The purpose of this study was to use multivariate analysis to evaluate contributing factors affecting deep infection and nonunion of open femoral fractures treated with locked intramedullary nailing (IMN). We examined 89 open femoral fractures (88 patients) treated with immediate or delayed locked IMN in static fashion at the Kitasato University Hospital from 1988 to 2001. Multiple regression models were derived to determine predictors of deep infection and nonunion. The following predictive variables of deep infection were selected for analysis: age, sex, Gustilo type (I+II or III), fracture grade by AO type (A or B+C), fracture site (proximal site+distal site or middle site), timing or method of IMN, reamed or unreamed nailing (R versus UR), debridement time (< or =6 h or >6 h), existence of polytrauma (ISS<18 or ISS> or =18), and existence of floating knee injury (+ or -). The predictive variables of nonunion selected for analysis were the same as those for deep infection, with the addition of deep infection (+ or -). Five fractures (5.6%) developed deep infections: one Gustilo type II and four type III. Multivariate analysis revealed that only Gustilo type significantly correlated with occurrence of deep infection (p<0.05). Nonunion occurred in 12 fractures (14.1%). Multivariate analysis revealed that only fracture grade by AO type significantly correlated with occurrence of nonunion (p<0.02).