While the first case was uncomplicated, our second case went into nonunion, requiring conversion to internal fixation. A nonunion rate of 5-17% has been noted by other authors versed in this technique (Table 1) [1–5] comparable with rates of nonunion (up to 20%)  in traditional external fixation. Nonunion in Case 2 can be attributed to the nature of LCP application and characteristics of the LCP that make it stand apart from traditional external fixation. First, while traditional external fixation employs introduction of half pins prior to cross-bar connection, LCP external fixation requires drilling and screw placement through predetermined plate holes while the plate is suspended above bone. During plate application, both plate and bone fragment can move independently, making accurate screw placement difficult as small shifts at the plate translate to great deviations at the level of bone. Second, with a single screw in place, plate movement is confined to rotation in one plane and once two or more screws are placed, alterations in plate position are no longer possible. Third, unlike the more forgiving traditional fixator, the monoaxial nature of the locking-head screw trajectory reduces the ability to compensate for imperfect placement, making it mandatory that anatomical reduction be achieved prior to placement of the first screw. Should adjustment be required following application, it may be necessary to abandon either the drilled bone hole or the selected plate hole. Fourth, the small space beneath the plate makes it difficult to apply vascularized soft tissue cover. Flap inset on top of a plate might lead to tension on the pedicle and pose problems for later hardware removal. To site the fixator away from the open wound in Case 2 in anticipation of later soft tissue coverage, the LCP was twisted to achieve a proximal-anteromedial, distal-anterior plate siting (Fig 3b) instead of a fully anteromedial placement. Another strategy to facilitate dressing changes and soft tissue coverage involves plate twisting or incorporating a "wave" design . This must be done with caution so as to avoid disruption of plate threads, thereby precluding screw placement. Fifth, while traditional constructs can be strengthened by stacking cross-bars, this is not possible for LCP external fixation. A more rigid construct can be created by reducing the moment arm with a thinner spacer (fewer folded towels during plate application), increasing overall screw number, placing screws closer to the fracture, and increasing the distance between screws in each screw group [6, 8, 14]. Alternatively, Kloen's strategy of double fixation (two LCPs on the tibia) may be attempted to surmount this problem . Sixth, screw placement is absolutely limited to the available screw holes of the chosen plate. Valgus drift in Case 2 may have been potentiated by inadequate screw purchase in the small distal fragment and compounded by screw loosening in increasingly osteopenic bone.
Certain considerations must be borne in mind when concentric bones such as humerus and femur are involved. Bicortical engagement may not always be possible owing to limitations on available screw length. If so, this must be compensated by use of more unicortical screws, bearing in mind that unicortical configurations have 50% less rigidity than bicortical purchase [6, 8].
Finally, additional cost is incurred if initial fixation is with a conventional external fixator (such as in both above cases). This is not the case if LCP external fixation is used primarily , although we have no experience with this.