2.1

Specimen fabrication

A CAD-based three-dimensional (3D) model of a mandibular first molar was designed (SolidWorks 2013; SolidWorks Corporation, Concord, MA, USA). A uniform reduction on the occlusal surface was performed to produce by boolean operations, three CAD models of restorations with different thicknesses (0.5, 1.0, and 1.5 mm). For the milling parameters of the restorations, a marginal gap of 0 μm and an internal gap of 50 μm were established between the tooth and the restoration. The restorations were milled (Ceramill Motion 2; Amann Girrbach, Koblach, Austria) in two different materials (n = 21/material): resin nanoceramic (Lava Ultimate; 3 M Oral Care, St Paul, MN, USA) and polymer-infiltrated ceramic network (Enamic; Vita Zahnfabrik, Bad Sackingen, Germany), totaling 126 occlusal veneers. Finishing and polishing were performed according to the manufacturers’ recommendations. The CAD-modeled preparation replicas, simulating advanced wear of the occlusal surfaces were milled in fiber-reinforced epoxy resin (n = 21/group) (G10; Protec, São Paulo, SP, Brazil) due to its similarity in terms of adhesion (9 MPa) and elastic modulus (18.6 GPa) to hydrated dentin (6.5 MPa and 18 GPa, respectively) [ , , , ].

2.2

Cementation procedure

All OV were cemented to aged (stored in distilled and deionized water at 37 °C for 21 days) [ , ] fiber-reinforced epoxy resin preparation preparations using a dual-cure resin cement (RelyX Ultimate; 3 M Oral Care, St Paul, MN, USA). Epoxy resin preparations were etched with 10% hydrofluoric acid (Condac porcelana; FGM Produtos Odontológicos Ltda, Joinville, SC, Brazil) for 120 s [ ]. The pretreatment of the OV followed the manufacturers’ instructions: RNC was sandblasted with aluminum oxide 50 μm grain size (Óxido de Alumínio, Bio-Art, São Carlos, SP, Brazil), at a distance of ∼1 cm for 5 s at 2 bar, and PICN was etched using 5% hydrofluoridric acid (Condac porcelana; FGM Produtos Odontológicos Ltda, Joinville, SC, Brazil) for 60 s [ , ]. Then, both epoxy resin preparation and OV were cleaned in an ultrasonic bath with distilled water for 2 min. The adhesive (Single Bond Universal; 3 M Oral Care, St Paul, MN, USA) was applied on both ceramic and epoxy resin preparation for 20 s, and then resin cement was applied on the internal OV surface. After the restoration-to-preparation setting, the assembly was maintained under a load of 10 N [ , ] to standardize the cementation process thus promoting a uniform cement spreading and thickness within the intaglio surfaces. The cement excess was removed with a micro-brush. The margins were light-cured (radii-cal; SDI Limited Bayswater, Victoria, Australia) for 20 s on each surface. After that, the set was vertically embedded in acrylic resin (Self-curing acrylic resin, Artigos Odontológicos Clássico Ltda, Campo Limpo Paulista, SP, Brazil).

2.3

Step stress accelerated life testing (SSALT)

Three specimens per group were subjected to the single load-to-failure (SLF) test in a universal testing machine (Instron 4411, Corona, CA, USA) with a load applied axially through a tungsten carbide indenter on the central fossa of the occlusal surface of the OV using a 5 kN load cell at a loading rate of 1 mm/min. Based on the SLF data, three different step-stress profiles, mild, moderate, and aggressive, were designed.

The remaining specimens (n = 18/group) were allocated to the three fatigue profiles: mild (n = 9), moderate (n = 6) and aggressive (n = 3), following the distribution ratio of 3:2:1, respectively, considering that the accuracy of the statistic prediction is inversely proportional to its cycling length [ ].

The SSALT test was performed in an electrodynamic fatigue testing machine (ElectroPlus E3000 Linear-Torsion Test InstrumentTM; Instron, Norwood, GA, USA) in the presence of distilled water at 20 Hz [ ]. In each cycle, the tungsten carbide indenter contacted 0.5 mm lingual to distobuccal cusp tip of the occlusal veneer in an axial direction. The test was considered finished when samples failed (considered as chipping or delamination) or when suspended for no event being detected throughout the designed cycles and 1500 N maximum load.

Based on the SSALT failure distribution, the use level probability Weibull curves were calculated and plotted (Synthesis 9, Alta Pro; Reliasoft, Tucson, AZ, USA) using the Weibull distribution and the inverse power law life-stress relationship for damage accumulation. The reliability was calculated for completion of a mission of 100,000 cycles at 200, 300, 400, 500 and 600 N, and the differences between groups were identified based on the non-overlap of the 90% two-sided confidence interval (CI). This analysis provides the beta (β) value, which describes the behavior of the failure rate over time (β <1: values indicate that the failure rate has decreased over time, β ∼ 1: failure rate does not vary over time, and β> 1: means that the failure rate has increased over time) [ ]. If the calculated beta value was <1 for any group, then a Weibull 2-Parameter Contour plot (Weibull modulus – m vs. characteristic strength eta – η) was calculated using the final load to failure data of all groups.

2.4

Fractographic analysis

Representative fractured specimens were sputter-coated with gold (Emitech K650; Emitech Products Inc, Houston, TX, USA) for scanning electron microscopy (SEM) analysis (JSM-5600LV, Jeol, Boston, MA, USA) and fractographic evaluation.

2.5

Finite element analysis

The CAD models of preparation replicas and OV were imported into the finite element analysis software (Ansys Workbench 15.0; Ansys, Inc, Canonsburg, PA, USA). A convergence test of 10% [ ] was performed to guarantee that no further mesh refinement was required to achieve accurate numerical results of the FEA model and, subsequently, the element size was determined (0.5 mm). The three-dimensional solid models meshed with tetrahedral and hexahedral elements were automatically generated. The number of elements and nodes varied depending on the model (33,744 to 36,853 elements and 81,738 to 88,153 nodes). All structures were considered isotropic, homogeneous, and linearly elastic [ ]. To simulate the influence of different structures, physical and mechanical properties (modulus of elasticity and Poisson’s ratio) of each structure of the model were obtained from the literature ( Table 1 ) and introduced on the software. All contacts were considered perfectly bonded. An axial load of 600 N was applied as a normal surface pressure load on the distobuccal cusp tip of the occlusal veneer models. The stress distribution was evaluated according to the criteria of maximum principal stress (σ _max ) on the OV restoration and maximum shear stress (τ _max ) on the cement layer. A scale of values was determined to compare stress patterns in the models.