The objective of this work was to formulate new oral insulin-loaded nanoparticules using the response surface methodology. The insulin nanoparticles were prepared by a water-in-oil-in-water emulsification and evaporation method. The polymers used for the encapsulation were blends of biodegradable poly-epsilon-caprolactone (PCL) and of positively-charged, nonbiodegradable polymer (Eudragis RS). A central composite design has been built to investigate the effects of three controlled variables: ratio of polymers (PCL/RS ratio), volume, and pH of the aqueous solution of polyvinyl alcohol. The nanoparticles were characterized by measuring the amount of entrapped insulin, the particle size, the polydispersity of the obtained particles, the zeta potential, and the amount of insulin released after 7 hours. A second-order model was evaluated by multiple regression and was statistically tested for each of the studied controlled variable. The obtained polynomials proved efficient to localize an optimal operating area highlighted by the use of three-dimensional response surfaces and their corresponding isoresponse curves. An interesting formulation given by the models was selected, prepared, and evaluated. The corresponding quantity of entrapped insulin was 25 IU per 100 mg of polymer, and the particle size was 350 nm with a polydispersity of 0.21. The quantity of released insulin was 4.8 IU per 100 mg of polymer after 7 hours and the zeta potential was +44 mV. All these collected values were in perfect accordance with values estimated by the models. Finally, the results suggested that PCL/RS 50/50 nanoparticles might represent a promising formulation for oral delivery of insulin.