Molecular diagnosis of vascular access device-associated infection in children being treated for cancer or leukaemia
Abstract
Blood samples were collected for quantitative 16S rDNA analysis from the vascular access device (VAD) of patients presenting with fever at participating centres of the UK Children’s Cancer and Leukaemia Group. In total, 260 of 301 episodes of fever were evaluable and were classified as probable, possible, unlikely or unclassifiable VAD-associated infection. The sensitivity of the 16S rDNA assay declined concomitantly with delays from time of presentation to sampling. The sensitivity with >0.125 pg of bacterial DNA/μL of whole blood was 80% for the 20 probable VAD-associated infections diagnosed with samples collected on the day of or day following presentation. The specificity rose with increasing amounts of bacterial DNA, from 93% with >0.125 pg, to 98% with 0.25–0.5 pg, and to 100% with >0.5 pg/μL blood. The positive predictive value (for probable or possible) was 88% (95% CI 70–98%) with 0.25 pg/μL, and 100% (95% CI 83–100%) with >0.5 pg/μL. All 18 (6.8%) episodes with >0.5 pg of bacterial DNA/μL blood were associated with positive blood cultures. Identifications derived from the DNA sequence were consistent with the blood culture identifications for 15 of the 17 episodes with a DNA sequence identification. The VAD was removed because of suspected infection in six (2.8%) of 216 episodes with <0.125 pg of bacterial DNA/μL, in one (5%) of 20 episodes with 0.125–0.25 pg/μL, in one (16.7%) of six episodes with 0.25–0.5 pg/μL, and in nine (50%) of 18 episodes with >0.5 pg/μL. A bacterial DNA concentration of >0.5 pg/μL in blood drawn through a central venous catheter at the time of fever presentation had a high positive predictive value for VAD-associated infection and predicted an increased risk of VAD removal because of suspected infection.
Introduction
Vascular access devices (VADs) include transcutaneous catheters and implantable ports. These devices have become essential for the management of certain groups of patients, e.g., those requiring intravenous nutrition, treatment for cancer or intensive care. However, infectious complications arise in 5–26% of patients with a VAD [1], with >50% of hospital-acquired bloodstream infections in the UK being device-related, most frequently involving a central venous catheter [2–4].
Children undergoing treatment for cancer usually require the use of a VAD, e.g., an implanted port, for infusion of anti-cancer drugs, complex drug and hydration schedules, blood product support and, frequently, for parenteral nutrition. The rate of VAD-associated infection in children undergoing treatment for cancer varies from 1.7 to >5.0/1000 VAD days [3,5–7]. Biomedical device infections can be diagnosed when a patient with few other risk-factors for infection develops signs and symptoms of infection associated with inflammation at the site of the device, or fever or rigor following VAD manipulations, or following isolation of commensal skin bacteria from multiple blood cultures. The diagnosis is most difficult in immunocompromised patients, for whom the clinical presentation may be non-specific. Even a tunnel infection may be difficult to diagnose, becoming apparent only when the neutrophil count recovers.
Traditional methods used for the diagnosis of VAD-associated infection have relied on clinical features combined with quantitative microbiology [8–10]. The highest levels of sensitivity and specificity for quantitative microbiological methods are achieved when blood is collected from both the VAD and a peripheral blood vessel, so that the numbers of bacteria/unit volume of blood can be compared [11]. Reliance on paired blood samples is problematical in children with cancer, because of resistance (by staff, patients and parents) to the routine collection of peripheral blood samples. Children undergoing treatment for cancer receive frequent and extended courses of antibiotics, both for prophylaxis and for treatment of infection; thus, an additional problem is that concurrent antibiotic therapy reduces the reliability of diagnostic methods based on laboratory culture.
Improved methods for diagnosis of VAD-associated infection have the potential to reduce unnecessary removal of VADs and to facilitate prompt implementation of targeted treatment. This study describes an evaluation of a novel molecular method that uses quantitative 16S rDNA detection for the diagnosis of VAD-associated infection [12]. The evaluation was carried out in children receiving treatment for cancer.
Patients and methods
The study was coordinated through the Supportive Care Group of the Children’s Cancer and Leukaemia Group (CCLG), and involved a number of UK centres. The protocol for the study was agreed by the CCLG and received ethical approval through the Trent Multi-centre Research Ethics Committee (ref. no. 05/MRE04/23). Each participating centre received local site-specific ethical and research approval. The centres that participated in the study were Belfast, Bristol, Great Ormond Street (London), Liverpool, Newcastle, Nottingham, Royal Marsden and University College Hospital (London).
Children, adolescents or young adults, aged 0–18 years, who were undergoing treatment for cancer/leukaemia, or who were immunosuppressed with a severe haematological disorder, were eligible for the study if the routine standard of care required a tunnelled single-, double- or triple-lumen VAD or implanted vascular port, and if the VAD or port would be required for a minimum of 3 months. Patients who had an indwelling VAD in situ were eligible if they had not received intravenous antimicrobial therapy during the preceding 2 weeks. Written informed consent was required from the parent/guardian, or from the patient where appropriate. Patients who failed to meet these criteria, and those with untunnelled or short-term VADs, were excluded. Eligible patients were invited to participate soon after insertion of a VAD, or at a later outpatient visit or inpatient stay (in the case of patients with existing devices).
When a recruited patient presented with a febrile episode, samples were collected through the VAD for 16S rDNA analysis. A febrile episode was defined as a temperature of either >38°C for >4 h, or >38°C on two occasions >4 h apart within a 24-h period, or >38.5°C on one occasion, or was established according to the presenting centre’s definition of fever.
Samples for quantitative 16S rDNA analysis
Venous blood was collected in 2-mL vacutainer tubes (Vacuette K3E; Becton Dickinson, Oxford, UK) from each lumen of the VAD when patients presented with fever. It is routine practice in many CCLG centres to withdraw and discard a small volume of blood before collecting blood for culture or other analyses. This ‘discard’ blood was accepted as a suitable sample for 16S rDNA analyses. Samples were stored at participating centres at less than or equal to –20°C until collected in batches for transport on dry ice to the laboratory at Barts and the London NHS Trust (London, UK). Routine samples were also collected at the time of presentation, including blood for culture. Centres were encouraged to send VAD tips for quantitative culture, particularly if a VAD was removed for suspected VAD-associated infection.
Samples were analysed for bacterial 16S rDNA when they had been collected within 24 h before or 72 h after the start of antibiotic treatment. The date of sampling was recorded so that delays in sampling could be taken into account in the analysis.
Clinical data collection
Clinical data were collected at baseline (within 72 h of fever presentation), and at 4 weeks after presentation, using standard questionnaires. The baseline data sheet at 72 h requested information concerning diagnoses, samples collected for laboratory analyses, VAD details (e.g., number of lumens), antibiotics administered, and symptoms and signs at presentation (including chills, fever, rigors or hypotension associated with VAD manipulations).
The data sheet completed at 4 weeks requested the results of laboratory investigations, details of antibiotics prescribed, duration of fever, clinical response to treatment, details of VAD management (including whether the VAD was removed as part of the management of suspected VAD-associated infection), other sources of infection, specific agents of infection identified, and the impression of and classification of the episode by the clinician responsible for the patient’s care.
Clinical data sheets were returned to the CCLG data centre in Leicester, where the data were extracted and placed in an Excel database. The molecular test results and clinical databases were merged for the analysis of test performance.
Definitions of VAD-associated infection
Febrile episodes were classified as probable, possible, unlikely or unclassifiable bacterial VAD-associated infections. The classification of the fever episodes was carried out at the CCLG data centre by staff who were unaware of the results of the 16S rDNA analyses. The definitions were agreed by clinical collaborators in CCLG centres, and broadly reflected the criteria used in the CCLG centres for defining VAD-associated infection.
Episodes were classified as probable if any of the following criteria were met: (i) two or more blood cultures collected within 72 h of presentation that were culture-positive for a skin commensal, e.g., a coagulase-negative staphylococcus (including positive blood cultures from different lumens of the same VAD on the same or different occasions of sampling); (ii) a positive blood culture from a patient with signs or symptoms of infection, and an isolate with the same identification and antibiotic susceptibility profile as that of an isolate from the VAD tip culture; (iii) fever, chills and/or hypotension associated with VAD manipulations, together with a response to treatment appropriate for central venous catheter (CVC)-associated infection (intravenous antibiotic treatment, and/or VAD removal); response to treatment was defined as resolution of fever within 5 days of the initiation of treatment, with no recurrence within 5 days of discontinuing antibiotic treatment, and antibiotic treatment for VAD-associated infection required that all of the lumens were exposed to antibiotic treatment; or (iv) inflammation extending at least 2 cm along the tunnel from the VAD exit site in a patient with systemic signs or symptoms of infection.
Episodes were defined as possible when there was a VAD in place and the patient’s clinical condition resolved in response to appropriate (for episodes with positive microbiological identification, e.g., a blood culture isolate) and specific treatment for bacterial VAD-associated infection. Specific treatment for VAD-associated infection required that all of the lumens were exposed to antibiotic treatment, and/or the VAD was removed. A complete response to treatment was defined as resolution of fever within 5 days of the initiation of treatment, and no recurrence of fever within 5 days of discontinuing antibiotic therapy.
Episodes of unlikely bacterial CVC-associated infection were those in which the patient showed a complete resolution of symptoms without specific treatment for bacterial CVC-associated infection. These episodes may or may not have been associated with positive blood cultures. Episodes in which the VAD was removed for suspected fungal CVC-associated infection were classified as unlikely (bacterial) CVC-associated infection.
Unclassifiable episodes were defined as those that did not fit the definition of probable, possible or unlikely bacterial VAD-associated infection. These included episodes for which there was insufficient information to classify an episode, episodes in which a patient remained febrile with or without specific treatment of VAD-associated infection for >2 weeks, and episodes in which there was recurrence of fever within 5 days of discontinuing systemic antibiotic therapy.
Episodes that were unclassifiable using the above definitions were reclassified using the classifications probable, possible, unlikely and unclassifiable, as recorded by the clinician responsible for patient care at 4 weeks after episode presentation. Only those episodes unclassifiable according to the predefined criteria and clinician judgement were considered to be unclassifiable in the final analyses. Clinicians had access to the definitions used in the formal classification.
Molecular methods
The development of these methods has been described previously by Warwick et al. [12]. For the purposes of this study, all extractions were performed as described below, although subsequent work is now performed using automated DNA extraction methods.
DNA extraction from clinical and control samples. DNA was extracted from 200-μL aliquots of EDTA-anticoagulated whole blood. Each sample was mixed with 1200 μL of freshly prepared 0.17 M ammonium chloride and incubated at room temperature for 30 min. Following centrifugation at 11 600 g for 10 min, the pellet was washed twice with 500 μL of sterile saline (0.9% w/v) and then extracted using a QIAmp DNA minikit (Qiagen, Hilden, Germany). The pellet was resuspended in 180 μL of Qiagen ATL buffer (containing EDTA and SDS) and exposed to six freeze–thaw cycles (cycling between −70°C and 50°C), with vortexing between cycles, before being heated in a boiling water bath for 10 min. The remainder of the extraction procedure was performed according to the manufacturer’s protocol. DNA was eluted in 50 μL of buffer and stored at –20°C until analysis.
Several controls were run routinely with each batch of tests. These included blood samples from a healthy individual with and without spiking with bacteria. An extraction control of blood spiked with 103 CFU of Staphylococcus epidermidis/μL was found to yield DNA levels close to the lower limit of detection. Bacterial DNA controls containing known amounts of bacterial DNA extracted from Enterococcus faecalis (100 pg to 100 fg) and a negative control (with template DNA omitted to detect reagent contamination) were also included in each run.
PCR conditions (TaqMan assay). Real-time PCRs were performed using the ABI Prism 7900HT sequence detection system (Applied Biosystems, Warrington, UK) in optical 384-well plates. Reaction mixtures contained (1× dilution) TaqMan universal PCR mastermix (Applied Biosystems), 300 nM each of the forward and reverse primers, 100 nM fluorescent probe, 2 μL of template DNA, and water to a final volume of 20 μL. The cycling conditions comprised 50°C for 2 min and 95°C for 10 min, followed by 40 cycles of 95°C for 125 s and 60°C for 1 min. The primer sequences were: forward primer 5′-TCCTACGGGAGGCAGCAGT; reverse primer 5′-GGACTACCAGGGTATCTAATCCTGTT; and probe sequence, 5′-CGTATTACCGCGGCTGCTGGCAC [12].
The threshold cycle (ct) value, which is inversely proportional to the log of the amount of target DNA initially present, was calculated using SDS software v.2.0 (Applied Biosystems). All samples were run in triplicate. The median cycle result was used to calculate bacterial DNA concentrations by comparison with a DNA standard curve constructed from the results obtained using DNA standards.
Identification using DNA sequencing. When a sample contained >0.5 pg of bacterial DNA/μL of blood, it was possible to amplify a c. 1300-bp 16S rRNA gene fragment directly from the DNA extracts using oligonucleotide primers 5′-TCAGATTGAACGCTGGCGGC (forward) and 5′-CCCGGGAACGTATTCACCG (reverse). Each PCR was performed in a total volume of 25 μL containing 0.2 μM each primer, 2 mM MgCl2, 1 U Taq DNA polymerase (Promega, Southampton, UK), 1× PCR buffer (Promega) and 2 μL DNA extract. PCRs comprised 95°C for 3 min, followed by 30 cycles of 95°C for 10 s, 58°C for 20 s and 72°C for 30 s using a Palm Cycler (Corbett Research, Sydney, Australia). PCR products were sequenced, using the forward primer and the internal primer 5′-TGCCAGCAGCCGCGGTAATA, on an ABI Prism 3700 DNA Analyzer (PE Biosystems, Warrington, UK). The sequences were aligned using the Clustal W algorithm to produce a consensus sequence. This was analysed using the BLAST algorithm at the NCBI site [13].
Results for samples containing 0.125–0.5 pg of bacterial DNA/μL of extracted whole blood were only reported as positive when the concentration was >0.125 pg/μL on repeat testing. All of the results of the molecular tests were entered into an Excel spreadsheet for statistical analysis.
Statistical methods
Sensitivity, specificity and positive and negative predictive values were calculated, together with exact binomial 95% CIs. Stata v.9 software (Stata Corp., College Station, TX, USA) was used for the analyses. When multiple lumens were present, the highest bacterial DNA concentration was used in the episode analysis.
Results
Samples and clinical data sheets were collected from 301 episodes of fever in 207 children. The numbers recruited by each centre were: Belfast 15, Bristol 51, Great Ormond Street 2, Liverpool 63, Newcastle 63, Nottingham 19, Royal Marsden 19, and University College Hospital 31. Forty-one episodes were excluded from analysis. The reasons for exclusion were: incomplete clinical data collection, problems with consent, inappropriate sample storage or loss of samples, or antibiotics given intravenously during the 14- to 3-day period before the onset of fever (ten episodes); failure to collect samples from all lumens (26 episodes); and development of signs and/or symptoms of VAD-associated infection >48 h after fever presentation (five episodes). The proportion of episodes excluded from analysis ranged from 0% to 33.3% for each centre. There was no significant difference among centres in the proportion of episodes excluded. The five episodes of late presentation of VAD-associated infection were diagnosed 5–23 days after initial presentation. Four of these five episodes were associated with positive blood cultures, and one episode was a tunnel infection. In one of these episodes, a sample was collected for 16S rDNA analysis at the time of fever recurrence (5 days after initial presentation), and this sample gave a bacterial DNA concentration of 0.34 pg/μL blood, while blood cultures taken at the same time grew Stenotrophomonas maltophilia.
Overall, VAD tips were sent for culture from 16 (84%) of 19 episodes in which the VAD was removed, and from 14 (82%) of 17 episodes in which the VAD was removed for suspected VAD-associated infection.
The patient had received oral antimicrobial agents in the previous 2 weeks in 133 (51.1%) of the 260 evaluable fever episodes, with 125 (48.1%) receiving an antibacterial agent. In 117 episodes, these antibacterial agents were prophylactic (trimethoprim–sulphamethoxazole in 110 episodes, and ciprofloxacin in seven episodes). In 17 episodes, oral antibacterial agents were being administered for treatment at the time of fever presentation (with or without prophylactic agents). Nine patients were receiving both prophylactic and therapeutic antibacterial agents.
The date on which the blood for 16S rDNA was collected was the date of fever presentation (day 0) in 189, day 1 in 46, day 2 in 21 and day 3 in four of the 260 episodes. In those episodes where the date of collection was on day 0 or 1 of fever, 67 patients had been started on intravenous antibiotics before the DNA sample was collected. The classification of fever episodes is shown in Table 1.
| VAD infection | A | B | C |
|---|---|---|---|
| Eligible, all lumens sampled and excluding hospital-acquired VAD infection (%) | Category A, plus collected on first or second day of fever (%) | Category B, plus intravenous antibiotics not given on days before DNA sample (%) | |
| Probable | 26 (10) | 20 (8.6) | 15 (7.8) |
| Possible | 43 (16.5) | 39 (16.7) | 29 (15.0) |
| Unlikely | 190 (73) | 174 (74.4) | 148 (76.7) |
| Unclassifiable | 1 (0.4) | 1 (0.4) | 1 (0.05) |
| Total | 260 (100) | 234 (100) | 193 (100) |
The bacterial DNA detection method had a relatively high detection level of c. ten genome copies (0.1 pg)/μL of blood. In some cases, the bacterial DNA concentration exceeded 100 pg/μL, which is equivalent to 107 genomes/mL. Blood samples were aspirated directly from the VAD without flushing the VAD, so these very high levels may be a consequence of undiluted accumulated biomass from the VAD. However, the results were reproducible with repeated testing and were calculated against a standard curve constructed using known DNA standards.
Sensitivity and specificity for detecting line infection with various concentrations of bacterial DNA in blood are shown in Table 2. The sensitivity improved substantially when the blood samples for bacterial culture and bacterial DNA analysis were collected temporally close together, and even more so if they were collected before intravenous antibiotics had been given. The specificity was high for a cut-off point at 0.25 pg of bacterial DNA/μL, and reached 100% with 0.5 pg of bacterial DNA/μL.
| Cut-off point (pg/μL) | A | B | C |
|---|---|---|---|
| Eligible, all lumens sampled and no subsequent hospital-acquired line infection | Category A, plus collected on first or second day of fever | Category B, plus intravenous antibiotics not given on days before DNA sample | |
| Sensitivity (probable line infection) (%) | |||
| 0.125 | 65 (44–83) | 80 (56–94) | 87 (60–98) |
| 0.25 | 46 (27–67) | 60 (36–81) | 73 (45–92) |
| 0.5 | 42 (23–63) | 55 (31–77) | 67 (38–88) |
| n | 26 | 20 | 15 |
| Specificity (unlikely line infection) (%) | |||
| 0.125 | 93 (89–96) | 93 (8–96) | 94 (89–97) |
| 0.25 | 98 (95–100) | 98 (95–100) | 98 (94–100) |
| 0.5 | 100 (98–100) | 100 (98–100) | 100 (97.5–100) |
| n | 190 | 174 | 148 |
The centre protocols for management of febrile episodes specified treatment for VAD-associated infection for patients who had unresolved pyrexia after 48–96 h of empirical antibiotic treatment; thus, many of the episodes of possible VAD-associated infection were effectively protocol-defined, and included a significant proportion of episodes associated with mucositis, viral infection or other potential causes of fever. When samples were collected on day 0 or day 1, the sensitivity of a bacterial DNA concentration of 0.125 pg/μL for possible VAD-associated infection was 33.3% (13/39). The four probable episodes that yielded negative bacterial DNA results when samples were collected on day 0 or day 1 were all defined as probable, based on the presence of fever, chills and/or hypotension associated with VAD manipulations and a response to treatment appropriate for CVC-associated infection. The positive predictive value of a probable or possible episode vs. unlikely or unclassifiable infection was 88% (21/24) (95% CI 68–97%) at >0.25 pg, and 100% (18/18) (95% CI 81–100%) at >0.5 pg.
Each of the 18 episodes with a bacterial DNA concentration of >0.5 pg/μL was associated with a positive blood culture. Three of the six episodes with 0.25–0.5 pg/μL were associated with positive blood cultures, as were nine of 20 episodes with 0.125–0.25 pg/μL. There were 22 episodes with positive blood cultures and bacterial DNA concentrations of <0.125 pg/μL. Of these, nine were positive with a skin bacterium in only one of multiple blood cultures (suggesting blood culture contamination), and seven had blood collected for DNA testing at least 1 day after the start of intravenous antibiotics and blood culture; the sample for one case was collected 2 days after blood culture.
Sequencing of the bacterial DNA in samples with >0.5 pg/μL was performed following amplification of 16S rDNA from DNA extracts. The sequence identifications obtained are summarised in Table 3.
| Bacterial DNA (pg/μL blood) | Sequence identification | Blood culture identification |
|---|---|---|
| 0.7 | Staphylococcus spp. | Coagulase-negative staphylococcus |
| 0.7 | Staphylococcus epidermidis | Coagulase-negative staphylococcus |
| 1.1 | Acinetobacter spp. | Acinetobacter spp. |
| 1.1 | Staphylococcus aureus | S. aureus |
| 1.4 | Enterobacter spp. | Enterobacter cloacae |
| 1.6 | S. epidermidis | Coagulase-negative staphylococcus |
| 1.6 | Klebsiella oxytoca | K. oxytoca |
| 2.9 | Acinetobacter baumannii | Acinetobacter spp./ Pseudomonas aeruginosa |
| 5.6 | S. aureus | S. aureus |
| 9.7 | S. epidermidis | Coagulase-negative staphylococcus |
| 11.25 | Vibrio harveyi | V. harveyi |
| 12.8 | A. baumannii | A. baumannii |
| 13.1 | Bacillus cereus | Bacillus spp. |
| 13.1 | K. oxytoca | K. oxytoca |
| 21.3 | Escherichia coli | Enterobacter spp. |
| 21.6 | Corynebacterium tuberculostearicum | Coagulase-negative staphylococcus |
| 160 | Unreadable sequence | Mixed Staphylococcus spp. |
| 425 | P. aeruginosa | P. aeruginosa |
In 17 (6.5%) of the 260 evaluable episodes, VADs were removed for suspected VAD-associated infection. Only two episodes in which the VAD was removed (and where the bacterial DNA samples were collected on the date of presentation or the day after, and intravenous antibiotics had not been given on the previous days) yielded a bacterial DNA concentration of <0.125 pg/μL. One of these two episodes was associated with diarrhoea and Clostridium difficile toxin-positive stool tests, and the other was associated with isolation of parainfluenza virus 3 from a nasopharyngeal sample.
The proportion of VADs removed within 4 weeks of fever presentation because of suspected infection increased as the bacterial DNA concentration increased. The VAD was removed because of suspected infection in six (2.8%) of 216 episodes with <0.125 pg/μL, one (5%) of 20 episodes with 0.125–0.25 pg/μL, one (16.7%) of six episodes with 0.25–0.5 pg/μL, and nine (50%) of 18 episodes with >0.5 pg/μL. The survival (retention) curve for VADs is shown in Fig. 1. In comparison, the VAD was removed because of suspected infection in 15 (28.8%) of 52 episodes associated with positive blood cultures.
Survival (retention) curves for vascular access devices in the 4-week period following presentation with various bacterial DNA concentrations in blood.
There was no relationship between C-reactive protein (CRP) levels, measured at the time of presentation, and bacterial DNA concentrations in the 149 episodes in which the CRP level was measured. Episodes classified as probable or possible VAD-associated infection had a higher rate of CRP positivity (>10 mg/L) than those classified as unlikely VAD-associated infection (97.5 vs. 82.7%; p 0.016). There was no correlation between the duration of use of intravenous antibiotics and bacterial DNA concentrations, but cases with negative bacterial DNA assays had a more variable duration of intravenous antibiotic usage than those with positive bacterial DNA assays (SD 8.3 with bacterial DNA <0.125 pg/μL; SD 5.4 with bacterial DNA >0.125 pg/μL).
Discussion
There is a need for improved diagnostic methods for the diagnosis of VAD-associated infection in children with cancer, particularly because of the reluctance to collect peripheral blood samples routinely. The method reported here has a relatively high detection level of c. ten genome copies/μL of blood. This relatively high detection level probably explains the episodes with positive blood culture and negative bacterial DNA test results, although the possibility of blood culture contamination cannot be excluded. Nevertheless, the method yielded sensitivity (for episodes defined as probable VAD-associated infection), specificity and positive predictive values similar to those reported for paired quantitative blood cultures [11]. Unlike many reported evaluations, this study was performed by laboratory staff working at a distant site who were unaware of the clinical details of individual patients, and the results were achieved despite the frequent exposure of patients to oral antibiotics in the 2-week period preceding fever presentation.
The method described in this study has a number of advantages over paired quantitative blood cultures, including a requirement for vascular access samples only, the use of small volumes of discard blood, and the potential for automation. The manual DNA extraction method described in this study is time-consuming, but subsequent evaluations have obtained comparable results using automated DNA extraction systems, with considerable savings in technician time (results not shown). The quantitative bacterial DNA method used in the present study does not generate a product that is sufficiently informative to allow bacterial identification. When the bacterial DNA concentration was >0.5 pg/μL, it was possible to identify bacteria by amplification of a discriminatory 16S rDNA region, followed by sequencing of the amplified product. The majority of identifications according to molecular and conventional laboratory methods were consistent. Discrepant identifications probably reflect the limitations of routine laboratory standard operating procedures.
Current practice in paediatric oncology in the UK and much of Europe [14] is to retain VADs if removal can be avoided; thus, in the present study, VAD removal occurred >7 days after fever presentation in 50% of cases, precluding the use of catheter segment culture as a reference method [15]. Collection of peripheral samples for blood culture is not routine practice in paediatric oncology, despite being recommended in policy documents [16], so the use of paired quantitative blood cultures was precluded as a reference standard. Instead, clinical and laboratory criteria was used to define VAD infection categories. This approach has the disadvantage of defining some episodes as probable VAD-associated infection based on clinical criteria alone, without any requirement for corroborating laboratory evidence. Episodes defined as probable VAD-associated infection, but with negative bacterial DNA test results, were all placed in the category defined by ‘fever, chills and/or hypotension associated with VAD manipulations’.
Novel methods for the diagnosis of VAD-associated infections have been reported previously, but these share many of the problems of traditional approaches, or have other disadvantages, and have therefore not been universally adopted. Methods relying on microbial culture, e.g., differential time-to-positivity [17], are unreliable in patients treated with antibiotics. This method also requires the collection of a sample from a peripheral vessel, which may be unacceptable with some patients, especially children. Methods that require invasive sampling of the lumen of the VAD [18] cannot be used in narrow bore VADs, e.g., those used in infants, and there is the potential for complications arising from disturbance of thrombi attached to the lumen or tip of the VAD. This methodology is time-consuming and requires specialist equipment. Newer radiological approaches, e.g., positron emission tomography [19], the use of labelled antimicrobial peptides [20], and the detection of specific immune responses, may also have a role in the diagnosis of device-associated infections [21], but none of these methods has been evaluated for the diagnosis of VAD-associated infection. These techniques are also likely to be expensive and available only in specialist centres.
Previous reports have suggested a link between time-to-positivity (a marker of bacterial load) and outcome for both Staphylococcus aureus [22] and Streptococcus pneumoniae [23] bloodstream infections. In the present study, increasing bacterial DNA load in blood samples drawn through the VAD was associated with an increasing risk of VAD removal for suspected infection. Information was only collected for 4 weeks after fever presentation. Prolonging the period of data collection might allow a better assessment of the implications for outcomes in patients with high bacterial load VAD-associated infection. Bacterial load is an important determinant of the efficiency of sterilisation and disinfection processes, so it is perhaps not surprising to find a relationship between the effectiveness of antimicrobial treatment of VAD-associated infection and bacterial load. The increased variability in duration of intravenous antibiotic treatment in patients with and without detectable levels of bacterial DNA is consistent with the hypothesis that detection of bacterial DNA identified a discrete group of episodes within an aetiological category (VAD-associated infection).
If confirmed in other studies, these results have implications for the management of patients with VAD-associated infection, and also for studies of interventions designed to prevent or treat VAD-associated infection. The bacterial DNA test can give results within a few hours of sample collection, and therefore has the potential to allow early specific interventions targeted at optimising the outcome for patients with VAD-associated infection. Further work should examine the persistence of bacterial DNA, particularly as a marker of response to treatment.
Acknowledgements
This study was funded by the Health Technology Programme of the Department of Health, and was supported by the Children’s Cancer and Leukaemia Group. The views and opinions expressed therein are those of the authors and do not necessarily reflect those of the Department of Health. The authors declare that they have no conflicting interests in this study.
