To assess the biomechanical strength of the bone-implant interface, the following parameters were assessed: (a) maximum removal torque (N cm) (primary outcome measure), obtained during the unscrewing process (primary outcome measure); (b) connection stiffness (N cm/rad), corresponding to the ratio between removal torque and angular displacement (secondary outcome measure); and (c) removal energy (× 10–2 J), corresponding to the energy (workload) necessary to completely unscrew the implant (secondary outcome measure).

The removal torque test was conducted on a Shimadzu universal testing machine.^{Footnote 4} This equipment was adapted in order to determine the referred properties (Fig. 2a, b). A horizontal shaft, supported by two ball bearings, with Allen keys socket on one end and a rotation sensor on the other end, was connected with a steel string to the mobile span of the Shimadzu universal testing machine, in such a way that the linear motion was converted to a rotational motion. The dog’s tibia bone block containing the implant was placed in alignment and inside the Allen keys socket and fixed with an adjustable clamp, in order not to rotate during the test. The upper span speed, at which the string was attached, was adjusted to produce a shaft rotation speed of 0.005 rad/s. During the test, both torque (N cm) and angular displacement (rad) were acquired using a sampling rate of 10 samples/s (Fig. 3a, b). In order to calculate the connection stiffness (N cm/rad), the tangent method was applied to the data after obtaining the linear correlation coefficient (R²) compared to the secant method, revealing the absence of mathematical discrepancy between the application of both methods (Fig. 4).

Calibration of one blinded examiner (R.N.R.J) and repeated measurements for data reproducibility was performed under supervision and prior to performing the removal torque test and respective calculation of connection stiffness.

A general linear statistical model with torque, energy, and stiffness as dependent variables and implant surface and time in vivo as independent variables was used at 95% level of significance and performed by one-way ANOVA followed by Tukey’s post hoc test. The Spearman rank correlation test was taken in order to test the association concerning the investigated dependent variables. Since sample size calculation, as already mentioned, was made to allow evaluation of possible differences in bone-to-implant contact between groups, a post hoc analysis was performed to define the minimum detectable difference between groups regarding the parameters assessed herein, with a power of 80% and an alpha error of 0.05%. The IBM® SPSS® Statistics software^{Footnote 5} was used.