In the present study, cyclic tensile and compressive loading were applied to cantilever superstructure. In the clinical situation, both compressive and tensile force was exerted to implant-supported prosthesis, but no previous study has discussed the microleakage using model study under this situation. In case of conical connection, compressive force may have promoted a higher penetration of the abutments into implant body, which may eliminate microgap . But in the clinical situation, tensile force may also be applied to abutment–implant interface as indicated above; the model employed in the present study may be pertinent.
Some previous model studies which measured the extent of microleakage from implant–abutment interface employed microorganisms [6,7,8,9]. These studies focused on bacteria itself using visible solution cloudiness test , scanning electron microscopy , bacteria viability test , or checkerboard DNA–DNA hybridization method  and did not mentioned about the bacterial toxin. Toluidine blue employed in the present study can easily be measured using absorptiometry and it was reported to be similar to bacterial toxins in its molecular size . In addition, trend in microleakage was reportedly similar between bacterial leakage model and dye leakage model .
In the present study, removal toque of abutment screw after the cyclic loading showed no statistically significant difference among the groups. Generally, conical abutment is believed to be better in fit and stability than non-conical connection . One possibility of this discrepancy may be due to the deformation of abutment screw. Actually, implant–abutment connection after the removal of abutment screw was still tight in CC group, but they were easily divided in other two groups. This may indicate that implant–abutment connection in CC group was almost sound and intact after the loading. In contrast, in EP and IP groups, abutment screw may be deformed which lead to the increase of RTV of abutment screw. The fracture of abutment screws in EP and IP groups may support this speculation. In addition, axial force is strongly affected by the interfacial friction coefficient . In the present study, interfacial friction was supposed to be largest in CC group because the contact area between implant body and abutment was smallest in EP group and both EP and IP groups had parallel walls at the interface with gaps and voids [13, 14]. This may be one reason for the larger standard deviation of RTV in both EP and IP groups.
In the present study, chronological increase of the amount of microleakage was observed in all three groups. This is in consistent with previous studies [5, 13, 15, 16]. Sigmoid curves of microleakage in all groups meant that the amount of leakage was large at the early stage of loading. This was in agreement with previous reports. Harder et al. reported that bacterial toxin leak occurred within 5 min of incubation using in vitro experimental model study, even without application of loading [15, 16].
The comparison of microleakage among the groups at every 500-cycle load indicated that there were no significant differences among the groups at 500-cycle loading but were statistically significant differences in those between IP and CC groups. The reason for microleakage in EP group having no significant differences indicated between EP and IP or CC groups may be due to the largeness of standard deviation in the value of EP group.
The limitation of the present study was the number of samples. We believed the sample number (seven in each group) was sufficient to obtain the trends of microleakage, but it may be better to investigate using large number of samples to analyze the nature of microleakage at the implant–abutment interface in detail. Another limitation was that the fastening torque applied to screw was not necessarily accurate. To mimic clinical situation, we used dedicated beam-type toque wrench delivered from manufacturer. According to the previous studies, the wrench of Nobel Biocare only reportedly demonstrated the target torque value falling within the 95% confidence interval for the true population mean among four kinds of wrenches . In addition, significantly lower deviations of torque values for beam-type wrenches were reported than for coil and toggle-style wrenches . But toque wrench still has an inaccuracy because the scale printed on the wrench is not fine and it seems to be inappropriate to apply precise torque value. In the present study, we used new torque wrench, but in the clinical settings, variability of toque value may expand because torque wrench is repeatedly used, with sterilization procedure, which decrease the accuracy .