Does Primary Closure Increase Surgical Site Infection after Intestinal Stoma Reversal? A Retrospective Cohort Study

Patients and Methods

This retrospective cohort study was conducted at the Department of Surgery, The Aga Khan University Hospital, Karachi, Pakistan. A list of patients, who underwent reversal of stoma (ileostomy and colostomy) from January 2005 to May 2011 was retrieved using International Classification of Diseases (ICD) codes 46.51 and 46.52 from the Hospital Information Management System (HIMS), and consecutive files were reviewed until the specified sample size of 71 in both groups (incisions left open and incisions closed primarily) was achieved. Adult patients (15–80 years of age) who underwent elective reversal of loop or double-barreled stomas through peristomal incisions were included. Exclusion criteria were reversal of stoma through the midline (i.e., Hartmann pouch reversal); conversion of peristomal reversal to laparotomy; other concomitant abdominal surgery (i.e., incisional hernia repair, definite tumor surgery); perioperative steroid use; post-operative course complicated by anastomotic leak; missing or incomplete records; and any histopathology-proved malignant tumor.

The main exposure variable was surgical site management. Following review of the operative notes, each patient was classified as either exposed (surgical sites closed at the time of reversal) or unexposed (surgical sites left open for healing by secondary intention). The outcome variable was SSI, assessed according to criteria of the U.S. Centers for Disease Control and Prevention (CDC) [14] within at least 30 d post-operatively or until complete healing of the site.

The stoma closures were done using an almost uniform technique. Patients were started on a clear liquid diet one day before surgery, and a purgative (polyethylene glycol solution) was given for clearance of the bowel. Oral antibiotics were not used during preparation. Intravenous prophylactic antibiotics, mostly triple (ceftriaxone, ampicillin, and metronidazole) and sometimes others (amoxicillin-clavulanic acid, amikacin, or piperacillin) were given at the time of induction.

After preparation of the skin with 10% povidone-iodine, an elliptical incision was created with a scalpel around the stoma and dissected free off the body wall. Anastomosis was done with either sutures (hand sewn) or a stapler. In a stapled anastomosis, an end-to-end functional bowel anastomosis was done with the Gastro Intestinal Anastomosis (GIA) stapler (Ethicon, Somerville, NJ), and the ends of the bowel were closed with a GIA or Transverse Anastomosis (TA) stapler (Ethicon). The muscle/sheath was closed with looped polydioxonone (PDS) suture (Ethicon). The site was irrigated and either left open (packed with 10% povidone-iodine soaked gauze) or closed with interrupted non-absorbable monofilament synthetic sutures with or without a drain according to the surgeon's preference.

Postoperatively, antibiotics were discontinued and a Penrose drain (if placed) was removed within seven days. Daily surgical site assessment was done by the chief resident of surgery (Year 5 resident), a consultant surgeon, or both until discharge of the patient from the hospital. Patients were discharged once they were mobilized, passing flatus/stool, and tolerating a soft diet. After discharge, there was a weekly followup at the outpatient clinic by the consultant surgeon until the incision was completely healed. If SSI was found in closed incisions the sutures were removed, and dressing was done daily by the patient, family, or a visiting home health care provider until the site had healed completely.

The World Health Organization (WHO) software was used to calculate the appropriate sample size. Based on results of Harold et al. (i.e., SSI of 8.3% in closed incisions and 0% in open incisions), a sample of 71 subjects per group was required at the 5% level of significance (one-sided) and 80% power.

The study protocol was approved by the Ethical Review Committee (1872-Sur-ERC-11) prior to the chart review. Data from the patient's file were recorded on a questionnaire. To obtain uniform and reliable data, the variables were grouped according to the most reliable source of information (i.e., the operative note, pre-operative assessment form, operating room record, or other notes in the record). An initial assessment form was provided to reviewers to scrutinize the patients for eligibility. Files were reviewed independently by two persons, a resident (year 4) and a fellow in general surgery. When a discrepancy was found, the file was reviewed jointly by the two surgeons and a consensus arrived at.

Data were checked for aberrant codes and internal consistency with frequency tables and cross-tabulations. Continuous variables were analyzed as means±standard deviation for data with normal distribution and median with interquartile ranges for skewed data. Categorical variables were analyzed as proportions and percentages for each exposure group. The rate of SSI was compared in the groups by the χ² test or the Fisher exact test, whichever was applicable. Subgroup analysis was done to compare SSI in incisions closed over a Penrose drain with those closed without a drain. The kappa test was used to determine the agreement between the reviewers for the main outcome variables.

Covariates were analyzed with the Cox proportional hazard regression by the purposeful selection technique [15]. Briefly, the initial multivariable model consisted of statistically significant variables at a 20% level of significance (on univariable analysis), as well as clinically meaningful (but statistically insignificant) variables. After removing insignificant variables and checking for interaction terms, the final model comprised the main exposure variable, significant confounders, and significant interaction terms. The association of exposure; i.e., incision closure with SSI is reported as adjusted relative risk (ARR) with the 95% confidence interval (CI).

Results

A total of 271 patients who underwent stoma reversal between January 2005 and May 2011 were identified. The patient files were evaluated consecutively until a sample size of 71 patients in each group was achieved. The groups were similar for the baseline variables (i.e., age, gender, body mass index (BMI), smoking, and diabetes mellitus (DM)) (Table 1). Patients with closed surgical sites were younger (mean 36±15 [standard deviation] years) than those with open surgical sites (41±15 years), with a male preponderance in both groups. However, there was a greater proportion of patients with large bowel stomas, exteriorized stomas, restoration of intestinal continuity with sutures, and peri-operative use of triple antibiotics in the open surgical site group (Table 1). Of 71 patients with primary incision closure, 36 underwent closure over a Penrose drain.

Fifteen patients were found to have SSI; three of 71 (4.2%) in the open and 12 of 71 (16.9%) in the closed incisions category. The incidence proportion of SSI in open incisions was 12.9% less than that in closed incisions (p=0.013). Surgical site infections were assessed by two independent reviewers, and good agreement [16] between the reviewers (kappa=0.71) was observed. Subgroup analysis revealed that the incidence of SSI in incisions closed without a drain was higher than that in the incisions closed over a drain (25.7% vs. 8.3%; p=0.05).

Continuous variables were categorized as age in years (0=≤50, 1=≥51), construction-to-closure period in days (0=<90, 1=≥91), and BMI (0–30, 1=≥31). Surgical site, smoking, anastomosis type, and antibiotics were significant on univariable analysis. In the initial model, these variables were kept along with the main exposure variable; i.e., surgical site closure and clinically important (but statistically insignificant) variables (i.e. age, gender, BMI, pathology, chemoradiotherapy, site of stoma, prophylactic antibiotics, site of stoma, and construction-to-closure interval). All the covariables were insignificant.

After removing the least significant variables, the subsequent model was tested for interaction terms, and none of them was found to be significant. So, the final model (Table 2) consisted of the main exposure variable (surgical site) and significant confounders (site, gender, chemoradiotherapy, BMI, and pathology). After adjusting for confounders, the risk of SSI in closed incisions was 5.8 times greater than that of open incisions (Table 2).

Discussion

The findings and statistical analysis supported the study hypothesis, and it was observed that there was a 12.9% greater incidence of SSI in closed surgical sites than in open incisions, with an RR of 5.8 adjusted for gender, BMI, site of stoma, malignant disease, and chemoradiotherapy.

Several steps were taken to enhance the internal validity of the study. Outcome was assessed according to a standard tool (i.e., the CDC criteria). The data for each variable were reviewed from the single but most reliable source in the patients' record. To obtain homogenous groups regarding procedure and perioperative patient condition (physiologic changes, operative time and blood loss, optimum antibiotic concentration, and pre-operative contamination), only loop and double-barreled stomas were included. Patients with malignant disease (other than the one for which the stoma was formed) and patients receiving steroids during the perioperative period were excluded, as both of these conditions increase the risk of SSI by decreasing site healing and patient immunity [1]. Furthermore, a multivariable analysis with Cox regression to adjust for the effect of confounders was performed. Although 80% power was retained on post hoc analysis of SSI, the multivariable model seems to suffer from over-fitting secondary to the paucity of observations of SSI as well as entry of clinically important (but statistically insignificant) variables.

One of the limitations of this study is differential misclassification bias. A CDC criterion for SSI is opening of the site within 30 days after surgery. Obviously, this is possible only for closed sites, and because closed incisions are inspected and documented vigilantly, there is a greater likelihood that a SSI will be identified. Also, because this was a retrospective review, it suffers from missing-data bias, although this has been reduced by using the CDC criteria, which encompass different aspects of SSI.

The issue of incision closure after stoma reversal is fraught with controversy. Three randomized studies have had divergent results [2,7,8]. Berne et al. [7] conducted a study in 1985 to compare SSI with three techniques of wound management; i.e., closure without a drain (n=38), closure with a drain (n=29), and delayed primary closure (n=38), finding SSI rates of 2.6%, 3.4%, and 7.9%, respectively. These results were confirmed by another randomized study in 2005 by Lahat et al. [2], who compared primary closure with delayed primary closure, placing 20 patients in each arm and reporting 20% and 10% rates of SSI, respectively. Taking the argument forward, Reid et al. [8] compared primary with purse-string closure. Also called circumferential subcuticular incision approximation (CSIA), the latter is a modified way of managing open incisions, whereby the opening is narrowed with a suture. Contrary to the prior two randomized studies, these investigators reported a higher SSI rate with linear site closure compared with CSIA (38.7% vs. 6.6%) with 31 participants in each arm. The issue of stoma closure is embarked on by a few descriptive and retrospective reviews with diverging results [3,4,6,10,12,13,17]; some have reported higher SSI rates in closed than in open incisions [9,13], whereas others have reported the opposite [10,12,18].

Berne et al. [7] and Lahat et al. [2] reported higher SSI rates in closed than in open surgical sites; however, they considered a purulent discharge the sole evidence for SSI, disregarding other features such as tenderness, warmth, redness, and induration. In all the studies in which the CDC criteria were used, including this study, a higher incidence of SSI has been found in closed than in open incisions [3,8,10].

Reid et al. [8] reported a higher SSI with linear site closure than with CSIA (38.7% vs. 6.6%). This modification; i.e., CSIA, has been studied in other retrospective reviews [6,10,12] with no SSIs, making it the method with the lowest reported incidence of SSI. Several retrospective reviews and case series have reported diverse results, although few have found higher SSI rates in closed than in open incisions [10,17,18]. Except in the series from Akiyoshi et al. [3], none of these studies has included a multivariable analysis to establish the independent risk factors after controlling confounders.

Does closure increase the risk of SSI after stoma reversal? The question can be answered using Hill's criteria [19]. A strong association between closure and SSI was found, with a 5.8 ARR. These findings are consistent with those of other studies [3,4,8,10,12,18]. There is a biologic gradient (Tables 3 and 4). The association also is plausible biologically, as closure may lead to incorporation of bowel contents/debris and pathogens in the incision, making it susceptible to inflammation and infection. The findings from experimental designs are diverse; a higher SSI rate in open incisions was reported by Berne et al. [7] and Lahat et al. [2], but Reid et al. [8] reported contrary results. Reid et al. had a methodologic edge over the former trials; the experimental evidence was stronger than that from the others.

What should be the standard of management? According to the analysis of the results in this study, the incidence of SSI in closed incisions was 12.9% higher than for open sites. Although this is statistically significant, the clinical significance has to be determined; i.e., subjecting 12.9% patients to SSI (managed conservatively) vs. 100% of patients managed by the open incision method (with high cost, health care burden, prolonged healing time, and poor quality of life). Therefore, there is a need to identify the technique that offers the best possible outcomes. Literature reviews (summarized in Tables 3 and 4) and review of experience indicates that the lowest to highest incidence of SSI is reported in CSIA, closure over a drain, open incision, and closure without a drain. In CSIA, the incision is neither completely closed nor entirely open, permitting the secretions to exit and the cosmetic results to be equivalent to those of closed incisions [8]. On the basis of these benefits and the quality of the evidence (class 1a), CSIA is the technique of choice.

The controversy can be resolved by conducting a well-designed randomized study on these four techniques simultaneously, in which the outcome should be assessed using more than one set of criteria; i.e., CDC and ASEPSIS score with photographs of the surgical sites to be evaluated independently by a blinded assessor. Outcomes such as pain, cosmesis, total healing time, and cost should be included.

Conclusion

Clearly, the risk of SSI in closed incisions is greater than in open ones. Closure over a drain or CSIA can be an alternative, as suggested in the literature; however, further prospective research on all four techniques, with evaluation of the financial impact (on the individual and the healthcare system) and psychosocial outcomes, is warranted to construct firm guidelines.