This study used Titanium Elastic Nails (Synthes Inc., Paoli, PA. USA). These are flexible titanium alloy nails. The nails used measured 4.0 mm in diameter by 440 mm in length. Synthes’ literature does not recommend the use of their elastic nails in unstable fractures such as long, oblique spiral fractures and multifragmentary fractures (Synthes® Website) [13].

The models were produced using third generation Sawbone^® femora (Pacific Research Labs Inc., Vashon Island, WA). Studies on such composite bones have shown biomechanical similarity to cadaveric human bone (Heiner & Brown 2001) [3]. The length of the femur is 380 mm with a 12 mm canal diameter, which corresponds to the upper range of our target population. The segmental fracture pattern was created by using a power saw to cut two fracture lines at 45˚ from the longitudinal axis of the femur, both in the same plane. The fracture lines divided the shaft into three equal segments. Four femora were divided into two groups, each containing two femora. One group received two nails and the other received three nails. One femur was left intact as a control.

The insertion technique for the retrograde nails involved an entry point approximately 2 cm proximal to the growth plate bilaterally, as would be the case in vivo. A 4.5 mm drill bit was used at 45˚ to the axis of the shaft toward the medullary canal. The elastic nails were pre-contoured to three times the diameter of the medullary canal. The apex of the bend was at the level of the centre of the middle segment. The nail tips sat at the level of the greater trochanter laterally and at the midpoint of the femoral neck medially.

One group of femora then received an additional nail inserted antegrade. The entry point for this was at the tip of the greater trochanter. Again a 4.5 mm drill bit was used at 45˚ but the nail was inserted without pre-bending. The antegrade nail was advanced until it reached the distal crossing point of the retrograde nails. Final position of the nails was verified with radiographs (Fig. 1).

Figure 2

Figure 1

Biomechanical tests were conducted in the Centre for Orthopaedic Biomechanics at the University of Bath in the UK. Each model was tested in bending (antero-posterior [AP], medio-lateral [ML] and latero-medial [LM]), torsion (internal rotation [IR] and external rotation [ER]) and axial loading. Results were recorded in terms of deflection (mm) against load (N). Each test was repeated five times, allowing the bone model to rest between tests.

Axial loading was carried out using Instron^® 4301 and 4303 tensile and compression testing machine (Instron, High Wycombe, Bucks, United Kingdom). All samples were loaded to 500 N. In order to examine the effect that the middle femoral segment had on the stability of the construct, one femur in each group was chosen to have its middle segment removed to mimic a highly comminuted fracture. The model was then subjected to repeated axial loading. Load against displacement data were obtained in text format from the Instron machine, which by using the Instron Software Series 9, was imported to Excel spreadsheets (Microsoft Corporation). One femur from each nail group was then tested to failure. As in previous work (Fricka et al 2004) [1], failure was defined as 5mm shortening.

Bending and torsion tests were performed using a jig specifically designed for each test similar to that in Miles et al study (Miles et al 1994) [10]. In both tests the proximal femur was secured in an aluminum block machined to accommodate the shape of the head and greater trochanter. The gap was filled with Wood’s metal, a low temperature casting alloy, to ensure a secure fixation. In the bending tests, a dial gauge was attached at a standard point (fig. 2a) and load applied by weights on a hanger attached to the distal femur. Deflection in mm was measured against load applied. In the torsion tests, a metal plate was fixed to the femoral condyles (fig. 2b) and secured at the center of rotation with a metal post that allowed rotation but prevented bending. The angle of twist (Ө) in degrees was measured and torsional stiffness expressed in Nm/degree. . The angle of twist Ө was calculated as angle Ө = tan^-1 (d/A) where d is the vertical deflection, measured at a distance A from the axis of rotation. Stiffness = F.B/Ө (Nm/degree).

Both in bending and torsion tests, load were applied manually in five increments to a maximum load of 42 N.

Figure 3

Figure 2 (a & b)

Load against displacement data from the above tests were entered into spreadsheets using Microsoft Excel. From these, a load-displacement graph was produced for each test, as well as obtaining the mean displacement and the standard deviation. The gradient from each graph was calculated using the Excel software. This gradient represented the stiffness in N/mm for bending and axial tests and the torsional stiffness in Nm/degree for the torsion tests. These figures were used to compare stability between models.

Generalising estimating equations (GEE) were used to fit a linear model to the data to examine the relationships between the nail systems and control group. This form of analysis allowed the data to be grouped by sample to take into account the clustering from the five repetitions of each test. For each GEE model, an equal-correlation model was assumed for the within-group correlation and robust standard errors were used. The GEE models were constructed within a standard statistical software package (STATA v8.2, StataCorp LP, College Station, TX, USA). In each case the conventional exchangeable correlation structure was used to adjust for within sample correlation.

Paired t-tests were applied to investigate the differences between samples of the same nail system, and between axial loading of models with and without the middle segment.

The 3-nail system had significantly higher bending stiffness compared to the 2-nail system (Table 2): AP bending increases 1.14 N/mm (95% CI 0.69-1.59) (p<0.001); ML bending increases 1.60 N/mm (95% CI 0.73-2.47) (p<0.001); LM bending increases 1.92 N/mm (95% CI 1.17-2.66) (p<0.001).

As would be expected, both the 2-nail and 3-nail systems showed significantly lower bending stiffness compared to the intact sample in AP, ML and LM bending (p<0.001 for all comparisons).

Figure 4

Table 2: AP Bending Stiffnes Test results (N/mm).

The 3-nail system (Table 3) had IR torsion stiffness of 0.206 N/degree (95% CI 0.137-0.275) greater than the 2-nail system (p<0.001).

The 3-nail system had an ER torsion stiffness of 0.260 N/degree (95% CI 0.078-0.441) greater than the 2-nail system (p=0.005).

Again, both systems had significantly less stiffness compared to the intact sample in internal and external rotation (p<0.001 for all comparisons).

Figure 5

Table 3: Torsional (IR) Stiffnes results (N/m/deg.)

There was no evidence of a statistical difference in the axial stiffness between samples fixed with the 2- and 3-nail systems (p=0.669) (Table 4).

Figure 6

Table 4: Axial Stiffnes Test results (N/mm)

Both the 2-nail and 3-nail systems had significantly lower axial stiffness compared to the intact sample (p=0.002, p<0.001 respectively).

No failure (5 mm shortening) was encountered in any sample, testing up to 500 N of axial compression, representing the full weight of a 49 kg child. In all fractured samples tested with the middle segment present, the maximum shortening was only 1.0 mm at 500 N. In addition, when the middle segment was removed to represent a highly comminuted fracture, the maximum shortening was 2.27 mm when loaded to 500 N.

When testing the models to failure (i.e. a shortening of 5 mm), which only achieved after removal of the middle segment of the shaft from the samples, 760N was required in the 2-nail model and 1164 N for the 3-nail model.

Removing the middle segment of the shaft in 2-nail and 3-nail system shows significant reduction in axial stiffness, for example a paired t-test showed that sample 4 which is a part of the 3-nail system had a 195 N/mm (95% CI 157-233) reduction in axial stiffness after the removal of the middle segment of the femoral shaft (p<0.001). This indicates that the presence of the middle segment played an important part in the axial stability, a factor that might be difficult to replicate in vivo.

There were significant intra-group differences in axial stiffness between models fixed with the same system (Table 4). For the 2-nail system, one femur had a reduced axial stiffness of 420 N/mm (95% CI 338-501) compared to the other (p<0.001). For the 3-nail system, one femur had a reduced axial stiffness compared to the other of 264 N/mm (95% CI 214-315) (p<0.001). The causes of this variation in shortening are likely two-fold. The first is due to longitudinal closure of the fracture gaps followed by transverse gliding of the segments at the oblique fracture lines as load increases (Fig. 3). Subtle variations in the surface roughness of the fracture line potentially lead to differing engagement of the fracture surfaces and altered resistance to further load. The second is the fact that bending deformity cannot be entirely discounted, especially considering the natural bowing of the femur.

Figure 7

Figure 3

An interesting behaviour was observed during testing to failure of the models after the middle segment was removed. An audible noise was noted that coincided with spikes on the deflection/load graphs. The most plausible explanation was that during axial loading the curvature of the nail increased. This caused greater contact of the nails with the inner cortex, which then seemed to be suddenly reduced due to minute slippage of the nails at the entry points, thus causing further shortening of the sample. This process seemed to repeat itself as indicated by the regular audible noise, which continued until the experiment terminated. This has been named the ‘Pistoning Phenomenon’ by the authors and is detailed in (Figure 4). It should be noted that this effect only occurred at forces above normal physiological load.

Figure 8

Figure 4