Artificial neural networks for bilateral prediction of formulation parameters and drug release profiles from cochlear implant coatings fabricated as porous monolithic devices based on silicone rubber.

OBJECTIVES

The coating of cochlear implants for topical delivery of drugs, for example, corticosteroids, or antibiotics is a novel approach to manage post-surgical complications associated with cochlear implantation surgery like inflammation or infections. Many variables, including formulation parameters, can be changed to modulate the amount and duration of drug release from these devices. Mathematical modeling of drug release profile from a delivery system may be helpful to accelerate formulations in a more cost-efficient way. To attain specific in vitro drug release characteristics, a model should be capable to provide good estimates on the initial formulation parameters, for example, composition, geometry and drug loading vice versa. Here, artificial neural networks (ANNs) are used to predict dexamethasone (DEX) release profile and formulation parameters, bilaterally, from cochlear implant coatings designed as porous, monolithic silicone rubber-based matrices.

METHODS

The devices were fabricated as monolithic dispersions of DEX in a silicone rubber matrix containing porogens. A newly developed mathematical function was fitted on the experimental DEX release curves, and the function coefficients were fed into the network as input variables to simulate drug release profile from the porous devices. Formulation variables consisted of drug loading percentage (0.05-0.5% w/w), porogen type (dextran (dext) or sodium chloride particles) and porogen content (5-40% w/w). The ANN was also examined to determine optimal levels of the formulation parameters to provide a specifically desired drug release profile.

KEY FINDINGS

The results showed that DEX release profile from porous cochlear implant devices can be modelled accurately and precisely using ANN in order to predict optimal levels for the formulation parameters to provide a specific drug release profile vice versa.

CONCLUSIONS

The developed ANNs were used to achieve shorter formulation development process, and to provide tailor-made drug delivery regimens. ANNs were also successfully simulated non-linear relationships present between the initial formulation variable(s) and predict the subsequent drug release patterns.