An implantable closed-loop insulin delivery device (INsmart device) containing a glucose responsive gel has been developed within the INsmart research group, over a period of 10 years, to mimic pancreas. In this thesis, the reliability and performance capability of the INsmart device was studied for future clinical use. Investigations into the device material compatibility with insulin solution, assessed by monitoring insulin loss and degradant formation over a period of 31 days using RP-HPLC have shown that stainless steel and titanium are the most compatible materials. Polycarbonate contributes to insulin loss after 11 days, resin might not be the best material and polyurethane is the least compatible for future device designs. To study insulin delivery mechanism and kinetics from the device, fluorescently labelled human insulin (FITC-insulin) was synthesised and characterised using RP-HPLC and MS, to produce a product with predominantly di-labelled conjugate (>75%) with no unreacted FITC or native insulin. Clinically used insulin analogues were also fluorescently labelled to produce predominantly di-labelled FITC-insulin conjugate with potential future biological and in vitro applications. The drug release mechanism from the glucose sensitive gel held in the INsmart device, studied using fluorescein sodium was determined as a Fickian diffusion controlled release mechanism. The diffusion coefficient (D) for FITC-insulin in the non-polymerised dex2M-conA gel (NP gel) determined using mathematical models, QSS and TL slope methods was 1.05 ± 0.02 x 10-11 m2/s and in the cross-linked dex500MA-conAMA gel (CL gel) was 0.75 ± 0.06 x 10-11 m2/s. In response to physiologically relevant glucose triggers in the NP gel, the diffusivity of FITC-insulin increases with increasing glucose concentrations, showing a second order polynomial fit, device thus showing glucose sensitivity and graded response, mimicking pancreas. Rheological measurements further confirmed the gel glucose responsiveness demonstrated by a third order polynomial fit between FITC-insulin D and the NP complex viscosity in response to increasing glucose concentration. The knowledge of FITC-insulin diffusion kinetics in the gel has aided in making some theoretical predictions for the capability and performance of the INsmart device. Alternate device geometry and design optimisation is also explored.