OBJECTIVE: To identify risk factors for failure and bone loss of implants in a large study sample on the basis of multivariate analyses. MATERIALS AND METHODS: Patient files of all patients referred for implant treatment from November 2004 to December 2007 were scrutinized, and information on implant- and patient-related factors was collected. The study sample in this retrospective cohort study consisted of both partially dentate and fully edentulous patients referred for various indications. The only inclusion criterion was a follow-up of at least 2 years. Implant survival and bone loss were assessed by an external investigator (SV) comparing digital periapical radiographs taken during recall visits with the postoperative ones. Univariate and multivariate tests were adopted to identify possible risk indicators for implant failure and peri-implant bone loss. RESULTS: Twenty-one of 1,320 (1.6%) implants were lost in 19 of 376 (5.1%) patients (210 female, 166 male; mean age 56, range 17-82) after a mean follow-up of 32 months (range 24-62). Based on multivariate analysis, only smoking (p = .001) and recall compliance (p = .010) had a significant influence on implant failure, with smokers more prone to failure. The overall mean bone loss was 0.36 mm (SD 0.68, range 0.00-7.10). Smoking (p = .001) and jaw of treatment (p = .001) affected peri-implant bone loss. More peri-implant bone loss was observed in smokers and in the maxilla. A clear discrepancy was found between univariate and multivariate analysis with regard to identification of risk factors. CONCLUSION: Multivariate analysis demonstrated that implant-related factors did not affect the clinical outcome, but smoking was identified as a predictor for implant failure. Predictors for peri-implant bone loss were smoking and jaw of treatment.