Amoxicillin duration and dose for community-acquired pneumonia in children: the CAP-IT factorial non-inferiority RCT

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 13/88/11. The contractual start date was in March 2016. The draft report began editorial review in October 2020 and was accepted for publication in June 2021. 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.

Copyright statement

Copyright © 2021 Barratt et al. This work was produced by Barratt et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This is an Open Access publication distributed under the terms of the Creative Commons Attribution CC BY 4.0 licence, which permits unrestricted use, distribution, reproduction and adaption in any medium and for any purpose provided that it is properly attributed. See: https://creativecommons.org/licenses/by/4.0/. For attribution the title, original author(s), the publication source – NIHR Journals Library, and the DOI of the publication must be cited.

2021 Barratt et al.

Background

Antibiotics are among the most frequently prescribed medicines for children worldwide. 2,3 In the UK, Italy and the Netherlands, almost 50% of children have received antibiotics by their second birthday. Annually, it is estimated that 30% of children aged 2–11 years receive antibiotics. 3

Of the possible indications in children aged < 5 years, the most common are acute respiratory tract infections, including community-acquired pneumonia (CAP). 4–6 CAP is one of the most common serious bacterial childhood infections. Although the majority of pneumonia deaths occur in low- and middle-income countries, CAP is a major cause of morbidity in Europe and North America. 5,7 In the UK, 62% of all antibiotics prescribed for community-acquired infections are for CAP. 8 In the USA, respiratory symptoms, fever or cough are responsible for one-third of all childhood medical visits, and 7–15% of these children will be diagnosed with CAP. 9,10

Emergency department (ED) attendances and hospital admissions of children with respiratory complaints have increased in recent decades, mostly in preschool children. 9,11,12 According to Hospital Episode Statistics,13 children aged 0–4 years accounted for around 2.11 million ED attendances in 2017–18. More than 11,000 children aged < 15 years were admitted to hospitals in England with a diagnosis of bacterial pneumonia in 2008, and 9000 1- to 4-year-old inpatients with non-influenza pneumonia were recorded in 2012/13. 13,14

The bacterial pathogen most commonly associated with childhood CAP is Streptococcus pneumoniae, including in countries where pneumococcal conjugate vaccine (PCV) is routinely administered. 7,15–17 In 2010, PCV13 (which covers 13 S. pneumoniae serotypes) was introduced in the UK, with almost 95% uptake in young children. 18,19 However, despite an observed impact on invasive pneumococcal disease, a decrease in CAP-related hospital admissions in young children has not been observed. 11,14,20,21

What are the current challenges in the management of childhood community-acquired pneumonia?

There is no test capable of accurately distinguishing between bacterial and viral CAP. 22 Interobserver agreement for chest radigoraphic findings is poor, casting doubt on the usefulness of chest radiographs for identifying bacterial CAP, and culturing of microbiological samples, such as sputum, has low diagnostic value and samples are often difficult to take from young children. 23–25 Diagnosis of bacterial CAP presents a challenge for treating clinicians, who rely largely on clinical criteria. 22 Children presenting with fever, raised respiratory rate, focal chest signs and other respiratory signs and symptoms (such as cough) are commonly ascribed a diagnosis of bacterial CAP,10,26–28 whereas wheezing is associated with the absence of radiographic pneumonia and failure to detect bacteria in clinical samples. 26,29 If bacterial CAP is considered the likely diagnosis, treatment with antibiotics is instituted. 10,30 This diagnostic challenge is particularly problematic in secondary care, where the proportion of children presenting with serious bacterial infections is higher than in primary practice. 31,32

A further challenge for clinicians is severity assessment. Available validated predictive scoring systems for CAP severity include the Pneumonia Severity Index and the CURB-65 (confusion, urea, respiratory rate, blood pressure, and 65 years of age or older) score, but these are not applicable to children. 33,34 Pneumonia mortality risk scores for children have been developed in low-resource settings, but do not differentiate between viral and bacterial pneumonia. 35,36 Low oxygen saturation in room air is included as one component in these risk scores, and is an important factor for differentiating between non-severe and severe pneumonia. 37–39

Finally, assessing the efficacy of childhood CAP treatment is complex. Key measures in studies assessing efficacy early in the treatment course include lack of improvement or worsening of clinical symptoms and signs, such as respiratory rate and oxygen saturation. 40 According to British Thoracic Society (BTS) guidance, such criteria should trigger clinical review of children treated with oral antibiotics for CAP,22 including where the following features are present at 48 hours: (1) persistent high fever, (2) increasing or persistently increased effort of breathing and (3) persistent or increasing oxygen requirement to maintain saturations ≥ 92%. 22 Approximately 15% of children with CAP receive further antibiotics within 28 days of starting treatment because of symptoms that concern parents. 41–44 However, only half of children show recovery from symptoms of acute respiratory illness by day 9 or 10, and 90% of children recover by 3.5 weeks after symptom onset. 45–47

What are the current management recommendations for childhood community-acquired pneumonia?

Amoxicillin is the drug of choice for the treatment of childhood CAP according to the British National Formulary for Children (BNFc) and BTS and National Institute for Health and Care Excellence guidelines, as well as several international guidelines,22,48–51 as it can effectively target and treat S. pneumoniae in the absence of high-level penicillin resistance. As a result, amoxicillin accounts for a very high proportion of overall oral antibiotic use among young children in many settings. Despite this, there is insufficient evidence to inform optimal treatment dose or duration.

What is the impact of antimicrobial resistance in childhood community-acquired pneumonia?

In the UK, the rates of penicillin non-susceptibility of S. pneumoniae are relatively low, at approximately 15% for respiratory samples (mainly from adults) and 4–6% for blood culture isolates. 59 Penicillin resistance [i.e. minimal inhibitory concentration (MIC) > 2 μg/ml] has not been observed in blood culture isolates and has been found in < 1% of respiratory S. pneumoniae isolates in the UK since 2010. 59 However, some worrying trends are observed in resistance of gut bacteria, and this situation will be exacerbated in a setting where antibiotics are used injudiciously. 60

The relationship between MIC (an in vitro phenomenon) and clinical outcome in CAP is complex, and data on the level of S. pneumoniae antimicrobial resistance that reduces amoxicillin effectiveness are limited. Harmonisation of European breakpoints (i.e. the MIC at which an isolate is considered susceptible, intermediate or resistant) attempts to provide a link between clinical impact and in vitro observation of resistance. 61 Clinical breakpoints are determined based on a variety of data, in addition to efficacy studies. This includes pharmacokinetic/pharmacodynamic data, which for penicillin usually take T > MIC of 40% as the key exposure measure.

Children have high rates of bacterial colonisation, which often represents an increased level of carriage of resistant organisms62,63 These may be passed on to others in the community, especially within child-care settings. 64,65 Interventions to maintain a low level of antimicrobial resistance among colonising bacteria may, therefore, have population implications.

The limited existing data on the specific impact of duration and dose of antibiotic treatment on subsequent colonisation with resistant bacteria in vivo suggest a complex and dynamic relationship. 62–73 Experimental models suggest that insufficiently high dosing could promote selection of resistant pathogens. In addition, although most of the effect on bacterial load is achieved early during antibiotic exposure, resistant isolates emerge after 4 or 5 days. 74–78 RCTs assessing the effect of antibiotic duration and dose have been called for, as they will probably provide the strongest evidence for the relationship between antibiotic exposure and colonisation with resistant bacteria. 79 One such RCT found that higher-dose shorter-duration amoxicillin therapy for childhood CAP led to less colonisation with resistant bacteria after 4 weeks, and was associated with better treatment adherence. 72 However, mathematical modelling indicates that this may come at the price of selecting isolates with higher levels of resistance, and clinical efficacy was not addressed in the trial. 72,78

Trial rationale

Despite the reduction in incidence of invasive pneumococcal disease since the introduction of the conjugate vaccine,20 CAP remains one of the most commonly identified and treated childhood infections in the UK. Although there is clear agreement that amoxicillin should be the first-line treatment, there are insufficient data to inform selection of dose and duration, and the impact that different regimens have on antimicrobial resistance is unknown.

Effectiveness and resistance-outcome data pertaining to dose and duration of amoxicillin could inform antimicrobial stewardship strategies in the large group of children with a high likelihood of bacterial CAP targeted by CAP-IT (Community-Acquired Pneumonia: a protocol for a randomIsed controlled Trial). A better understanding of the relationship between dose and duration of antibiotic treatment, and the impact on clinical outcomes and antimicrobial resistance, would make it possible to formulate improved evidence-based treatment recommendations for childhood CAP.

Objectives

The main objective of CAP-IT was to determine the following for young children with uncomplicated CAP treated after discharge from hospital if:

• a 3-day course of amoxicillin is non-inferior to a 7-day course, determined by receipt of a clinically indicated systemic antibiotic other than trial medication for respiratory tract infection (including CAP) in the 4 weeks after randomisation up to day 28

• lower-dose amoxicillin is non-inferior to higher-dose amoxicillin under the same conditions.

Secondary objectives were to evaluate the impact of lower-dose and shorter-duration amoxicillin on antimicrobial resistance, severity and duration of parent/guardian-reported CAP symptoms and specified clinical adverse events (AEs).

Trial design

The CAP-IT study was a multicentre clinical trial with a target sample size of 800 participants in the UK and Ireland. In design, it was a randomised double-blind placebo-controlled 2 × 2 factorial non-inferiority trial of amoxicillin dose and duration in young children with CAP (Figure 1).

FIGURE 1.

The CAP-IT schema.

Trial setting

Participants were recruited from 28 UK NHS hospitals and one children’s hospital in Ireland:

1. Alder Hey Children’s Hospital NHS Foundation Trust (Liverpool, UK)

2. Barts Health NHS Trust (London, UK)

3. Birmingham Women’s and Children’s NHS Foundation Trust (Birmingham, UK)

4. Brighton and Sussex University Hospitals NHS Trust (Brighton, UK)

5. Chelsea and Westminster Hospital NHS Foundation Trust (London, UK)

6. Children’s Health Ireland (Dublin, Ireland)

7. City Hospitals Sunderland NHS Foundation Trust (Sunderland, UK)

8. Countess of Chester Hospital NHS Foundation Trust (Chester, UK)

9. County Durham and Darlington NHS Foundation Trust (Darlington, UK)

10. Guy’s and St Thomas’ NHS Foundation Trust (London, UK)

11. Hull and East Yorkshire Teaching Hospitals NHS Trust (Hull, UK)

12. Imperial College Healthcare NHS Trust (London, UK)

13. King’s College Hospital NHS Foundation Trust (London, UK)

14. The Leeds Teaching Hospitals NHS Trust (Leeds, UK)

15. Manchester University NHS Foundation Trust (Manchester, UK)

16. Nottingham University Hospitals NHS Trust (Nottingham, UK)

17. Oxford University Hospitals NHS Foundation Trust (Oxford, UK)

18. Southport and Ormskirk Hospital NHS Trust (Southport, UK)

19. Royal Hospital for Children (Glasgow, UK)

20. Sheffield Children’s NHS Foundation Trust (Sheffield, UK)

21. South Tees Hospitals NHS Foundation Trust (Middlesbrough, UK)

22. St George’s University Hospitals NHS Foundation Trust (London, UK)

23. University Hospitals Bristol and Weston NHS Foundation Trust (Bristol, UK)

24. University Hospitals of Derby and Burton NHS Foundation Trust (Derby, UK)

25. University Hospitals of Leicester NHS Trust (Leicester, UK)

26. University Hospitals Lewisham (London, UK)

27. University Hospital Southampton NHS Foundation Trust (Southampton, UK)

28. University Hospital of Wales (Cardiff, UK).

Participating sites were tertiary or secondary hospitals with paediatric emergency departments (PEDs) and inpatient facilities, and were selected in collaboration with Paediatric Emergency Research in the UK & Ireland80 on the basis of clinical and research infrastructure, experience in clinical research and likely eligible population size.

Participants

Patients presenting to participating hospitals were identified in PEDs, assessment/observation units or inpatient wards. Potential participants were screened as early as possible during the initial clinical assessment. Informed consent was sought from a parent/guardian once eligibility had been confirmed, but only after full explanation of the trial aims, methods and potential risks and benefits. Discussions regarding the trial took place between families and clinical teams when the child’s clinical condition was stable, to minimise distress. Extensive information and recruitment materials were available for recruiting sites, including printed and video materials [accessible at URL: www.capitstudy.org.uk (accessed 29 July 2021)]. CAP-IT information film was designed to assist research teams in the recruitment process and provided information to parents/guardians about the purpose of the trial, the use of placebo and trial procedures. Parents/guardians could watch the film in their own time while in hospital, and research teams reported that the film was a useful tool during the recruitment process. The film was made with input from the trial patient and public involvement (PPI) representative and featured a site principal investigator and research nurse, as well as graphics to aid explanation of trial procedures. [It can be viewed at https://vimeo.com/217849985 (accessed 29 July 2021).] Families were able to decline participation in the trial at any time without providing a reason and without incurring any penalty or affecting clinical management.

Interventions

The investigational medicinal product (IMP) for treatment at home was provided as a powder to be suspended on the day of randomisation. Children received oral amoxicillin suspension twice daily, commencing on the day of randomisation. All children were weighed during eligibility screening and was used to determine dose volume according to seven weight bands (Table 1).

Participants were randomised to receive either a lower (35–50 mg/kg/day) or a higher (70–90 mg/kg/day) dose, concealment of which was achieved by using amoxicillin products of two different strengths (125 mg/5 ml and 250 mg/5 ml). Therefore, children in each dose arm in the same weight band were administered the same volume of suspension .

Participants were simultaneously randomised to receive either 3 or 7 days of amoxicillin treatment at home. A placebo manufactured to match the characteristics of oral amoxicillin suspension was used to blind parents/guardians and clinical staff to the duration allocation. Both active drug and placebo formed a yellow-coloured similar-tasting suspension. However, because of difficulties in exactly taste-matching the placebo suspension to amoxicillin, one brand of amoxicillin was used for the first 3 days of treatment followed by a second brand for days 4–7 when duration of treatment was 7 days. Parents were instructed to expect a taste change between bottles, but they did not know whether this was due to moving to placebo or to a new brand of amoxicillin. Allocated treatment duration to be given after discharge from hospital was fixed at 3 or 7 days independently of any antibiotics received before randomisation, with up to 48 hours of oral or parenteral beta-lactam treatment permitted before enrolment.

This resulted in four treatment arms, as shown in Figure 2.

FIGURE 2.

Treatment arms.

The hypothesis is that higher doses of amoxicillin given for a longer duration are non-inferior to lower doses of amoxicillin given for a shorter duration for the treatment of children attending hospital with CAP in terms of antibiotic retreatment.

The objective is to conduct a RCT in children attending hospital with CAP comparing higher and lower doses of amoxicillin given for 3 or 7 days.

Trial assessments and follow-up

Participants were screened as described in Participants, and, following receipt of informed consent, randomisation was performed at the point of discharge from hospital. Following randomisation, all participants were followed up for 29 days for evaluation of the primary and secondary end points described in Outcomes. The timing and frequency of assessments are summarised in the trial schedule (Table 2) and described below.

Randomisation

Eligibility was confirmed by CAP-IT site investigators through completion of an eligibility checklist. Patients were randomised simultaneously to each of the two factorial randomisations in a 1 : 1 ratio. Randomisation was stratified by group (PED and ward) according to whether or not they had received any non-trial antibiotics in hospital before being enrolled.

A computer-generated randomisation list was produced by the trial statistician based on random permuted blocks of eight. Each block contained an equal number of the four possible combinations of dose and duration in random order. The IMP supplier packaged the trial medication into kits that were grouped into blocks of eight, in accordance with to the randomisation list specification. Blinded IMP labels were applied to each kit, which contained the kit identifications (IDs). Kit IDs were made up of four numerical digits, the first three of which represented the block ID and fourth specified the kit ID within the block. Blinded randomised blocks of IMP were delivered to trial sites and participants were randomised by dispensing the next sequentially numbered kit within the active block.

Blinding

All treating clinicians, parents/guardians and outcome assessors [including End-Point Review Committee (ERC) members] were blinded to the allocated treatment. The use of placebo, as well as the permuted block randomisation strategy and blinded drug kits, ensured that parents and clinic staff remained blinded to amoxicillin duration and dose.

Access to the randomisation list was restricted to trial statisticians and IMP repackagers, and unblinded data were reviewed confidentially only by the Independent Data Monitoring Committee (IDMC) (annually) and trial statisticians. The Trial Management Team remained blinded until after the trial end and completion of the statistical analysis in accordance with the prespecified statistical analysis plan (SAP).

Unblinding was possible in situations where a treating clinician deemed it necessary, for example in the case of a significant overdose. This could be performed using an emergency unblinding system accessible through the CAP-IT website. Only the treating clinician would then be informed of the child’s allocation, maintaining the blinding of the trial team.

Outcomes

Sample size

The sample size was based on demonstrating non-inferiority for the primary efficacy end point for each of the duration and dose randomisations. Although inflation factors have been advocated for factorial trials to account for interaction between the interventions, or a reduction in the number of events, this is not necessary if either randomised intervention (dose or duration) has a null effect (i.e. the underlying hypothesis with a non-inferiority design), as marginal analyses can then be conducted.

The expected antibiotic retreatment rate was originally assumed to be 5%. However, data emerging during the enrolment phase suggested that the primary outcome event rate was considerably higher, at approximately 15%. This necessitated a change in the non-inferiority margin, which was increased from 4% to 8%. This is still lower than the European Medicines Agency’s recommendation of a 10% non-inferiority margin for adult CAP trials. 83 Assuming a 15% event rate, 8% non-inferiority margin (on a risk difference scale) assessed against a two-sided 90% confidence interval (CI) and 15% loss to follow-up, the sample size was calculated as 800 children to achieve 90% power.

Statistical methods

Interim analyses

The trial was reviewed by the CAP-IT IDMC. They met three times over the course of the trial: once at a joint meeting with the Trial Steering Committee (TSC) in June 2017 and twice in strict confidence in January 2018 and January 2019. The IDMC reviewed unblinded safety and efficacy data and made recommendations through correspondence to the TSC following each meeting.

Patient and public involvement

Parents of young children were involved during the development and delivery of CAP-IT. A PPI representative was a member of the TSC, contributing at meetings and in an ad hoc fashion when required. When considering the research question, the trial team were advised by parents that shorter antibiotic courses would be welcomed if equally effective, because of difficulties in giving medicine (due to palatability or challenges with day care and daytime doses). For the same reasons, parents supported the twice-daily dosing of the CAP-IT. Multiple PPI representatives reviewed and provided input on the patient information materials, including the CAP-IT information film, to ensure that they were clear, easy to understand and not off-putting to parents, while still providing sufficient detail to allow informed consent. Valuable input was provided from the PPI representative on the CAP-IT TSC on the plan for dissemination of the CAP-IT results.

Protocol amendments

The CAP-IT protocol v.2.0 was active when recruitment to CAP-IT commenced in January 2017. Two protocol amendments were completed subsequently, with version 3 implemented in September 2017 and version 4 in December 2018. Amendments were largely in relation to selection criteria (see Exclusion criteria) and the SAP (to which three significant updates were made on the basis of accumulating trial data). First, a stratified analysis was originally planned based on the PED and ward groups. This was changed to a joint analysis in protocol version 3 because of significant clinical overlap (see Appendix 1 for more details) Second, the primary end-point definition was made more specific in protocol version 3 and further refined in version 4 (see Changes to primary end point). Finally, the non-inferiority margin was adjusted, as the primary end-point event rate had been substantially underestimated. The trial and all substantial amendments were approved by the London – West London & GTAC Research Ethics Committee (reference 16/LO/0831).

Participant flow

Between 1 February 2017 and 23 April 2019, a total of 2642 children were assessed for eligibility and 824 were randomised. Ten patients were randomised but received no trial medication (owing to, for example, a change of mind by parent/guardian or administrative error) and were, therefore, excluded from the analysis, resulting in an analysis population of 814 patients.

A total of 591 participants had no pre-treatment antibiotic at trial entry. A total of 223 (mainly following admission to assessment units of wards) had received beta-lactam antibiotic pre-treatment for no more than 48 hours. The final follow-up visit occurred on 21 May 2019, which was considered the trial end date.

Six participants were randomised in error but were included in the analysis in accordance with the ITT principle. Of these participants, five did not have all the required symptoms to fulfil the criteria for CAP diagnosis (see Box 1). One patient did not have a cough reported in the previous 96 hours at presentation, two patients did not have a reported fever in the previous 48 hours at presentation and two patients lacked documentation of signs of laboured/difficult breathing and/or focal chest signs at presentation. In one of the final two patients, chest radiography was suggestive of lobar pneumonia, prior to this being added to the inclusion criteria as part of protocol version 4.0, and in the other participant pneumonia was diagnosed on chest radiography, but was documented as patchy infiltrate, which did not fulfil the inclusion criteria. The final patient randomised in error received an antibiotic other than a beta-lactam (clarithromycin) before discharge (Table 3).

Participants were well distributed between arms, with 208 (25.6%) participants receiving 3 days of lower-dose treatment, 202 (24.8%) participants receiving 7 days of lower-dose treatment, 205 (25.2%) participants receiving 3 days of higher-dose treatment and 199 (24.4%) participants receiving 7 days of higher-dose treatment (Figure 3 and Table 4).

FIGURE 3.

A CONSORT (Consolidated Standards of Reporting Trials) flow diagram. a, Inpatient stay > 48 hours and treated with non-beta-lactam antibiotics as inpatients; b, these children have been excluded from all analyses; and c, follow-up included up to time of withdrawal or no further contact.

Baseline

Parent/guardian-reported community-acquired pneumonia symptoms

Parent/guardian-reported symptom severity at trial entry is shown in Figure 4. The most common clinical symptom was cough, reported by 96.5% of participants. Fever and fast breathing were reported for 79.6% and 83.5% of participants, respectively, and the least common symptoms at baseline were vomiting and wheeze, reported in 41.1% and 51.8% of participants, respectively. Sleep disturbance, eating less and interference with normal activity were reported in between 80% and 90% of participants.

FIGURE 4.

Symptoms at trial entry.

Clinical symptoms in patients who received in-hospital antibiotics prior to trial entry (i.e. the ward group) were reported by parents/guardians at presentation (pre trial) and at baseline (trial entry). Figures 5 and 6 show parent/guardian-reported clinical symptom severity both pre trial and at trial entry for the ward group and at trial entry only for the PED group. For the ward group, the proportion of participants with presence of symptoms at any level of severity decreased between pre trial and trial entry for all symptoms except wet cough (phlegm). The greatest proportional decrease was for fever, for which the proportion of participants with a severity of slight/little or greater decreased from 87.9% to 50.2%.

FIGURE 5.

Clinical symptoms (i.e. fever, cough, phlegm and breathing fast) at trial entry, by group.

FIGURE 6.

Clinical symptoms (i.e. wheeze, sleep disturbance, vomiting, eating less and abnormal activity) at trial entry, by group.

Community-acquired pneumonia symptoms at trial entry, by stratum, are shown in Appendix 2, Table 27.

Follow-up

Of the 814 patients included in the analysis, 642 (79%) completed the final assessment. Where possible, this final assessment was carried out face to face at hospital or at home, but if this proved impossible (e.g. if parents/guardians were unable to attend an appointment), then the assessment was completed by telephone. Overall, 25% of final assessments were performed by telephone, 74% were performed in hospital and 1% were performed at home. In 172 (21%) participants, the final assessment was not conducted with the family. Of these 172 participants, 11 had withdrawn consent and a further 161 could not be contacted. However, 150 of these participants (87%) had provided consent for collection of the primary outcome via hospital and GP records, and primary outcome data were successfully collected in 144 of these participants. This ensured that primary outcome data were available for 786 (97%) participants, and only 28 participants (3%) were considered withdrawn or lost to follow-up (Table 12).

Follow-up data were also collected by telephone at days 3, 7, 14 and 21 (Table 13). Follow-up rates were 88% at day 3, 75% at day 14 and 76% at day 21. A total of 443 (54%) parents/guardians of participants completed all telephone calls and the final visit, with 153 (19%) parents/guardians of participants missing one follow-up visit, 95 (12%) parents/guardians of participants missing two follow-up visits, 51 (6%) parents/guardians of participants missing three follow-up visits and 48 (6%) parents/guardians of participants missing four follow-up visits. Twenty-four (3%) parents/guardians of participants missed all telephone calls and visits.

A symptom diary was to be completed daily by parents/guardians for the first 14 days after trial entry. Completed diary data were available for 406 (49.9%) participants and no diary data were available for 227 (27.9%) participants. Parents/guardians were assigned to complete symptom diaries either electronically (42.5%) or on paper (57.5%) using pseudorandomisation. Summary data on diary completion are presented in Table 14.

Adherence

A total of 240 (29.5%) participants deviated from the prescribed IMP regimen for reasons including taking fewer doses or a lower volume, taking too many doses or a greater volume, or deviation in timing (Table 15).

For dose randomisation, there was no evidence of an overall difference in adherence deviation between the two arms (p = 0.21). However, a greater proportion of participants in the lower-dose arm (7.3%) than in the higher-dose arm (4%) did not take bottle B/C as prescribed (p = 0.038).

For duration randomisation, 134 (32.4%) participants in the shorter-duration arm deviated, compared with 106 (26.4%) participants in the longer-duration arm (p = 0.06). A greater proportion of participants in the shorter-duration arm (13.3%) than in the longer-duration arm (9.4%) did not complete trial treatment (p = 0.015) (see Table 15).

Primary outcome

Analysis of primary end point

Overall

Overall, 100 participants in the analysis population (n = 814) met the criteria for a primary end point during the follow-up period (i.e. a cumulative proportion of 12.5%, 90% CI 10.7% to 14.6%, as estimated with Kaplan–Meier methods).

Dose randomisation

The observed number of primary end points was similar in the lower-dose arm (n = 51, 12.6%) and in the higher-dose arm (n = 49, 12.4%). The estimated risk difference at day 28 was 0.2% (90% CI –3.7% to 4.0%), meeting the criterion for non-inferiority (Figure 7).

FIGURE 7.

Kaplan–Meier curve for primary end point: dose randomisation.

Duration randomisation

A total of 51 (12.5%) participants experienced a primary end point in the shorter-duration arm and 49 (12.5%) participants experienced a primary end point in the longer-duration arm. The estimated risk difference at day 28 was 0.1% (90% CI –3.8% to 3.9%), again satisfying the non-inferiority criterion (Figure 8).

FIGURE 8.

Kaplan–Meier curve for primary end point: duration randomisation.

Interaction effects

The outcomes for the analyses of interaction effects between the two randomisations (i.e. dose and duration), between pre-exposure to antibiotics and dose randomisation and between pre-exposure to antibiotics and duration randomisation are shown in Figures 9–11, respectively.

FIGURE 9.

Kaplan–Meier curve for analysis of interaction between the two randomisations.

FIGURE 10.

Kaplan–Meier curve for analysis of interaction between pre-exposure with antibiotics and dose randomisation.

FIGURE 11.

Kaplan–Meier curve for analysis of interaction between pre-exposure with antibiotics and duration randomisation.

There was no evidence of an interaction between either of the two randomisation arms (p = 0.625), between the dose randomisation arm and pre-exposure to antibiotics (p = 0.456) or between the duration randomisation arm and pre-exposure to antibiotics (p = 0.592). This justifies analysis of the ‘main effects’ for the two randomisations (see Figures 7 and 8).

Primary end-point sensitivity analyses

The results of the sensitivity and subgroup analyses are summarised in Figures 12 and 13. Non-inferiority was demonstrated for all sensitivity analyses for both dose and duration comparisons.

FIGURE 12.

Forest plot summarising sensitivity and subgroup analyses outcomes in terms of difference in proportions of retreatment by day 28 for the dose randomisation.

FIGURE 13.

Forest plot summarising sensitivity and subgroup analyses outcomes in terms of difference in proportions of retreatment by day 28 for the duration randomisation.

On-treatment analyses

The on-treatment analyses gave very similar results to the primary analysis. For both the dose and the duration comparison, the upper 90% CI limit of the estimated difference at day 28 was lower than the non-inferiority margin of 8% for both definitions of non-adherence (see Appendix 3, Figures 21–24).

Subgroup analyses

Participants with severe community-acquired pneumonia

This a priori subgroup analysis repeated the primary analysis, limited to participants defined as having severe CAP. Table 18 shows the total number (%) of participants with each abnormality by randomisation group. Only 155 (19%) participants had none of these abnormalities at presentation; 291 (35.7%) participants had one, 341 (41.9%) had two and 27 (3.3%) had three. A total of 368 (45.2%) participants were included in the subgroup analysis.

TABLE 18 - Abnormalities at presentation considered for subgroup analysis for severe CAP

Abnormality	Treatment arm, n (%)		p-value	Treatment arm, n (%)		p-value	Total (N = 814), n (%)
	Lower dose (N = 410)	Higher dose (N = 404)		Shorter duration (N = 413)	Longer duration (N = 401)
Chest retractions	239 (58.4)	244 (60.4)	0.57	239 (58.0)	244 (60.8)	0.41	483 (59.4)
Oxygen saturation < 92%	18 (4.4)	25 (6.2)	0.25	18 (4.4)	25 (6.3)	0.23	43 (5.3)
High respiratory rate	270 (65.9)	258 (64.3)	0.65	262 (63.7)	266 (66.5)	0.41	528 (65.1)
Number of abnormalities			0.62			0.47
0	75 (18.3)	80 (19.8)		82 (19.9)	73 (18.2)		155 (19.0)
1	155 (37.8)	136 (33.7)		154 (37.3)	137 (34.2)		291 (35.7)
2	168 (41.0)	173 (42.8)		166 (40.2)	175 (43.6)		341 (41.9)
3	12 (2.9)	15 (3.7)		11 (2.7)	16 (4.0)		27 (3.3)
> 1	180 (43.9)	188 (46.5)	0.45	177 (42.9)	191 (47.6)	0.17	368 (45.2)

In total, 56 (15.4%) participants experienced a primary end point. There was no significant difference between the arms for either the dose comparison (p = 0.283) or the duration comparison (p = 0.821). For duration randomisation, the estimated risk difference at day 28 was 1.2% (90% CI –5.0% to 7.4%) (Figure 14). For dose randomisation, the estimated difference at day 28 was 3.8% (90% CI –2.4% to 10.0%). This is consistent with no effect, although the 90% CI crossed the non-inferiority margin (Figure 15).

FIGURE 14.

Kaplan–Meier curve for severe CAP subgroup primary analysis for duration randomisation.

FIGURE 15.

Kaplan–Meier curve for severe CAP subgroup primary analysis for dose randomisation.

Streptococcus pneumoniae carriage and resistance

Carriage and resistance to penicillin of S. pneumoniae isolates were assessed by analysis of nasopharyngeal samples collected from participants at baseline, at the final visit and at any unscheduled visits during follow-up.

Community-acquired pneumonia symptoms

Parent/guardian-reported symptom data were elicited at follow-up telephone calls and through parental/guardian completion of a daily diary up to day 14. The proportion of participants for whom parent/guardian-reported symptom data from any source were available reduced from days 3 (93%), 7 (88%), 14 (83%) and 21 (76%) to day 28 (75%) (Figure 16).

FIGURE 16.

Availability of symptom data over time, by data source. a, No data available because telephone call/visit did not happen or no data were reported.

Time to resolution of community-acquired pneumonia symptoms: overall

Severity was elicited for nine CAP symptoms, each of which was analysed separately in terms of time to resolution. As multiple comparisons were performed, the p-value from each individual analysis needs to be interpreted cautiously. Participants were included in the analysis only if a symptom was present at trial entry. Cough had the longest median time to resolution (11 days), followed by the related symptom wet cough (phlegm) (6 days). An estimated 20% of participants still had cough symptoms at day 28 (Figure 17). Vomiting and fever both resolved rapidly, in a median of 1 day and 2 days, respectively. The remaining symptoms had a median time to resolution of between 3 and 5 days.

FIGURE 17.

Kaplan–Meier curves for time to symptom resolution across all randomisation arms. Participants excluded if symptom not present at enrolment.

Time to resolution of community-acquired pneumonia symptoms: duration randomisation

For duration randomisation, there was no significant difference between groups for seven symptoms (log-rank p > 0.05). However, there was a difference in time to resolution of cough (p = 0.040) and sleep disturbed by cough (p = 0.026), with a significantly faster time to resolution in the longer-duration arm in both cases (Figures 18 and 19).

FIGURE 18.

Kaplan–Meier curve for time to resolution of cough in the duration treatment arms. Participants excluded if symptom not present at enrolment.

FIGURE 19.

Kaplan–Meier curve for time to resolution of sleep disturbed by cough in the duration treatment arms. Participants excluded if symptom not present at enrolment.

Adverse events

Specified clinical adverse events (diarrhoea, thrush and skin rash)

Presence and severity of diarrhoea, thrush and skin rash were elicited from parents at trial entry and throughout follow-up. Diarrhoea was the most common clinical AE and was present in 345 (43.6%) participants after baseline. Skin rash was present in 193 (24.4%) participants and oral thrush in 57 (7.2%) participants after baseline. For both dose and duration randomisations, there was no evidence of a difference between the randomised arms in terms of overall prevalence of diarrhoea and oral thrush after baseline (Table 24). For skin rash, there was some evidence of a difference between shorter- and longer-duration arms in terms of prevalence after baseline, with the number of participants ever having skin rash after baseline being 106 (27.4%) in the longer-duration arm and 87 (21.5%) in the shorter-duration arm (p = 0.055).

TABLE 24 - Prevalence of diarrhoea, oral thrush and skin rash after baseline

Excludes all participants with diarrhoea at trial entry.

Excludes all participants with thrush at trial entry.

Excludes all participants with rash at trial entry.

Prevalence of diarrhoea, oral thrush and skin rash	Treatment arm, n (%)				Total (N = 814), n (%)
	Lower dose (N = 410)	Higher dose (N = 404)	Shorter duration (N = 413)	Longer duration (N = 401)
First diarrhoea after baselinea
No	276 (78.0)	252 (76.4)	259 (75.1)	269 (79.4)	528 (77.2)
Yes	78 (22.0)	78 (23.6)	86 (24.9)	70 (20.6)	156 (22.8)
p-value	p = 0.62		p = 0.18
New diarrhoea after baseline or worse than at baseline
No	303 (75.6)	288 (73.8)	296 (73.3)	295 (76.2)	591 (74.7)
Yes	98 (24.4)	102 (26.2)	108 (26.7)	92 (23.8)	200 (25.3)
p-value	p = 0.58		p = 0.34
Ever diarrhoea after baseline
No	234 (58.2)	213 (54.6)	217 (53.7)	230 (59.3)	447 (56.4)
Yes	168 (41.8)	177 (45.4)	187 (46.3)	158 (40.7)	345 (43.6)
p-value	p = 0.31		p = 0.11
First thrush after baselineb
No	386 (96.3)	381 (96.0)	390 (96.8)	377 (95.4)	767 (96.1)
Yes	15 (3.7)	16 (4.0)	13 (3.2)	18 (4.6)	31 (3.9)
p-value	p = 0.83		p = 0.33
New thrush after baseline or worse than at baseline
No	385 (96.0)	374 (95.9)	390 (96.5)	369 (95.3)	759 (96.0)
Yes	16 (4.0)	16 (4.1)	14 (3.5)	18 (4.7)	32 (4.0)
p-value	p = 0.94		p = 0.40
Ever thrush after baseline
No	374 (93.3)	360 (92.3)	379 (93.8)	355 (91.7)	734 (92.8)
Yes	27 (6.7)	30 (7.7)	25 (6.2)	32 (8.3)	57 (7.2)
p-value	p = 0.60		p = 0.26
First rash after baselinec
No	310 (86.6)	317 (86.8)	329 (88.4)	298 (84.9)	627 (86.7)
Yes	48 (13.4)	48 (13.2)	43 (11.6)	53 (15.1)	96 (13.3)
p-value	p = 0.92		p = 0.16
New rash after baseline or worse than at baseline
No	348 (86.8)	331 (84.9)	354 (87.6)	325 (84.0)	679 (85.8)
Yes	53 (13.2)	59 (15.1)	50 (12.4)	62 (16.0)	112 (14.2)
p-value	p = 0.44		p = 0.14
Ever rash after baseline
No	307 (76.6)	291 (74.6)	317 (78.5)	281 (72.6)	598 (75.6)
Yes	94 (23.4)	99 (25.4)	87 (21.5)	106 (27.4)	193 (24.4)
p-value	p = 0.52		p = 0.055

In addition, when considering skin rash severity during the treatment period only, there was evidence of a difference between the shorter- and longer-duration arms. Participants in the longer-duration arm experienced greater skin rash severity than participants in the shorter-duration arm at days 3 (p = 0.019) and 7 (p = 0.005) (Figure 20). There was no evidence of a difference between dose randomisation arms in terms of skin rash severity during the treatment period, and there was no evidence of a difference between the dose and duration randomisation arms in severity of diarrhoea or oral thrush during the treatment period (see Table 24).

FIGURE 20.

Skin rash severity during treatment period: duration randomisation.

Health-care services

Utilisation of health-care services was unrelated to randomisation arm. Hospital admissions and visits to the ED without admission were reported in 46 (5.7%) and 43 (5.3%) participants, respectively, whereas a larger proportion of participants reported using any health-care service (n = 304, 37.3%) (Table 25).

Daily activities and child care

Data on daily activities and child care were available from parent/guardian-completed diaries for 441 participants (Table 26). No differences in reported disruption to daily activities and child care were found between randomisation arms. In total, 73.9% of participants reported that the child had missed school, day care or nursery, and the median number of days missed was 4 (IQR 0–6) days. In addition, 63.8% of parents reported missing work, with a median of 3 (IQR 0–5) days missed, and 34.9% of parents reported requiring additional care for the child.

Limitations

In contrast to the majority of trials addressing optimal antibiotic treatment duration and dose of a single drug for childhood pneumonia, CAP-IT was conducted in a high-income setting where the expected mortality, even from moderate to severe CAP, is low. We selected antibiotic retreatment for respiratory tract infection during a follow-up period of 28 days as a clinically relevant and ascertainable event with limited risk of bias in a placebo-controlled trial. 84

To further guard against bias, an independent ERC, comprising experienced clinicians, adjudicated all antibiotic retreatments during the trial period, regarding the reason (i.e. respiratory tract infection or other) and clinical indication. Of note, the primary end point could be ascertained in 97% of CAP-IT participants either at final follow-up or through contact with the GP. Therefore, we consider the impact of loss to follow-up negligible.

We aimed to exclude children in whom antibiotics would not be expected to have any beneficial effect, primarily those likely to have obstructive airway disease only. However, a mixed picture was common for hospitalised children, with 16% of children receiving either salbutamol or steroids during their hospital stay. Mostly, this affected children with pre-existing hyper-reactive airway disease, and treatment was discontinued in a majority of cases by the time children were discharged home. Compared with the 48% bronchodilator use observed in the most recent UK paediatric pneumonia audit85 the use of salbutamol or steroids was low in CAP-IT, indicating that there was a strong clinical suspicion of CAP likely to benefit from antibiotics in enrolled children.

We observed no relevant impact of either amoxicillin duration or dose on pneumococcal penicillin non-susceptibility at 28 days, but did not assess pneumococcal resistance at other time points. We did not obtain end-of-treatment samples on all children for resistance analysis for several reasons. First, an additional face-to-face visit would have been a major barrier to participation for many families. Second, penicillin colonisation rates at, or shortly after, the end of antibiotic treatment are expected to be very low, whereas significant recolonisation or regrowth was expected (and observed) by 28 days. Finally, we considered penicillin-resistant pneumococcal colonisation at final follow-up to be the most relevant population- and individual-level resistance marker, as children colonised at this time point could transmit resistant pneumococci to others and would be at higher risk of potentially more difficult to treat respiratory tract infections in the future. 86

Generalisability

Children were enrolled in the trial based on clinically diagnosed pneumonia requiring antibiotic treatment with amoxicillin, and are typical of children treated for pneumonia with amoxicillin in PEDs. We included children discharged from hospital within 48 hours of admission for observation or initial clinical management, as hospital stays for acute respiratory tract infections, including pneumonia, are mostly of very short duration. 87,88 Data from the pilot phase confirmed that these children could be considered part of the same spectrum of disease as those discharged directly home from the ED. Only 13% of screened children were not approached because of physician preference for an antibiotic other than amoxicillin at discharge. This is in keeping with guidance suggesting that amoxicillin is used as the first-line antibiotic for oral treatment of uncomplicated childhood pneumonia in the community.

We excluded children with complicated pneumonia requiring prolonged hospitalisation, and those receiving non-beta-lactam treatment. Our findings, therefore, cannot be directly generalised to more severely ill children or those treated for atypical pneumonia. However, it is highly likely that our observations are relevant to children with mild to moderate pneumonia seen in primary care, who would be treated with oral amoxicillin at home. In primary care, the acuity of disease is generally lower and a lower rate of pneumonia likely to benefit from antibiotic treatment is expected.

No nasopharyngeal penicillin-resistant pneumococcal isolates were observed in the trial, either at baseline or at final follow-up, which is consistent with reported low penicillin resistance levels in northern Europe. 89 Therefore, our findings for the effectiveness of lower compared with higher amoxicillin dose, and impact on resistance, may be of limited generalisability to children with pneumonia in other high-income settings with higher pneumococcal penicillin resistance prevalence.

Twice-daily dosing of amoxicillin in line with WHO and other international recommendations was used in CAP-IT, rather than administration in three daily doses, as recommended by the BNFc. This was selected as it is known to maximise adherence, which would be particularly important in children allocated to the lower-dose and shorter-duration arms. In addition, patient representatives involved in the design phase indicated this approach to be particularly family friendly, as an additional mid-day dose is difficult to give to children who attend day care. Consequently, our findings, especially for antimicrobial resistance outcomes, may not be generalisable to children being treated with a thrice-daily amoxicillin regimen. However, participants in CAP-IT had rates of antibiotic retreatment and secondary or rehospitalisation similar to those described in observational studies conducted in settings with standard administration of amoxicillin in three doses. 41,87,90,91

Interpretation

To the best of our knowledge, few head-to-head comparisons of the same antibiotic in different dosing or duration regimens have been conducted in children being treated for pneumonia. Most of the existing literature reports on trials conducted in low- and middle-income settings prior to the widespread availability of PCVs and in an era with lower pneumococcal penicillin resistance. 92,93 Two recent relevant trials94,95 conducted in Malawi investigated 3-day compared with 5-day amoxicillin treatment and 3-day amoxicillin treatment compared with placebo in young children with non-severe pneumonia and not infected with human immunodeficiency virus. In summary, 3-day treatment at a dose corresponding to the higher total daily dose in CAP-IT was found to be non-inferior to 5-day treatment for early treatment failure, but this was not the case for placebo compared with 3-day treatment. The same trial identified the number needed to treat for children with non-severe fast-breathing pneumonia to be 33. These trials used high-sensitivity, but low-specificity eligibility criteria appropriate for a high-mortality setting. Evidence specific to high-income settings is lacking and has led some guideline-setting bodies to question the generalisability of findings from large trials in low- or middle-income countries to high-income settings. The persisting evidence gap for children identified as having pneumonia applying higher specificity clinical criteria in high-income settings has now been addressed by CAP-IT.

A relatively high retreatment rate of 12.5% was observed in the CAP-IT cohort. This is consistent with similarly high retreatment rates in primary care reported in large observational studies, but has not previously been described for children with CAP seen in EDs or discharged from hospital after a short stay. Similarly, the secondary or rehospitalisation rate of around 5% was similar to that described for children with pneumonia in observational studies.

We observed remarkably similar retreatment rates for respiratory tract infections between the 3- and 7-day treatment durations, despite 2-day slower resolution of mild cough, on average, in the shorter-duration arm. We did not identify any differences between the lower- and higher-dose treatment arms. Antibiotic retreatment for respiratory tract infection during the follow-up period could be related to true failure of the initial treatment or could be linked to persistent symptoms unlikely to be responsive to amoxicillin because they were mainly triggered by a viral (co-)infection or new respiratory tract infection episodes.

Children and parents in the 3-day randomisation arm were not reported to have spent a longer time away from day care or school and work, making it unlikely that cough had a major impact on children’s usual routines. Slightly longer time to symptom resolution in placebo arms or placebo-controlled shorter-duration arms has been reported for acute otitis media. 96 However, it is unclear how children being mildly symptomatic for longer is weighed against the benefits of shorter treatment by children and their families. When symptoms are minor, shorter treatment is likely to be a key factor in allowing children to return to usual activities and will maximise adherence. 97,98

Antimicrobial resistance was a key secondary outcome in CAP-IT. Colonisation by penicillin-non-susceptible pneumococci at 28 days was similar for both randomisation arms. In general, the observed prevalence of pneumococcal penicillin non-susceptibility and the complete absence of penicillin-resistant pneumococci was in line with the UK being a low-resistance setting. Pneumococcal penicillin resistance alone is unlikely to reflect the full impact of amoxicillin dose and duration on the child nasopharyngeal microflora, including the presence of resistance genotypes. Next-generation sequencing approaches could provide in-depth information about differential changes in the microbiome and resistome with higher or lower amoxicillin dose and shorter or longer treatment duration. However, the interpretation of such analyses is likely to be complex, and will need to take account of the interactions between different pneumococcal subpopulations, as well as between pneumococci and other bacteria in a densely populated niche. An analysis of nasopharyngeal samples obtained in CAP-IT using next-generation sequencing approaches is ongoing.

Several other trials have generated results that complement CAP-IT findings, or are expected to in the near future. In the UK, this includes the primary care-based ARTIC PC (Antibiotics for lower Respiratory Tract Infection in Children presenting in Primary Care) study,99 a randomised placebo-controlled trial investigating the benefit of a 7-day course of oral amoxicillin in children with possible lower respiratory tract infection (but not considered to have pneumonia clinically). The SAFER (Short-Course Antimicrobial Therapy for Pediatric Community-Acquired Pneumonia) trial100 in Canada and SCOUT-CAP (Short Course Outpatient Therapy of Community Acquired Pneumonia) study in the USA both target children presenting to EDs but not admitted to hospital, the former comparing 5- and 10-day treatment courses with amoxicillin and the latter a selection of beta-lactams. The SCOUT-CAP study is expected to report on results at the end of 2021. Finally, a Canadian open-label RCT (NCT03031210) is investigating twice-compared with thrice-daily amoxicillin dosing in children treated for pneumonia. The total daily dose in this trial corresponds to the higher total daily dose investigated in CAP-IT.

Implications

For clinical practice, CAP-IT supports routine use of shorter, 3-day, oral amoxicillin courses at current doses for children presenting to hospital with uncomplicated clinically diagnosed CAP for community-based treatment after discharge from acute care. A slightly longer time to resolution of mild cough can be expected in children treated for 3 days, compared with children treated for 7 days.

For research, existing systematic reviews and meta-analyses should be updated to include CAP-IT and other high-income setting trials. A series of relevant trials includes studies already completed or about to complete. Their inclusion, for example in existing Cochrane reviews, would ensure that key reference systematic reviews are relevant globally.

The question of the comparison between two and three times daily dosing of amoxicillin needs to be addressed. However, this may best be tackled by modelling and simulation based on high-quality pharmacokinetic data analysed using modern pharmacometric approaches. Such data are needed from a variety of settings, including low/high prevalence of pneumococcal penicillin resistance, varying pneumococcal vaccine coverage and low-, middle- and high-income settings characterised by varying prevalence of important covariates, such as malnutrition and obesity. Data from adults suggest that gut amoxicillin absorption may be saturable, limiting the expected utility of high-dose regimens. 101

A proportion of children screened for CAP-IT were identified to be ineligible because the managing clinician was planning treatment with an antibiotic other than amoxicillin. Trial data supporting the use of macrolides (targeting atypical pathogens) or alternative beta-lactams, such as amoxicillin/clavulanate (co-amoxiclav, targeting Gram-negative respiratory pathogens producing beta-lactamases), are lacking.

Trial Steering Committee

Elizabeth Molyneux (chairperson), Chris Butler, Alan Smyth and Catherine Pritchard (PPI).

Independent Data Monitoring Committee

Tim Peto (chairperson), Simon Cousens and Stuart Logan.

Independent End-Point Review Committee

Alasdair Bamford (chairperson), Anna Turkova, Anna Goodman and Felicity Fitzgerald.

Trial Management Group

Mike Sharland (chairperson), Julia Bielicki, Mark Lyttle, Colin Powell, Paul Little, Saul Faust, Adam Finn, Julie Robotham, Mandy Wan, David Dunn, Wolfgang Stöhr, Lynda Harper, Sam Barratt, Nigel Klein, Louise Rogers, Elia Vitale and Diana Gibb.

Investigator sites

Alder Hey Children’s Hospital NHS Foundation Trust, Liverpool: Dan Hawcutt (principal investigator), Matt Rotheram, Liz D Lee and Rachel Greenwood-Bibby.

Birmingham Children’s Hospital: Stuart Hartshorn (principal investigator), Deepthi Jyothish, Louise Rogers and Juliet Hopkins.

Bristol Royal Hospital for Children: Mark Lyttle (principal investigator), Stefania Vergnano, Lucie Aplin, Pauline Jackson, Gurnie Kaur, Beth Pittaway, Michelle Ross, Sarah Sheedy, Alice Smith and Yesenia Tanner.

Chelsea and Westminster Hospital NHS Foundation Trust, London: James Ross (principal investigator), Poonam Patel, Nabila Burney, Hannah Fletcher, Jaime Carungcong and Kribashnie Nundlall.

City Hospitals Sunderland NHS Foundation Trust: Niall Mullen (principal investigator), Rhona McCrone and Paul P Corrigan.

Countess of Chester Hospital: Susie Holt (principal investigator), John Gibbs, Caroline Burchett, Caroline Lonsdale and Sarah De-Beger.

Darlington Memorial Hospital: Godfrey Nyamugunuduru (principal investigator), John Furness (former principal investigator) and Dawn Eggington.

Derbyshire Children’s Hospital, Derby: Gisela Robinson (principal investigator), Lizzie Starkey and Mel Hayman.

Evelina London Children’s Hospital: Anastasia Alcock (principal investigator), Dani Hall (former principal investigator), Ronny Cheung, Alyce B Sheedy and Mohammad Ahmad.

James Cook University Hospital, Middlesbrough: Arshid Murad (principal investigator), Katherine Jarman and Joanna Green.

John Radcliffe Hospital, Oxford: Tanya Baron (principal investigator), Chris Bird (former principal investigator I), Shelley Segal and Sally Beer.

King’s College Hospital, London: Fleur Cantle (principal investigator), Hannah Eastman, Raine Astin-Chamberlain, Paul Riozzi and Hannah Cotton.

Leeds Children’s Hospital: Sean O’Riordan (principal investigator), Alice Downes, Marjorie Allen and Louise Conner.

Leicester Royal Infirmary: Damian Roland (principal investigator), Srini Bandi and Rekha Patel.

Nottingham University Hospitals NHS Trust: Chris Gough (principal investigator), Megan McAulay and Louise Conner.

Southport and Ormskirk Hospital NHS Trust: Sharryn Gardner (principal investigator), Zena Haslam and Moira Morrison.

Our Lady’s Children’s Hospital, Crumlin: Michael Barrett (principal investigator) and Madeleine Niermeyer.

Royal Alexandra Children’s Hospital, Brighton: Emily Walton (principal investigator), Akshat Kapur and Vivien Richmond.

Royal Hospital for Children, Glasgow: Steven Foster (principal investigator), Ruth Bland, Ashleigh Neil, Barry Milligan and Helen Bannister.

Royal London Hospital: Ben Bloom (principal investigator), Ami Parikh, Olivia Boulton and Imogen Skene.

Royal Manchester Children’s Hospital: Katherine Potier (principal investigator), Fiona Poole and Jill L Wilson.

Sheffield Children’s NHS Foundation Trust: Judith Gilchrist (principal investigator), Noreen West, Jayne Evans and Julie Morecombe.

St George’s Hospital, London: Paul Heath (principal investigator), Yasser Iqbal, Malte Kohns, Elena Stefanova and Elia Vitale.

St Mary’s Hospital, London: Ian Maconochie (principal investigator), Suzanne Laing and Rikke Jorgensen.

University Hospital Lewisham: Maggie Nyirenda (principal investigator), Anastasia Alcock, Samia Pilgrim and Emma Gardiner.

University Hospital of Wales, Cardiff: Jeff Morgan (principal investigator), Colin VE Powell (former principal investigator), Jennifer Muller and Gail Marshall.

Southampton Children’s Hospital: Katrina Cathie (principal investigator), Jane Bayreuther, Ruth Ensom and Emily Cornish.

Contributions of authors

Sam Barratt (https://orcid.org/0000-0003-3265-9486) (Clinical Trials Manager) contributed to compiling data and drafting the report.

Julia A Bielicki (https://orcid.org/0000-0002-3902-5489) (Consultant in Paediatric Infectious Diseases) drafted the report.

David Dunn (https://orcid.org/0000-0003-1836-4446) (Professor of Medical Statistics) performed statistical analyses.

Saul N Faust (https://orcid.org/0000-0003-3410-7642) (Professor of Paediatric Immunology and Infectious Diseases) contributed to drafting the report.

Adam Finn (https://orcid.org/0000-0003-1756-5668) (Professor of Paediatrics) performed resistance analyses.

Lynda Harper (https://orcid.org/0000-0003-3430-0127) (Clinical Project Manager) contributed to drafting the report.

Pauline Jackson (https://orcid.org/0000-0001-6461-953X) (Paediatric Nurse) contributed to drafting the report.

Mark D Lyttle (https://orcid.org/0000-0002-8634-7210) (Consultant in Paediatric Emergency Medicine) contributed to drafting the report.

Colin VE Powell (https://orcid.org/0000-0001-8181-875X) (Consultant in Paediatric Emergency Medicine) contributed to drafting the report.

Louise Rogers (https://orcid.org/0000-0002-5037-847X) (Clinical Research Sister) contributed to drafting the report.

Damian Roland (https://orcid.org/0000-0001-9334-5144) (Consultant in Paediatric Emergency Medicine) contributed to drafting the report.

Wolfgang Stöhr (https://orcid.org/0000-0002-6533-2888) (Senior Medical Statistician) performed statistical analyses.

Kate Sturgeon (https://orcid.org/0000-0003-2209-141X) (Senior Research Nurse) contributed to drafting the report.

Elia Vitale (https://orcid.org/0000-0001-5750-8436) (Senior Paediatric Clinical Research Nurse) contributed to drafting the report.

Mandy Wan (https://orcid.org/0000-0001-8802-0425) (Paediatric Clinical Trials Pharmacist) contributed to drafting the report.

Diana M Gibb (https://orcid.org/0000-0002-9738-5490) (Professor, Epidemiology) drafted the report.

Mike Sharland (https://orcid.org/0000-0001-8626-8291) (Professor of Paediatric Infectious Diseases) drafted the report.

Data-sharing statement

All data requests should be submitted to the corresponding author for consideration. Access to anonymised data may be granted following review.

Patient data

This work uses data provided by patients and collected by the NHS as part of their care and support. Using patient data is vital to improve health and care for everyone. There is huge potential to make better use of information from people’s patient records, to understand more about disease, develop new treatments, monitor safety, and plan NHS services. Patient data should be kept safe and secure, to protect everyone’s privacy, and it’s important that there are safeguards to make sure that it is stored and used responsibly. Everyone should be able to find out about how patient data are used. #datasaveslives You can find out more about the background to this citation here: https://understandingpatientdata.org.uk/data-citation.

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 and Social Care. 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 and Social Care.