The aim of the present retrospective study was to evaluate longitudinal alterations in radiographic bone topography at proximal sites of three-unit implant-supported fixed partial prostheses during the first 3 years after bridge installation, in relation to vertical and horizontal inter-unit distances. The subjects were partially dentate patients who had received implant-supported fixed partial prostheses during the year 1995 at the Brånemark Clinic, Göteborg, Sweden. For inclusion in the study, the patient had to have a three-unit bridge construction supported by three implants in the posterior area of the jaw. Twenty-eight patients having 35 screw-retained prostheses on Brånemark standard implants fulfilled the inclusion criteria. Radiographs obtained at bridge installation and at 1- and 3-year follow-ups were assessed for implant positions, contact point level, bone level at implants and adjacent tooth and mid-proximal bone crest level. The data were analysed with respect to two proximal units: tooth/implant units (n=35) and implant/implant units (n=70). Multiple regression analyses were used to evaluate the influence of various factors on the peri-implant and periodontal bone level changes during the 3 years of follow-up. At the tooth/implant units, the mean bone loss over the 3 years was 0.5 mm at the implant and 0.4 mm at the tooth. Multiple regression analysis failed to identify significant explanatory factors for the peri-implant/periodontal bone level changes at the tooth/implant units (R2=0.28). At the implant/implant units, the peri-implant bone loss was 0.6-0.7 mm and was significantly influenced by the vertical inter-implant distance (P<0.01), the difference in bone level at baseline between two neighbouring implants (P<0.001) and the bone level changes at the opposed implant surface (P<0.001) (R2=0.49). Furthermore, the magnitude of apical displacement of the inter-implant bone crest level during the 3 years of follow-up was negatively associated with the horizontal inter-implant distance (P<0.05). The results of the study demonstrated that both vertical and horizontal differences in implant positions might influence bone alterations in the inter-implant area during the first 3 years of loading, while the data failed to show corresponding relationships for the bone changes at the proximal area between the implant and the neighbouring tooth.