The purpose of this study was to investigate three relevant aspects of intramedullary nailing in terms of their effect on the occurrence of local infection. In an infection model on the rabbit tibia, the following were compared: a hollow and a solid nail (Experiment I), a reamed with an unreamed technique (Experiment II), and a steel with a titanium nail (Experiment III). In order to minimize the number of animals required, a grouped sequential procedure combined with an "up and down" dosage technique was applied. Microbiological evaluation was both qualitative and quantitative. The results in Experiment 1 (n = 44) indicated an infection rate for the hollow nail (59%) almost double that of the solid nail (27%) (P < or = 0.05). Experiment II (n = 44) produced an infection rate of 50% for the unreamed technique compared to 64% for the reamed technique, a difference which, on the basis of the number of bacteria present, was also statistically significant (P < or = 0.05). In Experiment III (n = 44) an infection rate of 82% was recorded for the steel nail compared to 59% for the titanium nail (P < or = 0.05). The results of the three experiments are only partially comparable with each other because of the grouped sequential procedure and the different inocula used. Nonetheless it would seem that the dead space inherent in the design of the hollow nail represents a considerable risk with regard to the occurrence of local infection. Reaming the medullary cavity with the attendant reduction in local vascularity and necrosis and the lesser biocompatibility of steel compared to titanium may be additional risk factors which should not be overlooked.