Antibiotic or silver versus standard ventriculoperitoneal shunts (BASICS): a multicentre, single-blinded, randomised trial and economic evaluation.

BACKGROUND: Insertion of a ventriculoperitoneal shunt for hydrocephalus is one of the commonest neurosurgical procedures worldwide. Infection of the implanted shunt affects up to 15% of these patients, resulting in prolonged hospital treatment, multiple surgeries, and reduced cognition and quality of life. Our aim was to determine the clinical and cost-effectiveness of antibiotic (rifampicin and clindamycin) or silver shunts compared with standard shunts at reducing infection.
METHODS: In this parallel, multicentre, single-blind, randomised controlled trial, we included patients with hydrocephalus of any aetiology undergoing insertion of their first ventriculoperitoneal shunt irrespective of age at 21 regional adult and paediatric neurosurgery centres in the UK and Ireland. Patients were randomly assigned (1:1:1 in random permuted blocks of three or six) to receive standard shunts (standard shunt group), antibiotic-impregnated (0·15% clindamycin and 0·054% rifampicin; antibiotic shunt group), or silver-impregnated shunts (silver shunt group) through a randomisation sequence generated by an independent statistician. All patients and investigators who recorded and analysed the data were masked for group assignment, which was only disclosed to the neurosurgical staff at the time of operation. Participants receiving a shunt without evidence of infection at the time of insertion were followed up for at least 6 months and a maximum of 2 years. The primary outcome was time to shunt failure due the infection and was analysed with Fine and Gray survival regression models for competing risk by intention to treat. This trial is registered with ISRCTN 49474281.
FINDINGS: Between June 26, 2013, and Oct 9, 2017, we assessed 3505 patients, of whom 1605 aged up to 91 years were randomly assigned to receive either a standard shunt (n=536), an antibiotic-impregnated shunt (n=538), or a silver shunt (n=531). 1594 had a shunt inserted without evidence of infection at the time of insertion (533 in the standard shunt group, 535 in the antibiotic shunt group, and 526 in the silver shunt group) and were followed up for a median of 22 months (IQR 10-24; 53 withdrew from follow-up). 32 (6%) of 533 evaluable patients in the standard shunt group had a shunt revision for infection, compared with 12 (2%) of 535 evaluable patients in the antibiotic shunt group (cause-specific hazard ratio [csHR] 0·38, 97·5% CI 0·18-0·80, p=0·0038) and 31 (6%) of 526 patients in the silver shunt group (0·99, 0·56-1·74, p=0·96). 135 (25%) patients in the standard shunt group, 127 (23%) in the antibiotic shunt group, and 134 (36%) in the silver shunt group had adverse events, which were not life-threatening and were mostly related to valve or catheter function.
INTERPRETATION: The BASICS trial provides evidence to support the adoption of antibiotic shunts in UK patients who are having their first ventriculoperitoneal shunt insertion. This practice will benefit patients of all ages by reducing the risk and harm of shunt infection. FUNDING: UK National Institute for Health Research Health Technology Assessment programme.

RCT Entities:   Population Interventions Outcomes

Introduction

Hydrocephalus affects one in every 500 births. It also affects children and adults of all ages and can be secondary to different causes including haemorrhage, trauma, infection, and intracranial tumours. A systematic review and meta-analysis reported the prevalence of hydrocephalus to be 88 cases in 100 000 in children (aged ≤18 years), 11 cases in 100 000 in adults (aged 19–64 years), and 175 cases in 100 000 in elderly people (≥65 years). The commonest treatment for hydrocephalus is a ventriculoperitoneal shunt, which comprises proximal (ventricular) and distal (peritoneal) silicone catheters joined by a valve to drain CSF from the ventricles into the peritoneal cavity. Insertion of a ventriculoperitoneal shunt for hydrocephalus is one of the commonest neurosurgical procedures worldwide. Failure of this shunt due to infection occurs in 7–15% of patients.5, 6 Episodes of infection have a major impact on patients and are expensive to manage, requiring prolonged hospital treatment, antibiotics, and surgery to remove the infected shunt and to place a new one once the infection has been treated. Nevertheless, a health economic analysis from a UK perspective has not been done so far. Shunt infection affects health-related quality of life, cognitive function, and survival, with the number of infections per patient throughout their lifetime being an independent predictor of death.

Evidence before this study

In a systematic review comparing antibiotic-impregnated with standard shunts for patients with hydrocephalus, we identified only one randomised trial, one prospective cohort study, and ten retrospective studies; none were adequately powered to detect a difference in infection rates between different shunt types. There were no randomised trials of silver-impregnated versus standard shunts. Neurosurgeons were using antibiotic and silver shunts to reduce infection despite a lack of firm evidence to support this practice and increased financial cost.

Added value of this study

BASICS is, to our knowledge, the largest randomised trial evaluating infection risk of ventriculoperitoneal shunts in patients with hydrocephalus. Antibiotic shunts significantly reduce the risk of infection compared with standard shunts in patients of all ages. Silver shunts are associated with the same number of infections as standard shunts. From the perspective of the UK National Health Service health-care system, the use of antibiotic shunts saves £135 753 per infection avoided.

Implications of all the available evidence

From both the patient's and the clinical perspectives, every effort to reduce shunt infections should be made and technologies such as impregnated shunts, which have the potential to reduce infections, deserve proper evaluation through appropriately planned and powered trials. Having shown a marked reduction in such infections, which could have potentially catastrophic and life-changing health sequalae, the BASICS trial has provided sound evidence to support the adoption of antibiotic shunts in UK patients who are having their first ventriculoperitoneal shunt insertion. The increased upfront cost of the antibiotic shunt is offset by the added health economic benefit. The benefits and implications, both from an efficacy and health economic standpoint, are most pronounced in young patients. The global implications of these findings require consideration of their generalisability across different health-care systems.

Antibiotic-impregnated shunt catheters have been introduced as a means to reduce infection in addition to standard procedures (eg, prophylactic intravenous antibiotics) to prevent infection at the surgical site. There are three types of shunt catheter: standard, antibiotic-impregnated (0·15% clindamycin and 0·054% rifampicin), and silver-impregnated. Systematic reviews and meta-analyses9, 10 did not find any high-quality evidence to support the comparative effectiveness of these catheter types at reducing shunt infection. Consequently catheter choice varies globally, with selection based on the preference of the neurosurgeon and costs.

We coordinated the British Antibiotic and Silver Impregnated Catheters for ventriculoperitoneal Shunts (BASICS) multicentre randomised controlled trial to assess the clinical efficacy and cost-effectiveness of antibiotic and silver shunts at reducing shunt failure due to infection, compared with standard shunts in patients undergoing insertion of their first ventriculoperitoneal shunt for hydrocephalus.

Methods

Study design

In this parallel, multicentre, single-blind, randomised controlled trial we compared standard, antibiotic, and silver shunts in patients undergoing insertion of their first ventriculoperitoneal shunt for hydrocephalus. Trial sites were 21 regional adult and paediatric neurosurgery centres in the UK and Ireland (appendix pp 5–7). Ethics approval was obtained from the North West Greater Manchester Research Ethics Committee (reference number 12/NW/0790). The trial protocol is available online and has been published previously (substantial amendments are detailed in the appendix pp 56–57).

Participants

To participate in randomisation in the trial, patients could be of any age and have hydrocephalus of any aetiology requiring a first ventriculoperitoneal shunt. Patients with failed primary endoscopic third ventriculostomy, previous indwelling external ventricular drain, and indwelling ventricular access device were included. Patients were excluded if they had evidence of active and ongoing CSF or peritoneal infection; a previous indwelling ventriculoperitoneal shunt; multiloculated hydrocephalus requiring multiple shunts or neuroendoscopy; known allergy to rifampicin, clindamycin, or silver; or if a ventriculoatrial or ventriculopleural shunt was planned. Written informed consent was obtained for all patients; minors provided written consent whenever possible. Consent for adults lacking capacity was obtained from a consultee, usually the next of kin, or an independent health-care professional, and it was later sought again from the participant once capacity was regained.

Randomisation and masking

Patients were randomly assigned to receive standard shunts (standard shunt group), antibiotic-impregnated shunts (antibiotic shunt group), or silver-impregnated shunts (silver shunt group) in random permuted blocks of three and six (1:1:1). The randomisation sequence was generated by an independent statistician not otherwise involved in the trial and was stratified by neurosurgical unit and age group (adult or paediatric, defined according to unit practice). The randomisation was disclosed in the operating theatre at the time when the shunt was required using opaque, tamper-proof, sealed envelopes that were opened by tearing perforated edges. Envelopes were prepared and sealed by the independent statistician and were then sent by the trial team to the study site, where they were stored in the operating theatre, ready for use. The shunts had different colours, so it was not possible to mask the neurosurgeon and operating staff. Shunt type was not recorded in the operating record and was not disclosed outside the operating room. All investigators and statisticians who recorded and analysed data were masked to shunt assignment. Training on non-disclosure of shunt type was provided to all investigators. All shunt types were used in accordance with the manufacturer's instructions for their intended purpose. Patients were also masked to the type of shunt inserted.

Procedures

Data were collected at baseline, preoperative assessment, randomisation (first surgery), early postoperative assessment, first routine postoperative assessment, 12-weekly follow-up assessments, subsequent routine postoperative assessments, and, where applicable, at unscheduled visits and admissions, and at shunt revision and removal (appendix p 8). All patients received prophylactic antibiotics at the time of shunt insertion as per standard neurosurgical practice. All other parameters related to the surgical technique of shunt insertion (eg, choice of skin preparation, hair removal or not, number and seniority of operating neurosurgeons, rank on the operating list) were recorded but not standardised and were undertaken according to the practice of each participating neurosurgical centre. For patients requiring a first shunt revision after insertion, sites recorded data on clinical presentation (eg, temperature, headache, lethargy, meningism, consciousness level, wound erythema), peripheral white blood cell count, C-reactive protein concentrations, microbiological analysis of CSF (microscopy and culture), and type of treatment initiated (eg, antibiotics prescribed, shunt removed). Patients were followed up for a minimum of 6 months and a maximum of 2 years, depending on the time of randomisation. The types of data collected at each stage and methods used are detailed in the online study protocol.

Outcomes

The primary outcome was time to shunt failure due to infection, assessed by a masked central review panel comprised of our study's chief investigator (CLM or delegate [MDJ] for participants treated by the chief investigator) and trial microbiologist (JCH). On the basis of data on clinical presentation and type of treatment given, the shunt failure was classified as being due to infection or not on the basis of five infection definitions (panel).

Definite (culture-positive)—growth of organisms from CSF on primary culture or repeated (more than one) subculture (growth identified following enrichment of primary sample by overnight broth incubation and subculture into fresh medium), with or without clinical signs of infection, and managed by shunt removal and antibiotic treatment

Probable (culture-uncertain)—growth of organisms from CSF on one subculture only, with or without clinical signs of infection, with CSF pleocytosis, or organisms on Gram stain, or both, and managed by shunt removal and antibiotic treatment

Probable (culture-negative)—no organisms grown but with CSF pleocytosis, or organisms on Gram stain, or both (irrespective of clinical signs of infection), and managed by shunt removal and antibiotic treatment

Possible (culture-uncertain)—no signs of infection, no CSF pleocytosis, no organisms seen on Gram stain, growth after enrichment in one CSF sample only, and managed by shunt removal and antibiotic treatment

Shunt deep incision infection—infection of the deep surgical wound and subcutaneous shunt without any evidence of CSF infection, and managed by shunt removal and antibiotic treatment

The secondary outcomes were time to removal of the first shunt due to suspected infection as defined by the treating neurosurgeon at the time of first revision; time to shunt failure by any cause; reason for shunt failure as classified by the treating neurosurgeon (infection; mechanical [blockage of any component such as the valve or catheters]; patient [unrelated medical condition such as appendicitis]; or functional [change of valve for symptomatic overdrainage or underdrainage of CSF, such as change from a fixed-pressure valve to a programmable valve]); types of bacterial shunt infection; time to shunt infection after first clean revision as classified by central review; and quality of life measured using the Hydrocephalus Outcome Questionnaire. The secondary health outcomes for the economic analysis were incremental cost per shunt failure by any cause averted and per quality-adjusted life-year (QALY) gained, using the EQ-5D-3L, EQ-5D-3L (proxy), or EQ-5D-3L-Y versions of the EQ-5D health-related quality-of-life questionnaire. Data on complications and serious adverse events were collected (see the online protocol).

Statistical analysis

A trial steering committee, mostly comprising independent members who viewed reports in which treatment assignment was concealed, and an independent data monitoring committee viewing unblinded reports reviewed the trial regularly to assess conduct, progress including rates of shunt infection, and safety. The sample size estimate for the primary outcome was done via the method described by Pintilie with the following assumptions: (1) shunt failure due to infection was the event of interest, with all other reasons for failure representing a competing risk; (2) the incidence of infection would be 8% in the standard shunt group and 4% in each of the impregnated shunt (antibiotic or silver) groups; (3) the competing risk event rate would be 30%; and (4) 5% of patients would be lost to follow-up. On the basis of these assumptions, a total sample size of 1200 patients with 119 shunt failures due to infection showed sufficient statistical power (88%) with leverage for a reduced event rate. An interim analysis was planned after 50% of the total events had been observed, in accordance with the Haybittle-Peto boundary. The incidence of infection showed that the majority of events occur within 1 month of shunt insertion (ie, was not exponentially distributed), and that the incidence of infection, competing risk, and loss to follow-up were lower than expected. The independent data monitoring committee reviewed the sample size calculations and recommended increasing recruitment to a target population of 1606 patients with 101 shunt failures due to infection to provide 80% power; the trial steering committee agreed. The early occurrence of events and assumption of exponential risk were managed in the Pintilie assumptions by reducing the event accrual and follow-up rates to 1 month.

The analysis was done according to a prespecified statistical analysis plan. Amendments to the plan were considered and implemented by a masked statistician, which included masked collection and summary of the data before database lock. Outcomes were analysed according to the intention-to-treat principle and safety analyses according to the type of shunt in situ. To adjust for the three treatment groups, a p value of 0·025 was considered significant and 97·5% CIs were used throughout. Outcome analysis, with shunt failure due to infection as the event of interest, used Fine and Gray survival regression models with cause-specific hazard ratios (csHR) and subdistribution hazard ratio (sHR) presented.16, 17 Cox regression models were used to analyse time to shunt failure of any cause. The assumption of proportionality for time to event outcomes was checked using Schoenfeld residuals. Reason for shunt failure is presented descriptively and with a χ2 test. Types of organisms cultured from CSF are presented descriptively with frequency tables, no formal statistical analyses were done. Quality-of-life outcomes were analysed using mixed models for repeated measures. All survival models were adjusted for the age category of the recruiting site (paediatric or adult), with adult sites further categorised by age above 65 years. Due to the dependency between age group and study site, adjusting models by both covariates was not possible. Instead, age group was used in preference to study site because of its prognostic value. Analyses for the primary outcome and safety were validated by independent programming from the point of raw-data extraction. All analyses were done with the SAS software (version 9.4) with SAS/STAT package 14.3. The trial was registered with ISRCTN 49474281.

Economic analysis

The economic analysis (appendix pp 21–55) adopted the perspective of the National Health Service in the UK to estimate the incremental cost per first shunt failure (due to any cause) averted for antibiotic, silver, and standard shunts. Within-trial use of health-care resources was based on responses to questionnaires, routine hospital data via patient-level information and costing systems, and entries in case report forms. Unit costs for 2016–17 were taken from standard sources (appendix pp 23–26).18, 19, 20, 21 The cost of a silver shunt was £361·62, an antibiotic one was £384·00, and a standard one was £172·00. In the base-case economic analysis, costs and outcomes incurred in the second year after shunt placement were discounted at a rate of 3·5%, and any missing data were multiply imputed with the method of chained equations. Total costs were analysed by use of linear regression with the stratifying variables, time in study, intervention group, site, and treatment failure as predictors. Mean outcome by intervention group was analysed in the same way, but total cost was substituted by treatment failure as the dependent variable. Sensitivity analyses considered (1) applying different discount rates (0%, 1·5%, and 6% per annum for both costs and outcomes), (2) using observed data for costs (no multiple imputation), and (3) using a generalised linear model for analysing costs. Alternative forms of cost-effectiveness and a cost-utility analysis relating to participants aged at least 5 years were also done. A stratified analysis was undertaken to estimate cost-effectiveness by age group in paediatric patients, adults aged 18–65 years, and those aged 65 years and older.

Role of the funding source

The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Results

Between June 26, 2013, and Oct 9, 2017, we assessed 3505 patients for eligibility and randomly assigned 1605 to the study groups (536 to the standard shunt group, 538 to the antibiotic shunt group, and 531 to the silver shunt group; figure 1). One patient was erroneously randomised twice and so data from the first randomisation only were used. 53 participants subsequently withdrew from follow-up, of whom 24 continued to provide routinely collected data. The characteristics of the three groups were similar at baseline (table 1, appendix pp 11–12). 1601 (99·8%) patients had a shunt inserted and 1585 (98.8%) received the correct allocated shunt (figure 1). Patients who did not receive a shunt (n=4) or who had an infection at insertion (n=7) were not included in the primary analysis. The number of patients included in the primary analysis was 533 in the standard shunt group, 535 in the antibiotic shunt group, and 526 in the silver shunt group. The median follow-up time for patients assessed for the primary outcome was 22 months (LQ–UQ 10–24, min to max 0–24; n=1594).

Figure 1

Trial profile

Table 1

Baseline characteristics of the intention-to-treat population

	Standard shunt (n=536)	Antibiotic shunt (n=538)	Silver shunt (n=531)	Total (n=1605)
Age at randomisation, years
Median (LQ–UQ)	42·5 (0·8–69·7)	43·9 (1·1–70·8)	41·1 (0·5–68·8)	42·5 (0·8–69·6)
Range	0·0–90·3	0·0–88·9	0·0–91·1	0·0–91·1
Age category
Paediatric	200 (37%)	201 (37%)	198 (37%)	599 (37%)
Adult (<65 years)	174 (32%)	156 (29%)	172 (32%)	502 (31%)
Adult (≥65 years)	162 (30%)	181 (34%)	161 (30%)	504 (31%)
Gender
Female	246 (46%)	260 (48%)	282 (53%)	788 (49%)
Male	289 (54%)	278 (52%)	249 (47%)	816 (51%)
Missing	1 (<1%)	..	..	1 (<1%)

Data are number of participants (%), unless otherwise specified.

398 (25%) of the 1594 followed up patients had revision operations, with 75 (5%) being centrally classified as having shunt infections (table 2). Compared with the standard shunt, the antibiotic shunt decreased the incidence of shunt failure due to infection over time (csHR 0·38, 97·5% CI 0·18–0·80, p=0·0038; table 3). Silver shunts were comparable with standard shunts (0·99, 0·56–1·74, p=0·96) in this respect. Figure 2 shows the cumulative incidence of failure due to infection by shunt group. 53 [71%] of 75 centrally assessed infections were classified as definite (culture-positive).

Table 2

Summary and reasons for revision of first shunt according to catheter type and assessor

			Standard shunt	Antibiotic shunt	Silver shunt	Total
Surgeries
Patients eligible for primary outcome*			533	535	526	1594
No shunt removal or revision			403 (76%)	403 (75%)	390 (74%)	1196 (75%)
Shunt removal or revision (for any cause)			130 (24%)	132 (25%)	136 (26%)	398 (25%)
Reason for revision as classified by central review
Patients revised for infection			32 (6%)	12 (2%)	31 (6%)	75 (5%)
	CSF or peritoneal infection
		Definite (culture-positive)	22/32 (69%)	6/12 (50%)	25/31 (81%)	53/75 (71%)
		Probable (culture-uncertain)	1/32 (3%)	..	2/31 (6%)	3/75 (4%)
		Probable (culture-negative)	3/32 (9%)	3/12 (25%)	1/31 (3%)	7/75 (9%)
		Possible (culture-uncertain)	1/32 (3%)	..	1/31 (3%)	2/75 (3%)
		Clinically classified infection†	1/32 (3%)	..	..	1/75 (1%)
		Shunt deep incisional infection	4/32 (13%)	3/12 (25%)	2/31 (6%)	9/75 (12%)
Patients revised for other reason (no infection)			98 (18%)	120 (22%)	105 (20%)	323 (20%)
Reason for shunt revision as classified by treating neurosurgeon
Suspected infection			33 (6%)	15 (3%)	30 (6%)	78 (5%)
Revision for other reason (no infection)			97 (18%)	117 (22%)	106 (20%)	320 (20%)

Data are n, n (%), or n/N (%) of patients.

Randomised participants who did not receive a shunt (n=4) and had infection at time of insertion (n=7) were excluded from the primary outcome set (figure 1).

In one case the committee was unable to classify the infection, so the infection was clinically identified from the case report forms.

Table 3

HR estimates from multivariate regression modelling

			Events (n)	csHR (97·5% CI; p value)	sHR (97·5% CI; p value)
Primary outcome*
Shunt
	Standard		32	..	..
	Antibiotic		12	0·38 (0·18–0·80; 0·0038)	0·38 (0·18–0·80; 0·0037)
	Silver		31	0·99 (0·56–1·74; 0·96)	0·99 (0·56–1·72; 0·95)
Age group
	Paediatric		47	..	..
	Adult (<65 years)		23	0·55 (0·31–0·97; 0·018)	0·56 (0·32–0·99; 0·021)
	Adult (≥65 years)		5	0·12 (0·04–0·34; 0·0019)	0·12 (0·04–0·35; 0·0022)
Secondary outcomes
Time to removal of first shunt due to infection (assessed by neurosurgeon)*
	Shunt
		Standard	33	..	..
		Antibiotic	15	0·45 (0·23–0·91; 0·011)	0·45 (0·23–0·91; 0·011)
		Silver	30	0·93 (0·53–1·64; 0·77)	0·92 (0·53–1·61; 0·74)
	Age group
		Paediatric	50	..	..
		Adult (<65 years)	23	0·51 (0·29–0·91; 0·0085)	0·53 (0·30–0·93; 0·011)
		Adult (≥65 years)	5	0·11 (0·04–0·31; 0·0018)	0·12 (0·04–0·33; 0·0021)
Time to removal or revision of the first shunt for any reason*†
	Shunt
		Standard	130	..	..
		Antibiotic	132	1·01 (0·77–1·33; 0·94)	..
		Silver	136	1·08 (0·82–1·42; 0·54)	..
	Age group
		Paediatric	226	..	..
		Adult (<65 years)	118	0·57 (0·44–0·74; <0·0001)	..
		Adult (≥65 years)	55	0·25 (0·18–0·35; <0·0001)	..
Time to failure of second shunt due to infection, following clean revision‡
	Shunt
		Standard	9	..	..
		Antibiotic	6	0·55 (0·17–1·81; 0·26)	0·55 (0·17–1·75; 0·25)
		Silver	5	0·47 (0·13–1·63; 0·17)	0·48 (0·14–1·67; 0·19)
	Age group
		Paediatric	10	..	..
		Adult (<65 years)	9	1·64 (0·58–4·61; 0·28)	1·72 (0·62–4·81; 0·24)
		Adult (≥65 years)	1	0·34 (0·03–3·64; 0·14)	0·38 (0·04–3·91; 0·14)

csHRs are calculated from the multivariate Cox model with infection as the event of interest and both shunt and age groups as covariates. sHRs are calculated from the multivariate Fine-Gray model with infection as the event of interest, revision for a reason other than infection as a competing risk, and both shunt and age groups as covariates. HRs are calculated from the multivariate Cox model with revision or removal of the shunt as the event of interest and both shunt and age groups as covariates. csHR=cause-specific hazard ratio. sHR=subdistribution hazard ratio. HR=hazard ratio.

Median follow-up time from insertion of the first shunt was 22 months, LQ–UQ 10–24, range 0–24 (standard); 21 months, IQR 10–24, range 0–24 (antibiotic); and 22·5 months, IQR 8–24, range 0–24 (silver).

Reported as HR (97·5% CI; p value).

Median follow-up time from insertion of the first shunt was 9 months, LQ–UQ 2–19, range 0–24 (standard); 7 months, IQR 2–18, range 0–24 (antibiotic); and 7 months, IQR 2–19, range 0–24 (silver).

Figure 2

Cumulative incidence plots of infection (A) and competing risk (B) by shunt type

The numbers of patients at risk apply to both graphs. csHRs and sHRs for infection incidence are in table 3. For competing risk, the ratios are: csHR 1·22 (97·5% CI [0·90, 1·65]) and sHR 1·26 (97·5% CI [0·93, 1·70]) for antibiotic shunts, and csHR 1·11 (97·5% CI [0·81, 1·51]) and sHR 1·10 (97·5% CI [0·81, 1·50]) for silver shunts. csHR=cause-specific hazard ratio. sHR=subdistribution hazard ratio.

Of the 398 revision operations, 78 (5%) were defined by the treating neurosurgeon as being caused by a suspected infection (table 2). Antibiotic but not silver shunts were associated with a significant decrease of failure due to infection when compared with standard shunts (table 3). The reason for revision was classified by central review (primary outcome) and treating neurosurgeon (secondary outcome) as infection or no infection, and this classification was the same in 381 (96%) revisions (appendix p 13).

Kaplan-Meier curves for time to shunt failure for any cause showed no significant difference between antibiotic or silver and standard shunts (table 3, appendix p 13). The number of shunt failures was similar between the three groups (table 2); however, the underlying reason differed when comparing patients who had standard shunts with those who had antibiotic shunts (p=0·024), but not those with silver shunts (p=0·71; appendix p 14). Patients with antibiotic shunts had fewer infections but a higher frequency of mechanical shunt failure than the other two groups (appendix p 14).

Staphylococcus aureus (17 [30%] of 56 infections) and coagulase-negative staphylococci (22 [39%]) accounted for most organisms cultured from infected shunts (appendix p 15). Culture results showed a reduction in staphylococcal and Gram-positive infections in patients with antibiotic shunts compared with those who had standard or silver ones. Patients with all three shunt types had a similar number of Gram-negative infections. The proportion of culture-positive infections was lowest in patients with antibiotic shunts compared with patients from the other two groups (table 2). The remaining infections were identified by the central review panel on the basis of CSF white-cell counts, clinical features, and blood parameters (table 2).

In patients whose first revisions were not due to infection (n=323) as assessed by central review, the overall incidence of subsequent revisions for any reason (infection and no infection) was 40% (n=126; appendix p 16), higher than in the full follow-up group (table 2). The overall incidence of infection was higher in this subgroup (20 [6%] of 323; appendix p 16) than in the full follow-up group. There was no significant difference in time to infection following the first clean revision when comparing either antibiotic or silver (table 3) with standard shunts.

The proportion of revisions of the first shunt for any cause (225 [38%] of 592 children, 118 [24%] of 499 adults younger than 65 years, and 55 [11%] of 503 adults aged 65 years and above; appendix p 17) and for infection (47 [8%], 23 [5%], and five [1%]; appendix p 17) varied by participant age group. Compared with children, over time adults younger than 65 years and adults aged 65 years and above had a significantly lower rate of shunt failure due to infection (<65 years csHR 0·55, 97·5% CI 0·31–0·97, p=0·0018; ≥65 years 0·12, 0·04–0·34, p=0·0019; table 3, appendix pp 18–19).

Schoenfeld residuals supported the assumption of proportionality for time to event outcomes. There were no serious adverse events. 654 adverse events were reported in 413 (26%) of all patients who received a shunt. The proportion of patients experiencing an event was highest for those with silver shunts (134 [36%]), followed by those with standard shunts (135 [25%]), and was lowest for those with antibiotic shunts (127 [23%]; appendix p 20). Common adverse events were ventricular catheter obstruction (96 events in 79 patients), shunt valve obstruction (65 in 52), and valve change for symptomatic overdrainage (54 in 50). All of these were expected events in the context of re-admission for shunt revision.

The level of missing cost data was balanced across the three intervention groups (appendix pp 31–32). Disaggregated resource use and costs are shown in the appendix (pp 34–35). Mean total costs for the duration of the study period (24 months) were £18 707 (97·5% CI 13 888–26 966) for standard shunts, £14 192 (12 450–17 786) for antibiotic shunts, and £17 385 (14 649–22 355) for silver shunts (appendix p 37). In the base-case analysis, the total costs relating to both silver and antibiotic shunts were less than those relating to standard shunts (appendix pp 44–45). Incrementally, silver shunts saved £62 358 for each additional first shunt failure for any reason compared with standard shunts, and antibiotic shunts saved £638 600 per additional failure compared with silver (appendix pp 44–45). In sensitivity analyses, the incremental cost-effectiveness ratios were stable to changes in discount rate and choice of regression modelling but, on the basis of observed data, antibiotic shunts were superior to silver shunts and saved £56 771 for each additional failure compared with standard shunts (appendix pp 44–45). Antibiotic shunts were most cost-effective in children, with mean savings of £5312 and 0·004 fewer shunt failures during the study period (appendix p 47). A cost-effectiveness analysis based on the incremental cost per infection averted indicated that silver shunts were dominated by standard shunts, whereas antibiotic shunts were superior to the other two types, saving £4059 per 0·030 fewer infection-related shunt failures (appendix p 49). There were insufficient data to formally analyse Hydrocephalus Outcome Questionnaires, and results are therefore presented descriptively (appendix pp 51–52). In the cost-utility analysis, silver shunts were cheaper than antibiotic shunts (appendix p 49). Compared with standard shunts, silver shunts were £183 more expensive and yielded 0·096 additional QALYs overall, resulting in an incremental cost-effectiveness ratio of £1904 per QALY gained, and a probability of 0·52 of being cost-effective at a willingness to pay of £20 000 per QALY (appendix p 53).

Discussion

In this trial of patients with hydrocephalus undergoing insertion of a first permanent ventriculoperitoneal shunt, 6% of those receiving standard shunts, 2% of those receiving antibiotic shunts, and 6% of those receiving silver shunts had an infection. Compared with standard shunts, antibiotic shunts were associated with a significantly lower incidence of infection, whereas silver shunts were not. This effect was present across all age categories. The risk of shunt infection was highest in children, reducing in adults, and being particularly low in the elderly. There are significant economic benefits for every shunt infection averted, although cost-effectiveness is greatest in those at highest risk.

The BASICS trial provided sound evidence on the use of antibiotic shunts to reduce infection. A previous randomised trial compared antibiotic with standard shunts, but was underpowered and did not show a significant difference in the risk of infection (relative risk 0·38, 95% CI 0·11–1·30; p=0·12). Additionally, systematic reviews and meta-analyses9, 10 did not find any high-quality evidence to support the comparative effectiveness of antibiotic shunts at reducing infection. Silver catheters have only been evaluated for use in temporary external ventricular drains, not permanently implanted shunts. A randomised trial of external ventricular drains (silver vs standard) reported a reduction in infection from 21% (30 of 140 patients with a silver drain) to 12% (17 of 138 patients with a standard drain; p=0·042), although this proportion is much higher than the UK's national reported infection rate (9%). The BASICS trial was therefore conceived to evaluate both antibiotic and silver shunts, which might otherwise have been widely introduced into routine clinical practice despite a lack of firm evidence of their efficacy. The results of our trial show that antibiotic shunts have good clinical and cost-effectiveness and will inform neurosurgery practice and shunt choice for the benefit of patients.

Correctly diagnosing shunt infections when the CSF is culture-positive is straightforward, but this situation only applies to about 70% of cases. When the CSF is culture-negative, the treating neurosurgeon must consider other parameters including CSF white-cell counts, clinical symptoms and signs, and previous treatment with antibiotics. In these circumstances, removal of the shunt and antibiotic treatment often resolve the presumed infection and the patient recovers. The classification of shunt infection in our study was determined by the central review committee (table 2), and the proportion of culture-positive infections was 69% in patients with standard shunts, 50% in those with antibiotic shunts, and 81% in those with silver shunts. There was a reduced incidence of culture-positive infections with antibiotic shunts. Our analysis allowed for culture-negative infections to be included when there was sufficient supporting clinical evidence of shunt infection. This inclusion was possible because we postulated that the presence of antibiotic and possibly silver shunts might reduce the ability to culture organisms from infected shunts. Our study showed an even greater effect favouring antibiotic shunts when only culture-positive infections were analysed. The reduction of infections seen is consistent with the expected microbiological spectrum inhibited by the antibiotic shunts, which are especially active against Gram-positive organisms, and were designed to prevent infection by Staphylococcus species. The culture results show a large reduction in staphylococcal infection when antibiotic shunts were present, compared with standard and silver shunts, which account for most of the reduced infections. Patients with all three shunt types had a similar number of Gram-negative infections, supporting the biological plausibility of our results.

It should be noted that the overall incidence of shunt failure was the same for all groups, although infection was reduced in patients with antibiotic shunts. When infection is discounted as a cause of shunt failure, clean, non-infected revisions were slightly more frequent in patients with antibiotic shunts. The cause is unclear, but one hypothesis is that the antibiotic catheters might convert a true infected shunt revision into an apparently clean shunt revision. This masking might occur because pathogens with low virulence are restricted to a biofilm in the valve (which is not impregnated) that does not cause detectable changes in the CSF (such as increased white-cell count), as there is no ventriculitis and the bacteria are few or unable to grow in the presence of the eluted antibiotics. However, changes in CSF composition and flow (such as debris or high protein) might block the intricate valve mechanism. Our study was not powered or designed to answer this question directly, but analysis of the CSF samples collected in BASICS might lead to improved pathogen detection. Nevertheless, from the patient's perspective, although mechanical shunt revision still requires surgery which could impact on quality of life, the hospital admission is short, prolonged antibiotics are not required, and patients recover faster with fewer long-term neurological sequelae than if their shunts become infected.7, 8

Complications associated with shunt failures are expensive to manage.25, 26, 27 Economic analyses suggest that the use of impregnated shunts that result in fewer complications, even if more expensive to purchase, could be cost-effective or yield cost savings.10, 28, 29 The cost-effectiveness analysis in BASICS estimated that although antibiotic shunts cost twice as much as standard shunts, they are expected to be cost-effective overall. Nevertheless, on the basis of the primary economic outcome, this estimate might be at the expense of additional cases of shunt failures (for any reason) associated with the use of silver and antibiotic shunts compared with standard ones. The secondary economic outcome based on the incremental cost per shunt infection averted is relevant because a reduced incidence of infection is expected to be associated with reduced need of further surgery and prolonged hospital care. Standard shunts were superior to silver shunts but were inferior to antibiotic shunts, saving £4059 per 0·030 fewer infection-related failures, equating to £135 753 per infection avoided. The cost-utility analysis was limited by missing data and by the exclusion of participants who were at highest risk of shunt infections.

The strengths of our study are that (1) infections were centrally classified by researchers masked to treatment allocation, thereby removing the risk of bias by the treating neurosurgeon; (2) participant retention was very high because of the nature of the intervention and the primary outcome (patients with infected shunts are always re-admitted to hospital); (3) patient withdrawal was low (53 [3%]) so it is unlikely that shunt failures due to infection were undetected; (4) participants were recruited across the whole of the UK and Ireland to encompass all ages and socioeconomic classes; (5) the study population was large; and (6) the results have wide generalisability because we did not mandate a specific surgical technique for shunt insertion.

Some limitations of our trial should be noted. First, it was not possible to mask the treating neurosurgeon to the shunt type because the physical appearance of the shunts is distinctive. The shunt type was concealed from the patient and was not recorded in the patient records. Most shunt revisions and removals for infection happen as emergencies and are managed by the emergency neurosurgery team. Therefore, the likelihood of the same neurosurgeon who inserted the shunt being involved in the decision to remove it was low, considering the work rotas of neurosurgical staff. Furthermore, there was high agreement between shunt infections classified by the treating neurosurgeon and central assessment (96%), suggesting that any bias coming from the treating neurosurgeon did not affect the conclusions. Second, ventriculoatrial and ventriculopleural shunts were excluded, although we postulate that the results are translatable to patients with those shunt types. Finally, the low proportion of patient-reported outcomes restricted the analysis of the effect of shunt infection on patients, and the reliability of the cost-utility analysis.

The BASICS study is the largest prospective randomised study of patients with shunts for hydrocephalus worldwide. The blood and CSF samples from study participants will be used for future research into biomarkers for infection and host response. Data on hydrocephalus aetiology, surgical techniques, types of valves, and technology used for shunt insertion will be analysed and used to develop recommendations and health-care policy for patients receiving ventriculoperitoneal shunts. Antibiotic shunts would reduce the risk of infection and be substantially cost-effective, and thus they should be the first choice for patients with hydrocephalus undergoing insertion of their first ventriculoperitoneal shunt.

This online publication has been corrected. The corrected version first appeared at thelancet.com on September 18, 2019

Data sharing

Individual participant data, after deidentification, will be made available, along with the study protocol, statistical analysis plan, and consent forms. Data will be available beginning 9 months and ending 3 years after publication of the Article to researchers whose proposed use of the data is approved by the original study investigators. Proposals should be directed to the corresponding authors and requesters will need to sign a data access agreement.