Effect of an Implantable Gentamicin-Collagen Sponge on Sternal Wound Infections Following Cardiac Surgery:  A Randomized Trial FREE

Despite the use of prophylactic systemic antibiotics, postoperative sternal wound infection continues to be a serious problem after cardiac surgical procedures, especially in the increasing population of patients with diabetes and/or obesity. Sternal wound infection is associated with significant suffering, additional expense,^1- 2 lengthened hospital stay,^3- 4 and increased mortality. The gentamicin-collagen sponge was developed to prevent and/or treat wound infections by providing high local gentamicin concentrations without the high systemic concentrations associated with nephotoxicity.⁵ The sponge's collagen matrix is biodegradable. The gentamicin-collagen sponge (Innocoll Technologies Ltd, Roscommon, Ireland) received marketing approval in Germany in 1985, and is currently approved for use in another 53 countries. To date, more than 2 million sponges have been used to treat more than 1 million patients outside the United States across a broad range of clinical indications (Innocoll Technologies Ltd, unpublished data, October 2007).

In a large (n = 2000) investigator-initiated, 2-center trial conducted in Sweden, cardiac surgery patients randomized to insertion of 2 sponges demonstrated a 53% decrease in sternal wound infection (4.3% vs 9.0%; P < .001) and a 46% decrease in surgically treated sternal wound infection (2.1% vs 3.9%; P = .02).⁶ The current phase 3 trial was designed to confirm these promising data and support regulatory approval in the United States.

We conducted a phase 3, single-blind, prospective, randomized controlled trial of the gentamicin-collagen sponge in patients undergoing cardiac surgery. Patients were enrolled at 48 sites throughout the United States.

After institutional review board approval from each site, and individual written informed consent, patients meeting the following criteria were eligible for enrollment: (1) male or female patients aged 18 years or older, (2) scheduled to undergo nonemergent coronary artery bypass graft and/or valve repair or replacement surgery through a full median sternotomy, and (3) at higher risk for sternal wound infection, defined as the presence of diabetes mellitus (treated with either oral agent or insulin) and/or obesity, defined as body mass index (calculated as weight in kilograms divided by height in meters squared) greater than 30.

The exclusion criteria were (1) known history of hypersensitivity to gentamicin or bovine collagen, (2) emergency surgery, (3) significant concomitant surgical procedure, (4) minimally invasive or thoracic surgical approach, (5) pregnancy, (6) preoperative mechanical assist device or intraaortic balloon pump if inserted for shock or low output syndrome, (7) active and significant systemic infection, (8) antibiotic therapy within 2 weeks preoperatively, (9) preoperative serum creatinine level greater than 3 mg/dL (to convert to μmol/L, multiply by 88.4) or renal failure requiring dialysis, (10) malignancy except for squamous or basal cell carcinoma of the skin, (11) major organ transplantation, (12) significant drug or alcohol abuse, (13) receiving systemic immunosuppressive drugs, including steroids (at a dose >10 mg of oral prednisone daily), or current immunosuppressive condition (eg, symptomatic human immunodeficiency virus infection), (14) scheduled to receive stress doses of glucocorticoids, (15) postsurgical life expectancy of 90 days or less, (16) participation in another experimental drug or device study, and (17) refusal to accept medically indicated blood products.

Each 100 cm² (5 × 20 cm) sponge contained 280 mg of collagen and 130 mg of gentamicin. Study patients received 2 sponges inserted between the sternal halves along the full length of the sternum immediately before closure of the sternum, as described previously.⁶ The control group did not receive gentamicin-collagen sponges. The protocol called for patients randomized to the gentamicin-collagen sponge group and requiring reexploration (eg, due to bleeding) within 1 week after surgery to receive 2 new sponges inserted at the time of closure of the reoperation. The protocol required all surgeons to undergo a training and certification process that included viewing a video outlining proper handling and insertion of the gentamicin-collagen sponge.

Randomization occurred after surgical incision using a central randomization interactive voice response system, thus providing complete allocation concealment. The randomization scheme was stratified by site and random block sizes were used. Patients and members of the clinical events adjudication committee were blinded to study group assignment. Surgeons also were asked to determine the presence or absence of infection. To avoid possible bacterial growth on an unimpregnated sponge, patients in the control group did not receive a placebo sponge. As a result, surgeons were not blinded to study group assignment.

Preoperatively, the use of nasal mupirocin prophylaxis was allowed but not required. Consistent with published guidelines,^7- 9 the protocol called for antibiotic regimens with cefazolin or cefuroxime to be initiated 60 minutes prior to skin incision. Vancomycin was administered to individuals who were allergic to cephalosporin or penicillin, or those at increased risk of methicillin-resistant Staphylococcus aureus colonization. It was permissible for ciprofloxacin to be added to vancomycin if greater gram-negative coverage was desired. Dosing was weight based and was to be continued for at least 24 hours, but not more than 48 hours.^7- 9 The use of other topical antibiotics was prohibited in both the gentamicin-collagen sponge group and the control group.

A previous trial indicated that the effectiveness of the gentamicin-collagen sponge was enhanced if rigid fixation of the sternum, defined as use of more than 6 wires, was achieved.¹⁰ Therefore, our protocol called for the use of at least 7 sternal fixation wires. Standard preoperative demographics and intraoperative variables were collected as well as the following variables that were thought to have a possible role in sternal wound infection: diabetic control (level of hemoglobin A_1c), body mass index, administration of insulin in the perioperative period, perioperative serum glucose levels, perioperative core temperature, method of surgical skin preparation, and duration of surgery. Overall risk for mortality and morbidity was assessed using the validated Parsonnet risk score.¹¹

To comply with US Food and Drug Administration regulations, information regarding each individual's race and ethnicity was collected on the study case report form, which was completed by site personnel from the individual's medical record. The intent-to-treat population included all randomized individuals. The per-protocol population, as defined in the statistical analysis plan, included all randomized patients who completed the study and had none of the following prespecified major protocol deviations: (1) did not meet eligibility criteria unless waiver obtained prior to enrollment, (2) received treatment different from that to which they were randomized or received incorrect dosing of gentamicin-collagen sponge, or (3) did not receive prophylactic antibiotics prior to skin incision.

The primary study outcome was the incidence of sternal wound infection occurring within the period from surgery through postoperative day 90. Key secondary efficacy end points included the incidence of deep sternal wound infections, superficial sternal wound infections, surgically treated sternal wound infections (defined as any type of surgical intervention for sternal wound infection including opening the wound), postoperative hospital length of stay, and score for additional treatment, presence of serous discharge, erythema, purulent exudate, separation of the deep tissues, isolation of bacteria, and duration of inpatient stay (ASEPSIS) through 90 days postoperatively.^12- 14 The validated ASEPSIS score assigns points for 9 variables including antibiotics, drainage of pus under local anesthesia, debridement of wound under general anesthesia, isolation of bacteria, duration of hospital stay longer than 14 days, and daily assessments involving the portion of wound affected for 4 separate variables (serous discharge, erythema, purulent exudate, and separation of deep tissues).^12- 14 The minimum score is 0, higher scores are worse, and there is no theoretical maximum score. We also assessed (1) change in serum creatinine level from baseline to peak through postoperative day 7 or hospital discharge if earlier, (2) assessment of pain and wound healing in individuals based on a structured wound healing questionnaire administered at 30, 60, and 90 days postoperatively, (3) emergency department or surgical office visits secondary to wound complaints, (4) rehospitalization and mortality rates, (5) pain assessments from surgery through postoperative day 7 and amount of pain medication administered in the first 3 days postoperatively, and (6) serum gentamicin levels in a subset of sites. Blood was withdrawn at baseline (≤2 hours before implantation of 2 sponges), and then at of 2 ± 0.5 hours, 4 ± 0.5 hours, 8 ± 0.5 hours, 12 ± 1 hours, and 24 ± 2 hours after closure of the surgical wound for core laboratory determination of serum gentamicin levels. A cost analysis was planned, but was not performed due to the study's negative results.

The independent clinical events classification committee was composed of 3 infectious disease experts blinded to the treatment assignments. Possible infections were identified by triggered events in the electronic case report form, including signs or symptoms of possible infection, administration of postoperative antibiotics, rehospitalization, and death. After review of all pertinent medical records, the clinical events classification committee determined the presence or absence, extent, and severity of all possible infections using standardized criteria including those from the Centers for Disease Control and Prevention^{12- 13,15- 16} and the ASEPSIS scoring system.^12- 14 The clinical events classification committee also adjudicated the following secondary end points of (1) rehospitalization for sternal wound infection within 90 days postoperatively, (2) presence of superficial or deep incisional sternal wound infection, and (3) presence of surgically treated sternal wound infection. Infections not related to the sternum (eg, infection at saphenous vein harvesting sites, intravenous catheter, or pneumonia) were not included.

All statistical analyses were conducted by the independent Duke University statistical trial team. Sample size calculation during the planning stage indicated that a total of 1450 patients (randomized at a ratio of 1:1) would be required to detect a 50% relative reduction of sternal wound infections occurring within the period from surgery through postoperative day 90 in the gentamicin-collagen sponge group compared with the control group, with a power of 85% and a 2-sided type I error rate of .05. Based on previous trials, we assumed a 7% sternal wound infection rate in the control group.

The primary statistical analysis was performed using standard intention-to-treat methods. Unless otherwise indicated, all results in the text, tables, and figures reflect the intent-to-treat population. As a prespecified secondary analysis, we also performed a per-protocol analysis, which included all randomized patients who completed the study and had no prespecified major protocol deviations. Comparisons of the primary end point (incidence of sternal wound infections) between the 2 study groups were analyzed with 2-sided χ² testing that used data from across all sites after checking for treatment × site interaction. For all secondary efficacy and subgroup analyses, a nominal P value of less than .05 (2-sided) was adopted to indicate statistical significance, and the results were considered descriptive.

Descriptive statistical comparisons between the treatment groups were performed using the χ² test or the Fisher exact test as appropriate for the categorical secondary efficacy end points. The t test or Wilcoxon rank sum test were used as appropriate for comparison of continuous secondary efficacy end points. The log-rank tests were used to compare the time to first sternal wound infection between the 2 study groups. Kaplan-Meier survival curves of time to first sternal wound infection also were presented.

No formal interim analysis was planned in this study. An independent data and safety monitoring committee monitored the trial on an ongoing basis. All statistical analyses were performed using SAS software version 9.2 (SAS Institute Inc, Cary, North Carolina).

Overall, 1502 individuals were enrolled at 48 US sites between December 21, 2007, and March 11, 2009. Figure 1 shows the flow of individuals from enrollment through randomization, to 90-day follow-up. Of 753 patients randomized to receive 2 gentamicin-collagen sponges, 741 received 2 sponges (98.5%), 1 received 1 sponge (0.1%), and 10 did not receive any sponges (1.3%). One individual (0.1%) randomized to the gentamicin-collagen sponge group received 4 sponges intraoperatively; 2 were inserted and then the surgeon needed to reevaluate the bypass grafts and enough time had elapsed so he removed the 2 sponges and inserted 2 new gentamicin-collagen sponges prior to closing the sternum. Of note, only 31 individuals (2.1%) were lost to follow-up at 90 days (13 in the gentamicin-collagen sponge group and 18 in the control group). Only 12 of 1502 individuals did not contribute information to the primary end point. Due to the small effect made by these 12 patients (0.8%), in the final primary analysis with the intent-to-treat population we treated them as not having sternal wound infection.

Of the 1502 individuals in the study population, there were 1060 men (70.6%). Patients had a median body mass index of 32.9 and 1006 had diabetes (67.0%). The control group (n = 749) and the gentamicin-collagen sponge group (n = 753) were balanced with regard to baseline characteristics including age, weight, diabetes, and smoking history (Table 1). Perioperative variables including preincision antibiotics, type and duration of surgery, and number of internal mammary arteries used were also well balanced by treatment group (Table 2).

There was no significant difference in sternal wound infections in patients randomized to the gentamicin-collagen sponge group (63/753; 8.4% [95% confidence interval {CI}, 6.4%- 10.3%]) compared with patients in the control group (65/749; 8.7% [95% CI, 6.7%-10.7%]; P = .83) (Table 3). No significant differences were observed between the sponge group and the control group, respectively, in superficial sternal wound infections (49/753 [6.5%; 95% CI, 4.8%-8.3%] vs 46/749 [6.1%; 95% CI, 4.4%-7.9%]) and deep sternal wound infections (14/753 [1.9%; 95% CI, 0.9%-2.8%] vs 19/749 [2.5%; 95% CI, 1.4%-3.7%). Kaplan-Meier curves for adjudicated sternal wound infection in the intent-to-treat population appear in Figure 2. Analysis using the principal investigator's assessment of sternal wound infection for each individual US site demonstrated results similar to the clinical events classification committee's adjudicated events for sternal wound infection within 90 days postoperatively for the gentamicin-collagen sponge group (148/753 [19.7%] compared with the control group (150/749 [20.0%]; P = .86). In addition, analyses performed in the per-protocol population (n = 1460) revealed similar results to the intent-to-treat population (Table 3).

Exploratory analyses for the primary end point (adjudicated sternal wound infection in the intent-to-treat population) were conducted in several subgroups related to our predefined eligibility criteria for the sponge group vs the control group, respectively, of patients who were obese and had diabetes (35/316 [11.1%] vs 45/327 [13.8%]; P = .30), patients who had diabetes but were not obese (7/177 [4.0%] vs 10/186 [5.4%]; P = .52), and patients who were obese but did not have diabetes (21/258 [8.1%] vs 10/236 [4.2%]; P = .07).

Potential pathogens were isolated with equal frequency from patients with sternal wound infections in the sponge group (27/753 [3.6%]) and in the control group (32/749 [4.3%]; eTable 1). Except for coagulase negative staphylococci (involving 10 patients in the control group vs 3 patients in the sponge group), all pathogens were similarly distributed in the 2 groups. This included both sensitive and resistant S aureus, gram-negative bacilli, and streptococci including enterococcus. Only 3 organisms from each group were found to be resistant to gentamicin; 4 of the 6 were Staphylococcus epidermidis (eTable 2).

Peak serum gentamicin levels in 68 individuals ranged from 0.7 μg/mL to 4.6 μg/mL (mean, 2.3 μg/mL) and decreased to a mean (SD) level of 1.0 (0.8) μg/mL by 24 hours after sponge insertion (Figure 3). The percentage increase in serum creatinine from baseline to peak level through postoperative day 7 or hospital discharge was similar in both groups (mean [SD] change of 34.0% [47.7%] for the gentamicin-collagen sponge group vs 33.3% [57.6%] for the control group).

Overall, 49 individuals required reexploration of their surgical wound (21 in the sponge group and 28 in the control group). There were 46 deaths observed from randomization through 90 days postoperatively (19 in the sponge group [2.6%] and 27 in the control group [3.6%]). There was no difference in the frequency of adverse events (eTable 3). No overt differences were observed in wound healing between the 2 groups at 30, 60, or 90 days postoperatively based on data from a structured subject questionnaire (eTable 4). The median opioid equianalgesic dosing¹⁷ in the first 3 days postoperatively was similar in both groups (sponge group: 34.3 mg; control group: 35 mg) as were daily pain scores (data not shown).

Our primary finding in this phase 3 randomized controlled trial is that the gentamicin-impregnated collagen sponge did not reduce the rate of sternal wound infections in patients undergoing cardiac surgery. These findings directly contradict the data previously available on the efficacy of this technology in wound infection prevention.⁶

The gentamicin-collagen sponge, developed to deliver a high concentration of gentamicin to the skin and soft tissue surrounding postoperative wounds, has undergone testing for use in cardiac surgery throughout northern Europe. Initially, in a single center trial in Finland, 542 cardiac surgery patients randomized to insertion of 1 sponge (130 mg of gentamicin) during sternal closure vs control exhibited a nonsignificant reduction in sternal wound infection (4.0% vs 5.9%; P = .20).¹⁸ In a larger (n = 2000) investigator-initiated trial conducted in 2 centers in Sweden, cardiac surgery patients randomized to insertion of 2 sponges (total of 260 mg of gentamicin) showed a 53% decrease in sternal wound infection (4.3% vs 9.0%; P < .001) and a 46% decrease in surgically treated sternal wound infection (2.1% vs 3.9%; P = .02).⁶ This study served as strong preliminary data for our trial.

For our trial, we enrolled individuals with diabetes, obesity, or both, to study patients at higher risk for sternal wound infection because the unmet medical need is greatest in these individuals. Indeed, using these criteria we observed an event rate of 8.7% in the control group, which was slightly higher than the 7% rate assumed in our sample size and power calculation and demonstrates that the lack of effectiveness observed was not due to a deficiency of treatable outcome events. Our data do not allow us to identify with certainty the cause of the lack of efficacy we observed. One potential explanation that does not appear to apply in this trial is that sternal infections were due to gentamicin-resistant organisms. However, we can speculate that gentamicin may elute too rapidly to add efficacy to systemic preoperative antibiotics. In support of this theory is the data from Friberg et al⁵ demonstrating low wound and local levels of gentamicin 12 hours after insertion of 2 sponges (total of 260 mg of gentamicin) between the sternal halves. Arguing against this explanation, however, is the fact that these data were obtained in a subset of patients enrolled in the trial with very positive results,⁶ so it appears that gentamicin levels were indeed adequate to prevent 53% of all sternal wound infections in that trial.⁶

Because the collagen-gentamicin sponge was not effective in our trial, one may ask why an earlier large study suggested such a strong treatment benefit,⁶ especially because the dosing of the sponge was the same in both studies and our protocol mandated use of rigid sternal fixation, identified by the previous trial as an important factor associated with sponge efficacy.¹⁰ Important differences between the Swedish study and our study include several of the following important quality-control measures that were not incorporated in the previous study: onsite monitoring and source data verification, central adjudication of outcomes by an independent blinded committee, and the inclusion of a large number of hospitals (48 vs 2).

The importance of large validation trials is evident from our results and from the observation that findings from positive single-center trials are often not confirmed in larger multicenter trials.¹⁹ Furthermore, ethnic and regional differences may have produced differing results between this US-based trial and the previous study conducted in Sweden.²⁰ For example, the distribution of bacterial pathogens can vary between countries.²¹ In the Swedish trial there were no cases of methicillin-resistant S aureus,²² whereas in our trial some patients, albeit a small percentage (6.3%), had methicillin-resistant S aureus growth from sternal wound infection (eTable 1), which may not be sensitive to gentamicin even at high levels. Thus, variations in the distribution of bacterial pathogens among countries could affect efficacy in a trial of infection prevention such as the present one.

In conclusion, despite approval of the gentamicin-collagen sponge in 54 countries outside of the United States and positive results from a large Swedish trial, our large multicenter US trial did not find the gentamicin-collagen sponge to be effective at preventing sternal wound infection in the setting of cardiac surgery.