Do Triclosan Sutures Modify the Microbial Diversity of Surgical Site Infections? A Systematic Review and Meta-Analysis

Randomised controlled clinical trials (RCTs) report a lower incidence rate of surgical site infections (SSIs) with triclosan sutures (TSs) compared with non-triclosan sutures (NTSs). Do triclosan sutures modify the microbial diversity of culture-confirmed SSIs (ccSSIs)? If so, this would support the association between TS antimicrobial activity and the SSI incidence rate. This prospective systematic literature review (PROSPERO CRD42019125099) was conducted according to PRISMA. RCTs that compared the incidence of SSIs with TSs and NTSs and reported microbial counts from SSI cultures per suture group were eligible. The microbial species were grouped by genus, and the association between genera and sutures was tested. The pooled relative risk (RR) of ccSSIs was also calculated. Twelve RCTs were eligible. No publication bias was identified. The microorganism count was 180 in 124 SSIs with TSs versus 246 in 199 SSIs with NTSs. No significant difference in microbial diversity was found, but statistical power was low for test results to support or challenge the association between the antimicrobial activity of TSs and the reduced rate of SSIs. The RR of the ccSSIs was significant and consistent with comprehensive meta-analyses. The certainty of the pooled RR was moderate.


Introduction
Surgical site infections (SSI) are diagnosed up to 30 days postoperatively, although some guidelines extend the duration up to one year in prosthetic surgery. SSIs are superficial incisional (skin and subcutaneous tissue), deep incisional (fascia and muscle), and organ/space [1,2]. SSIs may extend across the three domains. SSI surveillance networks report a wide range of incidence rates across operations; e.g., from 0.5% [0. 2, 2.7] in prosthetic knee surgery to 10.1% [4.1, 16.9] in laparotomic colon surgery [3].
The precursor of SSI is microbial contamination, and the conceptual relationship of SSI risk has three factors (Formula (1)) [4]: SSI risk = bacterial dose × virulence resistance of the host patient (1) Virulence refers to disease severity associated with a microorganism. One proposed definition is "the proportion of clinically apparent cases that are severe or fatal" [5]. Virulence varies across microorganisms [6][7][8]. Microorganisms involved in SSI have been reported to originate mainly from the skin, surrounding tissues of the incision, or operated organs with microbial flora such as the bowel [9]. Concerning the bacterial dose, surgical sites contaminated with more than 10 5 /grammes of tissue have a significantly increased risk of SSI [10]. Much lower doses can produce an SSI when foreign material is inside the

Search Strategy
PubMed, Embase, Web of Science, and the Cochrane Library (including CENTRAL) were searched using the following string: "triclosan AND (suture OR sutures OR ligation OR ligations) AND (surgery OR surgeries OR surgical OR operation OR operations) AND ((systematic AND review) OR random* OR RCT OR guide* OR recom* OR meta-analy* OR metaanaly*)" [29,30]. No exclusion filter was applied. The extraction was up to date on 18 August 2021. Appendix A displays the search strategy as implemented in each database (Tables A1-A4).

Eligibility Criteria
Prospective parallel-group RCTs that met the PICO specifications were eligible. Posters, abstracts, communications, and studies that did not report institutional review board or ethics committee approval and patient informed consent were excluded.

Study Selection, Data Extraction, and Risk of Bias Assessment
Two reviewers (F.D. and M.C.) independently conducted the three steps, and the differences were adjudicated by a third reviewer (N.M.). All references were imported into a repository (EndNote X8, Clarivate Analytics, Philadelphia, PA, USA). Eligibility was determined by reading titles, abstracts, and full text. Duplicates were flagged, and multiple publications about the same study were grouped for joint review. Additional studies were identified from the references of RCTs and previous SLRs. Automated queries were used for post hoc verification.
Potentially relevant RCTs were exported to Review Manager (RevMan) 5.4 software (The Cochrane Collaboration, 2020). The individual RCT risk of bias was assessed using the seven items of the built-in risk of bias (RoB) tables [47].

Extracted Data
Data were extracted in standardised tables:

1.
Study characteristics: Design, committee approval and informed consent, study registration, statistical methods including power calculation, screening methods, treatment allocation and blinding details, sponsor details, enrollment period and sites, inclusion sites, patient inclusion and exclusion criteria, patient demographics, clinical indication, type of surgery, suture material by suture group, SSI prevention details, and additional patient groups.
Number of patients with a ccSSI by suture type and list of microorganisms per culture or the aggregate count of each microbial designation. When microbial percentages were reported, counts were calculated using the corresponding total number.

Microbial Data Analysis
For descriptive analysis, microbial counts were summed in a spreadsheet according to designation and suture group. The relative frequency of each cell was calculated.
The counts by original designation were then summed according to genus and suture group in a contingency table. Microorganisms that could not be traced to their genus and genera with an expected count of less than n = 5 per cell were excluded from the analysis.
The independence of genera and sutures was tested with Pearson's chi-squared and Fisher's exact tests. The significance threshold was p < 0.05. The measure of association between genera and sutures was Cramér's V (0-0.29 weak association, 0.3-0.59 moderate, 0.6-1 strong) [48].
The robustness to sensitivity analysis was tested by iteratively repeating the contingency table analysis with the data of one study removed.

Consistency with Clinical Outcomes
The consistency of microbiological findings with clinical outcomes was assessed by comparing results with the eligible studies' RRs of ccSSIs (TSs over NTSs). A risk of publication bias was suspected if the funnel plot of the RR was asymmetrical or if Harbord's test for binary variables was significant (i.e., p < 0.05) [49,50].
The heterogeneity of the distribution of the RCTs' RRs was tested with Cochran's Q-test (threshold: p ≥ 0.05) and the I 2 statistic, the percentage of variation across the RCTs' RRs due to heterogeneity rather than chance. The heterogeneity was considered high if I 2 > 25% [51][52][53][54][55][56]. The robustness of test results was assessed with a sensitivity analysis.
The contingency table analysis, sensitivity analysis, power calculation, and Harbord's tests were computed in STATA 17 (StataCorp LLC, College Station, TX, USA). The overall bias summary, stratified pooling of RR, heterogeneity analysis, and figure creations were performed in Review Manager 5.3. The risk of bias of the individual RCTs was summarised graphically with Review Manager's automated table coupled with a forest plot of the RRs. The level of certainty of the pooled RR of the ccSSIs was rated according to GRADE [57].

Study Identification and Selection
A total of 49 records concerning 33 RCTs were in the clinical scope; 20 of them concerning 12 RCTs fulfilled the PICO specifications and were included in the pooled analysis ( Figure 1). between genera and sutures was Cramér's V (0-0.29 weak association, 0.3-0.59 moderate, 0.6-1 strong) [48].
The robustness to sensitivity analysis was tested by iteratively repeating the contingency table analysis with the data of one study removed.

Consistency with Clinical Outcomes
The consistency of microbiological findings with clinical outcomes was assessed by comparing results with the eligible studies' RRs of ccSSIs (TSs over NTSs). A risk of publication bias was suspected if the funnel plot of the RR was asymmetrical or if Harbord's test for binary variables was significant (i.e., p < 0.05) [49,50].
The heterogeneity of the distribution of the RCTs' RRs was tested with Cochran's Qtest (threshold: p ≥ 0.05) and the I² statistic, the percentage of variation across the RCTs' RRs due to heterogeneity rather than chance. The heterogeneity was considered high if I² > 25% [51][52][53][54][55][56]. The robustness of test results was assessed with a sensitivity analysis.
The contingency table analysis, sensitivity analysis, power calculation, and Harbord's tests were computed in STATA 17 (StataCorp LLC, College Station, TX, USA). The overall bias summary, stratified pooling of RR, heterogeneity analysis, and figure creations were performed in Review Manager 5.3. The risk of bias of the individual RCTs was summarised graphically with Review Manager's automated table coupled with a forest plot of the RRs. The level of certainty of the pooled RR of the ccSSIs was rated according to GRADE [57].

Study Identification and Selection
A total of 49 records concerning 33 RCTs were in the clinical scope; 20 of them concerning 12 RCTs fulfilled the PICO specifications and were included in the pooled analysis ( Figure 1).
The summary of characteristics of the eligible studies showed that half of them were about abdominal surgery (mainly digestive, but also pilonidal and others). The others focused on cardiovascular operations, knee arthroplasty, and neurosurgery ( Table 2). Polyglactin sutures were the most frequently compared (83% of the studies), followed by polydioxanone (33%) and polyglecaprone 25 (once). One-third of studies compared associations of TSs. The counting of microorganisms was straightforward in all but two studies. In Jüstinger 2013, counts were calculated by multiplying the number of ccSSIs by the corresponding percentages of the microorganisms and then rounding decimals to the nearest integer. In Isik 2012, the random allocation ratio was 1 TS to 2 NTSs, thus unbalancing the microbial and SSI counts.
The risk of bias varied significantly across RCTs. The RoB tables of the included RCTs with the supportive information used to rate each item are displayed in Appendix B (Tables A5-A16).

Microbial Diversity
Microbial diversity consisted of 34 reported species, including remarkable strains (e.g., MRSA) and genera (e.g., Staphylococcus spp.) ( Table 3). The individual counts were too low to compare the relative frequencies between TSs and NTSs. E. coli was the most frequent species, and the only one with a significant RR of 0.58 [0.37, 0.92], with fewer cases in TSs.  The microorganisms were grouped in the contingency table according to eight phylogenetic genera ( Table 3). The genera that were excluded due to an expected count of less than five per cell were Proteus, Citrobacter, Morganella, Corynebacterium, Moraxella, Serratia, and Peptostreptococcus. Thirty cases designated as polymicrobial or "other bacteria" were also excluded.
The 2-by-8 contingency table had 375 microorganisms, 39% in the TS arm and 61% in NTS ( Table 4). The association between genera and sutures was weak (Cramér's V = 0.11) and nonsignificant (chi-squared p = 0.72). The power calculated post hoc was low (1 − β = 0.28). The sensitivity analysis (Supplementary Materials) did not change the conclusions of the overall table and the subtables, so no RCT was identified as a significant cause of bias in the microbial diversity analysis.
The null hypothesis was not rejected.

Clinical Outcomes
The funnel plot ( Figure 2) showed moderate asymmetry, and Harbord's test was nonsignificant (p = 0.27). Therefore, no publication bias was detected.

Clinical Outcomes
The funnel plot ( Figure 2) showed moderate asymmetry, and Harbord's test was nonsignificant (p = 0.27). Therefore, no publication bias was detected.  The visual display of RoB for each item and each included RCT is next to the forest plot of the pooled RR ( Figure 3).    The overall RR was robust to the sensitivity analysis (Supplementary Materials). The level of certainty of the evidence underlying the overall pooled RR of the cultureconfirmed SSIs was rated moderate according to GRADE (Table 5). The overall RR was robust to the sensitivity analysis (Supplementary Materials). The level of certainty of the evidence underlying the overall pooled RR of the cultureconfirmed SSIs was rated moderate according to GRADE (Table 5).

Discussion
This review tested if SSI microbial diversity differed between the TS and NTS groups. The protocol assumed that if the TS antimicrobial activity reduced the incidence of SSIs, then SSI cultures' microbial counts would reflect the microorganism's triclosan susceptibility.
The absence of a significant difference in the SSIs' microbial diversity after TSs and NTSs should challenge the association between the difference in the incidence rate of SSIs after TSs and NTSs. However, the statistical power of the chi-squared was low (28%), so the test results could have resulted from chance, and both hypotheses remain plausible.
The power calculation showed that multiplying all cells of the contingency table by 3.5 with the observed proportions would result in a significant chi-squared test result, with p = 0.03 and a power of 84%. Such an increase would require a total microbial count of n = 1309. However, such a scenario would still challenge the association between TS and SSI incidence reduction, because the contribution of Pseudomonas to the overall lower microbial count in the TS column, with a 0.64 ratio, would be confirmed. Therefore, adding more microbial counts from RCTs would need to show a significant shift of the Pseudomonas ratio towards one to demonstrate that Pseudomonas were equally frequent in the SSIs of the TS and NTS arms, whereas triclosan-susceptible species remained fewer.
The 12% of excluded culture results were insufficient to bias the contingency table significantly. The designated species are usually intrinsically triclosan-susceptible, and the unspecified cases had an expected 10% triclosan-resistant microorganisms.
No similar study was previously published, so the differences in microbial diversity between the TS and NTS groups of this study could not be compared with other sources.
However, the overall microbial diversity in this study was consistent with the European 2017 SSI surveillance report, in which Staphylococcus and Escherichia were the most frequent genera, and P. aeruginosa represented 4.7% of microorganisms [3]. The microbial diversity was also reasonably consistent with a study of retrieved sutures from SSIs in which Staphylococcus was the most frequent genus, and P. aeruginosa represented about 5% of microorganisms [75].
The CI of the overall pooled RR of ccSSIs overlapped with the CI of the most comprehensive meta-analysis of RCTs published (Ahmed 2019) [32]. The two studies also agreed in rating the level of evidence as moderate. These similarities suggested that the evidence used here represented the evidence used in Ahmed 2019.
The two limitations of the quality of the evidence in the 12 pooled RCTs; i.e., (1) the minority of significant studies and (2) the uncertain or high risk of bias in about half of the rated points, along with the nonconclusive test of the primary criterion, suggested implementing the WHO conditional guideline with caution. One approach could be making TSs available in routine surgeries for patients with a high risk of SSI or severe SSI complications. Systematically collecting SSI culture details in priority patient groups operated with TSs or NTSs with a minimal clinical dataset incorporated in current surveillance programs would enable an analysis of real-life practice data with evidence from RCTs. That would give those patients a chance to reduce SSI risk with an acceptable risk of adverse suture effects and enable the gathering of evidence to assess the impact of TSs on SSI microbial diversity and ecology. Close monitoring of triclosan-resistant microorganisms such as the Pseudomonas genus and mutant strains of usually triclosan-susceptible genera require specific focus.

Conclusions
This systematic literature review of randomised controlled clinical trials did not show a significant difference in the microbial diversities of surgical site infections after closure using sutures with or without triclosan. However, the amount of evidence was insufficient to support or challenge the relationship between the antimicrobial activities of sutures with triclosan and the incidence rate of surgical site infections.
The meta-analysis of the relative risk of culture-confirmed surgical site infections favoured sutures with triclosan and was consistent with comprehensive meta-analyses. The certainty of the pooled RR was confirmed as moderate.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/microorganisms10050927/s1. UBordeaux13042022. Supplement. List of excluded randomised clinical trials; Table S1: Source data-microbial count suture treatment arm and per study; Table S2. Sensitivity analysis of the relative risk of culture-confirmed SSIs; Table S3. Sensitivity analysis of the association between genera and suture types.

Conflicts of Interest:
The authors declare no conflict of interest.

Appendix A. Executed Search Strategies
All tables below were set with the same string with the wildcard "*"meaning that the search engine should also retrieve words that are followed with any character. For example: "random*" sets the search engine to retrieve: random, randomised, randomized, randomly, etc. This applies to all four search engines. In Table A1, Pubmed has replaced the "*" with all available explicit words. In Tables A2-A4, the search engine did not.         dual-inclusions between the two suture groups was not accurately reported, and two observations in the same patient were not statistically independent.