Additive risk of surgical site infection from more than one risk factor following craniotomy for tumor

This study seeks to expound upon risk factor etiologies for surgical site infection (SSI) and investigate their combinatorial effects on infection rate following craniotomy for neuro-oncologic pathology. Patients who underwent neuro-oncologic craniotomy between 2006 and 2020 were included. Medical records were reviewed to identify the occurrence of wound infection at ≤ 3 months postoperatively. Potential risk factors for infection included tumor pathology, location, anesthesia type, indication, ventricular entry, foreign body, brachytherapy, lumbar drain, prior operation, prior cranial radiation, prior infection, bevacizumab, and medical comorbidities (hypertension, obesity, diabetes, hyperlipidemia, other cancer, cirrhosis). Logistic regression was implemented to determine risk factors for SSI. Chi-square tests were used to assess whether the number of risk factors (e.g., 0, ≥ 1, ≥2, ≥ 3, ≥4) increases the risk of SSI compared to patients with fewer risk factors. The relative increase with each additional risk factor was also evaluated. A total of 1209 patients were included. SSI occurred in 42 patients (3.5%) by 90 days after surgery. Significant risk factors on multivariate logistic regression were bevacizumab (OR 40.84; p < 0.001), cirrhosis (OR 14.20, p = 0.03), foreign body placement (OR 4.06; P < 0.0001), prior radiation (OR 2.20; p = 0.03), and prior operation (OR 1.92; p = 0.04). Infection rates in the combinatorial analysis were as follows: ≥1 risk factor = 5.9% (OR 2.74; p = 0.001), ≥ 2 = 6.7% (OR 2.28; p = 0.01), ≥ 3 = 19.0% (OR 6.5; p < 0.0001), ≥ 4 = 100% (OR 30.2; p < 0.0001). Risk factors in aggregate incrementally increase the risk of postoperative SSI after craniotomy for tumor.


Introduction
The incidence of surgical site infection (SSI) after craniotomy ranges from 2.2 to 19.8% [1]. SSI significantly contributes to patient morbidity and has been associated with increased health care expenditures, reoperations, antibiotic treatments, and prolonged hospital stays [1,2]. Numerous studies have attempted to identify pre-and intraoperative Theodore H. Schwartz schwarh@med.cornell.edu

Methods
After obtaining Institutional Review Board approval, a prospective database of consecutive neuro-oncologic craniotomies performed by a single surgeon (THS) between January 2006 and June 2020 at a single institution (NewYork-Presbyterian Hospital-Weill Cornell Medicine) was retrospectively reviewed. Patient consent was not required (IRB 0606008601). Cases involving burr holes, endoscopic procedures, epilepsy surgeries, or transsphenoidal and decompressive craniectomies were excluded.
Medical records were reviewed to identify potential risk factors for infection. These included tumor pathology, tumor location (intraparenchymal versus extraparenchymal), cranial location (supratentorial versus infratentorial), anesthesia type (general versus awake), indication (biopsy versus resection), and potential risk factors for infection including: ventricular entry, presence of a foreign body (ventriculoperitoneal shunts (VPS), external ventricular drains (EVD), brachytherapy, intracranial pressure (ICP) monitors), lumbar drains, prior operation/biopsy (not limited to the same operative site), prior cranial radiation, prior infection, use of bevacizumab, and medical comorbidities including hypertension, obesity, diabetes, hyperlipidemia/hypercholesterolemia, other cancer (not neuro-oncologic), and cirrhosis.
The first step was to determine which patients developed wound infections. Both outpatient and inpatient medical records were reviewed to establish the occurrence of postoperative wound infections at either 1-or 3-months after surgery. Infection was established through reoperation and confirmed with subsequent bacterial cultures. All patients with suspected SSIs were taken for immediate reoperation and no patients were managed with antibiotics alone. All patients received prophylactic intravenous antibiotics consisting of cefazolin (or vancomycin if allergic to penicillin and its derivates) within an hour of surgery that continued for 24 h after surgery.
A multivariate logistic regression was subsequently implemented to determine which potential risk factors incurred a statistically significant risk of wound infection. The next step was to evaluate whether the number of risk factors (e.g., none, one or more, two or more, three or more, four or more) increases the risk of SSI. The relative increase with each additional risk factor was also assessed. Given the decrease in sample size that occurs when assessing patients with greater numbers of concomitant risk factors, we opted to establish groups based on the overall quantity of these variables, rather than their particular etiologies. Risk factor groups were therefore compared to corresponding groups with fewer risk factors (e.g., ≥ 1 vs. none, ≥ 2 vs. ≤ 1, ≥3 vs. ≤ 2, ≥4 vs. ≤ 3). Only variables found to be significantly correlated with SSI on multivariate logistic regression were used to categorize patients into risk factor groups. All p-values are two-sided with statistical significance evaluated at the 0.05 alpha level. Analyses were performed in SAS Version 9.4 (SAS Institute, Inc., Cary, NC).

Results
A total of 1375 cases were reviewed, of which 148 cases were excluded due to the operative report not being available (n = 16), use of a burr hole (n = 98), or lack of a cranial operation (n = 34). The final cohort was comprised of 1209 cases.
We subsequently assessed whether patients with multiple risk factors in combination exhibited an increased infection rate compared to those with fewer risk factors. Although CSF leak was significantly associated with SSI (OR 7.03, p < 0.001), we herein classified it as a postoperative risk factor, excluding it form the combinatorial analysis of potential pre-and intraoperative risk factors.
Patients with one or more risk factors exhibited an infection rate of 5.9% and were 2.74 times more likely to develop an infection than patients without risk factors (p = 0.001). Patients with two or more risk factors had an infection rate of 6.7% and were 2.28 times more likely to develop and infection compared to patients with either one or no risk factors (p = 0.01). Patients with three or more risk factors had an infection rate of 19.0% and were 6.5 times more likely to develop an infection than patients with two or fewer risk factors (p < 0.0001). Patients with 4 or more risk factors had a 100% infection rate and were 30.2 times more likely to develop an infection compared to patients with 3 or fewer risk factors (p < 0.0001) ( Table 3; Fig. 1). Additionally, patients with ≥ 1 risk factor exhibited a significantly shorter time to SSI occurrence compared to patients with no risk factors (29.9 ± 13.5 vs. 44.2 ± 22.3 days, p = 0.02). No differences in timing of SSI occurrence were observed between the other risk factor groups (p > 0.05).

Discussion
Precisely defining preoperative and intraoperative risk factors for wound infection in neurosurgical patients is crucial for mitigating associated health care expenditures, developing individualized prevention strategies, and maximizing patient quality of life [2,4]. However, prior literature has primarily focused on individual demographics, comorbidities, and perioperative characteristics, without considering the influence of multiple risk factors in combination. The present study is the first, to our knowledge, investigating the relative increase in risk of SSI caused by the presence of more than one risk factor. Notwithstanding etiology, we herein show that patients with a greater number of risk  of ventricular opening, foreign body placement, and utility of CSF surveillance would be beneficial. Of note, current neurocritical care guidelines suggest no more than a single dose of prophylactic antibiotics at the time of drain or shunt insertion, as well as limiting postoperative antibiotic duration to < 24 h [7,9]. The association of bevacizumab with infection and wound breakdown is also well documented. Clark et al. show that patients receiving bevacizumab developed significantly more healing complications relative to non-bevacizumab treated patients [10]. Saran et al. report greater infection rates among glioblastoma patients treated with bevacizumab (54.4% vs. 39.1%), citing myelotoxicity as a likely underlying cause [11]. It is reasonable to posit that prior radiation and brachytherapy compromise healing via a similar mechanism [11][12][13]. The infection rate among our patients who received brachytherapy (n = 68) was 8.8%, which is notably higher than the overall cohort (3.5%).
Fang et al. [4] and Hussein et al. [14] both report CSF leak as the most significant risk factor observed for SSI. Our group determined that CSF leak was significantly associated with SSI (OR 7.03; p < 0.0001). However, CSF leak was not included in our current analysis as it is a postoperative risk factor that cannot be identified preoperatively. Literature investigating the underlying mechanism suggests that CSF leak leads to retrograde movement of bacteria through the peritubular gap or lumen [15]. Thus, while not beneficial for predicting SSI risk preoperatively, the need to prevent CSF leak through careful surgical planning and early definitive management is evident.
Opening of the ventricle, lumbar drain placement, prior infection, as well as comorbidities including hypertension, diabetes, additional non-neurologic cancer, hyperlipidemia/ hypercholesterolemia, and obesity were not significant independent risk factors for SSI in our cohort. Aside from the lack of literature investigating the additive risk of multiple risk factors, these findings also suggest that there is substantial variability in regard to the assessment of individual risk factors for SSI. We now show that the presence of more than one risk factor may confer an additive risk of postoperative SSI that is dependent on risk factor quantity alone. Overall, by assessing risk factors as composites rather than individual entities, health care professionals will have the capacity to better identify patients at greater risk for surgical site factors may have a higher incidence of infection due to risk factor quantity alone, and that this incidence increases as more risk factors are added. Our results also indicate that patients with ≥ 1 risk factors exhibit a significantly faster time to SSI compared to those with no risk factors for infection.
Variables associated with increased risk of infection in our study had some overlap but were not identical to those identified in prior literature. A systematic review and metaanalysis of 26 studies reported the following risk factors for SSI among craniotomy patients: CSF leak (OR 7.82), > 1 operation (OR 2.35), operation duration > 4 h (OR 1.77), venous sinus entry (OR 4.02), ASA score (OR 1.40), and male sex (OR 1.47). Age, operative site, antibiotic prophylaxis, ICP monitoring, foreign body placement, hypertension, and diabetes were not associated with increased risk of wound infection [4]. Importantly, these studies solely investigated the association of individual risk factors with SSI and did not account for any relative increase in risk caused by the presence of more than one risk factor.
Patients with a foreign body in our cohort exhibited 4-times greater odds of postoperative infection. Buchanan et al. likewise found placement of a foreign body to be a significant predictor of 30-day readmission for SSI (OR 1.47; p = 0.02) [7]. The presence of external ventricular drains or ventricular shunts has been shown to predispose neurosurgical patients to colonization and subsequent infiltration of bacteria into the CSF [8]. Ventricular entry intraoperatively did not increase the risk of infection among our patients. Further prospective studies exploring the duration   infections and tailor care accordingly through direct patient counseling, optimization of antimicrobial prophylaxis dosing, and augmented surveillance for high-risk patient populations [16][17][18]. The study has several limitations. The retrospective, single-center design reduces the level of evidence and generalizability of our findings. Additionally, the limited sample size of patients with each potential combination of risk factors required us to investigate the additive effect of risk factor quantity overall, rather than particular risk factor etiologies. In the present analysis, patients were not distinguished by type of prior radiation. However, a prior study by Potter et al. found that intensity-modulated radiation therapy (IMRT) was a significant risk factor for SSI, while stereotactic radiosurgery (SRS) was not [19]. As such, future studies are needed to further investigate the differential impact of IMRT and SRS on SSI when present in combination with other risk factors. Other potential risk factors such as smoking and entry of the frontal sinus were not herein assessed, and likewise warrant further investigation.

Conclusions
To the best of our knowledge, this study is the first to investigate the combinatorial influence of risk factors on SSI. Our results suggest that risk factors in aggregate incrementally increase the risk of post-operative SSI. Additional multicenter prospective studies investigating the combinatorial influence of risk factors on SSI are needed to create more effective risk stratification systems that can individualize prevention strategies and operative planning among patients undergoing craniotomy for neuro-oncologic pathology.