Two hundred seventy-one tibial nonunions of average duration of 23.5 months (range, 9-69 months) were treated with direct current (167 patients), capacitive coupled electrical stimulation (56 patients), or bone graft surgery (48 patients). Logistic regression analysis was used to compare heal rates among the 3 treatment methods, to identify risk factors adversely affecting the heal rate, and to predict the probability of successful healing of a nonunion of any given risk profile treated with each of the 3 forms of therapy. Seven risk factors were identified: duration of nonunion, prior bone graft surgery, prior electrical treatment, open fracture, osteomyelitis, comminuted or oblique fracture, and atrophic nonunion. When no risk factors were present, there were no significant differences among the 3 treatment methods. As progressively more risk factors were present, the predicted heal rates decreased significantly regardless of the treatment method. Some differences among the treatment groups did appear in the heal rates: bone graft surgery yielded a worse heal rate when there was a previous bone graft failure, and capacitive coupling had a worse heal rate in the presence of an atrophic nonunion.