BATCH is a multi-centre, prospective, individually randomised, parallel, two-arm, controlled trial. The trial will assess the effectiveness of an additional PCT test in children (aged between 72 h and 18 years) hospitalised with suspected or confirmed bacterial infection to guide antimicrobial prescribing decisions.

Co-primary outcomes (antibiotic duration and a composite safety outcome) will be used to answer the primary research objectives of the study. Differences in antibiotic duration will be investigated for superiority of the intervention over standard care, while differences in the composite safety outcome will be investigated for non-inferiority [6].

Intervention

In children randomised to the intervention arm, a blood sample will be sent to the hospital laboratory for a PCT test at baseline/randomisation and every 1–3 days while still on IV antibiotics to align with clinical workflow and routine laboratory testing where possible. PCT results feed into an algorithm [7] that provides both definitive and advisory guidelines, based on thresholds of PCT concentration and PCT change [8]. Clinicians can overrule the algorithm if they feel it is appropriate to do so. Children in the control arm will not have the PCT test performed. The trial protocol [3] includes further detail on the implementation of the intervention.

Randomisation

Participants will be randomised in a 1:1 ratio to receive either current clinical management alone (control) or clinical management with the addition of PCT test guidance (intervention). Patients will be randomised by minimisation [9], with site and age group as minimisation factors and a random element to reduce predictability.

Participants will be randomised remotely using a secure 24-h web-based randomisation programme controlled centrally by the Centre for Trials Research at Cardiff University. Details of the age group cut-offs and random element are documented in a separate randomisation protocol and will be concealed from the treating teams.

Sample size

The trial has two co-primary outcomes [6] (“Co-primary outcomes” section), and the overall sample size is determined by both. The focus for the intervention is on moving the step down from IV to oral therapy earlier, and therefore, the time until this step down is our primary outcome on antibiotic usage (overall usage across both oral and IV is a secondary outcome). The study is powered to detect if PCT-directed care is superior to standard care on time until switch from IV antibiotics. The size of potential shortening of time to detect an effect has been taken from a systematic review [10]. The safety co-primary outcome is a composite measure reflecting various outcomes which represent deterioration or lack of clinical response in the child, and it would therefore be expected to increase if IV antibiotics were being withdrawn inappropriately early.

A 1-day reduction [10] in IV antibiotic duration from an estimated median of 5 days [11] in the control arm implies a hazard ratio of 1.25 (assuming proportional hazards and an exponential survival distribution). At 5% two-sided significance level with 90% power, 844 participants with observed IV antibiotics duration are needed. In terms of the event rates of safety elements, we estimate an admission/re-admission rate of 8.8% [11]. In critically ill patients, up to 3% reinstating IV antibiotic therapy rate, and 4% mortality were reported [10, 12]. With some overlaps considered, we estimate an overall rate of about 15% for the composite safety outcome. A previous trial on PCT-guided antibiotic therapy in adults used a non-inferiority margin of 8% risk difference for mortality [13]. Given the lower expected rate of safety outcomes in this population, we have chosen a similar (relative) non-inferiority margin of 5% (absolute) risk difference for the composite safety outcome. This means that an increase of no more than 5% (from 15% to 20%) using PCT-guided therapy would be considered non-inferior. To test non-inferiority with a one-sided significance level of 0.05 and 90% power would therefore require 874 participants per arm. Overall, with 1748 effectively recruited participants, we would have 99% power to detect an antibiotic duration decrease corresponding to a hazard ratio of 1.25 and 90% power to test non-inferiority in safety separately. This means that the power of the combined analysis would be at least 89% (if the co-primary outcomes are independent) and at most 90% (if the co-primary outcomes are colinear) [14]. Allowing for up to 10% loss to follow-up, our target sample size is inflated to 1942.

Data collection schedule

Participant data are collected at the following time points:

At baseline (baseline characteristics and admission data)

Daily post-randomisation until discharged home (antibiotic use, adverse events, and clinical data)

Day 28 telephone follow-up (healthcare utilisation and quality of life questionnaire)

Definition of non-adherence

There will be multiple measures of non-adherence, reflecting different stages of the clinical decision-making. Reasons for non-consideration of the PCT result or non-adherence to the algorithm can be broken down into three steps (Table 1). Steps 2 and 3 are only relevant when step 1 has been adhered to, i.e., when a PCT result is available. In cases where the PCT result was available and was considered, this will be considered adherence to the intervention policy specified in the trial protocol, regardless of the actual clinical decision [3].

Table 1 Types of non-adherence

For those patients whose clinical reviews were adherent to the intervention policy (steps 1 and 2), we will investigate step 3a by comparing the actual clinical decision with the recommendation given by the PCT-guided algorithm. For patients whose procedures were adherent to step 1, we will separately investigate step 3b by comparing the actual clinical decision with the recommendation given by the PCT-guided algorithm, regardless of adherence to step 2.

In this clinical context, contamination of the control arm is very unlikely, i.e., we do not expect that any patient in the control arm will have a PCT test done.

Analysis populations

All randomised participants will remain in their originally assigned groups, regardless of protocol deviations or non-adherence, and will be included in all analyses if outcome data are available.

In one of the planned secondary analyses (“Sensitivity analyses” section), we will estimate the complier average causal effect (CACE) [15] to account for departures from the randomised intervention. For the purposes of this sensitivity analysis, we will define different analysis populations depending on the level of adherence with the PCT-guided algorithm (as defined in the “Definition of non-adherence” section):

Patients whose PCT result was available at the time of the clinical review

Patients whose PCT result was available and was considered by the clinician (per protocol)

Patients whose PCT result was available and was considered and where the actual clinical decision agreed/did not agree with the recommendation of the PCT-guided algorithm

Patients whose PCT result was available and where the actual clinical decision agreed/did not agree with the recommendation of the PCT-guided algorithm, regardless of whether it was considered by the clinician

We will also consider adherence longitudinally (“Additional exploratory analyses” section). Most patients will have several PCT measurements taken at different points in time and availability/consideration/adherence may be different at different time points for the same child.

Reporting

Final analysis of the primary and secondary outcomes will take place when all randomised participants have completed their day 28 telephone follow-up, all forms have been received, and the datasets have been locked. The trial report will follow the guidelines of Consolidated Standards of Reporting Trials (CONSORT) for reporting randomised controlled trials [16] and its extension to non-inferiority designs [17].