CATheter Infections in CHildren (CATCH): a randomised controlled trial and economic evaluation comparing impregnated and standard central venous catheters in children

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 08/13/47. The contractual start date was in March 2010. The draft report began editorial review in May 2015 and was accepted for publication in October 2015. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

Michael Millar was a member of the National Institute for Health Research Health Technology Assessment Diagnostic Technologies and Screening Panel for the duration of the CATCH study.

Copyright statement

© Queen’s Printer and Controller of HMSO 2016. This work was produced by Harron et al. under the terms of a commissioning contract issued by the Secretary of State for Health. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

Use in practice

Central venous catheters (CVCs) are widely used for patients of all ages who need intensive or high-dependency care to provide venous access for resuscitation, drug delivery, intravenous feeding, monitoring and blood sampling. CVCs are associated with an increased risk of bloodstream infection (BSI), which is hypothesised to be caused by organisms tracking along the CVC from the skin or from the external parts of the CVC to colonise the CVC tubing and tip. 1–5

Risk factors for BSI include catheter dwell time, the frequency of ‘breaching’ the line for medication or sampling, multiple compared with single-lumen CVCs and infusion of lipid solution as part of parenteral nutrition. 6–10 The risk of BSI is reduced by strict adherence to aseptic procedures during CVC insertion and whenever the CVC is breached. 11–13 To help ensure staff follow aseptic procedures, audited checklists (called CVC bundles) have been introduced in several countries. 14–19

In this report we focus on children who need a CVC as part of their intensive care treatment. Paediatric intensive care units (PICUs) have one of the highest reported rates of hospital-acquired BSI of any clinical specialty20–23 and BSI is an important cause of adverse clinical outcome and health-care costs in critically ill children. 21,24–26 We estimate that approximately 60% of the 16,000 children admitted to 23 PICUs each year in England require insertion of a CVC as part of their acute care. 27 We do not include CVCs used for very preterm babies in neonatal intensive care or long-term CVCs, which are widely used to administer medication or parenteral nutrition for children with conditions such as cancer, cystic fibrosis, renal failure or short gut syndrome.

Rationale

Central venous catheter impregnation with anti-infective substances has been used for over 25 years. 28 Recent systematic review evidence from 48 randomised controlled trials (RCTs) and cost-effectiveness analyses including 11,586 patients demonstrated substantial benefits of impregnated compared with standard CVCs for catheter-related BSI (CR-BSI). 2,5,28–30 One of the most recent systematic reviews included a meta-analysis of direct and indirect comparisons of different types of impregnated and standard CVCs. 28 Heparin-bonded or antibiotic-impregnated CVCs were found to be the most effective options, being associated with similar reductions (70–80%) in the risk of CR-BSI. Heparin bonding acts by reducing thrombus formation and bacterial adherence to thrombi, but the bonding agent, benzalkonium chloride, also has anti-infective properties. Antibiotic-impregnated CVCs act by preventing biofilm formation and thereby prevent bacterial colonisation.

Despite the large number of RCTs and the substantial reductions seen in the risk of BSI in adults, impregnated CVCs have not been recommended for children in US or UK guidelines and their use in UK practice has been limited. 3,15,31,32 A recent survey showed that impregnated CVCs had been adopted for some or all children by less than half of British PICUs surveyed. 32 Lack of implementation in PICUs relates to (1) gaps in the evidence relating to children; (2) concerns about the quality of previous trials; and (3) uncertainty about the generalisability of RCT findings to settings where improved infection control strategies have been associated with steep declines in BSI rates. 33,34

In children, there is a lack of evidence on the most effective type of CVC and on the expected effect size. According to the network meta-analysis by Wang et al. ,28 heparin-bonded and antibiotic-impregnated CVCs are the most effective options, having similar effects compared with standard CVCs. However, there is a lack of evidence on which type of CVCs would be most effective as there have been no adequately powered, direct ‘head-to-head’ comparisons of these options. 28 In the UK, the additional costs of heparin-bonded or antibiotic-impregnated CVCs are similar and so the decision on which type to adopt depends on their relative benefits and adverse effects. Only one of the eight RCTs comparing antibiotic-impregnated CVCs with standard CVCs (n = 2073 patients) included children and this study was terminated early because of a lower than expected event rate. 35–42 As CVCs for children are much narrower than adult CVCs and the risk of thrombus formation, bacterial adhesion and infection is much higher, it is hypothesised that the relative effect of antibiotic-impregnated compared with standard CVCs may differ in children and adults. Evidence is stronger for the benefits of heparin-bonded CVCs, as two of the three RCTs comparing heparin-bonded CVCs with standard CVCs (n = 472) included children. 43–45

Several systematic reviews have raised concerns that the poor quality of previous studies means that the benefits of impregnated CVCs may have been overestimated. 5,29,46,47 First, few trials have reported good concealment of treatment allocation or blinding of clinicians to the intervention and many have failed to account for losses or withdrawals, all factors that could lead to overestimation of the effect. 5,29 Second, all previous trials relied on CR-BSI as the primary outcome measure, which requires positive cultures from the blood and catheter tip. This measure is highly susceptible to bias, as the tip can be easily contaminated during removal and residual antibiotic in the catheter tip may inhibit culture in the laboratory. Aside from the potential biases in measuring CR-BSI, impregnated CVCs may impact on all BSIs after CVC insertion, not just on CR-BSIs, and on the risks of mortality, complications and increased length of stay associated with BSI.

Few trials have determined the effect of impregnated CVCs on all BSIs in PICU in the context of ongoing reductions in BSI rates associated with the introduction of CVC care bundles. 11,13,14,33,48 Neither of the two trials of heparin-bonded CVCs in children and few of the trials of antibiotic-impregnated CVCs in adults have been conducted in the context of these strenuous efforts to reduce BSI. It is not known whether or not the relatively large reductions in relative risk (RR) and absolute risk seen in trials predating CVC care bundles would be sustained in PICUs where rates of infection have already been reduced by improved CVC care. 34 Even though a UK cost-effectiveness analysis estimated that impregnated CVCs would be cost-effective given baseline rates of CR-BSI as low as 0.2%,29 there remains the question of whether or not the relative effect of impregnation would be less given improved catheter care.

Risks and benefits

Prevention of BSI is undoubtedly a clinically important outcome. Although evidence on attributable mortality varies, BSI is clearly associated with a longer stay in hospital and more intensive support. 20,21,24–26,49 For children in intensive care, CR-BSI has been associated with an additional 9–21 days’ stay in hospital (6.5–15 days in PICU). 24–26 In adults, the additional acute health-care costs attributable to a BSI are an estimated £9148 per patient and could range between £2500 and £71,000. 29 The few studies of the costs of BSIs in PICU patients have found a difference of US$33,039–39,219 in PICU direct costs between infected and uninfected patients. 21,25 However, quantifying the effects of BSI are complicated by the time-dependent exposure: BSI increases hospital stay; increased length of stay is a risk factor for BSI. 50 Estimates of the attributable length of stay are subject to this time-dependent bias, leading to potentially overestimated BSI costs in previous studies. 51,52 On the other hand, no study has taken into account the long-term costs associated with BSI in children.

Potential adverse effects of CVCs are rare. Heparin bonding could theoretically trigger an allergic response, leading to heparin-induced thrombocytopenia, although no case has been reported to the manufacturers. Antibiotic impregnation could potentially lead to antibiotic resistance, although a systematic review showed no increased risk of resistant organisms isolated from blood cultures. 1

Overview of aims and research questions

From a policy perspective, there could potentially be significant gains in terms of children’s health and health-care costs across the NHS if impregnated CVCs could be confirmed to substantially reduce rates of BSI. We compared both types of impregnated CVC previously shown to be most effective (antibiotic and heparin) with standard CVCs to determine the effectiveness of CVC impregnation in children. Secondary analyses were conducted to evaluate the effectiveness of each type of CVC.

We aimed to inform NHS policy regarding impregnated CVCs for intensive care of children by undertaking a large pragmatic RCT to determine (1) clinical effectiveness; (2) cost-effectiveness of impregnated compared with standard CVCs; and (3) the generalisability and cost impact of adopting impregnated CVCs for all children who need them.

The main objectives and data sources for the three parts of the study were:

1. Clinical effectiveness:

   • to determine the effectiveness of impregnated compared with standard CVCs for reducing BSI in children admitted to intensive care

   • to determine which type of CVC is most effective, based on three-way comparisons of measures of BSI, mortality and adverse events

   • data source: clinical outcomes captured on case report forms (CRFs) in the RCT.

2. Cost-effectiveness:

   • to determine the cost-effectiveness of impregnated compared with standard CVCs for reducing BSIs, based on incremental acute health-care costs per BSI avoided

   • data sources: clinical outcomes captured on CRFs in the RCT and records of health-care use captured by linkage of RCT data with hospital administrative data.

3. Generalisability and cost impact:

   • to estimate the net cost impact to NHS PICUs given a policy to adopt impregnated CVCs for all children who need them

   • data sources: national data on PICU admissions (Paediatric Intensive Care Audit Network; PICANet) linked with infection surveillance data collated by Public Health England (PHE) and costs from the economic evaluation.

The specific objectives, methods and results for each of the three phases of the study are reported in Chapters 2–5. We discuss the implications of our findings for policy and recommendations for future research in Chapter 6.

Trial design

We conducted a parallel, three-arm RCT. 53 Children admitted to 14 PICUs in England between December 2010 and November 2012 were randomised to CVCs impregnated with antibiotics or heparin or to standard CVCs in a ratio of 1 : 1 : 1.

Setting and participants

Children aged < 16 years were eligible if they were admitted to a participating PICU or were being prepared for PICU admission by an emergency retrieval team and were expected to require a CVC for ≥ 3 days. Children who had already participated in the trial were ineligible.

Interventions

We used polyurethane CVCs manufactured by Cook Medical Incorporated (Bloomington, IN, USA). Sizes used were French gauge 4 (double lumen), 5 or 7 (triple lumen). Both types of impregnation involve internal and external surfaces. Cook Medical Inc. reports a concentration of 503 μg/cm of minocycline and 480 µg/cm of rifampicin for their antibiotic-impregnated CVC, which reduces biofilm formation. 54 Heparin bonding reduces thrombus and thereby biofilm formation and uses benzalkonium chloride as an anti-infective bonding agent. 5,55

Randomisation and consent

For children admitted to the PICU following elective surgery, we sought prospective parental consent during preoperative assessment. Randomisation took place in theatre or in the anaesthetic room prior to entry into theatre. For children who required a CVC as an emergency, we sought parental consent after randomisation and stabilisation (deferred consent) to avoid delaying treatment, which was usually within 48 hours of randomisation. Children who required a CVC as part of their emergency care or resuscitation were randomised at the bedside in the PICU or at another hospital, where they were randomised by the PICU retrieval team prior to transfer to the PICU. Further details are given in the protocol [see www.nets.nihr.ac.uk/projects/hta/081347 (accessed 20 November 2015)].

At randomisation, the clinician or research nurse opened a pressure-sealed, sequentially numbered opaque envelope containing the CVC allocation. Randomisation sequences were computer generated by an independent statistician in random blocks of three and six, stratified by method of consent (deferred or prospective), site and envelope storage location within the site to facilitate easy access to envelopes (e.g. for insertion in theatre and in the PICU).

Parents consented to the use of their child’s data for the trial, to follow-up using routinely recorded clinical data and to 0.5 ml of blood being collected whenever a blood culture was clinically required. 56 Samples were sent for polymerase chain reaction (PCR) testing for 16S ribosomal ribonucleic acid (rRNA) of bacterial ribosome protein to detect bacterial infection.

We also sought consent to link data from PICANet57 to the child’s study data to categorise the primary reason for admission and the Paediatric Index of Mortality (PIM2)58 score on admission and to link to administrative hospital data for the economic analyses and death registration data from the Office for National Statistics (ONS) [see www.ons.gov.uk (accessed 4 January 2016)] to determine mortality after discharge from the PICU.

Blinding

Central venous catheter allocation was not blinded to the clinician responsible for inserting the CVC (because of the different colour strips for antibiotic and heparin CVCs) but, as the CVCs looked identical whilst in situ, allocation was concealed from patients, their parents and PICU personnel responsible for their care. Labels identifying the type of CVC were held securely in a locked drawer in case unblinding was required. Participant inclusion in analyses and occurrence of outcome events were established prior to release of the randomisation sequence for analysis.

Comparisons and outcomes

The primary analysis for the trial compared antibiotic or heparin CVCs with standard CVCs. Secondary analyses compared antibiotic with standard CVCs, heparin with standard CVCs and antibiotic with heparin CVCs.

The primary outcome was time to the first BSI based on blood cultures taken between 48 hours after randomisation and 48 hours after CVC removal (or prior to death). We used time to event analyses as the risk of BSI increases the longer a CVC is in place. This time interval was intended to capture BSIs related to the type of CVC. All blood culture samples were clinically indicated, defined by removal of the CVC because of suspected infection or other recorded evidence of infection (one or more of temperature instability, change in inotrope requirements, haemodynamic instability or poor perfusion). Any positive blood culture was accepted for a non-skin organism, but for skin organisms two or more positive cultures of the same organism were required within 48 hours of each other. A clinical committee reviewed all primary outcomes involving positive cultures.

We conducted a sensitivity analysis for potentially missing microbiology data by assuming that children with a record of clinical indication but no sample taken in the primary outcome time window did actually experience the primary outcome.

The main secondary outcomes were:

• CR-BSI: based on the same organisms cultured from blood and the CVC tip between 48 hours after randomisation and 48 hours after CVC removal; or differential positivity of cultures from multiple CVC lumens on two or more occasions; or BSI and exit site infection or BSI and CVC removed for suspected infection

• rate of BSI per 1000 CVC-days, based on one or more BSI between randomisation and CVC removal

• time to a composite measure of BSI consisting of the primary outcome or a negative blood culture combined with (1) a positive 16S PCR result for bacterial rRNA; (2) removal of the CVC because of suspected infection; or (3) start of antibiotics or change in type of antibiotics on the same or next day.

Other secondary outcomes were:

• time to CVC thrombosis (defined by two episodes within 5 days of each other of difficulty flushing the CVC or drawing back blood from the CVC, one episode of swollen limb, CVC removal because of thrombosis or a positive ultrasound indicating thrombosis)

• time to CVC removal

• mortality by 30 days

• length of PICU admission

• length of hospital stay (up to 6 months post randomisation)

• type of bacteria or fungi isolated from the BSI included in the primary outcome.

CVC-related outcomes evaluated in the safety analyses were:

• CVC-related adverse events (unexplained thrombocytopenia after insertion of the CVC, exit site infection, hypersensitivity, trauma from line insertion, line displacement, line breakage/mechanical problem/manufacture complication)

• mortality recorded up until hospital discharge

• antibiotic resistance to minocycline (> 0.5 µg/ml) or rifampicin (> 1.0 µg/ml).

Antibiotic resistance outcomes were based on Etest^® strips [see www.biomerieux-diagnostics.com/etest (accessed 20 November 2015)] applied to organisms isolated from the BSI included in the primary outcome. Incomplete laboratory testing and reporting prevented analysis of resistance in cultures from the CVC tip (as specified in the protocol).

Sample size

We based the sample size calculation for the primary analysis on a RR. We assumed that detection of a RR of 0.5 in patients with a baseline risk of 10% would change policy. We assumed that the RR would remain relatively constant across baseline risks whereas the absolute risk difference would be more variable. A total of 1200 children were required in a 2 : 1 ratio (impregnated : standard) to achieve 80% power to detect a RR of 0.5 at a 5% level of significance, based on an estimated BSI rate of 10% and allowing for 5% loss to follow-up. A lower than expected BSI rate of 5% would have 62% power to detect a RR of 0.5 or 80% power for a RR of 0.32.

The Independent Data and Safety Monitoring Committee (IDSMC) recommended continuation of the study after (1) reviewing the first 209 children; (2) an interim analysis of 650 children using the Peto–Haybittle stopping rule for the primary outcome; and (3) recruitment had reached the original target of 1200 pre schedule in June 2012, before exhausting available funding (see Acknowledgements, Trial Oversight Committees and Table 23).

Statistical methods

Outcome data were analysed according to the intention-to-treat principle, meaning that children who were consented and randomised were analysed according to the type of CVC randomised, regardless of whether or not CVC insertion was attempted or the type of CVC received. Safety analyses included the subset of children for whom CVC insertion was attempted, grouped by CVC actually received.

The statistical analysis plan was developed prior to analysis and is available in Appendix 1. A 5% level of statistical significance and 95% confidence intervals (CIs) were used throughout. Absolute risk differences were calculated for proportions. Time-to-event outcomes were analysed using Kaplan–Meier curves and the log-rank test. Cox regression was used to adjust the primary analysis of time to BSI for the use of prospective or deferred consent and suspected infection at baseline. Poisson regression was used to analyse the secondary outcome of rate of BSIs (defined as the total number of BSIs per 1000 CVC-days occurring between randomisation and CVC removal). All analyses were conducted using SAS software version 9.3 (SAS Institute Inc., Cary, NC, USA).

Post hoc analyses evaluated competing risks from death or time to first BSI, using cumulative incidence curves. We applied Gray’s test to detect whether or not there was a difference between impregnated and standard CVCs for the primary outcome. 59 This analysis was conducted using R statistical software (The R Foundation for Statistical Computing, Vienna, Austria).

Study oversight and role of funders

The Research Ethics Committee for South West England approved the study protocol. The manufacturer Cook supplied CVCs to participating units at a 20% discounted price. Neither the manufacturer nor the funder (the National Institute for Health Research) had any role in the design of the study, the collection or interpretation of the data or the reporting of the results. The CATCH trial is registered with ClinicalTrials.gov (NCT01029717). The protocol is available at www.nets.nihr.ac.uk/projects/hta/081347 (accessed 20 November 2015) and the statistical analysis plan is provided in Appendix 1.

Study population

In total, 1859 children were randomised, of whom 501 children were randomised prospectively and 1358 were randomised as an emergency. Of those randomised as an emergency, 984 subsequently provided deferred consent for inclusion in the analyses (Figure 1; see Appendix 2, Figures 11 and 12 for participant flow by emergency/elective randomisation). Reasons for non-consent in the deferred consent group included not approached [n = 180 (48%), mainly because of transfer to a non-participating unit or early discharge from the PICU], no response [n = 17 (4.5%)] or consent refused [n = 177 (47%)]. Detailed reasons for non-consent are reported elsewhere. 60 Numbers enrolled by site and by month are provided in Appendix 2 (see Table 24 and Figure 13).

FIGURE 1.

Consolidated Standards of Reporting Trials (CONSORT) flow diagram for all trial participants. a, Based on clinically indicated blood culture sample taken ≥ 48 hours after randomisation and < 48 hours after CVC removal. ITT, intention to treat.

Comparison of interventions

The intention-to-treat sample included 1485 children, of whom 1345 children received the allocated CVC. Threats to validity because of protocol deviations are provided in Appendix 2 (see Table 26). Very few children had a clinical indication but no blood culture taken in the primary outcome time window (see Figure 1). Timings of samples for positive BSIs included in the primary and secondary outcomes are provided in Table 1.

Primary outcome

The number of blood samples contributing to the primary outcome is shown in Appendix 2 (see Figure 14). Blood cultures were taken between 48 hours after randomisation and CVC removal for 40% of those randomised (593/1485; see Figure 1). BSI was recorded for 42 children [standard 18/502 (3.6%); antibiotic 7/486 (1.4%); heparin 17/497 (3.4%)]. Gram-positive organisms accounted for the majority of BSIs (Table 5).

TABLE 5 - Primary outcome (absolute measures) and type of organism isolated, according to CVC allocation

IQR, interquartile range.

n number of participants by randomised CVC unless otherwise stated.

Includes one mixed BSI pathogen and skin organism.

Subtotals add to more than the total in the heparin group because of multiple types of organisms isolated on the same occasion in some patients.

Includes skin bacteria.

Outcome	Standard (n = 502)		Antibiotic (n = 486)		Heparin (n = 497)
	n a	%	n a	%	n a	%
BSI	18	3.6	7	1.4	17	3.4
Median time to first BSI in days (IQR)	7.5	(4.5–11.2)	6.9	(6.0–8.0)	4.2	(3.1–8.4)
Organism type
Non-skin	15b	2.99	6	1.23	16	3.22
Skin	3	0.60	1	0.21	1	0.20
Organism groupc
Gram positived	10	1.99	3	0.62	10	2.01
Gram negative	6	1.20	4	0.82	5	1.01
Candida	2	0.40	0	0.00	3	0.60

Figure 2 shows the Kaplan–Meier curve for the primary outcome of time to first BSI. There was no significant difference in time to first BSI when comparing any impregnated CVC (antibiotic or heparin) with standard CVCs (Table 6). However, the risk of BSI was significantly lower for antibiotic compared with standard CVCs [hazard ratio (HR) 0.43, 95% CI 0.20 to 0.96] and for antibiotic compared with heparin CVCs (HR 0.42, 95% CI 0.19 to 0.93). The direction of these results was robust to the sensitivity analysis (see Appendix 2, Table 27). Regression analysis showed no significant effect of prespecified variables (type of consent and suspected infection at randomisation) and the effect of type of CVC was similar after adjusting for these variables (Table 7).

FIGURE 2.

Kaplan–Meier curve for time to first BSI by CVC allocation. 53

TABLE 6 - Risk difference for first BSI and HR for time to first BSI according to CVC allocation

Bold indicates differences that are significant at the 5% level.

Analysis	Comparison	Risk difference (95% CI)	HR (95% CI)	p-value
Primary	Any impregnated (n = 983) vs. standard (n = 502) CVC	–1.14 (–3.04 to 0.75)	0.71 (0.37 to 1.34)	0.29
Secondary	Antibiotic (n = 486) vs. standard (n = 502) CVC	–2.15 (–4.09 to –0.20)	0.43 (0.20 to 0.96)	0.04
	Heparin (n = 497) vs. standard (n = 502) CVC	–0.17 (–2.45 to 2.12)	1.04 (0.53 to 2.03)	0.90
	Antibiotic (n = 486) vs. heparin (n = 497) CVC	–1.98 (–3.90 to –0.06)	0.42 (0.19 to 0.93)	0.03

TABLE 7 - Regression results for the primary outcome

HRs at p < 0.05 are in bold.

Participants with prospective consent were admitted electively and participants with deferred consent were admitted as an emergency.

Analysis	Variable (n with outcome)	Comparator (n with outcome)	HRa (95% CI)	p-value
Primary	Antibiotic or heparin CVC (24)	Standard CVC (18)	0.71 (0.38 to 1.33)	0.29
	Deferred consentb (30)	Prospective consentb (12)	0.87 (0.40 to 1.90)	0.73
	Suspected infection(18)	No suspected infection (24)	0.69 (0.33 to 1.42)	0.31
Secondary	Antibiotic CVC (7)	Standard CVC (18)	0.40 (0.17 to 0.96)	0.04
	Heparin CVC (17)	Standard CVC (18)	1.05 (0.54 to 2.05)	0.89
	Deferred consentb (30)	Prospective consentb (12)	0.87 (0.40 to 1.90)	0.35
	Suspected infection (18)	No suspected infection (24)	0.68 (0.33 to 1.40)	0.30
Secondary	Antibiotic CVC (7)	Heparin CVC (17)	0.39 (0.16 to 0.95)	0.04
	Deferred consentb (30)	Prospective consentb (12)	0.85 (0.30 to 2.45)	0.76
	Suspected infection (18)	No suspected infection (24)	0.99 (0.40 to 2.43)	0.98

Competing risk analysis using Gray’s test indicated no difference between the treatments for either competing risk (p-values of p = 0.29 for BSI and p = 0.89 for death; Table 8).

TABLE 8 - Competing risk analysis for primary outcome of time to first BSI

Outcome	HR (95% CI)	Gray’s test p-value
Time to first BSI (hours)	0.71 (0.39 to 1.31)	0.29
Time to death (hours)	1.08 (0.63 to 1.85)	0.89

Introduction

Central venous catheter infections are a substantial and preventable cause of iatrogenic morbidity, mortality, excess length of stay and health-care costs. In the setting of the PICU, BSIs related to CVCs have been reported to occur in 3–8% of all CVC insertions. As approximately two-thirds of the 16,000 admissions to English PICUs each year57 require CVCs, the overall impact represents a major burden to patients and the NHS. 20,21

Impregnated CVCs are nearly twice as expensive as standard CVCs, requiring decisions on their use to be informed by evidence of their cost-effectiveness. However, current economic evaluations are limited in their transferability to the PICU setting in the UK as they all relate to adult populations and, with one exception,29 apply to different health-care systems (Australia,61 Germany62 and the USA63–65). Although care pathways and costs may differ in the UK setting, these studies consistently demonstrated antibiotic-impregnated CVCs to be cost saving while yielding improved outcomes.

Hockenhull et al. 29 modelled the cost-effectiveness of impregnated CVCs compared with standard CVCs in adult patients. The cost of managing CR-BSIs, estimated as £9148, was taken from a systematic review of economic studies. Based on a systematic review of RCTs, impregnated CVCs were estimated to reduce the incidence of CR-BSIs from 3% to 1.4%. The incremental cost-effectiveness ratio (ICER) of £8530 saved for each CR-BSI averted was calculated as the additional cost of the impregnated CVC less the expected cost per patient of managing excess CR-BSIs divided by the absolute risk reduction. Although intuitively simple, the model did not consider mortality effects or discriminate between different types of impregnated CVCs and the authors recommended that decision-makers interpret the results with caution.

Halton et al. 61 used a Markov decision model to compare the cost-effectiveness of a range of antimicrobial-coated CVCs, including minocycline- and rifampicin-coated catheters, relative to uncoated catheters in adult intensive care unit patients. Simulations suggested that antibiotic CVCs prevented 15 CR-BSIs per 1000 CVCs placed, with a corresponding gain of 1.6 quality-adjusted life-years (QALYs). The model predicted that 32 intensive care unit (ICU) bed-days and 95 general ward bed-days would be released, with a cost saving of AU$130,289 per 1000 CVCs.

Frank et al. 62 performed a case–control analysis of resource use and costs among 30 adults who developed a CR-BSI and 108 control subjects, each in an ICU setting. The marginal cost per infectious episode was estimated as €231, but the calculation and meaning of the ICER presented for silver-impregnated CVCs were unclear.

Marciante et al. 63 developed a series of decision models with patient-level clinical trial data to determine whether or not minocycline- and rifampin-impregnated CVCs are cost-effective in adults. Cost-effectiveness was indeterminable for CVCs inserted for ≤ 1 week as no infections had occurred during this time. Antibiotic CVCs were modelled to be cost-effective for longer periods of insertion, with expected savings of US$67 and gains of 0.009 QALYs per patient.

Shorr et al. 64 presented another decision-analytic model based on a hypothetical cohort of 1000 adult patients requiring a CVC. The incidence of CR-BSIs, excess lengths of ICU and ward stays and associated costs were selected from published studies. Compared with standard CVCs, minocycline- and rifampin-impregnated CVCs were estimated to reduce the incidence of CR-BSIs from 3.3% to 1.4%, resulting in a saving of US$9605 for each CR-BSI averted.

Veenstra et al. 65 used data from RCTs, meta-analyses, and case–control studies within a decision-analytic modelling framework to estimate the incremental cost-effectiveness of antiseptic-impregnated CVCs in a hypothetical cohort of hospitalised patients at high risk for CR-BSIs. Modelling the use of chlorhexidine/silver sulfadiazine-impregnated compared with standard CVCs resulted in a 2.2% decrease in the incidence of CR-BSIs, a 0.33% decrease in the incidence of death and a saving of US$196 per CVC used.

An important limitation of these studies was that each analysis modelled the costs and consequences of BSIs using data from disparate sources and as such relied heavily on assumptions relating to attribution of hospital lengths of stay (the main cost driver) and mortality to BSIs. The only UK-based economic evaluation considered an adult population and assumed that a patient with a CR-BSI spends 6 additional days in the ICU and 5 additional days in a general medical ward. 29 A recent study of 1339 cases of CR-BSI sampled from a US paediatric population and matched to control subjects by propensity score revealed a higher mean attributable length of stay of 19 days. 66 Although this is comparable with the 21-day excess length of stay estimated for paediatric haematology/oncology patients,67 these estimates are reliant on retrospective observational data and are susceptible to bias.

Aim

We aimed to assess the cost-effectiveness of antibiotic, heparin and standard CVCs in an English PICU setting using data from the CATCH RCT. Although the primary comparison showed no evidence that impregnated CVCs (antibiotic or heparin) were more effective than standard CVCs, important differences in secondary comparisons among the three CVCs suggested that an economic evaluation was warranted to inform decisions on resource allocation. This would be especially relevant if one type of CVC were to reduce total costs, be associated with shorter periods in the PICU or reduce the length of ward stays.

Methods

Although cost–utility analyses, based on QALYs, are more appropriate for informing decisions concerning allocative efficiency, there are practical and methodological challenges in estimating utility values in children, especially very young children in the PICU setting. These include difficulties in responding to or understanding questions on health-related quality of life, whether for reasons of age, illness or consciousness; the limitations of using proxy utilities; the low event rate for the primary end point; and the inclusion of a wide range of clinical conditions. A cost-effectiveness analysis was therefore performed, which allowed for an assessment of technical efficiency (i.e. determination of the most efficient CVC for reducing the incidence of BSIs). The study methods were consistent with those used in other economic evaluations of CVCs. 62

Resource use

The perspective of the analysis was that of the NHS in England, with the expectation that the main cost driver of inpatient hospital care would represent the greatest proportion of direct medical costs. The principal cost components were PICU, high-dependency unit (HDU) and ward stays (including readmissions), outpatient clinic visits, accident and emergency (A&E) admissions and the CVCs. The time horizon of the base-case analysis was selected to include the costs associated with managing BSIs and any sequelae within the 6-month period from randomisation. Shorter time horizons were examined in sensitivity analyses.

The measurement of resource use required complementary approaches using data collected as part of the trial and as part of routine care. Patients’ use of hospital services was obtained from the following sources (Figure 3):

1. The trial CRFs. Research nurses completed the relevant sections of the CRF to record the dates during which patients were in neonatal intensive care units (NICUs) or PICUs, HDUs and paediatric wards within the hospitals participating in the CATCH trial. Data recorded on CRFs were used for the dates of hospital discharge, transfer to another hospital and CVC removal.

2. Hospital Episode Statistics (HES) data from the Health and Social Care Information Centre. 68 HES data contain details of all admissions to NHS hospitals in England and provide Healthcare Resource Groups (HRGs) for the type of care patients receive at a ward level, outpatient visits and A&E admissions, but do not provide details on ICU and HDU stays. HES data were used for estimating HRGs for ward stays, outpatient and A&E attendances.

3. The PICANet data set. 57 This data set includes all ICU length of stays for paediatric patients in the UK and allows for the tracking over time of patients who have been transferred between ICUs in different hospitals. PICANet data were used for the national schedule of reference costs HRGs for HDU and ICU stays69 and for checking hospital admission, transfer and discharge dates.

4. Hospital Patient Administration Systems (PASs) of CATCH-participating hospitals. These were accessed for patient lengths of stay on ICUs and wards and for relevant HRGs. These were used to supplement data that were missing from other sources.

FIGURE 3.

Flow diagram of the methods employed for the economic evaluation.

Results

Resource use and total costs

Complete cost data were available for all patients. In the 6 months preceding randomisation, the mean costs (length of stay) of ICU/HDU admissions were £6026 (3.19 days) for the standard CVC group, £5188 (2.76 days) for the antibiotic CVC group and £6616 (3.47 days) for the heparin CVC group. The mean total hospital costs for the corresponding period were £15,588, £16,933 and £16,722, respectively. Neither ICU/HDU costs nor total hospital costs differed by intervention group.

Patients randomised to antibiotic-impregnated CVCs spent a mean of 10.8 days (95% CI 9.3 to 12.5) in the PICU in the 6 months following randomisation compared with 9.9 days (95% CI 8.6 to 11.4) for those in the heparin-bonded CVC group and 10.5 days (95% CI 9.2 to 11.9) for those in the standard CVC group (Table 15). There were no significant differences between groups in length of stay in the PICU (p = 0.70), HDU (p = 0.43) or ward (p = 0.52). The mean total hospital stay in the 6 months after randomisation was 34.8 days (95% CI 31.2 to 38.5 days) for antibiotic CVCs, 31.4 days (95% CI 28.2 to 34.7 days) for heparin-bonded CVCs and 31.7 (95% CI 28.8 to 34.8 days) for standard CVCs. The six most significant HRGs (of 349 in total) accounted for 50% of ward costs. These related to surgical correction of congenital malformations, cardiac surgery or disorders of the lower respiratory tract.

TABLE 15 - Patient length of stay and count of dominant HRGs relating to inpatient stays from randomisation to 6 months (including readmissions) according to place and intensity of care and intervention group

ECLS, extracorporeal life support; ECMO, extracorporeal membrane oxygenation.

Unit	Antibiotic		Heparin		Standard
	Mean (median)	95% CI	Mean (median)	95% CI	Mean (median)	95% CI
Days on ICU	10.79 (5.00)	9.28 to 12.48	9.91 (5.00)	8.57 to 11.44	10.50 (5.00)	9.17 to 11.93
Paediatric Critical Care, Intensive Care, ECMO/ECLS (XB01Z)	0.31 (0.00)	0.07 to 0.72	0.39 (0.00)	0.09 to 0.80	0.41 (0.00)	0.17 to 0.72
Paediatric Critical Care, Intensive Care, Advanced Enhanced (XB02Z)	0.16 (0.00)	0.09 to 0.26	0.12 (0.00)	0.09 to 0.15	0.16 (0.00)	0.10 to 0.26
Paediatric Critical Care, Intensive Care, Advanced (XB03Z)	0.77 (0.00)	0.51 to 1.05	0.62 (0.00)	0.43 to 0.83	0.65 (0.00)	0.46 to 0.87
Paediatric Critical Care, Intensive Care, Basic Enhanced (XB04Z)	2.30 (0.49)	1.92 to 2.72	2.69 (0.78)	2.09 to 3.44	2.76 (0.00)	2.14 to 3.54
Paediatric Critical Care, Intensive Care, Basic (XB05Z)	6.96 (2.00)	5.65 to 8.45	5.63 (2.00)	4.75 to 6.59	6.40 (2.95)	5.42 to 7.47
Neonatal Critical Care, Intensive Care (XA01C)	0.29 (0.00)	0.10 to 0.55	0.46 (0.00)	0.13 to 1.03	0.11 (0.00)	0.04 to 0.20
Days on HDU	2.00 (0.59)	1.48 to 2.62	1.60 (0.59)	1.28 to 1.99	1.73 (0.00)	1.44 to 2.05
Paediatric Critical Care, High Dependency, Advanced (XB06Z)	1.28 (0.00)	0.94 to 1.70	1.09 (0.00)	0.80 to 1.45	1.22 (0.00)	0.98 to 1.49
Paediatric Critical Care, High Dependency (XB07Z)	0.72 (0.00)	0.42 to 1.16	0.51 (0.00)	0.40 to 0.64	0.51 (0.00)	0.40 to 0.64
Days on ward	22.01 (9.13)	19.26 to 24.80	19.85 (9.00)	17.40 to 22.40	19.48 (8.57)	17.12 to 21.94
Total days in hospital	34.80 (20.00)	31.21 to 38.48	31.36 (17.00)	28.18 to 34.65	31.72 (17.97)	28.75 to 34.81
Count of non-PICU/-HDU inpatient HRGs
Complex Congenital Surgery (EA24Z)	100		103		109
Intermediate Congenital Surgery (EA25Z)	68		70		72
Major Complex Congenital Surgery (EA23Z)	45		39		37
Cardiac Conditions with Complication and Comorbidity (PA23A)	109		102		74
Lower Respiratory Tract Disorders without Acute Bronchiolitis with Length of Stay ≥ 1 day with Complication and Comorbidity (PA14C)	95		78		105
Implantation of Prosthetic Heart or Ventricular Assist Device (EA43Z)	2		2		4
Other inpatient HRGs	1103		1055		964

Total and disaggregated costs are presented in Table 16. The mean 6-month cost was £44,503 (median £28,952, range £1786–360,983, 95% CI £40,619 to £48,666) for standard CVCs, £45,663 (median £29,793, range £2189–442,365, 95% CI £41,647 to £50,009) for antibiotic-impregnated CVCs and £42,065 (median £27,621, range £2638–382,431, 95% CI £38,322 to £46,110) for heparin CVCs (Figure 4). These costs were not statistically significantly different between intervention groups (p = 0.46) or when disaggregated according to bundled costs (p = 0.43) and unbundled costs (p = 0.73).

TABLE 16 - Disaggregated and total costs (£) by intervention group from randomisation to the end of the 6-month time frame

ECLS, extracorporeal life support; ECMO, extracorporeal membrane oxygenation.

National Schedule of Reference Costs 2012–13. 69

Top six (of 349) HRGs ranked by cost, together contributing 50% of overall inpatient costs.

2012–13 national tariff HRGs: < 1% taken from bed-day rates.

Costs supplied by CVC provider (Cook Medical Inc.).

Unit	Antibiotic		Heparin		Standard
	Mean (median)	95% CI	Mean (median)	95% CI	Mean (median)	95% CI
Paediatric critical care, intensive care
ECMO/ECLS (XB01Z)	1358 (0)	310 to 3159	1703 (0)	386 to 3509	1796 (0)	723 to 3156
Advanced Enhanced (XB02Z)	388 (0)	207 to 636	289 (0)	216 to 371	395 (0)	228 to 620
Advanced (XB03Z)	1545 (0)	1031 to 2124	1250 (0)	872 to 1674	1318 (0)	933 to 1752
Basic Enhanced (XB04Z)	4861 (1023)	4060 to 5738	5675 (1646)	4418 to 7260	5822 (0)	4512 to 7460
Basic (XB05Z)	12,137 (3486)	9855 to 14,730	9822 (3486)	8274 to 11,489	11,159 (5133)	9440 to 13,025
Neonatal Critical Care, Intensive Care (XA01C)	325 (0)	113 to 613	517 (0)	142 to 1150	125 (0)	42 to 225
Paediatric critical care, high dependency
High Dependency, Advanced (XB06Z)	1709 (0)	1254 to 2271	1450 (0)	1972 to 1940	1629 (0)	1301 to 1992
High Dependency (XB07Z)	635 (0)	372 to 1025	454 (0)	354 to 567	456 (0)	356 to 566
Transportation (XB08Z)	1158 (0)	1022 to 1293	1258 (0)	1109 to 1413	1208 (0)	1068 to 1353
Subtotal (PICU/HDU/NICU)a	24,115 (12,201)	20,824 to 27,764	22,417 (11,903)	19,429 to 25,771	23,907 (12,495)	20,989 to 27,049
Inpatient stayb
Complex Congenital Surgery (EA24Z)	3011 (0)	2445 to 3593	2908 (0)	2363 to 3481	3144 (0)	2565 to 3753
Intermediate Congenital Surgery (EA25Z)	2166 (0)	1670 to 2699	1934 (0)	1470 to 2440	2044 (0)	1583 to 2545
Major Complex Congenital Surgery (EA23Z)	1865 (0)	1315 to 2481	1915 (0)	1310 to 2603	1466 (0)	1013 to 1960
Cardiac Conditions with Complication and Comorbidity (PA23A)	1277 (0)	818 to 1845	1173 (0)	831 to 1558	739 (0)	495 to 1025
Lower Respiratory Tract Disorders without Acute Bronchiolitis with Length of Stay ≥ 1 day with Complication and Comorbidity (PA14C)	858 (0)	593 to 1157	668 (0)	454 to 913	943 (0)	657 to 1268
Implantation of Prosthetic Heart or Ventricular Assist Device (EA43Z)	273 (0)	0 to 684	298 (0)	0 to 762	548 (0)	103 to 1155
Other inpatient HRG costs	10,316 (4017)	8616 to 12,231	8803 (3058)	7524 to 10,106	9930 (3259)	7860 to 12,409
Subtotal (inpatient)	19,766 (14122)	17,934 to 21,755	17,700 (13,716)	16,308 to 19,182	18,814 (13,748)	16,649 to 21,327
Other
A&E costc	89 (0)	76 to 104	85 (0)	73 to 99	91 (0)	78 to 104
Outpatient costc	1615 (883)	1412 to 1838	1784 (837)	1496 to 2109	1648 (881)	1453 to 1871
CVC costd	78.28	78 to 78	78.25	78 to 78	42.91	43 to 43
Total cost (full 6 months)	45,663 (29,793)	41,647 to 50,009	42,065 (27,621)	38,322 to 46,110	44,503 (28,952)	40,619 to 48,666

FIGURE 4.

Ranking of 6-month total costs by intervention group, indicating patients who experienced a BSI. (a) Antibiotic CVCs; (b) heparin CVCs; and (c) standard CVCs.

Incremental and uncertainty analysis

As heparin CVCs were shown not to be clinically effective compared with standard CVCs there is no case for an incremental analysis: a clinically ineffective intervention cannot be cost-effective according to the same measure of BSI. The calculation of the ICER was therefore limited to the comparison between antibiotic and standard CVCs, which resulted in an ICER of £54,057 per BSI averted (Table 19).

TABLE 19 - Incremental analysis of unadjusted costs (mean values with 95% central range)

As heparin CVCs were not deemed to be clinically effective in reducing BSI rates, they cannot be cost-effective for the same outcome measure.

Cost saving.

Analysis	Antibiotic	Heparin	Standard
Base-case analysis (6-month time horizon)
Total cost (£)	45,663 (41,647 to 50,009)	42,064 (38,322 to 46,110)	44,503 (40,619 to 48,666)
Incremental cost (vs. standard) (£)	1160 (–4743 to 6692)	–2438 (–8164 to 3359)	–
BSI (%)	1.44 (0.4 to 2.5)	3.42 (1.8 to 5.0)	3.59 (2.0 to 5.2)
Incremental BSI (vs. standard) (%)	–2.15 (–4.1 to –0.2)	–0.17 (–2.5 to 2.1)	–
ICER (vs. standard) (£)	54,057 per BSI averted	–a	–
Sensitivity analysis (index hospitalisation)
Total cost (£)	33,073 (30,047 to 36,337)	32,245 (29,013 to 35,823)	35,165 (31,864 to 38,670)
Incremental cost (vs. standard) (£)	–2093 (–6919 to 2583)	–2920 (–7833 to 2180)	–
BSI (%)	1.44 (0.4 to 2.5)	3.42 (1.8 to 5.0)	3.59 (2.0 to 5.2)
Incremental BSI (vs. standard) (%)	–2.15 (–4.1 to –0.2)	–0.17 (–2.5 to 2.1)	–
ICER (vs. standard) (£)	–97,543 per BSI avertedb	–a	–

The CEAC yielded probabilities of 0.38, 0.49 and 0.62 of antibiotic CVCs being cost-effective at (arbitrary) thresholds of £10,000, £50,000 and £100,000 per BSI averted respectively (Figure 5). The probability of antibiotic CVCs dominating standard CVCs was estimated as 0.35.

FIGURE 5.

Cost-effectiveness acceptability curve based on a 6-month time horizon presenting the probabilities of antibiotic and standard CVCs being cost-effective for given values of ceiling ratio expressed as cost per BSI averted.

Sensitivity analysis

The mean number of days in hospital during the index hospitalisation was substantially shorter (e.g. 22.1 days for antibiotic CVCs) than the mean number of days in hospital during the 6 months from randomisation (e.g. 34.8 days for antibiotic CVCs; Table 20 and see Table 15). Considering only the index hospitalisation, total costs tended to be lower in the antibiotic CVC group (£33,073, 95% CI £30,047 to £36,337) and in the heparin CVC group (£32,245, 95% CI £29,013 to £35,823) than in the standard CVC group (£35,165, 95% CI £31,864 to £38,670). The unadjusted incremental cost saving for antibiotic compared with standard CVCs was –£2093 (95% CI –£6919 to £2583) and for heparin compared with standard CVCs was –£2920 (95% CI –£7833 to £2180).

TABLE 20 - Patient length of stay for hospitalisation episode from randomisation by intervention group

Unit	Antibiotic		Heparin		Standard
	Mean	95% CI	Mean	95% CI	Mean	95% CI
Days on ICU	9.31	8.09 to 10.70	8.93	7.71 to 10.32	9.79	8.60 to 11.03
Days on HDU	1.70	1.25 to 2.25	1.39	1.09 to 1.76	1.51	1.24 to 1.80
Days on ward	11.13	9.19 to 13.18	10.32	8.59 to 12.18	10.79	9.03 to 12.70
Total days in hospital	22.14	19.48 to 24.89	20.65	18.27 to 23.16	22.09	19.76 to 24.51

Based only on the costs of the index stay, antibiotic CVCs dominated standard CVCs with a saving of £97,543 per BSI averted (see Table 19).

An analysis of the cumulative mean costs over the course of the 6 months (Figure 6) shows that costs in the heparin CVC group were lower overall, whereas costs in the antibiotic CVC group were variably cost incurring and cost saving in comparison to costs in the standard CVC group.

FIGURE 6.

Relation between total costs (cumulative) and time since randomisation according to intervention group.

The resulting ICER for antibiotic compared with standard CVCs fluctuated considerably (Figure 7), ranging from £82,204 saved per BSI averted by day 50 post randomisation to being cost neutral by day 122 and to the base-case cost of £54,057 per BSI averted by 6 months.

FIGURE 7.

Relation between the ICER for antibiotic CVCs compared with standard CVCs and time since randomisation. Positive ICERs are cost incurring and negative ICERs represent incremental savings per BSI averted.

Introduction

The CATCH trial was the largest trial carried out in PICUs to date, recruiting 1485 children within 14 PICUs in 12 NHS trusts in England, corresponding to 5% of children admitted to all PICUs in England and Wales during the trial period (2010–12). However, if antibiotic-impregnated CVCs were to be adopted, it is likely that these CVCs would be bulk purchased and used for all children requiring CVCs in PICUs, not just children like those in the trial. Those making decisions on whether or not to purchase antibiotic-impregnated CVCs therefore need to take into account the generalisability of benefits to all children who need a CVC and the cost impact of purchasing the more expensive impregnated CVCs.

In terms of generalisability, trial populations may have different characteristics and outcomes from the characteristics and outcomes of those who receive the intervention in practice, for a variety of reasons. 76 In the CATCH trial there were two specific reasons why those recruited might differ from those likely to receive impregnated CVCs outside the trial setting. First, children recruited to the CATCH trial were expected to require a CVC for ≥ 3 days and would therefore have a higher risk of BSI than those staying for < 3 days. Second, the introduction of CVC care bundles and ongoing improvements in infection control in recent years have been associated with rapidly decreasing rates of BSI over the past decade, meaning that the background BSI rate may be lower now than it was at the start of the trial. 33,34

In terms of budget impact, impregnated CVCs are approximately twice as expensive as standard CVCs. However, the additional costs might be outweighed by the number of BSIs averted through using the more effective CVCs and the associated reduction in the use of health-care resources.

We determined the generalisability of the CATCH trial findings by estimating risk-adjusted trends in the rate of BSIs for children expected to require CVCs in PICUs, based on a data linkage study including children not participating in the CATCH trial. 77 We determined the budget and cost impacts of adopting antibiotic-impregnated CVCs for all children requiring a CVC in the PICU by synthesising the following evidence: (1) the estimated risk of BSI using standard CVCs (derived from the data linkage study); (2) the number of BSIs potentially averted by using antibiotic-impregnated CVCs (based on the relative treatment effect in the trial); (3) the additional costs associated with purchasing impregnated CVCs for all children expected to require a CVC (numbers of CVCs based on PICU survey data); and (4) the value of the health-care resources associated with each BSI (from the CATCH cost-effectiveness analysis).

Methods

Results

Rate of bloodstream infections using standard central venous catheters

Of the 2488 admissions in the CVC audit data, 1431 (58%) required a CVC. The best-fitting prediction model included length of stay, vasoactive agent, admission from ward, renal support and invasive ventilation (see Appendix 4, Table 37). With a probability cut-off of 0.57, the sensitivity of the predictive model for capturing admissions requiring a CVC was 61%, specificity was 82%, the positive predictive value was 82% and the negative predictive value was 61%. The predictive model identified 80% of the CATCH admissions as requiring a CVC.

Survey responses for the types of CVCs used prior to the CATCH trial were obtained for 18 of the 23 PICUs in England (see Appendix 4, Table 36). Only two PICUs reported not using standard CVCs for any admissions (both used heparin CVCs). BSI rates were estimated based on linked data from the remaining 16 English PICUs.

Applying the predictive model to the 16 PICUs in the linked data set identified a subset of 21,381 admissions most likely to have received a standard CVC between 2003 and 2012. The characteristics of these admissions (based on PICANet data) are provided in Appendix 4 (see Table 38). Risk-adjusted rates of BSI using standard CVCs decreased steadily between 2003 and 2012 and were greater for CATCH PICUs (5.27, 95% CI 5.06 to 5.49 per 1000 CVC-days in 2012) than for non-participating PICUs (2.09, 95% CI 1.60 to 2.58 per 1000 CVC-days in 2012) (Figure 8). Of the subset of admissions predicted to receive a CVC in 2012, 103 out of 3021 (3.4%) experienced a BSI, corresponding to an overall BSI rate using standard CVCs of 4.58 (95% CI 4.42 to 4.74) per 1000 CVC-days (Table 21). This was non-significantly lower than the rate observed during the trial (8.24, 95% CI 4.7 to 11.8 per 1000 CVC-days; see Table 9), partially because of the inclusion of all children with CVCs (not just those requiring CVCs for ≥ 3 days). Further explanations for this difference are the potentially incomplete reporting of BSIs to the national infection surveillance system, use of bed-days instead of CVC-days in the estimated rate and the increased frequency of sampling in trial PICUs during the CATCH trial.

FIGURE 8.

Risk-adjusted rates of BSIs for children expected to have a CVC based on linked PICANet–LabBase2 data for 16 PICUs in England. Symbols = observed rates; lines = smoothed adjusted rates (log scale).

TABLE 21 - Parameter estimates for the cost impact and sensitivity analyses

Variable	Base case	Source	Sensitivity analysis
BSI rate using standard CVCs in 2012 per 1000 CVC-days	4.58 (95% CI 4.42 to 4.74)	3021 admissions in 15 PICUs – subset of admissions identified as most likely to have received a standard CVC by applying the predictive model to the linked data set. Admissions identified by survey responses as receiving non-standard (heparin or antibiotic) CVCs were excluded	Random sample taken with replacement from the linked data set for the number of admissions expected to require a CVC
Rate ratio	0.40 (95% CI 0.17 to 0.97)	Trial clinical effectiveness analyses (see Table 10)	Log-normal distribution with parameters (–0.913, 0.415)
Estimated BSI rate using antibiotic CVCs in 2012 per 1000 CVC-days	1.83; worst case 4.29, best case 0.81	Rate ratio from the CATCH trial applied to the estimated BSI rate using standard CVCs for PICUs in England	Derived from (i) the BSI rate using standard CVCs and (ii) the rate ratio
Number of admissions requiring CVCs in 2012	8831	Average survey estimates for the percentage of emergency (60%) and elective (50%) admissions requiring CVCs, applied to all admissions in PICANet in 2012 (15,739 admissions in 23 PICUs)	Beta distributions with stated parameters: emergency: beta(60,40); elective: beta (50,50)
Number of CVC-days in 2012	85,971	Average bed-days per admission in the subset of admissions identified as most likely to have received a standard CVC by applying the predictive model to the linked data set, multiplied by the number of admissions requiring a CVC in 2012	Random sample taken with replacement from the linked data set for admissions expected to require a CVC
Number of BSIs averted in 2012	232	BSI rate applied to CVC-days for admissions requiring a CVC in 2012	Derived from (i) the number of admissions requiring a CVC in 2012 and (ii) the estimated BSI rate using antibiotic CVCs
Additional cost of antibiotic CVCs	£36	Difference in costs between standard (£43) and antibiotic (£79) CVCs (conservative case assuming triple-lumen CVCs used for all children)	Fixed at £36
Costs associated with managing each BSI	£10,975 (95% CI –£2801 to £24,751)	CATCH trial cost-effectiveness analysis (see Table 18)	Normal distributions with parameters (£10,975, £7,023)

Cost impact: value of additional costs associated with managing BSI

Based on each BSI being associated with a mean cost of £10,975 (95% CI –£2801 to £24,751; see Table 18), over 6 months the value of resources made available in 2012 through averting BSIs associated with standard CVCs (i.e. the total costs of managing these BSIs) would have been 232 × £10,975 = £2,541,397, with best- and worst-case scenarios of –£648,606 and £5,731,401 based on CIs for both estimates. The probabilistic sensitivity analysis provided a 95% uncertainty interval of –£66,544 to £5,557,451 for the total resources made available through averting BSIs in 2012. There was a probability of 0.90 that the value of resources made available would be more than the additional costs of purchasing antibiotic CVCs (Figure 9).

FIGURE 9.

Probability distribution for the value of resources made available by averting BSIs using antibiotic CVC in all PICUs in England during 2012. In total, 90% of the distribution represented costs greater than the additional cost of purchasing antibiotic CVCs.

The estimated cost impact for a typical PICU with 350 admissions per year is shown for a range of BSI rates in Table 22. Figure 10 shows that the cost of purchasing antibiotic CVCs for all children who require them will be less than the cost of managing BSIs with standard CVCs for PICUs with BSI rates > 1.2 per 1000 bed-days. This break-even value is substantially lower than the BSI rate observed in the standard arm of the trial (8.24, 95% CI 4.7 to11.8 per 1000 bed-days) or in the linked data set for PICUs in England (4.58, 95% CI 4.42 to 4.74 per 1000 bed-days).

FIGURE 10.

Cost impact: number of BSIs averted and value of resources made available using antibiotic in place of standard CVCs for a range of baseline BSI rates, assuming that each BSI is associated with a mean cost of £10,975.

Introduction

We aimed to inform NHS policy regarding impregnated CVCs for the intensive care of children. To address the question of whether or not impregnated CVCs should be adopted by PICUs in England and Wales, we undertook a large pragmatic RCT to determine the clinical effectiveness and cost-effectiveness of impregnated compared with standard CVCs. To determine the implications of adopting impregnated CVCs for all children who need them, we conducted a generalisability and cost impact study, using linked data from two national sources.

Clinical effectiveness

The primary analysis showed no evidence of a statistically significant difference in time to first BSI between any impregnated CVCs (antibiotic-impregnated and heparin-bonded CVCs combined) and standard CVCs. However, secondary analyses showed that antibiotic impregnation reduced the hazard of BSIs by 57% compared with standard CVCs and by 58% compared with heparin-bonded CVCs. Antibiotic-impregnated CVCs were associated with an absolute risk reduction of 2.15% compared with standard CVCs, meaning that 47 children would need to be treated with an antibiotic-impregnated CVC instead of a standard CVC to prevent one case of BSI.

Our choice of any BSI as a clinically important primary outcome and a recognised quality indicator is an important strength of our study, avoiding the biases inherent in measuring CR-BSI. 2,46,88,89 CR-BSI requires positive cultures from the blood and catheter tip and is highly susceptible to bias, as the tip can be easily contaminated during removal and residual antibiotic in the catheter tip may inhibit culture in the laboratory. 55,88

A further strength of the study is the restriction to positive blood cultures that were clinically indicated. This increased the clinical relevance of the primary outcome, but diminished the sensitivity of the study to detect bacteraemia, as only 40% of children had a blood culture taken in the relevant time window. A third strength is the representativeness of the study population in terms of children admitted to the 14 largest PICUs (of 23) across England. We were able to enrol a similar proportion of emergency patients (two-thirds) as seen in practice, enabled by the inclusion of retrieved children and the use of deferred consent. 90

A potential limitation of the study is that clinicians inserting the CVCs could not be blinded to allocation. However, we found no evidence of differential sampling by trial arm (see Figure 1). The number of children who received their allocated CVC was slightly higher for those in the standard arm, probably reflecting the fact that standard CVCs were the default CVC used in many units. 32 A second limitation is that, because of the lower than expected BSI rate in the standard arm of the trial, we had limited power to detect differences in the primary outcome comparing impregnated with standard CVCs. This choice of primary outcome was justified by the best available evidence to date – a systematic review and meta-analysis of direct and indirect comparisons of different types of impregnated and standard CVCs28 – which showed that heparin-bonded and antibiotic-impregnated CVCs resulted in similar reductions in the risk of CR-BSI (70–80%). A third limitation is that resistance testing was not standardised across sites. This reflects local laboratory administration and processing, which centralised testing of all positive cultures could have mitigated. When reported, resistance occurred in all trial arms, predominantly in Gram-negative isolates, as expected. The low rates are consistent with the previous lack of evidence for the emergence of resistance. 1

Few previous trials have reported the effectiveness of impregnated CVCs for any BSI. 45 However, the superiority of antibiotic-impregnated CVCs in children was consistent with the most recent systematic review reporting a pooled odds ratio for CR-BSI of 0.18 (95% CI 0.08 to 0.34). 28 Although our finding of a clinically important reduction in any BSI with antibiotic-impregnated CVCs (HR 0.25, 95% CI 0.07 to 0.09; p = 0.04) was based on a secondary comparison and should be viewed as exploratory, this result does add important evidence of the overall effectiveness of antibiotic-impregnated CVCs.

The finding that heparin CVCs were not effective for BSIs or CR-BSIs contradicts past evidence showing a pooled odds ratio for CR-BSI for heparin-bonded compared with standard CVCs of 0.20 (95% CI 0.06 to 0.44). 28 The difference in findings may reflect poor data quality in previous trials, highlighted by systematic reviews. Only one of the three trials comparing heparin with standard CVCs reported adequate concealment of randomisation, and this trial did not state whether or not clinicians were blinded to the intervention. 2 A further explanation for the discrepancy may be the low baseline event rate observed in the CATCH trial, which was conducted after implementation of CVC care bundles in PICUs to improve aseptic procedures during CVC insertion and maintenance. 32 It is conceivable that heparin CVCs are most effective in the context of high rates of surface colonisation, as they prevent thrombosis, which aids organism adherence to the CVC. Finally, the pairwise comparisons used to determine the most effective type of impregnation were not adequately powered to detect the anticipated small differences between antibiotic and heparin CVCs. However, our results suggest that antibiotic-impregnated CVCs can achieve further reductions in BSI rates, over and above that achieved by CVC care bundles. 33,34

Cost-effectiveness

The ICER of antibiotic-impregnated CVCs compared with standard CVCs was £54,057 per BSI averted over the 6 months after randomisation. Assuming that the health impact of a BSI is no greater (on average) than a reduction of 1 year of full health (i.e. 1 QALY), then, at the cost-effectiveness threshold of £30,000 per QALY, antibiotic CVCs may not represent a cost-effective alternative to standard CVCs in a PICU setting. However, there is considerable uncertainty surrounding this estimate, which is driven mainly by the time horizon of analysis.

The sensitivity analysis in which costs were restricted to the index hospital stay resulted in antibiotic CVCs dominating standard CVCs, with £97,543 saved for each BSI averted. Antibiotic CVCs therefore appear highly cost-effective when considering events and costs accruing over comparable periods.

A secondary analysis of the CATCH trial indicated that heparin CVCs were not clinically effective, with a risk difference for a first BSI of –0.17 (95% CI –2.45 to 2.12) compared with standard CVCs. It follows, therefore, that heparin CVCs cannot be cost-effective by the same measure. Theoretically, a cost-minimisation analysis might apply, to assess whether or not heparin CVCs are less costly overall than standard CVCs. However, heparin CVCs are more expensive than standard CVCs (in terms of unit prices) and, as the only difference among CVCs can be in BSI rates, any difference in total cost (which was not statistically significant) was caused by random variation. A cost-minimisation analysis might therefore lead to an erroneous conclusion that heparin CVCs are more cost-effective than standard CVCs.

Our economic evaluation benefits from being conducted alongside a pragmatic clinical trial, which is representative of current practice in the UK PICU setting. The evaluation utilises data from a definitive and unbiased comparison of impregnated and standard CVCs and was conducted robustly according to accepted methods of trial-based economic evaluations. 74 We used patient-level HES data to reflect the reimbursement costs for hospitals and multiple data sources to measure hospital resource use to ensure that cost data were complete.

However, there are limitations that affect the strength of our findings. First, the CATCH trial was not powered to determine cost differences between each of the three CVCs. As a consequence, the results are susceptible to random variation in costs between trial arms. Although hypothesis testing may be considered less relevant to decision making in the context of net benefits, the non-statistically significant differences in costs between groups translated to uncertainty in the joint distribution of costs and benefits such that, in the base-case analysis, antibiotic CVCs had a probability of 0.35 of dominating standard CVCs. 91 Mean total costs associated with heparin CVCs were lower than those for both antibiotic and standard CVCs, despite their ineffectiveness in avoiding BSIs compared with standard CVCs. Being a rare event, BSI costs were diluted compared with the overall costs relating to the intensive care of patients.

Second, the economic evaluation did not consider QALYs, which is the standard metric for informing decisions on resource allocation. This was because the estimation of utilities in paediatric ICU populations is empirically and conceptually challenging92,93 and because the main long-term consequence of BSIs, the impact on neurological outcomes, is poorly measured in children and was not measured in this trial. Short-term outcomes not considered in our economic analysis include mortality, antibiotic resistance and other adverse events. However, antibiotic resistance to minocycline or rifampicin did not differ by CVC allocation. There were no also no differences in 30-day mortality for antibiotic compared with standard CVCs (HR 0.96, 95% CI 0.61 to 1.51) or for heparin compared with standard CVCs (HR 0.65, 95% CI 0.40 to 1.07) and no differences in adverse events (see Table 11).

Assumptions regarding the time horizon of analysis represent a third limitation. The base-case, 6-month analysis was selected to include the costs of hospital readmissions in addition to the costs of the index hospitalisation and transfers that may have occurred subsequently. This was intended to capture the costs of managing any longer-term complications from BSIs, but, as the economic outcome was chosen to align with the primary clinical outcome, the health impacts of these complications were not included in the ICER. Consequently, as costs accrue over time with no corresponding change in the number of BSIs (these all occurred within 30 days), the ICER continued to increase over time.

Our findings are consistent with those of other studies in terms of the estimation of the costs associated with the management of BSIs. However, our ICER differs considerably and is inconclusive with regard to determining the cost-effectiveness of antibiotic CVCs. Published economic evaluations, including those that adopted a lifetime horizon of analysis, suggest dominance of antibiotic-impregnated CVCs over standard CVCs. One explanation for this discrepancy is in the methods of analysis. A decision-analytic model, based on a synthesis of data from various sources, is fundamentally different from a prospective RCT, in which differences between intervention groups are less evident, particularly in the context of rare events such as BSIs. In the evaluation by Hockenhull et al. ,29 for instance, the incremental cost saving of £138.20 per patient receiving an impregnated CVC was calculated as the additional cost of the antibiotic CVC less the expected cost per patient of managing excess BSIs. The equivalent calculation based on CATCH data for antibiotic CVCs results in a value of £200.08 saved for each antibiotic CVC used [(£78.28 – £42.91) – (£10,975 × 2.15%)]. Extending this further to calculate the ICER gives a value of £9326 saved per BSI averted (£200.08/2.15%), which differs appreciably from our base-case result [the differences between the ICER stated and the ICER calculated from the numbers in the text is due to rounding (difference in risk of BSI is 2.1453%)]. However, by analysing the data as a cohort study, separating the apparent costs of BSIs from the total costs relating to each intervention group, biases are likely to arise from assuming that the cost of managing BSIs is independent of CVC type.

In conclusion, the results of the cost-effectiveness analysis indicate that a policy of replacing standard CVCs with antibiotic-impregnated CVCs in paediatric ICUs would be more beneficial in terms of fewer patients developing BSIs. Given the low BSI rate, the variation in costs between arms and the sensitivity of analyses to the specified time horizon, there remains considerable uncertainty whether or not use of antibiotic CVCs represents a cost-effective use of NHS resources.

Generalisability and cost impact

We explored the generalisability of the CATCH trial results and the cost impact of changing practice in PICUs across England based on the trial results. In terms of generalisability, the observed rates of BSI using standard CVCs declined steadily over the past decade, including the period when children were enrolled into the CATCH trial. 34,94 In addition, children participating in the CATCH trial had a higher risk of BSI than all children receiving CVCs in practice, as they were expected to require a CVC for ≥ 3 days. This means that children currently receiving CVCs in PICUs are likely to have a lower BSI risk than those participating in the trial. This was reflected in the higher rate of BSIs observed in the standard arm of the trial (8.24 per 1000 bed-days) compared with linked administrative data from 16 PICUs in England for 2012 (4.58 per 1000 bed-days; see Figure 8).

In terms of budget impact, antibiotic CVCs are more expensive than standard CVCs. If adopted in PICUs, antibiotic CVCs would likely be bulk purchased for all children (including those with a lower risk of BSI than the risk for children participating in the trial). By estimating the number of BSIs potentially averted using antibiotic CVCs for all children (including those with a low risk of BSI), we showed that the additional cost of purchasing antibiotic CVCs is less than the value of resources associated with managing excess BSIs associated with using standard CVCs. A limitation of this study was that estimated BSI rates using standard CVCs relied on a predictive model for identifying children most likely to have required a CVC. Another limitation was the possible error in estimating CVC-days: we assumed that, for children in the linked data set likely to have required a CVC, the CVC would remain in place for the entire PICU stay. There is no clear direction of bias as we may have over- or underestimated CVC-days, but our assumptions are reasonable based on the subset of CATCH participants. Finally, we relied on survey responses to estimate the number of CVCs required in PICUs, but we addressed this and the uncertainty in other parameter estimates by performing sensitivity analyses. 95,96

The generalisability of the RCT results can be assessed by accounting for differences in subgroup treatment effects, for example by reweighting treatment effects based on population distributions. 97,98 In the CATCH trial, the event rate was low and there was limited power to assess variation in the treatment effect according to the duration of CVC insertion. However, because of the nature of the intervention, we assumed that the treatment effect would be constant across groups and would be the same in children who were not enrolled into the trial, as there was no a priori reason for an interaction.

Our results suggest that the benefits of using impregnated CVCs apply even for PICUs with BSI rates as low as 1.2 per 1000 CVC-days. These finding are consistent with systematic review evidence on the cost-effectiveness of impregnated CVCs in adults, which indicates that implementation of impregnated CVCs would be cost-effective for a range of RRs and for a baseline incidence of CR-BSIs as low as 0.2%. 29 The CATCH trial is the first trial to assess the effectiveness of antibiotic-impregnated compared with standard CVCs in children, and our results adds to the strong evidence of effectiveness in adults. Furthermore, as our cost estimates consider only use of hospital resources, the true cost of BSIs and the benefits of antibiotic CVCs may be even greater when longer-term outcomes of BSIs are taken into account.

Other conclusions

Implications for practice

Although our primary outcome showed no difference between the different types of CVC, secondary analyses demonstrated strong evidence of effectiveness of antibiotic-impregnated CVCs compared with standard and heparin-bonded CVCs in children. Based on these results, we consider that use of antibiotic-impregnated CVCs for children admitted to PICUs could result in clinically important reductions in BSI rates. We expect that the benefits of antibiotic-impregnated CVCs would be likely to apply even for PICUs with low BSI rates. At the cost-effectiveness threshold of £30,000 per QALY, antibiotic-impregnated CVCs may not represent a cost-effective alternative to standard CVCs in a PICU setting. However, there is considerable uncertainty surrounding this estimate, which is driven mainly by the time horizon of analysis. Careful monitoring of implementation would help to build up further evidence.

Recommendations for future research

Our research suggests that adoption of impregnated CVCs in PICUs should be considered. Implementation could be monitored through continued linkage of electronic health-care data and information on PICU practice. Such monitoring could allow routine feedback to PICUs and could be enhanced by routine capture of CVC insertion and removal dates in hospital records.

We do not recommend any further trials of antibiotic-impregnated or heparin-bonded CVCs compared with standard CVCs for children or adults in intensive care. However, further trials could be justified to determine whether or not antibiotic CVCs would be similarly effective in preterm neonates (for whom smaller line sizes are required, with potentially different mechanisms for BSI) or in those with long-term CVCs (to determine whether or not the effect of impregnation remains for longer periods). The NHS should work with industry to evaluate different types of impregnation for specific patient groups.

Use of linked administrative data should be considered for future trials to determine the generalisability of interventions in contexts in which outcomes are likely to change substantially over the lifetime of the trial and to monitor implementation of effective interventions. 117

Trial oversight committees

We would like to thank all of the members of the trial oversight committees for their support and work throughout the study and the recommendations in Table 23: Trial Steering Committee – Robert Tasker (Chair), Stephen Playfor (Chair), Andy Vail, Derek Roebuck, Jim Gray and Hazel Greig-Midlane; IDSMC – Paul Ewings (Chair), Mike Sharland, Neena Modi; and the Trial Management Group – Ruth Gilbert (Chair and Chief Investigator), Carroll Gamble, Kerry Dwan, Tracy Moitt, Rachel Breen, Colin Ridyard, Angie Wade, Dyfrig Hughes, Quen Mok, Liz Draper, Shane Tibby, Michael Millar, Oliver Bragshaw, Padmanabhan Ramnarayan, Julia Harris and Darren Hewett.

All members of the Trial Management Group, the Trial Steering Committee and the IDSMC were invited to comment on the final draft of the report.

Ethics

For PICANet, collection of personally identifiable data was approved by the National Information Governance Board for Health and Social Care (formerly the Patient Information Advisory Group) [see www.nigb.nhs.uk/s251/registerapp (accessed 25 November 2015)] and ethical approval was granted by the Trent Medical Research Ethics Committee (reference number 05/MRE04/17). PICANet also has specific permission from the National Research Ethics Service for linkage with PHE laboratory data on BSIs using personal identifiers and to share PICANet data with PHE. An exemption under Section 251 of the NHS Act 2006121 (previously Section 60 of the Health and Social Care Act 2001122) allows PHE to receive patient-identifiable data from other organisations without patient consent to monitor infectious disease. Specific permission for the PICANet–PHE linkage has been granted by the National Information Governance Board for Health and Social Care.

Funding

This work was supported by funding from the National Institute for Health Research Health Technology Assessment (NIHR HTA) programme (project number 08/13/47). The views and opinions expressed therein are those of the authors and do not necessarily reflect those of the HTA programme, NIHR, NHS or Department of Health. No funding was provided by the manufacturer (Cook Medical Inc.) of the CVCs, although participating units could purchase CVCs at a discount of 20% during recruitment to the study. Neither the funder nor the manufacturer had any involvement in the study design, interpretation of the results or writing of the report.

Contributions of authors

Katie Harron (research fellow), Dyfrig A Hughes (Professor of Pharmacoeconomics) and Ruth E Gilbert (Professor of Clinical Epidemiology) prepared drafts of the report, which were reviewed and amended by coauthors [Quen Mok (Consultant in Paediatric Intensive Care), Kerry Dwan (Statistician), Tracy Moitt (senior trials manager), Michael Millar (Consultant in Infection), Padmanabhan Ramnarayan (Consultant in Paediatric Intensive Care and Retrieval), Shane M Tibby (Consultant in Paediatric Intensive Care), Berit Muller-Pebody (Consultant in Infection Control) and Carrol Gamble (Professor in Medical Statistics)].

Statistical analyses were conducted by Kerry Dwan (research associate) and Carrol Gamble (Professor of Medical Statistics).

Colin H Ridyard (research officer in Health Economics) performed the cost-effectiveness analyses.

Berit Muller-Pebody co-ordinated the linkage of PICANet data to laboratory surveillance data.

The end-point review for the primary outcome was carried out by Quen Mok (Consultant in Paediatric Intensive Care), Michael Millar and Ruth E Gilbert (Professor of Clinical Epidemiology).

All authors were invited to comment on the final draft of the report.

Publications

Data sharing statement

All available data can be obtained, subject to data sharing agreements, from the Medicines for Children Clinical Trials Unit.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health.

Predictive model identifying children most likely to require a central venous catheter in the paediatric intensive care unit

The PICANet database does not record insertion or removal of CVCs. However, through the use of CVC audit data from two PICUs, it was possible to create a predictive model to identify admissions in the PICANet data set most likely to have required a CVC.

Methods

Results