This chapter contains reproduced material from Waldron et al.1,2 These are Open Access articles distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) licence, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The text below includes minor additions and formatting changes to the original text.
Aims and objectives
The aim of the PRECISE study was to determine if there are specific subgroups of patients, for whom a PCT-guided antibiotic algorithm may be beneficial, harmful or ineffective. It aimed to identify endotypes or sub-phenotypes of infection to facilitate optimisation of antibiotic dosing and duration (when, by how much and for how long). The embedded mechanistic substudy within an existing RCT allows the theranostic exploration of for host–response biomarkers to guide judicious antibiotic use.
Primary objective
The primary objective was to determine, within the BATCH trial, if there were specific subgroups of patients (based on baseline host response and organ dysfunction) in whom the addition of PCT testing to current best practice based on the NICE AMS guidelines could safely allow a reduction in duration of antibiotic therapy in hospitalised children with suspected or confirmed bacterial infection compared to current best practice alone. To meet this objective specifically, we assessed duration of IV antibiotics.
Secondary objectives
The secondary objective was to assess the effect of additional PCT testing to AMS best practice in specific subgroups within the BATCH trial (based on baseline host response and organ dysfunction) on:
- total duration of antibiotics (oral and IV)
- time to switch from broad spectrum to narrow spectrum antibiotics
- time to discharge from hospital
- suspected ADRs (SADR) (defined using the Liverpool Causality Assessment Tool)39
- hospital acquired infection up to day 28.
Design
Embedded mechanism of action substudy within a multicentre, prospective, individually randomised, open-label two-arm RCT comparing PCT-guided antibiotic algorithm versus UC.
Setting
Participants in the BATCH trial were recruited into the PRECISE substudy from paediatric wards or PICUs within UK children’s hospitals (n = 4) taking part in the host BATCH trial in the UK.
Site recruitment
The PRECISE study was open to participant recruitment from December 2020, following ethical approval of a substantial amendment for the addition of the substudy as part of the main BATCH trial protocol. Recruitment was completed in October 2022.
Four children’s hospitals in England who were taking part in the BATCH trial (Alder Hey Children’s Hospital, Bristol Royal Hospital for Children, Southampton Children’s Hospital, Oxford Children’s Hospital) recruited into the PRECISE study and were led by clinician principal investigators who were PRECISE co-applicants. The Clinical Research Network supported site recruitment.
Participant selection
The PRECISE study population comprised children who were recruited into the BATCH trial, aged between 72 and 18 years old admitted to hospital for confirmed or suspected SBI, in whom IV antibiotics were commenced, and who were expected to remain on IV antibiotics for more than 48 hours, and whose parents/legal guardians consented to additional blood samples being taken for the study.
Eligibility criteria
Children were eligible to join the BATCH trial and PRECISE study if they met all the inclusion criteria and did not meet any of the exclusion criteria described in Table 1.
| Inclusion criteria | Exclusion criteria |
|---|---|
| All children aged between 72 hours old and up to 18 years old admitted to hospital for confirmed or suspected bacterial infection, in whom IV antibiotics are commenced, and who are expected to remain on IV antibiotics for more than 48 hoursConditions including (but not limited to): bacteraemia, central line-associated bloodstream infections, uncomplicated bone and joint infections (such as single site infection, osteomyelitis with adjacent septic arthritis or septic arthritis with adjacent osteomyelitis), discitis, empyema, pneumonia, pyelonephritis, sinusitis, retropharyngeal abscess, pyomyositis, uncomplicated culture-negative meningitis, intra-abdominal infections, lymphadenitis, cellulitisFirst time in the BATCH trial | Preterm infants < 37 weeks corrected gestational age, < 72 hours old or > 18 years of ageChildren admitted moribund and not expected to survive more than 24 hoursChildren who had a predicted duration of IV antibiotics of < 48 hoursChildren not expected to survive at least 28 days because of a pre-existing conditionChildren who had bacterial meningitis,a bacterial endocarditis or brain abscess.Children who had complicated bone and joint infections.bChildren who received antibiotics for surgical prophylaxisChildren who had chronic comorbidities, such as cystic fibrosis, chronic lung disease, bronchiectasis where there is already a predefined length of course of antibioticsChildren who were severely immunocompromised (e.g. chemotherapy, stem cell transplant, biological therapy for inflammatory or rheumatological conditions).Children who, in the opinion of the local investigator, were unsuitable for randomisation due to high probability of requiring sustained IV therapyChildren who had a presence of existing directive to withhold life-sustaining treatmentInborn infants admitted to neonatal ICUs (NICUs), neonatal HDUs, special care baby units (SCBUs) or postnatal wards |
Study procedures
Figure 1 shows the BATCH and PRECISE participant flow and study schema.
FIGURE 1 Participant flow and study schema. a, Based on best practice for antimicrobial prescribing ‘Start Smart-Then Focus’ Public Health England, March 2015. https://www.gov.uk/government/publications/antimicrobial-stewardship-start-smart-then-focus. Reproduced from Waldron et al.2 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) licence, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The figure includes minor additions and formatting changes to the original text. (Figure not reproduced here; see the original article.)
Participant recruitment
Parents/carers of BATCH participants were asked to read a supplementary information sheet for the PRECISE study (see Appendix 1). If they agreed to take part, they were asked to sign the informed consent section at the bottom of the supplementary information sheet.
Parents/carers were advised that where possible, leftover blood taken from routine blood tests would be salvaged to use for the PRECISE study; however, in instances where there was not enough leftover blood, additional samples may have to be taken. If any additional samples were needed, they were told that their permission would be sought first and would use the method that is most preferable to them and their child for example, finger prick, venepuncture, arterial line (if already in place).
Parents/carers were also informed that they would receive a £20 high street shopping voucher for the inconvenience of taking part in the study, which was unconditional and not dependent on whether they gave permission for any additional samples to be taken. This is because in the BATCH trial, only children in the intervention arm had the additional PCT test, to help guide antibiotic duration (which may benefit the child), whereas in the PRECISE study, the additional tests could not provide any additional benefit to the individual child.
Payment of the voucher did not over-ride the principles of freely given and fully informed consent. The parents were told before they agreed to participate, that they could withdraw from the study at any time without losing the voucher.
Screening logs
Sites were asked to keep an anonymous screening log of participants who were eligible but had not consented to take part in the PRECISE study or who were not approached. These were used to inform adjustment of recruitment strategies and trial processes and assess potential selection bias.
Registration and withdrawal
Registration and withdrawal procedures are described in the BATCH report.1
Randomisation
Randomisation procedures are described in the BATCH report.1 Briefly, eligible participants were allocated in a 1 : 1 ratio using covariate minimisation with a random element, using age and centre as balancing covariates.
Safety monitoring and protocol deviations
Safety monitoring and protocol deviation procedures are described in the BATCH report.1
Training
All staff involved in the trial, including clinicians and research nurses/coordinators at sites were provided with written standard operating procedures (SOPs) and received study specific training in study procedures and good clinical practice prior to commencing the study.
Data collection
Data collection procedures are described in the BATCH report.1
Sample collection
Blood samples were taken for the PRECISE study (following randomisation) at baseline in both the BATCH intervention and control arms for MR-proADM, TRAIL, IP-10 and CRP analysis.
Samples were collected at the same time as routine bloods were taken where possible, however an additional sample may have been collected at a separate time point if routine blood tests were not due, or there was not enough routine blood left over to store for the PRECISE study. Where serial blood samples were being collected for PCT (in the intervention arm), for routine care, or the parent/carer was happy for additional samples to be collected, then serial study samples were also taken for MR-proADM, TRAIL, IP-10 and CRP analysis.
The PRECISE biomarker tests required 0.5–1 ml of additional blood. This amount is easily available from a finger prick or venepuncture, even in an infant. The sample volume required for the MR-proADM test was 26 µl of plasma and the TRAIL, IP-10 and CRP tests required a minimum of 35 µl of plasma. The plasma samples were collected, centrifuged and stored at −80 °C at the local sites, then periodically transferred to the research laboratory at the Department of Clinical Infection, Microbiology and Immunology at the University of Liverpool. The MR-proADM, TRAIL, IP-10 and CRP assays were performed in batches at the end of the BATCH trial.
Additional samples were also included from participants recruited into the main BATCH trial prior to commencement of the PRECISE study who gave consent for their samples to be stored and used for future research. These samples were from participants in the PCT arm only, as no extra blood samples were taken and stored from those recruited in the UC arm prior to the PRECISE substudy receiving ethical approval.
The results were not fed back to treating clinicians. Figure 2 shows the PRECISE study sample collection flow diagram.
FIGURE 2 PRECISE study sample collection flow diagram. (Figure not reproduced here; see the original article.)
Data management
Data management, data monitoring and data cleaning procedures in the BATCH trial are described in the BATCH report.1
Study management
A specific PRECISE Study Management Group (SMG), consisting of a subset of the BATCH Trial Management Group (TMG), met monthly by teleconference. It included the CoChief Investigators, all other co-applicants and the Cardiff University Centre for Trials Research (CTR) team. The PRECISE SMG developed study procedures/documents and advised on the conduct of the study. The BATCH Trial Steering Committee (TSC) and Independent Data Monitoring Committee (IDMC) reviewed the PRECISE protocol and monitored study progress.
Research governance
The BATCH trial had ethical approval granted by a NHS Research Ethics Committee (REC), recognised by the UK Ethics Committee Authority (UKECA). The initial approval was granted by the North West Liverpool East REC on 12 April 2018, reference number 18/NW/0100. Approval for an amendment to include the PRECISE study was granted by the North West Liverpool Central REC on 9 October 2020. NHS Research and Development confirmation of capacity and capability was sought from the respective NHS relevant organisations.
The PRECISE study was assigned the International Standard Randomised Controlled Trial Number (ISRCTN) ISRCTN14945050 (registered 17 December 2020).
Outcome measures
Primary outcome measure
Time from starting to stopping IV antibiotic therapy was the primary (time-to-event) end point, as in the BATCH trial.
Secondary outcome measures
Aligned to the BATCH trial:
- total duration of antibiotics (oral and IV)
- time from starting broad spectrum to switching to narrow spectrum antibiotics
- time from admission to discharge from hospital
- suspected ADRs
- hospital-acquired infection up to day 28.
Sample size
Time until IV antibiotic therapy is discontinued following randomisation was used as primary (time-to-event) end point, as in the BATCH trial. The aim was to test whether the effect of the intervention (PCT vs. UC) on the primary end point was different across the predefined baseline biomarker subgroups (Figure 3), that is, whether there exists an interaction between intervention arm and baseline biomarker subgroup.
FIGURE 3 Hypothesised effects of stratification in PCT-guided antibiotic arm compared to control arm, and assumed proportions of participants falling into each subgroup. (Figure not reproduced here; see the original article.)
For baseline MR-proADM, the assumptions were median antibiotic durations of 2 versus 4 days (UC vs. PCT) in the low MR-proADM subgroup, 7 versus 5 days (UC vs. PCT) in the intermediate MR-proADM subgroup, and 10 versus 5 days (UC vs. PCT) in the high MR-proADM subgroup (see Figure 3). To detect an arm-by-subgroup interaction effect of this size with 90% power while controlling the type I error level at 5%, 25 participants were needed per biomarker subgroup in each trial arm, as calculated with the method of Peterson and George assuming the antibiotic durations follow an exponential distribution (see Appendix 1 for R code).42 To achieve this sample size, given the assumed proportions of participants falling into each MR-proADM subgroup (as given in Figure 3), and also inflating for 5% dropout, a total of 266 participants were needed, with usable baseline blood samples.
For baseline TRAIL, IP-10 and CRP, the assumptions were median antibiotic durations of 1 versus 3 days (UC vs. PCT) in the high probability of viral infection subgroup, 7 versus 4 days (UC vs. PCT) in the intermediate probability of bacterial infection subgroup, and 10 versus 5 days (UC vs. PCT) in the high probability of bacterial infection subgroup (see Figure 3). Based on similar calculations as for MR-proADM above, 13 participants per biomarker subgroup in each trial arm were needed, and thus a total of 138 when taking into account the assumed proportions of participants in each TRAIL, IP-10 and CRP subgroup (as given in Figure 3), and inflating for 5% dropout. Consequently, the overall sample size for the study was 266 baseline blood samples.
Analysis
Final analysis of the primary and secondary outcomes took place when all the samples were processed and analysed in the research laboratory in the Department of Clinical Infection, Microbiology and Immunology at the University of Liverpool and the data securely transferred to the CTR.
Laboratory analysis
At the research laboratory, each box of samples was defrosted overnight at 4 °C immediately prior to testing. All samples were kept on ice while being analysed.
Analysis of each sample depended on sample size:
For samples of more than 1 μl but < 25 μl, only CRP analysis was carried out. For samples of more than 25 μl but < 300 μl, CRP and IP-10 and TRAIL were carried out. For samples of > 300 μl, all analyses were carried out (CRP, IP-10, TRAIL and pro-ADM).
Tumour necrosis factor-related apoptosis-inducing ligand, interferon gamma-induced protein 10 and C-reactive protein
The U-PLEX Biomarker Group 1 (Human) Multiplex Assays and V-PLEX Biomarker Group 1 (Human) Multiplex Assay protocols were followed (www.mesoscale.com/products_and_services/assay_kits/u-plex_gateway/u-plex_technology).
The principle of the Uplex assays is that biotinylated capture antibodies are coupled to U-PLEX Linkers, which self-assemble onto unique spots on the U-PLEX plate. Analytes in the sample bind to the capture reagents; detection antibodies conjugated with electrochemiluminescent labels (MSD GOLD™ SULFO-TAG, Rockville, MD, USA) bind to the analytes to complete the sandwich immunoassay. Once the sandwich immunoassay is complete, the U-PLEX plate is loaded into an MSD® (Rockville, MD, USA) instrument where a voltage applied to the plate electrodes causes the captured labels to emit light. The instrument measures the intensity of emitted light (which is proportional to the amount of analyte present in the sample) and provides a quantitative measure of each analyte in the sample.
Interferon gamma-induced protein 10 and TRAIL were analysed using U-PLEX Human IP-10 immunoassay and U-PLEX Human TRIAL immunoassay respectively [provided by Meso Scale Diagnostics (MSD)]. Plates were coated at the beginning of the week, these were then used from the following day onwards.
C-reactive protein was detected using the V-PLEX Human CRP immunoassay (provided by MSD). V-plex assays were already coated (www.mesoscale.com/en/products_and_services/assay_kits/v-plex).
Unfortunately, due to calibration issues with the MSD plate readers, many of the sample results could not be interpreted and, therefore, were unable to be included in the data set for statistical analysis. Analysis was repeated where there was enough sample available. All data from the MSD plate readers was sent to John Butler (Meso Scale Discovery Applications Support) who assisted with extracting the data from samples with the same participant identifier.
Proadrenomedullin
The University of Liverpool’s Department of Clinical Infection, Microbiology and Immunology Form B·R·A·H·M·S Kryptor Compact SOP was followed. MR-pro ADM was detected using the B·R·A·H·M·S Kryptor Compact (provided by Thermo Fisher, www.thermofisher.com/uk/en/home/clinical/diagnostic-testing/brahms/kryptor/models/kryptor-compact-plus.html).
The measurement principle of B·R·A·H·M·S KRYPTOR compact is based on TRACE™ Technology (Time-Resolved Amplified Cryptate Emission, Waltham, MA, USA), which measures a fluorescence signal that is emitted from an immunocomplex. The basis of the TRACE™ Technology is a non-radiative energy transfer from a donor (europium cryptate) to an acceptor (cyanine). The proximity of donor and acceptor when they are part of an immunocomplex and the spectral overlap between donor emission and acceptor absorption spectra intensify the fluorescence signal of the donor and extend the life span of the acceptor signal, permitting the measurement of temporally delayed fluorescence. When the sample is excited with a nitrogen laser at 337 nm, the donor emits a long-lived fluorescence signal in the millisecond range at 620 nm, while the acceptor generates a short-lived signal in the nanosecond-range measured at 665 nm. When the two components are bound in an immunocomplex, both the signal amplification and the prolongation of the life span of the acceptor signal occur at 665 nm, so that it can be measured over microseconds. Unfortunately, some samples for MR-proADM analysis had clots and analysis needed to be repeated, these samples were centrifuged and carefully pipetted out into new tubes.
Statistical analyses
Statistical software
R version 4.4.1 (R Core Team. R: A Language and Environment for Statistical Computing. In R Foundation for Statistical Computing, editor. Vienna, Austria; 2022) was used for data cleaning, statistical analysis, reporting and visualisation of results.
Descriptive analyses
Screening, recruitment, sample collection and sample analysis
Summary statistics on screening, recruitment, sample collection and sample analysis are reported in the PRECISE Consolidated Standards of Reporting Trials (CONSORT) flow diagram (Figure 4).
FIGURE 4 PRECISE CONSORT diagram. a, Issues with the laboratory processing and analysis were encountered. Some samples received at the lab could not be identified, due to labelling issues. During the first analysis of samples, invalid results due to calibration issues were recorded. Samples (where there was sufficient volume of sample) were re-analysed. (Figure not reproduced here; see the original article.)
Descriptive summary
Descriptive summaries of baseline and serial biomarker measurements and of the number of observations per patient were generated.
Baseline characteristics
Participant characteristics were reported as frequencies and percentages, means and standard deviations, or medians and interquartile ranges, as appropriate. Baseline characteristics were reported by trial arm for all patients included in the PRECISE study. There was no statistical comparison (e.g. using hypothesis tests) of baseline characteristics.
Analysis population
All PRECISE participants remained in the trial arm assigned by randomisation in the host trial, regardless of protocol deviations or non-adherence, and were included in analyses if outcome and biomarker data were available.
Missing data
Missingness within the PRECISE study was reported for each biomarker and outcome. Patients with missing biomarker data were not included in the analysis involving that biomarker.
Missing data on IV antibiotic stopping time will lead to censoring at the patient’s last available clinical review. Patients with missing outcome data can therefore still be included in the Cox regression model, under the assumption that this censoring is non-informative.
Statistical analyses
The MeMed Ltd. ImmunoXpertTM assays were not able to be performed as planned because of the blood sample collection method (plasma instead of serum). Instead, the individual values of TRAIL, IP-10 and CRP were measured. Since CRP at baseline was also measured and analysed locally at sites as part of the main BATCH trial, analyses involving CRP values obtained from the entire BATCH data set were also performed.
A participant’s baseline biomarker value was defined as the one closest to their randomisation date and within ± 1 day of randomisation. If multiple biomarker measurements were recorded for the same participant on the day of randomisation, or if multiple CRP measurements were recorded 1 day before or after randomisation for the same participant, the mean across these measurements was used (because only the date but not the time of randomisation was recorded).
The interaction between BATCH treatment arm and biomarker was tested, separately for each of the biomarkers (TRAIL, IP-10, CRP, MR-proADM) at baseline. The interaction was assessed separately for the continuous biomarker value (either on its original scale or log-transformed, as appropriate) and dichotomised using different cut-offs. CRP was categorised according to the following cut-offs: > 25 mg/l, > 50 mg/l and > 100 mg/l.43 MR-proADM was categorised according to the following cut-offs: > 0.5 nmol/l and > 0.7 nmol/l.7 TRAIL and IP-10 were treated as continuous because there are no well-established cut-offs.
Primary outcome
For the primary outcome analysis, a Cox proportional hazards model with time was fitted until IV antibiotic therapy was discontinued as dependent variable and both intervention arm and biomarker (or biomarker subgroup) at baseline as independent variables (fixed effects). The minimisation factors were also included as covariates, with age as a fixed effect and site as a random effect. Then a second model was fitted with an additional interaction term between the fixed effects of trial arm and biomarker (or biomarker subgroup). The models with and without the interaction were compared using a likelihood-ratio test (LRT). We reported the LRT χ2 statistic and corresponding p-values. This analysis was performed separately for each biomarker (CRP, TRAIL, IP-10, MR-proADM) at baseline and, where applicable, biomarker cut-off. To correct for multiple testing, the Benjamini–Yekutieli adjustment44 to the p-values was applied as implemented in the R function p.adjust, with the aim of controlling the false discovery rate at 5% across outcomes, biomarkers and thresholds.
Secondary outcomes
Analyses of secondary outcomes followed the same approach as outlined above, using regression models depending on the type of outcome variable: Cox regression for time-to-event outcomes (time until discontinuation of any antibiotics, time until discontinuation of broad-spectrum antibiotics, time until discharge from hospital), and logistic regression for a binary outcomes (SADRs, HAIs). Full details of all analyses were set out in the BATCH and PRECISE statistical analysis plan which was finalised and published prior to any data being analysed.45