Effects of incision closure method on infection prevalence following tibial plateau leveling osteotomy in dogs

Abstract

The goal of this study was to retrospectively investigate the effect of incisional closure with either stainless steel skin staples or intradermal poliglecaprone 25 on the prevalence of surgical site infection following tibial plateau leveling osteotomy in dogs. Medical records were reviewed for dogs treated with unilateral tibial plateau leveling osteotomy at Memphis Veterinary Specialists between 2006 and 2013. Procedures (n = 306) from 242 dogs were included in the study. The association of potential risk factors with the occurrence of postoperative infection was assessed using logistic regression. A value of P < 0.05 was considered significant. Weight and administration of postoperative antimicrobials were found to significantly influence surgical site infection prevalence. No significant association was noted between closure method and prevalence of postoperative infection.

Introduction

Surgical site infections can be a significant cause of postoperative morbidity and mortality, they may delay wound healing and cause additional financial burden for clients (1,2). Infection prevalence for clean orthopedic surgical procedures in dogs ranges from 3.6% to 10.0% (2–5). Reported risk factors for surgical site infection/inflammation include duration of anesthesia, duration of postoperative intensive care unit stay, wounds classified as “dirty” (based on the United States National Research Council wound category), number of persons in the operating room, patient weight, and concurrent endocrinopathies (6,7). Of the various techniques available to treat cranial cruciate ligament injury in dogs, tibial plateau leveling osteotomy (TPLO) is one of the most common and well-described. Numerous risk factors have been described for postoperative infection after TPLO procedures; however, the infection prevalence remains high and ranges from 3.0% to 13.0% (2,3,8–10).

One of the potential risk factors for postoperative infection following TPLO surgery is the method of incision closure, and 2 commonly used closure methods include stainless steel skin staples and absorbable intradermal suture. Reported advantages of an intradermal suture pattern in human studies include decreased incidence of wound disruption and smaller scar width (11,12). Advantages of stainless steel skin staples include reduced surgical time for closure, reduced pain, and improved overall cosmesis (13–15). Several studies in human medicine have compared the infection prevalence of incisions closed with either stainless steel skin staples or intradermal suture, with often mixed results (11,16–18). Closure method can affect the ease with which the wound is disrupted (11), and can also affect inflammation of the local area (19), which could affect the ability of the wound to resist infection. Some studies have reported an increased prevalence of infection in wounds closed with skin staples (11,16). Other studies have found no difference in the prevalence of infection in wounds closed with skin staples or intradermal suture (17,18). In a recent study in dogs, stainless steel skin staples were found to have a lower incidence of inflammation than external skin sutures, but an intradermal suture pattern was not directly evaluated (19).

The objective of the present study was to compare prevalence of surgical site infection for stainless steel skin staples and intradermal poliglecaprone 25 for incisional closure in dogs undergoing elective unilateral TPLO. We hypothesized that the use of intradermal poliglecaprone 25 would result in lower surgical site infection prevalence after TPLO procedures compared with stainless steel skin staples.

Materials and methods

Case selection

Medical records for dogs that underwent TPLO procedures at Memphis Veterinary Specialists Referral Hospital between 2006 and 2013 were reviewed. For all dogs, the diagnosis of cranial cruciate ligament (CCL) rupture was made on the basis of rear limb lameness, pain on extension of the stifle, palpable effusion, presence of cranial drawer motion, cranial tibial thrust, and radiographic evidence of arthritis and joint effusion. For incisional closure, either stainless steel skin staples (Covidien, Mansfield, Massachusetts, USA) or poliglecaprone 25 (Ethicon, Somerville, New Jersey, USA) in a continuous intradermal pattern was used at the discretion of the surgeon. To be eligible for the study, dogs had to be evaluated at Memphis Veterinary Specialists or by the referring veterinarian at 10 to 14 d after surgery and until radiographic evidence of osteotomy healing was obtained. Dogs that had concurrent procedures, such as mass removals, performed at the time of TPLO were excluded from the study. Dogs that had TPLO procedures performed on both rear limbs were included in the study if the procedures were performed under different anesthetic episodes.

Medical records review

Continuous data recorded included age, body weight, duration of surgery (the time from initiation of skin incision to the end of incisional closure), and duration of anesthesia (the time from intubation to extubation). Categorical data recorded included:

• experience level of the surgeon (resident or ACVS diplomate),

• presence of a preoperative infection (e.g., urinary tract, skin, or other location),

• induction agent (propofol or other),

• whether any prior surgeries had been performed on the stifle,

• whether or not a TPLO jig was used during the procedure,

• whether or not an arthrotomy was performed (for CCL or meniscal evaluation),

• type of TPLO plate used,

• method of skin closure,

• development of surgical site infection,

• whether postoperative antimicrobials were administered,

• whether an impermeable iodine-impregnated adhesive drape (3M Health Care, St. Paul, Minnesota, USA) was used,

• whether an impermeable leg drape (GEPCO, Philadelphia, Pennsylvania, USA) was used,

• bacterial species isolated from aerobic culture and sensitivity of draining tracts or removed implants (performed at the discretion of the attending clinician), and

• whether or not the TPLO plate was removed.

Various definitions of surgical site infection have been described in the literature. Some studies have classified surgical sites as infected if purulent drainage, an abscess, or a fistula was recorded or when 3 or more of the following were present: redness, swelling, signs of pain, heat, serous drainage, or incisional dehiscence (7,10). Another study defined an infection as an incision with 1 or more of the following: a positive culture, drainage > 48 h after the end of surgery, abscess, fistula, or dehiscence (19). In the present study, the outcome variable of surgical site infection was defined based on recent veterinary studies (3,4,19) that are an adapted form of the definitions of surgical site infection put forth by the United States Center for Disease Control and Prevention (20). A surgical site was considered infected with the presence of purulent discharge, an abscess or fistula, dehiscence of the incision, or with pain on palpation of the surgical site with any combination of the following: redness, swelling, heat, or fever, or when a positive culture result was obtained. A second outcome variable of plate removal was considered. Patients that presented with draining tracts or abscesses had samples of wound exudate aseptically collected with a cotton swab and submitted in a culturette. Patients that underwent implant removal had a screw from the plate submitted in a culturette in addition to a tissue sample collected from beneath the plate.

Anesthetic and surgical procedures

In all dogs surgical anesthesia was induced with either propofol (Propoflo; Abbott, North Chicago, Illinois, USA), 6 mg/kg body weight (BW), IV, to effect or a combination of ketamine (Ketamine HCl; Putney, Portland, Maine, USA), 5 mg/kg BW, IV, and diazepam (Diazepam; Hospira, Lake Forest, Illinois, USA), 0.25 mg/kg BW, IV, followed by intubation and maintenance of anesthesia with either isoflurane (IsoFlo; Abbott) or sevoflurane (SevoFlo; Abbott) in oxygen. All dogs received cefazolin (Cefazolin for Injection; WG Critical Care LLC, Paramus, New Jersey, USA), 22 mg/kg BW, IV before surgery and then every 90 min during anesthesia, followed by injections every 8 h for the first 24 h after surgery.

The affected rear limbs were clipped and aseptically prepared after induction of anesthesia. The TPLO procedure was performed by 1 of 6 surgeons (4 ACVS Diplomates, 2 residents) based on a previously described procedure (21) with minor variations depending on preferences of the attending clinician. Plate type chosen was based on surgeon preference and was either a non-locking plate (VOI, St. Augustine, Florida, USA) or a locking plate (Synthes Vet, Paoli, Pennsylvania, USA). The surgical sites were closed in 3 layers. In all cases, the pes anserinus and subcutaneous tissue were closed with polydioxanone (PDS II; Ethicon) suture in a simple continuous or simple interrupted pattern. The skin was closed with intradermal poliglecaprone 25 (Monocryl; Ethicon) or stainless steel skin staples (Appose ULC; Covidien).

Administration of antimicrobials and anti-inflammatory medications after surgery was based on surgeon’s preference. Antimicrobials, if prescribed, were administered for 7 d after surgery. Patients that received postoperative antimicrobials were administered cephalexin (Cephalexin Capsules; Karalex Pharma, Princeton, New Jersey, USA), 22 mg/kg BW, PO, q12h, cefpodoxime (Cefpodoxime Proxetil; Putney), 10 mg/kg BW, PO, q24h, enrofloxacin (Baytril; Bayer Healthcare, Shawnee Mission, Kansas, USA), 10 mg/kg BW, PO, q24h, or amoxicillin/clavulanate (Clavamox; Pfizer Animal Health, New York, New York, USA), 13.75 mg/kg BW, PO, q12h. All patients received either tramadol (Tramadol Hydrochloride; Amneal Pharmaceuticals, Glasgow, Kentucky, USA), 4 to 6 mg/kg BW, PO, q8h or transdermal fentanyl (Fentanyl Transdermal; Apotex Corp, Weston, Florida, USA), and carprofen (Rimadyl; Pfizer Animal Health), 4.4 mg/kg BW, PO, q24h, firocoxib (Previcox; Merial, Duluth, Georgia, USA), 5 mg/kg BW, PO, q24h, or meloxicam (Metacam; Boehringer Ingelheim Vetmedica, St. Joseph, Missouri, USA), 0.1 mg/kg BW, PO, q24h for postoperative pain control. All incisions were covered with an adhesive bandage and an ointment containing bacitracin, neomycin, and polymixin B (Triple Antibiotic Ointment; Taro Pharmaceuticals, Hawthorne, New York, USA) after surgery, and a modified Robert-Jones bandage was applied for the first 12 to 24 h after surgery. All patients were discharged with instructions for strict cage confinement for the first 6 wk after surgery, followed by a rehabilitation regime after evidence of osteotomy healing was obtained on radiographs. At a minimum, patients were evaluated at 10 to 14 d after surgery for an incision check/staple removal and again at 6 wk after surgery for radiographs. Patients diagnosed as having surgical site infection were seen for recheck examinations until resolution of clinical signs.

Sample size

A total of 306 cases met the criteria and used either stainless steel skin staples or intradermal poliglecaprone 25 sutures. Assuming an alpha level of 0.05, a power of 0.80, and the postoperative infection prevalence of 8.9% for intradermal poliglecaprone 25 sutures found in the present study, the sample size was sufficient to detect increased odds of infection of 2.4 or greater for stainless steel skin staples compared to intradermal poliglecaprone 25 sutures (22).

Statistical analysis

Histograms using PROC UNIVARIATE were used to evaluate the distribution of the continuous variables of dog age, dog body weight, duration of surgery, and duration of anesthesia. PROC TABULATE was used to calculate the appropriate descriptive statistics by infection status and plate removal status for these continuous variables. The frequencies of variable responses by infection status and plate removal status were calculated using PROC FREQ for the categorical variables.

The association of potential risk factors with the 2 outcome variables, postoperative surgical site infection and plate removal, were assessed using logistic regression (PROC LOGISTIC; SAS for Windows v 9.3; SAS Institute, Cary, North Carolina, USA). The association of the outcome with each explanatory variable was assessed individually in a univariable analysis. Explanatory variables that were associated with the outcome with a likelihood ratio test P-value ≤ 0.25 were retained as candidates for a multivariable model. Colinearity among candidate variables that were continuous was evaluated by assessing correlation using PROC CORR. Variables that had a Pearson’s Correlation Coefficient ≥ 0.80 were not included in the model at the same time. The multivariable model was developed by first including all the candidate variables in the model and then following a manual backward selection process. This was accomplished by fitting the model and then removing the explanatory variable with the highest P-value. The model was refit and the process continued until only explanatory variables with Wald Chi-Square P-values ≤ 0.05 remained in the model. Interactions among remaining main effects were assessed and retained in the model if found to be significant. The fit of the final multivariable model was evaluated by the Hosmer and Lemeshow goodness-of-fit test. The predictive power of the final model was evaluated by the area under the Receiver Operating Characteristic (ROC) curve. Additional analyses were undertaken to determine if there were associations between suture material and the explanatory variables in the multivariable model. A two-sample t-test was conducted using PROC TTEST to compare mean body weight for the dogs in which the different types of suture material were used. A Chi-square test was conducted using PROC FREQ to compare the proportions of postoperative antibiotic use between the dogs in which the different suture materials were used. An alpha level 0.05 was used to determine the statistical significance of associations.

Results

Dogs (n = 242) met the inclusion criteria for the study, with 64 of those having TPLO procedures performed on both stifles under different anesthetic episodes. The overall prevalence of infection was 10.8% (33/306), and the overall plate removal prevalence was 2.9% (9/306). Seventeen stifles were considered to have surgical site infection based on positive culture results. The remaining 16 were treated as surgical site infection based on clinical signs alone. Sixteen patients presented with draining tracts, and 1 presented with an abscess at the incision. Six of the 9 patients that had the TPLO plate removed had positive bacterial growth. The other 3 cases in which the plate was removed yielded negative culture results of the implants and surgical sites. Of the patients with surgical site infection, 18.2% required plate removal (6/33).

Results of the univariable analysis for the postoperative infection outcome are listed in Table 1 and the plate removal outcome in Table 2. Dog body weight, surgery duration, and anesthesia duration were approximately normally distributed. Although the ages of dogs were also approximately normally distributed, the distribution was somewhat more skewed than the other variables. The mean age of dogs at the time of surgery was 5.5 ± 2.6 y [± standard deviation (SD)]; the median age was 5.0 y [Interquartile range (IQR) 4.0 y]. The mean body weight at the time of surgery was 33.1 ± 11.5 kg. Increasing body weight was significantly associated with increased odds of postoperative infection. Body weight also had a statistically significant association with plate removal in the univariable analysis. The mean surgical and anesthesia times were 74.4 ± 23.1 min and 181.8 ± 31.3 min, respectively. All patients received perioperative antimicrobials.

Table 1.

Association of each variable with postoperative infection prevalence

Parameter	Unit or referent	OR	95% CI	P-value
Continuous variables
Age	1 year	0.88	0.76–1.02	0.08
Body weight	kg	1.04	1.01–1.08	0.01
Surgical time	min	1.01	0.99–1.02	0.35
Anesthesia time	min	1.00	0.99–1.01	0.85
Categorical variables
Resident	ACVS	1.14	0.41–3.14	0.80
Propofol for induction	Other	1.49	0.59–3.77	0.40
Preoperative infection	None	2.63	0.69–10.09	0.16
Jig	None	1.34	0.64–2.78	0.44
Arthrotomy	None	1.08	0.52–2.24	0.85
Adhesive drape	None	0.94	0.43–2.06	0.87
Leg drape	None	0.94	0.43–2.06	0.87
Poliglecaprone 25	Staples	0.58	0.28–1.21	0.14
Postoperative antibiotics	None	0.44	0.21–0.93	0.03
Plate type	VOI	1.77	0.84–3.69	0.14

OR — odds ratio, CI — confidence interval, ACVS — American College of Veterinary Surgeons.

Table 2.

Association of each variable with plate removal outcome

Parameter	Unit or referent	OR	95% CI	P-value
Continuous variables
Age	1 year	0.73	0.52–1.01	0.06
Body weight	kg	1.06	1.00–1.11	0.04
Surgical time	min	1.01	0.98–1.04	0.47
Anesthesia time	min	1.00	0.98–1.02	0.86
Categorical variables
Resident	ACVS	1.84	0.37–9.15	0.46
Propofol for induction	Other	0.63	0.15–2.58	0.52
Preoperative infectiona	None
Jig	None	0.49	0.10–2.41	0.38
Arthrotomy	None	0.40	0.08–1.98	0.26
Adhesive drape	None	0.50	0.13–1.91	0.31
Leg drape	None	0.50	0.13–1.91	0.31
Poliglecaprone 25	Staples	1.83	0.37–8.97	0.46
Postoperative antibiotics	None	1.71	0.42–6.98	0.45
Plate type	VOI	0.26	0.03–2.15	0.15

^aOdds ratio (OR) was not estimable because none of the dogs that had a plate removed had a preoperative infection.

CI — confidence interval; ACVS — American College of Veterinary Surgeons.

The infection prevalence for TPLO procedures performed by residents and ACVS diplomates were 11.9% (5/42) and 10.6% (28/264), respectively. The infection prevalence for dogs induced with propofol was 11.6% (27/232) compared with 8.1% (6/74) in dogs that were not induced with propofol. Dogs with a preoperative infection at a site other than the surgery site had an infection prevalence of 23.1% (3/13) compared with 10.2% (30/293) for those without a preoperative infection. Of those dogs, 8 were treated for pyoderma, 1 was treated with antibiotics for dental disease, 1 was treated for a urinary tract infection, and 3 had previously removed extra-capsular sutures that were treated for infection based on culture results of those implants. A TPLO jig was used in 36.3% of the TPLO procedures (111/306). Cases in which a jig was used had an infection prevalence of 12.6% (14/111) compared with 9.7% (19/195) when a jig was not used. An arthrotomy was performed in 40.8% of patients (125/306). Patients that had an arthrotomy had an infection prevalence of 11.2% (14/125) compared with 10.5% (19/181) in patients that did not receive an arthrotomy. Infection prevalence rates for patients that did or did not have an iodine-impregnated impermeable adhesive drape or an impermeable leg drape placed over the operated limb and surgical site were 10.6% (23/217) and 11.2% (10/89), respectively. The prevalence of infection for VOI TPLO plates was 8.5% (17/199), while the infection prevalence for Synthes plates was 14.2% (15/106).

Poliglecaprone 25 in a continuous intradermal pattern was used for incisional closure in 66.0% of TPLO procedures (202/306), and stainless steel skin staples were used for incisional closure in 34.0% of cases (104/306). Infection prevalence was 8.9% (18/202) in cases in which poliglecaprone 25 was used for incisional closure compared with 14.4% (15/104) for cases in which stainless steel skin staples were used.

Postoperative antimicrobials, administered more than 24 h after surgery, were used in 54.2% (166/306) of patients. The use of postoperative antimicrobials was significantly associated with decreased odds of postoperative infection. The surgical site infection prevalence was 7.2% (12/166) in patients that were administered postoperative antimicrobials compared with 15.0% (21/140) in patients that were not administered antimicrobials postoperatively. Cefpodoxime was used in 150 cases, and 6.0% (9/150) developed surgical site infections. Amoxicillin/clavulanate was used in 13 cases, and 23.1% (3/13) developed surgical site infections. One patient was administered cephalexin, and 2 were administered enrofloxacin. None of these developed surgical site infections.

Cultures were collected and submitted between 10 and 330 d after surgery (median 22 d). Bacteria cultured from infected surgical sites in order of decreasing frequency were methicillin- resistant Staphylococcus pseudintermedius (MRSP; 9 dogs), Staphylococcus pseudintermedius (6 dogs), Corynebacterium sp. (1 dog), and Staphylococcus aureus (1 dog). Of the surgical site infections with available culture results, patients that were administered postoperative antimicrobials developed only 5 MRSP infections.

Multivariable analysis

Variables from the univariable analysis that met the screening criteria and were retained as candidates for a multivariable model included body weight, use of postoperative antibiotics, age of dog, presence of preoperative infection, plate type, and suture material used for incisional closure. In addition to body weight, dog age, plate type, and performing an arthrotomy also met the screening criteria and were considered as candidates for a multivariable model for plate removal outcome. However, multiple variables were not retained during the model development process. A multivariable model was developed that included body weight and postoperative antibiotics for the outcome variable of surgical site infection (Table 3). Body weight was associated with increased odds of infection. Postoperative antibiotics were associated with decreased odds of infection. The Hosmer and Lemeshow Goodness-of-Fit Test indicated the model fit the data (P = 0.80). The area under the ROC curve was 0.69 suggesting the discriminatory ability of the model was fairly poor (23). An interaction term between body weight and use of postoperative antibiotics was added to the multivariable model but was not significant (P = 0.443) and was not retained. Although the variable describing the material used for incisional closure was not retained in the multivariable model, additional analyses were undertaken to determine if there were associations between suture material and the explanatory variables in the multivariable model. The mean body weight of dogs in which intradermal poliglecaprone 25 sutures were used (32.3 ± 11.5 kg) was not significantly different (P = 0.089) from the mean weight (34.7 ± 11.4 kg) of dogs in which stainless steel sutures were used. The proportion of cases in which postoperative antibiotics were used was 52.5% (106/202) and 57.7% (60/104) for dogs in which intradermal poliglecaprone 25 sutures and stainless steel sutures were used, respectively. There was no significant difference between the 2 proportions (P = 0.386).

Table 3.

Multivariable model for occurrence of surgical site infection following tibial plateau leveling osteotomy in dogs

Variable	OR	95% CI	P-valuea
Body weight (in units of 1 kg)	1.05	1.02–1.09	0.003
Post-operative antibiotics (yes versus no)	0.37	0.17–0.80	0.012

^aProbability of a greater than Wald Chi square.

OR — odds ratio; CI — confidence interval.

Discussion

The main finding of this study was that a statistically significant association was not detected between postoperative infection prevalence in TPLO surgical sites and closure with stainless steel skin staples compared with intradermal poliglecaprone 25 sutures. Our findings are similar to studies performed in human medicine that showed no difference in infection prevalence between incisions closed with either suture or staples (17,18,24).

The overall infection prevalence for this study was 10.8%, which is comparable to the prevalence of infection following TPLO procedures in previous studies (8,19,25). One factor that could falsely elevate the infection prevalence is the inherent difficulty in differentiating between inflammation and true infection of surgical sites in the postoperative period. It is possible that many of the cases that were treated as having a surgical site infection were not truly infected. This could be supported by the low plate removal prevalence in our study (2.9%; 9/306), which is comparable to the 4.8% plate removal prevalence found in another study (26). In the present study, 81.8% (27/33) of cases classified as infected had resolution of clinical signs without plate removal. This could suggest that many of the infections resolved with oral antimicrobial administration, or that not all of those cases were truly infected. It is also possible, however, that many of these cases were still harboring subclinical infections that had not manifested by the end of the study period.

Patient body weight was significantly associated with the postoperative surgical site infection prevalence. Patients with a greater body weight had a greater prevalence of surgical site infection, odds ratio (OR) = 1.04, P = 0.01. Body weight was also associated with an increased odds of infection (OR = 1.05, P = 0.003) in the multivariable model, which included postoperative antibiotics as well. These findings are similar to two studies in which increasing body weight was found to increase the risk of surgical site infections (3,7), but in contrast to previous studies (10,19) that did not find a significant correlation between body weight and surgical site infections following TPLO. One limitation to this finding is that body condition score was not recorded for most patients at our hospital; therefore, a correlation between surgical site infection and obesity or malnutrition cannot be made.

In our study, patients that were administered postoperative antimicrobials had a lower incidence of surgical site infection than those that did not receive them in a univariable model (OR = 0.44, P = 0.03) as well as a multivariable model (OR = 0.37, P = 0.01) which included body weight. This finding is similar to that of previous studies (7,10,27) which found a decreased infection prevalence in patients administered oral antimicrobials in the postoperative period. One interesting finding in the present study is that all patients that received oral antimicrobials and developed surgical site infection were infected with methicillin-resistant Staphylococcus species. This finding is in agreement with previous studies that show that antimicrobial use carries an increased risk of developing multi-drug and methicillin resistant infections (28,29). Therefore, while the surgical site infection prevalence was lower with administration of oral antimicrobials, this benefit must be weighed against the risk of developing multi-drug resistant infections.

Our study did not show a significant difference in surgical site infection prevalence between patients that were induced with propofol or other induction agents. This is in contrast to another study (30), which showed an increased prevalence of infection with administration of propofol. There was not a significant association between TPLO plate type and surgical site infection prevalence, and this is in agreement with another study that found no difference in infection prevalence between plate types used (26). Another interesting finding in our study is the lack of association between total anesthetic and surgical times and postoperative surgical site infection prevalence. This is in agreement with a previous study (19), but in disagreement with previous studies that found that surgical and anesthesia times were longer in patients that developed postoperative surgical site infections (6,7). One potential explanation is the lower sample size in our study (n = 306) compared with the study by Eugster et al (7) in which n = 1010. In light of this finding, further investigation is warranted.

The limitations of this study include those inherent with any retrospective study and the reliance on medical records that sometimes are vague and incomplete. Body condition scores were not recorded for most patients, and this variable could have had a significant impact on postoperative surgical site infection prevalence. Culture results were not available for all cases of suspected infection, and this could potentially have influenced the overall infection prevalence and treatment recommendations for affected patients. The presence of concurrent diseases that could have affected the development of surgical site infection, including endocrinopathies, was not always recorded. Also, patients could have been evaluated and treated for suspected infections by referring veterinarians, and this information may not have been present in the medical record.

In conclusion, no significant difference in the prevalence of postoperative surgical site infection following TPLO surgery was found between incisions closed with intradermal poliglecaprone 25 and those closed with stainless steel skin staples. Patient body weight and the administration of postoperative antimicrobials significantly influenced the prevalence of TPLO surgical site infection. Limitations associated with the case definition of surgical site infection in the present study could have affected these findings. Further studies are indicated to evaluate the effects of incisional closure method on surgical site infection prevalence following TPLO in dogs. CVJ