Prophylactic Topical Antibiotics in Fracture Repair and Spinal Fusion

Introduction The objective of this systematic review with meta-analysis is to determine whether prophylactic local antibiotics prevent surgical site infections (SSIs) in instrumented spinal fusions and traumatic fracture repair. A secondary objective is to investigate the effect of vancomycin, a common local antibiotic of choice, on the microbiology of SSIs. Methods An electronic search of PubMed, EMBASE, and Web of Science databases and major orthopedic surgery conferences was conducted to identify studies that (1) were instrumented spinal fusions or fracture repair and (2) had a treatment group that received prophylactic local antibiotics. Both randomized controlled trials (RCTs) and comparative observational studies were included. Meta-analysis was performed separately for randomized and nonrandomized studies with subgroup analysis by study design and antibiotic. Results Our review includes 44 articles (30 instrumented spinal fusions and 14 fracture repairs). Intrawound antibiotics significantly decreased the risk of developing SSIs in RCTs of fracture repair (RR 0.61, 95% CI: 0.40–0.93, I2 = 32.5%) but not RCTs of instrumented spinal fusion. Among observational studies, topical antibiotics significantly reduced the risk of SSIs in instrumented spinal fusions (OR 0.34, 95% CI: 0.27–0.43, I2 = 52.4%) and in fracture repair (OR 0.49, 95% CI: 0.37–0.65, I2 = 43.8%). Vancomycin powder decreased the risk of Gram-positive SSIs (OR 0.37, 95% CI: 0.27–0.51, I2 = 0.0%) and had no effect on Gram-negative SSIs (OR 0.95, 95% CI: 0.62–1.44, I2 = 0.0%). Conclusions Prophylactic intrawound antibiotic administration decreases the risk of SSIs in fracture surgical fixation in randomized studies. Therapeutic efficacy in instrumented spinal fusion was seen in only nonrandomized studies. Vancomycin appears to be an effective agent against Gram-positive pathogens. There is no evidence that local vancomycin powder is associated with an increased risk for Gram-negative infection.


Introduction
Surgical site infections (SSIs) are a significant source of morbidity and cost for patients undergoing orthopedic procedures. SSIs are challenging to treat because of the potential formation of a bacterial biofilm, an extracellular matrix that can attach to implants and protect pathogens from host immunity and systemic antibiotics [1]. Both instrumented spinal procedures and fracture surgeries have in common the use of metallic hardware, and each suffer from a nontrivial rate of SSIs, ranging from 9.4% of noninstrumented spinal trauma cases [2] to over 30% in lower extremity fracture cases [3,4]. SSIs lead to delayed healing, nonunion, irreversible loss of function, or amputation of the infected limb [3,4]. e current standard of care for SSI prevention is systemic antibiotics [5,6]. However, parenterally administered antibiotics have the disadvantages of delivering a reduced concentration of antibiotics to the targeted site, failing to reach poorly vascularized tissues, and potentially causing systemic toxicity. Alternatively, locally delivered antibiotics can achieve a high local concentration with low systemic levels, thereby avoiding dangerous side effects such as nephrotoxicity or ototoxicity [1].
Despite these strengths associated with local antibiotic therapy, there are also concerns. One is that a high concentration of antibiotics can potentially inhibit new bone formation [7]. Another concern is the development of antibiotic resistance [8] or the emergence of pathogens not covered under the narrow spectrum of commonly used antibiotics such as vancomycin [9][10][11]. Additionally, prophylaxis with local antibiotics is an off-label usage and may cause unforeseen adverse events.
Previous systematic reviews have examined the effect of intrawound antibiotics in either instrumented spinal procedures or open limb fractures [9,[12][13][14][15][16]. Because both fields share a high risk of infection and the objective of achieving bony union, which may be affected by local antibiotics, we believe there is value in reporting pooled outcomes of local antibiotics comprehensively from both specialties. e outcomes of local antibiotics in extremity fracture treatment may therefore be generalizable to spinal fusion surgery and vice versa. erefore, the aim of this systematic review is to assess the efficacy of locally administered antibiotics in preventing SSIs in instrumented spinal procedures and fracture surgeries and to investigate the effect of local vancomycin powder on SSIs caused by Gram-negative organisms.

Search Strategy.
We searched the PubMed, EMBASE, and Web of Science databases for our systematic review and meta-analysis. Keywords and phrases that guided our search strategy included "intrawound OR local" and "fracture OR fusion" and "prophylaxis OR prevent." e following MeSH terms were developed from key articles and used on PubMed: "anti-infective agents, local" and "antibiotic prophylaxis" (full search strategy in Appendix A). All relevant abstracts presented at major orthopedic conferences (Orthopaedic Trauma Association Conference, American Academy of Orthopaedic Surgeons Conference, North American Spine Society, Scoliosis Research Society) and available on the conference databases were included. Two additional articles were identified from the bibliographies of included articles and relevant review papers. Our initial search was performed in 2019; a secondary search was performed in 2021 to identify updated, relevant articles.

Eligibility Criteria.
Criteria for inclusion in this systemic review were studies that (1) included patients undergoing acute fracture repair or spinal fusion with instrumentation, (2) had a treatment group that received prophylactic local antibiotics, and (3) reported SSI as a primary outcome. We wanted to assess local antibiotic prophylaxis in instrumented procedures meant to achieve bone healing in adult studies.
We excluded studies of only pediatric patients. We included studies whose patient age range spanned children and adults because the age means with standard deviations of these articles indicated that the majority of the patients were adults. Studies of craniofacial surgeries were excluded because of the unique bacterial flora and vasculature of this anatomic region [17]. We excluded case series and studies without a control group. Studies of spinal decompression procedures without fusion (e.g., laminectomy) were excluded. For articles that investigated both decompression and/or instrumented procedures, only the infection results of instrumented cases were included. Studies in which the treatment group received other experimental therapies (e.g., antiseptics) in addition to local antibiotics were also excluded. Furthermore, there were some groups that published multiple studies on the same patient cohort, either at different time points during one collection period or with different subsets of the same data. In such cases, only the article with the greatest sample size was included. Finally, any articles that did not report patient data were excluded, such as narrative reviews, pharmacokinetic studies, and articles on novel antibiotic delivery systems. Any conflicts were resolved through discussion and consensus.

Study Identification and Data Extraction.
Two authors (EK and MT) individually conducted a title and abstract screening with DistillerSR (Evidence Partners, Ottawa, Canada).
e full text articles of selected studies were separately assessed by two authors (EK and CD). e data were extracted independently by two authors (EK and CD), including an assessment of patient population, local antibiotic of choice, sample size, method of controlling for bias, number of infections, and the culture results, if available. e Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram details the number of articles retrieved and excluded at each stage of the review ( Figure 1).

Meta-Analysis and Subgroup Analysis.
To study whether prophylactic local antibiotics reduce the risk of SSIs, separate meta-analyses were performed for the instrumented spinal procedure and fracture repair. Further, RCTs and observational studies were pooled separately, with subgroups of study design among nonrandomized studies (propensitymatched cohort study or nonpropensity-matched cohort study) (Figures 2-5 ). Risk ratios (RRs) and odds ratios (ORs) were calculated in the meta-analyses of RCTs and observational studies, respectively.
To evaluate the effect of vancomycin on Gram-positive and Gram-negative infections, data were pooled from articles that studied the use of vancomycin powder and reported SSI culture results. ese data were grouped into meta-analyses of Gram-positive vs. Gram-negative organisms with subgroup analysis for spine and fracture studies (Figures 6 and 7). Negative or mixed polymicrobial results (both Gram-positive and Gram-negative pathogens cultured in the same infection) were omitted. When multiple organisms that were either all Gram-positive or all Gramnegative were cultured in one SSI, they were counted as one. Studies that provided bacterial data for only a portion of the SSIs were excluded from this subgroup analysis.

Statistical
Analysis. STATA 16 software (Statacorp, College Station, TX) was used to conduct random-effects meta-analyses using the admetan command, which is built on the Mantel and Haenszel model to develop RRs and ORs for binary and continuous data [18,19]. e heterogeneity of the included studies was quantified with the I 2 statistic.

Spinal Instrumentation.
irty spinal instrumentation studies were included, with five RCTs, two prospective observational studies, and 23 retrospective studies (Table 1) . A total of 17,756 patients were included. Among the nonrandomized studies, three studies were propensitymatched [23,29,35]. ere was a wide variety in the type of instrumented procedure performed, reflecting a range of diagnoses across studies. Twenty-two of the 30 studies exclusively included fusion cases. e treatment group in 29 of 30 studies received vancomycin with varying dosages (0.5-2 g) and methods of application. e study that did not report the use of vancomycin did not specify either the antibiotic type or the dosage [23]. e primary outcome of included studies was SSI.
e pooled RR of infection in the treatment group compared to the control group across five spinal instrumentation RCT studies was 1.03 (95% CI: 0.56-1.91,
ere was considerable clinical heterogeneity among the studies, such as fracture location, antibiotic type, and definition of SSI. Data from Bibbo and Patel [50] did not contribute to the meta-analysis because both the treatment and control groups had a 0% SSI incidence.
e pooled RR of infection in the treatment group in the three RCTs was 0.61 (95% CI: 0.40-0.93, I 2 � 32.5%) ( Figure 4). Subgroup analyses by the study type (propensity-matched cohort or non-propensity-matched cohort studies) among observational studies were also performed. Because there was only one propensity-matched cohort study, it was not pooled. e pooled odds ratio (OR) of infection in the treatment group compared to the control group across 10 observational fracture studies was 0.49 (95% CI: 0.37-0.65 I 2 � 43.8%) ( Figure 5). In the subgroup analysis by study type, the pooled OR for nine non-propensity matched cohort studies was 0.51 (95% CI: 0.39-0.68, Because the fracture studies used a variety of antibiotics, subgroup analysis by antibiotic type was conducted, which pooled studies of all designs together (Appendix B). For the five vancomycin studies, the pooled OR was 0.71 (95% CI: 0.51-0.98, I 2 � 0.0%). e pooled OR for the three tobramycin studies was 0.31 (95% CI: 0.19-0.50, I 2 � 27.4%). Two studies used both vancomycin and tobramycin and showed a pooled OR of 0.41 (95% CI: 0.18-0.94, I 2 � 0.0%). e study by Bibbo and Patel [50] was not included because both the treatment and control groups had a 0% SSI incidence. ree remaining studies [52,54,55] each had a unique antibiotic regimen and were not pooled with other studies.

Microbiology in Instrumented Spinal Procedures and Fracture Surgeries at Used Vancomycin.
To address the concern that the use of local vancomycin powder can affect the incidence of Gram-negative infection, two metaanalyses spanning 20 studies that studied local vancomycin and reported culture data were conducted. e pooled OR for an infection caused by Gram-positive bacteria in the vancomycin group compared to the control group was 0.37 (95% CI: 0.27-0.51, I 2 � 0.0%) ( Figure 6). Subgroup analysis by spine and fracture cases revealed an OR of 0.33 (95% CI: 0.22-0.50, I 2 � 12.6%) in instrumented spinal procedures and OR of 0.46 (95% CI: 0.26-0.83, I 2 � 0.0%) in fracture repairs. e pooled OR for Gram-negative infection in the vancomycin group was 0.95 (95% CI: 0.62-1.44, I 2 � 0.0%) (Figure 7). e subgroup analysis showed that the OR was 0.83 (95% CI: 0.50-1.39, I 2 � 0.0%) for spinal instrumentation and 1.23 (95% CI: 0.60-2.55, I 2 � 0.0%) for fracture surgeries. ree studies were excluded in the meta-analysis of Gramnegative infections because both control and treatment groups had no Gram-negative SSIs. Subgroup analyses by study type showed a similar trend towards greater effect size among nonrandomized studies.

Discussion
We performed a meta-analysis of 44 studies evaluating the effect of locally administered antibiotics on rates of infection after instrumented spine and fracture surgeries. Notable findings include a significant reduction in the pooled incidence of infection in both patient populations, but this effect was weaker or absent with more rigorous study designs. e pooled effect of vancomycin was significant for the reduction in Gram-positive infection and did not show any association with Gram-negative infection compared to no local antibiotics.

Spinal Instrumentation.
Previous systematic reviews have demonstrated the benefit of local antibiotics and antiseptic prophylaxis. Dodson et al. [9] pooled 21 studies (2 RCTs and 19 observational studies) and found that  Surgical procedures with instrumentation have a higher risk of biofilm formation [1]. erefore, our study focused exclusively on instrumented procedures in adult patient populations and yielded a result consistent with other systematic reviews [9,12]. Neither the meta-analysis of the five RCTs nor the metaanalysis of the three propensity-matched cohort studies showed the same significant reduction that occurred with the pooling of cohort studies. Many observational studies used a "before-andafter" study design that is prone to confounding bias, which may explain the greater effect size observed in non-propensitymatched studies [64]. e blinded RCT remains the methodological gold standard for proving the efficacy of therapeutic intervention; it is important that any future observational studies incorporate design and analytical methods to control for bias, such as propensity score adjustment [65].     a We combined the two treatment groups into one intervention group in our analysis. b e treatment group received just antibiotic beads, and the control group received just parenteral antibiotics. Not included are the nonrandomized third cohort that received antibiotic beads + IV. is group of patients all had Grade 3 Gustilo-Anderson open fractures. c We included only deep SSI, which was the primary study outcome. Superficial SSI was excluded because the sample sizes for superficial SSI did not match those for deep SSI. d is study conducted analyses using two separate methods of matching: nearest-neighbor matching and propensity score matching. It also had both prospective and retrospective control cohorts. We included the data from propensity scores matching with the prospective control cohort. ‡ Only deep infections were reported in this study. * Studies showed a significant difference between the control and treatment groups.   Our meta-analysis showed a pooled benefit of prophylactic intrawound antibiotics in both randomized and nonrandomized studies. Similar to our analysis of spine instrumentation studies, bias reduction from randomization revealed that the magnitude of the effect is likely to be smaller than previously thought yet trending towards a protective effect. Subgroup analysis by antibiotic type showed that local vancomycin reduced SSI in fracture repair, which is in line with the established coverage pattern of vancomycin and common infectious organisms. Larger studies and trials that combine vancomycin with agents with Gram-negative coverage may be required to achieve the precision and possible added magnitude of effect to prove the impact of local antibiotics for the prevention of SSI in this population.

Microbiology with the Use of Vancomycin Powder.
Our meta-analyses assessed the microbiology of SSIs by pooling culture data from studies that used vancomycin powder. We specifically addressed vancomycin because of its widespread use and concern that its selective coverage of Gram-positive pathogens may increase the incidence of Gram-negative infection [1]. In a study of 2802 patients undergoing spinal surgery, Chotai et al. [11] observed a lower incidence of deep SSIs in the vancomycin group. However, there was a higher percentage of SSI caused by Gram-negative organisms in the vancomycin group than in the control group (28% vs. 12.5%).
We demonstrated that vancomycin reduces Gram-positive infection and has no effect on Gram-negative infection in both instrumented spinal procedures and fracture surgeries. e effectiveness of vancomycin against Gram-positive pathogens is consistent with its established antibacterial spectrum covering some of the most commonly cultured organisms in SSI of both instrumented spinal procedures and   fractures, including Staphylococcus aureus [8,10,11]. Furthermore, our results are reassuring to orthopedic surgeons who are apprehensive about the potential for vancomycin to increase the incidence of Gram-negative infections.

Adverse Events.
No studies reported any adverse events attributable to local antibiotics. e majority of the included studies had a single sentence denying side effects. Some studies explicitly reported that intrawound antibiotics did not impact the rates of nonunion, addressing the concern that topical antibiotics can impede bone healing [7,31,42,50,54,57]. It is important to note that most of the included studies were not powered to detect differences in pseudarthrosis.

Limitations
e primary limitation of this study is the pooling of cohort studies in the meta-analysis. We intentionally included both RCTs and observational studies because there are very few RCTs that investigate the prophylactic effect of intrawound antibiotics, but the RCTs and nonrandomized studies were analyzed separately. e majority of the current evidence is from observational studies. A high degree of heterogeneity in study design, outcome assessment, treatment protocols,   and definition of SSI existed among the included studies. ere was a hierarchy of study designs among observational studies across which differences in effect were identified.

Conclusion
Prophylactic topical antibiotics are associated with decreased risk of surgical site infection after both instrumented spine and fracture surgeries in much of the published literature on the topic. Although the effect is weak or absent in more rigorous study designs in the instrumented spinal fusion literature, pooling of fracture repair RCTs revealed that intrawound antibiotics significantly reduced SSIs. ere is no evidence to suggest a higher incidence of Gramnegative infection or other adverse events among patients treated with local vancomycin, irrespective of study quality.
ese results do not support the use of local antibiotics in patients undergoing spinal fusion but suggest therapeutic efficacy in patients undergoing fracture repair.
Data Availability e search criteria used for article discovery are included as Supplementary Table A. All data included in this systematic review and meta-analysis were aggregated from previously published sources and have been collected from anonymous subjects.