Process-based approach to modeling recurrent-event data explicated on the basis of occurrences of tooth losses in two different prosthetic treatment concepts.

<bold>Background: </bold>In studies comparing different prosthetic treatment concepts the repeated loss of teeth was chosen as the primary outcome. The resulting data appear to represent a data structure of recurrent events. However, the application of an existing method for recurrent events is far from straightforward. Often only the first event or the final state is analyzed using Kaplan-Meier survival statistics, thereby giving a great deal of information away.<bold>Methods: </bold>The paper presents a strategy for the analysis of recurrent data using a previously published study on the influence of different prosthetic treatment concepts for the shortened dental arch on tooth loss. A method based on cumulative sample history functions of recurrent events was adjusted for tooth loss. The shapes of these cumulative functions suggest a time dependency of the recurrence rate. To keep the model as simple as possible, a tripartite Poisson process (which assumes piecewise time-independent rates) was fitted to the cumulative mean functions stratified by treatment.<bold>Results: </bold>Within the middle interval of the three-phasic process, the treatment effects differ significantly, which is interpreted as a delay of tooth loss due to the use of one type of prosthesis (fixed) compared with the other (removable).<bold>Conclusions: </bold>An analysis based on cumulative history functions is based on process, therefore, temporally changing characteristics are better captured than in methods for survival analyses. The presented approach offers useful new insight into the temporal behavior of ongoing tooth loss after prosthetic treatment.<bold>Trial Registration: </bold>The trial has been registered at controlled-trials.com under ISRCTN97265367 (registration date 4 April 2008). [ABSTRACT FROM AUTHOR]