Increasing working length by placing a limited number of screws at each end of the plate has been recommended as a strategy to decrease construct stiffness and therefore allow more motion at the fracture gap [1, 7, 12, 18, 19]; however, previous mechanical studies have yielded conflicting results [8, 9, 11, 12]. Based on our results, we cannot support our hypotheses that constructs with fewer screws and a longer plate working length would be more flexible, have greater gap motion and be more resistant to cyclic fatigue failure than constructs with more screws and a short plate working length. We suspect that the inconsistencies in the results amongst reported studies might be attributed to the manner in which the plate is applied to the bone. In constructs that utilize non-locking plates [11, 14], the contact between the plate and the bone segments causes the bending moment to concentrate between the ends of the bone segments regardless the positioning of the screws. Therefore, the functional plate working length was not equal to the distance between the screws placed closest to the fracture gap, but rather to the unsupported area of the plate, which corresponded with the length of the fracture gap in this study. In contrast, the physical offset of a locking plate which is applied without the bone and plate in intimate contact enables a locking plate to bend along the entire segment of the plate between the two most centrally positioned screws . If our suppositions are confirmed, the concept of short and long plate working length as the distance between the screws closest to the fracture should only be applied to true functional locking plate constructs and should not be applied to conventional non-locking plate constructs.
The results of our cyclic fatigue testing are not consistent with previous studies which found differences between constructs with short and long plate working lengths [11, 12]. These contrasting results might be ascribed to differences in the fracture models employed and methods of loading. In our study the applied load was based on the dogs’ body weight in an attempt to replicate clinical situations. We used a similar approach to previous mechanical tests performed in canine femora. Goh et al. tested plated femora using 20%, 40%, and 60% of body weight to simulate progressively increased weight bearing in the early postoperative period, using values based on kinetic studies in dogs [20, 21]. As 40% to 50% of body weight is achieved in the hind limbs of clinically normal dogs during walking, we would expect that 100% of body weight would be a reasonable load to challenge the implant resistance to cyclic fatigue . In other studies the upper load threshold was set to induce implant failure within a predetermined number of cycles  or was established using displacement control based on plate strain . Establishing significance may have been masked in our study because we did not apply high enough loads as none of our constructs failed during cyclic testing. The magnitude of the load applied in diaphyseal fracture models is more complex than simply applying a load equivalent to mean body weight. In human patients which have been instrumented with distal femoral diaphyseal prosthesis with telemetric strain gauges, load values as high as three times body weight and bending forces up to ten times body weight have been recorded . Other loading protocols such as progressive loading  could have been used to shorten the test protocol and may have yielded significant differences between stabilization techniques.
Both the short and long plate working length stabilization constructs sustained 180,000 cycles with the equivalent of full body weight loaded in axial compression without failure, which approximates 2 to 4 months of weight-bearing during normal ambulation . This finding suggests that both constructs would likely provide acceptable clinical stability as most fractures are expected to obtain union within 12 weeks of stabilization [25, 26]. The yield load for both stabilization techniques ranged from 453 to 655 N, which corresponds to 2 to 4 times the hind limbs peak vertical force measured from a normal 20 kg dog ambulating at a trot [20, 27]. Although kinetic values measured with gait analysis are only an approximation of the forces tolerated by implants during loading, our results suggest that it is unlikely that either of our constructs would fail catastrophically in the early postoperative convalescent period. However, when interpreting our results it should be considered that we did not apply torsional loads, which may have contributed to earlier failure, especially in the long plate working length stabilization group.
Cadaveric studies have numerous limitations. When performing cyclic testing designated to resemble a clinical environment after fracture fixation, the loading plane should be considered . We selected an offset axial loading to simulate loading of a plated femoral fracture. Our testing methodology had the limitation of being isolated to a single plane, without considering more complex forces such as a combination of bending and torsional forces. In the diaphyseal region of the femur, however, axial and bending forces predominate and these forces were replicated in our testing protocol . Additionally, constraining the distal femur, as previously described , may have influenced the mode of failure. We, however, utilized uniform testing conditions allowing for valid comparisons between treatment groups. Another limitation was that we did not test a construct which utilized only locking screws. This omission limits our ability to make conclusions regarding the effect of plate working length of a locking plate employed in its intended application [9, 12]. A hybrid construct in which both locking and non-locking screws were used was selected because these constructs are commonly used in dogs . Furthermore, previous studies have shown that placement of a single locking screw in each of the major fracture segments increases a construct’s axial and torsional stability [29–32]. Another limitation that should be considered is that constructs had differences in the number of screws, in addition to differences in plate working length. Results of previous mechanical studies suggest that the number of screws may be a less important variable than the position of the screws and plate length [4, 12]. Our intent was to evaluate two constructs that represented contrasting approaches to the stabilization of long bone fractures in dogs [33, 34].