ORIGINAL RESEARCH

Preoperative Versus Extended PostoperativeAntimicrobial Prophylaxis of Surgical Site InfectionDuring Spinal Surgery: A Comprehensive SystematicReview and Meta-Analysis

Blaine T. Phillips . Emma S. Sheldon . Vwaire Orhurhu .

Robert A. Ravinsky . Monika E. Freiser . Morteza Asgarzadeh .

Omar Viswanath . Alan D. Kaye . Marie Roguski

Received: March 21, 2019 / Published online: May 15, 2020� The Author(s) 2020

ABSTRACT

Introduction: Surgical site infection (SSI) fol-lowing spinal surgery is a major source of post-operative morbidity. Although studies havedemonstrated perioperative antimicrobial pro-phylaxis (AMP) to be beneficial in the preven-tion of SSI among spinal surgery patients,

consensus is lacking over whether preoperativeor extended postoperative AMP is most effica-cious. To date, no meta-analysis has investi-gated the comparative efficacy of these twotemporally variable AMP protocols in spinalsurgery. We undertook a systemic review andmeta-analysis to determine whether extendedpostoperative AMP is associated with a differ-ence in the rate of SSI occurrence among adultpatients undergoing spinal surgery.Methods: Embase and MEDLINE databaseswere systematically searched for clinical trialsand cohort studies directly comparing SSI ratesamong adult spinal surgery patients receivingeither preoperative or extended postoperativeAMP. Quality of evidence of the overall studypopulation was evaluated using the Grading of

Digital Features To view digital features for this articlego to https://doi.org/10.6084/m9.figshare.12205415.

Electronic supplementary material The onlineversion of this article (https://doi.org/10.1007/s12325-020-01371-5) contains supplementary material, which isavailable to authorized users.

B. T. Phillips � E. S. Sheldon � V. Orhurhu �R. A. Ravinsky � M. E. Freiser � M. Asgarzadeh �M. RoguskiHarvard T.H. Chan School of Public Health, HarvardUniversity, Boston, MA, USA

B. T. PhillipsHarvard Medical School, Harvard University,Boston, MA, USA

R. A. RavinskyDivision of Orthopedic Surgery, Department ofSurgery, University of Toronto, Toronto, ON,Canada

M. E. FreiserUniversity of Miami Miller School of Medicine,University of Miami, Miami, FL, USA

V. Orhurhu (&)Department of Anesthesia Critical Care and PainMedicine, Massachusetts General Hospital, HarvardMedical School, Boston, MA, USAe-mail: vwo569@mail.harvard.edu

O. ViswanathDepartment of Anesthesiology, CreightonUniversity School of Medicine, Omaha, NE, USA

A. D. KayeDepartment of Anesthesiology, Louisiana StateUniversity Health Sciences Center, New Orleans, LA,USA

M. RoguskiDepartment of Neurosurgery, Tufts Medical Center,

Adv Ther (2020) 37:2710–2733

https://doi.org/10.1007/s12325-020-01371-5

Recommendations Assessment, Developmentand Evaluation (GRADE) Working Groupapproach. Random effects meta-analyses wereperformed utilizing both pooled and stratifieddata based on instrumentation use.Results: Five studies met inclusion criteria. Noindividual study demonstrated a significantdifference in the rate of SSI occurrence betweenpreoperative and extended postoperative AMPprotocols. The GRADE quality of evidence waslow. Among the overall cohort of 2824 patients,96% underwent lumbar spinal surgery. PooledSSI rates were 1.38% (26/1887) for patientsreceiving extended postoperative AMP and1.28% (12/937) for patients only receiving pre-operative AMP. The risk of SSI developmentamong patients receiving extended postopera-tive AMP was not significantly different fromthe risk of SSI development among patientsonly receiving preoperative AMP [RR (risk ratio),1.11; 95% CI (confidence interval) 0.53–2.36;p = 0.78]. The difference in risk of SSI develop-ment when comparing extended postoperativeAMP to preoperative AMP was also not signifi-cant for both instrumented (RR, 0.92; 95% CI0.15–5.75; p = 0.93) and non-instrumentedspinal surgery (RR, 1.25; 95% CI 0.49–3.17;p = 0.65). There was no evidence of hetero-geneity of treatment effects for all meta-analyses.Conclusion: Preoperative AMP appears to pro-vide equivalent protection against SSI develop-ment when compared to extendedpostoperative AMP. Prudent antibiotic use isalso known to decrease hospital length of stay,healthcare expenditure, and risk of complica-tions. However, until higher-quality evidencebecomes available regarding AMP in spinal sur-gery, surgeons should continue to exercise dis-cretion and clinical judgment when weighingthe effects of patient comorbidities and com-plications before determining the optimalduration of perioperative AMP.

Keywords: Antimicrobial prophylaxis; Meta-analysis; Spinal surgery; Surgical site infection;Systematic review

Key Summary Points

Surgical site infection (SSI) followingspinal surgery is a major source ofoperative morbidity

Studies have demonstrated perioperativeantimicrobial prophylaxis (AMP) to bebeneficial in the prevention of SSI amongspinal surgery patients

However, consensus is lacking overwhether preoperative or extendedpostoperative AMP is most efficacious

Our meta-analysis investigated whetherextended postoperative AMP is associatedwith a difference in the rate of SSIoccurrence compared to preoperativeAMP among adult patients undergoingspinal surgery

Preoperative AMP appears to provideequivalent protection against SSIdevelopment when compared to extendedpostoperative AMP in patients undergoingspine surgery

INTRODUCTION

Surgical site infection (SSI) following spinalsurgery is a major source of operative morbidity.Its incidence ranges from 0.22% to 16.4%,demonstrating substantial variation conditionalupon numerous patient, disease, and perioper-ative factors [1–19]. Treatment requires pro-longed antibiotic therapy and often requiresreoperation. Although perioperative antimicro-bial prophylaxis (AMP) has proven effective inthe prevention of SSI following spinal surgery[18, 19], uncertainty remains regarding itsoptimal duration.

Preoperative and extended postoperativeAMP have been shown to be clinically equiva-lent in major surgery [20, 21]. Accordingly, theCenters for Disease Control and Prevention(CDC) recommends maintaining AMP

Adv Ther (2020) 37:2710–2733 2711

throughout the operation until, at most, a fewhours after incision closure [22]. Although nospinal surgery study to date has presented sig-nificant evidence that extended postoperativeAMP is superior to preoperative AMP, spinalsurgeons still hesitate to assimilate this evidencein the provision of care [22–31] and often con-tinue AMP into the postoperative period[32, 33].

The rationale for this skepticism remainsunclear, but proposed explanations suggest thatthe relative quality and number of availablestudies investigating the efficacy of temporallyvariable perioperative AMP protocols in spinalsurgery have left surgeons reluctant to general-ize broad results to their subspecialty beforeadditional high-quality evidence becomesavailable [27, 28]. Another possible explanationcould come from disagreement among majorsurgery guidelines, discrepancies between sub-specialty practices [34–36], and the lack of rec-ommendations specific to spinal surgery. Inparticular, performance measures endorsed bythe Centers for Medicare and Medicaid Servicesare contingent upon discontinuation of AMPwithin 24 h after the end of major surgery[36–39]. Meanwhile, only recently have for-malized guidelines specific to spinal surgerybeen provided by the North American SpineSociety (NASS), suggesting that AMP be dis-continued with intraoperative dosing for adultpatients undergoing typical, uncomplicatedspinal surgery [23–25].

If extended postoperative AMP does notprovide a concomitant benefit in the preven-tion of SSI after spinal surgery, continuation ofAMP postoperatively may unnecessarily exposepatients to adverse drug reactions, additionalcosts, and prolonged hospital stays whilesimultaneously promoting development ofantimicrobial resistant bacterial (AMR) strains[19, 26] and Clostridium difficile infection.Therefore, this systematic review and meta-analysis aims to assess the comparative efficacyof extended postoperative AMP versus preoper-ative AMP in the prevention of SSI among adultpatients undergoing spinal surgery.

METHODS

Study Selection

We conducted a systematic review of Embaseand MEDLINE databases through February 1,2019. The study selection process was tailoredto identify published randomized controlledtrials (RCT) and cohort studies comparing therelative efficacy of preoperative and extendedpostoperative AMP protocols in the preventionof SSI among adult patients undergoing spinalsurgery. Only those studies that used SSIoccurrence as their primary outcome of interestwere considered. Our search terms included‘‘spine surgery’’ or ‘‘spine injury’’ or ‘‘foramino-tomy’’ or ‘‘laminectomy’’ or ‘‘spine fusion’’ or‘‘antibacterial agent’’ and ‘‘antibiotics’’, or ‘‘ce-fazolin’’ or ‘‘clindamycin’’ or ‘‘cefuroxime’’ or‘‘vancomycin’’ and ‘‘postoperative period’’ or‘‘perioperative care’’ or ‘‘postoperative care’’ or‘‘intraoperative period’’ or ‘‘preoperative period’’or ‘‘intraoperative period’’. SupplementalTable 1 provides more details on the searchmethodology in the present investigation. Sinceidentification of SSI involves interpretation ofclinical and laboratory findings, included stud-ies were required to provide evidence thatdiagnosis was at least partially commensuratewith current CDC guidelines to ensure consis-tency [22, 40]. Given that retrospective studieswere included in this review, the only parame-ter that was afforded exception in meeting SSIdefining criteria was the length of the observa-tional window in which diagnosis could bemade. Postoperative systemic infections andremote coexisting and complicating infectionsdid not meet SSI criteria [40] and were notincluded. This method of study selection issimilar to that implemented by McDonald et al.[20]. Secondary outcomes included the inci-sional type of SSI and the microbiology andresistance patterns of positive wound cultures.

Included studies needed to directly comparerates of spinal SSI between preoperative andextended postoperative AMP protocols. Studiesthat compared one perioperative AMP protocolwith placebo and those that solely assessedvariations of extended postoperative AMP were

2712 Adv Ther (2020) 37:2710–2733

excluded. Studies contrasting different treat-ment arms of preoperative and extended post-operative AMP regimens were included but weremanipulated a priori in this analysis so thatthere was only one preoperative AMP group,one extended postoperative AMP group, andone risk ratio (RR) for each study. Pharmacoki-netic studies evaluating the comparative effi-cacy and role of two or more differentantibiotics in spinal SSI prevention were exclu-ded. There were no inclusion criteria condi-tional upon the dosage and type of antibioticsadministered, and perioperative changes madeto these parameters for clinical reasons werepermitted.

The different AMP regimens utilized byincluded studies were obligated to fulfill specificcriteria. Preoperative AMP regimens requiredadministration of a single antibiotic dose priorto surgery and allowed for intraoperativeredosing; however, antibiotic administration atthe conclusion of the procedure or any timepostoperatively was not permitted [20]. Mean-while, criteria for extended postoperative AMPregimens were identical to those for preopera-tive AMP protocols, but also allowed for con-tinued antibiotic administration into thepostoperative period for an indefinite length oftime. Intraoperative redosing and its timing forboth prophylactic regimens needed to be con-tingent upon several factors, including opera-tive duration; half-life of the antimicrobialagents administered; patient characteristicssuch as renal function, body habitus, and age;amount of blood loss and volume of subsequenttransfusions and fluid therapy; and rate of dif-fusion into the wound, in order to maintainbactericidal concentrations of the antibiotic[20, 22, 30, 41–43]. The terms preoperative andextended postoperative AMP were used in thisstudy instead of single-dose and multiple-doseprophylaxis in an attempt to avoid confusion.Single-dose prophylaxis can, by convention,include multiple doses of antibiotics if admin-istered intraoperatively [20], and can easily bemisinterpreted as AMP characterized by multi-ple doses.

Included studies were required to have sam-ple sizes greater than or equal to 20 patientswith at least 10 patients in each treatment arm.

Case–control studies were also excluded toavoid including a potentially large source ofbias. All studies that were not accessible inEnglish were excluded.

There were no specific exclusion criteriaplaced upon adult age and comorbidity statussince this would eliminate the majority ofstudies without access to individual patient datafor pooled analysis. However, studies primarilyassessing spinal SSI among pediatric patientpopulations were excluded. There were also noexclusion criteria surrounding details of thespinal operation performed, including the typeof procedure and its elective or emergent nat-ure, surgical complexity and duration, instru-mentation use, anatomic level of surgery andwhether surgeons operated at multiple spinallevels, revision surgery, and use of surgical siteirrigation with antibiotics.

Data Extraction

Using a standardized data extraction form,information was independently collected bytwo teams of investigators. Discrepancies wereresolved for each team by a dedicated thirdparty through review of original articles andgroup discussions. Information collected fromeach study included the study period, overallsample size and the number of patients in pre-operative and extended postoperative AMPtreatment arms, percentage of male patients,descriptive statistics for cohort age, patientcomorbidities, antibiotics administered andtheir perioperative timing and duration, ana-tomic level and type of spinal surgery, instru-mentation status, number of SSIs along withtheir incisional type and microbiology, compa-rable SSI prevalence estimates, and sources ofconfounding. The antibiotics recorded were thepredominant form of pharmacologic therapyadministered in each study. Antibiotics utilizedfor exceptional cases, such as patients withpenicillin or cephalosporin allergies, were notdocumented.

Information regarding quality was alsoextracted and summarized for individual studiesand for the combined study population as awhole. Included RCTs were determined to have

Adv Ther (2020) 37:2710–2733 2713

a high, unclear, or low risk of bias through useof the Cochrane Collaboration’s tool forassessing risk of bias. Meanwhile, the quality ofincluded cohort studies was assessed using theNewcastle–Ottawa Scale (NOS) in which theycould receive anywhere from 0 to 9 stars basedon their performance in certain measurableareas. The Grading of Recommendations,Assessment, Development, and Evaluations(GRADE) Working Group approach was thenused to determine the overall quality of evi-dence for the entire study population as awhole. The GRADE score was calculated foreach study using a composite measure based ontype of evidence, quality, consistency, direct-ness, and effect size [44, 45].

Statistical Analysis

Studies with multiple treatment arms for dif-ferent variations of either preoperative AMP orextended postoperative AMP were combined toform one group for each study. Basic percentagecalculations were made without p values andconfidence intervals (CI) during the systematicreview. Risk ratios and 95% CIs were calculatedfor each study and then combined using a ran-dom effects model to determine whetherextended postoperative AMP reduced SSIdevelopment when compared to preoperativeAMP protocols through a summary risk ratio. Acumulative meta-analysis was performed andvisualized through a forest plot to determine if asignificant difference existed at any point intime and, if appropriate, to see when conclu-sions of statistical significance could have firstbeen reached. Based on the availability of data,post hoc subgroup meta-analyses were con-ducted to determine whether the effect of pre-operative AMP and extended postoperativeAMP on the rate of SSI was conditional uponother variables through confounding or effectmodification. Stratified analysis was only con-ducted on patients undergoing instrumentedand non-instrumented spinal surgery. All anal-yses were performed using Stata 12.0 software(StataCorp LP, College Station, TX, USA). Alltests were two-sided with an alpha value of 0.05.

Heterogeneity, Bias, and QualityAssessment

The degree of heterogeneity among the riskratios was evaluated using the I2 statistic. Theprobability of publication bias and other suchstudy effects were visually assessed throughobservation of funnel plot asymmetry andEgger’s test of linear regression between theprecision of included studies and the standard-ized effect. Selection bias and information biaswere qualitatively evaluated.

The Cochrane Collaboration’s tool forassessing risk of bias, the NOS, and the GRADEWorking Group scoring system were utilized toassess quality of evidence for each includedstudy and for the overall study population. TheCochrane Collaboration’s tool for assessing riskof bias decides whether RCTs have a higher,unclear, or low risk of bias based on a study’sperformance in key domain areas [46]. The NOSawards up to 9 stars for quality features ofobservational studies [47, 48]. Meanwhile, theGRADE approach collectively evaluates qualityof evidence by awarding points in five differentcategories: type of evidence, quality of studymethodology, consistency among and withinstudies, directness or generalizability of thestudy population and outcome to the researchquestion, and effect size among all studies.Potential quality of evidence scores can rangefrom 7 to - 4 and are categorized as high (C 4),moderate (3), low (2), and very low (B 1).Meanwhile, the GRADE approach appraisesstrength of recommendation by collectivelyassessing the quality of evidence (GRADE score),desirability of effects, patient values and pref-erences, and resourcefulness of intervention.GRADE strength of recommendation scores aredetermined as either strong or weak. The nom-inal group technique will be utilized within ourgroup dynamic to reach conclusions of thesesubjective decisions for quality and risk of biasassessment. Finally, we used the nominal grouptechnique [49, 50] to achieve consensusregarding judgments made for risk of bias andquality assessment.

2714 Adv Ther (2020) 37:2710–2733

Compliance with Ethics Guidelines

This article is based on previously conductedstudies and does not contain any studies withhuman participants or animals performed byany of the authors.

RESULTS

Systematic Review

The primary search identified 388 potentialstudies. After removing 61 duplicate reports andscreening the titles and abstracts of the 327studies that remained, 21 full-text articles werereviewed in detail (Fig. 1). Five studies met

inclusion criteria: one RCT [27] and four retro-spective cohort studies [28, 33, 51, 52].

Pertinent qualitative and quantitative detailsfrom each of these five studies are presented andsummarized. Included studies followed CDCguidelines for SSI diagnosis to varying degrees.Each study identified SSI through the clinicalobservation of the surgeon or attending physi-cian, while some studies also utilized positivewound cultures and other diagnostic measuresto supplement diagnosis. The most notable in-consistency across included studies in meetingCDC criteria was the time period in which SSIdiagnosis was made. The observational periodranged from within 30 days [28] to more than6 months [33] after the operation. Hellbuschet al. [27] was the only study that did not elu-cidate their observational window for diagnos-ing SSI. The one included RCT was found tohave a high risk of bias [27] after evaluationwith the Cochrane Collaboration’s tool forassessing risk of bias [46]. Meanwhile, the NOS[47] quality scores of the four remaining cohortstudies ranged from 6 stars for two studies[33, 52] to 7 stars for two other studies [28, 51].

Hellbusch et al. [27] conducted an RCTassessing the rate of SSI among 269 consecutive,uninfected patients undergoing clean, instru-mented lumbar fusion (97 posterior long seg-ment fusions [PLSFs], 94 transforaminal lumbarinterbody fusion and posterior long segmentfusions [TLIF/PLSFs], 21 posterior lumbar inter-body fusion and posterior long segment fusions[PLIF/PLSFs], 16 minimally invasive sextantfusions, and 5 anterior lumbar interbodyfusions [ALIFs]) for degenerative disease, with orwithout spinal stenosis, from 2000 to 2003.Potential study participants with a cephalos-porin or severe penicillin allergy were excluded.Included patients were randomized to one oftwo AMP protocols: a single preoperative doseof cefazolin or an extended 10-day postopera-tive regimen of cefazolin and cephalexin. BothAMP protocols received intraoperative dosing.Only 233 patients (male [M] = 102, female[F] = 131; age range = 21–82 years) completedthe entire study: 6 patients withdrew secondaryto complications or adverse reactions and 30patients were eliminated from the study afterdeviation from their assigned AMP protocol.Fig. 1 Flow diagram of search strategy

Adv Ther (2020) 37:2710–2733 2715

From the 30 eliminated patients, 27 receivedadditional postoperative antibiotic doses, pri-marily for infectious reasons such as fever ofunknown origin and urinary tract infection,while 3 had missed antibiotic doses. Informa-tion regarding the length of the observationalwindow for SSI diagnosis was not provided, butall SSIs occurred within 21 days of surgery.

The overall rate of SSI was 3.0% (= 7/233).Two superficial SSIs were reported among the116 patients receiving extended postoperativeAMP, and 5 superficial SSIs were observedamong the 117 patients receiving preoperativeAMP. Difference testing failed to establish astatistically significant difference in the rate ofSSI between patients who received extendedpostoperative AMP and those who receivedpreoperative AMP (2/116 = 1.72% versus5/117 = 4.27%; p[ 0.25). Two one-sided test(TOST) equivalence testing also failed to pro-duce a statistically significant result. Multivari-ate analysis included a variety of patient andoperative characteristics thought to be associ-ated with SSI, but no statistically significantrelationships were reported. Hellbusch et al.hesitated to put forth a clinical recommenda-tion for or against the administration of exten-ded postoperative AMP as a result of insufficientstudy power. However, they indicated that itseemed unlikely that extended postoperativeAMP regimens lasting less than the 10 dayswould provide additional protection against SSIdevelopment given their null result. TheCochrane Collaboration’s risk of bias summaryassessment for the outcome of SSI from Hell-busch et al. was high. Meanwhile, Hellbuschet al.’s GRADE evidence quality score wasmoderate, while its strength of clinical recom-mendation was found to be strong.

Regarding their study’s low power, Hellbuschet al. noted that 1400 patients would berequired to detect a statistically significant dif-ference in SSI as a result of the small effect sizedifference in proportions (ES = 0.074) and thevery low overall rate of infection. As surgicalpractices improve and SSI becomes less com-mon, the low rate of occurrence undermines thefeasibility of conducting adequately poweredtrials. This is an artifact of surgical practice, andwe do not believe that this study by Hellbusch

et al. should be overly penalized for encoun-tering this difficulty. However, it is worth not-ing that eliminating patients who receivedadditional postoperative antibiotics may intro-duce bias in either direction. The distribution ofthese patients between treatment arms was notreported, making it difficult to account forpotential systematic differences between elimi-nated patients and adherent patients withoutindividual level data. The authors removedthese subjects rather than conducting an as-treated analysis or otherwise statisticallyaddressing this possible source of bias.

The designs of the other four included stud-ies were retrospective. Dobzyniak et al. [52],Kanayama et al. [51], and Kakimaru et al. [28]each took advantage of changes to AMP proto-cols within their respective institutions toevaluate the effects of different dosing regimenson the rate of SSI in spinal surgery. Theseinstitutional changes were motivated by thenull result of an in-house animal study assessingextended postoperative AMP on spinal infec-tious [29] for Dobzyniak et al. and CDC rec-ommendations [22] for Kanayama et al. andKakimaru et al. In each case, existing AMPprotocols were reduced from extended postop-erative AMP to preoperative AMP. These tem-poral variations in prophylaxis could introducebias, as other uncontrolled factors in the insti-tutional setting and patient population alsomay have changed over the study period.Nonetheless, the purpose of this paper is toformulate recommendations for institutions,and studies that review the effects of institu-tional protocol adjustments provide valuableinsight to the current review.

Dobzyniak et al. [52] conducted a retrospec-tive cohort study evaluating the rate of SSIamong 636 patients (M = 393, F = 243; meanage = 43 years) undergoing hemilaminotomyand discectomy for herniated lumbar disc from1993 to March 1999. Patients in the preopera-tive AMP group only received a single dose ofAMP. Information regarding the duration of theextended postoperative AMP regimen or whe-ther intraoperative dosing was administered foreither group was not provided. Cefazolin wasthe primary antibiotic administered in bothgroups. There were 26 patients who did not

2716 Adv Ther (2020) 37:2710–2733

receive preoperative AMP during this time per-iod that were excluded from the final statisticalanalysis of the remaining 610 patients. None ofthese eliminated patients developed SSI. Nopatients were lost to follow-up during the6-week period of postoperative observation.

The overall rate of SSI was 1.3% (= 8/610).Five SSIs were reported among the 418 patientsreceiving extended postoperative AMP, while 3SSIs were reported among the 192 patientsreceiving preoperative AMP. Informationregarding the incisional type of SSI was notprovided. The rate of SSI for the extendedpostoperative AMP group was not significantlydifferent from the preoperative AMP group (3/192 = 1.20% versus 5/418 = 1.56%; p = 0.71).Dobzyniak et al. concluded that a single pre-operative dose of AMP is equally effective in theprevention of SSI following lumbar discectomyas extended postoperative AMP. The NOSquality of evidence score for Dobzyniak et al.accumulated six stars.

Kanayama et al. [51] conducted a retrospec-tive cohort study of 1597 consecutive, unin-fected patients (male = 912, female = 685;mean age = 55.4 years) undergoing instru-mented and non-instrumented posteriordecompression lumbar surgeries for numerousdiagnoses (686 disc herniations, 340 degenera-tive spondylolistheses, 259 spinal stenoses, 73failed lumbar surgeries, 52 degenerative scol-ioses, 48 isthmic spondylolistheses, 34 spinaltraumas, 27 foraminal stenoses, 27 spinaltumors, and 51 miscellaneous cases) from Jan-uary 1999 to September 2004. Patients in thepreoperative AMP group received a single doseof antibiotics, while those in the extendedpostoperative AMP group received an extended5- to 7-day postoperative regimen. Both proto-cols provided intraoperative dosing and surgicalsite irrigation with first-generation cephalos-porins. Only SSIs occurring within 6 months ofsurgery were included and no subjects were lostduring the observational period.

The overall rate of SSI was 0.7% (= 11/1597).Two deep SSIs were observed among the 464patients receiving preoperative AMP, while 9SSIs (2 superficial and 7 deep) were reportedamong the 1133 patients receiving extendedpostoperative AMP. Among the 665 cases that

utilized instrumentation, there was 1 SSI amongthe 182 patients receiving preoperative AMPand 7 SSIs among the 483 patients receivingextended postoperative AMP. Meanwhile,among the 932 patients undergoing non-in-strumented, posterior decompression, there was1 SSI reported among the 282 patients receivingpreoperative AMP and 2 SSIs observed amongthe 650 patients receiving extended postopera-tive AMP. Information regarding SSI incisionaltype for stratified analyses based on instru-mentation was not provided. The rate of SSI didnot differ significantly between the extendedpostoperative AMP group and the preoperativeAMP group (9/1113 = 0.8% versus2/464 = 0.4%, p[0.05). Only deep SSIs wereobserved in the preoperative AMP group (2/464 = 0.4%); the rate of deep and superficial SSIwas 0.62% (= 7/1133) and 0.18% (= 2/1113),respectively, in the extended postoperative AMPgroup. Among cases utilizing instrumentation,the difference in rate of SSI between the twostudy arms was again insignificant (7/483 = 1.4% versus 1/182 = 0.5%, p[0.05).Kanayama et al. concluded that a single preop-erative dose of AMP, with intraoperative dosing,is effective in the prevention of SSI during bothinstrumented and non-instrumented lumbarspine surgeries. The NOS quality of evidencescore for Kanayama et al. had 7 stars.

Kakimaru et al. [28] conducted a retrospec-tive cohort study of 284 patients (M = 183,F = 101; mean age = 63 years, agerange = 16–84 years) undergoing microscopic,non-instrumented spinal decompression sur-gery (115 lumbar fenestrations, 88 cervicallaminoplasties, 60 lumbar discectomies, and 21other procedures) from October 2003 to August2009. All patients with infectious spondylitiswere excluded. There were three differenttreatment arms in this study: one preoperativeAMP group with 143 patients and two tempo-rally variable extended postoperative AMPgroups with a collective total of 141 patients.The patients in the preoperative AMP armreceived a single preoperative dose of antibioticsand intraoperative dosing. Meanwhile, patientsin the first extended postoperative AMP groupreceived a single preoperative and postoperativedose on the operative day, with further AMP

Adv Ther (2020) 37:2710–2733 2717

dependent upon the attending surgeon’s dis-cretion (average duration = 2.7 postoperativedays), but they did not receive intraoperativedosing. In contrast, patients in the secondextended postoperative AMP group received asingle preoperative dose, intraoperative dosing,and a single postoperative dose after skin clo-sure on the operative day, but did not receiveadditional postoperative AMP. These twoextended postoperative AMP protocols werecombined by the study’s investigators to permitstatistical analysis between the 143 patientsreceiving preoperative AMP (M = 94, F = 49,mean age = 64 years) and the 141 patientsreceiving extended postoperative AMP (M = 89,F = 52, mean age = 62 years). Cefazolin was theprimary antibiotic administered in both groups.There was no mention of patients being lost tofollow-up during the 30-day period of postop-erative observation.

The overall rate of SSI was 2.1% (= 6/284).There were 4 SSIs (3 superficial and 1 deep) inthe extended postoperative group and 2 super-ficial SSIs in the preoperative AMP group. Therewas no significant difference in rate of SSIbetween patients receiving extended postoper-ative AMP and those receiving preoperativeAMP (4/141 = 2.8% versus 2/143 = 1.4%,p = 0.335). The rate of superficial and deep SSIwas 2.13% (= 3/141) and 0.71% (= 1/141),respectively, in the extended postoperative AMPgroup; only superficial SSIs were observed in thepreoperative AMP group (2/143 = 1.4%). Kaki-maru et al. concluded that the SSI rate amongtheir cohort of patients while undergoingmicroscopic non-instrumented, spinal decom-pression surgery was not related to AMP dura-tion. The NOS quality of evidence score forKakimaru et al. had 7 stars.

Khan et al. [33] conducted a retrospectivecohort study of 100 patients (M = 40, F = 60;age range = 18–70 years) undergoing clean,elective, and non-instrumented lumbardecompression surgery (66 fenestrations, 28laminectomies, and 6 combined operations) fordisc prolapse, spinal stenosis, or both diagnosesfrom January 2006 to March 2008. Patients withfractures, trauma, and Pott’s disease wereexcluded. Three different perioperative AMPregimens were evaluated in separate treatment

arms: one preoperative AMP group and twotemporally variable AMP groups. The 21patients in the preoperative AMP group onlyreceived a single preoperative dose. The firstextended postoperative AMP group had 59patients and received a single preoperative dose,intraoperative dosing, and two postoperativedoses. Meanwhile, the second extended post-operative AMP group had 20 patients andreceived a single preoperative dose and at leastthree postoperative doses until the patientdemonstrated improvement. Informationregarding the overall duration of proceduresand the specific duration for each of theextended AMP protocols was not provided.First- or third-generation cephalosporins werethe primary antibiotics administered. There wasno information provided regarding patientsbeing lost to follow-up during the observationalperiod for SSI diagnosis, which lasted for morethan 6 months after surgery.

The overall rate of SSI was 6% (= 6/100).There were no SSIs in the preoperative AMPgroup, 1 superficial SSI in the first extendedpostoperative AMP group with intraoperativedosing and only two postoperative doses, and 5superficial SSIs in the second extended postop-erative AMP group with at least three postop-erative doses. There were no significantdifferences in rate of SSI when comparing thepreoperative AMP group to either the extendedpostoperative AMP group receiving exactly twopostoperative doses or to the extended postop-erative AMP group receiving at least threepostoperative doses (0/21 = 0% versus1/59 = 1.7%, p = 1.00; 0/21 = 0% versus5/20 = 25%, p = 0.54). Meanwhile, the differ-ence in the rate of SSI between the two exten-ded postoperative AMP groups was found to bestatistically significant (1/59 = 1.7% versus5/20 = 25%, p = 0.003). However, the reportedSSI rate of 25% among patients in the extendedpostoperative AMP groups receiving three ormore postoperative doses is considerably higherthan other published rates of SSI following non-instrumented spinal surgery [13, 28, 52]. Thisstudy is retrospective, and patients were notrandomized. Therefore, patients may havereceived more than three postoperative doses ifthey were perceived by their attending surgeon

2718 Adv Ther (2020) 37:2710–2733

to be at a greater risk of developing SSI. Theyalso may have been more likely to undergooperations that carry a higher risk of complica-tion, such as multilevel laminectomy, as thedistribution of procedures among patientgroups was not specified. Khan et al. also notedthat patients in this extended postoperativeAMP group receiving at least three postopera-tive doses were continued on AMP until theyimproved postoperatively. However, they didnot further specify whether improvementreferred to mobility and function or whetherthis group of patients experienced postoperativecomplications. Since there were no deep SSIs,the rate of superficial SSI was equal to theoverall SSI rate of 6%. Khan et al. concludedthat a single preoperative dose of AMP wasequally effective as extended postoperative AMPin preventing SSI among clean, elective lumbarsurgeries. The NOS quality of evidence score forKhan et al. was 6 stars.

Combined Study Characteristics

Data on patient demographics and SSI devel-opment from the five included studies was col-lected (Table 1). Male patients comprised 55%of the 2824 patients making up the overallcohort. All included studies utilized samplepopulations that were either completely orpredominantly composed of adult patients.Average patient age was reported by Dobzyniaket al., Kakimaru et al., and Kanayama et al. andranged from 43 [52] to 63 years [28]. Mean-while, age ranges were provided by Kakimaruet al., Hellbusch et al., and Khan et al. andextended from 16 to 84 years [28]. Studies wereinconsistent in their reporting and analysis ofmedical comorbidities, including tobacco use,chronic steroid use, and body mass index (BMI)as a measure of obesity.

There were 2717 (= 96.2%) patients whounderwent lumbar spinal surgery, 109 (3.8%)patients who underwent thoracic spinal

Table 1 Characteristics included comparative studies evaluating the effects of preoperative versus extended postoperativeAMP on SSI development during spinal surgery on adult patients

Study Year Patients(no.)

Preoperativedosing (no.)

TotalSSIs(no.)

SCSSIs(no.)

DeepSSIs(no.)

Meanage(years)

Percent malepatients, (M/F)

Level of spinalsurgery (no.)

Extendedpostoperativedosing (no.)

C TX LS

Khan et al.

[33]

2009 100 21 0 0 0 N/A 40%, (40:60) 0 0 21

79 6 6 0 0 0 79

Hellbusch

et al. [27]

2008 233 117 5 0 5 N/A 44%,

(102:131)

0 0 117

116 2 0 2 0 0 116

Kakimaru

et al. [28]

2010 284 143 2 2 0 62 64%,

(183:101)

42 1 100

141 4 3 1 60 4 77

Dobzyniak

et al. [52]

2003 610 192 3 N/A N/A 43 64%,

(393:243)

0 0 192

418 5 N/A N/A 0 0 418

Kanayama

et al. [51]

2007 1597 464 2 0 2 55 57%,

(912:685)

0 0 464

1133 9 2 7 0 0 1133

AMP antimicrobial prophylaxis, SSI surgical site infection, SC superficial, C cervical, TX thoracic, LS lumbosacral, no.number, M/F ratio of male to female patients, N/A data not provided

Adv Ther (2020) 37:2710–2733 2719

surgery, and 5 (\0.1%) patients who under-went cervical spinal procedures. The distribu-tion of the two AMP protocols across spinalsurgery level was not provided. Overall, therewere 26 SSIs observed in 1887 patients whoreceive extended postoperative AMP (cruderate = 1.4%), and there were 12 SSIs observed in937 patients who received preoperative AMP(crude rate = 1.3%). Studies were inconsistentin their reporting and analysis of operativevariables, such as procedural duration, thespecific procedure performed, and complica-tions, but meaningful stratified analysis wasconducted on instrumentation use and a crudeanalysis was performed for SSI incisional type.

Hellbusch et al. and Kanayama et al. were theonly studies that included patients undergoinginstrumented spinal fusion, but only the formerdetailed the specific fusion procedures per-formed. The mean number of spinal levels fusedwas not reported in either study. After stratify-ing the overall cohort by instrumentation, therewere 15 SSIs among 888 patients in the instru-mented group and 17 SSIs among 1926 patientsin the non-instrumented group. The rate of SSIin patients undergoing instrumented spinalsurgery was nearly twice that observed in non-instrumented spinal surgery (15/888 = 1.69%versus 17/1926 = 0.9%). Among patientsundergoing instrumented spinal surgery, therewere 9 SSIs among 599 patients receivingextended postoperative AMP and 6 SSIs among299 patients receiving preoperative AMP (1.5%versus 2%, respectively). In contrast, amongpatients undergoing non-instrumented spinalsurgery, there were 17 SSIs among 791 patientsreceiving extended postoperative AMP and 6SSIs among 638 patients receiving preoperativeAMP (1.3% versus 0.94%, respectively).

Rates of superficial and deep incisional SSIswere differentiated in each study except forDobzyniak et al. After stratifying the fourremaining studies by SSI incisional type, therewere 20 superficial and 10 deep SSIs among2214 patients. Among 1469 patients receivingextended postoperative AMP, there were 13superficial and 8 deep SSIs (0.89% and 0.54%,respectively). Meanwhile, among 745 patientsreceiving preoperative AMP, there were 7superficial and 2 deep SSIs (0.94% and 0.27%,

respectively). There was no discernable differ-ence in rate of SSI when stratified by the inci-sional type secondary to crude analysis.

Although each included study reported onthe microbiology of positive cultures to varyingdegrees, inconsistencies in reaching diagnosisthrough wound culture and the reporting ofthese findings, both across and within studies,prevented us from ascertaining the true per-centage of patients from the overall cohort withpositive wound cultures and their individualmicrobiology. We were also unable to reliablydetermine what causative organisms were mostcommonly encountered among the two differ-ent AMP protocols and whether any associa-tions existed between instrumentation use orthe incisional type of SSI with specific types ofmicroorganisms. In particular, Dobzyniak et al.[52] did not specify the absolute number SSIsattributable to identified causative bacteria, butnoted that the majority of positive culturesgrew Staphylococcus species. Since most patientseither had negative cultures or were diagnosedsolely through clinical observation, only 20individual patients from the overall cohort werereported to have positive wound cultures[27, 28, 33, 51]. Causative organisms includedStaphylococcus aureus (n = 10) [28, 33, 51],coagulase-negative Staphylococcus species (n = 5)[27, 28, 51], methicillin-resistant S. aureus(MRSA) (n = 3) [28, 51], Enterococcus faecalis(n = 1) [51], and gram-positive bacillus (n = 1)[51].

There were also inconsistencies in thereporting of AMR strains and bacteria-inducedpostoperative complications. Kanayama et al.,Kakimaru et al., and Khan et al. commented onantibiotic resistance patterns of culturedorganisms. Among the 19 patients with positivecultures in these three studies, resistant organ-isms were cultured in 6 of 15 patients receivingextended postoperative AMP (40%), and in 0 of4 patients receiving preoperative AMP[28, 33, 51]. In addition, two cases of pseu-domembranous colitis secondary to C. difficilewere noted in the extended postoperative AMPgroup while no cases were noted in the preop-erative AMP group [28].

The overall combined GRADE quality ofevidence score was low. Additionally, the

2720 Adv Ther (2020) 37:2710–2733

included studies had an average GRADE qualityof evidence score of low when this evaluationcriteria was applied to them individually.

Meta-Analysis

The overall risk of SSI development in patientswho received extended postoperative AMP was1.11 times the risk of SSI development inpatients who only received preoperative AMP,though this effect was not significant (RR =1.11, 95% CI 0.53–2.36, p = 0.78) (Fig. 2a).There was no evidence of heterogeneity oftreatment effects among the included studies(I2 = 0%; p = 0.50), and cumulative meta-anal-ysis failed to reveal a significant risk ratio at anytime.

Publication bias was assessed using funnelplots (Fig. 2b), and the study estimates appearsymmetric about the estimated true effect size.However, the small number of studies limits theinterpretation of a funnel plot. Egger’s test forsmall-study effects was also found to be statis-tically insignificant (p = 0.81). Taken together,these results support the hypothesis that the SSIrate after spine surgery is not different betweenextended postoperative AMP and preoperativeAMP.

Stratified meta-analysis was performed bywhether instrumentation was implanted duringsurgery. Among patients who underwentinstrumented spinal surgery, no significant dif-ference in risk of SSI development was observedbetween patients who received extended post-operative AMP and those who received preop-erative AMP (RR = 0.92, 95% CI 0.15–5.75,p = 0.93) (Fig. 3a). No notable benefit or harmwas observed from extended postoperative AMPin terms of occurrence of SSI. Only two studies,Hellbusch et al. and Kanayama et al., con-tributed to the instrumented cohort in thisstratified meta-analysis; the results of neitherstudy were statistically significant indepen-dently, and their estimated effects actuallytrended in opposite directions (RR = 0.40 andRR = 2.64, respectively). Therefore, someheterogeneity could be expected and is indi-cated by an I2 of 49.1%, but this was found notto be statistically significant (p = 0.16). The

interpretation of the funnel plot is extremelylimited because of a very small sample size(Fig. 3b).

Four studies reported data on outcomes fol-lowing non-instrumented spinal surgery. Thepooled risk of SSI following non-instrumentedspinal surgery in patients who received exten-ded postoperative AMP was not significantlydifferent from the risk of SSI in patients whoreceived preoperative AMP (RR = 1.25, 95% CI0.49–3.17, p = 0.65) (Fig. 4a). There was no evi-dence of heterogeneity of treatment effectsamong the included studies (I2 = 0%; p = 0.70).Again, the study estimates appear symmetri-cally dispersed about the estimated true effectsize, suggesting absence of publication bias,although any conclusions drawn are limited bysmall study numbers (Fig. 4b).

DISCUSSION

Guidelines over the optimal timing of periop-erative AMP in spinal surgery have been subjectto clinical debate for over half a century. Priorto publication of Burke’s [53] landmark bovinestudy in 1961, the prophylactic efficacy ofperioperative antibiotic utilization was poorsecondary to administration that was com-monly confined to the postoperative window[54, 55]. Succeeding publications, includingClassen et al.’s [30] seminal 1992 observationalstudy, substantiated Burke’s findings over thenext three decades [29, 56–63] and highlightedthe importance of administering preoperativeAMP just prior to first incision. Horwitz andCurtin [6] are often credited with first demon-strating the efficacy of initiating AMP preoper-atively in spinal surgery in 1975 and,ultimately, these authors recommended exten-ded prophylaxis. Despite this evidence, RCTs[64–69] had so far failed to demonstrate clinicalefficacy of perioperative AMP in preventingspinal SSI. Barker [19] performed a meta-analy-sis of these RCTs in 2002 and found results thatrobustly reinforced administration of AMP withat least a single preoperative dose during spinalsurgery, even when anticipated rates of SSIwithout AMP are low.

Adv Ther (2020) 37:2710–2733 2721

Fig. 2 a Forest plot and pooled relative risk of SSIdevelopment in patients who received both preoperativeand extended postoperative AMP compared to a patientswho only received preoperative AMP. b Funnel plot of allstudies with non-zero rates of SSI demonstrating a

symmetric distribution of study estimates around theestimated true effect size, suggesting no evidence forpublication bias. RR relative risk, CI confidence interval,SSI surgical site infection, AMP antimicrobial prophylaxis

2722 Adv Ther (2020) 37:2710–2733

Fig. 3 a Forest plot and pooled relative risk of SSIdevelopment in patients undergoing instrumented spinalsurgery who received both preoperative and extendedpostoperative AMP compared to patients who onlyreceived preoperative AMP. b Funnel plot of all studies

with non-zero rates of SSI demonstrating a symmetricdistribution of study estimates around the estimated trueeffect size. The small sample size limits the interpretationof this plot. RR relative risk, CI confidence interval, SSIsurgical site infection, AMP antimicrobial prophylaxis

Adv Ther (2020) 37:2710–2733 2723

Fig. 4 a Forest plot and pooled relative risk of SSIdevelopment in patients undergoing non-instrumentedspinal surgery who received both preoperative andextended postoperative AMP compared to patients whoonly received preoperative AMP. b Funnel plot of allstudies with non-zero rates of SSI demonstrating a

symmetric distribution of study estimates around theestimated true effect size. The small sample size limits theinterpretation of this plot. RR relative risk, CI confidenceinterval, SSI surgical site infection, AMP antimicrobialprophylaxis

2724 Adv Ther (2020) 37:2710–2733

Although research has heavily focused onwhen to start AMP in spinal surgery, questionsover its overall duration remain. In 2003,Dobzyniak et al. [52] and Beiner et al. [70] werethe first to call attention to the lack of consis-tent recommendations over how long, if at all,to continue AMP postoperatively. Clinicalstudies have since not only assessed the efficacyof extended postoperative therapy to preopera-tive AMP alone [27, 28, 33, 51, 52] but have alsocompared the therapeutic benefit of extendedAMP regimens to others that differ in postop-erative duration [32]. The NASS [24, 25] hasmade recommendations for optimal antibioticduration for both uncomplicated and compli-cated spinal surgeries after qualitativelyreviewing these studies in 2013, but a meta-analysis of comparative clinical trials has stillnot yet been conducted.

The main meta-analysis performed in thisstudy combined the results of one RCT and fourretrospective cohort studies. We did not findevidence that the administration of an exten-ded AMP regimen was favorable over theadministration of a single preoperative dose asindicated by an overall risk ratio of 1.11 and95% confidence interval of 0.53–2.36 (p = 0.78).Heterogeneity analysis did not reveal variabilityamong the studies and did not provide anyevidence that the studies could not be com-bined. Funnel plots did not reveal any clearpublication bias, and Egger’s test for small-studyeffects was found to be non-significant. Thesefindings support the hypothesis that combinedpreoperative and extended postoperativeantimicrobial prophylaxis does not reduce SSIrates in adult patients undergoing spinal sur-gery as compared to a single preoperative dose.

The broad confidence interval of the riskratio and the insignificant test statistic implythat the meta-analysis is likely underpowered todetect small differences in the risk of SSI afterspinal surgery between these two AMP regi-mens. However, the above finding is consistentwith previously published studies. As men-tioned previously, Barker performed a meta-analysis of six RCTs in order to evaluate the roleof AMP in spinal surgery. A subgroup analysis ofthese studies failed to demonstrate a greaterbenefit of prophylaxis when postoperative doses

were added to a single preoperative dose [19].Furthermore, a large meta-analysis of 28 RCTscomparing preoperative and extended AMP inthe setting of major surgery also failed todemonstrate superiority of extended AMP overpreoperative only regimens [20]. Although thefindings of this meta-analysis and the previousliterature cannot definitively answer the ques-tion of whether extended postoperative AMPprovides additional benefit over preoperativeonly AMP during spinal surgery, it contributessignificantly to the existing state of knowledgeby providing more evidence that preoperativeAMP alone may be sufficient to decrease the riskof SSI.

This is a valuable observation related to thehigh costs of SSI, the association of antibioticuse with increasing antimicrobial resistance,and related morbidity such as C. difficile colitis.Attributable costs per spinal SSI range fromUS$15,000 to upwards of US$100,000[7, 11, 71–78]. Notably, the costs incurred fromSSI after spinal surgery are greater than thereported additional expenditures required forthe average surgical patient who experiences SSI[79, 80].

In terms of SSI and the potential for AMR,most infections reported in the included studieswere, unsurprisingly, due to organisms relatedto skin flora, and included infections withS. aureus and coagulase-negative Staphylococcusspecies. Infections with E. faecalis and gram-positive bacillus were also observed. Interest-ingly, among those studies that commented onantimicrobial patterns, resistant organisms,such as MRSA, were cultured in six of fifteenpatients treated with extended AMP and in zeroof four patients who only received preoperativeAMP. Clearly, the numbers of patients in thesetwo groups are low, and, thus, meaningful sta-tistical inference is impossible. Furthermore, thepatients were not randomized and it is possiblethat patients with increased risk for developinginfection from resistant organisms may havealso been more likely to receive extended AMP.However, despite these limitations, this obser-vation is concerning given emerging antibioticresistance patterns. The potential to reduceAMR infections by minimizing the duration of

Adv Ther (2020) 37:2710–2733 2725

perioperative AMP should continue to encour-age additional, high-quality research.

It is incumbent upon all surgeons to endea-vor to prevent SSI and AMR emergence byexercising conservative caution when usingperioperative AMP. To date, surgeon adherenceto recommended guidelines often deviates sig-nificantly in practice, especially with regard tohow long AMP is extended into the postopera-tive window [81–83]. This variation in prophy-lactic care is secondary to numerous factors,including instrumentation use, surgical com-plexity, and patient comorbidity status[9, 16, 52], but can also result from institutionalguidelines or surgeon discretion [28, 52]. Thislack of homogeneity in AMP in spinal surgery isunsurprising given the historically contradict-ing recommendations made regarding exten-sion of AMP into the postoperative window[6, 16, 29, 52, 53, 67], and highlights the needfor this systematic review and meta-analysis. Itis important to weigh the real risk of antimi-crobial resistance against a perceived, yet sci-entifically unsupported, benefit of continuingantibiotics postoperatively. These considera-tions aided the authors in developing the rec-ommendation that AMP should not becontinued postoperatively in uncomplicated,adult patients during spinal surgery.

However, it is difficult to generalize theresults of this meta-analysis to all spinal surgerypatients because of the complex interactionbetween patient-related risk factors, surgicalcomplexity, and other unmeasured factors,particularly in complicated cases. In order tohelp address this uncertainty, we performedstratified meta-analyses. Most of the studies didnot report the rates of SSI among various relatedmedical comorbidities; therefore, the types ofstratified analyses that could be carried out werelimited. However, stratification was possibleaccording to whether instrumentation wasimplanted. Patients undergoing instrumentedspinal fusion are subjected to a higher risk ofpostoperative infection due to longer operativetime, increased procedure complexity, and theimplantation of a foreign body [15, 16, 61]. Asexpected, the rate of SSI in patients followinginstrumented spinal surgery was nearly twicethat observed in patients following non-

instrumented spinal surgery. However, we didnot find a significant benefit of extended AMPover preoperative only AMP in patients eitherundergoing instrumented spinal surgery or inthose undergoing non-instrumented spinalsurgery.

Overall, the strengths of this study make itsfindings noteworthy for spine surgeons. First,the inclusion criteria concerning the studydesign were strict. In particular, each study wasrequired to have a near equivalent comparisongroup which allowed for more accurate effectestimates between the two AMP protocols. Thisis highlighted in Kanayama et al.’s [51] cohortstudy in which the investigators identified apotential self-imposed limitation to their study:the definition of SSI required surgical treatmentof the wound. Since the majority of SSIs weresuperficial and managed with conservativetherapy, the reported rate of SSI for both treat-ment arms may be underestimated. However,since the same definition was utilized for bothtreatment arms, the relative rate of SSI betweenthe two groups may actually be accuratelyreflected, despite the conservative nature of thedefinition itself. But, this may not necessarily bethe case. Preoperative AMP may have a differenteffect on SSI requiring surgical interventionwhen compared to extended postoperativeAMP, and this presents an opportunity for biasin this meta-analysis. Also, since Kanayamaet al. provided data for the primary analysis andboth subgroup meta-analyses, this confoundingcould extend throughout the study.

The internal validity of this review was highsecondary to assessment with funnel plots andEgger’s test, which both failed to identify pub-lication bias. Since this meta-analysis found anon-significant result, the publication of furthernon-significant studies on this topic would onlylend greater support to our current findings,making publication bias of little concern to theauthors. Additionally, there was no hetero-geneity present among the five studies. Takentogether, these results support the internalvalidity of the results reached in this meta-analysis. With that said, however, the externalvalidity of this study may be called into ques-tion. It is important to note that weaknesses ofthis meta-analysis and its methodology are

2726 Adv Ther (2020) 37:2710–2733

mutually exclusive to the limitations of theincluded studies. One weakness of the presentstudy is the generalizability of results. Thisstudy aimed to report on the efficacy of tem-porally variable perioperative AMP in spinalsurgery at any level. However, 96.2% (2717/2824) of patients underwent spinal operationsconfined to the lumbar spine. Stratification ofthe 107 patients undergoing either cervical orthoracic operations was not possible because ofthe small sample size. Although unlikely, it ispossible that cervical, thoracic, or coccygealprocedures experience different SSI rates underdifferent perioperative AMP protocols. Addi-tionally, there was not enough data provided toconduct further subgroup analyses on patientcomorbidity status, which may impact SSI out-comes concerning AMP duration. We were ableto stratify based on instrumentation use, butthese subgroup analyses were hindered by thesame limitations detracting from the primarymeta-analysis. There were also differencesbetween studies in the length of AMP prescribedbased on individual surgeon preference. Thus,although there was no evidence supporting theextensions of AMP postoperatively for eitherinstrumented or non-instrumented patientsubgroups, decisions regarding a patient’s risk ofSSI and whether antibiotic prophylaxis is nee-ded following surgery must be made on anindividual basis until more data becomesavailable.

There were several limitations of the studiesincluded in this review and meta-analysis thatshould be considered separately from the meta-analysis itself. The GRADE quality of evidencescore for the overall included data ranged fromlow [48] to moderate [46] quality, which wasprimarily related to their observational andoften retrospective nature [10, 19–21], as theauthors were unable to control for unmeasuredconfounding and bias. Since the RCT [27] wasnot blinded, investigators were aware of theAMP protocol administered to each patient andbias could arise from increased observation andsurgical performance. The majority of includedstudies were underpowered with relatively smallsample sizes, increasing the likelihood of type IIerror and limiting the ability to detect signifi-cant differences in SSI rates between the

different AMP protocols. Most studies failed toinclude explicit and objective definitions of SSI,including those for superficial and deep SSIs;the criteria in which SSI was diagnosed; or whowas making these decisions. One study followedguidelines set forth by the CDC [28]. Anotherstudy required positive wound cultures and/orthe attending surgeon’s clinical impression oftypical infectious signs that led to additionalsurgical intervention [51]. Meanwhile, othersdid not mention their criteria for defining anddiagnosing SSI [33]. Although some studiesstated what individuals were making theseclinical decisions, possible sources of bias,especially for the prospectively collected data-bases and RCT, could arise. In particular, if theprospective studies failed to have a third-partyrecord their data, observer bias could occur ifthe investigators knew which patients receivedwhich AMP protocol. At the same time, theretrospective studies could be biased since theauthors had no control over data collection.Similarly, the follow-up observational period forSSI identification demonstrated considerablevariation among studies and ranged fromwithin 30 days [47] to 6 months [50] and longer[48] to unspecified follow-up times [50]. Thisreview did not exclude studies based on antibi-otic type used or changes in antibiotic admin-istered during treatment. The timing of initialpreoperative AMP administration and thelength of postoperative AMP duration in theextended postoperative groups also demon-strated significant temporal heterogeneity bothwithin and between studies, ranging fromwithin 30 min prior to skin incision to 10 dayspostoperatively [27, 28, 51]. The timing ofperioperative AMP was either unknown or notreported by Dobzyniak et al. and Khan et al.[33, 52]. Hellbusch et al. [27] statistically asses-sed the impact of demographic factors, comor-bidities and past surgical history, and operativevariables on the outcome of SSI development,finding no associations of statistical signifi-cance. No multivariate analyses were conductedin any of the included studies. The considerablevariability in defining, diagnosing, and report-ing SSI; the inconsistency in the postoperativeobservational period; and the inconsistency inpreoperative and extended postoperative

Adv Ther (2020) 37:2710–2733 2727

extended AMP somewhat diminish the validityof this meta-analysis.

These strengths and weaknesses were suc-cinctly enumerated and tallied by the GRADEWorking Group scoring system in the resultssection. As a result of the relative novelty of theGRADE approach and the very low to moderatequality of evidence scores awarded to theincluded studies, the GRADE scoring systemdeserves further mention here. The GRADEapproach is currently regarded as one of themost effective and unequivocal scoring meth-ods to systematically connect evidence-qualityevaluations to degree of efficacy for clinicalrecommendations [44]. This underlies one ofthe primary reasons why the GRADE approachis valuable in improving clinical practice: itsscoring system transparently separates decisionsregarding evidence quality from decisions eval-uating recommendation strength [45]. In otherwords, studies with high quality evidence donot always indicate strong clinical recommen-dations, and strong clinical recommendationscan sometimes be drawn from studies with lowquality evidence [45]. Therefore, despite theaverage low quality of evidence scoring assignedto the included studies through the GRADEsystem in this review, the statistically null resultreported in the present analysis and its associ-ated clinical recommendation to provide spinalsurgery patients with preoperative AMP onlyshould give pause to spinal surgeons,nonetheless.

The authors also believe that certain qualitiesof the GRADE scoring system led to artificiallylow scores among the included articles. Firstly,the GRADE scoring approach is partial towardsstudies that demonstrate statistical significance.All included studies failed to demonstrate sta-tistically significant results, and this con-tributed to their overall low quality of evidencescores through poor scoring in two of the fivecategories of the GRADE scoring system: effectsize and consistency. Effect size awardsincreasing points when every included studyreports large, unidirectional magnitudes ofeffect. Meanwhile, consistency assigns points tostudies that are characterized by dose responsepatterns, either between or within studies.Magnitude of effect and dose response are two

of Hill’s criteria for causality [84] and are desir-able if, in reality, there is a difference in effectbetween treatment arms. However, their inclu-sion in the GRADE scoring methodologyassumes that the true and desirable outcome isalways statistically significant. It is pertinent toremember that statistical significance and clin-ical significance are distinguishable, and statis-tically insignificant results can impact care inclinically meaningful ways. Given the clinicalsignificance of our null result and its potentialto reduce unnecessary medical expendituresand decrease risk of AMR development, it isappropriate here to highlight theseshortcomings.

An additional issue brought on by theGRADE approach that contributes to the overalllow quality of evidence scores is the explicitvalue placement on study design. Specifically,the GRADE scoring system allocates four pointsto RCTs and two points to observational studies,reflecting the old adage that RCTs are the ‘‘goldstandard’’ of research. However, the sample sizenecessary for an RCT to establish statistical sig-nificance is prohibitively large in SSI studies inspinal surgery. For example, the investigators[27] who conducted the only RCT included inthis meta-analysis hesitated to make clinicalrecommendations secondary to their smallsample size of 233 patients. Utilizing the low SSIrates experienced by patients in their study andthe small effect size difference in proportions(ES = 0.074), they performed a power calcula-tion to determine the sample size that wouldhave been required to obtain statistically sig-nificant results. They found that 1400 patients,a study population six times greater than theiroriginal sample size, would be required toestablish statistical significance in comparingSSI rates in patients receiving preoperative AMPto those receiving extended postoperative ther-apy at the desired power. Additionally, Dimicket al. [26] further justified the impracticality ofconducting adequately powered RCTs on SSIdevelopment and perioperative AMP in spinalsurgery and the notable risk of making a type IIerror when not enough patients are enrolled.They demonstrated that 474, 988, and 2518patients would be required to detect baseline SSIrates of 10%, 5%, and 2%, in each treatment

2728 Adv Ther (2020) 37:2710–2733

group, respectively, of an RCT with 80% powerto detect a 50% reduction in SSI through peri-operative AMP in spinal surgery. Therefore, itwas highly unlikely that RCTs would have beenavailable for this meta-analysis because of theexorbitant costs required to enroll enoughpatients to ensure adequate power. Thus, therelative ease of identifying patients for a largeobservational study may, in the case of investi-gating SSI rates in spinal surgery, make obser-vational data more meaningful than data fromsmall RCTs. However, this is not reflected in theGRADE approach, which broadly penalizesobservational studies when, in this clinicalapplication, they are actually the best option forfinding statistically significant results, especiallygiven the low incidence of SSI in spinal surgery[85]. This meta-analysis benefited from the rel-atively large sample sizes provided by themajority of the included observational studies.The GRADE scoring system aims to objectivelyevaluate a research process in which investiga-tors must invariably make subjective decisions.Researchers and practitioners must consider theneed for flexibility in the interpretation of suchevaluative methods in order to adapt to theunique challenges of conducting research inparticular medical and surgical specialties.

In light of these deficiencies, the authors ofthis review endorse the validity of the presentanalysis despite the low GRADE evidence qual-ity scores assigned to the included studies. Thisis evidenced by the strong strength of recom-mendation GRADE scores awarded to four ofthe five included studies, emphasizing theimportance of the research context and thediscrepancy between statistical and clinical sig-nificance. Therefore, the evidence utilized forthis review and analysis, although evaluated tobe of moderate to poor overall quality, stillprovides useful, evidence-based clinical recom-mendations that have the potential to impactcurrent practices as spinal surgeons still com-monly extend perioperative AMP based onpersonal discretion. Therefore, the recommen-dation set forth by this systematic review andmeta-analysis is strong according to the GRADEscoring system.

CONCLUSION

A meta-analysis of comparative clinical trialsinvestigating the development of SSI in 2824adult spinal surgery patients, receiving eitherpreoperative or extended postoperative AMP,revealed that SSI occurred in 1.4% of all cases.Analysis of the collective data demonstrated nostatistical difference between the temporallyvariable AMP protocols, suggesting that preop-erative AMP without extended postoperativetherapy is clinically equivalent to extendedpostoperative AMP. Subgroup analyses alsofailed to demonstrate statistical significance inSSI development between the different periop-erative AMP protocols between instrumentedand non-instrumented spinal surgery and in thedermal level of SSI origin. Postoperative AMPappears unnecessary for spinal surgery given itsincreased medical costs and risk for AMRdevelopment from gratuitous prophylaxis.However, higher-quality data are needed todetermine the theoretically optimal periopera-tive AMP regimen. As a result of the low inci-dence of SSIs, the high percentage of patientswho underwent lumbar region surgery, theinability to stratify results by patient comor-bidity, and heterogeneity between study proto-cols regarding AMP use, the results found inthese studies may not be generalizable to a lar-ger patient population.

Surgeons should use their clinical judgmentin prescribing the correct prophylaxis for theirpatients while keeping in mind the real medi-cal, financial, and societal costs of unnecessaryantibiotic use. Related to the low incidence ofSSI in spinal surgery, future studies of periop-erative AMP duration in this surgical subspe-cialty should either be adequately poweredRCTs or large prospective observational studies.

ACKNOWLEDGEMENTS

Funding. No funding or sponsorship wasreceived for this study or publication of thisarticle.

Adv Ther (2020) 37:2710–2733 2729

Authorship. All named authors meet theInternational Committee of Medical JournalEditors (ICMJE) criteria for authorship for thisarticle, take responsibility for the integrity ofthe work as a whole, and have given theirapproval for this version to be published.

Disclosures. Blaine T. Phillips, Emma S.Sheldon, Vwaire Orhurhu, Robert A. Ravinsky,Monika E. Freiser, Morteza Asgarzadeh, OmarViswanath, and Marie Roguski have nothing todisclose. Alan D. Kaye is a member of the jour-nal’s editorial board.

Compliance with Ethics Guidelines. Thisarticle is based on previously conducted studiesand does not contain any studies with humanparticipants or animals performed by any of theauthors.

Data Availability. The datasets generatedduring and/or analyzed during the currentstudy are available from the correspondingauthor on reasonable request.

Open Access. This article is licensed under aCreative Commons Attribution-NonCommer-cial 4.0 International License, which permitsany non-commercial use, sharing, adaptation,distribution and reproduction in any mediumor format, as long as you give appropriate creditto the original author(s) and the source, providea link to the Creative Commons licence, andindicate if changes were made. The images orother third party material in this article areincluded in the article’s Creative Commonslicence, unless indicated otherwise in a creditline to the material. If material is not includedin the article’s Creative Commons licence andyour intended use is not permitted by statutoryregulation or exceeds the permitted use, youwill need to obtain permission directly from thecopyright holder. To view a copy of this licence,visit http://creativecommons.org/licenses/by-nc/4.0/.

REFERENCES

1. Blam OG, Vaccaro AR, Vanichkachorn JS, et al. Riskfactors for surgical site infection in the patient withspinal injury. Spine (Phila Pa 1976). 2003;28(13):1475–80.

2. Browne JA, Cook C, Pietrobon R, Bethel MA,Richardson WJ. Diabetes and early postoperativeoutcomes following lumbar fusion. Spine (Phila Pa1976). 2007;32(20):2214–9.

3. Carreon LY, Puno RM, Lenke LG, et al. Non-neu-rologic complications following surgery for adoles-cent idiopathic scoliosis. J Bone Jt Surg Am Vol.2007;89(11):2427–32.

4. Culver DH, Horan TC, Gaynes RP, et al. Surgicalwound infection rates by wound class, operativeprocedure, and patient risk index. National Noso-comial Infections Surveillance System. Am J Med.1991;91(3b):152s–7s.

5. Fang A, Hu SS, Endres N, Bradford DS. Risk factorsfor infection after spinal surgery. Spine (Phila Pa1976). 2005;30(12):1460–5.

6. Horwitz NH, Curtin JA. Prophylactic antibiotics andwound infections following laminectomy for lum-ber disc herniation. J Neurosurg. 1975;43(6):727–31.

7. McGirt MJ, Parker SL, Lerner J, Engelhart L, KnightT, Wang MY. Comparative analysis of perioperativesurgical site infection after minimally invasive ver-sus open posterior/transforaminal lumbar inter-body fusion: analysis of hospital billing anddischarge data from 5170 patients. J NeurosurgSpine. 2011;14(6):771.

8. National Nosocomial Infections Surveillance Sys-tem. National Nosocomial Infections Surveillance(NNIS) System Report, data summary from January1992 through June 2004, issued October 2004. Am JInfect Control. 2004;32(8):470–85.

9. Olsen MA, Nepple JJ, Riew KD, et al. Risk factors forsurgical site infection following orthopaedic spinaloperations. J Bone Jt Surg Am Vol. 2008;90(1):62–9.

10. Omeis IA, Dhir M, Sciubba DM, et al. Postoperativesurgical site infections in patients undergoingspinal tumor surgery: incidence and risk factors.Spine (Phila Pa 1976). 2011;36(17):1410–9.

11. Parker SL, Adogwa O, Witham TF, Aaronson OS,Cheng J, McGirt MJ. Post-operative infection afterminimally invasive versus open transforaminallumbar interbody fusion (TLIF): literature reviewand cost analysis. Minim Invasive Neurosurg MIN.2011;54(1):33–7.

2730 Adv Ther (2020) 37:2710–2733

12. Pull ter Gunne AF, Cohen DB. Incidence, preva-lence, and analysis of risk factors for surgical siteinfection following adult spinal surgery. Spine(Phila Pa 1976). 2009;34(13):1422–8.

13. Smith JS, Shaffrey CI, Sansur CA, et al. Rates ofinfection after spine surgery based on 108,419procedures: a report from the Scoliosis ResearchSociety Morbidity and Mortality Committee. Spine(Phila Pa 1976). 2011;36(7):556–63.

14. Watanabe M, Sakai D, Matsuyama D, Yamamoto Y,Sato M, Mochida J. Risk factors for surgical siteinfection following spine surgery: efficacy of intra-operative saline irrigation. J Neurosurg Spine.2010;12(5):540–6.

15. Weinstein MA, McCabe JP, Cammisa FP Jr. Post-operative spinal wound infection: a review of 2,391consecutive index procedures. J Spinal Disord.2000;13(5):422–6.

16. Wimmer C, Gluch H, Franzreb M, Ogon M. Pre-disposing factors for infection in spine surgery: asurvey of 850 spinal procedures. J Spinal Disord.1998;11(2):124–8.

17. Abdul-Jabbar A, Takemoto S, Weber MH, et al.Surgical site infection in spinal surgery: descriptionof surgical and patient-based risk factors for post-operative infection using administrative claimsdata. Spine (Phila Pa 1976). 2012;37(15):1340–5.

18. Schnoring M, Brock M. Prophylactic antibiotics inlumbar disc surgery: analysis of 1,030 procedures.Zentralbl Neurochir. 2003;64(1):24–9.

19. Barker FG 2nd. Efficacy of prophylactic antibiotictherapy in spinal surgery: a meta-analysis. Neuro-surgery. 2002;51(2):391–400 (discussion -1).

20. McDonald M, Grabsch E, Marshall C, Forbes A.Single- versus multiple-dose antimicrobial prophy-laxis for major surgery: a systematic review. Aust NZ J Surg. 1998;68(6):388–96.

21. DiPiro JT, Cheung RP, Bowden TA Jr, MansbergerJA. Single dose systemic antibiotic prophylaxis ofsurgical wound infections. Am J Surg. 1986;152(5):552–9.

22. Mangram AJ, Horan TC, Pearson ML, Silver LC,Jarvis WR. Guideline for prevention of surgical siteinfection, 1999. Hospital Infection Control Prac-tices Advisory Committee. Infect Control HospEpidemiol. 1999;20(4):250–78 (quiz 79–80).

23. Watters WC 3rd, Baisden J, Bono CM, et al.Antibiotic prophylaxis in spine surgery: an evi-dence-based clinical guideline for the use of pro-phylactic antibiotics in spine surgery. Spine J.2009;9(2):142–6.

24. Shaffer WO, Baisden JL, Fernand R, Matz PG. Anevidence-based clinical guideline for antibioticprophylaxis in spine surgery. Spine J. 2013;13(10):1387–92.

25. Shaffer WO, Baisden J, Fernand R, Matz PG. Evi-dence-based clinical guidelines for multidisci-plinary spine care: antibiotic prophylaxis in spinesurgery. USA: North American Spine Society; 2013.

26. Dimick JB, Lipsett PA, Kostuik JP. Spine update:antimicrobial prophylaxis in spine surgery: basicprinciples and recent advances. Spine (Phila Pa1976). 2000;25(19):2544–8.

27. Hellbusch LC, Helzer-Julin M, Doran SE, et al. Sin-gle-dose vs multiple-dose antibiotic prophylaxis ininstrumented lumbar fusion—a prospective study.Surg Neurol. 2008;70(6):622–7 (discussion 7).

28. Kakimaru H, Kono M, Matsusaki M, Iwata A, UchioY. Postoperative antimicrobial prophylaxis follow-ing spinal decompression surgery: is it necessary?J Orthop Sci. 2010;15(3):305–9.

29. Guiboux JP, Cantor JB, Small SD, Zervos M,Herkowitz HN. The effect of prophylactic antibi-otics on iatrogenic intervertebral disc infections. Arabbit model. Spine (Phila Pa 1976). 1995;20(6):685–8.

30. Classen DC, Evans RS, Pestotnik SL, Horn SD,Menlove RL, Burke JP. The timing of prophylacticadministration of antibiotics and the risk of surgi-cal-wound infection. N Engl J Med. 1992;326(5):281–6.

31. Page CP, Bohnen JM, Fletcher JR, McManus AT,Solomkin JS, Wittmann DH. Antimicrobial pro-phylaxis for surgical wounds. Guidelines for clinicalcare. Arch Surg. 1993;128(1):79–88.

32. Takahashi H, Wada A, Iida Y, et al. Antimicrobialprophylaxis for spinal surgery. J Orthop Sci.2009;14(1):40–4.

33. Khan IU, Janjua MB, Hasan S, Shah S. Surgical siteinfection in lumbar surgeries, pre and postoperativeantibiotics and length of stay: a case study. J AyubMed Coll Abbottabad. 2009;21(3):135–8.

34. Steinberg JP, Braun BI, Hellinger WC, et al. Timingof antimicrobial prophylaxis and the risk of surgicalsite infections: results from the Trial to ReduceAntimicrobial Prophylaxis Errors. Ann Surg.2009;250(1):10–6.

35. Bratzler DW, Houck PM, Richards C, et al. Use ofantimicrobial prophylaxis for major surgery: base-line results from the National Surgical InfectionPrevention Project. Arch Surg. 2005;140(2):174–82.

Adv Ther (2020) 37:2710–2733 2731

36. Bratzler DW, Houck PM. Antimicrobial prophylaxisfor surgery: an advisory statement from theNational Surgical Infection Prevention Project. ClinInfect Dis. 2004;38(12):1706–15.

37. Berenguer CM, Ochsner MG Jr, Lord SA, SenkowskiCK. Improving surgical site infections: usingNational Surgical Quality Improvement Programdata to institute Surgical Care Improvement Projectprotocols in improving surgical outcomes. J AmColl Surg. 2010;210(5):737–41, 41–3.

38. Bratzler DW, Hunt DR. The surgical infection pre-vention and surgical care improvement projects:national initiatives to improve outcomes forpatients having surgery. Clin Infect Dis. 2006;43(3):322–30.

39. Stulberg JJ, Delaney CP, Neuhauser DV, Aron DC,Fu P, Koroukian SM. Adherence to surgical careimprovement project measures and the associationwith postoperative infections. JAMA. 2010;303(24):2479–85.

40. Horan TC, Gaynes RP, Martone WJ, Jarvis WR,Emori TG. CDC definitions of nosocomial surgicalsite infections, 1992: a modification of CDC defi-nitions of surgical wound infections. Infect ControlHosp Epidemiol. 1992;13(10):606–8.

41. Swoboda SM, Merz C, Kostuik J, Trentler B, LipsettPA. Does intraoperative blood loss affect antibioticserum and tissue concentrations? Arch Surg.1996;131(11):1165–71 (discussion 71–72).

42. Polly DW Jr, Meter JJ, Brueckner R, Asplund L, vanDam BE. The effect of intraoperative blood loss onserum cefazolin level in patients undergoinginstrumented spinal fusion. A prospective, con-trolled study. Spine (Phila Pa 1976). 1996;21(20):2363–7.

43. Alexander JW, Solomkin JS, Edwards MJ. Updatedrecommendations for control of surgical site infec-tions. Ann Surg. 2011;253(6):1082–93.

44. BMJ Group. What is GRADE? BMJ 2014; 2012 [up-dated Sept 20, 2012]. https://clinicalevidence.bmj.com/x/set/static/ebm/learn/665072.html. AccessedMar 14, 2014.

45. Guyatt GH, Oxman AD, Vist GE, et al. GRADE: anemerging consensus on rating quality of evidenceand strength of recommendations. BMJ.2008;336(7650):924–6.

46. Higgins JPT, Altman DG, Stern JAC. Chapter 8:assessing risk of bias in included studies. In: HigginsJPT, Green S, editors. Cochrane Handbook for Sys-tematic Reviews of Intervenions Version 510 (up-dated March 2011): The Cochrane Collaboration;2011.

47. Wells GA, Brodsky L, O’Connell D, et al. An eval-uation of the Newcastle Ottawa Scale: an assess-ment tool for evaluating the quality of non-randomized studies. In: Abstract, XI InternationalCochrane Colloquium Book of Abstracts. 2003;O-63.

48. Wells GA, Brodsky L, O’Connell D, et al. TheNewcastle-Ottawa Scale (NOS) for assessing thequality of nonrandomised studies in meta-analyses.2014. https://www.ohri.ca/programs/clinical_epidemiology/oxford.asp. Accessed Nov 15, 2014.

49. Delbecq AL, van de Ven AH. A group process modelfor problem identification and program planning.J Appl Behav Sci. 1971;7:466–92.

50. Delbecq AL, van de Ven AH, Gustafson DH. Grouptechniques for program planning: a guide to nom-inal group and Delphi processes. Glenview: ScottForesman; 1975.

51. Kanayama M, Hashimoto T, Shigenobu K, Oha F,Togawa D. Effective prevention of surgical siteinfection using a Centers for Disease Control andPrevention guideline-based antimicrobial prophy-laxis in lumbar spine surgery. J Neurosurg Spine.2007;6(4):327–9.

52. Dobzyniak MA, Fischgrund JS, Hankins S, Herkow-itz HN. Single versus multiple dose antibiotic pro-phylaxis in lumbar disc surgery. Spine (Phila Pa1976). 2003;28(21):E453–E455455.

53. Burke JF. The effective period of preventive antibi-otic action in experimental incisions and dermallesions. Surgery. 1961;50:161–8.

54. Hong HS, Chang MC, Liu CL, Chen TH. Is aggres-sive surgery necessary for acute postoperative deepspinal wound infection? Spine (Phila Pa 1976).2008;33(22):2473–8.

55. Sanchez-Ubeda R, Fernand E, Rousselot LM. Com-plication rate in general surgical cases; the value ofpenicillin and streptomycin as postoperative pro-phylaxis; a study of 511 cases. N Engl J Med.1958;259(22):1045–50.

56. Bernard HR, Cole WR. The prophylaxis of surgicalinfection: the effect of prophylactic antimicrobialdrugs on the incidence of infection followingpotentially contaminated operations. Surgery.1964;56:151–7.

57. Polk HC Jr, Lopez-Mayor JF. Postoperative woundinfection: a prospective study of determinant fac-tors and prevention. Surgery. 1969;66(1):97–103.

58. Stone HH, Hooper CA, Kolb LD, Geheber CE,Dawkins EJ. Antibiotic prophylaxis in gastric,

2732 Adv Ther (2020) 37:2710–2733

biliary and colonic surgery. Ann Surg. 1976;184(4):443–52.

59. Burdon DW. Principles of antimicrobial prophy-laxis. World J Surg. 1982;6(3):262–7.

60. Kaiser AB. Antimicrobial prophylaxis in surgery.N Engl J Med. 1986;315(18):1129–38.

61. Massie JB, Heller JG, Abitbol JJ, McPherson D,Garfin SR. Postoperative posterior spinal woundinfections. Clin Orthop Relat Res. 1992;284:99–108.

62. Savitz MH, Malis LI, Meyers BR. Prophylacticantibiotics in neurosurgery. Surg Neurol. 1974;2(2):95–100.

63. Pavel A, Smith RL, Ballard A, Larsen IJ. Prophylacticantibiotics in clean orthopaedic surgery. J Bone JtSurg Am Vol. 1974;56(4):777–82.

64. Rubinstein E, Findler G, Amit P, Shaked I. Periop-erative prophylactic cephazolin in spinal surgery. Adouble-blind placebo-controlled trial. J Bone Jt SurgBr Vol. 1994;76(1):99–102.

65. Pavel A, Smith RL, Ballard A, Larson IJ. Prophylacticantibiotics in elective orthopedic surgery: aprospective study of 1,591 cases. South Med J.1977;70(Suppl 1):50–5.

66. Geraghty J, Feely M. Antibiotic prophylaxis inneurosurgery. A randomized controlled trial. J Neu-rosurg. 1984;60(4):724–6.

67. Young RF, Lawner PM. Perioperative antibioticprophylaxis for prevention of postoperative neuro-surgical infections. A randomized clinical trial.J Neurosurg. 1987;66(5):701–5.

68. Bullock R, van Dellen JR, Ketelbey W, Reinach SG. Adouble-blind placebo-controlled trial of periopera-tive prophylactic antibiotics for elective neuro-surgery. J Neurosurg. 1988;69(5):687–91.

69. Djindjian M, Lepresle E, Homs JB. Antibiotic pro-phylaxis during prolonged clean neurosurgery.Results of a randomized double-blind study usingoxacillin. J Neurosurg. 1990;73(3):383–6.

70. Beiner JM, Grauer J, Kwon BK, Vaccaro AR. Post-operative wound infections of the spine. NeurosurgFocus. 2003;15(3):E14.

71. Emohare O, Ledonio CG, Hill BW, Davis RA, PollyDW Jr, Kang MM. Cost savings analysis of intra-wound vancomycin powder in posterior spinalsurgery. Spine J. 2014;14(11):2710–5.

72. Levi AD, Dickman CA, Sonntag VK. Management ofpostoperative infections after spinal instrumenta-tion. J Neurosurg. 1997;86(6):975–80.

73. Parker SL, Shau DN, Mendenhall SK, McGirt MJ.Factors influencing 2-year health care costs inpatients undergoing revision lumbar fusion proce-dures. J Neurosurg Spine. 2012;16(4):323–8.

74. Godil SS, Parker SL, O’Neill KR, Devin CJ, McGirtMJ. Comparative effectiveness and cost-benefitanalysis of local application of vancomycin powderin posterior spinal fusion for spine trauma: clinicalarticle. J Neurosurg Spine. 2013;19(3):331–5.

75. Gold HS, Moellering RC Jr. Antimicrobial-drugresistance. N Engl J Med. 1996;335(19):1445–533.

76. Poelstra KA, Vaccaro AR, Albert TA, et al. Gentam-icin microspheres for infection prophylaxis in therabbit spine. In: 53rd orthopaedic research society(ORS) meeting; February 10–14; San Diego, 2007.

77. Stall AC, Becker E, Ludwig SC, Gelb D, Poelstra KA.Reduction of postoperative spinal implant infectionusing gentamicin microspheres. Spine (Phila Pa1976). 2009;34(5):479–83.

78. O’Neill KR, Smith JG, Abtahi AM, et al. Reducedsurgical site infections in patients undergoing pos-terior spinal stabilization of traumatic injuries usingvancomycin powder. Spine J. 2011;11(7):641–6.

79. Kirkland KB, Briggs JP, Trivette SL, Wilkinson WE,Sexton DJ. The impact of surgical-site infections inthe 1990s: attributable mortality, excess length ofhospitalization, and extra costs. Infect ControlHosp Epidemiol. 1990s;20(11):725–30.

80. de Lissovoy G, Fraeman K, Hutchins V, Murphy D,Song D, Vaughn BB. Surgical site infection: inci-dence and impact on hospital utilization andtreatment costs. Am J Infect Control. 2009;37(5):387–97.

81. Maki DG, Schuna AA. A study of antimicrobialmisuse in a university hospital. Am J Med Sci.1978;275(3):271–82.

82. Larsen RA, Evans RS, Burke JP, Pestotnik SL, Gard-ner RM, Classen DC. Improved perioperativeantibiotic use and reduced surgical wound infec-tions through use of computer decision analysis.Infect Control Hosp Epidemiol. 1989;10(7):316–20.

83. Shapiro M, Townsend TR, Rosner B, Kass EH. Use ofantimicrobial drugs in general hospitals: patterns ofprophylaxis. N Engl J Med. 1979;301(7):351–5.

84. Hill AB. The environment and disease: associationof causation? Proc R Soc Med. 1965;58:295–300.

85. Platt R. Methodologic aspects of clinical studies ofperioperative antibiotic prophylaxis. Rev Infect Dis.1991;13(Suppl 10):S810–S814814.

Adv Ther (2020) 37:2710–2733 2733