Serial Neuroendoscopic Lavage for the Treatment of Elevated Cerebrospinal Fluid Protein Levels in Infants with Gram-Negative Rod Ventriculitis

Gram-negative rod (GNR) bacterial shunt infection is a severe complication of shunt-dependent hydrocephalus, often requiring an extended and invasive treatment course. Accumulation of proteinaceous exudate limits the circulation of antibiotics and prevents the clearance of microorganisms [1‒3], and may be the reason that shunt infection presents as shunt malfunction in 36–50% of cases [1, 4‒7]. The standard of care for GNR shunt infection requires shunt removal, placement of an external ventricular drain (EVD), and several weeks of intravenous antibiotic therapy. The shunt is replaced after a course of antibiotics, serial negative cerebrospinal fluid (CSF) cultures, and normalization of CSF protein levels [8]. Preliminary studies suggest that application of neuroendoscopic intraventricular lavage in the setting of GNR ventriculitis may significantly reduce CSF protein levels and bacterial burden (and thereby, infection duration) [9‒15], as well as reduce the incidence of post-infectious complications of shunt obstruction and multiloculated hydrocephalus requiring a complex shunt construct [12, 13, 16, 17]. This report describes five infants with shunt-dependent hydrocephalus and GNR ventriculitis, which were effectively managed with serial neuroendoscopic lavage in addition to standard care measures. CSF protein measurements are presented throughout hospitalization, as well as percent postoperative decrease in CSF protein. This series presents a novel application of this technique in a population of infants at high risk for morbidity and mortality associated with ventriculitis.

Five infants with shunt-dependent hydrocephalus and subsequent GNR ventriculitis, treated with standard care measures and the addition of serial neuroendoscopic lavage, were identified. A retrospective chart review was performed to collect patient characteristics, shunt dependency, and shunt revisions within a year of ventriculitis resolution (Table 1).

All patients presented with signs of shunt infection, including fever, shunt malfunction, and positive CSF cultures, and underwent ventriculoperitoneal (VP) shunt removal, EVD placement, and initiation of broad-spectrum antibiotics. Antibiotics were narrowed with consultation to the infectious disease service after an organism and its sensitivities to antibiotics were identified. Suspicion for ventriculitis was confirmed by CSF culture and MRI, demonstrating enhancement of the ependyma with or without intraventricular accumulation of purulent debris (see Fig. 1). CSF labs and cultures were obtained daily and monitored for persistently elevated protein, which served as a primary indication for intraventricular lavage. CSF cultures and studies were also performed on CSF samples collected at the start of surgery prior to lavage. CSF protein levels following lavage are reported as the first collection within 24 h. See Figure 2 for a summary of CSF protein and timing of neurosurgical intervention during hospitalization. Shunts were reinserted following resolution of symptoms, completion of antibiotic course, and negative CSF cultures, in consultation with the Infectious Disease Service. See Table 1 for a summary of patient characteristics and CSF protein measures before and after lavage.

The endoscopic technique employed for these cases included targeting the ventricle showing the greatest concentration of debris on preoperative MRI (see Fig. 1) with a rigid neuroendoscope (Clarus Medical LLC) for intraventricular lavage with warmed (37°C) lactated Ringer solution. The endoscope was used to irrigate, mobilize, and clear purulent material adherent to the ependyma. Lavage continued until the fluid returned clear and no further debris could be mobilized. An alligator rongeur was used where necessary to gently separate purulent debris from the ependyma and to fenestrate septations. Meticulous care was taken to avoid damage to the ependyma and choroid plexus.

This series describes five complex cases of GNR ventriculitis and secondary hyperproteinorrachia in infants with shunt-dependent hydrocephalus, managed effectively with neuroendoscopic lavage in addition to standard of care measures. Lavage was initiated for frank purulent debris visualized on MRI, significant hyperproteinorrachia that failed to normalize after a week of external ventricular drainage, or both. Endoscopic fenestration of septations and adhesions was utilized where possible to facilitate CSF circulation and simplify the eventual shunt construct needed. Patients had developed GNR ventriculitis between 2 and 37 days from their most recent neurosurgical intervention, and CSF cultures remained positive between 6 and 18 days after diagnosis of ventriculitis. See Table 1 for a summary of ventriculitis and lavage characteristics. Serial neuroendoscopic lavage produced a decline in CSF protein levels across the course of ventriculitis and post-infectious hyperproteinorrachia (Fig. 2), with a mean decrease in CSF protein of 74% (range 22–99%) following each lavage (Fig. 3). Shunts were reinserted after 69 ± 47 days of treatment for ventriculitis, following at least 2 weeks of negative CSF cultures and in the setting of largely normalized protein levels. Length of hospitalization was 80 ± 50 days.

This infant was referred to neurosurgery for suspicion of obstructive hydrocephalus. An MRI demonstrated aqueduct stenosis, and they were taken to the operating room under the same anesthetic for elective placement of a frontal VP shunt, which required a single revision roughly 2 months later. They presented postoperatively with signs of shunt malfunction. CSF culture revealed moderate GNRs, and empiric doses of cefepime, metronidazole and vancomycin were initiated. The VP shunt was removed, and an EVD was placed. Intraoperative CSF cultures identified Klebsiella oxytoca as the causative organism. Antibiotics were narrowed to ceftriaxone as sensitivities became available. Because of persistently positive cultures, the patient underwent MRI, which showed ependymal enhancement of both the lateral and third ventricles consistent with ventriculitis, as well as a small collection of purulent debris layering dependently in the occipital horn of the right lateral ventricle (see Fig. 1). Neuroendoscopic lavage was performed twice to address persistent hyperproteinorrachia of 759 mg/dL and 362 mg/dL, respectively, with good response and no acute complications. A frontal VP shunt was inserted following 2 weeks of negative CSF cultures, clearance by the Infectious Disease Service, and a CSF protein of 260 mg/dL. This patient required no additional hospitalizations or interventions within the follow-up period.

This infant was diagnosed with hydrocephalus prenatally, and a postnatal MRI revealed aqueduct stenosis. They underwent placement of a frontal VP shunt in the days after their term delivery. Three weeks later, they presented with fever, a full anterior fontanelle, and pseudomeningocele over the VP shunt valve. The shunt system was removed and sent for culture, and an EVD was placed. MRI demonstrated diffuse enhancement of the ependyma, consistent with ventriculitis. CSF cultures showed GNRs, and empiric doses of vancomycin, cefepime, and metronidazole were initiated. Antibiotics were narrowed to meropenem and intrathecal gentamicin at the direction of the Infectious Disease Service after identification of the organism as Serratia marcescens. Neuroendoscopic lavage was performed a total of eight times, every 7–10 days beginning in the third week of hospitalization, to mitigate persistently elevated CSF protein levels (>1,000 mg/dL) despite apparent infection resolution (Fig. 1). Repeat imaging after four lavages showed persistent ependymal enhancement and purulent debris within the lateral ventricles. Of note, daily CSF labs were not acquired for a period of time as the patient’s EVD was removed due to scant CSF production; CSF samples were only obtained immediately pre-lavage upon placement of the endoscope into the ventricle. A temporal VP shunt was placed prior to discharge, in the setting of normalized CSF protein of 198 mg/dL. The patient required placement of a contralateral temporal VP shunt to address an enlarging loculated temporal horn 3 weeks after discharge. This patient required two shunt revisions within 1-year of discharge. Four months after insertion of this shunt, the patient developed a methicillin-sensitive Staphylococcus aureus shunt infection and underwent removal of the bilateral VP shunts with bilateral EVD placement. The both VP shunts were replaced prior to hospital discharge. This patient required one revision of his left VP shunt 7 months after this infection resolution. This patient has not required any additional neurosurgical interventions in the interim follow-up period.

This infant was diagnosed with hydrocephalus prenatally, and a frontal VP shunt was placed in the days after birth. Follow-up imaging revealed large bilateral acute-on-chronic subdural hematomas, which required surgical evacuation. These were deemed secondary to the patient’s scant parenchyma and massive macrocephaly and not related to non-accidental or accidental trauma. This patient presented with seizures, fever, and an incisional CSF leak 1 month after subdural hematoma evacuation. Imaging showed ependymal enhancement, suggestive of ventriculitis. The VP shunt was removed, and EVD was placed. Intraoperative CSF cultures were positive for GNRs, and empiric doses of metronidazole, cefepime, and vancomycin were initiated. The organism was identified as Escherichia coli, and antibiotics were changed to ceftriaxone and meropenem. Neuroendoscopic lavage was performed four times over the course of 2 months to mitigate persistent hyperproteinorrachia. A right frontal VP shunt was placed in the context of relatively normalized CSF protein levels (400 mg/dL) and negative CSF cultures. This patient required no revisions or other neurosurgical intervention and developed no additional infections. Years later, this patient unfortunately expired during a hospitalization for non-neurologic complications, and life support was ultimately withdrawn.

This very preterm infant (born less than 32 weeks) developed grade IV post-hemorrhagic hydrocephalus. Due to low birth weight and extreme prematurity, a subgaleal shunt was initially placed, and this was converted to a VP shunt once an adequate body weight was obtained. The shunt required revision due to a malfunction 1 year after insertion. Shortly after, the patient presented with symptoms of increased intracranial pressure, and imaging demonstrated progressive ventriculomegaly. The shunt was removed, and an EVD was placed. Culture of intraoperative CSF grew Pseudomonas aeruginosa. Ventriculitis was treated with gentamicin, cefepime, and meropenem following completion of cefepime course. Neuroimaging after 1 week of antibiotic initiation was consistent with ventriculitis, demonstrating purulent debris in the right lateral ventricle, left occipital horn, and fourth ventricle (see Fig. 1). CSF protein levels remained extremely high (>1,000 mg/dL) throughout hospitalization, and serial neuroendoscopic lavage was performed a total of five times to mitigate this. A VP shunt was placed after a final ventricular lavage. CSF protein levels remained elevated prior to shunt insertion at 1,659 mg/dL but improved postoperatively (205 mg/dL). This patient required one neurosurgical intervention within 1 year of follow-up, including and endoscopic fenestration of a loculated third ventricle with placement of a separate temporal VP shunt into the temporal horn to address progressive enlargement. This patient has not required any additional interventions.

This infant was born extremely preterm and developed hydrocephalus following grade IV intraventricular hemorrhage. Postnatally, a frontal subgaleal shunt was placed. At 1 year of age, endoscopic third ventriculostomy with choroid plexus cauterization (ETV-CPC) was performed. The ETV was found to not be patent at 1 month follow-up, and a right ventriculosagittal sinus shunt (VSS) was placed. A peritoneal, pleural, or arterial shunt terminus was contraindicated in this patient due to coexisting pathology. The VSS was converted to a VP shunt after the abdominal pathology had been addressed; however, the patient developed fever and symptoms of shunt malfunction on postoperative day two. Neuroimaging revealed diffuse enhancement of the lateral and third ventricular walls, indicative of ventriculitis. CSF culture via shunt tap revealed GNRs, and empiric treatment with cefepime, vancomycin, and metronidazole was initiated. The shunt was removed and EVD was placed. The infectious organism was identified to be extended-spectrum beta-lactamase (ESBL)-producing Enterobacter cloacae. The antibiotics were changed to meropenem and vancomycin at the discretion of the Infectious Disease Service. Neuroendoscopic lavage was performed a total of seven times to facilitate infection clearance and mitigate hyperproteinorrachia. A frontal VP shunt was implanted following resolution of symptoms and serially negative CSF cultures. Preoperative CSF protein at the time of final lavage and VP shunt reinsertion was 1,679 mg/dL. The patient required shunt removal after postoperative CSF cultures demonstrated ESBL Enterococcus following fenestration of the temporal horn. The VP shunt was never replaced, and ultimately the prior ETV-CPC, along with presumed scarring from prolonged and recurrent infection, was evidently adequate treatment for their hydrocephalus. This patient required no further neurosurgical intervention.

This case series supports a growing body of literature suggesting that proper timing of neuroendoscopic lavage, in combination with lysis of septations, may be key in resolving infection, evacuating purulent debris, and reducing infectious and post-infectious hyperproteinorrachia associated with ventriculitis [11]. A prospective case-control study by Al Menabbawy et al. [15] (2020) in 33 pediatric patients with cerebral ventriculitis found that ventricular lavage was associated with better outcomes and shorter hospital stays compared to conservative care. Similarly, Gaderer et al. [13] report a higher incidence of shunt revision and reinfection in patients treated with standard care compared to patients with neuroendoscopic lavage added as supplementary treatment. In the presented cases, CSF protein levels decreased an average of 74% following each lavage (Fig. 3), which is consistent with previous reports [12, 13]. Unique to the cases reported here, serial CSF protein studies revealed that hyperproteinorrachia persisted beyond the resolution of active infection in 3 of the 5 cases (cases 1–3), and serial lavage was used to normalize CSF protein prior to shunt reinsertion. Persistence of high CSF protein despite resolution of microbial burden is thought to reflect active host inflammatory response following infection resolution [18] and may differ between Gram-positive and Gram-negative shunt infections [19]. Figure 2 demonstrates widely variable patterns of CSF protein values among patients in reference to reported age-referenced ranges of typical CSF protein levels [20]. Furthermore, there is discordance in the literature regarding the appropriate timing of neuroendoscopic lavage in the course of infection. While some studies recommend neuroendoscopic lavage be performed after 2 weeks of antibiotic treatment with unsatisfactory results [17, 21], others suggest lavage be pursued immediately after diagnosis of ventriculitis to improve the circulation of antibiotics [12]. Future studies should seek to identify clinical markers for the timing of initiation of treatment with serial lavage, as well as timing between treatments, to properly assess its utility.

Considerations for VP shunt reinsertion after infection resolution included serial negative CSF cultures and CSF protein levels. Here age-based reference for normal CSF protein of 170 +/− 60 mg/dL were based on published ranges for the 4-month to 14-year-old age range using the corrected age for premature birth in patients 5 and 6 (gray plot area Fig. 2) [20]. However, further research is needed to define pre-operative CSF protein targets for lavage and target protein levels for VP shunt reinsertion. Research suggests that premature reimplantation, while CSF protein levels are still elevated, may correlate with a higher likelihood of subsequent shunt failure [15]. Assessing the best timing for reinsertion based on CSF infectious markers merits consideration of the age-related changes in CSF protein of healthy pediatric patients, which decrease dramatically across the first 4 months of life and stabilize as patients reach adulthood. Published reference ranges include CSF protein levels of 590 ± 210 mg/dL for patients younger than 2 months, 370 ± 150 mg/dL between 2 and 4 months, 170 ± 60 mg/dL between 4 months and 14 years, and 260 ± 60 mg/dL between 14 and 18 years [20]. Work by Yakut and colleagues in children with ventriculitis ages 1–204 months suggests that a CSF protein level greater than 100 mg/dL prior to shunt reinsertion is a risk factor for shunt failure and reinfection [22]; others report successful VP shunt reinsertion 1 week after negative CSF cultures regardless of protein levels [15]. Although no concrete guidelines exist, 200 mg/dL was considered a target for VP shunt reinsertion in the cases presented here, with consideration for age-related variation. Higher CSF protein concentrations were accepted if the patient was otherwise optimized for shunt reinsertion. All clinical markers of infection resolution should be considered in the interest of decreasing total hospitalization time. Future studies are needed to identify CSF protein targets for the timing of shunt reinsertion that may reduce the likelihood of shunt failure.

This report supports the utility of serial neuroendoscopic lavage in neonatal and infant patients with ventriculitis, in line with prior studies in children. See Table 2 for a summary of studies using neuroendoscopic lavage to treat pediatric patients with intraventricular empyema. These studies differ from the one presented here in age at time of ventriculitis, infectious organism, and ventriculitis etiology. Al Menabbawy et al. [15] (2020) present data from the first prospective cohort of ventricular irrigation for cerebral ventriculitis in 33 children and adolescents (n = 27 related to shunt infection), although it is difficult to compare their results and lavage protocol to those presented here given the heterogeneous sample of organisms and ventriculitis etiologies, in addition to lack of reported CSF laboratory studies, intraoperative irrigation volume, and timing of infection clearance and shunt reinsertion. Ventriculitis developed secondary to shunt infection, surgery, or meningitis, and was caused by a variety of infectious fungi and Gram-positive or -negative organisms, including Candida albicans, Acinetobacter, Pseudomonas, Klebsiella, MRSA, and coagulase-negative Staphylococcus. The decision to perform lavage was based on CSF leukocyte cell count >100/mL. Al Menabbawy et al. [15] did not find a significant difference in mortality rate between the lavage and the conventional management groups (25% vs. 53%). Here, a 0% mortality rate in a population of infants at high risk for morbidity and mortality associated with ventriculitis is reported. The Al Menabbawy et al. [15] study did find better outcomes in the lavage group, defined as a modified Rankin Score of 3 or less at follow-up, but they do not present follow-up data regarding need for additional surgical intervention for shunt revision or replacement for comparison to the data present in the current study.

Compared to other studies in similar populations, the technique presented here trends toward more rapid clearance of infection, although it is difficult to compare directly given the heterogenicity of samples and lack of standard data reporting. Gaderer et al. [13] (2018) present a similar population and lavage protocol, with use of 1–4 endoscopic lavages and an average of 33 days to negative CSF cultures, compared to 2–8 lavages and 12 days to negative cultures presented here, and similar length of hospitalization (average 80 vs. 84 days). Al Menabbawy et al. [15] (2020) report an average hospitalization of 21 days for children in the lavage treatment group (average age 6 years). Time to shunt reimplantation was partially dictated by normalization of CSF protein as part of ventriculitis resolution and application of serial lavage to speed this occurrence. This may explain the longer average days to VP shunt reported in the current study as compared to Gaderer et al. [13] (60 vs. 29 days). Within a year of follow-up, 60% (N = 3) of the patients in the current study required one revision of the shunt placed at the time of infection resolution, and 60% required placement of an additional shunt catheter into a loculated fluid collection. Gaderer et al. [13] report that 22% of patients required surgical revision in 2 years of follow-up, compared to 85% of control patients. While all patients in this study were discharged with VP shunt following infection, including 1 patient with ETV during treatment, 2 patients were ultimately treated successfully with ETV within the follow-up period and remained shunt-free thereafter. This suggests that serial neuroendoscopic lavage may be more useful in clearing infectious and post-infectious intraventricular debris than a single lavage. Publication of additional studies of similar patient populations (infants) will enable the standardization of the techniques presented here and determine the efficacy of serial versus single endoscopic lavage.

A potential alternative to the proposed approach may be similar to the technique employed in the drainage, irrigation, and fibrinolytic therapy (DRIFT) study, which treated progressive ventricular dilatation after intraventricular hemorrhage in preterm infants with intraventricular injection of fibrinolytic, followed by continuous lavage until the drainage was clear. The technique included continuous irrigation (for a median of 72 h) at a rate of 20 mL/h through a ventricular reservoir in the front of one lateral ventricle and an EVD in the back of the contralateral ventricle [23]. The protocol also included therapeutic concentrations of gentamicin and vancomycin to the irrigation fluid. This is certainly a potential complementary treatment or alternative to performing serial neuroendoscopic procedures and warrants further consideration. However, the benefit of neuroendoscopy in ventriculitis may be lost if this alternative is employed, including the ability to address formation of septations and lysis of adhesions to the ependyma, which may limit the circulation of CSF and antibiotics. Additionally, whereas the DRIFT protocol requires an additional burr hole and intracranial catheter, NEL can be performed through the burr hole in which the external ventricular drain for the treatment of shunt-associated ventriculitis is already placed.

The potential risks posed by neuroendoscopic lavage must be acknowledged, including repeated exposure to anesthesia, wound breakdown due to repeated opening of an incision, hemorrhage, and the potential introduction of additional infectious organisms. It is imperative that the risks of surgical intervention are weighed against the proposed benefits of serial neuroendoscopic lavage for the effective treatment of GNR ventriculitis in infants with hydrocephalus. Future studies should examine the cost-benefit and risk-benefit ratios of this procedure in comparison to standard prolonged external ventricular drainage. Additionally, there are limitations to interpreting the observations of this report due to its small sample size and lack of control group.

GNR ventriculitis is a rare but severe complication of shunt-dependent hydrocephalus for pediatric patients. Supplementation of standard care with serial neuroendoscopic lavage may alleviate the acute manifestations of infection in the CSF, including organisms, infectious debris, and inflammatory cells; minimize length of hospitalization and time to infection clearance; as well as treat persistent hyperproteinorrachia following resolution of severe ventriculitis [9, 10, 12, 13, 16, 17]. The ultimate shunt system required may be simplified by the fenestration of loculations during endoscopic lavage. In conclusion, neuroendoscopic lavage may facilitate the extirpation of purulent debris and lysis of septations forming the infectious pseudocapsule, effectively reducing CSF protein and expediting the time to shunt reimplantation. While this technique may be associated with fewer future shunt revisions, the added risk of multiple neuroendoscopic procedures to facilitate normalization of CSF protein prior to shunt reinsertion must be considered, and the risk-benefit analysis of this type of intervention warrants closer study to determine safety and efficacy.

This study protocol was reviewed and approved by the University of Pittsburgh Institutional Review Board, approval number 20050395. This study protocol has been granted an exemption from requiring written informed consent to publish the details of the medical cases. All directly and indirectly identifiable information has been removed or anonymized.

The authors do not disclose any real or perceived conflicts of interest.

The authors have no funding sources to disclose.

R.K.S., S.G., K.W.N., J.K., and J.L.H. contributed to the concept and design of this report. J.L.H. acquired and analyzed data, wrote the manuscript, and generated figures. R.K.S., K.W.N., J.K., and S.G critically edited the manuscript. K.W.N. performed subset of analyses and figure generation. All authors approve of the manuscript in its current form and agree to be accountable for all aspects of the work.

All data generated or analyzed during this study are included in this article. Further inquiries can be directed to the corresponding author.