Engineering a Programmable Delivery Platform for Ultra-Basal Therapeutics: Conceptual Framework and Computational Validation Using Temperature-Responsive Chitosan–PNIPAAm Hydrogels Repurposed from Ocular Drug Delivery

Diabetes mellitus imposes a substantial global burden: the International Diabetes Federation estimates that 589 million adults (20–79 years) were living with diabetes in 2024, with a projected rise to 853 million by 2050. Adherence to insulin therapy is poor—pooled at 55.4% (95% CI: 48.6–62.2) in meta-analytic evidence—contributing to glycaemic variability and an elevated risk of micro- and macrovascular complications. Contemporary basal insulins, including degludec (ultra-long profile, elimination half-life >25 h, duration of action beyond 42 h) and once-weekly icodec, reduce but do not eliminate injection burden.We propose a programmable subcutaneous delivery platform built on chitosan–poly(N-isopropylacrylamide) (PNIPAAm) thermosensitive hydrogels, repurposing an in situ gel-forming chemistry with an established precedent in ophthalmic sustained protein delivery. Below their lower critical solution temperature (LCST ≈ 32°C) these matrices are injectable liquids; on warming to body temperature they undergo rapid in situ gelation, providing sustained, near-zero-order release. We formulate a self-consistent coupled release–pharmacokinetic model in which the three physiological factors propagated in the uncertainty analysis—local temperature, pH, and volume of distribution (Vd)—explicitly enter the model output through Arrhenius-modulated erosion, chitosan-ionisation-modulated permeability, and first-order disposition, respectively.A fully reproducible pure-NumPy/SciPy pipeline (fixed seed) reproduces every reported number. At reference conditions the platform delivers 0.70 U kg⁻¹ day⁻¹, reaching a flat euglycaemic steady state of 15.0 µU mL⁻¹ within ≈28 min, with a modelled initial burst of 1.4% and near-zero within-subject temporal fluctuation. Under full physiological uncertainty the between-subject coefficient of variation of steady-state exposure is 12.8% (95% CI: 11.9–19.5 µU mL⁻¹; 0.04% of realisations outside the 10–25 µU mL⁻¹ window). Variance-based global sensitivity analysis over the same five uncertain inputs (Sobol indices via the Saltelli/Jansen estimator; N = 16,384 base samples, 114,688 evaluations) identifies Vd as the dominant driver of exposure variability (first-order S1 = 0.65), with inter-individual clearance (S1 = 0.16) and temperature (S1 = 0.12) secondary and loading tolerance and pH minor; parameter interactions are negligible (≈0.4% of variance). Thermal responsiveness is thus not the leading source of exposure variability—temperature ranks third, behind the two patient-specific pharmacokinetic parameters. This ranking is robust across defensible coupling assumptions but, as expected, is conditional on the assumed physiological spreads.This interdisciplinary framework bridging ophthalmology and endocrinology is presented as a hypothesis-generating design study. All predictive claims are explicitly falsifiable and require experimental validation; we specify a staged roadmap, success criteria, and refutation conditions. The platform is formally tunable and, in principle, extensible to GLP-1 receptor agonists and monoclonal antibodies, motivating murine pharmacokinetic studies as the essential next step toward investigational new drug (IND)-enabling development.