Risk estimates of root-filling quality on apical periodontitis, multilevel analyses

Objective: The aim of this study was to explore possible interdependence among root-filled teeth within randomly selected adult individuals with regard to the relationship between the quality of root-fillings (RF) and presence of apical periodontitis (AP).

Material: Randomly selected individuals 20-70 years of age were included from four crossectional studies.

Methods: AP was recorded using the PeriApical Index (PAI) and length and seal of rootfillings were measured per mm from the apex and lateral or apical voids, respectively.Binary logistic regression models were applied with the independent variable ‘Quality of root-filling' coded as 0= adequate and 1= inadequate root-filling. The dependent variable used was the PAI-score dichotomized into 0= no apical periodontitis (PAI score = 1-2) and 1= apical periodontitis (PAI-score 3-5). The observational unit in the analyses were teeth. However, due to intraindividual correlations among two or more root-filled teeth/apical periodontitis, i.e a significant portion of patients contributed two or more root-filled teeth in the statistical analyses, multilevel logistic analyses were performed in order to account for such correlations. The statistical computer program used in the analyses were SPSS version 14 and MLwin version 2.2.

Results: Model with teeth as unit showed an OR of 4.59, indicating that an inadequate root-filled tooth had a statistically significant increased risk of having an apical periodontitis, i.e a more than 4 times higher risk as compared to a tooth with an adequate root-filling. Multilevel differs from the previous model, specifically with regard to the OR and estimated standard errors and confidence intervals. However, the differences are not seemingly large. The OR was 4.79 and the variance for the random intercept 0.173. A statistical inference test for the intercept variance was performed with a Wald test, giving a test-statistic of 307.00 which evaluated with the chi-square distribution resulted in p<0.0001. This means that there was a significant difference between patients.

Conclusions: Our results show that there was a small but significant part of variance that was attributed to patients. Moreover, analysis revealed a statistically significant difference between patients.