Incidence of fractures and lifetime predictions of all-ceramic crown systems using censored data

To determine Kaplan-Meier survival estimates and Weibull lifetime predictions for four all-ceramic crown systems from long-term data (> 5 yrs).

Single unit crowns of Cerestore (n = 30), Dicor (n = 30), Hi-Ceram (n = 22) and In-Ceram (n = 68) were placed in 95 patients treated in a university clinic. They were cemented using glass-ionomer (GI) for Cerestore, zinc phosphate (ZP) for Dicor, and 75% ZP (n = 51), 20% GI (n = 13) and 5% resin-based cement (n = 4) for In-Ceram crowns. The follow-up times were 8 yrs for Cerestore, 7 yrs for Dicor, 6 yrs for Hi-Ceram and 5 yrs for In-Ceram. The statistical analyses were based on censored data sets. A progressively censored Weibull distribution allowing for lifetime predictions beyond the actual observation time was used as well as the Kaplan-Meier Survival Product Limit Estimate (PLE), which provides survival estimates up to the maximum time of follow-up.

Cerestore, Dicor and Hi-Ceram demonstrated molar fractures in the first 2 yrs, whereas for In-Ceram these occurred during the third and fourth year of the study. The Kaplan-Meier Survival PLE was 69% for Cerestore at 8 yrs, 86% for Dicor at 7 yrs, 81% for Hi-Ceram at 6 yrs, 92% for In-Ceram at 5 yrs. The predicted Weibull characteristic time T0 (time at which 63% of the restorations would have failed) was 23 yrs for Cerestore, 34 yrs for Dicor, 31 yrs for Hi-Ceram, and 16 yrs for In-Ceram. However, when using data sets arbitrarily limited to the three first years of follow-up, T0 decreased significantly for Cerestore, Dicor and Hi-Ceram due to the number of early fractures. Such Weibull lifetime data illustrate the risk of predicting long-term (> 5 yrs) survival using short-term (< or = 3 yrs) data on ceramic restorations, which exhibit fracturing in the initial years.