Method:
Ninety monolithic crowns made of lithium disilicate ceramics (IPS e.max press, Ivoclar, Liechtenstein) were made with a standardized procedure (IPS e.max Ceram). Three groups with 30 crowns each were built (untouched, ground, ground and polished) and seated (Vivaglass Cem, Ivoclar, Liechtenstein) on steel dies. The crowns were loaded with 2.4 million cycles (50N, 20Hz, thermocycling 5,000 times (5-55°C)). The actual effective force for each crown was constantly measured. A universal testing machine (Z010, Zwick/Roell, Ulm, Germany) was used for loading to failure. Statistical analysis was performed with univariate analysis of variance (IBM SPSS 20.0, USA; level of significance α=0.05).
Result:
The mean failure load was 2.498N (SD 332) for untouched restorations, 2.681N (SD 632) for ground and polished restorations and 3.027N (SD 680) for ground restorations. The latter differed significantly from the other two groups (p=0,002). After censoring the values of all specimens for which the effective mean cyclic load deviated more than 10% from the targeted load of 50N, the mean differences between the three groups were lowered accompanied by smaller 95% confidence intervals. A significant difference between the three groups was no longer found (p=0,054).
Conclusion:
Less variation across the results followed from censoring all specimens with a deviating effective load during dynamic cyclic loading and thermocycling before further analysis. In view of these results the influence of the chewing simulation itself ceramic restoration failure should be further investigated.