Identification of Patients

This study took place at the University of Crete. Patients were eligible if they were at least 18 years of age, underwent elective or emergency craniotomy between 2006 and 2008, and survived at least 7 days after surgery. Major craniotomies were included. Patients having only CSF shunt or external ventricular drain (EVD) implantations, bur hole trepanation, or stereotactic surgery were excluded. The patients were prospectively followed for the development of meningitis during the first 30 postoperative days and during the 1st postoperative year if a foreign body was implanted. It was assumed that the occurrence of meningitis would result in readmission; although it is possible that admission to another institution may have resulted in some missed cases, it is highly unlikely.

Data Abstraction

Data were abstracted from the medical charts to a standard database. We recorded comorbidities such as diabetes, malignancy, atherosclerotic cardiovascular disease, chronic renal failure, chronic obstructive pulmonary disease, and perioperative steroid use; prophylactic antibiotic use; characteristics of the procedure performed, such as procedure urgency (elective or emergent), surgery duration, multiple procedures performed at the same time (simultaneous orthopedic or abdominal surgery or facial reconstruction); and the presence of any postoperative CSF drainage, reoperations, or CSF leaks. An operation was considered a revision surgery whenever it was performed through the same incision as a previous operation, regardless of the interval between the 2 procedures. Postoperative CSF leak was recorded when CSF drainage was diagnosed after surgery on the basis of otorrhea, rhinorrhea, or leakage from the surgical wound.

Meningitis was diagnosed according to the definitions of the Centers for Disease Control.⁸ The definition of the disease was as follows: 1) organisms cultured from CSF, 2) at least 1 of the following signs or symptoms with no other recognized cause: fever (> 38°C), headache, stiff neck, meningeal signs, cranial nerve signs, or irritability and if diagnosis was made antemortem, attending physician instituted appropriate antimicrobial therapy, and at least one of the following: a) increased WBC count, elevated protein level, and/or decreased glucose level in CSF; b) organisms seen on gram stain of CSF; c) organisms cultured from blood; d) positive antigen test of CSF, blood, or urine; e) diagnostic single antibody titer (IgM) or 4-fold increase in paired sera (IgG) for pathogen. We recorded the postoperative day on which the diagnostic lumbar puncture was performed, the organisms identified by CSF culture, and all the infections (ventilator-associated pneumonia [VAP], urinary tract infection [UTI], pneumonia, bacteremia) that developed in this cohort. The data were transferred to a database and were analyzed using SPSS software (Version 16.0, SPSS Inc.). Continuous variables were compared using the Student t-test or the Mann-Whitney U-test, whereas categorical variables were compared using the Fisher exact test. In univariate analysis, odds ratios and 95% confidence intervals were calculated using the Mantel-Haenszel test. Stepwise multivariate logistic regression was used to model the interactions of those variables significantly associated with surgical site infections (SSIs) or meningitis in univariate analyses. A backward elimination model was used with P_enter = 0.20 and P_leave = 0.05 to identify independent predictors for meningitis. When a continuous variable such as the number of days of ventricular drainage, was shown to be a statistically significant risk factor, the relationship was modeled further using suitably coded dummy variables. Power calculations were performed using the PASS software (PASS 2008, https://www.ncss.com).

Study Population and Procedure Characteristics

The study included 334 procedures in 239 patients who met the inclusion criteria; 65.6% of the procedures were performed in men. Meningitis developed after 16 (4.8%) of the procedures. The median age was 48 years for the patients who developed meningitis (interquartile range [IQR] 22–58 years) and 51 (IQR 27–67 years) for those who did not. Traumatic brain injury was the most common reason for craniotomy (accounting for 49.7% of the procedures). The rest of the craniotomy indications included space-occupying lesions (in 34% of cases), vascular conditions (10%), abscess (1%), and a variety of other reasons (3%); 18.6% of the procedures were revision surgeries, and 49.7% were emergent procedures. Only 3.3% of the patients had concurrent procedures. Almost 40% of the patients who underwent craniotomy developed at least 1 infection. SSIs developed after 9% of the procedures. The overall mortality rate was 15%. There was no statistically significant difference in mortality rate between patients who did and did not develop meningitis. The median time from admission to surgery was 3 days. The median duration of surgery was 2 hours (IQR 2.0–4.0 hours) for the procedures after which meningitis did not develop and 4 hours (IQR 3.0–6.0 hours) for those after which it did. The median length of stay (LOS) was 27 days (IQR 12–79). Patients who developed meningitis had a much longer LOS (median 191 vs 24.5 days, p < 0.001). Patients who developed meningitis had a greater propensity to be admitted to the intensive care unit (ICU) (p < 0.001) and a prolonged intubation beyond the time of surgery (p < 0.001).

Meningitis Risk

The median time from surgery to lumbar puncture was 6.5 days in those cases in which meningitis was diagnosed. Table 1 lists all variables that were examined for their contribution to the risk of meningitis. Individual variables that were statistically significant at a p value less than 0.05 included female sex, perioperative steroid use, increased duration of surgery, presence of CSF leak, ICU admission, use of a central line, and use of ventricular drains. The presence of drains overall was not associated with the development of meningitis. The presence of another SSI was associated with meningitis risk (p < 0.001). Among the patients who had an SSI other than meningitis, 41% also developed meningitis. Emergency status of the procedure (i.e., emergent or not) was not associated with the development of meningitis. “Time to surgery” and the presence of skull fractures also were not associated with the development of meningitis. Overall, 14 patients (4%) had open wound trauma. Patients with open wound trauma were more likely to develop an SSI than those without (36% vs 11%, p = 0.02). Also, openwound trauma patients had a higher risk for meningitis (14% vs 5%), but this association failed to reach statistical significance (p = 0.16). Crossing a sinus during a procedure was not associated with the development of meningitis (p = 0.23).

Variables that were statistically significant at p ≤ 0.20 in the univariate analysis were included in a multivariate regression (Table 2). Presence of ventricular drainage (OR 70.52), CSF leak (OR 48.03), and perioperative steroid use (OR 11.55) remained in the model as independent predictors with p < 0.05.

CSF Features in Patients With Meningitis

Postoperative meningitis developed in 16 cases. The characteristics of the CSF in these cases are shown in Table 3.

Microbiology of Meningitis

Meningitis was documented by positive CSF culture in all 16 cases. Gram-negative organisms predominated. Acinetobacter spp. were the main isolates (present in 45% of cases). The other pathogens included: Klebsiella spp. (4 cases), Pseudomonas aeruginosa (2 cases), Enterobacter cloaceae (1 case), Proteus mirabilis (1 case), coagulasenegative staphylococci (1 case), Staphylococcus aureus (1 case), and Candida albicans (1 case). Acinetobacter spp. were fully resistant to β-lactams, aztreonam, aminoglycosides, and quinolones but they were fully sensitive to colistin. Klebsiella spp. were 100% resistant to β-lactams, cephalosporins, aztreonam, and quinolones. Pseudomonas aeruginosa isolates were fully resistant to the carbapenems, but they retained full susceptibility to ticarcillin/ clavulanate, piperacillin/tazobactam, cefepime, aztreonam, amikacin, gentamicin, colistin, and ciprofloxacin (Table 4).

Infections Other Than Meningitis

Ventilator-associated pneumonia was the most common infection in this population (22.5%) with 9% of patients developing an SSI other than meningitis, 9% a UTI, and 15% pneumonia not associated with mechanical ventilation. Among SSIs, wound infections predominated (75%). Postsurgical subdural and epidural abscesses represented only 11% of the SSIs. Ventriculostomy site, bone cement and bone infections, and subdural empyemas were rare, representing a total of 12% of the SSIs.

When the meningitis pathogens were compared with the catheter-associated bacteremia/blood stream infection (CAB/BSI) isolates, in 3 cases the same pathogens grew from the CSF and the blood (namely E. cloacae, Acinetobacter baumannii, and Klebsiella pneumoniae).

Microbiology of Other Infections

In VAPs, Acinetobacter spp. were the main pathogens, isolated in 44% of the cases. Pseudomonas aeruginosa was the second most prevalent isolate (25%). The resistance patterns are shown in Table 4. Of note, the Acinetobacter spp. retained full susceptibility to colistin. The resistance to imipenem approached 67%. Pseudomonas aeruginosa isolates were also fully susceptible to colistin. In UTIs, Klebsiella spp. were the most prevalent pathogens (31%). The isolates were 100% resistant to β-lactams, aztreonam, and quinolones and 75% resistant to imipenem and colistin.

In SSIs, Acinetobacter spp. predominated, with Klebsiella spp. being the second in prevalence. Gram-positive pathogens were the third most common category. In CAB/BSI, coagulase-negative staphylococci were the most prevalent pathogens. Gram-positive organisms generally predominated in these infections. Among the gram-negative organisms, A. baumanii was the most common.

Survival Analysis

After a cumulative follow up of 18,795 days (median 27 days, range 1–318 days), 50 patients had died (15%)— among them, 5 (29%) of the 16 patients who had meningitis, 1 (4%) of the 23 who had an SSI other than meningitis, 24 (24%) of the 98 who had other nosocomial infections, and 20 (10%) of the 194 who had no infection. Estimated survival rates at Day 30 were 0.94 (95% CI 0.65–0.99) for patients with meningitis, 1.0 for patients with other SSIs, 0.84 (95% CI 0.74–0.90) for patients with other nosocomial infections, and 0.90 (95% CI 0.84–0.94) for those without infections. The corresponding age- and sex-adjusted hazard ratios for patients with infection compared with those without infection were 0.80 (95% CI 0.22–2.84) for meningitis, 0.38 (95% CI 0.05–2.89) for other SSIs, and 1.43 (95% CI 0.77–2.63) for other nosocomial infections. The results of all between-group comparisons were also insignificant. Only age predicted an adverse outcome (HR for death 1.03 [95% CI 1.02–1.05] per year increase in age). In conclusion, there was no statistically significance difference between the study groups with respect to mortality rates (Fig. 1, Table 5).

Kaplan-Meier survival estimates for the patients undergoing craniotomy.

Prevalence and Microbiology of Post-Craniotomy Meningitis

The incidence of meningitis in this study was 4.8%, higher than incidences noted in some series^2,12,13 but not in all. Reichert et al.²⁸ in their prospective study reported a rate of 8.9%. In a recent retrospective study from the US, the incidence of post-craniotomy meningitis (PCM) was 5.5%.¹⁴ Gram-negative organisms predominated as meningitis pathogens in this prospective cohort. In studies published after 1993, Enterobacteriaceae and other gramnegative rods played a major role.^1,6,9,18,28 Our results seem to reflect this trend. In this cohort there were no culture-negative cases, which can make treatment decisions very difficult, and the cases were 100% culture proven.

Acinetobacter spp. were isolated in 45% of cases, with Klebsiella spp. and P. aeruginosa the next most common isolates. The increase in the rate of Acinetobacter PCM has been extensively described in the years following 2000, although in the previous years it was reported as a rare occurrence.^{4,5,18,22,28,32,36} The main issue with the Acinetobacter spp. is their increasing resistance to almost all antibiotic classes (including carbapenems), which has been a significant problem in the recent years.^{1,4,10,16,22,25,30,36} Klebsiella spp. were the second most commonly isolated pathogens in this cohort, with the isolates carrying a high degree of resistance to third-generation cephalosporins and carbapenems (Table 4). One hundred percent of the isolates were sensitive to tetracycline, but during the study period the sensitivity to tigecycline was not routinely recorded.

The median CSF glucose/serum glucose ratio was 0.40 (IQR 0.40–0.47). Since all the cases were culturepositive, it is safe to assume that this rate is very suggestive of microbial meningitis as it has been reported previously.¹⁶ The 100% culture positivity rate is a rare occurrence in the literature. A positivity rate of 32% was reported in another prospective study.²⁸ Retrospective studies by Kourbeti et al. have demonstrated a positivity rate ranging from 50%¹⁴ to 74% (unpublished data). In those studies there were not any differences in the CSF characteristics between culture-positive and culture-negative cases.¹⁴ The mean duration from operation to the onset of symptoms was 6.5 days, an interval similar to that reported before.^14,15,28

Risk Factors

The independent risk factors identified by the multivariate analysis were use of perioperative steroids, CSF leak, and the use of EVDs.

Craniotomies with an EVD carry a higher risk for meningitis. ^{15,19,20,24,28} Ventriculitis in association with ventricular drain placement was a frequent problem and was observed in 11% of general neurosurgical patients in the series by Mayhall et al.²⁰ There is little agreement among various series regarding the relation between duration of drainage and infection incidence; in one report, no relation was observed. ¹⁹ Lozier et al.¹⁷ have commented on the controversy regarding the significance of the duration of EVD use with respect to the development of infections. At the University of Crete there were no routine changes of EVDs. Leaving an uninfected drain in place may be the best policy.¹⁵ In our institution, dressings were changed daily, the catheters were replaced at the first sign of infection or malfunction. Antibiotic-coated catheters were not routinely placed; we used them after the first catheter change for any reason or in patients who already had signs of infection and needed an EVD placed because of intraventricular hemorrhage or obstructive hydrocephalus.

Previous studies demonstrated postoperative CSF leak to be a significant risk factor for the development of meningitis.^{2,11,12,23,26} In our series, the presence of CSF leak was associated with a 48-fold increase in the risk of meningitis, and this was statistically significant. This is much higher than the risk we had identified in a previous retrospective study,¹⁵ in which the CSF leak increased the risk of meningitis by a factor of 1.7 but was not a statistically significant risk factor. Surgeons should be very vigilant in correcting a CSF leak. When a patient with a post-craniotomy CSF leak develops a fever, the possibility of meningitis should be strongly considered.

Meningitis or positive CSF culture has been described as the most frequent complication of intracranial pressure (ICP) monitoring.²⁷ In this study, however, the presence of an ICP monitoring device did not affect the development of meningitis.

In contrast to findings of a previous prospective study,²⁶ in our study duration of surgery did not remain an independent risk factor in the multivariate analysis.²⁶ Age, sex, and diabetes were not significant risk factors for the development of meningitis, a finding consistent with previously reported data.¹² Repeat surgery, which has been a major risk factor in recent²⁸ and older²⁶ reports, was not a risk factor in our cohort. Scheduling the procedure as an emergency or not an emergency was not a risk factor for PCM, a finding that we had described before.¹⁵ Entering a sinus was not a risk factor in this prospective study, in contrast to our retrospective report,¹⁵ and the reports of others.^11,23

The use of perioperative steroids achieved a statistical significance in this prospective study. This has not been described before in studies regarding PCM and in most studies regarding postoperative wound infections in general neurosurgical patients.^{4,5,12–15,23,26,28,33} The effect of steroids on patients undergoing neurosurgical procedures has been variably reported.²¹ In certain studies, the steroid effect has not been specifically investigated.^4,12,13 Walcott et al.³⁵ mention that for rare events such as infections in neurosurgery, it may be difficult to accomplish proof for the association of certain variables. In the study by Erman et al.,⁵ perioperative steroid use did not remain an independent risk factor for the development of SSIs in the multivariate analysis. In the prospective study by Reichert et al.²⁸ and the retrospective study by Kourbeti et al.¹⁵ on the risk factors associated with PCM, steroid administration did not reach a statistical significance. In studies in which steroids were administered to patients undergoing implantation of subdural electrodes, steroids had no effect on the incidence of infection.^7,34 McPhee et al.²¹ demonstrated a significant increase in postoperative wound infection and breakdown with steroid use in the setting of spinal operations for metastases, but they did not mention the effect on the development of meningitis.²¹ Shinoura et al.³¹ did not find that steroid use was a significant factor in a population of patients undergoing craniotomy for brain tumors. In summary, this is a matter that is highly controversial in the literature. Well-designed prospective studies will be needed in order to define the safe period for steroid use in neurosurgical patients.