Prosthetic valve endocarditis may be reduced by the local delivery of antibacterial proteins from the Dacron sewing ring of a prosthetic heart valve. Dacron discs were treated with a carbon dioxide gas plasma to improve the hydrophilicity and thereby enabling homogeneous impregnation with gelatin type B. The gelatin samples were cross-linked to different degrees using various amounts of water-soluble carbodiimide (EDC) and N-hydroxysuccinimide (NHS). Lysozyme, a model protein for antibacterial proteins, was loaded into (non)-cross-linked gelatin gels incorporated in Dacron, or adsorbed onto non-treated and gas plasma-treated Dacron. The in vivo lysozyme release was measured after subcutaneous implantation of lysozyme-loaded samples in rats. The lysozyme content of the samples, and the lysozyme level of the surrounding tissue were determined at different explantation times (ranging from 6 h up to 1 week). For cross-linked gelatin gels, the lysozyme tissue level was elevated up to 2 days after implantation. In vitro release was measured using agarose medium or phosphate buffer. Lysozyme release in buffer solution under sink conditions was in good agreement with the in vivo lysozyme release profiles, and therefore considered a good model to describe in vivo release characteristics. The release was modelled with a solution of Fick's second law of diffusion using the appropriate boundary conditions. In this way the lysozyme concentration in the gel and the surrounding tissue as a function of time and distance was obtained. The presence of cross-linked gelatin in Dacron did lead to an increased uptake of lysozyme and a delayed release during 30 h after implantation, whereas a burst release took place from Dacron, gas plasma-treated Dacron, or Dacron containing non-cross-linked gelatin.