The stability and integrity of the abutment-implant connection, by means of a screw, is fallible from the moment the prosthetic elements are joined and is dependent on the applied preload, wear of the components and function. One of the main causes of screw loosening is the loss of preload. The loosening of the screw-abutment can cause complications such as screw fracture, marginal gap, peri-implantitis, bacterial microleakage, loosening of the crown and discomfort of the patient. It is also reported that loosening of the screw/abutment may lead to a failure of osseointegration. It is necessary to evaluate and quantify, with in vitro studies, the torque loss before and after loading in the different connections. Aim: evaluate the influence of implant- abutment connection design in torque maintenance after single tightening, multiple tightening and multiple tightening followed by mechanical cycling. Materials and Methods: 180 Klockner implants divided in 4 groups: 15 SK2 external connection, 25 Ncm tightening torque; 15 KL external connection, 30 Ncm tightening torque; 15 Vega internal connection, 25 Ncm tightening torque; 15 Essential internal connection, 30 Ncm tightening torque. In each group removal torque values (RTV) were evaluated with a digital torque meter, in 3 distinct phases: after one single tightening, 10 multiple tightenings and 10 multiple tightenings and cyclic loading (500 N × 1000 cycles). Results: After one single tightening, and for all connections, RTV were lower than those of insertion, but only for Essential and Vega internal connections this result was statistically significant. After multiple tightening, RTV were significantly lower in all connections. After repeated tightening followed by cyclic loading, mean RTV were significantly lower, when compared to insertion torque. The multiple tightening technique resulted in higher RTV than the single tightening technique, except for Vega implant. The multiple tightening followed by cyclic load, compared to the other phases, was the one that generated the lowest RTV, for all connections. Conclusions: The connection design, in our study, did not seem to influence the maintenance of preload. Loading influenced the loss of preload, in the sense that significantly decreased the removal torque values. The multiple re-tightening technique resulted in higher removal torque values than the single tightening technique. Clinically, our results recommend to retighten retaining screws, a few minutes after insertion.