The positive predictive value of radiographic identification of unstable Branemark implants was evaluated. Based on approximately 2,000 patients with implants followed during a 3-year period, a total of 84 patients were radiographically detected to have suspected failing implants. A total of 481 implants had been placed in these patients, but because of the study design and some patient dropouts, the evaluation was carried out in only 79 of the identified patients, who had 413 implants. Calculations were performed based on the number of occasions the implants were examined (n = 482) because some patients were assessed on more than one occasion. Results indicated a high positive predictive value (83%) to ra diographically identify failing implants, and only 5% were clinically found to be failing without having been detected radiographically. In nine jaws (11%), the fixed prost hetic restorations were unnecessarily detached because of inaccurate radiographic diagnosis. The study indicated that when examining patients on a long-term routine basis. The timing for these assessments should be annual during the first 3 years of follow up for the unexperienced clinician and may thereafter be individualized with more clinical experience.