Tretinoin-loaded niosomes were prepared from polyoxyethylene (4) lauryl ether, sorbitan esters and a commercial mixture of octyl/decyl polyglucosides, in the presence of cholesterol and dicetyl phosphate. Liposomes made of hydrogenated and non-hydrogenated phosphatidylcholine were also prepared as a comparison reference. A study was made of the influence of vesicle composition and preparation method on the vesicle structure (MLV, LUV, SUV), size distribution, entrapment efficiency and in vitro release of incorporated tretinoin. Results showed that in the presence of cholesterol all the amphiphiles used were able to form stable vesicle dispersions with or without tretinoin. Vesicle sizes were dependent on the preparation method, bilayer composition and drug load. Multilamellar (MLV) vesicles were larger than extruded (LUV) and sonicated (SUV) vesicles while drug-loaded vesicles were generally smaller than empty ones. Entrapment efficiencies of tretinoin were always very high especially for multilamellar (91-99%) and extruded (88-98%) vesicles. The in vitro release of tretinoin from the prepared vesicular formulations was studied using the vertical Franz diffusion cells. The rate of drug release through a Silastic membrane from a liposomal and niosomal tretinoin dispersion was generally faster than from a tretinoin solution. Release data showed that tretinoin delivery is mainly affected by the vesicular structure and that tretinoin delivery increased from MLVs to LUVs to SUVs.