaches for insulin delivery

Engineered bacteria present a platform for intestinal insulin delivery. Genetically modified probiotic strains expressing insulin-like growth factor-1 (IGF-1) could serve as living production systems within the intestinal tract, continuously generating insulin in proximity to absorption sites (Wang et al. 2024b). Bacteria engineered to secrete insulin in response to glucose fluctuations could create fully autonomous, closed-loop delivery systems. For such applications, promising chassis organisms include Lactobacillus and Bifidobacterium species (Huang et al. 2022), which have been extensively studied. Beyond insulin production, engineered bacteria could simultaneously express protective proteins that shield insulin from degradation (Mejia-Pitta et al. 2021). While still largely experimental, these approaches offer potentially transformative possibilities for sustained and controlled insulin delivery, which could overcome pharmacokinetic limitations.

5.4. Synbiotic formulations enhancing oral bioavailability

Synbiotic formulations combining probiotics with prebiotics offer multifaceted benefits for oral insulin delivery. Specific probiotic strains can modulate intestinal conditions to favour insulin stability, such as UV-killed Lactobacillus acidophilus, which inhibits the insulin-degrading enzyme, thereby protecting insulin from degradation (Neyazi et al. 2018). Prebiotics, such as fructooligosaccharides, can selectively promote the growth of beneficial bacteria while generating short-chain fatty acids that improve insulin sensitivity (Singh et al. 2025). They may also potentially enhance epithelial permeability through the controlled modulation of tight junctions (Wongkrasant et al. 2020). Synbiotics could create favourable microenvironments for absorption while simultaneously supporting overall metabolic health (Kim et al. 2018; Gomez Quintero et al. 2022), and could be utilised for oral insulin delivery. The approach recognises the integral role of gut microbiota in metabolic health, creating potential synergies between insulin delivery and comprehensive diabetes management.

5.5. Critical assessment of emerging paradigm

Microbiome-centred approach to oral insulin delivery represents a paradigm shift, but significant challenges remain. Interindividual variability in microbiome composition introduces unpredictable elements into therapeutic response, potentially necessitating personalised approaches (Gilbert et al. 2025). The selection of chassis and colonisation efficiency of engineered probiotics under varied physiological conditions remain concerns, as does the potential for horizontal gene transfer (Huang et al. 2022). Regulatory considerations for living therapeutics are complex and continually evolving (Cordaillat-Simmons et al. 2020). Careful evaluation is required for the long-term effects of microbiome modulation, particularly in patients with diabetes, who often have a baseline microbiome dysbiosis (Craciun et al. 2022). Despite these challenges, microbiome-based approaches offer unique advantages in terms of sustained production (Gelli et al. 2025) and potential for closed-loop systems, which may justify development complexities, particularly for improving management of chronic conditions like diabetes, where daily administration is required indefinitely.

6. Progress and challenges in oral insulin clinical research

Overcoming the formidable barriers of the gastrointestinal tract has been the central challenge in the development of oral insulin. As shown in Table 2, extensive research efforts have been dedicated to advanced formulation strategies, particularly various nanotechnology-based approaches, to enhance the systemic delivery of insulin following oral administration. These methods have achieved varying degrees of improvement in relative bioavailability, yet a substantial gap remains when compared to subcutaneous injection. Table 3, several technology platforms have been explored, each with distinct characteristics and varying levels of clinical success.

Table 2.

Comparison of the efficiency of nanotechnology approaches for oral insulin delivery.

Nanoparticle platform	Carrier material	Dose (IU/kg)	PA (%)	F (%)	Sample size (n)	Species	Ref
Polymer-based nanocarrier	PEGylated Zein	50	14.9	10.2	6	Male Wistar rats (180–220 g)	(Inchaurraga et al. 2020)
	HA-decorated Lys-aaPEAs	30	8.51	9.85	5	Male Kunming mice (~35 g)	(Han et al. 2023)
	HA-coated chitosan	20	13.8	/	6	Male Institute of Cancer Research (ICR) mice (16–18 g)	(Wu et al. 2022b)
	Chitosan-coated porous PLGA MPs	50	11.52	/	5	Male Sprague–Dawley rats (170–200 g)	(Eilleia et al. 2018)
Lipid-based nanocarrier	SMEDDS	50	3.23	/	5–7	Male Sprague Dawley rats	(Goo et al. 2022)
	Protein corona liposomes	75	10.9	11.9	6	Male Sprague–Dawley rats (180–220 g)	(Wang et al. 2019a)
	AINS-Lip-Gel (liposome-in-alginate hydrogels)	40	11.0	/	6	Female Institute of Cancer Research (ICR) mice (16–18 g)	(Wu et al. 2023)
	‘Oil-soluble’ reversed lipid nanoparticles	1.73	/	28.70	6	Male Wistar rats	(Wang et al. 2020)
Metal-organic framework	PCN-222 (I@P)	50	/	0.34	5	Male Sprague–Dawley rats (180–200 g)	(Zou et al. 2025)
	nCOF	50	24.1	/	3	Wistar rats (180–220 g)	(Benyettou et al. 2021)
Bio-inspired	Milk-derived exosomes	30	4.03	5.69	5	Male Sprague–Dawley rats (190–210 g)	(Wu et al. 2022a)
	Virus-mimicking nanoparticles	50	7.02	/	4	Wistar rats (180–220 g)	(Cheng et al. 2021)
Stimuli-responsive formulation	Layer-by-layer calcium phosphate nanoparticles (VitB12-Chi-CPNPs)	50	5.89	26.91	5	Male Wistar rats (105–135 g	(Verma et al. 2016)
	Net-neutral nanoparticles-extruded microcapsules	20	52	43.75	5	Male C57BL/6J (20−22 g)	(Lu et al. 2023)
	VB12-FU-ConA	50	13	/	5	Male Kunming mice (30−35 g)	(Zhou et al. 2024)
	PCB-PEA-FPBA (PPF)	50	/	18.9	5	Male C57BL/mice	(Ji et al. 2024)
	GOx-loaded NI-CYS-ALG conjugates	75	/	6.25	5	Male Sprague–Dawley rats	(Zhou et al. 2020b)
Osmotic enhancer-based nanocarrier	Sodium alginate-coated SDS@PCN-222	50	/	12.9	3	Male C57BL/6 mice	(Zhu et al. 2025b)
	TW-Zn-rhINS micelle capsules	30	/	7.88	4−5	Male ICR mice	(Xu et al. 2024)
	PC6-coated chitosan	50	7.6	16.22	4	Sprague–Dawley rats	(Zhou et al. 2020a)
	pHPMAs-1 coated CPP nanocomplex	75	6.61	3.02	5	Male Sprague–Dawley rats (180−220 g)	(Shan et al. 2015)
Transporter-based nanocarrier	Polyzwitterionzwitterionic/INS (M8-INS) capsules	20	/	16.2	4	Male rats	(Fang et al. 2023)
	Cholic acid- quaternary ammonium modified chitosan derivative/HPMCP	30	26.9	/	5	Female ICR mice (23−27 g)	(Zhang et al. 2018)
	Enteric-coated capsules with freeze-dried DNPs	30	/	15.9	6	Male Sprague–Dawley rats (180−200 g)	(Fan et al. 2018)
	DA-modified Heyndrickxia coagulans spores	30	/	15.1	6	Male Sprague–Dawley rat (180–220 g)	(Liang et al. 2025)
	UDCA-decorated zwitterionic nanoparticles	25	26.7	/	5	Male Balb/c mice	(Ma et al. 2024)
Receptor-based nanocarrier	Transferrin-coated acid-resistant UiO-68-NH2	20	/	29.6	6	Male Sprague–Dawley rats (180−200 g)	(Zou et al. 2022)
	FA-modified PCN-222 (I@F-P)	50	/	35.5	5	Male Sprague-Dawley rats (180–200 g)	(Zou et al. 2025)
	FA-modified PEGylated liposomes 1%	50	19.08	/	5	Wistar rats (35–40 g)	(Yazdi et al. 2020)
	FA decorated virus-mimicking nanoparticles (PBCA/FA12.51-CS/HA)	50	9.8	/	4	Wistar rats (180–220 g)	(Cheng et al. 2021)
	PG-FAPEP in HPMCP-coated capsules	45		14.3	6	Male Sprague–Dawley rats (180–220 g)	(Xi et al. 2022)
Binary mutual-assist	Pep/Gal-PNPs	75	10.1	7.7	6	Male Sprague–Dawley rats (200–220 g)	(Yang et al. 2022)
	VB12-CS-PLGA@IL	50	31.8	/	3	Male Kunming mice (30–35 g)	(Zhou et al. 2023)

AAC: the area above the blood glucose level−time curve. AUC: the area under the plasma/serum concentration−time curve; PA%: pharmacological availability; F%: relative bioavailability.

The relative bioavailability (F) was calculated by using the following equation:

$$F(\%)=\frac{\mathit{AUC}\left(\mathit{oral}\right)/\mathit{dose}\left(\mathit{oral}\right)}{\mathit{AUC}(s.c\text{.})/\mathit{dose}(s.c\text{.})}$$

The pharmacological availability (PA) was calculated by using the following equation:

$$\mathit{PA}(\%)=\frac{\mathit{AAC}\left(\mathit{oral}\right)/\mathit{dose}\left(\mathit{oral}\right)}{\mathit{AAC}(s.c\text{.})/\mathit{dose}(s.c\text{.})}$$

Table 3.

Characteristics and clinical status of different oral delivery technology platforms.

Technology platform	Mechanism	Advantages	Limitations	Clinical status
Polymer-based nanocarrier	• Encapsulating for protection • Endocytosis	• Tunable release profiles • Biodegradability	• Limited drug loading • Potential toxicity risks	Phase II (ORA2)
Lipid-based nanocarrier	• Encapsulating for protection • Endocytosis	• High biocompatibility • High loading potential	• Poor physical stability • Susceptible to enzymatic degradation	Phase III (HDV-I)
Inorganic carrier	•Nanoconfinement protects from enzymes	• Good stability • High loading capacity	• Poor biodegradability • Unknown long-term toxicity	Phase II (Oshadi Oral Insulin)
Bio-inspired carrier	• Mimicking biological penetration • Enabling intrinsic targeting	• Superior penetration and biocompatibility • Programmable functionality	•Immunogenicity requires control • Challenging production	Trials
Stimuli-responsive formulation	• Responding to physiological signals • Enabling site-specific release.	• Controlled release • Targeted delivery	• Complex synthesis • Need for precise response control	Trials
Osmotic enhancer-based nanocarrier	• Disrupting membrane lipids • Opening tight junctions.	• High permeability efficiency • Transient and reversible nature	• Risk of leaky gut • Potential mucosal irritation	Phase II (OI338GT)
Transporter-based nanocarrier	• Utilising membrane transporters • Enabling active absorption.	• No tight junction opening • High active transport efficiency	• High design barrier for carriers • Potential transporter saturation	Trials
Receptor-based nanocarrier	• Targeting specific receptors • Initiating transcellular transport	• Efficient transcellular transport • Crosses intact barrier	• Complex carrier engineering • Receptor downregulation from repeated dosing	Trials

The development of oral insulin is driven by its potential to mimic the physiological portal vein/liver first-pass effect, offer the convenience of non-invasive administration, and thereby potentially enhance patient compliance. However, evaluating its clinical value requires an evidence-based comparison with established subcutaneous injection regimens and careful consideration of the practical challenges associated with its low bioavailability, such as high-dose requirements, complex manufacturing processes, and increased costs.

Novo Nordisk's oral insulin OI338GT employed a strategy combining an acylated insulin analogue with absorption-enhancing agents, supplemented by GIPET, achieving glycemic control comparable to subcutaneous injection (Easa et al. 2019). However, its development was ultimately discontinued after Phase II clinical trials, primarily constrained by low oral relative bioavailability, which necessitated high dosing requirements.

Nevertheless, the value of oral insulin should not be assessed solely on relative bioavailability. A more critical aspect is whether its theoretical physiological advantages can translate into measurable clinical endpoints. The advancement of the hepatocyte-directed vesicle insulin (HDV-I) capsule to Phase III trials marks a significant breakthrough. HDV-I employs a unique liver-targeting delivery system co-loaded with insulin, hepatocyte-targeting molecules, and biotin-phosphatidylethanolamine, aiming to deliver insulin to the liver (Easa et al. 2019) specifically. This maximises the portal vein/liver first-pass effect and provides compelling evidence for validating its physiological superiority.

Concurrently, a series of innovative formulations designed to overcome gastrointestinal barriers has shown promise in mid-stage clinical trials. For instance, Biocon's Tregopil (IN-105) modifies the insulin molecule itself to improve stability and absorption, and remains in Phase II (Fonte et al. 2013; Easa et al. 2019). Furthermore, companies like Oshadi Drug Administration and Bows Pharmaceuticals AG are exploring systems based on silica nanoparticles/polysaccharide/oil solutions and dextran matrix, respectively, both currently in Phase II clinical studies (Easa et al. 2019). These diverse technological routes collectively enrich the research and development strategy of oral insulin.

The positioning of oral insulin is not to entirely replace injectable insulin but to serve as a differentiated therapeutic option addressing specific clinical needs and patient preferences. Although cost and manufacturing complexity remain challenges on the path to commercialisation, its potential comprehensive benefits in improving treatment adherence and achieving more physiological glucose control constitute a strong value proposition for clinical application.

7. Addressing pharmacokinetic challenges: precision and predictability

7.1. Strategies for reducing inter- and intra-patient variability

Variability in absorption represents a critical challenge for oral insulin therapy. Food effects significantly impact absorption (Cheng and Wong 2020), with high-fat meals potentially delaying insulin release from oral formulations. Approaches addressing include designing fasting-independent formulations through gastro-retentive systems (Mandal et al. 2016) or rapid-dissolving dosage forms (Maheshwari et al. 2024) that release insulin before substantial food interaction occurs. Circadian variations in gastrointestinal function, including enzyme activity and gastrointestinal motility (Segers and Depoortere 2021), necessitate either time-standardised administration or formulations resistant to these variations. Individual differences in gastrointestinal physiology, including pH profiles, enzyme levels, and transit times, contribute to interpatient variability (Abuhelwa et al. 2017). Standardised administration protocols, potentially linked to meals of a defined composition, could further reduce variability; however, such approaches must strike a balance between predictability and patient quality of life considerations.

7.2. Smart delivery devices for oral insulin delivery

An ingestible capsule device named the luminal unfolding microneedle injector (LUMI) was developed (Abramson et al. 2019b). It dissolved in the small intestine (pH ≥ 5.5), activating mechanical arms and pushing dissolvable drug-loaded microneedles into the intestinal wall, enabling efficient delivery of large-molecule drugs (such as insulin) (Abramson et al. 2019b). In the small intestine of a swine model, LUMI demonstrated rapid pharmacokinetics (peak concentrations were reached within 25 min) and achieved a relative bioavailability exceeding that of subcutaneous injection by over 10% (Abramson et al. 2019b).

By mimicking the structure of a leopard tortoise shell, an ingestible self-orienting millimetre-scale applicator (SOMA) for oral drug delivery was designed, featuring a shifted centre of mass, a high-curvature upper shell, and a low-curvature bottom shell (Abramson et al. 2019a). Drug-loaded microcolumns utilise high-pressure compression technology (550 MPa) to compact insulin with PEG into a solid mixture. Sucrose/isomalt dissolution triggered spring then enabled the microcolumn to precisely penetrate the gastric mucosa, ensuring highly efficient drug delivery (Abramson et al. 2019a).

A liquid-injecting self-orienting millimetre-scale applicator (L-SOMA), composed of a liquid medication, an injection needle, and a piston, was developed (Abramson et al. 2022). The L-SOMA system, an oral capsule for gastric submucosal injection of liquid drugs, was evaluated in swine. It delivered doses of up to 4 mg for monoclonal antibodies and GLP-1 analogues, and 0.14 mg for insulin, achieving peak plasma concentrations within 30 minutes. The absolute bioavailability of insulin delivered by L-SOMA reached up to 80% (with a mean of 51% ± 16%) relative to intravenous injection (Abramson et al. 2022). After completion of the injection, the needle automatically retracted and was eventually expelled via the digestive tract (Abramson et al. 2022).

A micromotor-based mini-tablet platform was fabricated by compressing insulin-loaded magnesium micromotors with excipients, followed by coating with an esterified starch layer (Liu et al. 2023). After the outer starch layer degraded in the colon, micromotors were released. Magnesium then reacted with water, propelling micromotors to actively penetrate the intestinal mucosa and achieve both drug protection and controlled release (Liu et al. 2023).

7.3. Chronotherapeutic considerations for insulin delivery

Chronotherapeutic approaches recognise that insulin requirements fluctuate throughout the day in response to circadian metabolic rhythms (Singh et al. 2010). Dawn phenomenon (elevated morning blood glucose resulting from nocturnal hormone surges) may require formulations with precisely timed predawn release. Conversely, increased insulin sensitivity observed in many patients during evening hours may necessitate reduced evening doses. Programmable delivery systems featuring sequential release components could adapt to these changing requirements without increasing administration frequency. Formulations incorporating melatonin could potentially synchronise insulin release with circadian metabolic rhythms (Espino and Pariente 2011). The implementation of chronotherapeutic principles requires a careful balance between optimised timing and practical patient adherence considerations, potentially leveraging digital health technologies to provide timing guidance while minimising patient burden (Forlenza and Lal 2022).

7.4. Analysis of current limitations in predictable absorption profiles

Despite advances in formulation technology, achieving predictable insulin absorption profiles remains challenging. The physiological environment of the gastrointestinal tract, including the pH gradient and transit time, influences the absorption and bioavailability of orally administered drugs (Abuhelwa et al. 2017). For example, the oral bioavailability of peptide drugs can be improved by adjusting the pH value in the stomach through the use of Sodium N-[8-(2-hydroxybenzoyl) amino] caprylate (SNAC) (Kommineni et al. 2023). Gastrointestinal transit time variability significantly impacts the residence time and absorption window of drugs in the small intestine (Yuen 2010). Current formulations demonstrate substantial pharmacokinetic variability. Multiple physiological factors contribute to variability, including the physicochemical properties of drugs, gastrointestinal pH values, gastrointestinal transit times, splanchnic blood flow, lymphatic transport, first-pass metabolism, and formulation characteristics (Abuhelwa et al. 2017). Mathematical modelling approaches incorporating these variables show promise for predicting population pharmacokinetics but require further refinement for individual prediction (Abuhelwa et al. 2017). Emerging evidence suggests that microbiome composition significantly influences pharmacokinetic variability (Liu et al. 2024b), adding another dimension to the complex challenge. Future approaches may require the integration of multiple complementary strategies, potentially including patient-specific formulations based on individualised physiological parameters, to achieve predictable absorption profiles necessary for optimal glycemic control.

7.5. Role of AI in oral insulin delivery

The integration of AI and machine learning into oral insulin delivery research represents a transformative advancement that addresses multiple challenges simultaneously, from molecular design to formulation optimisation and pharmacokinetic prediction. These computational approaches have accelerated the development process by enabling systematic exploration of vast chemical spaces and prediction of biological outcomes that would be impractical to evaluate experimentally. AI-based tools such as machine learning algorithms and molecular dynamics simulations can be harnessed to design nanoparticle systems while predicting with high accuracy their interactions with biological systems (Alshawwa et al. 2022). It should be noted that while AI has demonstrated significant success in broader pharmaceutical and drug delivery applications, its specific application to oral insulin formulation and process design remains in a nascent stage.

7.5.1. AI-driven molecular design and nanocarrier optimisation

Machine learning algorithms, particularly deep learning models and graph neural networks, have found widespread applications in peptide drug design. These computational approaches can predict the impact of structural modifications on insulin stability, receptor binding affinity, and permeability across intestinal epithelium with remarkable accuracy. Transformer-based models have been employed to screen vast chemical libraries, and identify optimal conjugation sites and modification strategies, and predict biological activity (Wu et al. 2024b; Wang et al. 2025a; Kandhare et al. 2025). For example, a study focused on oral peptide drug development directly applied machine learning to predict peptide stability in simulated gastrointestinal fluids (Wang et al. 2023b). The researchers extracted stability data for 109 peptides in simulated gastric fluid (SGF) and simulated intestinal fluid (SIF) from the literature and calculated over 200 physicochemical descriptors for each peptide as model inputs. By comparing various machine learning algorithms, they established classification models with promising predictive performance for stability (best model accuracy of 75.1% for SGF and 69.3% for SIF). Crucially, feature importance analysis revealed that a peptide's lipophilicity, molecular size, and conformational rigidity were the most critical physicochemical determinants of its stability in the gastrointestinal tract (Wang et al. 2023b). These findings provide a computational framework for prioritising insulin analogue designs with improved gastrointestinal stability. PepMimic, an AI platform that designs short peptide binders by mimicking the binding interface of known protein complexes (Kong et al. 2025). Experimental validation confirmed that PepMimic-designed peptides can achieve nanomolar affinity, demonstrating the potential of AI to actively engineer functional peptide therapeutics (Kong et al. 2025). This approach could be strategically applied to design insulin-mimetic peptides with optimised oral bioavailability.

Generative AI models, including generative adversarial networks, enable the de novo design of peptides by learning from existing structure-activity relationship data, significantly reducing the time and resources required for experimental screening (Goles et al. 2024; Kandhare et al. 2025). Structure-aware generative models are driving peptide design toward greater precision. For instance, the HYDRA model integrates diffusion models with a binding affinity optimisation algorithm, enabling the de novo generation of stable, high-affinity therapeutic peptide sequences based on the three-dimensional binding pocket information of target receptors. The model simultaneously considers key pharmaceutical attributes such as peptide stability and hydrophilicity/hydrophobicity during the design process, providing a powerful computational framework for designing next-generation insulin analogues with oral delivery potential (Choudhuri et al. 2024).

Neural network models and Bayesian optimisation approaches have demonstrated substantial utility in optimising nanoparticulate drug delivery systems for oral insulin. The complexity of formulation design, involving multiple interacting variables such as particle size, surface charge, drug loading, and release kinetics, presents an ideal application for machine learning-based optimisation (Chou et al. 2023; Bao et al. 2024). For instance, in the broader field of oral drug delivery, machine learning models trained on relatively small datasets of lipid-based nanoparticles have successfully predicted optimal formulations that enhance drug solubility by up to 3000-fold (Bao et al. 2024). Although this specific study focused on hydrophobic small molecules, the same data-driven optimisation framework could be adapted to oral insulin nanocarriers, where parameters such as particle size and surface charge must be simultaneously optimised to achieve adequate intestinal absorption. Deep learning architectures can forecast critical parameters including particle stability, encapsulation efficiency, and protein corona formation, which significantly influence the in vivo behaviour of oral insulin formulations (Chou et al. 2023; Bao et al. 2024). While the AI-PBPK modelling approach has focused on tumour-targeted nanoparticle delivery, this method of integrating machine learning with physiological modelling to predict nanoparticle biodistribution can be adapted to model the gastrointestinal absorption of oral insulin nanocarriers.

7.5.2. Formulation development and process optimisation

Quality by Design (QbD) principles, when combined with machine learning algorithms, have transformed the systematic optimisation of oral insulin formulations. This integrated approach identifies Critical Quality Attributes (CQAs) and formally links them to product performance and safety. Traditional Design of Experiments approaches, while valuable, explore only limited design spaces. In contrast, AI-guided QbD frameworks enable the identification of critical process parameters and the establishment of design spaces that ensure consistent batch-to-batch reproducibility during scale-up manufacturing (Zhu 2025a).

Recent applications in the nucleic acid delivery field have demonstrated the utility of combining high-throughput experimentation with machine learning to screen extensive libraries of ionisable lipids for efficient drug delivery (Hanna et al. 2025). Although developed for mRNA transfection optimisation, this high-throughput screening and machine learning paradigm is directly transferable to oral insulin formulation development. For oral insulin carriers, such approaches could systematically evaluate combinations of absorption enhancers, enzyme inhibitors, mucoadhesive polymers, and enteric coating materials, representing a multidimensional optimisation problem that is impractical to address through traditional experimental methods alone. By training models on transfection data from combinatorial chemistry libraries, researchers have identified high-performing candidates that would have been difficult to discover through traditional screening methods. Machine learning models can predict critical quality attributes of nanoparticles, including particle size distribution, encapsulation efficiency, and drug release kinetics, based on formulation parameters and processing conditions (Hanna et al. 2025).

7.5.3. Prediction of pharmacokinetic behaviour and bioavailability

Physiologically based pharmacokinetic (PBPK) modelling, enhanced by AI algorithms, offers powerful tools for predicting oral insulin absorption and addressing inter-patient variability. These models incorporate physiological parameters including organ blood flow, tissue partition coefficients, enzymatic activity, and transporter expression to simulate drug behaviour across diverse patient populations (Ozbek and Genc 2024; Talkington et al. 2025). Model-Informed Drug Development (MIDD) approaches leverage these quantitative tools to forecast clinical outcomes, optimise dosing regimens, and potentially reduce the need for extensive clinical studies (Sahasrabudhe et al. 2025). The hybrid models can account for interindividual variability in gastrointestinal physiology, enabling more accurate predictions of clinical outcomes from preclinical data (Djuris et al. 2024; Sahasrabudhe et al. 2025). The application of AI extends to predicting ADMET characteristics, enabling researchers to screen large libraries of potential carrier molecules before committing to costly experimental validation (Ferreira and Carneiro 2025). In the development of oral GLP-1 receptor agonists, AI models are actively employed to predict pharmacokinetic parameters and to screen optimal combinations of permeation enhancers and excipients by analysing large datasets. Furthermore, the integration of PBPK modelling with digital twin concepts is being explored to simulate drug absorption dynamics in virtual gut models. These approaches, while still evolving, provide a validated conceptual and technical roadmap for the systematic, AI-augmented optimisation of oral insulin formulations (Kim and Kim 2025).

7.5.4. Intelligent closed-loop delivery systems

The integration of AI with smart delivery devices represents a paradigm shift toward autonomous insulin management. It should be noted that current AI-driven closed-loop systems have been developed primarily for subcutaneous insulin pumps, where rapid insulin action allows for real-time glucose correction (Vettoretti and Facchinetti 2019; Zhu et al. 2021; Boughton and Hovorka 2021). The translation of these approaches to oral insulin delivery presents distinct challenges that remain largely unexplored. Machine learning algorithms analysing continuous glucose monitoring data can predict glycemic excursions and optimise the timing and dosing of oral insulin administration. Reinforcement learning approaches enable adaptive dosing algorithms that learn from individual patient responses, progressively improving glycemic control over time. These adaptive systems learn directly from observed data without requiring predefined physiological models, enabling personalised treatment strategies that improve long-term glycemic control (Zhu et al. 2021).

Computer models (simulators) can simulate the effects of different dosing strategies, enabling personalised oral insulin regimens tailored to each patient's metabolic profile, dietary patterns, and lifestyle factors (Zhu et al. 2021). The convergence of AI-optimised oral insulin formulations with digital health platforms creates opportunities for comprehensive diabetes management systems. Smart pill technologies incorporating sensors that confirm dissolution, release, and physiological parameters could address adherence monitoring challenges unique to oral formulations (Vettoretti and Facchinetti 2019; Zhu et al. 2021; Boughton and Hovorka 2021). When integrated with continuous glucose monitoring systems through synchronised data platforms, these technologies enable precise timing of oral insulin administration relative to glycemic trends.

The development of AI-driven closed-loop systems for oral insulin delivery faces unique challenges compared to injectable insulin pumps. The delayed and variable absorption kinetics of oral formulations require sophisticated predictive algorithms that anticipate glucose changes rather than simply responding to them. Specifically, the delayed onset of action (typically 30-60 minutes versus 5-15 minutes for rapid-acting injectable analogues), the lower and more variable bioavailability (typically less than 10% versus near-complete absorption), and the influence of food and gastrointestinal conditions create a substantially more complex control problem. Future AI systems for oral insulin will likely require integration of predictive meal detection algorithms, personalised absorption modelling, and potentially hybrid approaches combining oral insulin for basal-bolus replacement with rescue injectable doses for unexpected hyperglycaemia. Machine learning models that combine historical glucose data, meal information, and individual absorption patterns can potentially overcome these challenges, although substantial validation in clinical settings remains necessary (Vettoretti and Facchinetti 2019; Zhu et al. 2021; Boughton and Hovorka 2021).

7.5.5. Accelerating clinical translation and challenges

AI methodologies are increasingly employed to enhance clinical trial design and accelerate regulatory approval pathways. AI methodologies are increasingly employed to enhance clinical trial design and accelerate regulatory approval pathways. A compelling direct application in oral insulin development is the use of explainable AI and causal machine learning for patient stratification. Post-hoc analyses of Phase II and III oral insulin trial data identified specific responder profiles (e.g. patients with lower baseline BMI) who achieved significant HbA1c reductions (~1%), despite the trials not meeting their overall primary endpoints (Kidron et al. 2024; Eldor et al. 2025). These AI-derived insights are directly informing the eligibility criteria for new Phase III studies, demonstrating a powerful MIDD approach to overcome variability and improve clinical success rates. Natural language processing and machine learning can analyse vast repositories of scientific literature and patent databases to identify promising formulation strategies and potential safety signals (Vamathevan et al. 2019; Olawade et al. 2026). AI-assisted patient stratification based on biomarkers, genetic profiles, and microbiome signatures enables the identification of responder subpopulations most likely to benefit from specific oral insulin formulations. Additionally, real-world evidence analysis using AI can supplement clinical trial data, providing regulators with comprehensive safety and efficacy profiles to support approval decisions (Olawade et al. 2026). For oral insulin specifically, AI-assisted clinical trial design could address several unique challenges: identifying optimal dosing windows relative to meals, stratifying patients based on gastrointestinal characteristics (e.g. gastric emptying rate, intestinal permeability markers, microbiome composition), and developing adaptive trial designs that account for the high inter- and intra-individual variability characteristic of oral peptide absorption. Machine learning analysis of early-phase trial data could also help distinguish true non-responders from those who might benefit from modified formulations or dosing regimens.

Despite significant progress, several challenges must be addressed to fully realise the potential of AI in oral insulin delivery. The ‘black box’ nature of deep learning models raises concerns about interpretability and regulatory acceptance, as clinicians and regulatory agencies require understanding of how decisions regarding dosage and delivery are made. Explainable AI approaches, such as local interpretable model-agnostic explanations and Shapley additive explanations, can provide insights into the factors driving model predictions, although their implementation in drug delivery systems remains at an early stage (Linardatos et al. 2021).

Data quality and diversity present additional challenges. AI models trained predominantly on data from specific populations may not perform well across different ethnic backgrounds, age groups, or disease states. Ensuring diverse datasets and regularly auditing models for bias is essential for developing equitable AI-driven therapies (Topol 2019). The regulatory landscape for AI-enabled medical products continues to evolve, with agencies such as the FDA developing frameworks for validating adaptive algorithms that learn and change over time (U.S. Food and Drug Administration 2019a). The successful translation of AI-optimised oral insulin formulations will require continued collaboration between computational scientists, pharmaceutical developers, and regulatory bodies. Standardised protocols for model validation, risk assessment, and ethical approval must be established to ensure that AI-enhanced formulations can progress through development pipelines efficiently while maintaining appropriate safety standards (Topol 2019; U.S. Food and Drug Administration 2019a).

In summary, while AI and machine learning technologies have demonstrated remarkable success in adjacent pharmaceutical fields, their specific application to oral insulin formulation and delivery optimisation remains at an early, exploratory stage. The examples discussed in this section are drawn largely from related areas (optimisation of oral lipid-based nanoparticle formulation, general peptide drug design, PBPK modelling for various drug classes, and closed-loop systems for injectable insulin) to illustrate methodological approaches that hold significant promise for oral insulin development. As dedicated datasets for oral insulin formulations accumulate and as AI tools become more accessible to pharmaceutical scientists, we anticipate accelerated progress in this field. The successful integration of AI into oral insulin development will require close collaboration between computational scientists, formulation experts, and clinical researchers to ensure that these powerful tools are appropriately validated and translated into clinically meaningful improvements in oral insulin therapy.

8. Regulatory, manufacturing, and translational challenges

8.1. Regulatory framework and approval challenges for oral Insulin

Modified insulin products for oral delivery face distinct regulatory challenges compared to conventional insulin formulations. Demonstrating pharmaceutical equivalence becomes complex when the active moiety is structurally modified, necessitating careful characterisation of structure-activity relationships and comparative receptor binding studies (Akbarian et al. 2018). Regulatory agencies have yet to establish specific guidance for comparative bioavailability studies for oral insulin, creating uncertainty in development pathways. Safety evaluations must address excipients and absorption enhancers that are absent in traditional formulations (Heinemann and Jacques 2009), often requiring extensive toxicology studies. The nature of many delivery technologies necessitates an extensive safety evaluation beyond what might be required for new formulations of unmodified insulin, and including considerations of cost, scale-up feasibility, and controlled performance (Duran-Lobato et al. 2020).

8.1.1. Regulatory classification and approval pathways

Oral insulin formulations occupy a complex position within regulatory classification systems. In March 2020, FDA transitioned insulin products from New Drug Applications (NDAs) under the Federal Food, Drug, and Cosmetic Act to Biologics License Applications (BLAs) under the Public Health Service Act (George and Woollett 2019). This transition opened pathways for biosimilar and interchangeable insulin products, potentially increasing competition and reducing costs (U.S. Food and Drug Administration 2020b). For oral insulin, this regulatory shift introduces additional considerations, as developers must navigate both the biologics framework and the unique requirements for novel delivery systems.

Regulatory frameworks for combination products, which involve both a biological drug and a device-like delivery system, add another layer of complexity (Olsen et al. 2023). Many advanced formulations incorporate components that function as both drug and device, requiring navigation of overlapping regulatory pathways. Early and frequent interaction with regulatory authorities is crucial for establishing suitable development pathways for these innovative products.

In the United States, the FDA may classify oral insulin as a new drug application for novel formulations, a biologics license application for products where the insulin component is considered the primary mode of action, or a combination product requiring input from both the Centre for Drug Evaluation and Research (CDER) and the Centre for Devices and Radiological Health (CDRH) when device components are integral to delivery (U.S. Food and Drug Administration 2017; Olsen et al. 2023). The European Medicines Agency (EMA) applies similar considerations under its centralised procedure, with additional scrutiny for advanced therapy medicinal products when novel delivery technologies are employed (Medicines Agency 2014; U.S. Food and Drug Administration 2017). This regulatory ambiguity can result in prolonged review timelines and increased development costs.

8.1.2. Bioequivalence and clinical trial design challenges

Traditional bioequivalence approaches, developed primarily for immediate-release oral medications with predictable absorption, are inadequate for oral insulin products. The different pharmacokinetic profiles of oral versus subcutaneous delivery render conventional comparisons challenging. Oral insulin demonstrates delayed peak concentrations, different time-action profiles, and substantial inter- and intra-subject variability compared to injectable formulations (Heinemann et al. 2015). Regulatory agencies increasingly require pharmacodynamic bioequivalence studies using glucose clamp techniques or meal tolerance tests to demonstrate comparable glucose-lowering effects (Heise et al. 2012).

Controlling blood sugar levels is crucial for patient outcomes. Even if HbA1c levels are similar, the effects of different drugs vary, and multiple indicators need to be comprehensively evaluated (Lipska and Krumholz 2017). The different pharmacokinetic profiles of oral versus subcutaneous delivery render traditional bioequivalence approaches inadequate. Traditional endpoints, such as HbA1c reduction, may inadequately capture the benefits of improved adherence and quality of life associated with oral formulations. Endpoints that account for different delivery patterns while ensuring clinical efficacy must be developed and validated. Patient-reported outcomes, including treatment satisfaction and diabetes-related distress, deserve a more central role in trials and clinical practice, as they may better reflect the real-world benefits (Liarakos et al. 2024).

The high within-subject variability observed in oral insulin pharmacodynamics presents particular challenges for clinical development. The inherent variability in oral insulin pharmacodynamics necessitates larger sample sizes and more complex study designs to achieve adequate statistical power (Heinemann and Jacques 2009; Michels and Gottlieb 2018). Historical attempts at oral insulin development have consistently encountered this obstacle, with absorption variability making it difficult to establish predictable dose-response relationships. Regulatory agencies require demonstration of consistent therapeutic effects across diverse patient populations and conditions, including fed and fasted states, different times of day, and various disease severities (Heinemann and Jacques 2009).

8.1.3. Safety evaluation requirements for novel excipients and enhancers

Oral insulin formulations typically incorporate permeation enhancers, protease inhibitors, and other novel excipients that are absent in traditional insulin products. These components require extensive safety evaluation beyond what might be expected for formulations containing only established excipients. The FDA has indicated that chronic exposure to absorption enhancers warrants particular scrutiny, as repeated disruption of intestinal barrier function could theoretically increase the risk of pathogen entry or unintended absorption of dietary antigens (Maher et al. 2016). Permeation enhancers and tight junction modulators require extensive toxicological evaluation, including assessments of acute and chronic effects on intestinal barrier integrity, potential for enhanced absorption of endotoxins and pathogens, and reversibility of permeation enhancement (Maher et al. 2016).

Nanoparticle-based delivery systems face additional regulatory considerations related to their unique properties. The FDA published guidelines in April 2022 advocating for risk-based frameworks that emphasise physicochemical characterisation of nanomaterials to prioritise safety (U.S. Food and Drug Administration 2022). Minor variations in particle size, charge, morphology, stability, and surface chemistry can alter biological interactions and change biodistribution patterns (U.S. Food and Drug Administration 2022). The complexity of nanoparticle systems, particularly functionalized carriers designed for targeted delivery, creates challenges for establishing appropriate specifications and release criteria (U.S. Food and Drug Administration 2022; Clogston et al. 2024). For nanoparticle-based formulations, regulatory agencies require comprehensive characterisation of biodistribution, accumulation in tissues, and long-term fate of carrier materials (U.S. Food and Drug Administration 2022).

Long-term safety studies must address potential concerns including immunogenicity, local gastrointestinal effects, accumulation of non-biodegradable materials, and impacts of repeated barrier modulation. PEGylation, widely used to improve insulin stability, has emerged as a potential concern following recognition that anti-PEG antibodies are present in a substantial proportion of the general population, potentially affecting pharmacokinetics and causing hypersensitivity reactions (Yang and Lai 2015). These findings highlight the need for comprehensive immunogenicity assessments during development. The potential for immunogenicity is heightened with modified insulin analogues and novel excipients, necessitating anti-drug antibody testing and clinical monitoring for hypersensitivity reactions (Yang and Lai 2015; U.S. Food and Drug Administration 2019b).

8.1.4. Manufacturing and quality control standards

The production of oral insulin formulations, particularly those utilising advanced delivery technologies, presents significant challenges for meeting current Good Manufacturing Practice (cGMP) requirements. Structurally modified insulins require precise control over modification sites and stoichiometry, necessitating sophisticated analytical methods to ensure batch-to-batch consistency. Demonstrating batch-to-batch consistency is particularly challenging for bio-inspired delivery systems and microorganism-based carriers, where biological variability can impact product performance. Nanoparticle-based formulations require validated analytical methods for critical quality attributes, including particle size distribution, surface charge, morphology, drug loading, and encapsulation efficiency. Regulatory agencies expect robust process analytical technology (PAT) implementation to ensure real-time quality control during manufacturing (Tyner et al. 2015; Souto et al. 2020).

Regulatory agencies require demonstration of manufacturing processes that can reliably produce consistent products at commercial scale. Scale-up from laboratory to manufacturing scale while maintaining control over critical quality attributes represents a significant technical challenge for many promising oral insulin technologies. Process validation must demonstrate that manufacturing procedures reproducibly generate products meeting predetermined specifications across multiple batches (Tyner et al. 2015; Younis et al. 2022). Stability testing under various storage conditions must account for the potential for insulin degradation, aggregation, and carrier instability over the product shelf life (Tyner et al. 2015).

8.1.5. Regulatory harmonisation and global market access

Lack of harmonisation between major regulatory agencies creates significant barriers to global market access for oral insulin products. The FDA and EMA often hold differing views on assessment approaches for complex nanosimilar products. For example, regulatory divergence exists between the FDA, which emphasises manufacturing processes and quality control, and the EMA, which focuses more on a product's chemical composition and physicochemical properties (Choi and Han 2018). This difference in regulatory frameworks, combined with the widespread patent strategies (such as filing patents after approval and securing numerous patents on delivery devices) employed by insulin manufacturers to extend market exclusivity, creates a complex challenge for developing products like insulin biosimilars and novel delivery systems (Olsen et al. 2023). These conflicting force developers to navigate a fractured regulatory landscape, increasing costs and delaying patient access.

International Council for Harmonisation (ICH) guidelines provide some framework for global development, but specific guidance for oral biologics remains limited. While ICH provides guidance on pharmaceutical development (Q8-Q12) and safety testing, specific requirements for oral peptide delivery systems vary between the FDA, EMA, and other regulatory authorities (International Council for Harmonisation 2008; U.S. Food and Drug Administration 2022; European Medicines Agency 2025). Regional requirements for clinical trials, manufacturing standards, and post-marketing surveillance add complexity to global development programmes (U.S. Food and Drug Administration 2022; European Medicines Agency 2025). Companies developing oral insulin must design comprehensive regulatory strategies that anticipate requirements across multiple jurisdictions while maintaining efficient development timelines.

8.1.6. Regulatory precedents and guidance documents

While no oral insulin product has received marketing approval from major regulatory agencies, several guidance documents inform development strategies. The FDA's guidance on ‘Bioavailability and Bioequivalence Studies for Orally Administered Drug Products’ (2014) and ‘Liposome Drug Products’ (2018) provide relevant frameworks (U.S. Food and Drug Administration 2014; U.S. Food and Drug Administration 2018). The FDA's and EMA's guidelines on nanotechnology-based medicinal products offer additional direction (U.S. Food and Drug Administration 2022; European Medicines Agency 2025).

The approval of oral semaglutide (Rybelsus), which utilises the absorption enhancer sodium SNAC, has established an important regulatory precedent demonstrating that oral delivery of peptide therapeutics can achieve regulatory approval, albeit with extensive clinical programmes demonstrating safety and efficacy (Aroda et al. 2022). SNAC was granted Generally Recognised as Safe (GRAS) status by the FDA through the approval process for a VB12-SNAC coformulation, and this precedent has facilitated subsequent peptide development programmes (Aroda et al. 2022; Solis-Herrera et al. 2024). While oral semaglutide is a GLP-1 receptor agonist rather than insulin, the regulatory pathway established through its approval provides valuable insights for oral insulin developers regarding the level of evidence required to demonstrate safety and efficacy of oral peptide formulations containing permeation enhancers (Aroda et al. 2022).

8.1.7. Combination product considerations and post-marketing requirements

Many advanced oral insulin formulations incorporate components that function as both drug and device, qualifying them as combination products under FDA regulations. Device-like delivery systems, including smart pills with sensors, ingestible injectors, and microelectromechanical release systems, require navigation of overlapping regulatory pathways. The FDA's Office of Combination Products determines the primary mode of action and assigns products to the appropriate review centre, but the involvement of multiple review divisions can extend review timelines (U.S. Food and Drug Administration 2017). Postmarket requirements for combination products may include device-specific elements such as design controls, human factors testing, and device adverse event reporting in addition to standard drug or biologic requirements (Kramer et al. 2012). These additional obligations must be considered during development planning to ensure appropriate resources and timelines are allocated.

Given the novel nature of oral insulin delivery systems, regulatory agencies may impose enhanced post-marketing surveillance requirements. Risk Evaluation and Mitigation Strategies (REMS) or similar risk management plans may be mandated to monitor for long-term safety signals, particularly those related to gastrointestinal effects of permeation enhancers and potential consequences of chronic nanoparticle exposure (U.S. Food and Drug Administration 2020a). Real-world evidence collection through registries and electronic health record analysis may be required to supplement clinical trial data and confirm long-term safety and effectiveness in diverse patient populations (U.S. Food and Drug Administration 2020a). While oral insulin for general diabetes treatment does not qualify for orphan drug designation, certain specialised applications might be eligible for accelerated development pathways, including paediatric priority review or the FDA's Breakthrough Therapy designation if clinical evidence demonstrates substantial improvement over existing treatments for specific patient populations (Kesselheim et al. 2015).

Early and frequent engagement with regulatory agencies is essential for optimising development strategies. Pre-Investigational New Drug meetings, End-of-Phase 2 meetings, and Pre-BLA meetings provide opportunities to align on development plans, clinical trial designs, and regulatory expectations. For novel oral insulin technologies, these interactions can help identify potential issues early and reduce the risk of costly late-stage development failures. The path to regulatory approval for oral insulin requires careful coordination of preclinical characterisation, clinical development, and manufacturing validation activities. Developers must demonstrate not only that their products are safe and effective but also that they can be manufactured consistently at commercial scale. The complexity of these requirements, combined with the high cost of development, partially explains why oral insulin has remained elusive despite decades of research effort.

8.2. Manufacturing scalability and cost-effectiveness

Manufacturing considerations have a significant impact on the commercial viability of oral insulin technologies. Structurally modified insulins often require complex synthesis, which can substantially increase production costs compared to conventional insulin. Site-specific modifications, essential for maintaining biological activity, require precise control over modification sites and stoichiometry (Lieser et al. 2020), presenting significant challenges in manufacturing.

Advanced delivery systems, such as nanoparticles, often utilise complex fabrication processes with multiple critical parameters that affect product performance. Scaling these processes from the laboratory to commercial scale while maintaining tight control over particle size distribution, morphology, and drug loading presents formidable engineering challenges (Agrahari and Hiremath 2017). Achieving batch-to-batch consistency at the commercial scale remains a significant hurdle for many promising laboratory-scale technologies (Agrahari and Hiremath 2017).

Cost considerations are particularly important for diabetes therapies, which patients require for the remainder of their lives. Low biological potency requires a large amount of insulin, which significantly increases production costs (Heinemann and Jacques 2009). The additional complexity of oral insulin technologies increases manufacturing costs (Gutire 2024), requiring a careful analysis of the value proposition. Higher production costs might be justified by significant clinical advantages or patient preference, but the economic case must be compelling in increasingly cost-conscious healthcare systems. The path to approval for oral insulin involves rigorous safety and efficacy testing, as well as significant regulatory hurdles, a process known for being both lengthy and expensive (Gutire 2024).

8.3. Bridging the gap between laboratory and clinic

Translating promising laboratory findings to clinical success requires systematic approaches to address common failure points in oral insulin development. Early assessment of formulation performance under conditions that reflect physiological variability, including fed and fasted states, circadian variations, and disease-related alterations in gastrointestinal function, can identify robust candidates more likely to succeed in clinical settings.

Predictive modelling, integrating animal models, quantitative systems pharmacology, and experimental clinical trials, can guide rational optimisation before extensive clinical testing (Denayer et al. 2014). Physiologically based pharmacokinetic models, which incorporate detailed parameters of gastrointestinal function, metabolic and transport factors, and population variability parameters (Abuhelwa et al. 2017), enable the in silico prediction of clinical performance, thereby accelerating development while reducing resource requirements.

Clinical trial designs can more efficiently evaluate oral insulin candidates. Adaptive designs, which enable rapid dose optimisation and crossover studies with intensive pharmacokinetic/pharmacodynamic assessment, provide richer data than traditional approaches, thereby reducing the costs associated with drug development (Bahl 2023). Patient stratification based on factors likely to affect response, such as gastrointestinal transit time or enzyme expression profiles, can identify responsive subpopulations and guide personalised approaches.

Strategic partnerships between academic institutions, biotechnology companies, and established pharmaceutical firms can accelerate the translation of research findings into practical applications. Academic research often excels at concept generation but lacks resources for clinical development, while pharmaceutical companies offer development expertise and manufacturing capabilities but may be risk-averse to unproven technologies. Structured collaboration models with appropriate risk-sharing mechanisms can leverage the strengths of each partner, creating more efficient translation pathways.

9. Future perspectives: convergent technologies and paradigm shifts

9.1. Integration of oral insulin with digital health technologies

Convergence of oral insulin delivery with digital health technologies presents transformative opportunities for diabetes management. Smart pill technologies incorporating sensors that confirm dissolution, release, and physiological parameters could address adherence monitoring challenges unique to oral formulations (Jain 2024). Integration with continuous glucose monitoring systems through synchronised data platforms could enable the precise timing of oral insulin administration relative to glycemic trends (Esnafoglyu 2024). Machine learning algorithms analysing patterns of glycemic response to oral insulin could generate personalised dosing recommendations that account for individual absorption characteristics (Eghbali-Zarch and Masoud 2024). These integrated systems could address several limitations of current approaches, particularly in terms of dose timing and adjustment, creating comprehensive management platforms that maintain the convenience of oral delivery while incorporating the precision of digital health technologies.

9.2. Hybrid delivery systems combining multiple mechanisms

Future oral insulin approaches will likely combine complementary delivery mechanisms addressing distinct absorption barriers. Multicompartment systems featuring sequential barrier negotiation (Attarwala et al. 2018) could incorporate an outer layer resistant to gastrointestinal conditions, a middle layer that enhances permeation, and an inner core that protects insulin integrity until it is absorbed. Combinatorial approaches that integrate physical permeation enhancement through devices generating localised ultrasound, magnetic, or electrical stimulation with advanced formulations show particular promise, which can enhance drug absorption from the gastrointestinal tract (Luo et al. 2021). Recent advances in continuous glucose monitoring technology could complement these approaches by integrating with electronic or mechanically triggered release systems (Kim et al. 2024), creating hybrid closed-loop oral insulin systems that combine biological sensing with precise delivery mechanisms. These increasingly sophisticated integrated systems require multidisciplinary development teams combining expertise in pharmaceutical sciences, materials engineering, and device development. While complexity increases development challenges, these hybrid approaches may provide the comprehensive barrier management necessary for consistent, effective oral insulin delivery.

9.3. Personalised approaches based on patient phenotypes

Personalisation represents a frontier in oral insulin delivery, recognising significant interindividual differences in gastrointestinal physiology (Abuhelwa et al. 2017). Phenotypic stratification based on factors such as intestinal transit time, microbiome composition, and enzyme expression profiles can inform the selection of formulations (Abuhelwa et al. 2017; Zhao et al. 2023; Liao et al. 2024). Genetic factors influencing drug transporter expression, metabolism, and targets may predict response to specific formulation components. Additionally, environmental factors can also influence drug metabolism and clearance, potentially leading to individual variability in drug response (Zhao et al. 2023). Implementation would require the development of cost-effective phenotyping methods and flexible manufacturing approaches capable of producing customised formulations, presenting both technological and logistical challenges that must be balanced against the potential for significantly improved therapeutic outcomes.

9.4. Predictions for technology convergence and disruptive approaches

The next decade is likely to witness disruptive changes as various technologies currently under development converge. Tissue engineering approaches may create implantable intestinal organoids that express insulin in response to glucose (Nie 2023), thereby combining cell therapy with physiological glucose sensing. 3D-printed personalised dosage forms could integrate multiple complementary technologies in geometrically optimised structures tailored to individual patient parameters (Milian-Guimera et al. 2023). Microelectromechanical systems (MEMS) with programmed release profiles could reduce dosage administration (Villarruel Mendoza et al. 2020), transforming oral insulin from a multiple-daily-dose regimen to once-weekly administration. Advances in synthetic biology may enable the construction of autonomous, closed-loop therapeutic cells (Mahameed and Fussenegger 2022), creating complete insulin-delivery ecosystems within the gastrointestinal tract. While speculative, these approaches illustrate the transformative potential of technology convergence in addressing the multifaceted challenges of oral insulin delivery.

10. Conclusion

The quest for effective oral insulin delivery has evolved from simple protective formulations to sophisticated, integrated approaches that combine molecular engineering with advanced delivery technologies. Structural modifications that enhance stability and permeability while preserving biological activity represent a significant advancement, creating insulin molecules that are inherently more suitable for oral delivery. Complementary developments in protective nanocarriers, permeation enhancement strategies, and site-specific release technologies address multiple barriers encountered during gastrointestinal transit.

A critical analysis of the current research landscape reveals both promising advances and persistent limitations. While preclinical studies demonstrate significant improvements in bioavailability and pharmacodynamic profiles, translating these findings to consistent clinical efficacy remains challenging. Variability in absorption, manufacturing complexity, and long-term safety considerations present significant hurdles for commercial development.

Most promising approaches integrate complementary technologies that address multiple barriers simultaneously while maintaining manufacturability and clear regulatory pathways. No single technology is likely to overcome all challenges of oral insulin delivery, making strategic combination and integration of technologies essential for success. The persistent pursuit of the goal, despite setbacks and competing approaches, reflects both the substantial clinical advantages of oral insulin and the significant market opportunity it represents.

Effective oral insulin technology has the potential to transform diabetes management, improving adherence, quality of life, and potentially metabolic outcomes through more physiological delivery patterns. Benefits extend beyond convenience to potentially altering disease progression through earlier insulin initiation and improved adherence to the regimen. Transformative potential justifies continued investment despite substantial challenges.

Realising vision requires concerted effort across multiple domains. Fundamental research must deepen understanding of biological barriers to oral insulin delivery and identify approaches to overcome them. Translational research must bridge the persistent gap between promising laboratory results and clinical efficacy, developing predictive models and clinical trial designs to accelerate development.

Regulatory frameworks must evolve to accommodate these technologies, striking a balance between rigorous safety evaluation and pathways that recognise their transformative potential. Manufacturing technologies must advance to enable cost-effective production at a commercial scale, making these sophisticated therapeutics economically viable for healthcare systems.

Most importantly, effort requires collaboration across traditional boundaries, including those between disciplines, academia and industry, as well as between researchers and patients. The complexity of oral insulin delivery demands the integration of diverse expertise and perspectives, working toward the shared goal of transforming diabetes care through this long-sought but still elusive therapeutic advance.

Funding Statement

This work was supported by the National Key Research and Development Program of China (2021YFA0910500).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

Data sharing is not applicable to this article as no new data were created or analysed in this study.