Characteristics and management of ventricular shunt infections in children, 2000-2015: a single centre retrospective chart review

Introduction

Hydrocephalus is a common and serious neurosurgical condition. The incidence is estimated between 0.4-4 cases per 1000 births worldwide^1,2. In some cases there is a treatable underlying cause of hydrocephalus (e.g. a posterior fossa tumour) but in most cases cerebrospinal fluid (CSF) diversion is required. CSF diversion usually requires some kind of shunt, although endoscopic third ventriculostomy is an option in certain types of obstructive hydrocephalus. Different types of shunts are available, of which ventriculoperitoneal (VP) shunts are most frequently used. Other types of shunts include; ventriculoatrial (VA), ventriculopleural and lumboperitoneal (LP)^2–5.

Shunt insertion procedures are not without risk and infection is a frequent complication^6,7. Rates of infection per shunt vary widely, between 3 and 13%^3,6,8–12. Overall shunt infection rate in the UK is estimated at 5.2%⁷. Shunt infections have a mortality of 10.1% and worse Glasgow Outcome Scale scores, as well as worse school performance in the long-term¹³. From a public health perspective, shunt infections are associated with major hospital costs, with one episode costing approximately £29,000¹⁴. Therefore, prevention or prompt recognition and treatment of infection are imperative.

There are recognised factors known to increase the risk of shunt infection. In general, younger and premature children have a higher risk due to immaturity of their immune system and differences in skin flora compared to older children¹⁵. Premature infants who have had shunt surgery before 40 weeks gestational age have the greatest risk of all ages (Hazard radio (HR): 4.72, 95% CI 1.71-13.06)^9,10,14,16. Children who have multiple shunt revision procedures, either because of shunt blockage, malfunction, fracture or infection are also at greater risk as with each revision, the cumulative risk rises¹¹. There are technical aspects of shunt surgery that increase risk, including handling of the shunt system with sterile surgical gloves contaminated early in the procedure (HR 1.07, 95% CI 1.02-1.12, p=0.01). The risk is higher still if there is a post-operative CSF leak through the wound, allowing skin flora to access the shunt system (HR 19.16, 95% CI 6.96-52.91, P<0.0001)^4,16.

Recognition of shunt infections in children can be difficult as they present with a wide variety of non-specific symptoms such as fever, vomiting, drowsiness, headaches, irritability, seizures and abdominal pain^3,10,17. Up to 90% of shunt infections occur within the first three to nine months after shunt insertion. During this period there should be a low threshold for investigating patients who display any of these symptoms^3,4,7,12.

The organisms most frequently identified are skin flora, with staphylococcal species accounting for up to 80% of shunt infections^9,14,18. Staphylococcus epidermidis is the most frequent coagulase-negative staphylococci (CoNS)^7,12. Other common organisms are Staphylococcus aureus (5-26%) and other CoNS (36.1-53%)^9–11,17,19. Gram-negative organisms are also frequent findings, accounting for up to 54% of infections in some series²⁰. Gram-negative rods were identified in approximately 7–9% of infections^11,17.

Treatment of shunt infection requires complete removal of the device in combination with temporary drainage, unless infection is confined to the abdomen in which case there is a period of shunt externalisation before replacement^18,20. Early surgical intervention is especially important since staphylococci commonly form a biofilm after colonising an implant, making it more difficult for antibiotics to penetrate and eradicate infection²¹. Additionally, early and prolonged intravenous and/or intrathecal antibiotic therapy are needed to sterilise the CSF. It is recommended to wait until the cultures have been sterile for at least 72 hours before implanting a new shunt⁵. There is some evidence that in case of low-grade inflammation the patient can sometimes be treated with antibiotics alone, but success rates are reported to be low^3,4,13,14.

Optimal duration and choice of antibiotic treatment are unclear in the current literature. There is no correlation between recurrence of infection and the duration of antibiotic therapy given, suggesting that a shorter treatment course could be as effective as a longer one²². Due to the increasing concern of multidrug-resistant bacteria, the use of penicillin has largely been replaced by vancomycin¹⁸. Linezolid is a good alternative, effective against most Gram-positive microorganisms, including multidrug-resistant strains²³. There are currently no randomized studies, large case series, recent prospective studies or guidelines regarding the treatment of ventricular shunt infections in children¹⁴.

Study objectives

Given the severity of this condition and the lack of evidence-based management strategies, there is a need for evaluation of current practice to establish a standard of care. Therefore, this study aimed to retrospectively analyse the management of ventricular shunt infections in children in the regional referral centre for the North East of England, the Great North Children’s Hospital in Newcastle upon Tyne, UK. We aimed to describe the clinical presentation, diagnostic efficacy, as well as surgical and medical treatment regimens including antibiotic use. In addition, we hope to identify areas for improvement in clinical practice to reduce mortality, re-infection rates or neurological sequelae in these children.

Methods

Patient population

All children (≤18 years) diagnosed with shunt infections presenting to the Great North Children’s Hospital, Newcastle Upon Tyne, UK over the last 15 years were included in this study. Only children with a permanent shunt device were included. This could either be a VP shunt, a VA shunt, a ventriculopleural shunt or a LP shunt. Patients who had the infection before the year 2000 were excluded from the study, as were patients who were over 18 years old at the time of infection. A shunt infection was defined as any patient coded as ‘shunt infection’ at discharge or any patient who received treatment for suspected shunt infection, with or without microbiological confirmation.

Patient lists were obtained from clinical coding department of the hospital. Using broad search terms ‘shunt infection’, ‘CSF infection’ and patients aged ≤18 years old, 708 possible patients were identified. After screening for eligibility with a digital archive for discharge and referral letters, using the criteria listed above, 65 patients were included in the study (Figure 1).

Figure 1. Flowchart of patient inclusion and exclusion criteria for the study.

Results

Presentation to hospital

The median age at the first infection was 36 months (IQR 4-100 months) and 40.6% (n=26) had their first infection in the first year of life. Patients presented with a variety of non-specific symptoms. Of all the episodes of infection for which information was available (n=91), 61 (67.0%) presented with fever. Other common symptoms were irritability (n=27, 29.7%), vomiting (n=26, 28.6%), cutaneous manifestations, for example rashes or erythema over the shunt tract, (n=21, 23.1%) and abdominal pain (n=17, 18.7%). Infections with S. aureus presented more often with cutaneous manifestations (n=5, 14%), headache (n=3, 9%) and diarrhoea, (n=3, 9%), while CoNS infections presented more often with irritability (n=13, 18%) (Figure 2).

Figure 2. Distribution of symptoms documented at presentation of patients who were later diagnosed with shunt infections.

Medical and surgical management

Information regarding treatment regimens was available for 91 episodes. A total of 97 microorganisms could be identified in 73 (80.2%) of the episodes, with 31 different microorganisms causing shunt infection (Figure 3). The two largest groups were human skin flora, CoNS (n=46, 47.4%) and S. aureus (n=16, 16.5%). In 19.8% (n=18) of episodes no microorganisms were identified. Eighty-eight patients (96.7%) received antibiotic treatment. Complete information about the treatment course could be collected from 62 episodes from 45 patients. The median number of days these patients received antibiotics per episode was 18 days (IQR 14-25 days). The shortest period was one day and the longest 52 days. Most patients received more than one type of antibiotic during their treatment course. Only 4.8% (n=3) of episodes were treated with one antibiotic. The median number of different antibiotics received per episode was three, with the maximum of nine in one patient (Table 2).

Figure 3. Microorganisms identified from culture of CSF, shunt or wound and treated as the causative agent for ventriculoperitoneal shunt infection.

Table 2. Antibiotic management for 62 episodes of infection.

	Median	IQR	Min	Max
Duration of antibiotics (days)	18	14-25	1	52
Antibiotics used per infection (n)	3	3-5	1	9
Time from diagnosis to CSF sterilisation (days)	4	2-10	1	28
Time from sterilisation to shunt reinsertion (days)	11	5-17	1	56

In total 25 different types of antibiotics were used to treat 88 episodes. Vancomycin was used most frequently, in 58.0% (n=51) of episodes. Furthermore, linezolid was used in 48.9% (n=43), cefotaxime in 43.2% (n=38), meropenem in 38.6% (n=34) and rifampicin in 33.0% (n=29). Other antibiotics were used in less than 20% of the episodes (Figure 4). A schematic representation of the antibiotic treatment per patient was generated, which showed no overall trends in antibiotic use. Table 3 and Table 4 show the different combinations of antibiotics used for five days or more in S. aureus and CoNS infections respectively to give a clearer view of overall use. Empirical use of antibiotics at the start of treatment is highlighted for comparison. We have limited this information to CoNS and S. aureus infections due to a wide variation in other pathogens (n= 1-2 maximum).

Figure 4. Antibiotics used in the treatment of shunt infections, shown as the frequency of a given antibiotic used at any point in the treatment course, either as monotherapy or as part of combined therapy.

Table 3. Combinations of antibiotics used ≥5 days in 10 Staphylococcus aureus infections.

	Vanc	Line	Rifa	Fluc	Mero	Cotr	Cefo	Cefu	Metr	Teic	Ceft
1	x	x
2	x	x		x	x
3	x	x			x
4	x	x	x		x	x
5	x		x	x
6	x		x					x
7		x	x
8		x	x
9			x	x			x
10				x

X are antibiotics used for ≥5 days overall, highlighted are antibiotics were started empirically at diagnosis.

Vanc = vancomycin, Line = linezolid, Rifa = rifampicin, Fluc = flucloxacillin, Mero = meropenem, Cotr = co-trimoxazole, Cefo = cefotaxime, Cefu = cefuroxime, Metr =metronidazole, Teic = teicoplanin, Ceft =ceftriaxone

Table 4. Combinations of antibiotics used ≥5 days in 26 coagulase-negative staphylococcal infections.

	Vanc	Line	Rifa	Fluc	Mero	Cotr	Cefo	Cefu	Cipr	Metr	Teic	Ceft	Cefa	Cefta	Gent
1	x								x
2	x						x
3	x										x
4	x		x				x				x
5	x		x							x
6	x		x					x		x
7	x			x	x
8	x	x			x
9	x	x			x
10	x	x			x					x
11	x	x
12	x	x
13	x	x
14	x	x
15	x	x
16	x	x	x
17		x	x
18		x	x
19		x	x
20		x	x
21		x	x				x
22		x			x
23		x			x			x
24		x					x					x
25		x
26							x

X are antibiotics used for ≥5 days overall, highlighted are antibiotics were started empirically at diagnosis.

Vanc = vancomycin, Line = linezolid, Rifa = rifampicin, Fluc = flucloxacillin, Mero = meropenem, Cotr = co-trimoxazole, Cefo = cefotaxime, Cefu = cefuroxime, Cipr = ciprofloxacin, Metr = metronidazole, Teic = teicoplanin, Ceft = ceftriaxone, Cefa = cefalexin, Cefta = ceftazidime, Gent = gentamicin

Regarding surgical management, 67.0% (n=61) had shunt removal, 15.4% (n=14) had shunt externalisation and 26.4% (n=24) were treated without any surgical intervention. Date of sterilisation of the CSF was available for 57 episodes. The time between infection of the shunt and sterilisation of the CSF varied and was between 1 and 28 days. The median was 4 days (IQR 2-10 days). The median number of days between sterilisation of the CSF and insertion of a new shunt was 11 days (IQR 5-17 days).

Recurrent infections occurred in 18/64 patients (28.1%). In seven of these episodes the reinfection was with the same microorganism within 90 days of the previous infection. All seven patients were treated with appropriate antibiotics for the first infection and six of them had sterilised CSF after treatment. Three of the patients had no surgery, one had a shunt externalisation and three had shunt removal for this episode of infection.

Among these 92 infections in 65 patients, a total of 146 new shunts were implanted and a total of 178 shunts were revised, of which 120 revisions were not due to infection, but predominantly due to shunt blockage. Most of the infections took place shortly after insertion or revision of the shunt. 58.8% (n=47) took place within 100 days after insertion of a shunt and 70.4% (n=57) within 100 days of last contact with the shunt (either placement or revision) (Figure 5). 67.5% (n=54) of infections took place within nine months after placement of a new shunt and 58.8% (n=47) within three months. 74.1% (n=60) of the infections took place within nine months of last contact with the shunt and 70.4% (n=57) within three months. There was a significant positive correlation between the number of infections and shunts, Spearman rho 0.247 (p=0.05) and a non-significant correlation between the number of infections and revisions, Spearman rho 0.153 (p=0.23).

Figure 5. Frequency of shunt infections occurring by week after either placement of shunt or last surgical contact in the first 100 days.

Discussion

Infants and children born prematurely are known to have a higher risk of shunt infection^9,10,14,16. In our population 46% of the patients were born prematurely and 41% had the first infection before the age of one year. This reaffirms that in these groups the suspicion of a shunt infection should be promptly investigated and treated. Gender does not seem to be a risk factor for developing a shunt infection^11,13,16 and there was no difference between the number of males and females infected (55% male, p=0.19) in our study. In our population 68% of infections occurred within nine months of shunt insertion and 59% within three months. This suggests that the shorter the period after shunt insertion, the higher the suspicion of an infection. Our percentages are lower than those reported in the literature; 90% within three to nine months^4,7,12, which may suggest fewer episodes might be attributed to microbial colonisation during surgery in our cohort. It has been shown that the cumulative risk of infection increases with every revision¹¹. Our data also show a significant correlation between number of shunts and number of infections. However, we found no correlation between the number of revisions and the number of infections. This could be because the definition of ‘revision’ is complex and used for a large range of surgical procedures in our records.

Preventative strategies during surgery may influence the incidence of shunt infections. One of the most important actions shown to reduce shunt infections is perioperative prophylactic antibiotics, with the literature supporting their use in all patients undergoing a shunt insertion^2,24. A systematic review and meta-analysis by Ratilal et al. demonstrated that antibiotic prophylaxis can significantly decrease the rate of shunt infection (OR 0.51, 95% CI 0.36-0.73)²⁵. Only 65% of the patients in our populations had documented evidence of receiving antibiotics before their first shunt insertion. However, due to incomplete availability of records and a lack of documentation this number is very likely underestimated. Clearer documentation will be necessary in the future to be able to accurately determine this modifiable factor. In the Great North Children’s Hospital it is standard practice to use cefuroxime as prophylaxis before surgery. Cefuroxime covers for infections with S. aureus, but does not cover for infections with CoNS, which was the most frequently found organism causing shunt infection. It may therefore be beneficial to consider using teicoplanin as prophylaxis which is effective against both based on local resistance patterns.

Many different, non-specific symptoms can occur as a consequence of shunt infection^3,10,17. In this study, fever was the most common symptom (67% of the episodes), with cutaneous manifestations more frequent with S. aureus infections and irritability associated with CoNS infections. In children with a shunt who complain of fever, irritability, vomiting, cutaneous manifestations or abdominal pain the suspicion of an infection should be high.

S. epidermidis, S. aureus and other CoNS were the most frequently observed and together accounted for 54% of the organisms found. This correlated with findings from other groups from the United Kingdom, the United States of America, Taiwan, Korea, Spain and Germany^{7,9–12,14,17–20}. Another relatively frequent organism in the literature was Propionibacterium acnes^19,26,27 which was not identified in our cohort possibly due to the fact this organism is mainly found in patients over the age of 14–15 years old, whereas 88% (n=80) of the episodes in this study occurred before 14 years.

In this cohort, 67% underwent shunt removal and 15% underwent shunt externalisation (n=14, of which 9 went on to have shunt removal). In CSF infections, it is best practice to remove the infected device completely and insert an EVD until the CSF is sterilised^{18 ,20}. However, in infections very soon after insertion of a shunt, treatment with antibiotic alone could be justified, as the microorganisms have not had the time yet to form a biofilm, but there is a paucity of evidence for this at present and currently the safest option is to remove the infected device completely. There are no specific guidelines, but it is considered safe to insert a new shunt when the CSF has been sterile for 48–72 hours⁵. Since antibiotic treatment should be started as soon as possible, it may be beneficial to develop protocols for the use of empirical antibiotic therapy after CSF sampling at presentation. The agent chosen should cover for at least CoNS and S. aureus, since they were the most common findings. If there is abdominal pain, it may be advisable to choose an agent that also covers for Gram-negative bacteria. This agent should also be able to penetrate the CSF. A combination of linezolid and ceftriaxone would fulfil these requirements, with meropenem as a second line agent. However, specific local resistance patterns should be taken into account. Antibiotics can be rationalised once the CSF culture results are known. The optimal duration of treatment is still unclear,but should be at least seven to twelve days after the last positive culture, and depending on the microorganisms identified, according to expert opinion. It is best to start treatment IV, as this will allow CSF delivery of antibiotic. Once a patient with confirmed ventriculitis has an EVD, antibiotics should be switched to intrathecal administration when possible, since this seems to be the optimal method of antibiotic delivery to CSF^28,29.

Limitations of this study included, the incomplete availability of clinical records and incomplete documentation, resulting in information that was not possible to acquire retrospectively, in particular antibiotic prophylaxis, treatment courses and surgical notes. We were unable to obtain the total number of CSF shunts inserted during this period and were therefore unable to calculate the incidence of shunt infections in our region.

Data availability

Dataset 1: Data outlining the demographics and treatment course for each anonymised patient. 10.5256/f1000research.15514.d211612³⁰

The results presented here have previously been presented as a part of an abstract for the 2016 34^th Annual European Society for Paediatric Infectious Diseases meeting and can be found here.