Subject enrollment and randomization. *Sixteen patients did not receive the allocated interventions in the extended arm: 6 had fluids in excess of protocol; 3, fraction of inspired oxygen of less than 80%; 3, no heating blanket; and 5, no wound protector; 105 of 106 received at least 4 of 5 interventions.
Anthony T, Murray BW, Sum-Ping JT, Lenkovsky F, Vornik VD, Parker BJ, McFarlin JE, Hartless K, Huerta S. Evaluating an Evidence-Based Bundle for Preventing Surgical Site InfectionA Randomized Trial. Arch Surg. 2011;146(3):263–269. doi:10.1001/archsurg.2010.249
Surgical site infections (SSIs) are among the most common postoperative complications in patients who have colorectal operations. The occurrence of an SSI results in reduced quality of life, increased hospital length of stay, increased likelihood of mortality, and markedly increased cost.1- 4 Therefore, identifying and implementing evidence-based strategies designed to minimize SSI is an important clinical goal. Prior research has focused mainly on antibiotic choice and timing5,6 and improving these processes7; however, a number of adjunctive measures have also been individually evaluated in high-risk populations (including patients who have colorectal surgery) and have shown promise in reducing SSI. These adjunctive measures include (1) omission of mechanical bowel preparation8; (2) preoperative and intraoperative patient warming9,10; (3) the use of an increased concentration of inspired oxygen during and immediately after the procedures' conclusion11; (4) limiting intraoperative intravenous fluid volumes12; and (5) the use of wound barriers to protect the surgical wound from contamination during the procedure13 (Table 1). Despite the evidence supporting these measures, most have not been widely adopted to clinical practice. Furthermore, the effect on SSI rate that these interventions may have when used as a bundled intervention has not been defined.
The purpose of this study was to test the hypothesis that a series of evidence-based interventions not currently in widespread use, incorporated as a single bundle, would significantly decrease overall SSI rate for patients who have elective colorectal surgery.
This institutional review board–approved, nationally registered (NCT00953784), prospective, randomized clinical trial was designed to enroll patients who were having elective transabdominal colorectal procedures at a Veterans Administration teaching hospital. Both subjects undergoing laparoscopic and open procedures were eligible, as well as patients undergoing diverting or bypass procedures during which the colon or rectum required division or resection. This included the surgical creation or reversal of a colostomy. Patients undergoing emergency operations, transrectal procedures, or procedures involving only the small bowel or appendix were excluded from the study. Potential subjects were approached concerning the trial by the study coordinator. Subjects meeting eligibility criteria signed written consent and then were randomized. Randomization was by a block method (computer generated, 50 subjects per block; generated by principal investigator, T.A.). The randomization sequence was concealed prior to subject assignment by the study coordinator. Subjects were allocated 1 to 1 between the intervention (extended) and control (standard) arms of this study.
A review of the literature was conducted that focused on evidence-based measures that could be expected to reduce SSI and could be instituted in our practice setting. We identified evidence-based support (1 or more randomized controlled trial) for (1) omission of mechanical bowel preparation8; (2) use of preoperative and intraoperative warming designed to maintain normothermia9,10; (3) maintenance of increased concentration of inspired oxygen during and immediately after surgery11; (4) reduction of intravenous fluids during the operation12; and (5) the use of wound edge protection.13 We theorized that incorporation of these interventions as a bundle would reduce SSIs for subjects undergoing elective colorectal surgery. These measures taken together were designated as the extended arm of the study.
Subjects assigned to the extended arm of the trial did not receive either mechanical bowel preparation or oral antibiotics. (Subjects in both arms of the trial received a clear liquid diet for the calendar day prior to their operation and phosphosoda enemas if they were having operations involving the colon distal to the splenic flexure.) Preoperative warming was accomplished by placement of a forced-air heating blanket on the torso of the subject while in the surgical holding area. Once in the operating room, a conductive heating blanket was used beneath the subject in addition to the forced air unit. Warmed fluids were not used. Subjects received an increased concentration of inspired oxygen (80%) from the point of intubation until 2 hours after surgery. After extubation, subjects were given 80% face-mask oxygen. Restriction of intraoperative, intravenous fluid administration was accomplished according to an algorithm designed to favor early colloid administration compared with crystalloid. The final intervention was placement of a plastic wound edge protection device in the incision once the peritoneal cavity was entered.
The standard arm of this trial consisted of our current practices. These included mechanical bowel preparation with oral antibiotics, intraoperative forced air warming, maintenance of physiologic concentration of inspired oxygen after endotracheal intubation (target fraction of inspired oxygen, 30%), intravenous fluid delivered at the discretion of the anesthesiologist, and no wound edge protectors.
Both arms received intravenous antibiotics prior to the surgical incision according to Surgical Care Improvement Project guidelines. After approximately 1 year, the preferred prophylactic antibiotic was switched from cefoxitin to ertapenem based on evidence from a randomized controlled trial.15
Infections were classified according to their location relative to the surgical incision. Infections occurring at or above the fascia were classified as superficial incisional infections. Infections occurring below the fascia were designated as organ/space infections. The Center for Disease Control and Prevention criteria were applied to define infections in these locations.16 Multiple sites of infection in 1 subject were counted as a single occurrence. Subjects sustaining both superficial incisional infections and organ/space infections were counted as organ/space infections. The primary outcome measure for this trial was overall infection rate at 30 days after surgery (superficial incisional infection rate plus organ/space infection rate = overall SSI rate). Secondary outcome measures were the location of the SSI (superficial incisional or organ/space) and the time to identification of the infection. Subjects who completed the study were reviewed independently by blinded, certified infection control coordinators (K.H. and J.M.) who applied the study definitions to determine whether or not the subject had sustained an SSI, the timing of the identification of the SSI, and whether the SSI was a superficial incisional or organ/space in location. Evaluations were made by the infection control coordinators at the time of discharge and/or by electronic medical record review at 30 days after surgery.
Clinical and demographic information was collected including age, sex, indication for surgery, comorbid conditions, and preoperative laboratory values. Perioperative data collected included American Society of Anesthesiologists class, type and timing of intravenous prophylactic antibiotic given, skin preparation used, surgery performed, estimated level of contamination, wound classification, estimated blood loss, length of operation, blood work (including arterial blood gas), and type and volume of fluids given. In addition to the study end points, postoperative variables collected included subject temperature in the recovery unit and postoperative blood work. If an SSI was identified, data was collected on the timing from surgery and the location of the infection.
We theorized that the bundled interventions included in the extended arm of this study would significantly reduce overall SSI. Based on a 50% reduction in overall SSI from the baseline rate (ie, 25% baseline to 12.5% after intervention) with α set at .05 and power of 0.9, it was estimated that 334 subjects were required (167 in each arm) and, planning for a 5% attrition rate, the total sample size estimated to complete the study was calculated to be 350 subjects.
The trial opened in April 2007, closed in January 2010, and successfully recruited 211 subjects. The study was terminated after a planned second interim analysis suggested that continuing the study to its originally designated sample size was unlikely to result in a significant improvement in overall SSI rate. Conditional power analysis of the study results at this junction suggested a less than 1% chance of identifying a statistically positive effect of the extended arm interventions on SSI rate if the study was taken to its original accrual goal.
The resulting database was analyzed in SPSS version 16.0 (SPSS Inc, Chicago, Illinois). All analyses were performed according to intention to treat. Univariate analysis of categorical variables was performed using Fisher exact or Pearson χ2 tests. Normally distributed continuous variables were analyzed using the t test. Nonparametric variables were assessed using the Mann-Whitney U test. All P values were calculated with 2-tailed tests. All variables trending toward significance (ie, where P ≤ .20) in univariate analysis were included in multivariate analysis. A forward stepwise logistic regression was used to identify factors associated with occurrence of SSI.
The trial enrolled and randomized a total of 211 subjects. A single subject withdrew consent; thus, 210 consenting, randomized subjects underwent a surgical procedure. Thirteen subjects were excluded from the analysis because they did not have incisions at risk for SSI (12 subjects had their surgical wounds left open to heal by secondary intention [7 standard and 5 extended] and a single subject had transrectal resection of a rectal tumor). Thus, there were 197 subjects considered in the intent-to-treat analysis (Figure).
All but 9 subjects were male, with an average age of 63.8 years (range, 30-87). Of the characteristics present prior to surgery, only serum albumin, creatinine, and hematocrit were significantly different between the standard and extended arms of the study. Although statistically significant, these values were not felt to represent clinically important imbalances. Table 2 and Table 3 summarize preoperative and perioperative variables for subjects in the extended and standard arms of the trial. Of note, operations performed on patients in the standard arm where shorter in duration than operations performed on patients in the extended arm (mean [SD], 150  vs 170  minutes; P = .03). The most common indication for surgery was to remove an adenomatous polyp or colorectal malignancy (80% of operations overall). The most common surgical procedure performed was right hemicolectomy (39% of all operations) (Table 4). There was no mortality among the participants in this trial; therefore, 30 days of follow-up were completed for the entire cohort.
Overall compliance with Surgical Care Improvement Project guidelines for timing and appropriateness of antibiotic choice was 99% and 96%, respectively. There was not a significant difference in the rate of SSI based on the type of prophylactic antibiotic delivered. The overall rate of infection for subjects who received cefoxitin was 37% (32 of 87), and the rate for those receiving ertapenem as prophylaxis was 34% (34 of 99; P = .76). One hundred percent of subjects in both arms had their abdominal wall hair clipped. A chlorhexidine-alcohol preparation was used to clean the surgical site in 187 subjects (95%), and povidone-iodine was used in the remaining 10 subjects.
For subjects considered in the intention-to-treat analysis, complete compliance with all of the 5 interventions comprising the extended arm of the study was 84% (84 of 100). Ninety-nine percent of subjects received at least 4 of 5 of the extended interventions. Of the individual components of the extended arm interventions, compliance was most difficult for the fluid restriction algorithm (94%). Compliance with the other intended interventions was greater than 95% in each case. In 5 cases, subjects in the standard arm did not receive the intended mechanical bowel preparation; otherwise, there was no evidence of contamination of the standard arm with interventions intended for the extended arm.
Considering the intention-to-treat population, the overall rate of SSI identified was 35% (69 of 197). The intention-to-treat analysis identified a significantly higher rate of infection for subjects in the extended arm of the study (intention-to-treat rates: standard arm, 24% vs extended arm, 45%; P = .003). Most of this effect occurred owing to a difference in the rate of superficial incisional infections. The superficial incisional infection rate for subjects enrolled in the standard arm was 19%, whereas subjects enrolled in the extended arm had a rate of 36% (P = .004). The rate of organ/space infections was not significantly different between the 2 groups (standard arm, 6% vs extended arm, 9%; P = .59).
Univariate analysis failed to reveal any significant associations between overall SSI rate and a wide range of perioperative variables except for study arm assignment (Table 5 and Table 6). Logistic regression considering factors trending toward significant association with overall SSI in univariate analysis (American Society of Anesthesiologists class ≥3, laparoscopic approach, operation below peritoneal reflection, operative time, and body mass index) and study arm assignment demonstrated that only allocation to the extended arm of this trial was independently associated with SSI. Allocation to the extended arm of this study conferred a 2.49-fold (95% confidence interval, 1.36-4.56; P = .003) increased risk of developing a SSI.
The median time to identification of infection was the ninth postoperative day (3-24 days). Although there was no difference in time to identification of infection between the 2 groups (median time to infection: standard arm, 9 days [range, 3-20]; extended arm, 10 days [range, 3-24]; P = .71), more infections were identified after discharge in the extended arm. A total of 32 (46%) infections were identified after the subject had been discharged from the hospital. In the extended arm, 51% of infections were identified after discharge and in the standard arm, 38% (P = .006).
Infections after colorectal surgery result in substantial morbidity and mortality. Despite greater awareness of the negative consequences of SSI and the widespread adoption of process measures directed at reducing SSIs, they remain stubbornly persistent. This study was designed to build on the process improvements already in place in an attempt to reduce SSI rates among subjects who have colorectal surgical procedures. A previous retrospective review of our practice identified an SSI rate of 24.5% for subjects who had colorectal operations despite careful adherence to prophylactic intravenous antibiotic recommendations.17 Subsequent review of the available literature identified 5 additional evidenced-based practices that reduced SSI rates in subjects who had colorectal surgical procedures. While each of these practices had a modest ability to reduce SSI, we hypothesized that incorporating these strategies into a single treatment bundle would have a synergistic effect on reducing infection rates in colorectal surgery subjects. Not only did this prove to be false but, counterintuitively, the bundled intervention resulted in significantly more SSIs.
There are several limitations of the study that may explain this result. First, the design does not allow for the identification of the contribution of individual measures to the overall outcome. Thus, it is possible that 1 or more of the measures included in this study may, singly or in combination, have a favorable affect on SSIs. Although aware of this issue, we felt justified in designing a study in which multiple interventions were bundled in an attempt to improve an outcome because this closely resembles current practices in the era of process improvements and performance measures.
Second, 2 additional studies have been published since this trial was begun that call into question the utility of some of the measures included in the extended arm of this study. An updated Cochrane review and the report of the PROXI trial have suggested that omission of bowel preparation and the use of supplemental oxygen may not exert as positive an influence on SSI as was originally believed.14,18 A further limitation is the methodology used to define the main outcome. Although this was intended to be prospective and blinded, medical record review at 30 days was often used to define the main outcome; thus, some degree of observer bias was possible. It is also notable that several risk factors that have been traditionally associated with SSI (eg, body mass index, operative time, and National Healthcare Safety Network category) were imbalanced in the 2 arms of this study, perhaps contributing to the observed results.
Finally, although the lack of mortality in this population suggests quality perioperative care, the overall rate of SSI identified in this study is higher than previously described and brings into question the generalizability of the results. The overall rate of infection was 36%, with almost 80% of the infections occurring within the surgical incision and nearly 50% of these identified after the subject was discharged from the hospital. This rate is likely a reflection of many factors including the broad inclusion criteria (for example, the present study included subjects with stomas, a population typically excluded from studies of SSI); the general lack of physiologic reserve of the population (80% were American Society of Anesthesiologists class 3 or 4); complete subject follow-up allowing accurate capture of infections (especially the high number of infections occurring after hospital discharge); use of blinded evaluators; and rigid application of unambiguous definitions of SSI. The effect of additional factors such as the teaching nature of our program, the specific sex and age distributions of our subject population, variations in surgical practice, and some of the imbalances in risk factors (For example, in the extended arm there were more subjects with NHSN risk scores of 2 and 3, longer operative times, and greater number of operations performed below the peritoneal reflection.) may also explain this result. It should be noted, however, that despite the high rate of SSI overall, the rate within the standard arm was identical to our previous article and similar to the rate reported in several recent publications.15,17,19 Therefore, while the overall rate of infection reported in this study appears high, this is attributable to the bundled interventions in the extended-arm. We speculate that the higher-than-expected rate of infection among the subjects who received extended arm treatments is the result of unanticipated interactions between individual components that made up the extended arm of the trial. For example, perhaps increasing subject temperature coupled with reduced intraoperative fluids resulted in diminished blood flow and tissue oxygenation within the peri-incisional tissues, reducing immune surveillance and facilitating bacterial growth. It is also possible that the degree of diligence required to successfully adhere to multiple interventions diverted surgeon attention from the technical aspects of the procedures that are more important in limiting SSI.
Regardless of the specific reasons for the increased rate of infection in the extended arm of this study, these interventions seem to be ineffective when used as a bundle. However, we believe the more important implication of this study pertains to the common practice of adopting bundled interventions to improve outcome. The assumption that single measures identified by randomized controlled trials can be grouped into a bundle and effect an outcome in a predictable and positive fashion is not supported by the findings of this study. This, in turn, suggests a more cautious approach when it comes to adopting bundled measures that have not formally been tested in the target population.
In summary, we elected to take a scientifically unconventional but clinically common approach to the problem of SSI after colorectal surgery and attempted to show that an intervention bundle consisting of evidence-supported practices would positively affect SSI. The results of study demonstrated, however, that the bundle of interventions used (omission of mechanical bowel preparation, preoperative and intraoperative warming designed to maintain normothermia, increased inspired oxygen during and immediately after surgery, limiting fluid administration during surgery, and the use of surgical wound protectors) increased the risk of SSI 2.49-fold when compared with our standard practice. The use of this bundle of interventions is not warranted and raises significant questions concerning the general wisdom of adopting bundled approaches in other clinical situations to minimize negative outcomes when these bundles have not formally been tested in controlled clinical trials.
Correspondence: Thomas Anthony, MD, Department of Surgical Services, Veterans Affairs North Texas Health Care System, 4500 S Lancaster Rd, Dallas, TX 75216 (email@example.com).
Accepted for Publication: March 3, 2010.
Published Online: November 15, 2010. doi:10.1001/archsurg.2010.249
Author Contributions: Dr Anthony had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Anthony, Lenkovsky, and Huerta. Acquisition of data: Anthony, Murray, Sum-Ping, Lenkovsky, Vornik, Parker, McFarlin, and Hartless. Analysis and interpretation of data: Anthony, Sum-Ping, McFarlin, Hartless, and Huerta. Drafting of the manuscript: Anthony, Parker, Hartless, and Huerta. Critical revision of the manuscript for important intellectual content: Anthony, Murray, Sum-Ping, Lenkovsky, Vornik, McFarlin, and Huerta. Statistical analysis: McFarlin, and Hartless. Obtained funding: Administrative, technical, and material support: Anthony, Murray, Parker, and Huerta. Study supervision: Anthony, Murray, Sum-Ping, and Lenkovsky.
Financial Disclosure: Dr Anthony reports receiving honoraria from Merck; and Dr Sum-Ping, an educational grant from Hospira.