A multilevel analysis of factors affecting the longevity of fixed partial dentures, retainers and abutments

There is a methodological problem in analysis of data belonging to different hierarchical levels, e.g. patient, arch, and tooth. A method for multilevel modelling (MLM) that resolves this problem is available. This method which is presented in this article, was used to reanalyse previously published long-term treatment results concerning the longevity of fixed partial dentures, retainers and abutments. The main results were consistent, indicating that no great faults were committed when using the conventional logistic regression method. There were, however, differences when calculating the combined risks for various situations. The conventional method was found to over- and underestimate probabilities of survival in different cases, compared to the MLM method. Especially when the most negative factors were combined, the conventional method had a tendency to underestimate the risk. For example, if an abutment had the combination 'endodontically treated'; 'FPD placed by 'another dentist''; 'distal abutment'; and 'marginal bone loss > 50%'; the probability of the tooth remaining after 18-23 years was only 20% according to the MLM method but 35% according to the conventional method. It was concluded that MLM is the method of choice for many situations in dental research where data belonging to different levels are to be analysed.