Sample size calculation
The sample size was calculated using unpublished pilot data (n = 5) and the two-sided Welch's t-test for unequal variance at a significance level of P ≤ 0.05 and a power of 0.9 (nQuery Advisor version 7, Statistical Solutions, Cork, Ireland). The sample size was evaluated as n = 15 for each group. Considering some dropouts and a deviation of normality assumptions a sample size of n = 18 was used in the study.
Sample preparation
Extracted human maxillary lateral incisors with comparable dimensions were disinfected in accordance with the university’s policy. The teeth were cleaned with scalers and stored in 1% chloramine trihydrate solution. Crowns were removed using a diamond saw at slow speed (WOCO 50/Med, Conrad, Clausthal-Zellerfeld, Germany) to obtain roots of 13 mm length. A stereomicroscope (Stemi SV8, Zeiss, Oberkochen, Germany) at 12 x magnification was used to exclude any pre-existing dentin defects. After numbering the teeth, the cross sections of the roots were measured at the level of the cutting surface in the mesio-distal and bucco-palatal direction with a digital caliper (Garant, Hoffmann, Munich, Germany). The area of the ellipsed root cross section (A) was calculated according to: A = p/4 x a x b (where a and b were the mesio-distal and bucco-palatal dimension in mm). Roots of extreme size were excluded. The remaining samples were randomly distributed into five groups of 18 roots each according to a random numbers table. To imitate a human periodontium with uniform stress distribution, the roots were wrapped in one layer of latex rubber milk (Suter Kunststoffe, Jegenstorf, Switzerland) with a thickness of approximately 250 µm and embedded in acrylic resin (Technovit 4071, Heraeus Kulzer, Hanau, Germany) with the coronal third (4 mm) being exposed.
In the control group, the roots were left untreated. In the other groups, the canals were instrumented using single-length technique with nickel-titanium rotary files (EasyShape, Brasseler Komet, Lemgo, Germany) up to file size 40 and 0.04 taper. The working length was set to 12 mm. During instrumentation, canals were irrigated with 5 ml of 3% sodium hypochlorite (NaOCl) and 15% ethylenediaminetetraacetic acid (EDTA) solutions (Glyde File Prep, Dentsply Sirona, Ballaigues, Switzerland). After a flush with 5 ml distilled water, the canals were dried with paper points and obturated with the matched EasyShape gutta percha cone and EasySeal resin-based sealer (Brasseler Komet) using the single-cone technique.
Post placement
Drills supplied by the post manufacturer (Brasseler Komet) were used to prepare post spaces of 8 mm depth and size ISO 70. Gutta percha was removed with the pilot drill 183LB leaving 4 mm of the root filling in the apical portion. The root canals were enlarged with the reamer 196 and roughened with the form congruent diamond instrument 196D. After a flush with 5 ml distilled water, the post spaces were dried with paper points. Coated FRC posts (ER DentinPost Coated) and sandblasted titanium posts (ER Kopf-Stift) (TIT) in the same cylindroconical shape and size (ISO 70) were shortened from the coronal end to 7 mm with diamond drills under water cooling so that they could be placed into the post canals 1 mm below the cutting surface of each root. All posts were cleaned with 70% alcohol. For each post material, half of the samples were cemented (C), while the other half was adhesively luted (A). For cementation, zinc phosphate (Richter & Hoffmann, Berlin, Germany) was mixed in a creamy consistency in relation 1.5 g zinc oxide powder to 1 g phosphoric acid on a cooled glass plate and applied to the post surface. After post insertion, excess cement material was removed with brushs. For adhesive luting, the dual-curing luting system of the posts were applied following manufacturer's instructions. Post spaces were etched for 20 s with 37% phosphoric acid gel, rinsed with water spray, and dried with paper points. The etch-and-rinse adhesive DentinBond was applied to the root dentin and light-cured for 10 s. The composite DentinBuild was applied to the post surface directly from the automixing syringe. After post insertion, excess composite was removed with brushs. Light curing was performed for 20 s at 1000 mW/cm2 light intensity (Astralis 10, Ivoclar Vivadent, Schaan, Liechtenstein). The cervical 1 mm of the post space was temporarily filled (Cavit, 3M Espe, Seefeld, Germany).
Chewing simulation and VRF testing
After water storage for 24 h at 37°C, the samples were subjected to 1500 thermocycles in distilled water at 5-55°C with a dwelling time of 20 s in each bath (Haake W15, Willytec, Gräfelfing, Germany). Mechanical loading was performed according to the staircase method starting at a load of 25 N at an angle of 10° to the axial direction of the roots in a chewing simulator (Standard 2002, Willytec). Every 20,000 cycles at a frequency of 2 Hz, the load was increased in increments of 25 N until 120,000 cycles were reached. The unfilled cervical 1 mm of the post space ensured that the force applied by the coneshaped metal antagonist with an angle of 90° was transmitted to the root dentin rather than the post. The diameter of the truncated cone was dimensioned in such a way that the cone tip fitted exactly into the coronal unfilled post space.
After chewing simulation, the external root surfaces were examined under the microscope using a cold light source (Stemi SV8, Zeiss). Because of the latex milk, the roots could easily be removed from the acrylic blocks. Pictures were taken with a digital camera at 12-100 x magnification (3CCD Color Video Camera, Sony, Tokyo, Japan). Crack formation was analyzed per root section (cervical, middle, apical) as follows: (a) no defect, (b) craze line, (c) vertical crack, and (d) horizontal crack. Representative images of the dentin defects are shown in Figure 1. After microscopic analysis, the roots were returned to the acrylic blocks and subjected to VRF testing. The same antagonist as used for the chewing simulator was attached to the load cell of a universal testing machine (Zwicki 1120, Zwick, Ulm, Germany). The samples were loaded until fracture with a crosshead speed of 1 mm/min. The fracture load (N) was recorded when the force in the load-strain curve decreased by 30%.
Statistics
Statistical analysis was performed with IBM SPSS version 19 for Windows (Chicago, IL, USA). The significance level was set in advance at P ≤ 0.05. Differences in crack formation were evaluated by Fisher’s exact test. After confirming that the VRF resistance (P = 0.244) and the root cross sectional area (P = 0.651) met the assumptions of normality as indicated by the Kolmogorov-Smirnov test, differences between the means of the groups were compared with one-way ANOVA followed by post hoc Tukey’s test.