The treatment of an unstable proximal humeral fracture is challenging. Although the majority of proximal humeral fractures can be treated successfully with nonoperative treatment, some fractures require operative treatment to achieve outcomes that meet the patient's expectations.
Several techniques and devices have been employed to treat unstable proximal humeral fractures, including the locking plate1,2, intramedullary locking nails3, Kirschner wires4, and screw fixation5.
The proximal humeral locking plate has been shown to achieve relatively rigid fracture fixation6,7. However, after the initial enthusiasm for locking plates, several authors recently have raised concerns about these implants8,9.
Less-invasive techniques, such as Kirschner wire fixation, may provide adequate stability and require a less-invasive second operation for implant removal. However, this advantage must be balanced with the inferior biomechanical properties of Kirschner wires, with risks that include wire migration and loss of fracture reduction10-12.
New fracture-fixation devices have been introduced to minimize the disadvantages of Kirschner wires13, with the goal of redistributing the loads on the wires and transmitting the loads from the insufficient subchondral cancellous bone to the lateral cortex of the proximal part of the humerus. However, the biomechanical properties of these devices recently have been shown to be inferior to locking plates14. The reasons for these differences could be related to several variables, including the Kirschner wire dimensions, the length of the threaded portion of the Kirschner wire, the pin configuration used in the humeral head (including the positioning, spacing, orientation, and number of pins), the system for cortical fixation, or a combination of all of these factors.
In 2005, a new device was developed to augment percutaneous wire fixation both by using an external fixator to link the wires and by using wires with longer threads that engage the lateral cortex instead of only the subchondral bone. Although a clinical advantage has been demonstrated for this device15, its biomechanical properties are unknown. The first purpose of the present study was to investigate the biomechanical properties of the construct with use of different dimensions of wires and different wire-fixation configurations to identify the most secure construct. The second purpose was to investigate the hypothesis that the addition of an external fixator would significantly improve the stability of the constructs. The results associated with these constructs were compared with those associated with a locking plate-and-screws construct. We also hypothesized that the best of these wire constructs would not be mechanically inferior to a locking plate-and-screw construct.
Seventy-two fourth-generation synthetic composite humeri (Sawbones; Pacific Research Laboratories, Vashon, Washington) were separated into nine groups of eight humeri each. Two-part proximal humeral fractures were created and then were stabilized with use of one of nine different methods. The constructs were loaded in cyclic torsion and cyclic compression-plus-varus bending and then were loaded to failure in the compression-plus-varus bending modality with use of an Instron 8800 servohydraulic tension-torsion materials testing machine (Instron, High Wycombe, United Kingdom).
Simulated surgical neck fractures were created at the base of the greater tuberosity at a 20° oblique angle, directed in a medial and inferior direction with use of a 2-mm-thick manual saw. A plastic template was used to standardize the osteotomy, which started 30 mm distal to the greater tuberosity. The osteotomy initially did not involve the medial cortex, to prevent fragment displacement and to preserve a 2-mm-thick osteotomy gap. The fracture was then stabilized, and the osteotomy was then completed.
The fractures were fixed according to the following techniques (Fig. 1).
2.5-mm Fan Construct (Fig. 1, A)
Four 2.5-mm diameter, terminally threaded wires with the tips threaded for 25 mm were inserted in a convergent configuration. The wires were inserted in a retrograde fashion and engaged one cortex, lateral to the biceps groove, converging within the humeral head, 30 mm distal to the tuberosity. The outer two wires were 10 mm apart where they entered the bone.
3-mm Fan Construct (Fig. 1, B)
This construct was the same as described for the 2.5-mm fan construct but involved the use of 3-mm fully threaded pins.
3-mm Fan Plus External Fixator Construct (Fig. 1, C)
This construct was the same as the 3-mm fan construct but was augmented with a custom-made external fixator (Orthofix Orthopedics International, Verona, Italy). The pins were manually bent 100 mm from the lateral cortex. The two opposite pins were then locked together with use of a clamp. After the remaining two wires were locked together, the two clamps were bridged with use of an 8-mm-thick carbon-fiber connecting bar.
2.5-mm Box Construct (Fig. 1, D)
Four 2.5-mm terminally threaded wires were inserted in a retrograde fashion across one cortex in a parallel configuration that formed a 15-mm square with two proximal and two distal wires. The wires were inserted lateral to the biceps groove at the same entry level as described above.
3-mm Box Construct (Fig. 1, E)
This construct was the same as the 2.5-mm box construct but involved the use of 3-mm fully threaded wires.
3-mm Box Plus External Fixator Construct (Fig. 1, F)
This construct was the same as the 3-mm box construct but was augmented with an external fixator that linked the four wires together.
2.5-mm Up-Down Construct (Fig. 1, G)
Four 2.5-mm terminally threaded wires were used in an up-down configuration. Two bicortical wires were inserted in a proximal-to-distal direction, starting from the greater tuberosity and passing through the medial cortex. The other two unicortical retrograde wires were inserted lateral to the biceps groove and ended in the subchondral bone.
3-mm Up-Down Construct (Fig. 1, H)
Four 3-mm fully threaded wires were inserted as described for the 2.5-mm up-down construct.
Plate Construct (Fig. 1, I)
A locking plate was used to fix the fracture according to the technique recommended by the manufacturer (AxSOS; Stryker Trauma, Selzach, Switzerland). Three proximal 4-mm locked screws were used to fix the humeral head, and one 4-mm locked screw was interfragmentary. Two 4-mm locked screws and one 4-mm unlocked bicortical screw were used to stabilize the humeral diaphysis.
Wire Insertion and Humeral Orientation
A custom-made guide was used to insert the wires in a reproducible way in each configuration. Each guide was built with use of four drill sleeves (2.5 or 3 mm) cast into a polymethylmethacrylate bone cement mold that perfectly matched the lateral aspect of the humerus (Fig. 2).
The distal part of the humerus was transected to leave a length of 280 mm, and the shaft was potted with polymethylmethacrylate bone cement in a custom-made steel cylinder (50 mm in diameter by 100 mm in length). The humerus was oriented in 20° of abduction (Fig. 3).
Testing
Each construct was subjected to cyclic torsion, cyclic varus bending, and load to failure in the varus bending mode in order to evaluate torsional stiffness, maximum rotational displacement, varus bending stiffness, maximum axial displacement during varus bending, and ultimate load to failure in the varus bending mode.
Two different sets of jig fixation were used for the varus bending and torsional tests. For torsional stability testing, the proximal part of the humerus was similarly potted with polymethylmethacrylate bone cement in a custom-made steel tube. The most superior 25 mm of the humeral head was potted in a way that the cement did not impinge on the plate or wires. For varus stability testing, a 15-mm metal ball was positioned in a predrilled concavity in the proximal aspect of the humeral head to allow reproducible application of the force. In order to accommodate the large medial-lateral bending deflections that occurred during axial compression plus varus loading, the moving actuator of the test machine had a linear bearing with a shallow socket for the metal ball mounted below (Fig. 3).
The constructs were first subjected to a cyclic torsion test by applying a ±2-Nm sinusoidal waveform for 100 cycles at a frequency of 1 Hz, and the stiffness of the construct (Nm/deg) was calculated as the line of best fit from the graph of torque (Nm) versus rotation (deg). After torsional testing, the same humerus was tested for bending stiffness by applying a 0 to 1000-N compressive force sinusoidally at a frequency of 2 Hz for 1000 cycles, and the stiffness of the construct (N/mm) was calculated. During both torsion and bending, the maximum rotation and vertical displacement were measured. The construct was then tested for ultimate load to failure in the axial compression-plus-varus bending mode at a test speed of 50 mm/min.
Statistical Analysis
The data were analyzed with a single-factor repeated-measures analysis of variance with use of SPSS 13.0 for Windows (SPSS, Chicago, Illinois). A post hoc analysis with Bonferroni correction was done to compare the nine constructs. The level of significance was set at p < 0.05.
Source of Funding
The funds from the University of Turin Medical School were used to buy the materials used for this experiment.
Cyclic Torsion Test
In cyclic torsion, the 2.5 and 3-mm Kirschner wire fan constructs had significantly lower stiffness than all other constructs (p < 0.05) except the up-down configuration (p > 0.05) (Fig. 4). The diameter of the wire and the length of the thread did not influence the stiffness significantly in the box or fan configurations (p > 0.05). Adding an external fixator tended to increase the torsional stiffness of the 3-mm fan construct (with the mean value increasing from 1.2 to 2.6 Nm/deg [p = 0.12]) and significantly increased the torsional stiffness of the box construct (p = 0.03). The locking plate construct had the highest torsional stiffness (3.8 Nm/deg; 95% confidence interval, 3.4 to 4.1), but this value was not significantly higher than that for the 3-mm box plus external fixator construct (3.3 Nm/deg; 95% confidence interval, 2.9 to 3.8; p = 1).
The maximum rotation results during the torsion tests were similar to the torsional stiffness results with regard to the comparative data for the different constructs (Table I). The 2.5-mm wire fan construct performed significantly more poorly in torsion as compared with the plate, box, and up-down constructs, with a mean rotational displacement of 4.4°, compared with means of 1.3° to 3.0° for the other constructs (p < 0.05). The application of an external fixator to the 3-mm wire fan group tended to reduce the displacement, to a mean of 1.9° (p = 0.07).
The 3-mm wire box construct with an external fixator, with a mean maximum displacement of 1.3° (95% confidence interval, 1.1° to 1.4°) performed as well as the plate group did (p = 1).
Cyclic Axial Compression Plus Varus Bending Test
The box and fan constructs tended (p = 0.08) to have a higher stiffness in varus bending when they had been augmented with an external fixator. These constructs tended to be stiffer than the plate (p = 0.07) and up-down constructs. However, the only construct that was significantly weaker (p = 0.03) was the 3-mm up-down wire construct when compared with the fan plus external fixator construct (Fig. 5).
Significant differences were not found between the constructs for the peak displacement of the humeral head during cyclic loading (Table II). Three of the nine constructs in the 3-mm up-down group failed prematurely, before the end of the testing protocol, at 120, 460, and 634 cycles.
Load to Failure
Unexpectedly, the ultimate load to failure of some constructs was essentially the same as that of non-osteotomized Sawbones because the bone did not fail at the fixation site. Instead, failure occurred at the site where the diaphysis of the bone emerged from the mounting pot when the load to failure approached 2000 N, which was the site with the largest bending moment. This was true for all of the constructs in the plate group, five of the nine constructs in the 3-mm wire box plus external fixator group, four of the nine constructs in the 3-mm wire fan group, and four of the nine constructs in the 3-mm wire fan plus external fixator group.
The 2.5-mm fan construct and the 2.5 and 3-mm up-down constructs did not perform as well as the other fixation modalities, with mean ultimate loads of 1647, 1822, and 1630 N, respectively (Fig. 6). The 3-mm fan constructs with and without the external fixator and the 2.5-mm box construct showed intermediate properties (mean ultimate load, 1948, 1941, and 1908 N, respectively).
The last three constructs had a mean ultimate load that did not differ from the intact Sawbones (2000 N for the 3-mm box construct, 2068 N for the 3-mm box plus external fixator construct, and 2070 N for the plate construct, compared with 2020 N for the Sawbones).
The constructs that failed at the fracture fixation site varied in terms of the mode of failure. The 2.5-mm fan and box constructs failed by a combination of axial slip (pistoning) of the wires at the lateral cortex and wire bending. The 3-mm fan and box constructs failed because of bending of the wires and loosening of purchase in the subchondral synthetic bone but did not slip in the lateral cortex. The up-down constructs failed because of a fracture of the medial calcar, where the antegrade wires penetrated the cortical bone in eight of eight specimens in the 3-mm group and six of eight specimens in the 2.5-mm group.
Multiple devices and fixation techniques for humeral head fractures have been tested biomechanically in the past, and the results of these tests have helped the clinician to make decisions on the treatment of proximal humeral fractures11,14,16-18. As we had hypothesized, the addition of an external fixator to both of the 3-mm box and 3-mm fan wire constructs increased the stiffness in all three testing modalities of torsion, compression, and load to failure. The improvement was most marked in the fan groups, but the absolute values were higher for the box constructs. The box group of four fully threaded parallel 3-mm wires plus an external fixator was consistently the superior method of percutaneous fracture fixation in torsion and compression and, as we had hypothesized, was mechanically similar to the plate-and-screw constructs. We found no other similar comparisons of the biomechanical features in the literature. In our clinical experience15, the use of this construct has led to a significant reduction in osteoporotic-related complications compared with standard percutaneous Kirschner wire fixation.
The fan group performed particularly poorly in torsion, probably because the wires crossed close to the fracture site. The addition of an external fixator tended to improve its stiffness, but not significantly (p = 0.12). The fan group generally performed better in compression than torsion, but the 2.5-mm subgroup performed poorly when loaded to failure. This latter observation is probably because the 3-mm wires were stronger in bending and engaged the lateral cortex as well as the subchondral bone in the humeral head. However, in the model used, the length of the thread or the diameter of the wires did not appear to have a significant effect in cyclic testing.
The up-down technique is a useful method for securing the greater tuberosity in three and four-fragment fractures. However, in the present study, the up-down construct performed poorly in compression testing, particularly in the 3-mm-wire subgroup. Figure 1, H shows that the wires crossed in the coronal plane at the osteotomy site, which explains the low torsional stiffness. The 3-mm wires performed better than the 2.5-mm wires in torsion, reflecting the greater stiffness of the larger wires and the superior grip of both cortices. However, the 3-mm group performed particularly poorly in compression in comparison with the 2.5-mm group, probably because the 3-mm fully threaded wires engaged all cortices, whereas the 2.5-mm wires were able to advance through the proximal cortex. The tuberosity wires were parallel to the vector of the compressive force in 20° of abduction. The method of failure in this group was due to a fracture of the medial calcar, where the distal end of the tuberosity wires finished. By engaging both cortices, the 3-mm tuberosity wires transferred more compressive force onto the calcar.
Overall, the most resilient construct was the plate, but significant differences were not found in comparison with the wire constructs augmented with an external fixator. The plate constructs were the stiffest in torsion, and none failed at the fixation site on loading. No screws loosened on cyclical compression. The stiffness of the plate in compression was less than that of most of the other constructs, but this did not lead to failure.
The relative contributions of varus bending and torsional stress that lead to clinical failure are not fully understood and probably are largely related to the fracture configuration and the bone quality of the patient. We therefore chose to test each bone in three different modalities. It was decided that we would cause the constructs to fail in compression and not torque because we believe that compression is the predominant force causing failure of fixation in vivo. It was therefore necessary to perform the nondestructive torsion testing first, after which the jig was changed and the compressive tests were then performed, ultimately leading to a destroyed bone.
The present study had some obvious weaknesses. We were comparing methods of fracture fixation that commonly are used in osteoporotic bone in the elderly patient. However, we used Sawbones that withstood >2000 N of force, often without the wires loosening or backing out. In fact, the wires usually bent when the bone failed, which is rarely seen in vivo, and thus the results were optimistic. However, the artificial bones are manufactured to have as little variability as possible, and therefore different group results can be compared directly with a high degree of certainty, which is not possible when using cadaveric bone. The guides that were used when inserting the wires allowed reproducible and accurate fixation, which meant that all eight bones in each group were nearly identical. Because the failure mechanisms may differ in real osteoporotic bones, it would be desirable to partly repeat the present study to confirm the performance of the best methods in real bones. In addition, these tests could only be related to the time immediately after surgery in view of tissue healing in vivo, which was the reason for limiting the number of load cycles. It is unfortunate that the bones failed at the diaphysis in the strongest fixation groups. This meant that we were unable to compare the ultimate compressive strength of the plate and box constructs and the constructs augmented with an external fixator with any certainty. The bones failed at the rim of the mounting pot under axial compression plus varus load. If the shaft of the Sawbones had been cut shorter, then the bending moment at the failure point would have been reduced, but this length was necessary in order to accommodate the plate construct. This loss of data has little clinical importance because failure of fracture fixation usually results from repetitive nondestructive cyclic loading.
The use of artificial bones allowed a range of fracture fixation constructs to be compared in tests involving a large number of consistent specimens, but the key findings should be confirmed in real bones.
If the configuration of the wires caused them to converge through the fracture site, the resistance of the construct to torsion or bending displacement was reduced.
By linking together the lateral ends of the fixation wires, the external fixator frames led to increased stiffness and strength of the constructs.
When four 3-mm-diameter fully threaded wires were placed in the box configuration and their lateral ends were linked by an external fixator, the stiffness and strength of the construct did not differ significantly from that of the bone plate and locking screws constructs.