The aim of the study was to predict the marginal bone level at a 5-year follow-up based on the information available from an initial radiographic examination and to evaluate the precision of the prediction by comparing the predicted bone levels with those actually observed at the follow-up. In 1997, 616 randomly selected dentate individuals underwent a full-mouth radiographic survey. In 2003, 473 of those individuals (77%) participated in a second radiographic examination. Marginal bone level, caries lesions, fillings, crowns, root fillings and periapical status were recorded on all teeth. On the basis of data available from the first examination, a linear mixed model regression analysis with the tooth as the unit of analysis was used to predict the marginal bone level 5 years later. Number of teeth, smoking, and also presence of apical periodontitis and crowns were associated with bone loss and could be used as predictors of future marginal bone level. The analysis of all teeth showed that the number of tooth- and person-specific factors at the first examination influenced the prediction of the marginal bone level at the 5-year follow-up examination. However, the performance of the combined prediction model was less satisfactory.