A surgical tendonitis model in horses: Technique, clinical, ultrasonographic and histological characterisation

Histopathological evaluation of the SDFT was performed at the same levels as the sonographic examination. Results: Only mild clinical signs of tendonitis were observed. Ultrasonographic core lesions were 10–16 cm long and had a mean maximum cross-sectional area (CSA) of 18.25 ± 5.91% occurring at 17–23 cm distal to the accessory carpal bone, and a mean volume of 1.86 ± 0.26 cm 3 . Mean duration taken to achieve maximum lesion CSA and lesion volume was 35 ± 7 days. Histologically, the lesions were characterised by mild inflammation followed by fibroplasia. Conclusion: The reported surgical technique resulted in core lesions that were consistent in size and location, were readily evaluated with ultrasonography, and showed similarities with the ultrasonographic and histological progression of naturally occurring tendonitis lesions.


Introduction
Tendon injuries are common in all athletic activities in both humans and horses (1). Strains of the superficial digital flexor ten-don (SDFT) account for up to 46% of limb injuries in racing Thoroughbreds and were reported to be the most important reason for retirement of racehorses from racing (2,3). The most commonly used disease model for tendonitis of the equine SDFT is collagenase-induced injury (4,5). Because of the difficulty in producing experimental tendon strain by mechanical means, intratendinous injection of collagenase was introduced as a model of overload tendon injury to mimick the release of degradative enzymes by inflammatory cells following spontaneous injury (4)(5)(6). This model has been used extensively in the evaluation of both the pathology of tendonitis and the efficacy of various treatment modalities (4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15). However, the relevance of collagenase-induced tendonitis to mechanical overload injury is uncertain and has increasingly been questioned (9,16). The model results in a strong inflammatory response that extends to the peritendinous tissues, either due to leakage of collagenase through the injection sites or because it causes a 'melting' tendonitis that erodes from the center of the tendon outwards and destroys the superficial tendon layers and paratenon. This can be accompanied by marked pain, swelling and lameness for several days to weeks after injection (6). As a consequence, the location, size, shape and volume of core lesions can be difficult to control and standardise between different limbs injected with collagenase. Due to the absence of 'true' core lesions surrounded by a margin of intact, tendon tissue with normal echogenicity, therapeutic medications administered by intralesional injection, such as polysulphated glycosaminoglycans, insulin-like growth factor-1, mesenchymal stem cells or platelet-rich plasma, may not remain within the injured tendon.
The objective of this study was to determine if a core lesion could be created surgically in the middle to distal third of the metacarpal region of the SDFT that was reproducible and did not result in the disadvantages associated with collagenase-induced injury. Our hypothesis was that a core lesion could be created surgically in the SDFT that showed clinical, ultrasonographic and histopathological appearance and progression similar to naturally occurring tendonitis.

Technique
The experimental protocol was approved by the Institutional Animal Care and Use Committee of North Carolina State University. Four horses, which had been donated for reasons other than tendon injury, were used in this study. Two were Thoroughbreds and two were Thoroughbred crossbreds. Each horse was evaluated for forelimb lameness and underwent bilateral ultrasonographic examination of the SDFT with a 14 MHz linear array transducer a and an echolucent stand-off block b to rule out the presence of pre-existing tendon disease.
Horses were placed under general inhalation anaesthesia and positioned in lateral recumbency. A core lesion of 8 cm in length was created in the SDFT of each forelimb using a 3.5 mm synovial resector c under ultrasonographic guidance (ǠFig. 1). A 5 mm vertical skin incision was made with a number 15 scalpel blade d , approximately 20 mm proximal to the ergot in the sagittal midline on the palmar aspect of the fetlock. The incision was continued through the palmar annular ligament and the palmar mesotenon (vinculum) into the palmar aspect of the SDFT. The SDFT was incised to a depth of 2-3 mm, taking care not to penetrate the dorsal surface of the tendon. A 2.5 mm blunt obturator was then introduced and guided proximally through the central part of the SDFT parallel with the longitudinal fibre direction of the tendon under ultrasonographic guidance to create a tunnel over a distance of approximately 10 to 12 cm. Penetration of the dorsal margin of the SDFT was avoided by ultrasonographic monitoring of the position of the tip of the obturator. Following withdrawal of the obturator, a sideways-cutting 3.5 mm Razorcut™ Blade c was introduced along the previously created tunnel. When the entire length of the blade of the synovial resector (8 cm) was positioned in the tendon, suction was applied to the resector prior to activating the blade. With the blade and suction activated, the resector was withdrawn gradually from the tendon, while manually rotating the cutting blade at the tip in order to transect tendon fascicles dorsally, laterally and medially while protecting the thin palmar covering margin of the tendon from penetration. Suction and cutting were discontinued when the tip of the synovial resector was located 2.5 cm proximal to the stab incision in the skin. Following removal of the synovial resector, a 14-gauge teflon catheter e was positioned inside the tunnel and balanced electrolyte solution f as injected while the cross-sectional size of the defect was assessed ultrasonographically. This procedure was repeated until a consistent core defect of approximately 10% of the cross-sectional area (CSA) of the tendon was identified ultrasonographically. This required up to three passages of the synovial resector along the intra-tendinous tunnel. The synovial resector was then withdrawn and the deficit in the palmar annular ligament closed with a cruciate suture of 2-0 poliglecaprone 25 g suture, while the skin was closed with one or two stainless steel staples h . Limbs were bandaged routinely.

Postoperative Care
A three-day course of intra-muscular procaine penicillin G (20,000 IU/kg SID) was initiated prior to induction of anaesthesia. Horses were stall rested for 24 hours postoperatively and then exercised daily for 15-30 minutes at the trot and canter on the lunge for seven days or until they became consistently lame at the trot, whichever occurred earlier. Horses received phenylbutazone (2.2 mg/kg, PO, q12h) for a total of four days postoperatively. Apart from these initial exercise periods, horses were kept in stall confinement throughout the study. Body temperature, heart rate, respiratory   (17).
Hand walking exercise was initiated on the 21 st postoperative day starting at 10 minutes daily, and continued each day, with time being increased each week by five minutes, until the end of the experiment.

Ultrasonography
Ultrasonographic examination was conducted at weekly intervals starting on the first postoperative day. Care was taken to use identical gain settings, focus depth, magnification and frequency for all ultrasonographic examinations. The skin on the palmar aspect of the metacarpal region was clipped, aseptically prepared, and marked with indelible pen at 7,9,11,13,15,17,19,21,23 and 25 cm distal to the most prominent palmar protuberance of the accessory carpal bone, and acoustic gel was applied i . Ultrasonographic measurements of the CSA of the SDFT and the CSA of the core lesion were made at each marked level. Ultrasonographic core lesion size was calculated for each level as a percentage of the total tendon CSA. Tendon volume and core lesion volume were calculated by multiplying the CSA at each level with the distance between two consecutive CSA measurements (20 mm), and summing the volumes of all 20 mm sections between 9 cm and 23 cm distal to the accessory carpal bone or the volumes of all 20 mm sections with ultrasonographic evidence of a core lesion. A sonographic score was given at each level for echogenicity and linear fibre pattern of the core lesions on a scale of 0 to 5, adapted from a previously reported grading system, in which echogenicity score 0 means gen-erally anechoic, echogenicity score 5 is normoechoic, linear fibre pattern score 0 means a total absence of linear echoes, and a linear fibre score of 5 means normal, long parallel linear echoes (18). The largest anechogenic lesion for each limb was identified and the site marked on the skin for future reference.

Pathology
At each of the following time periods following surgical injury inducement, a horse was euthanatized: two, four, eight and 12 weeks. The forelimbs were removed within two hours of death through the middle third of the radius. Skin markings were transferred to the corresponding levels of the SDFT with the limb extended and the SDFT was removed. A series of transverse and longitudinal incisions were made at the level of each mark on the SDFT to include the core defect at the level of the previously imaged areas. Gross photographs were taken of the transversely cut surface of each tendon segment at each marking level to show the macroscopic appearance of lesions. Tendon segments were preserved in 10% neutral buffered formalin for 24 hours, followed by 24 hours in ethyl alcohol with five percent phenol. The tendon segments were embedded in paraffin, cut to 6 μm sections on a rotary microtome, and stained with haematoxylin and eosin and Masson trichrome stains. Sections were evaluated under plain and polarised light. As previously reported, lesions were classified as acute, traumatic (stage 1), early inflammatory (stage 2), late inflammatory (stage 3), fibroplasia (stage 4) or healed (stage 5) (16). Alignment and density of collagen fibres was graded qualitatively under polarised light as absent, little, moderate, good or excellent. Core lesion measurements (maximum CSA, core lesion volume, ultrasonographic echogenicity and linear fibre pattern) were compared between different measuring levels and different time points using a oneway ANOVA. For those tests that indicated significance, a post hoc Tukey's test was performed. The total core lesion volume and maximum CSA of lesions were compared between limbs at each time point with a chi square analysis to test for variance in the normal distribution of these measurements between limbs at each time point. Significance of all statistical tests was set at <0.05.

Clinical signs
Body temperature, heart rate, respiratory rate and appetite remained within normal limits throughout the study. There was no evidence of heat, pain or swelling of the SDFT on the first day after surgical injury for any of the horses. None of the horses were lame when walking or trotting in the first 24 hours after surgery. Two horses became mildly lame when trotting during the six-day period of exercise that started on the second postoperative day. There was mild heat, mild swelling and signs of mild pain on palpation of one or both SDFT for three of the four horses during the six-day period of exercise. Maximum lameness observed in any of the four horses throughout the study was a mild to moderate grade of 3 out of 5 and occurred only during and immediately following the six-day period of exercise. Lameness resolved rapidly within two to seven days after the cessation of exercise.

Ultrasonography
Anechogenic core lesions with disruption of the longitudinal fibre pattern were visible following distension of the intra-tendinous tunnel with balanced electrolyte solution during surgery (ǠFig. 2A). All core lesions were contained within the SDFT, surrounded at all levels by a continuous margin of normoechogenic tendon tissue. Twenty-four hours following surgical injury, core lesions were hypoechogenic rather than anechogenic (ǠFig. 2B) and well-defined from the normal surrounding tissue. There was minimal sonographic evidence of peritendinous oedema or swelling. Seven days after injury, core lesions became slightly larger and more anechogenic and there was mild to moderate peritendinous oedema in six of eight limbs (ǠFig. 2C). At two weeks, mild to moderate peritendinous oedema was present in all limbs. At two and four weeks, lesions continued to increase in size and were characterised by a mixture of anechogenic and hypoechogenic regions (ǠFig. 2D). At six weeks, there was slight increase in echogenicity of core lesions in all four remaining limbs. There were a few irregulary arranged, short linear echos representative of short, longitudinally oriented fibres in core lesions in both forelimbs of horse four. At eight weeks, a reduction in core lesion size was observed for the first time in three of four limbs. The margins of the lesions were less well defined. Echogenicity and longitudinal fibre pattern showed some further improvement in both forelimbs of horse four at weeks 10 and 12, but the longitudinal pattern was still poor. The mean echogenicity and linear fibre pattern scores of the core lesions were consistently low for the first six weeks of the study but echogenicity started to improve from six to eight weeks after injury (ǠFig. 3 and 4). Mean tendon volume, mean core lesion volume, mean maximum CSA size as percentage of tendon CSA and mean lesion length are shown with standard deviations at different time points of the study in ǠTable 1. The mean time to maximum core lesion volume and maximum core lesion CSA was 35 ± 7 days. There was a significant difference between the CSA of the lesion at different measurement levels within the tendon (p <0.001 at two weeks after injury, p <0.001 at four weeks and p <0.001 at eight weeks) (ǠFig. 3). The maximum lesion CSA occurred between 17 and 23 cm distal to the accessory carpal bone at two, four and eight weeks after injury. No significant differences in lesion CSA were identified between levels 17, 19, and 21 cm distal to the accessory carpal bone. There was a significant difference between the maximum CSA of lesions at different time points after injury (p <0.001). Maximum lesion CSA was significantly larger at three, four, five, six and seven but not eight weeks when compared to the first day following surgical injury (p = 0.001 -0.032). There was also a signficant difference between the volume of core lesions at different times after injury (p <0.001). The core lesion volume was significantly larger at weeks four and five than at weeks one and two after injury. At six and seven weeks after injury, the core lesion volume was also significantly larger than during the first three weeks, and week 11 after injury (p = 0.001 -0.032) (ǠFig. 4). The echogenicity score was significantly higher at week 10 than during the first four weeks and at week 12 than during the first eight  Twenty-four hours later, a hypoechogenic rather than anechogenic core lesion was present in the same SDFT. The core defect is well defined from the normal sur-rounding tissue. There was minimal evidence of peritendinous oedema or swelling. C) Seven days after injury of the same SDFT, this core lesion was hypoechogenic and slightly larger, and there was mild to moderate peritendinous oedema. D) At four weeks, the same core lesion is more hypoechogenic and enlarged. Mild to moderate peritendinous oedema is present.
weeks of the study (ǠFig. 5). The linear fibre score was significantly higher at weeks six and 10 than during the first two weeks of the study (ǠFig. 6). It was also higher in week 12 than in the first eight weeks of the study. Using the chi square analysis to look at the distribution of lesion sizes (maximum lesion CSA and lesion volume) at each time point of the study, there was no significant relationship between lesion size and limbs at any time point.

Histology
At two weeks after the surgical injury, lesions at the maximum CSA level were characterised by a central core in which tendon fascicles had been replaced by numerous proliferating fibroblasts, and small caliber blood vessels (ǠFig. 7A). Fibroblasts could be observed migrating from the connective tissue around the pre-existing, mostly intact tendon fascicles at the periphery of the core lesions. There were only a small number of inflammatory cells present, which were predominantly haemosiderin-laden macrophages and lymphocytes. Small scattered areas of haemorrhage were also observed centrally. Towards the periphery of core lesions, proliferating and migrating fibroblasts and small blood vessels were mostly confined to the endotenon around intact tendon fascicles. Any remaining necrotic tendinous tissue resulting from surgical injury appeared to have been removed and fibroblastic repair constituted the major component of the lesion at this stage. Two-week-old core lesions were classified as mid stage 3 (late inflammatory stage), based on the absence of acute inflammatory cells and persistence of only mild chronic inflammation. At four weeks after injury, minimal evidence of inflammation was confined to the perivascular spaces at the junction between the core lesion and the peripheral, largely intact tendon fascicles at the margin of the lesion. Compared to two-week-old core lesions, fibroblast numbers were moderately reduced at the centre of the core, although mitotic figures were still observed. Fibroblast density at the periphery of core lesions was similar to that observed in two-weekold lesions. There was increased deposition of non-collagenous matrix and immature collagen fibres (ǠFig. 7B) but fibres were much smaller than adjacent normal caliber tendon fibres. In less severely affected areas at the periphery of core lesions adjacent to normal tendon tissue, there was some organisation in collagen fibre orientation. Four-week-old core lesions were classified as late stage-3, based on the minimal evidence of inflammation. The appearance of eight-week-old core lesions was similar to that of four-week-old lesions, except for a mild increase in intercellular deposition of non-collagenous matrix and immature collagen at the centre of the core lesion, which resulted in more separation and less densely packed fibroblast nuclei in the healing tissue (ǠFig. 7C). There was still robust infiltration of the peripheral areas of the core lesion with fibroblasts and small caliber blood vessels. These blood vessels were larger, including differentiated ve- Core lesion volume was significantly larger at weeks four and five than at weeks one and two after injury. At six and seven weeks after injury, the core lesion volume was also significantly larger than during the first three weeks as well as at week 11 after injury. Each time point is identified by a letter below the line.  Table 1 The average tendon volume, average core lesion volume, average maximum cross-sectional area of the core lesion as a percentage of the cross-sectional area of the tendon, and average lesion length with standard deviations at different time points of the study following surgical induction of the lesions.
nules and arterioles in addition to capillaries. Lesions were classified as early stage-4 (fibroplasia) based on the increasing evidence of fibrogenesis. At 12 weeks after injury, there appeared to have been only a small increase in the amount of noncollagenous matrix and immature collagen compared to four-and eight-week-old lesions. There was increased collagen fibre thickness and some improvement in collagen fibre orientation at the periphery of lesions ( ǠFig. 8). Core lesions were still largely composed of fibroblasts with no reconstitution of normal tendon architecture. Twelve-week-old core lesions were classified as mid stage-4.

Discussion
Surgical models of tendonitis have been explored sporadically as alternatives to enzymatically induced tendon injuries in horses (19)(20)(21). These models have relied on either surgical removal of a segment of the SDFT by incising the paratenon and the outer surface of the tendon, or on complete transection of the SDFT (19)(20)(21). Surgical transection models of tendon injury have also been used experimentally in the Achilles tendon of rabbits and sheep, and in the deep digital flexor tendon of dogs (22)(23)(24). None of these techniques bear much resemblance to biomechanical overload injury as they bypass a major part of the naturally occurring processes such as traumatic fibre rupture, vascular disruption, necrosis, and demarcation of disrupted fibres (21). Such surgical models also maximise extrinsic healing mechanisms as evidenced by ensuing peritendinous fibrosis and adhesion formation, unlike naturally occurring core lesions that are fully contained within the tendon and rely in large part on intrinsic repair (18). The surgical model used in this study was designed in an attempt to mimick natural injury and repair more closely. The minimally invasive approach with a 3.5 mm Razorcut ® Blade allowed us to create traumatic fibre damage restricted to the core of the mid to distal metacarpal region of the SDFT only, and without disruption to the paratenon and the outer surface of the tendon except at the site of instrument entry. However, as the incision was made through the annular ligament while remaining within the confines of the vinculum, the DFTS was not entered during surgery in any limb, and the entry hole into the palmar surface of the SDFT was able to heal rapidly by sealing of the highly vascular vinculum, thereby avoiding any risk of open communication between the core lesion and the digital flexor tendon sheath (DFTS). In two horses, a small area of subcutaneous fibrous thickening developed at the site of incision, but effusion of the DFTS was not observed in any of the horses throughout the study. Peritendinous adhesions and fibrosis did not occur in any of the limbs in this study, suggesting that extrinsic healing played a minimal role in the echogenicity score of core lesions was significantly higher at week 10 than during the first four weeks, and at 12 weeks than during the first eight weeks of the study. The echogenicity score was adapted from a previously reported grading system, in which an echogenicity score of 0 means generally anechoic and an echogenicity score of 5 is normo-echoic (16). Each column is identified by a letter within the bar. The letters above each column indicate that a significant difference exists between that column and the column labelled by that particular letter.

Fig. 6
Mean linear fibre pattern score of core lesions up to 12 weeks after injury. The ultrasonographic linear fibre score was significantly higher at weeks six and 10 than during the first two weeks of the study. It was also higher in week 12 than in the first eight weeks of the study. The linear fibre pattern score was adapted from a previously reported grading system, in which a linear score of healing of these core lesions. Our technique thus produced a core lesion representative of moderate tendonitis, that may be considered suitable for the study of intralesionally delivered therapeutic agents. Even though severe tendonitis has been described as a lesion extending to approximately 40% of the CSA of the SDFT at maximum injury level, we aimed to create a core lesion with a CSA that occupied no more than 10 % of the SDFT at the time of surgery (25). The reasons for this approach were our desire to minimise postoperative pain by creating small lesions and our expectation that lesions would enlarge in the postoperative period through enzymatic degradation initiated by the surgical injury, as has been described for naturally occurring tendonitis (4). However, in a pilot study using this model, we failed to observe any clinicals signs of tendonitis or lesion enlargement in two horses that were confined to a stall without activity after surgical injury. Consequently it was decided to supplement the initial surgical injury with postoperative exercise in order to create biomechanical overload on the remaining healthy tendon fibres, thereby causing mechanical disruption around the surgically created core defect, lesion enlargement and clinical evidence of tendonitis. The extent of exercise needed to cause mild clinical signs of tendonitis varied between horses from two to seven days of lunging and only resulted in mild lameness and mild peritendinous swelling. Exercise caused a 50 % enlargement in the lesion CSA within one week and resulted in a maximum lesion size ranging from 13 to 24% of the total CSA at four weeks, upgrading all lesions from mild to moderate tendonitis. Therefore, the current level of severity of our model may still pose a problem for use in efficacy studies of various treatments of tendonitis. Several authors have shown the severity of tendonitis to be the main factor affecting clinical outcome as compared to the choice of treatment (26,27). It follows that a consistent model of more severe tendonitis with a larger core lesion size may be required in order to show a significant treatment effect when testing a therapeutic modality for tendonitis.
There was good agreement in the ultrasonographic appearance of surgically induced core lesions with regard to lesion size, lesion echogenicity and linear fibre pattern at different time points (ǠFig. 3 to 6), although in view of the small number of horses used in the study, it is difficult to make strong statistical conclusions with regard to a larger population. The ultrasonographic progression of core lesions in this study showed similarities with that reported for naturally occurring and collagenase-induced tendonitis, although variation exists in the timing of different stages of healing between different reports, de- tered areas of haemorrhage (*) were also observed centrally. B) At four weeks, lesions were characterised by increased fibroblast density (*) and pronounced interfascicular septa (black arrow) sprouting capillary proliferation (white arrows). C) The appearance at eight weeks was similar to that of four-week-old lesions except for more intercellular ground substance deposition resulting further separation of fibroblast nuclei (black arrows) and proliferation of larger blood vessels (white arrows).
pending on the severity and management of lesions (5,6,18,(28)(29)(30). Histologically, the reaction of tendon tissue to injury follows the general phases of wound healing (21). In our study, the histological characteristics of two-to 12-week-old lesions were condensed between the late inflammation phase and slow progression of the fibroplasia phase. Surprisingly, no signs of acute trauma and obvious inflammation were present in the limbs of the horse that was euthanatized at 14 days in our study, even though mild clinical tendonitis had not become evident until day seven after surgical injury when this horse had been exercised for six days. The absence of obvious inflammation at any stage of the study makes it difficult to explain how lesions continued to enlarge ultrasonographically until six weeks. A similar observation was made in a surgical excision model (21). The lack of inflammation in our model was not consistent with histological reports of both naturally occurring tendonitis nor collagenase-induced tendonitis in horses, in which signs inflammation were visible as late as three and six weeks after injury respectively (5,6,18,28). This suggests that our surgical model may be less inflammatory and therefore closer to the condition of human tendinosis (31). Although the number of horses was too small to allow for strong conclusions, it is possible that the acute phase of necrosis and inflammation resolved quickly in our surgical model as part of the torn tendon fibres were mechanically removed by a combination of suction and resection at the time of surgery.
Even though the reparative process began early in this model, there was minimal reconstitution of normal tendon architecture by 12 weeks. Other authors have observed a similarly slow time line for progression of healing in naturally occurring and experimental tendon injuries in horses (5,6,18,19,21). Normal tendon architecture was missing at six months in naturally occurring tendonitis, five and six months in surgically created defects and 10 and 14 months in collagenase-induced tendon injuries (5,6,18,21,32). It follows that the 12-week follow-up period used in our study was too short to determine the final outcome of the healing process, especially when trying to establish the efficacy of new therapies for tendonitis.
Limitations of this study were the small number of horses and the relatively short time period in which the surgically induced lesions could be followed. This resulted in a smaller number of data points being available for the later stages of the study. Therefore it would be useful if the current results could be confirmed in a larger number of horses with larger core lesions and a longer follow-up time. The surgical technique used in this study was able to induce standardised core lesions in the SDFT that showed similarities in ultrasonographic progression and histological characteristics to core lesions in naturally occurring tendonitis. At the same time, this model does not represent certain aspects of natural injury such as the effect of mechanical overload on the central injury region and the presence of a more obvious inflammatory phase of healing. In spite of these limitations, this model may prove useful for evaluation of therapies aimed at improving tendon healing in the horse.