The development of angle-stable locking screw fixation was the most significant advance in bone plate osteosynthesis of fractures since their original conception of the dynamic compression plate. Some potential advantages of locking plate fixation include the reduced need for accurate anatomical plate contouring, improved construct stability in osteoporotic bone and maintenance of bone blood supply in minimally invasive osteosynthesis. However, the requirement to insert locking screws at a predetermined fixed angle to the bone plate can be a ‘double-edged sword’, especially in juxta-articular locations. This is because the intraoperative bending of both straight locking plates and anatomically pre-contoured locking plates to follow the curved contour of the metaphysis can inadvertently direct locking screws into the articular surface or other internal fixation implants.
There are several possible solutions to this problem. A locking screw hole in the plate can either be left empty or filled with a cortical bone screw that is directed away from the joint. Alternatively, a shorter locking screw that does not reach the articular surface can be inserted. Each of these options might reduce the strength of the plate-construct in this region. Consequently, surgeons may elect to intentionally insert a locking screw in an off-axis direction – that is at a small angle away from the pre-determined angle of the plate hole. For locking plates designed to accept locking screws with a threaded head, this will result in cross-threading of the screw–plate junction. Cross-threading of the screw head into the locking plate hole can impair the stability of the fixation, resulting in screw loosening and screw ‘back-out’. This was observed in early clinical cases of human forearm and femoral fractures in up to 4% of patients, prior to complete healing of the fracture. Consequently, the potential for delayed or non-union of fractures due to locking screw loosening and fixation construct instability is a risk, albeit small.
During the fracture healing period, the biomechanical cyclic loading forces that act on long bone fractures repaired by locking plate osteosynthesis are a complex interaction of shear and axial loading, combined with bending loads. These in vivo loads are extremely difficult to replicate by ex vivo laboratory quasistatic mechanical testing. There are two main methods for ‘simplified’ ex vivo testing of locking screw constructs. One method is ‘screw pushout’ from the plate hole. One study published in this issue of the journal reported that the initial pushout forces of locking screws inserted axially with a torque of 1.5 Nm into 3.5-mm locking compression plates (4,356 ± 266 N) and 3.5-mm polyaxial locking plates (3,992 ± 84 N) were reduced by about half with five degrees of off-axis insertion. The pushout forces for the five degree off-axis screws were increased slightly by tightening the screws to 2.5 Nm.
The second method of ex vivo mechanical testing of locking screw-plate connections is to apply cantilever-bending load to the screw engaged in the plate hole with a force directed perpendicular to the screw long axis. This is considered to be the principal loading modality of locking screws in vivo. Also, it is the reason that locking screws are manufactured with a larger core diameter than the corresponding cortical bone screws – this increases cross-sectional area moment of inertia of the locking screw and consequently its bending stiffness. Some examples of studies reporting the performance of locking screws tested in this manner are as follows. 
The load at 1.5 mm of displacement for 3.0-mm threaded locking screws inserted at optimal angle into second-generation PC-Fix plates (Mathys, Bettlach, Switzerland) was 1480 ± 390 N, and for screws inserted at five degrees angulation it was 780 ± 110 N (p = 0.0001).
For 3.5-mm locking screws inserted at optimal angles into fully circumferential threaded holes in the Peri-Loc plate (Smith and Nephew, Memphis, Tennessee, United States) and loaded at a constant rate of 5 mm/min until failure in a cantilevered bending, failure occurred at approximately 5 N/m. The failure load of screws inserted at three degrees angulation was reduced by approximately 50%.
For 5.00-mm locking screws inserted into Combi holes of a 4.5/5.0 locking compression plate (Mathys, Switzerland), the load at 1.0mm displacement for screws inserted at optimal angle was up to 1,240 ± 210 N, and for screws inserted at five degrees angulation it was up to 790 ± 90 N.
These results of testing by push out and cantilever bending are not generalizable across different types and manufacturers of locking bone plates. However, they do suggest that deliberate off-axis insertion of a locking screw by five degrees results in about a 50% reduction in screw push-out strength and cantilever bending of the locking screws. This reduced strength of a deviated locking screws with cross threading is probably more than adequate to maintain construct stability in canine patients, although in vivo loads experienced by locking plates with complete cyclic loading are currently unknown. Certainly, the routine angulation of locking screws in locking compression plates or polyaxial plates is not being advocated, but it might be a solution to prevent articular penetration of a single locking screw at the end of locking plate, such as a tibial plateau levelling osteotomy plate. This is problem that warrants continued investigation.