Intravenous or oral antibiotic treatment in adults and children with cystic fibrosis and Pseudomonas aeruginosa infection: the TORPEDO-CF RCT

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 07/51/01. The contractual start date was in December 2009. The draft report began editorial review in August 2019 and was accepted for publication in December 2020. 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

© Queen’s Printer and Controller of HMSO 2021. This work was produced by Langton Hewer et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

2021 Queen’s Printer and Controller of HMSO

Scientific background

Cystic fibrosis (CF) is the most common life-limiting recessively inherited condition in white populations. It is a multisystem disorder in which the airways frequently become blocked with mucus, often associated with respiratory infections. These infections may lead to progressive respiratory failure and ultimately to death from breathing failure. Pseudomonas aeruginosa is a common infection in the lungs of patients with CF. The age-specific prevalence of P. aeruginosa in pre-school children is 9%, rising to 32% for 10- to 15-year-olds. 1 Early infection can be eradicated in the majority of patients. However, once chronic infection is established, P. aeruginosa is virtually impossible to eradicate and is associated with increased morbidity and mortality. 2 Long-term infection is associated with poor outcomes, including more rapid decline in lung function, such as the amount of air expired in 1 second of forced expiration [forced expiratory volume in 1 second (FEV₁)]. 3,4 New isolation of P. aeruginosa is treated with antibiotics in an attempt to eradicate the infection and to delay acquisition of chronic infection. 5 However, there is uncertainty about the best method to eradicate P. aeruginosa from the lower respiratory tract, and several different strategies are used, including oral quinolones such as ciprofloxacin, and intravenous (i.v.) and nebulised antibiotics. 6–10

There are clear differences between the available treatments in terms of the impact on the patient and their family, the use of resources and the cost of treatment. However, few studies have compared the efficacy and cost-effectiveness of these treatments.

Fourteen days of i.v. treatment will usually necessitate admission to hospital and siting of one or more i.v. lines for drug infusion. The siting of lines can be traumatic, especially for children and their families.

Intravenous aminoglycosides are commonly used and they require further blood tests for monitoring of plasma levels, and can be associated with kidney and inner ear damage. 11 No study has yet been conducted to investigate the therapeutic advantage of i.v. and oral treatment.

In 2005, the NHS National Institute for Health Research (NIHR) commissioned the Medicines for Children Research Network to assess the feasibility of conducting a randomised controlled trial to investigate the prevention of colonisation with P. aeruginosa in CF patients. This report was completed in 2007, having surveyed UK clinical practice, surveyed opinions of CF patients and families, and assessed the number of potentially eligible patients for such a trial. 12 The result of this feasibility study was to show that, generally, clinicians treat first or new growth of P. aeruginosa in accordance with the UK CF Trust guidelines13 and 95% of clinicians reported that they would consider i.v. treatment of first isolation of P. aeruginosa. In addition, 71% of clinician and 43% of consumer respondents would consider entry for themselves/their patients into a randomised controlled trial comparing oral with i.v. antibiotics. The conclusion of the study stated ‘[T]he clinical community are in equipoise when considering effectiveness of eradication therapy for treatment of P. aeruginosa in patients with cystic fibrosis’ and recommended that it is feasible to consider the initiation of a randomised controlled trial investigating eradication therapy to treat P. aeruginosa in patients with CF.

This study has been conceived and designed in response to addressing this clinical equipoise and has been commissioned by the NIHR.

Rationale for research

There is equipoise about the best method to eradicate P. aeruginosa from the lower respiratory tract. Several strategies exist to treat early infection with P. aeruginosa. This includes the use of inhaled antibiotics, such as colistin and tobramycin,9,14,15 oral quinolones, such as ciprofloxacin,6,10 and i.v. antibiotics, usually consisting of a combination of an aminoglycoside with a beta-lactam.

Antibiotic strategies for eradication of P. aeruginosa in people with CF have been investigated in a systematic review of randomised clinical trials,16 which concluded that there is an urgent need for well-designed and well-executed trials. The review made the specific recommendation that any future trial should investigate the hypothesis to see if antibiotic treatment of early P. aeruginosa infection prevents or delays chronic infection, and whether or not this then results in an appreciable clinical benefit to patients, without causing them harm. The systematic review made the recommendation that the following outcomes be considered for any future randomised controlled trial: spirometric lung function; nutritional status;11 and socioeconomic outcomes, including quality of life.

The UK CF Trust has published guidance for antibiotic treatment for CF, including treatment for eradication of newly acquired P. aeruginosa infection. 11 This guidance recommends energetic treatment for a patient who has isolated P. aeruginosa where cultures have previously been negative and the report commented that there is no evidence favouring any particular regimen for eradication. The guidance states that appropriate treatment in this situation will include oral ciprofloxacin for the duration of treatment up to 3 months or i.v. treatment such as a beta-lactam antibiotic (e.g. ceftazidime or meropenem) or an anti-pseudomonal penicillin in combination with i.v. tobramycin. 11 Intravenous antibiotics are usually administered for 10–14 days in patients with CF, although there have been no randomised controlled trials of shorter treatment durations.

The rationale for choosing 14 days of i.v. treatment and for choosing 3 months for oral treatment is that both of these regimens are standard practice in many UK CF centres identified in the feasibility study,12 both are standard recommendations within the published UK guideline11 and they are believed to represent current best practice.

Interventions

Participants recruited into the study were randomised to one of the following treatment groups.

Objective

This study aimed to establish the superiority of 14 days of i.v. therapy compared with 3 months of oral therapy.

Study design

This study was a Phase IV, multicentre, parallel-group, randomised controlled trial comparing 14 days of i.v. antibiotic therapy with 3 months of oral antibiotic therapy for participants with CF (Figure 1).

FIGURE 1.

The TORPEDO-CF study design. a, Sites that are unable to comply with the trial dosing regimen can use their current dosing regimen as long as the total daily dose administered is within national clinical guidelines; b, 3 months is defined as 12 weeks; and c, sample stored for genotyping.

Trial registration and ethics

The trial was registered on the European Union Drug Regulating Authorities Clinical Trials (EudraCT) database on 5 May 2009 (as EudraCT number 2009-012575-10) and received clinical trials authorisation from the Medicines and Healthcare products Regulatory Agency on 2 November 2009 (clinical trials authorisation reference 12893/0220/001). The trial and all subsequent protocol amendments were reviewed and authorised by the Medicines and Healthcare products Regulatory Agency.

The trial protocol was not initiated until it had received the favourable opinion of the National Research Ethics Committee (REC) (London REC reference 09/H0718/51) on 16 November 2009. It was then reviewed at the research and development offices at participating sites. All subsequent amendments were reviewed and approved by the National Research Ethics Committee, London REC.

The trial was listed on the International Standard Randomised Controlled Trial Number (ISRCTN) registry on 21 May 2009 as ISRCTN02734162.

Participant inclusion and exclusion criteria

Recruitment

The trial took place in 70 UK CF centres and two CF centres in Italy.

Recruitment commenced on 5 October 2010 and the final participant was randomised on 27 January 2017.

Informed consent

The trial recruited adults and minors (defined in statutory instrument 200418 No. 1031 as aged < 16 years). Informed consent procedures reflected the legal and ethical requirements for obtaining valid written informed consent in these populations.

Prior written informed consent was required for all trial participants. In obtaining and documenting informed consent, the investigator was required to comply with the applicable regulatory requirements, and adhered to the principles of good clinical practice (GCP) and to the ethical principles that have their origin in the Declaration of Helsinki. 19

Potential participants and their families were provided information regarding the trial both verbally and in writing using the ethics-approved patient information sheets and consent forms (PISCs). They were given the opportunity to discuss the trial with the site team. The PISCs took into account the age of minors and their assent was obtained, where appropriate. Potential participants and their families were provided with a clear overview of the trial and details of the procedures, and the potential risks and benefits of all trial medications were carefully discussed.

Adequate time to consider trial entry (generally 24 hours, although it was acknowledged that some patients/families came to a decision sooner) was allowed and all participants were given the opportunity to ask questions, had the opportunity to discuss the study with their surrogates and had time to consider the information prior to agreeing to participate. All PISCs used in TORPEDO-CF were made available in the native language of the countries participating in the trial (with the exception of Wales, where the majority language was used).

All of the recruiting investigators were experienced CF physicians familiar with imparting information to the relevant trial populations. All investigators requesting consent to participate had attended GCP training. During the screening process, if a potential patient was identified, then they/their parent or the person with parental responsibility were approached by the investigator or a designated member of the investigating team. The trial and its objectives were then described to them, at which point any questions could also be answered. The treatment schedule and trial visits were in line with standard clinical care. The potential risks and benefits of the trial interventions were discussed, as well as what would happen if they chose not to enter the trial or had to withdraw from the trial for any reason.

The right of the patient (non-minors) or parent/legal guardian (for minors) to refuse consent to participate in the trial without giving reasons was respected. After the patient had entered the trial, the clinician remained free to give alternative treatment to that specified in the protocol, at any stage, if they felt that it was in the best interest of the participant. However, the reason for doing so was recorded and the patient remained within the trial for the purpose of follow-up and data analysis according to the treatment option to which they had been allocated. Similarly, the patient remained free to withdraw from the protocol treatment and trial follow-up at any time without giving reasons and without prejudicing their further treatment.

Randomisation

Participants were randomised using a secure (24-hour) web-based randomisation programme, which was controlled centrally by the Liverpool Clinical Trials Centre (LCTC), to ensure allocation concealment. Randomisation lists were generated in a 1 : 1 ratio using simple block randomisation, with random variable blocks of length of two and four after an initial block of length of three to reduce predictability. Randomisation was stratified by site, but this was not disclosed in the protocol to further reduce predictability.

Participant treatment allocation was displayed on a secure web page and an automated e-mail confirmation was sent to the authorised randomiser and the principal investigator (PI) or co-investigator (where applicable) at the randomising site. It was the responsibility of the PI or delegated research staff to inform the pharmacy department at their centre of the potential participant prior to randomisation to ensure that there was a sufficient supply of the study drugs.

In the event of an internet connection failure between the centre and the randomisation system, the centre contacted the LCTC to resolve the problem. In the event that site access could not be promptly reinstated, LCTC would attempt to centrally randomise the patient using the web system. Where this was not possible, LCTC would randomise the participant using a back-up randomisation envelope. Of the 286 participants who were randomised, four were randomised utilising a back-up randomisation envelope.

Interventions

Participants recruited into the study were randomised to one of the following treatment groups.

Data collection and management

Sample size

Patient and public involvement

TORPEDO-CF was conceived as a NIHR-commissioned call and thus had patient and public involvement (PPI) through this process. In addition, it benefited from a feasibility study that explored the importance, and acceptability, of the research question for children, young people and their parents, and adults with CF.

In its set-up, TORPEDO-CF profited from PPI involvement through the Young Persons Advisory Group of the Medicines for Children Research Network, which provided input on the design of documentation intended for the use of trial participants (e.g. information sheets and promotional materials).

During its conduct, TORPEDO-CF received support from the CF Trust and PPI representation within the Trial Management Group (TMG) was in the form of ad hoc representation through the CF Trust Special Adviser on Research and Patient Involvement, who advised on activities to enhance engagement with the CF community generally and with potentially eligible individuals in particular.

An important element of PPI involvement was provided by the independent Trial Steering Committee (TSC). PPI representation on the committee provided reassurance that TORPEDO-CF remained of relevance to and in the interests of the CF community.

Changes to the protocol

The first site was opened to recruitment on 18 June 2010 and version 2.0 of the protocol was in operation at this time.

Over the course of the trial, eight substantial amendments were made to the protocol. Each amendment was assessed by the TMG and approved by the REC and by the Medicines and Healthcare products Regulatory Agency. The main corrections to the protocol included changes to the management of the trial and the addition of further guidance for participating sites. The definition of the primary outcome was also amended from ‘[S]uccessful eradication of P. aeruginosa infection at three months post randomisation, remaining infection free through to 15 months post randomisation’ to ‘[S]uccessful eradication of P. aeruginosa infection three months after allocated treatment has started, remaining infection free through to 15 months after the start of allocated treatment’ to eliminate any systematic bias due to the inevitable delay in starting i.v. treatment compared with oral treatment (caused by the need for the patient to be admitted to hospital to receive i.v. treatment).

The initial changes to the protocol were to provide clarification of the exclusion criteria and the secondary outcomes. Following site set up, several sites raised an issue with regard to differences in site dosing regimens compared with that specified in the protocol. The TMG confirmed that the changes were acceptable clinically and from a trial perspective.

Please refer to the trial protocol for a full list of protocol amendments (see NIHR Journals Library; www.journalslibrary.nihr.ac.uk). A summary of amendments follows.

Compliance with intervention

Participants’ compliance with trial treatment was monitored using participant-completed treatment diaries to record their daily treatment routine.

The majority of the trial treatment for participants in the i.v. antibiotic therapy group was anticipated to be administered during hospital in-patient visits, ensuring accurate monitoring of investigational medicinal product (IMP) compliance. Participants who had experience of administering i.v. antibiotics at home or those who preferred to have home i.v. were allowed to self-administer i.v. antibiotics at the discretion of their local PI. In this event, participants were asked to complete a home i.v. treatment diary each day to ensure compliance with the study medication. Any unused medication and the packaging of all used medication were to be returned at the next scheduled follow-up visit. All returned medication/packaging was destroyed as per local procedures.

Trial management and oversight

University Hospitals Bristol NHS Foundation Trust was the sponsoring organisation, and delegated responsibilities to LCTC, the University of Liverpool and the chief investigator. LCTC was responsible for co-ordination of the trial (see Appendix 1).

University of Liverpool was co-sponsor, having no responsibility for the conduct of the trial in the UK, but acting as sole sponsor for non-UK centres (i.e. the University of Liverpool was legally responsible for the non-UK conduct of the trial).

General statistical considerations

The analysis and reporting of the study were undertaken in accordance with the Consolidated Standards of Reporting Trials (CONSORT)25 and the International Conference on Harmonisation E9 Guidelines. 26 The main features of the statistical analysis plan are included here with a full and detailed statistical analysis plan (available at www.uhbristol.nhs.uk/media/3933191/torpedo_final_analysis_statistical_analysis_plan_v2.pdf; accessed 8 October 2021). All statistical analyses were undertaken using SAS^® software version 9.3 or later (SAS Institute Inc., Cary, NC, USA).

A two-sided p-value ≤ 0.05 was used to declare statistical significance for all analyses, and all confidence intervals (CIs) were calculated at the 95% level. There were no adjustments for multiple testing; rather, all secondary analyses were treated as hypothesis generating.

The primary analysis followed the ITT principle as far as practically possible; all randomised participants were analysed on the basis of the treatment to which they were randomised, regardless of whether or not they received it. If consent for treatment was withdrawn but the participant was happy to remain in the study for follow-up, then they were followed up until completion. However, if they decided to withdraw consent completely, then the reasons for withdrawal of consent were collected (when possible) and reported for both groups.

Analysis of baseline data

Demographic and baseline characteristics were summarised for each treatment group using descriptive statistics. No formal statistical testing was performed on these data. Descriptive statistics, including the number of observations, mean and standard deviation (SD) for continuous variables, and counts and percentages for discrete variables, were presented, as appropriate.

Analysis of compliance data

For i.v. antibiotic therapy, the number of daily doses that participants received varied by site. The protocol stated that patients should receive at least 10 days of treatment. For these reasons, compliance is presented in terms of the number of days on treatment rather than in terms of the number of doses received.

For oral antibiotic therapy, compliance is presented in terms of the number of doses received and a compliance rate [the percentage of the total doses (n = 168: two doses per day for three lots of 28 days) that were received].

For nebulised therapy, compliance is presented in the same way as oral antibiotic therapy. Data are presented split by treatment group.

All compliance data are presented descriptively [total (n), mean, SD, median, interquartile (IQR) range, range, minimum and maximum]. No formal statistical testing was undertaken.

Analysis of primary outcome

The primary outcome, successful eradication of P. aeruginosa 3 months after the start of treatment, and remaining infection free through to 15 months after the start of treatment, was a binary outcome. Participants were classified as a success if there was no record of P. aeruginosa on their microbiology case report form between 3 and 15 months after treatment commencement, and a failure if there was at least one record of P. aeruginosa on their microbiology report during this time.

To determine if a participant had eradicated P. aeruginosa at 3 months, they must have had a sample within 28 days on either side of their expected 3-month visit date (treatment commencement date + 84 days), which was at least 2 days after the last date of any anti-pseudomonal treatment. To determine if a patient had remained P. aeruginosa free at 15 months, they must have had either a positive sample within the primary outcome window (3–15 months) or a sample within 28 days either side of their expected 15-month visit date (treatment commencement date + 420 days). Participants who had not regrown P. aeruginosa and did not have a sample within the 15-month window (defined above) were not included in the primary analysis. If a patient was withdrawn before 15 months of follow-up was completed, they were included in the primary outcome analysis only if they had isolated P. aeruginosa; if there was no evidence that they had isolated P. aeruginosa prior to being withdrawn they were not included in the primary analysis.

The number and percentage of participants who were classified as a success and a failure for the primary outcome were presented for each treatment group. The difference between the groups was tested using the chi-squared test, and the relative risk and associated 95% CI are presented. The number and percentage of patients who had a positive sample at their 3-month follow-up visit are also presented for each treatment group. As a post hoc analysis, the difference between the groups at 3 months was also tested using the chi-squared test, the relative risk and associated 95% CI were calculated, and a sensitivity analysis was also carried out that extended the 3-month window to – 4 weeks/+ 10 weeks.

Six sensitivity analyses were conducted to determine the robustness of the results of the primary analysis:

1. All participants followed up past 3 months, but with no 15-month sample were classified as successes.

2. All participants followed up past 3 months, but with no 15-month sample were classified as failures.

3. All patients followed up past 3 months, but with no 15-month sample were classified as successes/failures according to the next sample taken after the end of the 15-month window.

4. Primary analysis adjusted for centre effect – a logistic regression model was fitted to investigate heterogeneity across centres and adjusted for centre as a random effect.

5. Analysis as per the primary analysis, but 3- and 15-month windows extended to 10 weeks either side of the expected visit dates: post hoc analysis.

6. Analysis as per the primary analysis, but 15-month window removed (any sample after 3-month window included): post hoc analysis.

Analysis of secondary outcomes

Time to reoccurrence of the original P. aeruginosa infection was presented graphically using Kaplan–Meier curves stratified by treatment. The difference between the two treatment groups was tested using the log-rank test. Two analyses of this outcome were performed as a result of blind review. The first assumed that all patients who had had a reoccurrence of P. aeruginosa but did not have both a baseline and a follow-up sample (between 3 and 24 months) sent for genotyping did not have a reoccurrence of the original P. aeruginosa strain. The second assumed that these patients did have a reoccurence of the original P. aeruginosa strain.

Reinfection with a different genotype of P. aeruginosa and admission to hospital were analysed using the chi-squared test. The relative risk and 95% CI are also presented.

Lung function [percentage predicted FEV₁, percentage predicted FVC and percentage predicted FEF_25–75 (not measured in participants < 5 years)], oxygen saturation and growth and nutritional status [height z-scores (children only), weight z-scores (children only), BMI z-scores (children only) and BMI (adults only)] were analysed using a repeated-measures random effects model with a special power covariance structure. The baseline measurement, treatment group and time were included in the model as coefficients and a treatment-by-time interaction was also fitted. The treatment difference at 15 months and 95% CI are presented.

For the number of pulmonary exacerbations and the number of days spent in hospital, a Mann–Whitney U-test was used to test whether or not the distribution was the same in each treatment arm.

Quality of life (as measured by the CFQ – Revised) was analysed using a mixed-effects model for repeated measures with an unstructured covariance matrix. The baseline measurement, treatment group and time were included in the model as coefficients and a treatment-by-time interaction was also fitted. The treatment difference at 15 months and 95% CI are presented.

Whether or not participants had grown at least one positive culture of MRSA, Burkholderia cepacia complex, Aspergillus or Candida was analysed using the chi-squared test (where there were sufficient events), and the relative risk and 95% CI are presented. Other sputum microbiology was presented descriptively; no formal analysis was undertaken.

Carer and participant burden was analysed in two ways. Whether or not carers/participants had been absent from work/education was analysed using the chi-squared test and presented with a relative risk and 95% CI. The Mann–Whitney U-test was used to determine whether or not there was a difference in the distributions of the number of days carers/participants were absent from work/education in each treatment group.

Analysis of safety data

The safety analysis data set contained all participants who were randomised and had commenced treatment. All events were coded using the Medical Dictionary for Regulatory Activities (MedDRA).

The number of occurrences of each AE [at the preferred term and System Organ Class (SOC) levels] and the number (and percentage) of patients experiencing each AE are presented for each treatment arm and overall. A similar table was produced for all SAEs reported under all versions of the protocol.

Each SAE reported under versions 1.0–5.0 of the protocol and each SAR reported under version 6.0 of the protocol onwards is presented in the form of line listings detailing:

• SAE number

• treatment

• preferred term

• SOC

• date of onset

• serious criteria (reporting of all that apply from seven possible options) (PI and chief investigator assessment)

• severity (mild/moderate/severe) (PI assessment)

• relationship to study drug (PI and chief investigator assessment)

• expectedness (chief investigator assessment)

• most likely cause (disease under study/other illness/prior or concomitant treatment/protocol procedure/lack of efficacy) (PI assessment)

• outcome.

There was no formal statistical analysis of any safety data.

Post hoc analyses

Time to P. aeruginosa reoccurrence is presented graphically using Kaplan–Meier curves stratified by treatment. The difference between the two treatment groups was tested using the log-rank test. Two analyses of this outcome were performed; the first included all positive samples after 3 months and the second included samples post 3 months and prior to 15 months. The hazard ratio (HR) and 95% CI were also calculated for the second analysis.

Additional analyses were performed on the lung function, oxygen saturation, growth and nutritional status, and quality-of-life outcomes. Where the interaction term was not significant in the original model, it was removed and an overall treatment difference and 95% CI are presented. Treatment differences at 3 and 24 months and the associated 95% CIs were also presented from the original model.

Use of variable number tandem repeat in a reference laboratory setting

Public Health England (PHE) has used VNTR analysis at nine loci as a method for P. aeruginosa strain typing since 2010 and, to date (May 2019), PHE has a national database of 31,028 profiles comprising CF, non-CF and environmental isolates from hospitals across the UK, providing a good national picture. Good concordance of this methodology was found on comparison with the gold standard, pulsed-field gel electrophoresis (PFGE). 31 In addition, as it is a polymerase chain reaction-based method, results can be generated rapidly and are portable between laboratories, making it preferable to PFGE, which is more laborious and not easily comparable between laboratories. Of the nine loci chosen for this scheme, the ninth is particularly discriminatory and may be used to evaluate relatedness between isolates from hospital outbreaks and in other settings among non-CF patients,33–36 but this locus is not always useful for CF isolate discrimination, as variation in copy number may be found in different colonies from the same sputum sample.

Examining the UK Pseudomonas aeruginosa population structure using variable number tandem repeat analysis

A study conducted by PHE in 2013 used VNTR analysis to examine the population structure of P. aeruginosa among UK CF patients, non-CF patients and hospital environmental isolates over a 2-year period. 37 In total, 3870 isolates were examined, comprising 726 environmental isolates from 47 hospitals and 2325 isolates from 2230 patients (of whom approximately 54% were CF patients) from 143 centres. The study highlighted the existence of VNTR profiles representing common clones present in both patient and environmental specimens, submitted from hospitals across the UK. Fourteen common types, each found in more than 24 patients within this 2-year period, were described. Among these were three well-characterised strains associated with transmissibility, which have been isolated almost exclusively from CF patients, namely the Liverpool, Manchester and Midlands 1 strains. 38,39 These were isolated from 6%, 1.7% and 1.5%, respectively, of 1204 CF patient isolates in the above-named study. 37

Four of the most common profiles were examined in more detail. Two of these correlated with previously well-established international clonal lineages, namely ‘PA14’ (VNTR profile: 12, 2, 1, 5, 5, 2, 4, 5, x, where the ninth locus is variable) and ‘clone C’ (VNTR profile: 11, 6, 2, 2, 1, 3, 7/8, 2/3, x, where the seventh, eighth and ninth loci may vary in copy number),40,41 which correspond to multilocus sequence types ST253 and 17, respectively. The other two types were ‘cluster A’ (VNTR profile: 8, 3, 4, 5, 2, 3, 5, 2, x), corresponding to ST27, and ‘cluster D’ (VNTR profile: 10, 3, 5, 5, 4, 1, 3, 7, x), where ‘x’ is variable (ST395). These four types were isolated from 6%, 5%, 3% and 2% of patients, respectively, and were geographically widespread, although they were not always isolated in large numbers in individual hospitals. PFGE analysis of representatives of each of these VNTR types from multiple hospitals separated the isolates into four broad clusters corresponding to their VNTR type with an overall similarity of approximately 70%. This led to the hypothesis that in most cases patients acquired these strains independently rather than by patient-to-patient transmission. Further evidence for the clonal nature of these four common profiles was found by examining the sequences for the intrinsic bla_OXA-50-like gene.

To examine the hypothesis that these strains were often acquired independently, whole-genome sequences for 12 isolates of ‘cluster A’ from nine hospitals comprising six CF, five non-CF and one environmental isolate from geographically distant sites and separated in time were examined. The presence and absence of genes from the accessory genome detected 9454 variable sequences across the 12 isolates and the six reference genomes that were compared. A set of seven accessory genes were chosen from these, which were found to be variably present among a larger set of ‘cluster A’ isolates. Representatives from patients within a single centre mostly had distinct accessory gene profiles, suggesting that these patients acquired the strain independently, whereas those with clear epidemiological links shared the same profile. Profiles also varied between representatives from different centres.

Challenges in variable number tandem repeat analysis of cystic fibrosis isolates

Variable number tandem repeat analysis of CF isolates often poses additional challenges compared with the analysis of non-CF outbreaks, particularly, although not exclusively, among chronically infected patients. In addition to the variation at the last locus found between different colony types from the same sputum, variation of up to two repeats at one out of the first eight loci may also be found in some patient samples. There may also be variation between sequential isolates over time. Caution needs to be applied when inferring strain relatedness solely using VNTR analysis in certain cases. For example, it is unlikely to be sufficiently discriminatory when assessing whether isolates with similar profiles pre and post antibiotic therapy represent reinfection from an environmental/other source or therapeutic failure. In these cases, the greater discrimination provided through whole-genome sequencing of isolates may be required. Despite these drawbacks, this method has been successfully used by PHE to type P. aeruginosa isolates from hospitals throughout the UK since 2010, and has proved invaluable for highlighting outbreaks and aiding cross-infection decisions in CF clinics.

Participant recruitment

The first participant was randomised on 5 October 2010 and the final participant was randomised on 27 January 2017, when the recruitment target was met. The last follow-up visit occurred on 10 April 2018.

Sixty-one out of the 72 sites (see Appendix 3) that screened patients randomised at least one participant. Twenty-two sites randomised at least five participants.

The flow of participants through the trial is presented in the CONSORT flow diagram (Figure 2). A total of 1308 patients screened for eligibility (patients could be screened on multiple occasions); 1022 did not undergo randomisation because they did not meet the eligibility criteria and 286 patients were randomised (137 to i.v. antibiotic therapy and 149 to oral antibiotic therapy). Detailed information on recruitment is provided in Appendix 4, Tables 28–32.

FIGURE 2.

The CONSORT flow diagram for all trial participants. a, Multiple screenings allowed; b, Patient randomised by all data completely removed as PI sign off could not be obtained.

Analysis populations

The ITT analysis population included 285 of the 286 participants (99.7%) who were randomised; one participant’s data could not be included as PI approval could not be obtained despite every effort being made by the TMG to do so. The safety population included 272 of the 286 randomised participants (95.1%); 13 patients did not receive any of their allocated treatment and one participant’s data could not be included as PI approval could not be obtained. The analysis set for the primary outcome included 255 participants: 125 in the i.v. antibiotic therapy group and 130 in the oral antibiotic therapy group.

All participants who withdrew consent for trial continuation contributed outcome data up until the point of withdrawal.

Premature discontinuations

Premature discontinuation of treatment (either IMP or non-IMP) occurred in 40 of the 137 participants in the i.v. antibiotic therapy group (29.2%) and 26 of the 148 participants in the oral antibiotic therapy group (17.6%). The most common reasons for premature discontinuation of treatment in the i.v. antibiotic therapy group were withdrawal of consent (n = 9; 22.5%), venous access problem (n = 7; 17.5%) and long-line failure (n = 5; 12.5%), and in the oral antibiotic therapy group the reasons were AE (n = 13; 50%) and withdrawal of consent (n = 4; 15.4%). Reasons for premature discontinuation of treatment are summarised in Table 1.

Withdrawal from follow-up occurred in 5 of the 137 participants in the i.v. antibiotic therapy group (3.6%) and 16 of the 148 participants in the oral antibiotic therapy group (10.8%). The most common reason in both groups was that participants were lost to follow-up [three participants in the i.v. antibiotic therapy group (60%); nine participants in the oral antibiotic therapy group (56.3%)]. Reasons for withdrawals from follow-up are summarised in Table 2.

Baseline characteristics

The demographic baseline data of the 285 randomised participants were comparable between the two groups. The proportion of infants and toddlers was slightly larger in the i.v. antibiotic therapy group (30.7%, compared with 18.9% in the oral antibiotic therapy group). The proportions of male and female participants were approximately the same across the two treatment groups, with slightly more female than male participants (Table 3).

Body mass index, lung function and oxygen saturation were similar across the two treatment groups. The proportions of patients who were naive or free from P. aeruginosa for the preceding 12 months and who had experienced a pulmonary exacerbation were also comparable across the two groups. There were fewer patients in the i.v. antibiotic therapy that had a diagnosis based on a homozygous delta f508 mutation (60.8%, compared with 51.1% in the oral antibiotic therapy group). Similar numbers of participants in each treatment group had other microorganisms detected at baseline (see Table 3).

Protocol deviations

Protocol deviations were monitored centrally by evaluating inclusion/exclusion criteria at trial entry and throughout the trial. A total of 284 participants (99.6%) had at least one major protocol deviation. The most common protocol deviations related to visits occurring outside the protocol-specified visit windows; 282 participants (98.9%) had a visit outside the window at 3 or 15 months (major deviation) and 280 participants (98.2%) had a visit outside the window at 6, 9, 12, 18, 21 or 24 months (minor deviation). All protocol deviations were agreed with the co-chief investigators prior to them seeing any unblinded results (Table 4).

Compliance

Primary outcome

The primary outcome (successful eradication of P. aeruginosa infection 3 months after allocated treatment has started and remaining infection free through to 15 months after the start of allocated treatment) was achieved by 55 of the 125 participants (44.0%) in the i.v. antibiotic therapy group and by 68 of the 130 participants (52.3%) in the oral antibiotic therapy group. Participants who were randomised to the i.v. antibiotic therapy group had a reduced chance of having successful eradication of P. aeruginosa 3 months after the start of treatment and remaining infection free through to 15 months (relative risk 0.84, 95% CI 0.65 to 1.09; p = 0.184). Results from five of the six sensitivity analyses to address missing data can be found in Table 9, which details the number of patients included in each analysis, the number in each treatment group in whom eradication was or was not observed, the relative risk, 95% CI and the p-value from the chi-squared test. The results from all five sensitivity analyses are consistent with the results from the primary analysis, indicating that the original results are robust with regard to the assumptions that were made. The additional sensitivity analysis not reported in Table 9 applied all of the same assumptions as the primary analysis, but used a logistic regression model to adjust for centre as a random effect. The adjustment to the model for centre did not significantly affect the results (p = 0.218), indicating that there was no statistically significant effect of centre on the primary results. Results from a post hoc subgroup analyses in P. aeruginosa naive and P. aeruginosa free patients are also reported in Table 9. A Mantel–Haenszel test for interaction was conducted and the relative risks were not significantly different in the two subgroups (p = 0.19).

The number of participants who did not eradicate P. aeruginosa 3 months after the start of treatment was 13 (11.8%) in the i.v. antibiotic therapy group and five (4.3%) in the oral antibiotic therapy group. Participants who were randomised to the i.v. antibiotic therapy had an increased risk of unsuccessful eradication of P. aeruginosa 3 months after the start of treatment (relative risk 2.74, 95% CI 1.01 to 7.44; p = 0.037) (post hoc analysis). The results of this analysis were supported by a sensitivity analysis in which the 3-month window was widened to – 4 weeks/+ 10 weeks of the expected 3-month visit date (relative risk 2.00, 95% CI 1.03 to 3.86; p = 0.035).

A post hoc analysis was performed on time to P. aeruginosa reoccurrence. Figure 3 is the Kaplan–Meier plot for time to P. aeruginosa reoccurrence up until the end of the 15-month window. This analysis included the 255 participants included in the primary analysis. Oral antibiotic therapy delayed time to P. aeruginosa reoccurrence compared with i.v. antibiotic therapy; however, the effect was not statistically significant (HR 1.31, 95% CI 0.93 to 1.85; p = 0.119). Figure 4 is the Kaplan–Meier plot for time to P. aeruginosa reoccurrence up until the end of the 24-month follow-up period.

FIGURE 3.

Time to reoccurrence of P. aeruginosa infection (any strain) up to the end of the 15-month window.

FIGURE 4.

Time to reoccurrence of P. aeruginosa infection (any strain) up to the end of the 24-month follow-up period.

Following database lock and the unblinding of the treatment allocations, it was found that one participant (who had withdrawn from the trial and was not included in the primary analysis) had a positive sample for P. aeruginosa on the day that they had stopped treatment. The definition in the statistical analysis plan stated that any samples taken in the 48-hour window following treatment cessation should be excluded and this was executed for the primary analysis shown above; however, this should have stated that only samples that were negative for isolation of P. aeruginosa in this period should be excluded – positive samples should have been included. Further investigation identified no other occurrences of this; in a post hoc analysis that included this participant, there was no impact on the primary analysis.

Secondary outcomes

Number of pulmonary exacerbations

A total of 283 participants were included in the analysis of the number of pulmonary exacerbations within the first 15 months of follow-up. In both treatment groups, most participants did not experience any pulmonary exacerbations (i.v. 72.3%; oral 64.4%) during this time, but the risk of experiencing at least one pulmonary exacerbation was lower in the i.v. antibiotic therapy group than in the oral antibiotic therapy group (relative risk 0.78, 95% CI 0.55 to 1.10; p = 0.155). The median (IQR) number of exacerbations experienced in each group was 0 (0–1). The distributions of the number of exacerbations experienced in each treatment group were not significantly different (p = 0.090). Table 12 provides the results of the additional analyses of this outcome over different time periods of interest. The results of these analyses were all consistent with the results over the first 15 months of follow-up. The analysis investigating the time to first pulmonary exacerbation found no significant difference between the two treatment groups (HR 0.73, 95% CI 0.50 to 1.06; p = 0.1); the Kaplan–Meier plot can be seen in Figure 5.

TABLE 12 - Secondary outcomes: number of pulmonary exacerbations and admissions to hospital

Data are n/N (%) for participants experiencing at least one occurrence.

Outcome	Time period	Treatment group, n/N (%)a		Relative risk (95% CI)	p-value
		i.v. antibiotic therapy	Oral antibiotic therapy
Number of pulmonary exacerbations	Up to 15 months	38/137 (27.7)	52/146 (35.6)	0.78 (0.55 to 1.10)	0.155
	Up to 3 months	8/135 (5.9)	15/144 (10.4)	0.57 (0.25 to 1.3)	0.173
	3–15 months	31/131 (23.7)	46/136 (33.8)	0.7 (0.48 to 1.03)	0.067
	15–24 months	14/99 (14.1)	26/103 (25.2)	0.56 (0.31 to 1.01)	0.048
Admission to hospital	Up to 3 months	25/135 (18.5)	15/143 (10.5)	1.77 (0.97 to 3.2)	0.057
	3–15 months	40/129 (31.0)	61/136 (44.9)	0.69 (0.5 to 0.95)	0.02
	15–24 months	33/105 (31.4)	41/107 (38.3)	0.82 (0.57 to 1.19)	0.293
Admission to hospital: sensitivity analysis	Up to 3 months	24/135 (17.8)	9/143 (6.3)	2.82 (1.36 to 5.86)	0.003
	3–15 months	39/129 (30.2)	63/136 (46.3)	0.65 (0.47 to 0.9)	0.007
	15–24 months	34/104 (32.7)	41/108 (38.0)	0.86 (0.6 to 1.24)	0.422

FIGURE 5.

Time to first pulmonary exacerbation.

Safety

Overview

A prospective economic evaluation was conducted alongside the randomised controlled trial to assess the cost-effectiveness of oral antibiotic therapy compared with i.v. antibiotic therapy. The primary analysis used an NHS and Personal Social Services (PSS) perspective for the collection and incorporation of resource use, as recommended by the National Institute for Health and Care Excellence (NICE). 42 The time horizon for the primary analysis was 15 months from randomisation.

The primary outcome within the economic evaluation was the same as that in the clinical study and measured the percentage of patients with successful eradication of P. aeruginosa 3 months after the start of treatment and who remained infection free through 15 months after the start of treatment. Analysis for the economic evaluation was based on the population with data for the primary outcome; missing health economic data within that population were multiply imputed.

The primary cost-effectiveness analysis measured the incremental cost per successful eradication of P. aeruginosa infection 3 months after allocated treatment had started, and of remaining infection free through to 15 months after the start of allocated treatment, in the oral antibiotic therapy group and the i.v. antibiotic therapy group. Regression analysis controlled for baseline differences between groups.

Secondary cost-effectiveness analyses were carried out and used the EQ-5D-3L (completed by a mixture of patients/carers, as appropriate) and QALYs to value health outcomes.

Where the EQ-5D-3L and QALYs were used to value health outcomes, cost–utility analysis was conducted to measure the incremental net benefit (INB) of treating patients in the oral antibiotic therapy group and the i.v. antibiotic therapy group. Sensitivity analyses explored factors that were identified a priori as likely to be key drivers of cost-effectiveness, including assumptions regarding the incorporation of costs for the intervention.

Resource use and costs

Total resource use and costs for each patient in the clinical trial were calculated. The main resource use and cost components for both the oral antibiotic therapy and i.v. antibiotic therapy arm comprised (1) the different modes of antibiotic intervention, information on which was collected through the use of case report forms and (2) any follow-up interactions with the health and social care system, which information was captured through patient diaries and collected at 3, 6, 9, 12 and 15 months. All costs were calculated in Great British pounds with the price year of 2016/17 used. Where possible, unit costs were sourced from national databases including the BNF43 for medicines; NHS reference costs44 for inpatient, outpatient and accident and emergency (A&E) visits; and Personal Social Services Research Unit (PSSRU)45 for primary care consultations (Tables 19 and 20 present the complete list of unit costs). Where relevant unit costs were not available for the correct price year of 2016/17, unit costs from other price years were inflated using the Hospital and Community Health Service (HCHS) pay and price inflation index. 45

Outcomes

The primary outcome measure was the percentage of patients with successful eradication of P. aeruginosa 3 months after the start of treatment and who remained infection free through 15 months after the start of treatment and this was calculated for both the i.v. antibiotic therapy arm and the oral antibiotic therapy arm.

The secondary outcome was the QALY, which is a composite of HRQoL, as measured by the EQ-5D-3L, and length of life. The EQ-5D-3L questionnaire has five domains (i.e. mobility, self-care, usual activities, pain and discomfort, and anxiety and depression) and three levels (i.e. no problems, some problems and extreme problems) and was completed by either patients or their carers (if < 13 years) at baseline, and at 3, 15 and 24 months. It is still unclear what the most appropriate method is to elicit HRQoL for very young children. 50,51 The EQ-5D-3L was selected in this study as it was available at the time of study design and provides consistency with measures of HRQoL frequently used in economic evaluations within adult populations. However, although there are now specific measures of HRQoL developed for children, none of the new measures has a robust evidence base for performance yet. 50 In the primary analysis, the time horizon of 15 months was used to match the collection of resource use data. In sensitivity analysis, the long-term 24-month horizon for QALYs was assessed. The EQ-5D-3L utility tariff was applied to the EQ-5D-3L questionnaire responses to generate utility scores for each time point for each patient. To calculate the QALYs, linear interpolation was assumed between the time points and the area under the curve method was applied. 52

Statistical analysis: bootstrapping and missing data

Missing data were multiply imputed (m = 25) through the use of chained equations. 53 The data were assumed to be missing at random. To account for non-normality in the distributions of costs and HRQoL, predictive mean matching was used. Missing cost data were imputed at the level of the category (inpatient care, outpatient visits, A&E visits, primary care, etc.) and total costs were passively imputed as a sum of the individual cost domains for those with missing data. For HRQoL, missing data were imputed for the utility score for each of the time points and QALYs were passively calculated. Predictors in both the cost and HRQoL multiple imputation models were patients’ age, sex, treatment arm, and costs at baseline. Owing to the large numbers of patients with missing data for wider resource use, missing data were not imputed and, instead, individuals were assumed to have zero use.

To account for statistical uncertainty and the correlation between costs and patient outcomes, the data were bootstrap sampled with replacement 2000 times. 54 For each bootstrap sample, missing data were imputed55 and incremental measures of cost and outcome were calculated using regression analysis, controlling for baseline HRQoL, age and treatment arm. The regression analysis used ordinary least squares, accounting for the multiply imputed data through the use of Rubin’s rules. 56 All regression analysis, including bootstrapping and multiple imputation, was conducted using Stata^® v14 (StatCorp LP, College Station, TX, USA).

Cost-effectiveness

In the primary analysis, mean incremental costs and outcomes were calculated using the binary treatment arm variable within the linear regression models. To calculate the uncertainty around mean incremental costs and outcomes, the percentile method was used, with the 97.5th and 2.5th percentile bootstrap iterations representing the 95% CIs. Each of the 2000 bootstrap iterations were plotted on a cost-effectiveness plane. If feasible, an ICER was calculated:

where OT is oral therapy and IVT is i.v. therapy.

A cost-effectiveness acceptability curve (CEAC) reflecting the likelihood of either technology being cost-effective for a decision-maker’s willingness to pay for the outcome variable was also plotted.

In the secondary analysis, mean incremental costs and QALYs were calculated, with 95% CIs estimated through the percentile method. The INB was estimated from the incremental costs and QALYs for i.v. antibiotic therapy compared with oral antibiotic therapy using the following formula:

with λ representing the opportunity cost of health-care resources used in the NHS, otherwise known as the cost-effectiveness threshold. NICE uses a threshold range between £20,000 and £30,000 per QALY. 57 A positive INB suggests that i.v. antibiotic therapy is the cost-effective option and a negative INB suggests oral antibiotic therapy is the cost-effective option. Cost-effectiveness planes and CEACs were also plotted for the secondary analyses.

Sensitivity analysis

A series of assumptions used in the economic evaluation were explored through one-way sensitivity analysis. Separate analyses explored:

1. full time horizon of QALYs (24 months) alongside the collection of resource use data (15 months)

2. using generalised linear models (GLMs) for the cost analysis with different link and family functions

3. using yearly health-care resource use codes for CF instead of treatment costs and inpatient stay per diem

4. using a societal perspective for the collection and incorporation of costs.

Generalised linear models

Owing to the skewed and zero-centred nature of costing data, some analysts have recommended using more complex models such as GLMs to assess incremental differences. 58 GLMs have an associated link function and family function. The base-case approach for both primary and secondary analyses involved using an identity link with Gaussian family, equivalent to ordinary least squares regression. For sensitivity analysis, a range of different family functions were also explored alongside an identity link and log-link, including inverse Gaussian, Poisson and gamma. The method of recycled predictions was used to estimate the incremental difference in costs.

Using cystic fibrosis Healthcare Resource Groups

Since 2013–14, the NHS has used specialised tariffs for patients with CF. 59 The CF currency uses a risk-adjusted banding system with seven bands of increasing complexity, with providers reimbursed on a yearly basis according to the risk-adjusted bands. The CF currency is intended to include all direct medical costs associated with treatment and bandings are split according to whether patients are adults (≥ 17 years) or children.

Patients were placed in bands according to their age, any length of stay associated with the initial intervention and the recorded length of stay for further hospitalisation.

Appendix 6, Table 40 includes the banding algorithm and Table 41 provides the unit costs for the bandings. Inpatient costs were then replaced in totality with banding costs for 1 year. All other costing estimates were included within this separate Healthcare Resource Group (HRG) analysis to calculate a total NHS and PSS cost.

Societal perspective for the inclusion of costs

A societal perspective included patients’ purchase of home aids or over-the-counter medicines, and the cost of transport to attend outpatient visits or inpatient stays, as well as work lost by patients or caregivers during the trial. Unit costs were attached to resource use using costs reported in Appendix 6. The total societal costs were calculated by adding these wider costs to the NHS and PSS costs.

Results

The analysis for health economics was based on a total of 255 patients, of whom 130 patients were randomised to the oral antibiotic therapy arm and 125 patients were randomised to the i.v. antibiotic therapy arm.

Tables 21 and 22 present resource utilisation and levels of missingness for NHS and societal resource use data, respectively. Most resource use was balanced between the arms. However, the resource use in the oral antibiotic therapy arm was suggestive of having a larger proportion of inpatient stays than the i.v. antibiotic therapy arm.

Table 23 presents data for the completion of the EQ-5D-3L for the different time points recorded. The domains most affected by CF were pain and discomfort, and self-care, with fewer patients having ‘no problems’ in these domains. There was no clear difference between EQ-5D-3L levels between arms of the trial or over the duration of the trial.

Incremental costs and outcomes

Unit costs used in the analysis are shown in Appendix 6.

In Table 24, mean costs and outcomes are calculated for the arms of the trial. The mean costs for the intervention were far higher for the i.v. antibiotic therapy arm (£7284.40) than for the oral antibiotic therapy arm (£264.30). Follow-up inpatient costs and outpatient costs were higher for the oral antibiotic therapy arm than for the i.v. antibiotic therapy arm. Overall, oral antibiotic therapy was less costly than i.v. antibiotic therapy, with the incremental difference in mean costs between the arms being –£5938.50 (95% CI –£7190.30 to –£4686.70) after adjusting for baseline covariates. Most of the total cost difference reflects the difference in intervention costs, with i.v. antibiotic therapy being associated with high inpatient stay costs. The inclusion of societal costs had a trivial impact on overall incremental cost differences between the arms of the trial, with the incremental societal cost, after adjusting for baseline covariates, being –£5938.50 (95% CI –£7190.30 to –£4686.70) for the oral antibiotic therapy arm compared with the i.v. antibiotic therapy group.

The primary outcome measure showed that the oral antibiotic therapy arm had a larger proportion of successful eradications of P. aeruginosa 3 months after the start of treatment and remaining infection free through 15 months after start of treatment than the i.v. antibiotic therapy arm. The incremental difference in effect after adjusting for baseline covariates was 0.091 (95% CI –0.034 to 0.22).

For a 15-month time horizon, patients in the oral antibiotic therapy arm gained 1.14 QALYs, with patients in the i.v. antibiotic therapy arm gaining only 1.05 QALYs. There was an overall unadjusted incremental difference in QALYs of 0.063 (95% CI 0.0074 to 0.12) after 15 months, with an adjusted incremental difference in QALYs of 0.035 (95% CI –0.015 to 0.086). After 24 months, patients in the oral antibiotic therapy arm gained 1.8 QALYs and patients in the i.v. antibiotic therapy arm gained 1.7 QALYs. There was an unadjusted incremental difference in QALYs of 0.098 (95% CI 0.013 to 0.18) and an adjusted incremental difference in QALYs of 0.058 (95% CI –0.020 to 0.14) for the oral antibiotic therapy arm compared with the i.v. antibiotic therapy arm.

Cost-effectiveness analysis

Table 25 shows incremental cost-effectiveness for the primary and secondary analysis, as well as the impact of sensitivity analyses on cost-effectiveness. In all scenarios, oral antibiotic therapy was associated with lower costs than i.v. antibiotic therapy, and was also more effective. Consequently, oral antibiotic therapy dominated i.v. antibiotic therapy in both the primary and the secondary analyses. In the secondary analysis, for a threshold of £20,000 per QALY, oral antibiotic therapy generated £6770.80 (95% CI £5027.40 to £7906.20) benefit per patient, compared with i.v. antibiotic therapy.

The use of specialised HRG costs led to a smaller difference in costs between the two arms, but the oral antibiotic therapy arm still had a lower incremental cost of –£653.08 (95% CI–£1197.80 to –£79.80) compared with the i.v. antibiotic therapy arm. Table 26 shows how patients were grouped into HRG bands. Patients in the oral antibiotic therapy arm were more likely to be in band 1 than patients in the i.v. antibiotic therapy arm, in which patients tended to be clustered in higher-risk/higher-cost bands. However, the highest band reported in the data (band 3) included a larger proportion of patients in the oral antibiotic therapy arm than in the i.v. antibiotic therapy arm as a result of a small number of patients in the oral antibiotic therapy arm having a long inpatient stay in the follow-up period. In the secondary analysis with CF HRG costs, the oral antibiotic therapy arm generated £653.10 (95% CI £79.80 to £1197.80) benefit per patient for a cost-effectiveness threshold of £20,000 per QALY.

Changes in model function for the cost data and the inclusion of societal costs led to only small changes in the incremental cost differences between the arms of the trial. The INB was large for oral antibiotic therapy compared with i.v. antibiotic therapy in each scenario.

Incremental cost-effectiveness planes and CEACs for the primary analysis, secondary analysis and associated sensitivity analyses are shown in Appendix 6, Figures 33–45.

For each scenario, oral antibiotic therapy was highly likely to be cost-effective, with virtually all of the bootstrap replicates falling within the south-east quadrant, demonstrating that the oral antibiotic therapy was both more effective and less costly than i.v. antibiotic therapy. For the secondary analysis involving a cost-effectiveness threshold of £20,000 per QALY, oral antibiotic therapy had a 100% chance of being cost-effective compared with i.v. antibiotic therapy in the base-case analysis, and findings were robust to assumptions regarding the use of CF HRG codes, costing model specification and inclusion of societal costs.

Contributions of authors

Simon C Langton Hewer (https://orcid.org/0000-0002-2711-8258) (chief investigator, lead applicant on grant, PI at Bristol Royal Hospital for Children, Consultant Respiratory Paediatrician) contributed to the conception and design of the trial, delivery and management of the trial, interpretation of data, drafting and revision of the report and approval of the report.

Alan R Smyth (https://orcid.org/0000-0001-5494-5438) (co-chief investigator, lead applicant on grant, PI at Nottingham Children’s Hospital, Consultant Respiratory Paediatrician) contributed to the conception and design of the trial, delivery and management of the trial, interpretation of data, drafting and revision of the report and approval of the report.

Michaela Brown (https://orcid.org/0000-0002-7772-271X) (Trial Statistician) contributed to the delivery and management of the trial, analysis and interpretation of data, drafting and revision of the report and approval of the report.

Ashley P Jones (Co-applicant on grant, Lead Statistician) contributed to the conception and design of the trial, delivery and management of the trial, analysis and interpretation of data, drafting and revision of the report and approval of the report.

Helen Hickey (https://orcid.org/0000-0003-0467-0362) (Head of Trial Management) contributed to the delivery and management of the trial, interpretation of data, drafting and revision of the report and approval of the report.

Dervla Kenna (https://orcid.org/0000-0002-3105-3637) (Clinical Scientist) contributed to the analysis of the laboratory samples, delivery and management of the trial, analysis and interpretation of data, drafting and revision of the report and approval of the report.

Deborah Ashby (https://orcid.org/0000-0003-3146-7466) (Co-applicant, Senior Statistician) contributed to the conception and design of the trial, delivery and management of the trial, analysis and interpretation of data, drafting and revision of the report and approval of the report.

Alexander Thompson (https://orcid.org/0000-0003-4930-5107) (Health Economist) contributed to the analysis and interpretation of data, drafting and revision of the report and approval of the report.

Laura Sutton (https://orcid.org/0000-0003-3327-5927) (Trial Statistician) contributed to the analysis and interpretation of data, revision of the report and approval of the report.

Dannii Clayton (https://orcid.org/0000-0003-3535-3466) (Trial Statistician) contributed to the analysis and interpretation of data, revision of the report and approval of the report.

Barbara Arch (https://orcid.org/0000-0001-6060-8091) (Trial Statistician) contributed to the analysis and interpretation of data, revision of the report and approval of the report.

Łukasz Tanajewski (https://orcid.org/0000-0002-4387-3093) (Health Economist) contributed to the analysis and interpretation of data and approval of the report.

Vladislav Berdunov (https://orcid.org/0000-0002-5743-2184) (Health Economist) contributed to analysis and interpretation of data and approval of the report.

Paula R Williamson (https://orcid.org/0000-0001-9802-6636) (Senior Statistician) was the Director of the Medicines for Children Clinical Trials Unit (now Liverpool Clinical Trials Centre) at the time the study was conceived, and was also involved in the feasibility study (https://trialsjournal.biomedcentral.com/articles/10.1186/1745-6215-11-11) supporting the trial. She provided statistical input to the main study design, grant application, and case report forms. She was on the Trial Management Group, provided input to the protocol, supervised and oversaw contributions from CTU staff (statistical monitoring and analysis, trial and data management) throughout. She undertook blind review of the data prior to final analysis. She contributed to, and approved the final report.

Publications

Smith CT, Williamson P, Jones A, Smyth A, Langton Hewer S, Gamble C. Risk-proportionate clinical trial monitoring: an example approach from a non-commercial trials unit. Trials 2014;15:127.

Langton Hewer SC, Smyth AR, Brown M, Jones AP, Hickey H, Kenna D, et al. Intravenous vs. oral antibiotics for eradication of Pseudomonas aeruginosa in cystic fibrosis (TORPEDO-CF): a randomised controlled trial. Lancet Respir Med 2020;8:975–86.

Data-sharing statement

All data requests should be submitted to the corresponding author for consideration. Access to anonymised data may be granted following review and appropriate agreements being in place.

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.

Trial Steering Committee

Independent members: Professor Jonathan Grigg, Dr David Stableforth, Professor Barry Plant, Professor Duncan Geddes, Mrs Jennifer Wederell, Dr Ranjit Lall, Ms Sophie Lewis and Mr Dominic Kavanagh.

Non-independent members: Dr Simon Langton Hewer and Professor Alan Smyth.

Independent Data and Safety Monitoring Committee

Dr Robert Dinwiddie, Professor Christiane De Boeck and Mrs Enid Hennessy.

Trial Management Group

Dr Simon Langton Hewer, Professor Alan Smyth, Professor Deborah Ashby, Professor Paula Williamson, Ms Jessica Bissett, Dr Ashley Jones, Mrs Michaela Brown, Ms Helen Hickey, Ms Sue Howlin, Ms Farhiya Ashoor and Dr Dervla Kenna.