SEARCH

SEARCH BY CITATION

Keywords:

  • nonunion;
  • model;
  • impaired healing;
  • rat tibia osteotomy;
  • fracture healing

Abstract

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

The purpose of this study was to develop an inexpensive and easily implemented rat tibial osteotomy model capable of producing a range of healing outcomes. A saw blade was used to create a transverse osteotomy of the tibia in 89 Sprague–Dawley rats. A 0.89 mm diameter stainless steel wire was then inserted as an intramedullary nail to stabilize the fracture. To impair healing, 1, 2, or 3 mm cylindrical polyetheretherketone (PEEK) spacer beads were threaded onto the wires, between the bone ends. Fracture healing was evaluated radiographically, biomechanically, and histologically at 5 weeks. Means were compared for statistical differences by one-way ANOVA and Holm–Sidak multiple comparison testing. The mean number of “cortices bridged” for the no spacer group was 3.4 (SD ± 0.8), which was significantly greater than in the 1 mm (2.3 ± 1.4), 2 mm (0.8 ± 0.7), and 3 mm (0.3 ± 0.4) groups (p < 0.003). Biomechanical results correlated with radiographic findings, with an ultimate torque of 172 ± 53, 137 ± 41, 90 ± 38, and 24 ± 23 N/mm with a 0, 1, 2, or 3 mm defect, respectively. In conclusion, we have demonstrated that this inexpensive, technically straightforward model can be used to create a range of outcomes from normal healing to impaired healing, to nonunions. This model may be useful for testing new therapeutic strategies to promote fracture healing, materials thought to be able to heal critical-sized defects, or evaluating agents suspected of impairing healing. © 2010 Orthopaedic Research Society. Published by Wiley Periodicals, Inc. J Orthop Res 29:109–115, 2011.

Despite the many advances in orthopaedics, 5–10% of fractures in the US continue to have delayed or impaired healing.1, 2 In order to accurately test potential treatments of delayed fracture healing, an easily reproducible small animal model is necessary. Many models of impaired fracture healing in rats have been proposed.3–10 Most commonly, these models impair healing by resection of periosteum, removal of bone marrow, creation of a segmental defect, unstable fixation, or the interposition of foreign material between the bone ends. Many of the models require expensive hardware for stabilization or have technical steps that are difficult to perform reproducibly/precisely. For example, when using an external fixator for stabilization, prior studies have shown that even minor variations in frame stiffness have a profound effect on healing.7, 8 Trying to accurately resect a defined segment of periosteum or bone is also inherently difficult and so significant variations in the amount of resected bone are common.

The purpose of this study was to develop an inexpensive and easily implemented rat tibia osteotomy model that produces a range of healing outcomes from normal healing to impaired healing, to reproducible nonunions. We believed that a model that is easily modified to increase the challenge to fracture healing would be desirable to test potentially useful treatments for delayed fracture healing. Many treatments aimed at improving fracture healing have limits in that they work to speed up healing in an environment that is slowly healing, but are unable to overcome a more challenged fracture environment where a nonunion is highly probable. An intervention that is shown to help fractures heal more quickly or robustly in an unchallenging environment (e.g., ultrasonic stimulation) might not be capable of inducing healing in a situation of impaired healing. Conversely, a treatment (e.g., bone grafting) that is capable of stimulating the healing of nonunions may not actually result in more rapid or robust healing of unimpaired normal fractures. Prior to using a proposed intervention indiscriminately on all fractures, delayed unions, impending nonunions, and established nonunions, the limits of the intervention should be more accurately defined in order to prevent misapplication.

We undertook this study to test our hypothesis that under standardized conditions, using a spacer to control fracture gap size, bone union will still be possible with a small gap, but bone healing will become increasingly difficult with longer spacers.

METHODS

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Experimental Design

We used a number of rat cadavers to determine that a 0.89 mm Kirschner wire filled the marrow canal of the average female retired breeder rat and that a bead with an outer diameter of 1.89 mm readily held the bone ends apart to establish a defect without covering the entire cut surface of the tibia in a fashion which might preclude healing. We evaluated a number of possible spacer beads including commercially available glass beads which were not consistently sized, and machined beads of polymethylmethacrylate which were too brittle, and finally settled on machining beads (inner diameter 0.89 mm, outer diameter 1.98 mm) out of a sheet of polyetheretherketone (PEEK, model 8504K34; McMaster-Carr, Atlanta, GA). Using a milling machine (model EVS-3VS; ACER Group, Piscataway, NJ) allowed for a high degree of accuracy, with the length, inner, and outer diameters being within 0.05 mm of goal measurements, which was verified with a digital caliper (model 01408A; Neiko Tools, Gardena, CA) for each spacer prior to implantation.

After obtaining approval from the UNC Institutional Animal Care and Use Committee (Protocol No. 08-232), 89 female retired breeder Sprague–Dawley rats were obtained from a commercial breeder. Based on our prestudy power calculation, group sizes of 20 would be enough to detect a difference of 27% difference between groups assuming an expected standard deviation of 25% of the mean. From our prior experience, we expected a few animals from each group would likely have to be excluded due to postoperative infection, systemic illness prior to or after the surgery, significant weight change during the study period, intraoperative death, wound dehiscence, or a nontransverse osteotomy. Due to these anticipated losses/exclusions, we ordered 10 additional animals (10% extra) so that we would still have adequate numbers in each group to detect a difference after exclusions. The animals' mean body weight was 299 g (SD 56 g).

All surgical procedures were performed under general anesthesia using 1.5–2.5% isoflurane inhalation and sterile technique. The animals were randomized into four groups (no defect, 1 mm, 2 mm, 3 mm defect). The right leg was clipped, and an approximately 2 cm incision was made over the anterior surface of the tibia. Curved hemostats were used to bluntly dissect the tissue around the tibia. The insertion of the hamstrings on the anterior cortex at the apex of the tibial bow was used as a reproducible landmark identifying where to make the osteotomy (Fig. 1). Prior to making the osteotomy, the hemostats were passed just posterior to the tibia to protect the soft tissues. A rotary saw-toothed steel blade (0.2 mm thick, radius 12.7 mm, 130 teeth; model XML6-SS42S; Widget Supply, Albany, OR) was then used to create a transverse osteotomy at the level of the hamstring insertion, which correlates with the junction of the proximal and middle thirds of the tibia. Making the osteotomy at the apex of the tibial bow enables a straight wire to be inserted as an intramedullary nail. The saw blade was replaced after every two osteotomies in order to ensure that it did not become dull and produce heat necrosis at the osteotomy site. Saline irrigation was used during cutting to cool the saw blade. The fibula was fractured manually (distal to the osteotomy). A 0.79 mm drill bit (model BLW 12; Widget Supply) was used to ream the medullary canal (both retrograde in the proximal segment and anterograde in the distal tibial segment). A 0.89 mm stainless steel wire (316LVM, model GWXX 0350-60; Small Parts Inc., Miramar, FL) was inserted retrograde in the medullary canal, exiting in the midsubstance of the patellar tendon. In the experimental groups, one of the custom-machined PEEK spacers was inserted on the distal end of the wire prior to advancing the wire into the distal segment to reduce the tibia. The wire was advanced as far distally as possible. The proximal end of the wire was then bent (170°) with pliers in order to resist the wire backing out of the tibia. Any excess wire beyond 5 mm above the bend was cut. The wound was closed with absorbable sutures. Rats were allowed free movement about their single-housed cages and were provided unrestricted access to food and water. The analgesia regimen included acetaminophen (160 mg/5 ml) drinking water for 3 weeks after surgery), buprenorphine (0.01–0.03 mg/kg given at the time of surgery and then twice daily for the 3 days following surgery), and bupivicaine (0.25% solution given as 0.5 ml splash block prior to wound closure). All animals were sacrificed after 35 days of healing.

thumbnail image

Figure 1. Intraoperative view showing insertion of hamstrings at apex of tibial bow used as a landmark for the level of the osteotomy.

Download figure to PowerPoint

Radiologic Evaluation

Standard anteroposterior and lateral radiographs (model MX-20, Faxitron X-ray LLC, Licolnshire, IL) of the potted samples were taken using a digital cassette and processor (model CR MD 4.0, CR 35-X; AGFA Healthcare, Greenville, SC). Radiographs of each tibia were assessed by eight blinded, independent observers to judge whether there was bridging callus on 0, 1, 2, 3, or 4 cortices. The scores (0–4) of the eight observers for each sample were averaged.

Biomechanical Evaluation

Immediately following euthanasia, the right tibia of each animal was dissected. A small cuff of soft tissue was left around the osteotomy site so as not to disturb the (fracture) callus. The proximal end of the (intramedullary) wire was cut just distal to the bend. The proximal and distal ends of the bone were then potted with a polymer resin (number 265 & 928, 3M Bondo Corp., Atlanta, GA) using custom-made molds. The intramedullary wire was used to center the bone in the mold. The samples were kept moist in saline-soaked gauze during this process. After potting, the samples were again wrapped in saline-soaked gauze prior to storing in (−40°C) freezer until day of torsional testing. The morning of mechanical testing, the bones were thawed at room temperature.

To assess the mechanical quality of the healed fracture, the stiffness, maximum torque to failure, and energy of the fracture callus was determined by torsional testing in external rotation on a uniaxial servohydraulic material testing machine (8500 Plus, Instron Corp., Norwood, MA) fitted with fixtures to convert the axial motion to rotary motion. Each potting mold with the specimen was connected inline with the torsional axis of the apparatus. A precision potentiometer (Series P2201, Novotechnik U.S Inc., Southborough, MA) was used to measure rotations about the torsional axis and a 350 N/mm torque cell (Model 2105-50, Honeywell Sensotec, Columbus, OH) was used to measure torques during the tests. A preload of 0.7 N/mm was applied to each specimen at the start of the test. The specimen was then torqued at a constant rate of 6°/s. This rate was selected based on prior research that shows that maximum torque capacity and stiffness are independent of angular velocity when angular velocity is in the range of 3–12°/s.11 Load was applied until failure at the osteotomy site occurred or displacement exceeded 25°. Displacement >25°/s without reaching a peak torque was felt to be indicative of fibrous nonunion tissue rather than bony union. The torque and deflection angle were recorded and torsional properties measured using customized computer software written in a graphical programming language (Labview 6.0, National Instruments, Austin, TX). The ultimate torque was defined as the maximum torque recorded prior to 25° of angulation being achieved. Stiffness was determined by the slope of the regression line of the torque–deflection angle curve between limits of 25% and 75% of the maximum torque.

Histological Evaluation

A subset of torsionally tested fractures were subsequently processed for qualitative histological evaluation of the extent of bone bridging, cartilage callus, and fibrous tissue at the fracture gap. Specimens were fixed in neutral-buffered formalin for 48 h, transferred to 70% ethanol, decalcified, embedded in paraffin, sectioned longitudinally in 5 µm slices, and stained with hematoxylin and eosin.

We remain dissatisfied with the use of histologic evaluation of fracture specimens in the literature because it remains very subjective and such results are susceptible to selection bias and sample errors. Therefore, we chose to utilize all of our specimens to increase the “n” of our objective mechanical data. We selected three specimens from each spacer group to evaluate histologically. We had not saved any specimens solely for histologic examination so we performed our histology on the specimens after they had been torsionally tested. While there was some damage to the sections due to the mechanical testing, they did show different extents of healing with the different spacer sizes. The specimen selected to undergo histologic evaluation were representative of each spacer group in that they were average-weight animals (not an outlier with a very high or low weight compared with the mean), who had experienced no perioperative complications, that had radiographic confirmation of a transverse osteotomy, and had been operated on in the middle of the study period (to control for any variation in surgical technique that might occur at the very beginning or end of the experiment). Additionally, because we had previously performed mechanical testing on these specimen, we were able to select specimen that had mechanical properties at or near the mean for each spacer size, in a further effort to select specimen that were representative of each group.

Statistical Analysis

Mean and standard deviations were calculated and a one-way ANOVA test followed by Holm–Sidak multiple comparison testing was used to determine statistical differences (p < 0.05) between the groups for biomechanical parameters. Differences in cortical bridging between the groups were evaluated by Kruskal–Wallis ANOVA on Ranks followed by multiple comparison testing by Dunn's method. Data were normally distributed with the exception of the energy to ultimate load measure, which was found to become normally distributed after the data were subjected to a square-root transformation.

RESULTS

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Exclusions

Of the 90 rats initially ordered, 72 tibial specimen were included in the biomechanical testing. Eighty-five rats underwent surgery and survived to postoperative day 35. Of the 5 early exclusions, 1 animal was euthanized a few days after arrival to our laboratory due to a large flank mass that seemed to be causing the animal to suffer; 1 was euthanized at the end of surgery due to a catastrophic defect inadvertently created intraoperatively; 3 others never awoke from anesthesia. An additional 13 were excluded from torsional testing because of variation in fracture pattern (11), an enlarging neck mass noticed 2 weeks postoperatively (1), or infection noticed at the time of euthanasia (1). Variation in fracture pattern means that at the time of surgery, when the osteotomy was created with a saw blade, the surgeon noted that the cut was oblique rather than the intended transverse cut, or a small butterfly fragment was incidentally created. These variations were confirmed on X-ray. Tibias with nontransverse osteotomies were excluded from torsional testing because spacers placed on a nontransverse osteotomy created a different size fracture gap than the length of the spacer itself.

Radiologic Evaluation

Radiographs taken after potting the tibia show transverse fractures at the junction of the proximal and middle thirds of the tibia (Fig. 2). The mean number of “cortices bridged” was found to be significantly decreased as the size of the spacer was increased (Fig. 3). Approximately 75% of the osteotomies in the no defect group healed, as evidenced by ≥3 cortices bridged, compared with only 44% in the 1 mm group and 0% radiographically healed in the 2 and 3 mm groups.

thumbnail image

Figure 2. Representative radiographs of potted tibias showing examples of cortical bridging and corresponding intraoperative photographs for the 0 mm, 1 mm, 2 mm, and 3 mm defects using PEEK spacers (note: intramedullary pin is 0.89 mm in diameter for scale reference).

Download figure to PowerPoint

thumbnail image

Figure 3. Bar graph displaying the mean number of cortices (0–4) bridged in each group (whiskers represent 1 SD). * Significant difference between groups at ends of bar (p ≤ 0.05).

Download figure to PowerPoint

Interobserver repeatability in determining whether the osteotomy was healed or not, based on three or more cortices bridged, overall was high. There was 100% agreement that no specimen were healed in the 3 mm group. There was also 100% agreement that 11/12 specimen were not healed in the 2 mm group. One observer thought that the one remaining specimen in the 2 mm group had three cortices bridged, whereas the remaining 7/8 observers thought only two cortices were bridged (88% agreement), and therefore that the fracture was not radiographically healed. There was 100% agreement that 7/16 of the 1 mm specimen were healed, 75% agreement that two more specimen in this group were healed, and 100% agreement that the rest of the group (7/16 specimen) were not healed. In the no spacer group, there was 100% agreement that 12/16 specimen were healed, 75% agreement that one other specimen was also healed, with three cortices bridged, and 100% agreement that the remaining three specimen were not healed.

Biomechanical Evaluation

Displacement >25° without reaching a peak torque was felt to be indicative of fibrous nonunion tissue rather than bony union. In a prior rodent study, intact femurs were found to reach ultimate torques by 16° and healing fractures were observed to reach ultimate failure at 25°.12 In our study, none of the specimen in the 0 mm group, one specimen in the 1 mm group, five in the 2 mm group, and eight specimens in the 3 mm group displayed a peak torque at greater than 25° of displacement. This increased incidence of displacement >25° before reaching a peak torque in the groups with larger defects correlates with the radiographic results that there were fewer cortices bridged in the groups with larger defects.

The mean stiffness of the healed fractures was found to decrease significantly as the size of the spacer was increased (Fig. 4). The no defect group and 1 mm group displayed significantly greater stiffness than the 2 or 3 mm groups. There was a trend towards the mean stiffness of the no defect group being greater than the 1 mm group, but this was not significant (p = 0.121).

thumbnail image

Figure 4. Bar graph comparing the mean stiffnesses (N/mm/deg) and ultimate load (N/mm) for each group (whiskers represent 1 SD). * Significant difference between groups at ends of bar (p ≤ 0.05).

Download figure to PowerPoint

The mean failure torque was found to decrease significantly with increasing spacer size (Fig. 5). The differences in the mean failure torques were significant between all groups.

thumbnail image

Figure 5. Representative decalcified longitudinal sections (H&E staining) of the healing fracture 5 weeks after injury for each of the different spacer size groups. Fracture surface is oriented horizontally and longitudinal bone axis oriented vertically in all micrographs. In the 3 mm section bone is only shown at the top of the field because of the size of the fracture gap. Scale bar = 500 µm.

Download figure to PowerPoint

The mean energy was 768.1 ± 331.3, 595.9 ± 195.3, 582.8, and 189.2 N/mm/deg for the no defect, 1 mm, 2 mm, and 3 mm groups, respectively. The differences in the mean energy was significant between the no defect and 3 mm groups, the 1 mm and 3 mm groups, and the 2 mm and 3 mm groups (p < 0.001).

Histological Evaluation

A total of 12 specimen were analyzed histologically. The images included in Figure 5 were representative of the three specimen from each group, as qualitatively there was not significant variability between the samples examined. The 0 mm section demonstrates completed bone bridging across the fracture gap. The 1 mm section shows bone bridging across the fracture gap that is nearly complete along with cartilage callus formation. The 2 mm section shows a decrease in the fracture gap with cartilage callus attempting to span the gap but with some fibrous tissue remaining in the gap. The 3 mm section shows bone only at the top of the field with an extended fracture gap below filled with fibrous tissue.

DISCUSSION

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Most previous impaired healing/nonunion models are technically challenging, utilize expensive hardware, or have been found to be difficult to reproduce. The present study demonstrates a normal through impaired healing to nonunion model that may be useful for studying interventions aimed at improving fracture healing or evaluating agents that might impair healing. The incremental healing found between the no spacer, 1 mm, 2 mm, and 3 mm spacer groups seems to indicate that the same model can be used to provide a predictably more and more challenging healing environment. Being able to “dial-in” the amount of challenge to fracture healing in the same model is a significant strength.

Limitations of this study include the obvious fact that it is a small animal model with a range of small bone defects that, while large in comparison to the size of the animal, are still small compared to human bone defects. It may be that agents or techniques used to stimulate or perhaps impair healing in rats will produce different outcomes than in humans; however, this model can serve as an excellent first step in such evaluations.

Strengths of this model include the low cost of the animal model and of the implants used to produce a wide range of fracture healing outcomes and the relative simplicity of the surgery. The cost of the hardware necessary to surgically create the defect and stabilize it for our model was approximately $5 and the time of the surgery from incision to closing was approximately 12 min for a surgical resident. In the described experiment, Dr. Bob Dennis and Steve Emmanuel graciously donated their time and the use of a Bridgeport 3-axis milling machine, belonging to the University of North Carolina Department of Biomedical Engineering, to manufacture the PEEK spacers at no cost for labor or machine usage. The 3-axis milling machine is one of the most common machine shop Computer Numerical Control (CNC) tool milling machines and is available at most Universities with a Physics or Engineering Department and is widely available in the private sector. In investigating some commercial options for machining PEEK spacers we found that the spacers could be manufactured for $3.50 each (Nationwide Plastics, Dallas, TX). Even if a researcher has to pay to obtain these spacers commercially, $3.50 per implant is still significantly cheaper than other fixation methods such as mini fragment plates. In our model, the fraction of animals excluded due to any aspect related to the surgery was 19% and this is similar to previous reports using internal (14%) and external fixation (20%) in delayed fracture healing models in the rat femur.13, 14 Although our model does not decrease the number of animals that have to be excluded due to a surgical factor compared with previous reports, we believe that the lower overall cost of implant and the ability to more precisely create and maintain a fracture gap of particular size make the model better than previous models. The application of our model to the tibia rather than the femur is also a clinical strength because of the higher incidence of tibial fractures and delayed unions reported clinically at this site.15 Furthermore, the use of intramedullary fixation with our model is nearer to clinical practice because of the common use of this type of fixation in the tibia compared to external fixation which has been more commonly used in rodent defect models. While the use of a PEEK spacer at the fracture site in our model may seem to have no clinical connection, it is important to remember that PEEK interbody fusion cages are commonly used clinically in the spine and function similarly to the spacer in our model. The use of a specific subset of animals is another strength. We selected retired breeder rats to ensure that all animals were perimenopausal, in order to decrease variability in bone density and fracture healing capacity that corresponds to age. We specifically did not want immature animals as the increased ability to heal fractures enables this group to overcome challenges to bone healing that adult animals could typically not overcome. Additionally, a model using full-grown animals is more clinically relevant since few fractures in children experience delayed or impaired healing. A further advantage of our model is that testing a new modality expected to positively or adversely affect fracture healing can be undertaken over a variable challenge to healing. Testing a modality over the entire gamut of this model would enable an investigator to determine whether normal fractures (0 mm defect) are impaired or stimulated relative to normal healing or whether mildly impaired healing (1 and 2 mm defects) can be overcome, or whether critical-sized defects (3 mm) can be healed.

A key to obtaining reproducible results with a model is assuring that the technique is done exactly the same each time and of course to always include a new control group. Even with attention to detail, small variations are often unknowingly or unconsciously introduced into the procedure. These small variations can yield significantly different results. For example, we experienced this firsthand in a subsequent (not yet reported) study using the 1 mm defect version of the model described above to evaluate a treatment modality, the healing response of the control animals was noted to be significantly less than the control animals of the current study. On retrospective review of the plain films from the two studies, it was noted that the osteotomy was significantly more distal than in the initial study. The mean level of osteotomy (length of bone proximal to the osteotomy/total length of bone) for the initial (presently reported) study was 0.32 (±0.06) compared with 0.50 (±0.05) for the later study, which was statistically significantly different (p < 0.001). This more distal osteotomy was in the mid-diaphyseal region, compared with this initial study osteotomy being more proximal at the metaphyseal–diaphyseal junction. It is accepted that diaphyseal fractures heal more slowly than fractures at the metaphyseal–diaphyseal junction, perhaps explaining why the more distal osteotomies had significantly less healing over the same period of time than more proximal osteotomies. More specifically, a prior study comparing the influence of fracture level on the biomechanical properties of healing rat tibial fractures showed that the maximum load and stiffness decreased the more distal the fracture was located.3 Another example of a seemingly insignificant change in technique producing vastly different results can be seen by comparing the rat tibial osteotomy model proposed by Kratzel et al.16 In contrast to our technique using a stainless steel 120-toothed saw blade, this study used a diamond disk to create the osteotomy. No segmental defect was created. One might expect that these animals would heal similarly to the no spacer group in our study. Instead, all of their animals had either delayed union or nonunions until 84 days postintervention, whereas 75% of our animals had healed radiographically (defined by ≥3 cortices bridged) by 35 days postintervention. By design, diamond disks cut by friction (compared with the actual cutting away of bone chips achieved by a toothed saw blade), and we postulate that this friction may have caused some heating of the tissues in this model, resulting in an additional challenge to fracture healing although we cannot be sure of this reasoning as bone necrosis at the fragment ends was not observed histologically in this study. Additionally, the osteotomy in the model described by Kratzel was mid-diaphyseal, which is consistent with our finding that more distal osteotomies heal less well and may also explain the difference in healing between the studies. Because variation is so easily introduced into a biologic model/technique, a new control group must be included each time a study/different set of animals is used.

In conclusion, this model appears to be promising for early testing of interventions aimed at improving fracture healing, as well as agents suspected of impairing fracture healing. Despite the inherent weakness of a small animal model, it seems to be a promising one for the early stages of testing, as it is inexpensive and technically straightforward. As with all models, one must be diligent throughout every step of the experiment to avoid introducing variability.

Acknowledgements

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

This research was supported by the Aileen Stock Research Fund, as well as the UNC Department of Orthopaedics Research Fund. The PEEK spacers were generously donated by Dr. Robert Dennis and Steve Emmanuel (UNC Department of Biomedical Engineering).

REFERENCES

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES