The association between preoperative epidural steroid injections and postoperative cervical and lumbar surgical site infections: A systematic review and meta-analysis

Background Is there a statistically significant association between preoperative epidural steroid injections (ESI) and postoperative cervical and lumbar spinal surgery infections (SSI)? Methods A systematic review and meta-analysis was completed of patients 18 years or older who underwent elective cervical or lumbar spinal surgery. Those who underwent surgery with preoperative ESI were compared to those without. We assessed for differences in postoperative SSI incidence. Electronic literature databases were searched through October 2022. Peer-reviewed publications that included raw data regarding epidural exposure and non-exposure were included. Case reports, case series, abstracts, editorials, or publications that did not include raw data were excluded. Odd's ratios (OR) were calculated from the raw data collected. Meta-analysis was done using RevMan v5 with a fixed effects model. Results We identified 16 articles for inclusion. When not controlling for the type of surgery and time from ESI to surgery, there was a statistically significant OR between preoperative ESI and postoperative SSI. The association persisted when the ESI was performed within 30 days or 31-90 days of the surgery. No association was discovered when evaluating only cervical spine surgeries. The evidence is assigned a “moderate” GRADE rating. Conclusions Our analysis shows a small, time-dependent, statistically significant association between preoperative ESI and postoperative lumbar SSI may exist. However, the OR produced, while statistically significant, are close enough to 1.0 that clinically, the effect size is “small.” The number needed to treat for an ESI in the appropriate clinical setting is, at worst, 3. The number needed to harm, meaning the number of patients who undergo an ESI at any time before their spine surgery and then develop a SSI, is 111 patients. Ultimately, the surgical sparing potential from an ESI outweighs the SSI risk based on our findings.


Introduction
There is inherently an increased risk of infection with any injection.The number of published articles on preoperative epidural steroid injections and postoperative surgical site infections is sparse.An insurance database review showed that 46.4% of patients had a lumbar epidural steroid injection within a year before lumbar spinal surgery due to a disc herniation or spinal stenosis [ 1 ].Another meta-analysis of patients undergoing cervical, thoracic, or lumbar spine surgery, regardless of ESI

Study selection
Two authors with formal training and certification in the principles of evidence-based medicine independently assessed the titles and abstracts for relevance and eligibility.The primary author resolved discrepancies.Subsequently, 3 authors independently reviewed the selected publications in full and evaluated them for inclusion.The primary author resolved discrepancies.Additionally, the references were reviewed to assess whether further studies should be included.Studies were included if they presented clinically relevant data on the relationship between preoperative epidural steroid injections and postoperative spinal surgical site infections.Moreover, the studies were excluded if they did not provide raw data which could be used for statistical analyses ( Fig. 1 ).

Data items and collection
Reviewers extracted the following data from each study: (1) bibliographic details, (2) study design, (3) the source of the data (for example, single site, multi-site, database), (4) the surgery performed, (5) sample size (6) raw data between the 2 groups of those who did receive an epidural steroid presurgical intervention and those who did not, (7) published Odds Ratio and statistical significance, and (8) any reported timing between when the epidural was given preoperatively.Additionally, commentary was kept regarding aberrancies in methodology or statistics, selected Current Procedural Terminology (CPT)/International Classification of Diseases (ICD) codes, patient populations, and outcomes that may suggest bias.

Data extraction
Articles that presented raw data regarding the number of patients with preoperative epidural steroid injections who had postoperative surgical site infections, patients with preoperative epidural steroid injections who did not have postoperative surgical site infections, patients without preoperative epidural steroid injections who had postoperative surgical site infections, and patients without preoperative epidural steroid injections who did not have postoperative surgical site infections had their data transferred into a password protected digital spreadsheet.
Articles that presented raw but incomplete data were evaluated to see if complete data could be curated by creating a 2 ×2 contingency table based on the published raw data in conjunction with published odds ratios, such that the missing data could be calculated by a reverse odds ratio calculation.
In scenarios in which time-matched exposure groups (0-30 days, 31-90 days, etc.) were compared against a single control group that was not time-matched for each respective cohort, the inclusion of such comparison data was deemed inappropriate for a subcohort time-based metaanalyses but appropriate for non-time matched meta-analyses.
Three authors independently reviewed each study for data.Discrepancies between authors were rectified by a collaborative meeting in which the data in question was reviewed between the parties to determine why the discrepancy occurred and then fix the discrepancy.

Summary measures and synthesis of results
The primary outcome of interest was the odds ratio association with preoperative epidural steroid injections and postoperative spinal surgical site infection.For statistical purposes, studies were to be grouped by ESI, timing from injection to surgery, location of surgery, type of surgery, and the etiology of the data acquired (i.e., database, cohort).

Risk of bias and methodological assessment
The body of evidence was evaluated using the Grades of Recommendation, Assessment, Development, and Evaluation (GRADE) appraisal system to determine the evidence's quality regarding the association.The GRADE system transparently evaluates the body of evidence in domains including risk of bias, imprecision, inconsistency, indirectness, and publication bias.GRADE provides an initial rating of quality based on the best available evidence.It allows for upgrading ( e.g., large magnitude of effect, dose-response gradient) or downgrading ( e.g., risk of bias, indirectness) of the evidence quality.Two authors independently reviewed the selected literature and provided assessments.A collaborative discussion between the 2 authors resolved disagreements regarding GRADE evaluation.

Statistical methodology
RevMan v5.4.1 was used for data synthesis and meta-analysis.Incidence and population data were extracted from published manuscripts.Meta-analysis was performed using fixed effects, and alpha was set at 0.05.

Systematic review
There were 1,303 articles initially identified.After removing duplicates (240), 1,036 articles were screened via abstract and title review.A total of 29 articles were deemed eligible for full review.14 articles were excluded due to a lack of raw published data or lack of clarity on preoperative epidural steroid injections and postoperative spinal surgical site infections (9), retracted (1), editorial/commentary (1), meta-analyses (2), and a narrative review (1).Thus, 15 articles were identified for inclusion in the Systematic Review.The PRISMA flow chart visualization of this process is Fig. 1 [ 1 , 7-19 ].Two studies were included in the systematic review but excluded from the meta-analysis due to statistical, methodological, or publication issues that could not be reconciled [ 12 , 20 ].

Retrospective cohort studies
Zusman et al.was a single-center retrospective study of patients who underwent elective thoracic and lumbar spine fusion surgery, not for trauma, tumor, or infectious reasons, from 2007 to 2010.Among other complications, they evaluated postoperative surgical site infections captured within 30 days of the index surgery.They define epidural steroid injection by patient report of preoperative "spinal injection " at any time point.They do not delineate if that injection was an epidural injection, facet injection, medial branch block, trigger point injection, radiofrequency ablation, intramuscular injection, or other injection.Despite this, all patient-reported spinal injections are referred to as epidu-ral steroid injections in the manuscript's title, abstract, results, and discussion.They found a nonstatistically significant increase in risk associated with preoperative patient-defined spinal injections and postoperative surgical site infection within 30 days of the index lumbar fusion.They did add that the complications did not affect outcome measures [ 7 ].
Hartveldt et al. was a dual-center retrospective study of patients who underwent lumbar spine surgery, both fusion and non-fusion, for degenerative lumbar spine conditions from 2005 to 2015 at Massachusetts General Hospital.They evaluated preoperative epidural steroid injections at 0-30 days, 30-90 days, and 0-90 days from surgical intervention to assess for postoperative surgical site infections.The identified preoperative epidural steroid injections using CPT codes.Among the selected CPT codes, 0217T was included.CPT 0217T is for ultrasound-guided facet injections, not lumbar epidural steroid injections (See Table 3 ).They defined surgical site infections as requiring reoperation or incision and drainage due to infection within 90 days of the index surgery.They identified no statistically significant association between preoperative steroid injections at 0-30 days, 30-90 days, and 0-90 days and postoperative surgical site infection within 90 days of the index surgery [ 8 ].
Ozturk et al. was a single-site study out of Turkey that evaluated patients who had undergone unilateral, single-level lumbar microdiscectomies due to extruded or sequestered disc material.While many variables were captured, they evaluated whether preoperative epidural steroid injections, defined as a mix of 80 mg triamcinolone and 3 mL of .5% bupivacaine injected into the anterior epidural space via a transforaminal approach, altered the risk of postoperative surgical site infections.Of note, it is the only study that defines the injectate's specifics.While they did not explicitly define surgical site infection, no infections were reported in the steroid or control groups.Given a lack of events, this individual study could not calculate an odds ratio.Ultimately, they were unable to find a difference between preoperative transforaminal epidural steroid injections given at 0-30 days, 30-60 days, 60-180 days, 180-365 days, or > 365 days and postoperative surgical site infections [ 13 ].
Kreitz et al. was a retrospective single-site study from Philadelphia, PA, which looked at the risk of preoperative ESI and postoperative SSI in patients who had undergone lumbar decompression or lumbar fusion performed for a diagnosis of lumbar radiculopathy and/or spinal stenosis with minimum 90 days follow up.They defined preoperative ESI via the following CPT codes: 62311, 64475, 64483, and 64493.However, 64475 and 64493 are CPT codes for facet injections (See Table 3 ).Therefore, the study's results included procedures that are not epidural steroid injections.They used ICD codes 996.67, 998.12, 998.31, 998.32, 998.59,T81.31XA, T81.32XA, and T84.7XXA to define surgical site infection.The study does not delineate the type of epidural performed, the type of steroids used, or the dosage of the injectate.There was no statistically significant difference between the ESI and control group for the lumbar decompression without fusion group.There was a statistically significant difference between the ESI and control group for the lumbar fusion group [ 15 ].Shakya et al. was a retrospective single-site study from India that evaluated patients aged 21-65 who had undergone lumbar discectomy for single-level disc herniations from 2017 to 20. 129 patients received a transforaminal epidural steroid injection at 0-90 days, 90-180 days, and > 180 days before moving onto surgery, while 186 patients went directly to lumbar discectomy.They do not define the injectate or dose for the transforaminal epidural steroid injection.They do not define SSI but report that identification of infection was determined by retrospective chart review.There was no statistically significant difference between the 2 groups concerning surgical site infection.Interestingly, all 129 patients who received a transforaminal epidural steroid injection preoperatively matriculated onto lumbar discectomy.Zero patients found "substantial relief " with preoperative transforaminal epidural steroid injection [ 17 ].

Retrospective database studies
Cancienne et al. is a retrospective analysis of a Medicare PearlDiver Patient Record Database that was used to compare postoperative infection rates within 90 days for patients who had undergone an anterior cervical discectomy and fusion (ACDF) or posterior cervical fusion (PCF) and had a preoperative cervical epidural steroid injection from 2005 to 2012.CPT Codes 22554, 22551, 22585, 22590, and 22600 were used to identify patients with either an ACDF or PCF.CPT codes 64479 and 62310 were used to identify patients with a preoperative cervical epidural steroid injection.The exposure group was controlled for 0-90 days, 91-180 days, and 181-365 days.The study defines surgical site infection as a postoperative infection within 90 days and uses the CPT codes 20005, 10180, 21501 or ICD-9: 998.5, 998.51, 998.59, 996.67, 996.69 to search the database.There was a statistically significant association for post-ACDF SSI when preoperative cervical epidural steroid injections were given 0-90 days before the operation.There was a statistically significant association for post-PCF SSI when preoperative cervical epidural steroid injections were given 0-180 days before the operation [ 19 ].
Singla et al. was a retrospective study that utilized the Medicare PearlDiver Database from 2005 to 12 to evaluate preoperative ESI association with postoperative SSI related to 1-2 level posterior lumbar fusion surgery.Lumbar surgeries were identified via CPT codes.Lumbar epidural steroid injections were identified via CPT codes.Surgical site infections were defined by ICD-9 codes within 90 days of the index lumbar fusion surgery.They found a statistically significant association between postoperative lumbar fusion surgery surgical site infection and preoperative lumbar epidural steroid injections at 0-30 days and 31-90 days, but not 91-180 days before the surgery [ 11 ].
Yang et al. was a retrospective study that utilized the Medicare Pearl-Diver Database from 2005 to 12 to evaluate preoperative lumbar epidural steroid injections association with preoperative SSI as it relates to 1-2 level lumbar decompression surgery.Lumbar surgeries were identified via CPT codes 63005, 63030, 63047.Lumbar epidural steroid injections were identified via CPT codes 64483 and 62311.Patients who received facet interventions were excluded.Surgical site infections were defined by ICD-9 codes 998.5, 998.51, 998.59, and CPT codes 20005 and 22015 within 90 days of the index lumbar decompression surgery.Surgeries that preoperatively involved infection or tumor were excluded.They found a statistically significant risk for postoperative surgical site infection associated with preoperative lumbar epidural steroid injections at 0-30 days and 31-90 days before the operation but found no statistically significant association when the injection was given > 90 days from the operation [ 9 ].Donnally et al. produced a nearly identical study to Yang et al.Donnally and Yang defined the surgery using the 63030 and 63047 CPT codes, but Yang added the 63005 CPT code for both 1 and 2level decompressions.Using the Medicare PearlDiver Database, Donnally reviewed from 2005 to 14, while Yang reviewed from 2005 to 12.They used identical codes for epidural steroid injections: 64483, 62311.They defined surgical site infections using ICD 9 codes 998.5, 998.51, 998.59,CPT Code 20005, and 22015.The only difference in the definition of a surgical site is that Donnally added 996.67, which is infection/inflammation from an orthopedic device, implant, or graft.Despite the similarities, Donnally 2018 found a statistically significant association between preoperative epidural steroid injections and postoperative surgical site infections for epidural steroid injections at 31-90 days and 91-180 days preoperative.Unlike Yang, Donnally found no statistically significant association with ESI given within 30 days of surgery.In summary, these 2 studies used similar search methods in the same database over a similar period.However, Yang found statistically significant risk at 0-30 days and 31-90 days, while Donnally found risk at 31-90 days and 91-180 days [ 9 , 12 ].The only agreement was between the increased risk associated with a preoperative epidural steroid injection at 31-90 days.
Given the overlap between the Donnally and Yang study, including both studies for the meta-analysis is inappropriate, given that an assumed large percentage of the data will then be counted twice without a way to differentiate such data.Including the Yang data eliminates 2 years of data obtained by the Donnally study.Including the Donnally study eliminates all 2-level fusion data from the Yang study.Maintaining the standard set by the Kazarian 2022 and Patel 2022 Meta-Analyses, we have elected to use the Yang data and not use the Donnally data for our Meta-Analyses [ 5 , 6 ].
Seavey et al. was a retrospective study that utilized the Military Health Systems Data Repository from 2009 to 14 to evaluate preoperative lumbar epidural steroid injections association with postoperative SSI related to 1-2 level lumbar decompression surgery.Lumbar surgeries were identified via CPT codes 63005, 63030, 63047, 63056, and 63005.Lumbar epidural steroid injections were identified via CPT codes 023T, 64483, and 64484.Patients who received facet interventions were excluded.Surgical site infections were defined by ICD-9 codes 998.51 and 998.59 within 90 days of the index lumbar decompression surgery.Surgeries that preoperatively involved infection or tumor were excluded.They found no statistically significant risk associated with preoperative lumbar epidural steroid injections at 0-30 days, 31-90 days, 0-90 days, 91-180 days, 181-365 days, or > 365 days.Given the differences in outcomes from this study relative to previously published studies, specifically Yang et al., they ran an additional cohort analysis that included only patients > 65 years old to isolate a Medicare-eligible population.There was an increase in infection rates in the > 65-year-old population who received lumbar epidural steroid injections compared to those who had not, but the difference was not statistically significant [ 10 ].
Pisano et al. was a retrospective study that utilized the Military Health Services Database from 2009 to 14 to evaluate preoperative ESI association with postoperative SSI related to lumbar fusion surgery.Lumbar surgeries were identified via CPT codes.To differentiate between the 2, ESI and facet interventions were identified via CPT codes, including the raw data.SSI was defined by ICD codes.They found no statistically significant risk associated with preoperative ESI and/or facet interventions with preoperative SSI at 0-30 days, 31-90 days, 91-180 days, 181-365 days, or > 365 days [ 14 ].
Koltsov et al. used the IBM MarketScan Database from 2007 to 15 to retrospectively evaluate all patients who underwent lumbar spine surgery, both fusion and nonfusion, for lumbar disc herniation or lumbar spinal stenosis.The defined ESI using CPT codes.They then evaluated those with a preoperative epidural steroid injection at 0-30 days, 31-60 days, 61-90 days, and 91-365 days to matched controls who had not undergone preoperative ESI.There was no statistically significant difference at any measured time in postoperative surgical site infections when comparing those who received a preoperative ESI to those who did not [ 1 ].
Wadhwa et al. was a retrospective study that utilized the IBM Mar-ketShare Database from 2007 to 16 to evaluate preoperative ESI association with postoperative SSI as it relates to cervical spine surgery for cervical degenerative disorders.They excluded patients < 18 years old, surgeries related to tumors, or surgeries related to trauma.They isolated the cohort via ICD-9 identification of those with cervical spinal degenerative disorders and then cross-referenced those codes with the CPT codes for 1-2 level cervical spine surgeries.They defined ESI using multiple CPT codes which were not specific to an ESI injection.They included patients in their exposure cohort who underwent cervical epidural steroid injections, lumbar epidural steroid injections, epidural catheter placement, and facet interventions.They evaluated preoperative injections at 3, 6, 12, 18, and 24 months.The evaluated postoperative SSI was determined by ICD codes for reoperation within 90 days from the index surgery.They then matched controls to the respective measured time periods in the injection cohort.They detected no statistically significant difference between those who received a preoperative spinal injection at any point and SSI within 90 days of index cervical spine surgery, except for a statistically significant difference identified among those receiving preoperative injections at 91-180 days before cervical spine surgery.However, this study did not draw its conclusion from epidural steroid injections alone.Therefore, we cannot definitively conclude that any specific injection is associated with the difference observed at the 91-180-day time point [ 16 ].The study looked at patients who developed SSI versus matched controls who did not.They then reviewed whether those patients had or had not received a spinal injection before their operation.The study does not delineate between epidural steroid injections, facet joint interventions, or other spinal injections.There was no statistically significant difference in SSI risk observed between those with a preoperative spinal injection and those without a preoperative spinal injection [ 21 ].

Prospective cohort studies
Li et al. prospectively examined patients from 2015 to 2019 who underwent posterior lumbar fusion at a single center in China.They compared those who did and did not receive a preoperative ESI before their operation within 0-30 days and > 30 days from their operation.In the 0-30 day injection group, 3.5 + /-1.0 levels were fused.In the > 30-day day injection group, 3.3 + /-.9 levels were fused.In the control group, 3.2 + /-.9 levels were fused.It is unclear from the publication how patients were selected for epidural injections.All patients who received a preoperative epidural steroid injection matriculated to surgical intervention.Furthermore, the article reports that some epidural injections were performed with steroids while others were performed with lidocaine alone.There is no explanation for why some might get steroids while others lidocaine alone.The published data does allow for the delineation of those who got an epidural steroid injection and those who received an epidural lidocaine injection.Of the studies reviewed, this study had the highest infection rate, with a control rate of 3.5% and a < 30-day ESI rate of 10.5%.They produced a statistically significant association between preoperative ESI when given < 30 days from surgery and postoperative SSI.The association did not exist in the < 30 days group when no steroid was used in the epidural injection.However, no association beyond that time point was found [ 18 ] ( Table 2 ).

I. ESI Only Studies Not Controlled for Time
Eight studies were included in the meta-analysis of postoperative infection risk following exclusively epidural steroid injections with no control for time.A total of 2.12% (3,383/159,295) of patients who underwent a preoperative epidural steroid injection experienced a postoperative infection compared to 1.23% (7,250/591,314) of controls.This represented a statistically significant increase in surgical site infection risk (OR = 1.08, 95% CI 1.03-1.13,p < .00001;heterogeneity: I2 = 88%).The Number Needed to Harm (NNH) is 111 patients.The control data was not controlled for each subcohort.We could not reconcile the difference in OR between the published OR and our manually calculated figure.Eight studies were included in the meta-analysis for postoperative infection risk stratified to include all lumbar studies controlled exclusively for epidural steroid injections received 30 days or less prior to lumbar surgery.A total of 2.23% (637/28568) of patients experienced postoperative infection, as compared to 1.41% (2,195/155797) of controls, which did constitute a statistically significant difference (OR = 1.28, 95% CI 1.15-1.43,p < .00001);heterogeneity: I 2 = 91%).The Number Needed to Harm (NNH) is 95 patients.

III. ESI only studies within 31-90 days from lumbar surgery
Seven studies were included in the meta-analysis for postoperative infection risk stratified to include all lumbar studies controlled exclusively for epidural steroid injections received 31-90 days prior to lumbar surgery.A total of 2.27% (1,296/57,125) of patients experienced postoperative infection, as compared to 1.63% (2,682/164,183) of controls, which did constitute a statistically significant difference (OR = 1.08, 95% CI 1.00-1.16,p < .00001);heterogeneity: I2 = 78%).The NNH is 102 patients.

IV. All studies not controlled for time
Fourteen studies were included in the meta-analysis of postoperative infection risk following spinal injections with no control for time.Studies that included non-ESI-specific data were included in this figure .A total of 1.85% (3,613/194796) of patients who underwent a preoperative epidural steroid injection experienced a postoperative infection compared to 1.2% (7,668/638866) of controls.This represented a statistically significant increase in surgical site infection risk (OR = 1.08, 95% CI 1.03-1.12,p < .00001;heterogeneity: I2 = 79%).This procedure is the injection of an anesthetic agent and/or steroid in the form of a transforaminal epidural injection into a single level (either cervical or thoracic) [ 3 ].64480 Wadhwa This procedure is the injection of an anesthetic agent and/or steroid in the form of a transforaminal epidural injection in the cervical or thoracic region.The code applies to each additional level after the initial level [ 4 ].64483 Kreitz,

V. All cervical studies only not controlled for time
This procedure is the injection of an anesthetic agent and/or steroid in the form of a transforaminal epidural injection into a single level (either lumbar or sacral) [ 5 ].64484 Hartveldt, Seavey, Koltsov, Pisano, Wadhwa This procedure is the injection of an anesthetic agent and/or steroid in the form of a transforaminal epidural injection in the lumbar or sacral region [ 6 ]. 64493 Kreitz.Pisano In this service, the provider injects a diagnostic or therapeutic agent into a facet joint, the joint connecting 2 spinal vertebrae together, at the lumbar or sacral level.Or he may perform the injection for nerves innervating that joint.He uses imaging guidance of either fluoroscopy or CT scan [ 7 ].996.67 Kreitz, Cancienne, Donnally, Singla Infection and inflammatory reaction due to other internal orthopedic device implant and graft [ 8 ]. (p67) 996.69 Cancienne Infection and inflammatory reaction due to other internal prosthetic device implant and graft [ 9 ]. (p69) 998.12 Kreitz Hematoma complicating a procedure [ 10 ]. 998.31 Kreitz Disruption of internal operation (surgical) wound [ 11 ]. (p31) 998.32 Kreitz Disruption of external operation (surgical) wound [ 12 ]. (p32) 998. 5 Cancienne, Yang, Donnally, Singla Postoperative infection not elsewhere classified [ 13 ].The provider incises the area of infection and drains any fluid collection, with the help of surgical instruments [ 23 ].
( continued on next page )

21501
Cancienne A provider performs an incision and drainage procedure in the deep tissues of the neck or chest to relieve pain and pressure from a pocket of blood or pus [ 24 ].22015 Yang, Donnally, Singla The provider incises and drains an abscess, or pocket of infection, in the deep tissues at the back of the lower spinal column to relieve pain and pressure [ 25 ].
2 studies were included in the meta-analysis of postoperative infection risk stratified out to include all cervical studies with no control for time.This meta-analysis included studies that included spinal injections beyond epidural steroid injections, namely facet interventions.A total of 0.57% (254/44,511) of patients who underwent a preoperative injection experienced postoperative infection compared to 0.85% (2,798/330,096) of controls.This represented no statistically significant association with surgical site infection risk (OR = 1.09, 95% CI 0.94-1.27,p = .10;heterogeneity: I2 = 57%).

VI. Additional subgroup analyses
Five studies were included in the meta-analysis for postoperative infection risk stratified to include studies that exclusively evaluated lumbar fusion with no control for time.This meta-analysis included studies that included spinal injections beyond epidural steroid injections, namely facet interventions.A total of 1.99% (386/19,390) of patients experienced postoperative SSI, as compared to 1.63% (1,277/78,571) of controls.This represented a statistically significant increase in surgical site infection risk (OR = 1.26, 95% CI 1.12-1.41,p < .0001;heterogeneity: I2 = 60%) [ 7 , 11 , 14 , 15 , 18 ].
Three studies were included in the meta-analysis for postoperative infection risk stratified to include studies that exclusively evaluated lumbar decompression with no control for time.This meta-analysis included studies that included spinal injections beyond epidural steroid injections, namely facet interventions.A total of 1.00% (226/22,499) of patients experienced postoperative SSI compared to 0.63% (756/118,423) of controls.This represented a statistically significant increase in surgical site infection risk (OR = 1.55, 95% CI 1.33-1.80,p < .00001;heterogeneity: I2 = 82%) [ 9 , 10 , 15 ].
Three studies were included in the meta-analysis for postoperative infection risk stratified to include only studies from the Medicare pearlDiver database with no control for time.This meta-analysis in-cluded studies that included spinal injections beyond epidural steroid injections, namely facet interventions.A total of 1.26% (640/50,776) of patients experienced postoperative infection compared to 0.93% (4,406/475,927) of controls.This represented a statistically significant increase in surgical site infection risk (OR = 1.30, 95% CI 1.19-1.42,p < .00001;heterogeneity: I2 = 85%) [ 9 , 11 , 19 ].
Four studies were included in the meta-analysis for postoperative infection risk stratified to include only studies from the IBM MarketScan and Military Health Services databases with no control for time.This meta-analysis included studies that included spinal injections beyond epidural steroid injections, namely facet interventions.A total of 2.07% (2,860/138,440) of patients experienced postoperative infection, as compared to 2.01% (2,931/145,648) of controls, which did not constitute a statistically significant increase in surgical site infection risk (OR = 1.00, 95% CI 0.95-1.06,p = .94;heterogeneity: I2 = 0%) [ 1 , 10 , 14 , 16 ].
Eleven studies were included in the meta-analysis for postoperative infection risk stratified to include only studies from the non-Medicare PearlDiver database studies with no control for time.This meta-analysis included studies that included spinal injections beyond epidural steroid injections, namely facet interventions.A total of 2.06% (2,963/144,152) of patients experienced postoperative infection compared to 2.00% (3,262/17,743) of controls.This did not represent a statistically significant increase in surgical site infection risk (OR = 1.01, 95% CI 0.96-1.06,p = .010;heterogeneity: I2 = 57%) [ 1 , 7 , 8 , 10 , 13-18 , 21 ].

GRADE assessment of the evidence regarding the association between epidural steroids and postspinal surgery surgical site infections
When applying GRADE, the resulting body of evidence is assigned a "moderate " GRADE quality of evidence rating.There were no disagreements that required third-party intervention.This rating is attributed to the variety of definitions of surgical site infection and the variety in definitions of what type of injection qualified as an "epidural steroid injection." Only 1 study identified the epidural technique, medication, and dose when discussing an epidural steroid injection.There was an inability to assess the specifics regarding the instrumentation used for fusion surgeries and the accuracy of the databases accessed for many of the large retrospective studies [ 22 ].We are moderately confident in the estimate based on the published literature.However, there is a possibility that prospective studies may yield different results.

Discussion
There appears to be a statistically significant association between preoperative ESI and postoperative lumbar spine SSI.The association was statistically strongest at ESI 0-30 days from surgery, less so at ESI 31-90 days from lumbar spinal surgery, and no longer statistically significant when the ESI was given > 90 days from the lumbar spinal surgery.These results suggest a timing association between when the ESI is given preoperatively and preoperative ESI.
The odds ratios are close enough to 1.0 regardless of stratification by surgery type, the database used, or specific injection that clinically the effect size is "small " or "weak " [ 23 , 24 ].For some providers, such an effect size may be irrelevant in decision-making.
There were 2 previously published Meta-Analyses by Kazarian and Patel.Kazarian recommended against all corticosteroid injections within 1 month of any spinal surgery.It should be noted that all corticosteroid injections are not synonymous with epidural steroid injections.While our study did not address "all corticosteroid injections, " some of our meta-analyses, which included non-ESI spine injections, found that there was a small, but likely clinically irrelevant, association between preoperative spinal injections and postoperative SSI regardless of time given prior to surgery (1.08, 95% CI 1.03-1.12).
However, in Kazarian's "leave-out " meta-analysis, they eliminated the results of Singla and Yang.The exclusion eliminated the association between CSI 0-30 days before surgery and postoperative infection, suggesting a weighted bias from those 2 studies.Kazarian concluded that an association between preoperative CSI 0-30 days before surgery and postoperative infection hinged solely on including Singla and Yang.
A key facet of the Singla and Yang studies is that they share the same data source, Medicare PearlDiver.We also found that when eliminating the Medicare PearlDiver database from our meta-analyses, a statistically significant association was no longer present, small as it may have been.
Our analysis further strengthens the signal that the Medicare PearlDiver database may uniquely capture a patient population more at risk for preoperative SSI than other sources.

Limitations of the evidence included in the review
Our conclusions are limited as the data is > 99% retrospective and lacks specificity regarding the injections performed to draw more helpful conclusions to guide clinical decision-making.Only 1 study identified the type of steroid, amount of injectate, injection approach, and localization of injectate.Any of these factors may affect the potential risk passed to patients.Interestingly, 3 studies exclusively reviewed the use of TFESI [ 10 , 13 , 17 ].The pooled data from those studies, which includes decompression and discectomy patients, produces no statistically significant difference (1.44 [0.74-2.79]p = .2815)in preoperative surgical site infection, which raises the question regarding whether technique plays a role.
The Medicare PearlDiver Database is a private national database comprising 41 billion patient records from commercial insurance, government claims, and other sources.The database is restricted to coding and billing of patients 65 or older and thus may misrepresent certain aspects of care given regional variability in coverage.The database evaluates potential associations between variables but should look to prospective studies to determine causal relationships [ 25 , 26 ].
The IBM MarketScan database comprises commercial claims from inpatient, outpatient, and pharmaceutical claims of over 75 million employees, retirees, and dependents.A substantial portion of the American population is covered by employer-sponsored insurance and thus represented in this database [ 16 ].
The Military Health Services Data Repository is made up of over 9 million military and civilian patients from the age of 18-64.While a pediatric population is included as well, for our database reviews, this portion of the database was not queried per the study's methodology [ 27 ].
In our meta-analysis, only the Medicare PearlDiver database detected a statistically significant risk associated with preoperative epidural steroid injection.Is this finding due to a confounding variable reporting error, or is it a real independent risk factor unique to the Medi-care PearlDiver patient population?Database reviews are limited by the accuracy of coding, the populations from which they are drawn, and the specificity of the coding.Their limitations are well documented [ 22 , 28 , 29 ].

Implications for practice, policy, and future research
This study is warranted because, in 2022, there were 2 different meta-analyses, which used 2 different sets of data and reached 2 different conclusions, with each suggesting changes to practice habits.Our effort is now the third entry, but it uses more studies and raw data and attempts to inform readers regarding the heterogeneity of studies published to date to best inform clinical decision-making.
While we also concluded a time-dependent statistically significant association between preoperative ESI and postoperative SSI, we cannot comment on the temporal relationship based on the type of spinal surgery (decompression vs fusion) based on statistical issues regarding controlling for time described in Table 1 .Based on the available evidence, there is a time-dependent statistically significant association between preoperative ESI and postoperative SSI.However, the data does not elevate to a level that should limit ESI use due to a risk passed along to the patient.
We state this because there is a surgical sparing benefit from ESI depending on the pathology, chronicity of symptoms, and the specific injection/injectate used.Some studies demonstrate that surgical sparing benefits may be as high as 80% [30][31][32][33][34].The number needed to treat (NNT) for an epidural steroid injection in the appropriate clinical setting is, at worst 3 [ 32 , 35 ].The number needed to harm (NNH), meaning the number of patients who undergo an epidural steroid injection then develop an SSI which per our study may be attributed to that epidural steroid injection is 111 patients.When balancing the surgical sparing benefit of an ESI to the OR identified in our research of postoperative SSI, readers should feel strengthened in our recommendation to consider ESI before moving forward with surgical intervention in the appropriately selected patients.
Interestingly, upon reviewing our selected articles with the knowledge of the potential surgical sparing benefit of ESI relative to the risk of postoperative SSI, it was discovered that somewhere between 9% and 46% of patients trialed an ESI before moving forward with surgical intervention in the studies which made such data available [ 7 , 8 , 10 , 14-19 , 30 ].These percentages felt low relative to what we expected, especially given that the patients tended to skew towards single-level disc herniations or stenotic lesions.We would opine that using even the most conservative estimates of the surgical sparing benefit of an ESI, the percentage of patients trialing an ESI should be higher.
There has yet to be a published prospective study from the United States on this topic.Despite this, preoperative ESI use is ubiquitous.To inform better practice and policy regulations involving the intersection of ESI and spinal surgery, researchers ask a more specific question and follow the results prospectively.If designed appropriately, such a study can alter practice habits.

Conclusion
Our analysis shows a small, albeit statistically significant, association between preoperative ESI and postoperative lumbar SSI may exist.The relationship may be time-dependent as the statistical strength of the association decreased with time from injection to surgery.However, the odds ratios produced, while statistically significant, are close enough to 1.0 regardless of stratification by surgery type, database used, or specific injection that clinically the effect size is "small " or "weak." At worst, the NNT for an ESI in the appropriate clinical setting is 3.The number need to harm (NNH), meaning the number of patients who undergo an ESI and then develop a SSI, which per our study, may be attributed to that ESI, is 111 patients.Ultimately, the surgical sparing potential from an ESI outweighs whatever SSI risk exists based on our findings.

Farshad
et al. prospectively examined patients who underwent lumbar spinal decompression with or without fusion surgery in the Swiss Lumbar Stenosis Outcome Study, a multicenter cohort study of patients with symptomatic lumbar spinal stenosis from hospitals in Switzerland.

Table 1
Included studies.
( continued on next page )

Table 1 (
continued ) ( continued on next page )

Table 2
Statistical summary.
( continued on next page )