The specimens attached to the parallelometer were perpendicularly recessed until only the upper coils of the implants were on view.

To produce replicas of the specimens from the experimental group, the crowns’ occlusal was cast using VPS Hydro Putty und VPS Hydro Light Body (Henry Schein Inc., New York, USA) before and after CS. The impression was grouted with Stycast 1266 (Loctite Henkel Electronic Materials, Westerlo, Belgium). The replicas could be scanned by X-ray micro-computed tomography (Micro-CT, Skyscan 1172-100-50, Bruker microCT, Kontich, Belgium). Table 1 shows the scanning parameters of the replicas before and after CS.

The specimens of the experimental group underwent long-term chewing simulation (SD Mechatronik GmbH, Feldkirchen-Westerham, Germany): 1,200,000 cycles, 50 N, and simultaneous thermocycling of 5500 cycles with changing temperatures of 4 and 56 °C for 60 s each. Hydroxyapatite steatite indenters (6.35 mm diameter) were used as antagonists and were replaced for each specimen. The indenter slid 1.5 mm down the inner cliff of the vestibular cusp and 0.5 mm horizontally toward the central fossa at a speed of 60 mm/s each. Five specimens underwent CS at the same time.

The specimens from the control group did not undergo CS. Failure was defined as fracture within the system (crown or implant) or loosening of the crowns during or after CS.

After CS, replicas were produced in the same way as before CS. A commercially available dough, aluminum holder (SEM Specimen Stubs, Agar Scientific, Essex, UK), and foam pellets allowed four replicas to be attached at the same time to the tubes of the Micro-CT (Fig. 2). One single specimen could not be scanned due to a mistake during grouting.

For the generation of 3D data sets from the scans of the Micro-CT, the program NRecon v.1.6.10.4 (Bruker microCT) was employed. The software could reduce ring artifacts by 20 (Ring Artifact Correction). Beam hardening correction was set to 60 %.