Impact of Antibiotic Prophylaxis on Surgical Site Infections in Cardiac Surgery

(1) Background: Cephalosporins (CA) are the first-line antibiotic prophylaxis recommended to prevent surgical site infection (SSI) after cardiac surgery. The combination of vancomycin/gentamicin (VGA) might represent a good alternative, but few studies have evaluated its efficacy in SSI prevention. (2) Methods: A single-centre retrospective study was conducted over a 13-year period in all consecutive adult patients undergoing elective cardiac surgery. Patients were stratified according to the type of antibiotic prophylaxis. CA served as the first-line prophylaxis, and VGA was used as the second-line prophylaxis. The primary endpoint was SSI occurrence at 90 days, which was defined as the need for reoperation due to SSI. (3) Results: In total, 14,960 adult patients treated consecutively from 2006 to 2019 were included in this study, of whom 1774 (12%) received VGA and 540 (3.7%) developed SSI. VGA patients had higher severity with increased 90-day mortality. Nevertheless, the frequency of SSI was similar between CA and VGA patients. However, the microbiological aetiologies were different, with more Gram-negative bacteria noted in the VGA group. (4) Conclusions: VGA seems to be as effective as CA in preventing SSI.


Introduction
Since the advent of antibiotics in the 1940s [1,2], antibiotic prophylaxis has played an important role in the prevention of infections, including cardiac surgery [3]. However, 80 years later, despite international guidelines [4], the modalities of the administration of antibiotic prophylaxis remain heterogeneous [5]. In cardiac surgery, the reference prophylaxis policy directs the administration of cephalosporins before skin incision and up to 24-48 h after, and the main alternative is the administration of vancomycin [4,6]. Indeed, in many cardiac surgery studies, vancomycin appears less effective for preventing surgical site infections (SSIs) than anti-staphylococcal penicillins or first/second-generation cephalosporins, especially those infections caused by methicillin-susceptible staphylococci (MSS) [7][8][9][10]. This finding might be explained by the reduced bactericidal effect of vancomycin [11,12] and less optimal penetration into mediastinal tissues during cardiac surgery under cardiopulmonary bypass (CPB) [13]. However, several studies did not identify major differences between these two antibiotic regimens, particularly when the duration of administration was considered [14,15]. In addition, some studies have found a significant effect of the timing of vancomycin injection, with an increased risk of SSI in cases of delayed administration [15,16]. Finally, in our centre, vancomycin is combined with gentamicin to enlarge the spectrum of prophylaxis and target Gram-negative bacterial infections [17]. Interestingly, few studies have compared the efficacy of the combination of vancomycin and gentamicin to that of cephalosporins.
Thus, the aim of our retrospective, single-centre study was to assess whether the combination of vancomycin and gentamicin (VGA) is a valuable alternative to cephalosporins (CA) to prevent SSI in cardiac surgery with CPB.

Study Design
This single-center retrospective study was performed in a tertiary care teaching hospital performing more than 1000 annual cardiac surgical procedures. Between 2006 and 2019, all consecutive patients who underwent cardiac surgery with cardiopulmonary bypass (CPB) were eligible. The exclusion criteria were a previous history of cardiac surgery in the past year and missing clinical information (including antibiotic prophylaxis).

Definitions
SSI was defined as the need for reoperation within 90 days after index surgery for local or systemic infection involving the sternotomy scar [18,19]. The sternal scar was assessed daily after surgery, and the date of SSI was the date of return to the operating room for SSI. Only the first reoperation for suspected SSI was considered. The validation of SSI was made during multidisciplinary regular meetings including a cardiac surgeon, an intensive-care physician, and a physician of the infection control unit. Deep SSI (dSSI) was defined as the need for sternal reopening with deep purulent discharge or destruction of the sternal bone/deep sternal osteomyelitis. Otherwise, the SSI was defined as a superficial SSI (sSSI) requiring reoperation. Obesity was defined as a body mass index >30 kg/m 2 . Acute infective endocarditis was defined as endocarditis under antibiotics at the time of surgery. Cardiac insufficiency corresponds to patients with a left ventricular ejection fraction of less than 30%.

Selection of Antibiotic Prophylaxis and Protocol of Administration
During the study period, first-and second-generation cephalosporins were used the first-line intraoperative antibiotic prophylaxis. Cefamandole was administered from 2006 to 2015, and cefazolin was administered from 2016 to 2019. Cephalosporins were administered according to the following protocol: 30 mg/kg at induction with a reinjection every 4 h during the operation and then a dose of 1 g per 6 h for 24 h. Moreover, 1 g was administered into the priming of the CB. The combination of vancomycin/gentamicin was the second-line antibiotic prophylaxis selected in cases of allergy to beta-lactam, prior antibiotherapy ≥ 72 h, prior hospitalization in a unit with a high prevalence of methicillinresistant Staphylococcus aureus (MRSA) and/or in patients known to be colonized by MRSA. VGA was administered after the induction and from 5 to 30 min before the surgical incision through a central line according to the following protocol: vancomycin was perfused over 1 to 2 h at a dose of 30 mg/kg, and 500 mg was added to the priming of CPB. Then, 1 g was administered 12 h after the end of surgery. Gentamicin was perfused over 30 min as a single dose of 5-6 mg/kg. All patients underwent preoperative nasal screening for S. aureus carriage and preoperative nasal decontamination with mupirocin for 5 days, most frequently started on the day before surgery [20].

Objectives
The primary outcome was SSI occurrence at D-90 after index surgery. Secondary outcomes were the type of SSI, microbiological features, death at D-90, and occurrence of SSI based on propensity score-matching.

Data Collection and Ethics
Data were extracted from three databases. The first registry, Cardiobase, prospectively includes all adult patients who underwent cardiac surgery with CPB. This database has been approved by the Paris-Nord Biomedical Research Project Evaluation Committee (CERB, IRB No. 00006477), the Advisory Committee on Information Processing in Health Research (Paris, France) and the Commission Nationale Informatique et Liberté (CNIL, Paris, France). It is registered on ClinicalTrials.gov (NCT03393169). The following data are collected in this registry: demographic data, medical history, type of surgery and perioperative data. The Cardiobase information on SSIs was combined with a second database collected by the Infection Control Unit [18,20]. Finally, survival data were extracted from public French data of Institut National de la Statistique et des Etudes Economiques (INSEE) from 2006 to 2021, where the date of death was recorded.

Patient Management
Chronic therapies were administered on the day of surgery except for angiotensinconverting enzyme inhibitors/angiotensin receptor blockers. CA was administered at the time of anaesthesia induction, whereas VGA was infused after induction when the central venous catheter was inserted and before the surgical incision. Otherwise, patient intraoperative management was performed as previously described [21]. In our centre, coronary artery bypass grafting (CABG) is performed only with the internal mammary arteries (IMA), even for multivessel disease. For this purpose, the two IMA are harvested using the "skeletonization technique" (without collateral veins or endothoracic fascia). Then, the right IMA is implanted as a free graft on the left IMA using the so-called Y or T techniques. Finally, the use of sequential distal lateral anastomoses allows for multiple coronary branches to be targeting while respecting a harmonious bypass flow path. Secondary angioplasty is reserved for rare cases where the downstream bed of the right coronary is not accessible to bypass. After surgery, patients were admitted to the cardiac surgical intensive care unit for at least 48 h.

Statistical Analysis
Continuous variables were expressed as medians with interquartile ranges (IQRs) and were compared with the Mann-Whitney U test or the Kruskal-Wallis test as appropriate. Categorical variables were expressed as absolute numbers and proportions and were compared with Fisher's exact test or the chi-square test, as appropriate.
Time-to-event analyses were estimated with Kaplan-Meier analyses. To assess whether SSI was an independent factor for death, multivariate analyses were performed using a Cox model. For all models, variables with nominal two-tailed p values less than 0.05 were entered into the multivariate model, except for variables with obvious collinearity. The final models were selected using backwards stepwise regression and evaluated using the AIC and Tjur's R 2 coefficient of discrimination.
To address confounding by indication of VGA and other sources of bias arising from the use of observational data, we estimated full propensity score matching without replacement [22] using the Matchit package [23], which calls functions from the optimatch package [24]. The propensity score was estimated using logistic regression of the likelihood of total VGA administration (see Figure S1), and covariate balance was assessed using the cobalt package [25]. All statistical analyses were performed using R software (R Core Team,  [26] (checklist available in Appendix A). the use of observational data, we estimated full propensity score matching without re-placement [22] using the Matchit package [23], which calls functions from the optimatch package [24]. The propensity score was estimated using logistic regression of the likelihood of total VGA administration (see Figure S1), and covariate balance was assessed using the cobalt package [25]. All statistical analyses were performed using R software (R Core Team, 2014, R Foundation for Statistical Computing, Vienna, Austria). Figures were produced using the 'ggplot2 package', and statistics were produced using the 'stat package'. Missing values were not imputed. A p value < 0.05 was considered statistically significant. Reports of statistical analyses respected the STROBE recommendations for cohort studies [26] (checklist available in Appendix A).

Patient Characteristics Based on the Type of Antibiotic Prophylaxis
Compared to patients who received CA, those who received VGA were younger and had a lower BMI, fewer cardiovascular risk factors (arterial hypertension, diabetes mellitus, and dyslipidaemia) and a lower proportion of ischaemic heart disease (52% vs. 32%,

Relationship between Antibiotic Prophylaxis and SSI in the General Population
The type of antibiotic prophylaxis was not associated with SSI occurrence (Table 1 and Figure 2), and the time to SSI was similar between VGA and CA (15 (12-26) vs. 18 (12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27) days, p = 0.77). The main independent parameters associated with SSI (Table 2 and  Table S1) included older age, female sex, obesity, diabetes mellitus, cardiac insufficiency, ischaemic heart disease, and end-stage renal disease. Furthermore, SSIs were associated with CABG, short CPB duration and catecholamine requirements after CB. In multivariate analysis, VGA was not associated with SSI (Table 2).    Of note, in univariate analysis, the occurrence of sSSI appeared to be lower in the VGA group than in the CA group (Table 1 and Figure 3A), without a difference in the timing of sSSI occurrence. However, this association did not persist in multivariate analysis (Supplementary Table S2). No association was found between dSSI occurrence and type of prophylaxis. Of note, in univariate analysis, the occurrence of sSSI appeared to be lower in the VGA group than in the CA group (Table 1 and Figure 3A), without a difference in the timing of sSSI occurrence. However, this association did not persist in multivariate analysis (Supplementary Table S2). No association was found between dSSI occurrence and type of prophylaxis.

Microbiological Characteristics of Patients According to the Type of SSI
The microbiological nature of the SSI was different according to the type of prophylaxis (Table 3). Gram-negative bacteria were more frequently noted in VGA than in CA (52% vs. 35%, p = 0.013). In contrast, Gram-positive bacteria were more frequently noted

Microbiological Characteristics of Patients According to the Type of SSI
The microbiological nature of the SSI was different according to the type of prophylaxis (Table 3). Gram-negative bacteria were more frequently noted in VGA than in CA (52% vs. 35%, p = 0.013). In contrast, Gram-positive bacteria were more frequently noted in CA (69% vs. 50%, p = 0.006), especially enterococci, representing 13% of SSI in the CA group. In contrast, only one enterococcal infection was observed in the VGA group.

Relationship between Antibiotic Prophylaxis and SSI in Propensisty Score Analsysis
Given the heterogeneity of the study population, three sensitivity analyses were performed. The first was a propensity score using the full matching method.
In this analysis, the main risk factors associated with SSI (Table 4), sSSI and dSSI (Supplementary Table S3) remained the same. However, acute infective endocarditis became a risk factor for dSSI, and cardiac transplantation was only associated with dSSI. Of note, in these analyses, end-stage renal disease (ESRD) was the risk factor with the highest OR. In all analyses, VGA was not associated with SSI.

Relationships between Antibiotic Prophylaxis and SSI in a Sensitivity Analysis Excluding Active Infective Endocarditis
The second sensitivity analysis excluded patients with active infective endocarditis. Indeed, these patients may have a lower risk of SSI due to concomitant antibiotic therapy. In addition, VGA was over-represented among these patients, with 62% receiving VGA, while they represented only 12% of the overall population. In this analysis, VGA patients still presented more comorbidities and severity risk factors with higher EuroSCORE II compared to CA patients (3 (2-7) vs. 2 (1-4), p < 0.001) (Supplementary Table S4). No difference was noted in the occurrence of SSI, sSSI and dSSI ( Figure 4). In multivariate analysis, VGA was not a risk factor of SSI (Table 5), sSSI or dSSI (Supplementary Table S5).

Relationships between Antibiotic Prophylaxis and SSI in a Sensitivity Analysis Excluding Active Infective Endocarditis
The second sensitivity analysis excluded patients with active infective endocarditis. Indeed, these patients may have a lower risk of SSI due to concomitant antibiotic therapy. In addition, VGA was over-represented among these patients, with 62% receiving VGA, while they represented only 12% of the overall population. In this analysis, VGA patients still presented more comorbidities and severity risk factors with higher EuroSCORE II compared to CA patients (3 (2-7) vs. 2 (1-4), p < 0.001) (Supplementary Table S4). No difference was noted in the occurrence of SSI, sSSI and dSSI ( Figure 4). In multivariate analysis, VGA was not a risk factor of SSI (Table 5), sSSI or dSSI (Supplementary Table S5).

Relationships between Antibiotic Prophylaxis and SSI in a Sensitivity Analysis Restricted to CABG
Finally, a sensitivity analysis was performed in the homogeneous subgroup of patients undergoing coronary artery bypass grafting, a population at high risk of developing SSI. In this subgroup, VGA was less frequently associated with patient severity (Supplementary  Table S6). In the VGA group, patients were more often female, with a higher frequency of insulin-dependent diabetes, end-stage renal failure and chronic obstructive pulmonary disease, and slightly greater EuroSCORE II (2 (1-3) vs. 1 (1,2), p < 0.001) (Supplementary Table S6).

Summary of the Main Results
In our long-term cohort, VGA was prescribed more frequently in the most severe patients, including those with acute infective endocarditis. In univariate analysis, VGA was associated with a lower risk of sSSI, but this association did not persist in multivariate analysis. In addition, no association was found with sSSI and dSSI, even when performing three sensitivity analyses. Therefore, our study suggests that VGA could represent a good alternative to cephalosporins. However, our study also underlines the microbiological differences according to the type of prophylaxis. Specifically, VGA was associated with more Gram-negative bacteria and fewer Gram-positive bacteria in SSI.

SSI Occurrence According to the Type of Antibiotic Prophylaxis
The similar efficacy of CA and VGA to prevent SSI is consistent with the large multicentre retrospective study on 21,396 patients [27] and the 2012 meta-analysis, which included nine studies, regardless of the duration of antibiotic prophylaxis, five of which had a similar duration of administration of vancomycin and cephalosporins [14].
In the overall analysis of the meta-analysis that included 6022 patients, vancomycin administration was associated with a slight increase in the frequency of SSIs (OR = 1.33 [1.08-1.64]), regardless of the duration of antibiotic prophylaxis, but this association disappeared in the subgroup analysis that included 2278 patients with a similar duration of antibiotic prophylaxis (OR = 0.81 [0.57-1.15]).

Prevention of Gram-Negative SSI by Gentamicin
In addition, our study suggests an increased incidence of Gram-negative bacteria in the VGA group. This association was not found in the 2012 meta-analysis, although this analysis found more postoperative Gram-negative pneumonia in the vancomycin group. This lack of association in this meta-analysis may be due to a lack of power, as our population was more than twice as large [14]. The increase in Gram-negative SSI is surprising and deceptive, given that gentamicin was associated with vancomycin to reduce Gram-negative bacteria infection. This finding is in contrast to that noted in the metaanalysis, where vancomycin was used alone [14]. From this perspective, gentamicin seems to have a minimal impact on the prevention of SSI. A first explanation for this could be the high prevalence of gentamicin resistance in a nosocomial environment [28,29]. However, active endocarditis patients, who are at high risk of bacterial resistance due to long term antibiotherapy, had a low prevalence of SSI. In addition, the better coverage of Grampositive bacteria, such as enterococci, offered by vancomycin could favour Gram-negative infections. Indeed, cephalosporins are not effective against this pathogen [30,31], which may explain why, out of the 63 enterococcal SSIs in our study, 62 occurred in the CA group, compared with one in the vancomycin group. Therefore, for patients at high risk for SSI who are not allergic to beta-lactams, our results suggest that the combination of vancomycin and cefazolin could be tested to provide better coverage of Gram-negative and Gram-positive bacteria, as used by some teams [16]. A prospective multicentre study evaluating this combination would provide answers [32]. Otherwise, broad-spectrum cephalosporins, such as ceftaroline [33,34], could represent another option, providing a more protective effect for patients allergic to beta-lactam. This antibiotic offers two advantages for our combination of VGA. First, in the absence of veinotoxicity, it could be administered before central-line insertion. Second, it is less nephrotoxic, and the prophylactic administration of vancomycin and, especially, the combination of vancomycin/aminoglycoside has been associated with an increased risk of acute kidney injury [27]. However, no data are available for this indication. The results from 2014, assessing ceftaroline for this indication, are not available (clinical trial number = NCT02307006) [35], and the risk of selecting new resistance and the alteration in gut microbiota may limit its use.

Timing of Antibiotic Prophylaxis and Role of Gentamicin
Nevertheless, for patients allergic to beta-lactams, gentamicin may reinforce the administration of vancomycin. Indeed, in our study, the exact start date of antibiotic prophylaxis was not collected. However, in our protocol, VGA started before the surgical incision and after the insertion of the central line. Therefore, our protocol is not in accordance with French and international recommendations [6], which suggest that the vancomycin infusion should end before the surgical incision. Indeed, some studies have shown an increased risk of SSI when vancomycin is delayed [15,16]. Garey et al. [15] reported that starting vancomycin 0-15 min before the surgery incision was associated with an increased risk of SSI with an OR of 11.6 (2.6-52.4) compared to patients where vancomycin was started 16-60 min before incision. Three hypotheses could explain the discrepancy with our study. First, the population of the Garey et al. study was small (15 patients in the 0-to 15-min group and 176 in the 16-to 60-min group) and may lead to an overestimation of the risk. Second, some patients began receiving vancomycin from 15 to 30 min before incision, which is the optimal time according to this study. Third, gentamicin was not used in this study. This antibiotic offered a fast bactericidal effect [36], which is increased by vancomycin [37]. Therefore, gentamicin may offer some advantages.

Limits and Strengths
This study is one of the largest studies on SSI in cardiac surgery and, more specifically, on the combination of vancomycin/gentamicin. The inclusion of patients and their surgical management were stable over time. Thus, the SSI rate was stable throughout the study. However, the use of VGA therapy decreased over time, which was possibly related to a decrease in the MRSA rate in our population.
Other limitations are present. First, this is a retrospective study, which may lead to collection bias. This bias is limited by the fact that much of the information was collected prospectively. Second, it is a monocentric study, which may limit generalization. This study found SSI rates and risk factors for SSI in line with previously published studies [38][39][40][41], and the results are consistent with previously published articles [14,27]. Thus, we assume that this bias is limited. Third, the exact timing of the injection of the antibiotic was not collected, which does not allow for us to take this factor into account in the occurrence of SSI. Fifth, the patients in the VGA group had a higher mortality rate than those in the CA group. Given this heterogeneity, we performed a propensity score analysis, which showed similar results. We used the "full matching" method, which has the advantage of retaining the whole population by assigning a weight to each patient. Then, we performed two additional subgroup analyses, first by excluding active endocarditis and then by selecting a homogeneous population of patients undergoing isolated CABG. The fact that the results of all three sensitivity analyses concurred with the full population analysis limits this severity bias.

Conclusions
In this prolonged cohort of consecutive patients, the incidence and risk factors for SSI were comparable to those reported in the literature. VGA was more commonly used in the most severe patients than in CA. However, our results did not find any significant increase in risk of SSI. Differences in the microbiological cultures of SSI were observed according to the type of antibiotic prophylaxis. Therefore, VGA represents a valuable alternative for SSI prevention. Further studies are needed to confirm these results.

Supplementary Materials:
The following supporting information can be downloaded: https://www. mdpi.com/article/10.3390/antibiotics12010085/s1, Figure S1: Covariate balance in the propensity matched samples; Figure S2: Characteristics of the patients during the inclusion period; Figure S3: 90-day mortality according to the type of prophylaxis; Table S1: Perioperative characteristics of patients according to SSI subtype occurrence; Table S2: Multivariate analysis of risk factors for superficial and deep SSI; Table S3: Multivariate analysis of risk factors for superficial and deep SSI in the propensity score analysis; Table S4: Perioperative characteristics of patients according to the type of prophylaxis in the sensibility analysis excluding patients with active endocarditis; Table S5: Multivariate analysis of risk factors for superficial and deep SSI in the sensi-tivity analysis excluding active endocarditis; Table S6: Perioperative characteristics of patients according to the type of prophylaxis in the sensibility analysis restricted to patients with isolated CABG; Table S7: Multivariate analysis of risk factors for superficial and deep SSI in the sensi-tivity analysis restricted to patients undergoing isolated CABG.

Introduction
Background/rationale 2 Explain the scientific background and rationale for the investigation being reported 2 Objectives 3 State specific objectives, including any prespecified hypotheses 2, 3

Study design 4
Present key elements of study design early in the paper 2 Setting 5 Describe the setting, locations, and relevant dates, including periods of recruitment, exposure, follow-up, and data collection 2 Participants 6 (a) Give the eligibility criteria, and the sources and methods of selection of participants. Describe methods of follow-up 2 (b) For matched studies, give matching criteria and number of exposed and unexposed -

Participants 13
(a) Report numbers of individuals at each stage of study-e.g., numbers potentially eligible, examined for eligibility, confirmed eligible, included in the study, completing follow-up, and analysed 4 and Figure 1 (b) Give reasons for non-participation at each stage Figure 1 (c) Consider use of a flow diagram Figure 1

Item No Recommendation Page
Descriptive data 14 (a) Give characteristics of study participants (e.g., demographic, clinical, social) and information on exposures and potential confounders Table 1 (b) Indicate number of participants with missing data for each variable of interest Table 1 (c) Summarise follow-up time (e.g., average and total amount) Supplementary Figure S1 Outcome data 15 Report numbers of outcome events or summary measures over time Figure 2 Main results 16 (a) Give unadjusted estimates and, if applicable, confounder-adjusted estimates and their precision (e.g., 95% confidence interval). Make clear which confounders were adjusted for and why they were included Other analyses 17 Report other analyses done-e.g., analyses of subgroups and interactions, and sensitivity analyses Tables 4 and S2

Discussion
Key results 18 Summarise key results with reference to study objectives 12

Limitations 19
Discuss limitations of the study, taking into account sources of potential bias or imprecision. Discuss both direction and magnitude of any potential bias 13 Interpretation 20 Give a cautious overall interpretation of results considering objectives, limitations, multiplicity of analyses, results from similar studies, and other relevant evidence 12, 13 Generalisability 21 Discuss the generalisability (external validity) of the study results 14

Funding 22
Give the source of funding and the role of the funders for the present study and, if applicable, for the original study on which the present article is based 14