|Home | About | Journals | Submit | Contact Us | Français|
Abundant evidence exists that fracture healing can be influenced by mechanical loading. However, the specific loading parameters that are osteogenic remain unknown. We hypothesized that the bone healing response in mouse tibial osteotomies would be different with a short delay before loading compared to immediate load application, as well as with higher and lower load magnitudes applied. Eighty 12-week mice underwent osteotomy of the left tibia followed by intramedullary nailing. Mice were divided into 6 groups based on days delayed until application of load (0 days or 4 days) and amplitude of cyclic load (0.5N, 1N, or 2N). Loading regimens were applied at 1 Hz for 100 cycles per day, 5 days per week for 2 weeks, using an external device that applied axial compression to the tibia. Bone healing was assessed by both microCT and four-point bend testing. A short delay followed by cyclic application of a relatively low load led to improved fracture healing, as determined by increased callus strength, but this enhancement disappeared as load amplitudes increased. Load initiation immediately following fracture inhibited healing, regardless of the magnitude of load applied. MicroCT measurements of calluses in the early healing stage did not predict the mechanical strength of the fractures. These findings confirm that controlled, noninvasive cyclic loading can improve the strength of healing callus. However, application of load immediately after fracture appears to be detrimental to healing. Load magnitude also plays a critical role, and must be taken into account in future studies and clinical applications. As the loading parameters necessary to enhance fracture healing become refined, external compression may be used as a potent stimulus for treating fractures with decreased biological capacity.
Most fractures heal uneventfully using traditional treatment methods, but more than 500,000 of the 5.6 million fractures that occur annually in the United States show evidence of impaired healing.17 Management of delayed healing and nonunions can be challenging. Abundant experimental and clinical evidence exists that the course of fracture repair can be influenced by mechanical stimuli. Previous studies, mostly in large animal models, have demonstrated that controlled loading can profoundly affect bone regeneration.1, 2, 5, 6, 9-11, 21-23, 25, 26 For example, cyclic compression of fractures leads to earlier and stronger callus formation,10, 11, 14, 15, 22, 26, 38 with early dynamization of external fixation leading to accelerated fracture healing compared to rigid fixation.10-12, 22 Conversely, static force across a fracture site, even when superimposed with cyclic compression, does not accelerate fracture repair.5, 6, 16, 29, 43
The optimal loading parameters to accelerate healing have not been defined. Paramount among these are the magnitude of the load and the time post fracture that loading is initiated. Several reports indicate that immediate27, 31, 32 or early3, 6 mechanical intervention is most effective, but other authors have suggested that a short delay may be necessary to allow initial neovascularization.9, 42 Similar uncertainty exists concerning load magnitude. For example, a wide range of load and strain magnitudes have been studied in sheep models,7, 10, 21, 30 but the effects of different peak loads on healing have not been examined.
We sought to address how both the time interval between fracture and initiation of loading and load magnitude affect the structural behavior of healing fractures. To answer this question, we employed a noninvasive external compression device, recently developed and validated to study adaptation of intact mice tibiae to mechanical stimulation.20 Loading parameters such as peak load, rate, frequency, and waveform can be applied precisely with this device. Using this device in a mouse model may have advantages over larger animal models for loading and fracture healing studies, including the feasibility of conducting large studies, the minimally invasive nature of load application, and the ability to take advantage of mouse genetic models.
Our goal was to use a mouse tibial osteotomy model with the external loading device to determine the effects of immediate loading vs. delayed loading, as well as load magnitude, on bone healing. We hypothesized that the timing of load initiation relative to fracture creation would have a significant impact on the extent of healing. We also hypothesized that the magnitude of cyclic load applied would be a factor in the efficiency of bone healing.
Eighty C57Bl/J6 male mice were obtained (Jackson Laboratories, Bar Harbor, ME) and acclimated in our facilities. The mice were 12 weeks old at the beginning of the experiment, which corresponds to early adult age. All mice underwent surgical tibial osteotomy and intramedullary nailing according to a protocol approved by our Hospital's Institutional Animal Care and Use Committee.
General anesthesia was induced with a ketamine (80 mg/kg) and xylazine (5 mg/kg) cocktail administered intraperitoneally. The dorsal aspect of one leg was shaved and sterilized using an iodine solution. Under a dissecting microscope, an incision was made over the knee and extended distally to the midshaft of the tibia. Dissection continued through the subcutaneous soft tissues, and the patellar tendon and tibial periosteum were identified. Just medial to the patellar tendon insertion site on the tibia, a 25-gauge needle was used to bore a hole in the proximal tibia, angling distally to enter the intramedullary canal. A 27-gauge needle was then inserted approximately 15 mm until resistance was encountered distally and then withdrawn slightly into the proximal tibia. Without stripping the periosteum, an oscillating saw was used to create a transverse osteotomy 5-6 mm distal to the patellar tendon insertion. The fragments were visibly aligned, and the needle was then reinserted into the distal fragment and cut to the appropriate length. The wound was irrigated and closed with several 4-O nylon sutures (Ethicon, Johnson & Johnson, Somerville, NJ). All fibulas were verified to be intact immediately postoperatively using X-ray. Buprenorphine was administered ad libitum for pain relief postoperatively, and no nonsteroidal anti-inflammatory medications were used. All mice were allowed full unrestricted cage activity following surgery. On the first postoperative day the animals did not appear to favor the non-operated limb.
Mice were randomly divided into six groups (n = 10 to 16 per group) based on the timing and load amplitude during postoperative loading. All groups underwent surgical osteotomy. In five groups, compression was applied to the ends of the tibia with a noninvasive external loading device (Fig. 1).19 Three groups had a 4-day delay prior to initiation of loading (“4d” groups), with load amplitudes of 0.5N, 1N, or 2N. In two additional groups, loading was initiated on the day of surgery (“0d” groups), and load amplitudes of either 0.5N or 1N were applied. The final group, the control group, was induced under anesthesia daily, and placed in the loading device without any load. This group was age-matched with the 4-day delay group.
A baseline preload of one half of the peak load was applied to maintain contact between the extremity and the loading device in all groups. All loading protocols were performed at 1 Hz for 100 cycles per day using a triangle waveform, for 5 days per week for 2 weeks under general anesthesia. The 2-week endpoint was chosen to detect early healing differences due to loading and is prior to complete fracture healing. The three different load amplitudes corresponded to loading rates of 1, 2 and 4 N/s (Fig. 2), which in intact tibia would give strain rates of 260, 520, and 1040 microstrain/s on the midshaft surface.33 Previous modeling of intact tibia loading demonstrated greater compressive than tensile strains at the region of the fracture site.13
Following completion of the 2-week loading regimen, the animals were killed by carbon dioxide inhalation. Immediately following euthanasia, the operated hindlimbs were disarticulated at the hip, the tibiae were dissected free of soft tissue, and the intramedullary pin was removed. Specimens were frozen in moist gauze at -20°C in preparation for microCT and mechanical analysis.
The formation of mineralized tissue was assessed by quantitative microCT (MS-8 small animal scanner, GE Healthcare, formerly EVS Corporation, Ontario, Canada). The tibiae were placed in a saline-filled scan tube and scanned at 80V and 80μA. Each scan included a phantom containing air, saline, and an SB-2 bone analogue (1.18 g/cc) for calibration of image Hounsfield Units to tissue mineral density in mg/cc. Reconstruction of the individual projections to CT volume data was accomplished with a modified Parker algorithm,18, 36 with an isotropic voxel resolution of 11.6 microns.
To distinguish mineralized tissue from marrow, water, and unmineralized callus, images were thresholded using 25% of the mineral attenuation value of the cortical bone for each specimen. In preliminary work, we determined that relative differences between groups were insensitive to different low-range threshold techniques. We chose a specimen-specific method to minimize any variability among the individual scans.41
Regional analyses of the thresholded scans were performed using the system software (MicroView, GE Healthcare Technologies, Waukesha, WI). To characterize calluses globally and locally, two volumes of interest (VOI) were selected (Fig. 3). First, a best-fit elliptical cylinder was placed around the entire callus, using the borders of the callus circumferentially and its proximal and distal extent. This VOI was used to determine the total mineralized volume of the callus. The second VOI was an elliptical cylinder centered at the osteotomy site with a height of 1 mm and of maximal cross-sectional area to fit entirely within the callus. Within this VOI, we calculated the total content of mineralized tissue (bone mineral content, BMC, in mg) and the mineral mass divided by the volume (bone mineral density, BMD, in mg/cm3). These measurements at the osteotomy site were used to judge the quantity and quality of newly formed bone.
Subsequent to microCT scanning, all bones were tested to failure using four-point bending on a precision electromagnetic-based load frame (EnduraTEC ELF 3200, Bose Corporation, Minnetonka, MN). Tibiae were placed with the flat anteromedial surface on the lower supports, which were set 8.4 mm apart. The upper load points were set 3.5 mm apart, and load was applied at 0.033 mm/sec until failure occurred. The bending moment to failure (N-mm) and bending stiffness (N-mm2) within the elastic range were calculated from the load-deflection curves and the 4-point dimensions.
For each measurement, the mean and standard deviation were calculated for each group. One-way ANOVA tests, followed by Bonferroni post hoc tests, were used to compare the means of each group to assess for statistical differences, using a p-value < 0.05. Regressions were examined and Pearson correlation coefficients calculated for each of the microCT measurements compared to the bending moment to failure in each group to determine the correlations between these two methods of callus evaluation.
The number of mice per group was based on a power analysis considered prior to the experiment. With 10 mice per group, there was over 80% power to detect a difference among groups of 1.2 times the pooled standard deviation using a one-factor analysis of variance design (alpha = 0.05). Power calculations were made using One-Way Analysis of Variance Panel, PASS 6.0 (NCSS, Kaysville, UT).
Increased callus strength was found in one loaded group relative to the control group. The group with a 4-day delay prior to loading and 0.5N load amplitude had 39% greater maximum bending moment (p < 0.05) compared to the control group (“4d / 0.5N”, Fig. 4). Bending stiffness was also significantly greater in this group compared to controls (Table 1). Despite the increased strength, neither callus bone volume, BMC, nor BMD was significantly different from the control group (Fig. 4, Table 1). On visual examination of the microCT's, the fracture calluses from this group appeared to generally more consolidated, bridged, and mature than the control group, though this was not formally quantified (Fig 5).
The load amplitude accentuated the effect of loading on callus formation among the 4-day delay groups. The 0.5N group was significantly stronger than both the 1N and 2N groups (Fig. 4). In addition, the 2N group had a significantly lower BMC and was less stiff compared to the 0.5N group (Table 1).
Immediate application of loading following surgery (the two “0d” groups) had an adverse effect on callus strength, regardless of the load magnitude. Load amplitudes of 0.5N and 1N led to callus strengths that were reduced by 68% and 57% relative to the control group, respectively, and stiffnesses that were reduced by 70% and 63% (Fig. 4, Table 1). However, in both groups the callus bone volumes were significantly larger (Figs. 4 and and5)5) and BMC was greater at the osteotomy site than for the controls (Table 1). No significant correlations were found between the microCT parameters measured at the callus and the bending moments to failure in any of the groups. However, these calluses visually appeared to be large and poorly bridged compared to the other groups.
In the search for a favorable mechanical environment to augment bone regeneration, a wide range of loading parameters have been identified that may contribute to the course of fracture healing. In our study, bone healing was enhanced by initiating daily application of a relatively low magnitude of intermittent cyclic axial load (0.5N) four days after the fracture occurred using a novel external compression device. The same loading protocol initiated immediately post-fracture reduced bone strength, despite increasing the mineralized volume of the callus. Similarly, higher magnitude loads applied after a four-day delay did not improve the load-bearing capacity of the tibia.
The ability to apply well defined, repeatable loading to the mouse tibia was critical to our hypothesis. We applied cyclic compressive axial loads, comparable to physiological loading of the tibia. The curvature of the mouse tibia may induce bending moments and may produce both compressive and tensile strains locally at the fracture site. The load magnitudes that inhibited healing were higher than the load that enhanced healing and may have produced strains that exceeded the strain threshold of the tissue for bone differentiation.28, 37 The strains induced at the middiaphysis with loading were characterized at the middiaphysis for the intact tibia of 10-week old C57Bl/6 mice.13, 20 A 4N peak load corresponded to approximately 1,040 microstrain on the anterior aspect of the tibia and even greater compressive strains were predicted on the posterior aspect.13, 20 In the future, finite element models based on microCT may be used to characterize the strains present at the fracture site and during healing with loading, and predict mechanical strength.39
The timing of load application was a significant factor in the strength of osteotomies after two weeks of healing. The importance of intervention during the inflammatory phase of healing is well accepted3, 5, 6, 27; however, within this healing window the precise timing needs to be established. The mechanical stability during early healing may determine whether bone is formed by endochondral or intramembranous ossification.32 We found that mechanical stimulation immediately post-fracture inhibited effective callus formation, regardless of the load magnitude applied. It is well established that callus volume does not predict mechanical strength,34, 45 as is seen in hypertrophic nonunions with inadequate stability, and occurred in the 0-day delay groups which had significantly larger but weaker calluses than controls. Although the mechanism is unclear, immediate loading may inhibit or disrupt early blood vessel ingrowth into the hematoma.9, 21, 23, 42, 44 When external fixator stiffness was varied in a sheep osteotomy model, a 25 percent decrease in fixator stiffness resulted in a fourfold decrease in corticomedullary blood flow after two weeks of healing.42 Although no direct correlations between the human and mouse fracture healing course have been reported, mouse fractures generally heal by 21 days, and remodel by 60 days.8 Thus the time points studied presently corresponded to the late chondral and early calcification stages of healing. The 4 day delay in our protocol is most equivalent to the hematoma and inflammation stages in the first 2 weeks of human fracture healing.17
The load magnitudes and strain rates also clearly affected callus formation, as evidenced by the progressive and significant decrease in callus strength among the 4-day delay groups as load amplitude increased from 0.5N to 1N to 2N. To maintain a constant loading frequency, increased load magnitude was coupled with higher strain rates. Based on the bone adaptation literature, higher strain rates would be expected to increase bone formation;33 therefore, we presume healing was modulated by strain magnitude in this study but cannot definitively distinguish the two effects. A load amplitude of 0.5N, which led to improved healing after a 4 day delay, corresponds to approximately one and a half times the mouse body weight. A direct comparison of our loads with those examined previously in sheep healing studies is difficult.1, 2, 6, 9-11, 21-23 These previous studies quantified the mechanical environment in terms of the applied strains, based on the size of the osteotomy gap and the displacement of the external fixator. The nature of the present study precludes quantification of the strain magnitude because the fracture gap is fixed with an axially unstable intramedullary nail, not an external fixator. No previous studies have examined specific loading parameters in a mouse fracture model. One study in rats used bending stimulation at an unspecified load, at 0.5 Hz beginning postfracture day 7, and showed the ability to modulate bone healing.40
Interpretation of these data is affected by several limitations of this study. The control group was age-matched with the 4-day delay group, so the groups loaded immediately following surgery had four fewer days of total healing time compared to the other groups. This four-day difference is unlikely to explain the significant deficit in callus mechanical performance relative to the controls, as minimal increases in bone strength are attained over the course of four days during this healing period.24 In fact, the difference in healing may have been underestimated, as the externally-applied load was superimposed on routine weight-bearing, which has been shown to dampen the effects of augmented load application.45 Regardless, the increased strength with the 4-day delay and 0.5N load amplitude was measured with respect to controls with a similar healing period. In addition, we examined a single time point of healing. Whether the differences seen after two weeks of healing translate into long-term effects remains to be determined. One technical limitation of this model is that the pin diameter which fits into the distal diaphysis often is too large for the proximal region, potentially causing slight translation of the fracture ends. Finally, in addition to varying the load amplitude, the preload applied with each load level also varied. This static load is not expected to influence healing,35, 43, 44 but was a variable nonetheless.
In summary, we were able to both enhance and inhibit fracture healing with different combinations of the timing of load initiation and cyclic load amplitude. When axial compression was applied with a short delay and at low load amplitude, healing was enhanced as determined by increased callus strength. However, applying a similar magnitude load too early after fracture significantly impeded healing. Future applications should include later time points to examine loading effects on callus remodeling later in the healing process. Confirming and clarifying the mechanical modulation of bone healing in the mouse is important, since this provides a basis for the use of genetically modified models to elucidate the molecular mechanotrandsuction mechanisms. The recent mapping of the mouse genome4 may also help identify the genes and cellular mediators involved the mechanosensitivity of fracture healing. Furthermore, knockout models of impaired healing may be used to assess the ability of mechanical loading to overcome a nonunion situation. Such research may further lead to directed mechanical and pharmacological therapies for fractures with impaired healing potential.
Funding provided by the Orthopaedic Research and Education Foundation, the Oxnard Foundation, the Kirby Foundation, the Clark Foundation, and NIH Core Grant P30AR046121