A high-dose preparation of lactobacilli and bifidobacteria in the prevention of antibiotic-associated and Clostridium difficile diarrhoea in older people admitted to hospital: a multicentre, randomised, double-blind, placebo-controlled, parallel arm trial (PLACIDE)

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 06/39/02. The contractual start date was in May 2008. The draft report began editorial review in January 2013 and was accepted for publication in April 2013. 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.

Declared competing interests of authors

Stephen Allen has undertaken research in probiotics supported by Cultech Ltd, Baglan, Port Talbot, UK; the Children and Young People’s Research Network, Cardiff University, Wales; the Knowledge Exploitation Fund, Welsh Development Agency and the National Ankylosing Spondylitis Society, UK. He has also been an invited guest at the Yakult Probiotic Symposium in 2011 and received research funding from Yakult, UK, in 2010.

Copyright statement

© Queen’s Printer and Controller of HMSO 2013. This work was produced by Allen et al. under the terms of a commissioning contract issued by the Secretary of State for Health. 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.

Background

Antibiotic-associated diarrhoea (AAD), defined as diarrhoea in association with antibiotic treatment and without an alternative cause, occurs in 5–39% people within 12 weeks of exposure to antibiotics. 1,2 The predominant mechanism underlying AAD is disruption of the commensal gut flora. This impairs colonisation resistance and facilitates the emergence of a range of gut pathogens. 2,3 The major diarrhoeal pathogen associated with antibiotic treatment is Clostridium difficile, which accounts for 15–39% of AAD cases. 4 Altered commensal flora may also result in diarrhoea through changes in the metabolism of carbohydrates, short-chain fatty acids and bile acids, and some antibiotics cause diarrhoea through direct effects such as increased gut motility. 2,5

Clostridium difficile is an anaerobic bacterium that produces resistant spores that persist long term in the environment. Transmission is faecal–oral and 4–21% patients may acquire the infection during admission to hospital through contact with colonised patients, contaminated fomites and the hands of health-care staff. C. difficile diarrhoea (CDD) occurs in both endemic and outbreak scenarios. 6 Since 2003, an increase in the frequency and severity of CDD in North America and Europe has been attributable to the emergence of the hypervirulent 027 strain, which may produce higher amounts of toxin. 7–9

Antibiotic-associated diarrhoea occurs typically in older people admitted to hospital. 1,2 ADD complicates treatment with many antibiotic classes but especially broad-spectrum penicillins, cephalosporins, clindamycin and long-duration antibiotic treatments. 10 Additional risk factors include prolonged hospital stay, previous hospital admission, previous gastrointestinal surgery and use of a nasogastric tube (NGT). 3 In a retrospective study of European hospitals, risk factors for CDD included age ≥ 65 years, severe comorbidity and recent treatment with cephalosporins and aminopenillin-beta (β)-lactamase inhibitor combinations. 11 In a prospective study of people aged > 18 years admitted to Canadian hospitals, age, exposure to antibiotics, treatment with proton pump inhibitors (PPIs) and previous hospital admission within the last 2 months predicted CDD. 12

The severity of AAD varies greatly. Although usually a mild, self-limiting illness, it is associated with prolonged hospital stay and increased health-care costs. C. difficile infection may remain asymptomatic, but intestinal pathology results from secretion of toxins A and B causing increased mucosal fluid secretion and inflammation. Symptoms range from mild, self-limiting diarrhoea to severe diarrhoea complicated by pseudomembranous colitis, toxic megacolon and death. 6,8

Estimates of the financial burden of C. difficile infection (CDI) to the health-care service have varied between $2454 and $16,464 for every health-care-acquired case in the USA,13–15 £4107 in the UK16 and €7147 in Germany. 17 The annual cost of health care for CDD in the USA has been estimated to be $3B. 18,19 Nosocomial CDI increases hospital stay by between 7 and 26 days,16,17,20 and prolonged hospital admission was identified as the main cost driver in most studies. 13,15,16 Furthermore, an increase in length of hospital stay due to more severe disease in recent years has resulted in a rise of the incremental cost of CDI. 21

Treatment

Uncomplicated AAD usually responds to withdrawal of the offending antibiotics. CDD usually responds to treatment with metronidazole or vancomycin, but 20–25% patients go on to suffer from a recurrent form of the disease. Novel modes of treatment include probiotics, immunoglobulin infusion and faecal transplant from healthy donors. 3,6

Prevention of antibiotic-associated and Clostridium difficile diarrhoea

The frequency and severity of CDD in hospitals in the industrialised world have led to comprehensive strategies for prevention. These include decontamination of the environment, hand washing and isolation of patients with diarrhoea. 6 Antibiotic stewardship programmes have also been effective in reducing infection rates. 6,22,23

Probiotics

Probiotics are defined as live microbial organisms which, when administered in adequate numbers, are beneficial to health. 24 Although clinical trials are undertaken to determine whether or not a microbial preparation has a health benefit in a specific population, the term ‘probiotic’ is commonly used to refer to the preparation being evaluated and it is used in this sense here. Bacteria used as probiotics are among the organisms ‘generally regarded as safe’ by the Food and Agriculture Organization of the United Nations. 25 Live bacteria from several genera and the yeast Saccharomyces boulardii have been administered to vulnerable groups such as preterm infants and people with human immunodeficiency virus (HIV) infection without adverse effects. 26 Adverse events occurring in clinical trials evaluating the prevention of AAD have not been ascribed to probiotic intake. 26 Administration of lactic acid bacteria has been associated in rare cases with septicaemia in immunocompromised people and endocarditis in people with artificial heart valves. 27 Despite the apparent low risk of adverse events, careful assessment of safety in clinical trials is recommended. 26

Rationale

Specific probiotic strains have been identified in vitro and in vivo to possess several mechanisms that may prevent or ameliorate AAD and CDD through enhanced barrier function of the gut epithelium. 28 First, potential pathogens are killed or inhibited by the secretion of antimicrobial peptides and probiotics compete for attachment sites on intestinal mucus and epithelium. Acidification of the gut contents through the production of lactic acid also inhibits pathogen growth. Second, mucosal immunity is enhanced through increased secretory immunoglobulin A production and stimulation of antimicrobial peptide secretion by host cells. Finally, through direct effects on the epithelium, probiotics increase the secretion of mucus, enhance tight junction integrity and decrease epithelial cell apoptosis.

At the time that probiotic lactobacilli and bifidobacteria in antibiotic-associated diarrhoea and Clostridium difficile diarrhoea in the elderly (PLACIDE) was developed, meta-analyses had suggested that probiotics may be effective in the prevention of AAD. McFarland pooled data from 25 randomised controlled trials (RCTs), which included a total of 2810 adults and children. 29 There was significant heterogeneity (χ² = 82.5; p < 0.001) in a fixed-effects model. In random-effects analysis, the relative risk (RR) for AAD was 0.43 [95% confidence interval (CI) 0.31 to 0.58] in participants receiving a probiotic. This meta-analysis included all of the trials included in previous systematic reviews, which had broadly similar findings. 30–33 Although meta-analysis has provided some evidence that probiotics may be effective in the prevention of AAD, the marked differences between trials in the microorganisms evaluated (single strains and mixtures of bacteria, the yeast S. boulardii), administration regimens (mode of delivery, number of organisms, probiotics combined with prebiotics), patient populations (age and exposure to antibiotics) and period of follow-up for AAD probably underlie the statistical heterogeneity in the results and weaken the evidence for probiotic effectiveness.

Much less evidence from clinical trials was available for probiotics in the prevention of CDD. A pilot study in elderly hospitalised patients reported that 30 out of 138 (21.7%) patients developed diarrhoea and 5 out of 69 (7.2%) in the placebo group compared with 2 out of 69 (2.9%) who had received a combination of Lactobacillus acidophilus and Bifidobacterium bifidum tested positive for C. difficile toxin. Stool culture suggested that the main effect of the probiotic was neutralisation of toxin rather than prevention of colonisation. 34 Thomas et al. had assessed Lactobacillus rhamnosus GG for the prevention of AAD in 267 adults who were monitored for an average of 21 days, but the number of patients from whom stool samples had been collected and in which C. difficile was detected was too small to assess probiotic effect. 35 In randomised trials of 19336 and 180 adult patients,37 the occurrence of CDD was similar in those receiving S. boulardii and those receiving the placebo. Although there was trial evidence for probiotics in the treatment of established CDD or the prevention of recurrence,29 we are not aware of any other studies that had assessed probiotics in the prevention of CDD in adults.

We are not aware of any studies that have assessed the effect of probiotic on quality of life (QoL) in patients at risk of AAD.

Selection of the probiotic preparation

Although several mechanisms whereby probiotic organisms enhance gut barrier function have been identified,28 it remains unclear which of these are most relevant to the prevention or amelioration of AAD and to what extent these mechanisms are common to many different probiotic organisms or are strain specific. Therefore, the scientific evidence to inform the selection of a specific probiotic intervention for the prevention of AAD is limited. The meta-analyses included trials that had evaluated many different bacterial preparations and the yeast S. boulardii. 29 The bacterial interventions included single strains, mixtures of different organisms and mixtures of probiotics with prebiotics. Dosages (number of organisms) and modes of administration also varied markedly between studies. Factors associated with greater efficacy in preventing AAD in meta-analysis included the use of S. boulardii or L. rhamnosus GG, mixtures of probiotics and preparations with high numbers of organisms. 29 Efficacy was similar for bacterial preparations [five trials conducted on 384 adults and children; odds ratio (OR) 0.34; 95% CI 0.19 to 0.61] and yeast preparations (four trials conducted on 830 participants; OR 0.39; 95% CI 0.25 to 0.62). 31

In an attempt to maximise gut colonisation and, therefore, colonisation resistance, we used a multistrain preparation of Lactobacillus sp. and Bifidobacterium sp., bacterial species that had been evaluated extensively in clinical trials, with a high number of viable bacteria (total 6 × 10¹⁰ organisms per day). We intended to undertake a pragmatic trial to assess the clinical effectiveness and cost-effectiveness of the probiotic preparation in older people receiving antibiotics in secondary health-care settings representative of those in industrialised counties and with the causes of diarrhoea determined by routine laboratory practice.

Trial design

Probiotic lactobacilli and bifidobacteria in antibiotic-associated diarrhoea and Clostridium difficile in the elderly was a prospective, multicentre, pragmatic, two-armed, double-blind, randomised, placebo-controlled trial with equal randomisation.

Approvals obtained

The Research Ethics Committee (REC) for Wales approved the study on 27 November 2008 (number 08/MRE09/18).

A clinical trial authorisation (CTA) was granted for the live bacterial preparation by the Medicines and Healthcare products Regulatory Agency (MHRA) on 6 October 2008 (CTA number AAD-CDD-001). The details of the REC for Wales, local RECs, competent authorities and research and development department approvals are provided in Appendix 1. The trial was assigned the International Standard Randomised Controlled Trial Number (ISRCTN) ISRCTN70017204 and EudraCT number 2007–002876–32.

Trial sites

Inpatients were recruited from medical and surgical wards of secondary care hospitals from two NHS regions: the Abertawe Bro Morgannwg University Health Board (ABMUHB) in South West Wales (Singleton, Morriston and Princess of Wales hospitals), and the County Durham and Darlington Foundation Trust (CDDFT; the University Hospital of North Durham and Darlington Memorial Hospital). Details of the trial sites are included in Appendix 2.

Participant inclusion criteria

People aged ≥ 65 years who were admitted to hospital and had been exposed to one or more antibiotics within the preceding 7 days, or were about to start antibiotic treatment, were eligible to join the study. Approval to invite the patient to participate in the study was secured from the patient’s consultant.

Participant exclusion criteria

People were excluded if they:

• already had diarrhoea, which was defined as three or more watery or loose stools (Bristol Stool Form Scale types 5–7)38 in the preceding 24 hours

• were sufficiently immunocompromised to require isolation and/or barrier nursing

• had a severe illness requiring high-dependency or intensive care

• had a prosthetic heart valve

• had suffered from CDD in the previous 3 months

• had inflammatory bowel disease that had required specific treatment in the previous 12 months

• had suspected acute pancreatitis, which was defined as abdominal pain with serum amylase or lipase greater than three times the institutional upper limit of normal

• had a known abnormality or disease of mesenteric vessels or coeliac axis that may compromise gut blood supply

• had a jejunal tube in situ or were receiving jejunal feeds

• had a previous adverse reaction to probiotics

• were unwilling to discontinue their existing use of probiotics.

Investigational medicinal products

The active preparation consisted of a vegetarian capsule containing 6 × 10¹⁰ live bacteria as lyophilised powder comprising two strains of L. acidophilus [CUL60, National Collection of Industrial, Food and Marine Bacteria (NCIMB) 30157 and CUL21, NCIMB 30156; 3 × 10¹⁰ colony-forming units (cfu)] and two strains of Bifidobacterium (B. bifidum CUL20, NCIMB 30153 and Bifidobacterium lactis CUL34, NCIMB 30172; 3 × 10¹⁰ cfu). Obsidian Research Ltd, Port Talbot, UK, prepared the investigational medicinal products (IMPs) according to the randomisation schedule. The organisms were available commercially through BioCare^® Ltd (Lakeside, Birmingham, UK) and Pharmax (Bellevue, WA, USA). The placebo capsules were identical to the probiotic capsules but contained inert maltodextrin powder. The dose was one capsule per day with food and, where possible, between antibiotic doses, for 21 days.

Each participant was provided with a bottle labelled with his or her unique study number and containing 21 capsules. The IMPs were given to participants on the day of recruitment and administration during the hospital stay was supervised by the research nurses in ABMUHB and by the ward nurses in CDDFT. To prevent cross-contamination, strict hygiene procedures were followed, for example where capsules were opened, and the contents were mixed with fluids for administration to participants with difficulty swallowing. Although the live bacteria had a shelf life of 2 months at room temperature, participants were instructed to keep the capsules in the refrigerator following discharge from the hospital and encouraged to complete the 21-day course.

In vitro antibiotic testing demonstrated that the lactobacilli and bifidobacteria were sensitive to broad-spectrum antibiotics. All four strains were sensitive to ceftriaxone, chloramphenicol, erythromycin, linezolid and tetracycline.

After recruitment of the first 50 participants, a research nurse perceived a slight difference in the colour of the powder contained in the probiotic and placebo capsules, which were transparent. Therefore, the IMPs were resupplied in opaque capsules according to the original random allocation sequence. No unmasking of participant allocation occurred.

To ensure the quality of the IMPs, unused capsules were collected from participants on an opportunistic basis (e.g. from participants who withdrew before completing the 21-day course). A laboratory independent of the research team performed quantitative bacterial culture. The findings were sent to the independent statistician to check against the randomisation sequence and for the total count of viable organisms.

Objectives

Primary objectives: to determine the clinical effectiveness and cost-effectiveness of a live preparation of two strains of lactobacilli and two strains of bifidobacteria in the prevention of AAD and CDD in people aged ≥ 65 years who were exposed to oral or intravenous antibiotics and who were representative of patients admitted to secondary care NHS facilities in the UK.

Secondary objectives: to assess the effect of the probiotic on the duration and severity of AAD, the acceptability of the probiotic preparation, serious adverse events (SAEs) of the probiotic preparation and its effect on QoL.

Outcomes

Primary outcomes: the occurrence of AAD within 8 weeks and CDD within 12 weeks of recruitment.

Secondary outcomes: the severity and duration of AAD; the severity and duration of CDD and incidence of recurrence within the follow-up period; the number of days with abdominal pain, bloating, flatus and nausea; the incidence of pseudomembranous colitis, the need for colectomy; well-being and QoL; duration of hospital stay; frequency of SAEs; the acceptability of the live bacterial preparation (to identify if the participants had any difficulty taking the bacterial preparation vs. the placebo); the viability of the bacterial preparation at point of administration and death.

Sample size

Based on a review of previous clinical trials, we estimated that AAD would occur in 20% and CDD in 4% of participants in the placebo arm. The detection of a 50% reduction in CDD to a frequency of 2% in the active arm with 80% power at the 5% significance level required 2478 subjects (1239 in each arm). This number of participants would provide a power of > 99% to detect a 50% reduction in the risk of AAD (to 10% frequency) and a power of 90% to detect a 25% risk reduction in AAD (to 15% frequency) in the active arm at the 5% significance level. We planned to recruit 2974 participants to allow for 10% drop-outs and 10% loss to follow-up due to deaths unrelated to diarrhoea.

Randomisation

The independent statistician prepared a random allocation sequence using blocks of variable sizes and stratified by hospital using SAS^® PROC PLAN, version 9.1 (SAS Institute Inc., Cary, NC, USA). The sequence allocated participants in a 1 : 1 ratio to the two arms of the study.

Blinding

The allocation sequence was not available to any member of the research team until databases had been completed and locked. In view of the established safety record of lactobacilli and bifidobacteria27 and the sensitivity of the organisms used in this study to broad-spectrum antibiotics, there was no provision for the emergency unblinding of participants. Therefore, copies of the random allocation sequence were not held at the recruitment sites.

Recruitment

The research nurses received training in ‘Good Clinical Practice’ and were fully conversant with all aspects of the trial. Specific training was given in assessment of patient eligibility, recruitment and obtaining informed consent, the trial protocol, reporting of adverse events and collecting information. Presentations were made to clinical, nursing and pharmacy staff to ensure their familiarity with the purpose and conduct of the trial. Permission to approach patients and invite them to join the study was obtained from hospital consultants.

Research nurses visited wards daily to identify eligible patients and provide them with verbal and written information (see Appendix 3). Patients were revisited either later the same day or the next day after they had had the opportunity to discuss with relatives/carers and health-care personnel whether or not they wished to join the study. For people unable to provide consent, information was provided to their relatives/carers and assent sought (see Appendix 3). Although reasons for not joining the study were requested, patients and their relatives/carers were free to decline to participate without giving a reason. Patients in whom consent or assent was obtained were allocated the next unique study number in the random sequence for that site and the research nurses provided them with the corresponding trial preparation.

Baseline assessment

Information recorded at recruitment included basic demographic characteristics, use of cigarettes and alcohol, source of admission, principal diagnosis or reason for admission, comorbid illnesses, duration of hospital stay prior to recruitment, non-antibiotic drug treatment, indication for antibiotic treatment and antibiotics prescribed (see Appendix 4).

Participant follow-up

Participants were visited daily by the research staff during hospital stay. Changes to antibiotic treatment, gastrointestinal symptoms (including the presence of diarrhoea), compliance, difficulties taking the IMPs and occurrence of SAEs were recorded onto standard forms (see Appendix 4). The same information was requested weekly, after discharge from hospital, via a telephone questionnaire and continued for 8 weeks post recruitment. In addition, participants were encouraged to contact a named member of the research staff to report potential SAEs at any time during follow-up. Review of laboratory data regarding stool assays was continued until 12 weeks after recruitment.

Trial completion

Participants had completed the trial when they:

• had completed follow-up

• had withdrawn and declined collection of further follow-up data

• were lost to follow-up

• had died.

A chart showing participant flow through the study is included as Appendix 5.

Measurement of primary outcomes

Diarrhoea was defined as three or more loose stools (stool consistency 5–7 on the Bristol Stool Form Scale)38 in a 24-hour period. Diarrhoea was also diagnosed in participants with frequent stools that they described as loose but who were unable to describe stool consistency using the Bristol Stool Form Scale. The presence of diarrhoea according to these criteria was confirmed by the research nurses during admission with either the patient, their carers or a member of the medical staff. After discharge, this was confirmed during telephone interview.

Stool samples for analysis were collected only during episodes of diarrhoea, including diarrhoea that occurred after discharge from hospital. Stools were analysed for diarrhoeal pathogens according to routine NHS laboratory practice. Analyses included bacterial culture for Salmonella spp., Shigella spp., Campylobacter and Escherichia coli 0157 and wet film for ova, cysts and parasites. Detection of viruses depended on the clinical context (e.g. suspected norovirus outbreak). In ABMUHB, C. difficile toxins were detected by an in-house tissue culture assay with confirmation by enzyme immunoassay Premier™ Toxins A&B (Meridian Bioscience, Inc., Cincinnati, OH, USA). In CDDFT, toxins were detected by the VIDAS^® C. difficile A&B (bioMérieux SA, Marcy l’Etoile, France). To improve the detection of CDD from June 2010, CDDFT also introduced detection of the C. difficile antigen glutamate dehydrogenase C. DIFF QUIK CHEK^® (TECHLAB^® Inc., Blacksburg, VA, USA) to be used in conjunction with the toxin assay. Further stools samples were requested if a cause of diarrhoea was not identified and especially if there was clinical suspicion of CDD. Stools positive for C. difficile toxin were cultured and the isolates sent to a central reference laboratory for ribotyping.

Antibiotic-associated diarrhoea was defined as occurring in association with antibiotic therapy but without diarrhoeal pathogens detected on routine laboratory analysis or an alternative explanation (e.g. laxative treatment). Among patients with AAD, CDD was diagnosed in those with stools testing positive for C. difficile toxins.

Measurement of secondary outcomes

Information regarding the severity and duration of AAD and CDD, number of stools per day and stool consistency, incidence of recurrence of CDD within the follow-up period, the number of days with abdominal pain, bloating, flatus and nausea, duration of hospital stay and acceptability of the live bacterial preparation was collected by the research nurses during follow-up (see Appendix 4). CDD was investigated and managed according to the current hospital practice and clinical and laboratory information from clinical records was recorded (see Appendix 4). Information regarding the occurrence of pseudomembranous colitis, the need for colectomy and death was also collected from the patient’s case records. The information from follow-up and patient case records was used to classify the severity of CDD according to UK national guidelines (see Appendix 6). 39 The severity of CDD was classified independently by two assessors, WH and SA, and differences resolved by discussion.

Quality of life was assessed by the generic Short Form questionnaire-12 items version 2 (SF-12 v2) administered by research nurses at baseline and at 4 and 8 weeks. SF-12 v2 QoL subscales (social function, role function–emotional, role function–physical, general health, mental health, bodily pain, vitality, physical functioning), physical component summary (PCS) and mental component summary (MCS) scores were calculated and quality assured according to the QualityMetric Incorporated guidance40 with imputation of missing scores using the SF-12 v2 Missing Data Estimator software where possible. 40 SF-12 v2 subdomain, PCS and MCS scores were allocated a value of 0 for the lowest/worse score and 100 for the highest/best score.

Serious adverse events were identified and reported according to standard guidelines. 41 Suspected unexpected serious adverse reactions were to be reported immediately to the independent safety monitor and, if appropriate, the ethics committee in accordance with local and national requirements, which were identified as:

• any manifestation of infection (e.g. abscess, bacterial endocarditis, bacteraemia) in which lactobacilli or bifidobacteria were isolated from pathological specimens

• the development of pancreatitis, which was defined as abdominal pain with serum amylase or lipase concentration equal to or greater than three times the institutional upper limit of normal

• the development of multiorgan failure

• the development of bowel ischaemia.

Additional data collected

Information was collected to identify subgroups of participants who may be at increased risk of AAD and CDD. This included potential risk factors at baseline, antibiotic treatment and duration of hospital stay.

Data management

All data were collected on standardised forms that were checked for missing values by the trial manager. Routine laboratory records were accessed to record results of stool analyses. Data were entered into Microsoft Access^® (Microsoft Corporation, Redmond, WA, USA) and included range checks and double entry. Databases were compared using Microsoft Excel^® (Microsoft Corporation, Redmond, WA, USA) to identify data entry errors.

Antibiotics were classified according to British National Formulary categories (see Appendix 7). 42 Indications for antibiotic treatment were classified according to the system organ class (SOC) terminology of the Medical Dictionary for Regulatory Activities (MedDRA). 43 Participants were allocated to each SOC for either suspected or proven infection of that organ or system. Antibiotic treatment for suspected infection where no system or organ was identified was classified as ‘suspected sepsis but site unclear’. SAEs were coded according to the most appropriate preferred terms (PTs) of the MedDRA. 43

After data cleaning, databases were locked and forwarded to the trial statistician for analysis. Initial analysis according to the randomisation sequence identified the two arms of the study as only ‘A’ or ‘B’. The identity of the two arms was only revealed after the Data Monitoring and Ethics Committee had reviewed these data and approved the analysis.

Statistical analysis

Demographic and baseline data were summarised by recruitment hospital and treatment group. Continuous variables were summarised using number of observations, median and interquartile range (IQR) and categorical variables by the number and percentage of events.

Analysis of primary and secondary end points was performed in a modified intention-to-treat (ITT) population that excluded a small number of subjects who withdrew shortly after randomisation and did not have follow-up data. The chi-squared test or Fisher’s exact test was used to compare proportions. Risk difference and RR together with the 95% CIs were calculated using a generalised linear model that included treatment as a single predictor. Similarly, CIs for the ORs of AAD and CDD were estimated from logistic regression models. Secondary continuous outcomes with no repeated measurements were summarised using number of observations, median and IQR. The t-test or Mann–Whitney method were used to compare continuous variables. Duration data were summarised by median and IQR and compared using the Mann–Whitney method. No transformation was used for any variables.

Analysis of primary end points was also performed by logistic regression model adjusting for the following prespecified baseline characteristics and potential risk factors for AAD that may be likely to affect the occurrence of the primary end point:

• centre

• age

• sex

• antibiotic class

• duration of antibiotic treatment

• antacid therapy (including PPI treatment)

• NGT in situ

• previous gastrointestinal surgery

• recent previous hospital admission

• duration of hospital stay.

We intended to include all 10 prespecified variables in the logistic model but some were not entered because their effects were inestimable. A per-protocol (PP) population excluded participants who did not receive any IMP doses or in whom compliance was unclear. Additional analyses also assessed probiotic effect on primary outcomes in participants who took all 21 doses, 14 or more doses, or seven or more doses of the IMP. Analysis methods for the PP population were similar to those described for the modified ITT population. All analyses were performed on both the modified ITT and PP populations.

SAS version 9.2 was used for data analyses.

Quality of life analysis

The main analysis compared the change from baseline in SF-12 v2 PCS and MCS composite scores at 4 weeks in the two study arms. We also compared SF-12 v2 PCS and MCS composite scores at 8 weeks and scores for individual SF-12 v2 subdomains (social function, role function–emotional, role function–physical, general health, mental health, bodily pain, vitality and physical functioning) at 4 and 8 weeks. Composite scores were compared using mixed model analysis using the SF-12 v2 score at baseline as a covariate, treatment, visit, interaction between the treatment and visit as fixed effects, and subject as a random effect. During the trial, some subjects dropped out, resulting in some incomplete observations. These incomplete observations were not computed but were assumed to be missing at random in the mixed model analysis. The treatment difference, together with the 95% CI at each visit, was derived from the mixed model.

Economic analysis

The cost-effectiveness evaluation was undertaken from the perspective of the UK NHS. Costs were assigned to the resources utilised by each participant. These consisted of the bacterial preparation, staff time involved in administering the preparation, treatment relating to potential adverse events, the assessment of cases of diarrhoea (stool collection and culture/toxin assay, endoscopy) and diarrhoea management costs (laundry, antibiotics, increased hospital stay and comorbidities). Unit costs (cost year 2011) were derived from published information44–46 and through discussions with relevant clinical and finance department staff. Missing data were addressed using the imputation-based method for QoL data and censored data relating to costs using the weighted cost method with known cost histories. 47 In view of the short timescale of the project, there was no discounting of the costs or benefits. Costs and benefits would have been discounted at the conventional rate of 3.5% if the time scale of the follow-up had exceeded 1 year.

Cost differences between the two arms of the study were determined and used in conjunction with differences in outcomes in undertaking a cost–consequences analysis, with cost per case of AAD averted as the primary outcome measure. We planned to undertake subgroup analyses to determine the relative cost-effectiveness of preventative strategies in different risk groups as indicated in the covariate analysis. In addition, a cost–utility analysis considered the differences in costs between the two study arms and differences in quality-adjusted life-years (QALYs) derived from European Quality of Life-5 Dimensions (EQ-5D) responses.

Resource use and costs

Inclusion and exclusion criteria

In practice, on assessment for eligibility, some patients were severely ill and not expected to survive for the intended period of follow-up; therefore, they were not approached regarding joining the study. Similarly, participants who were nil by mouth at initial assessment were not approached.

Follow-up

We had intended that diarrhoea outcomes would be assessed during antibiotic treatment and for 8 weeks after stopping antibiotics. However, prolonged follow-up for participants on long courses of antibiotics was not feasible. In practice, daily follow-up during hospital stay or weekly follow-up after discharge from hospital was continued to 8 weeks after recruitment. Review of laboratory data regarding stool assays was continued until 12 weeks after recruitment.

Assessment of quality of life

We considered modifying tools validated to measure QoL in treatment-induced diarrhoea in people with HIV54 and older patients with faecal incontinence. 55 However, we considered that completion of additional questionnaires would be too onerous for elderly inpatients and these were not pursued.

Participant characteristics according to intervention arm

Consent to participate in the trial was provided directly by 1398 patients (95.1%) in the probiotic arm and 1392 patients (94.6%) in the placebo arm. For patients unable to give consent themselves, assent was usually provided by a family member: daughter [in 24 cases (1.6%) in the probiotic arm and 34 cases (2.3%) in the placebo arm], wife [in 19 cases (1.3%) in the probiotic arm and 13 cases (0.9%) in the placebo arm], or son [in 15 cases (1.0%) in the probiotic arm and 18 cases (1.2%) in the placebo arm].

Baseline demographic and patient characteristics were similar in the two intervention arms except for a greater proportion of males than females in the probiotic arm and vice versa in the placebo arm (Table 2). The frailty of the study population is reflected in the median age of 77.1 years and common occurrence of comorbid illnesses: 54.6% of participants suffered from hypertension, 24.1% from chronic obstructive pulmonary disease (COPD) and 22.9% from diabetes. Participant age ranged from 65.0 to 107.5 years in the probiotic arm and from 65.0 to 104.4 years in the placebo arm. More participants were recruited during the winter than in the summer months. The majority of patients were admitted to hospital from home and approximately one-third had been admitted to hospital within the previous 8 weeks. Very few people of non-white ethnic origin were recruited. Cigarette smoking was uncommon, but approximately one in three participants drank alcohol. Recent consumption of foods containing live bacteria was uncommon among all participants and occurred with a similar frequency in both study arms.

Participant characteristics according to centre

Overall, 1873 (63.7%) inpatients were recruited in hospitals in ABMUHB (Singleton, Morriston and Princess of Wales) and 1068 (36.3%) in CDDFT (Durham and Darlington). In ABMUHB, recruitment began with a pilot study of 50 patients in Morriston Hospital on 1 December 2008 to evaluate the recruitment methods and data collection forms. Methods were found to be reliable and these patients were included in the final analysis. Recruitment continued until 28 February 2012 and a total of 1479 patients (50.3% of total) were recruited (see Table 2). Recruitment in Singleton Hospital began on 9 June 2009 but was terminated on 9 February 2011, after 203 (6.9%) patients had been recruited, because of falling numbers of eligible patients due to service reconfiguration. To maintain patient numbers, recruitment was undertaken at Princess of Wales Hospital from 5 May 2011 to 10 January 2012 and 191 (6.5%) patients were recruited. The start of recruitment was delayed in CDDFT for operational reasons. Darlington Memorial Hospital recruited 521 (17.7%) patients from 12 October 2009 to 27 February 2012 and University Hospital of North Durham recruited 547 (18.6%) patients from 17 November 2011 to 28 February 2012 (see Table 2).

Baseline participant characteristics were generally similar across the hospital sites (Table 3) with some exceptions. The greater proportion of males in the probiotic arm and females in the placebo arm occurred in all centres except for Singleton Hospital, where there were more females than males in the probiotic arm (data not shown). Participants recruited at Singleton Hospital were more likely to be female and tended to be older than participants from other hospitals. The period during which recruitment in each centre occurred was reflected in the lower proportion of patient recruitment during the winter months in Princess of Wales than in other hospitals. The frequency of COPD and hospital admission in the previous 8 weeks were both more common in hospitals in CDDFT than in ABMUHB.

Indications for initial antibiotic treatment

Indications for antibiotic treatment classified according to the MedDRA SOC43 were similar in the two study arms (Table 4). The most common indication was ‘respiratory, thoracic and mediastinal disorders’. Antibiotic treatment for suspected sepsis where the site was unclear was given to a small proportion of patients. About one in four patients in each arm of the study received antibiotics for prophylaxis rather than the treatment of infection and nearly all of these were for surgical and medical procedures.

In keeping with differences in the frequency of COPD according to centre, a greater proportion of the patients in hospitals in CDDFT than ABMUHB were treated for ‘respiratory, thoracic and mediastinal disorders’ (Table 5).

Antibiotic exposure

All of the participants were receiving one or more antibiotics when they started the IMPs. The date the participant began taking the antibiotics before recruitment was known in 1448 participants in the probiotic and 1443 in the placebo arm. The median (IQR) period of exposure to antibiotics before starting the IMP was 3.0 days (2.0–6.0 days) in both study arms (p = 0.38).

During the period 7 days before, and 8 weeks following, recruitment, the most commonly used antibiotic class was the penicillins, with over half of all participants receiving a broad-spectrum penicillin. About one in four participants were exposed to a cephalosporin. Antibiotic exposure was similar in the two study arms (Table 6).

Antibiotic exposure varied according to centre (see Appendix 9, Table 31). In hospitals in CDDFT, exposure to broad-spectrum penicillins was greater than in hospitals in ABMUHB (67.8–70.6% vs. 47.3–57.1%, respectively), but exposure to cephalosporins was lower (1.9–3.3% vs. 13.6–29.1%, respectively), as was exposure to quinolones (6.9–7.5% vs. 8.4–21.2%, respectively).

Fewer than 1 in 10 participants received only a single dose of an antibiotic and most received antibiotics for ≥ 7, with one-third treated for at least 14 days (Tables 7 and 8). The majority of participants were exposed to antibiotics from two or more classes. Exposure to combination therapy and duration of antibiotic therapy was similar in the two study arms.

Non-antibiotic drug treatment

Use of drugs other than antibiotics was common, with many participants receiving antihypertensive therapy, aspirin, PPIs or angiotensin-converting enzyme (ACE) inhibitor therapy. Non-antibiotic drug treatment was similar in the two study arms (Table 9).

Primary outcomes

Antibiotic-associated diarrhoea (including CDD) occurred with a similar frequency in the probiotic arm (159 participants, 10.8%) and placebo arm (153 participants, 10.4%; RR 1.04; 95% CI 0.84 to 1.28; p = 0.71; Table 10). This included 12 participants with frequent stools that they described as looser than normal but who were unable to describe stool consistency using the Bristol Stool Form Scale. 38

Clostridium difficile diarrhoea was uncommon and occurred in 12 (0.8%) participants in the probiotic arm and 17 (1.2%) participants in the placebo arm (RR 0.71; 95% CI 0.34 to 1.47; p = 0.35; see Table 10). Based on this effect size and the low prevalence of CDD, the number needed to treat to prevent one case is 295. This would be reduced to 95 for an effect size at the lower limit of the 95% CI (a threefold reduction in CDD in the probiotic arm). The corresponding number needed to harm (the upper 95% CI) is 267.

Secondary outcomes

Clostridium difficile was isolated from stools in two participants with mild loose stools (not meeting the study criteria for diarrhoea) in the probiotic arm. One participant in each arm had an episode of CDD after an initial episode of AAD that was not associated with CDI; the participant allocated to the placebo arm required surgery for CDD and the participant allocated to the probiotic arm had gallstones and died during the episode of CDD. One patient with known carcinoma of the head of the pancreas with a biliary stent in situ died during an episode of CDD that occurred after withdrawal from the trial.

The adjusted treatment effect on occurrence of AAD from covariate analysis was similar to the unadjusted effect after controlling for nine prespecified covariates. Covariate analysis identified that the occurrence of AAD could be predicted by the duration of antibiotic treatment, antacid therapy and duration of hospital stay (Table 11).

The frequency of AAD was similar in each centre: Morriston 162/1479 (11.0%), Singleton 20/203 (9.9%), Princess of Wales 21/191 (11.0%), Durham 56/547 (10.2%) and Darlington 53/521 (10.2%; p = 0.97). Subgroup analyses showed that the distribution of cases of AAD according to prespecified potential risk factors for AAD, including those identified as risk factors in covariate analysis, was similar in the two intervention arms and there was no evidence of a statistically significant interaction between prespecified potential risk factors for AAD and intervention arm (Table 12).

Most episodes of AAD (73.7%) occurred within 4 weeks of recruitment. On average, episodes of AAD lasted for 2 days with four stools in 24 hours and of consistency seven on the Bristol Stool Form Scale (Table 13). The most commonly associated symptoms were urgency, abdominal pain and nocturnal diarrhoea. The latter tended to occur more frequently in the placebo than the probiotic group (p = 0.051) and other characteristics of the diarrhoea episodes were similar in the two study arms (see Table 13). Most episodes of AAD were managed in hospital and stool samples were collected and tested for diarrhoeal pathogens in 58.6% of all cases. For many episodes of AAD, the short duration and occurrence after discharge from hospital complicated the collection of a stool specimen for testing for pathogens.

As with AAD, the adjusted treatment effect for CDD was similar to the unadjusted estimate (Table 14). Covariate analysis showed that duration of antibiotic treatment was associated with CDD.

The frequency of CDD was similar in each centre: Morriston 21/1479 (1.4%), Singleton 2/203 (1.0%), Princess of Wales 0/191 (0.0%), Durham 3/547 (0.5%) and Darlington 3/521 (0.6%; p = 0.15). In subgroup analysis, there was a statistically significant interaction between intervention arm and age (p = 0.0015; Table 15). In patients aged > 77 years, the frequency of CDD was significantly lower in the probiotic arm than in the placebo arm. In contrast, the frequency of CDD was similar in the two intervention arms for patients aged ≤ 77 years. In addition, the interaction between treatment group and duration of antibiotic treatment was of borderline statistical significance (p = 0.054). There was no evidence of a significant interaction between the intervention arm and other prespecified potential risk factors for CDD (Table 15).

The timing of onset of CDD was similar to that of AAD, with 75.8% cases occurring within 4 weeks of recruitment. On average, the duration of CDD was 6.5 days and duration was similar in the two study arms (Table 16). Bloating was less common in the placebo arm than in the probiotic arm (risk difference 40.7%; 95% CI 7.4% to 74.0%) and median stool frequency tended to be lower in the placebo arm than in the probiotic arm (see Table 16). Otherwise, gastrointestinal symptoms, clinical findings and investigations and classification of severity were similar in the two study arms. During follow-up, no patient was identified as having peritonitis, ileus, toxic megacolon or life-threatening CDD or as having died from CDD. The majority of patients in both study arms were managed in hospital.

Seven (0.5%) participants in the probiotic arm and 10 (0.7%) participants in the placebo arm had diarrhoea due to other causes (RR 0.70; 95% CI 0.27 to 1.84). In the probiotic arm, six had norovirus diarrhoea and one was diagnosed with non-specific colitis. In the placebo arm, six had norovirus diarrhoea, one had diarrhoea after taking laxatives, two patients attributed diarrhoea to drinking a large volume of fruit juice and one had melaena associated with abnormal clotting.

Overall, 2927/2940 (99.6%) participants took at least one dose of the IMP, with a similar proportion in the probiotic (1462/1469, 99.5%) and placebo arms (1465/1471, 99.6%; p = 0.78; compliance unknown for one participant in the probiotic arm). The median number of days that participants were observed or reported taking the IMP in the first 3 weeks was similar in the probiotic [n = 1469, 21 days (IQR 14–21 days)] and placebo arms [n = 1471, 21 days (IQR 14–21 days); p = 0.55; Figure 2]. The full 21-day course was completed by 52.5% of participants. Overall, 1076/2934 (36.7%) participants reported that they disliked taking the IMP, and this proportion was similar in the probiotic (529/1466, 36.1%) and placebo arms (547/1468, 37.3%; p = 0.51). Taking account of compliance in covariate analysis did not materially alter the risk of AAD (OR 1.02; 95% CI 0.80 to 1.30) or CDD (OR 0.66; 95% CI 0.30 to 1.47).

FIGURE 2.

Total number of days participants took the IMPs according to intervention arm.

Unused IMPs were collected opportunistically at three time points during the study, from participants who had withdrawn or died, for assessing correct identity according to active versus placebo and number of viable organisms in the probiotic preparation. Thirty-four probiotic capsules were tested and all contained ≥ 1.62 × 10¹⁰ viable bacteria. All of the 33 placebo capsules tested were sterile.

During the first 3 weeks while participants were taking the IMPs, the duration of hospital stay was similar in the probiotic [n = 1469, median 6 days (IQR 2–13 days)] and placebo arms [n = 1470, median 6 days (IQR 2–13 days); p = 0.65]. The most commonly reported gastrointestinal symptoms were nausea (14.9%), abdominal pain (13.4%) and diarrhoea (any loose stools reported by the participants; 12.3%; Table 17). The frequency of gastrointestinal symptoms was similar in the two study arms with the exception of flatus, which was marginally less common in the placebo than the probiotic arm (risk difference 2.3%; 95% CI 0.0% to 4.6%). Futhermore, although very few participants had a NGT in situ, this was significantly more common in the probiotic than placebo arm. With these two exceptions, the duration that symptoms were present was also similar in the two study arms (p > 0.17 for all comparisons).

There were no statistically significant differences in either the frequency or duration of gastrointestinal symptoms or other morbidity according to study arm during weeks 4–8 of the study (data not shown). Overall, average duration of hospital stay was known in 2899 participants and was similar in the probiotic [n = 1452, median 4 days (IQR 1–11 days)] and placebo arms [n = 1447, median 4 days (IQR 1–11 days); p = 0.87; Figure 3].

FIGURE 3.

Duration of hospital stay according to intervention arm.

Eighteen participants were excluded from PP analysis. Seven patients in the probiotic and six in the placebo arm declined to take any of the IMPs. Investigation of the IMP labelling error that occurred at one centre resulted in the IMPs being withdrawn before completion of the 21-day course in one participant in the probiotic arm and four in the placebo arm. Among these participants excluded from PP analysis, none developed CDD, but one allocated to the probiotic arm developed AAD. Analysis of primary outcomes in the PP population did not materially alter the assessment of the efficacy of the intervention (data not shown). In addition, the risk of developing AAD or CDD was as similar among those participants who took all 21 IMP doses, 14 or more doses or seven or more doses as it was in all participants (data not shown).

Serious adverse events

Serious adverse events were common in the study population with 578 (19.7%) participants experiencing one or more SAE (Table 18). The most common MedDRA SOC classifications for SAEs were respiratory, thoracic and mediastinal disorders, gastrointestinal disorders, and cardiac disorders.

Serious adverse events classified as gastrointestinal disorders occurred in 79 (2.7%) participants with a similar frequency in both study arms (see Table 18). With SAEs classified according to MedDRA PTs43 (see Appendix 9, Table 32), gastrointestinal haemorrhage occurred in 15 participants in the probiotic arm (specified as upper gastrointestinal haemorrhage in four participants and lower in five participants) and 11 participants in the placebo arm (specified as upper gastrointestinal haemorrhage in one participant and lower in seven participants). Peptic ulcer occurred in four participants in the probiotic arm (specified as duodenal ulcer in one participant and perforated peptic ulcer in two participants) and one participant in the placebo arm experienced a perforated duodenal ulcer. Abdominal pain occurred in four participants in the probiotic arm and three in the placebo arm. Gastroenteritis occurred in three participants in the probiotic arm and two in the placebo arm. Constipation occurred in one participant in the probiotic arm and two in the placebo arm. Peritonitis, volvulus and dysentery each occurred in one participant in the probiotic arm and appendix abscess, colostomy performed, diarrhoea, liver abscess and pancreatitis each occurred in one participant in the placebo arm. There was no occurrence of intestinal ischaemia.

The frequency of SAEs that were, or may have been, due to bacterial infection occurred with similar frequency in each arm (see Appendix 9, Table 32). Pneumonia occurred in 52 participants in the probiotic arm (specified as caused by Pseudomonas sp. in two of these participants) and 53 in the placebo arm (specified as caused by Pseudomonas sp. in three of these participants). Abscesses occurred in one participant in the probiotic arm (specifically, a groin abscess) and four participants in the placebo arm (specifically groin, mediastinal, liver and psoas abscesses). Urinary tract infection occurred in 15 participants in the probiotic arm and 12 in the placebo arm, wound infection or cellulitis occurred in 16 participants in the probiotic arm and nine in the placebo arm, infected implant site occurred in two participants in the probiotic arm and one in the placebo arm and infected haematoma occurred in one participant in the probiotic arm. Sepsis occurred in 10 participants in the probiotic and 12 in the placebo arm and organ failure in one participant in each study arm.

The most frequent SAEs classified according to MedDRA PTs43 were pneumonia (3.3%), obstructive pulmonary disorder (1.6%) and falls (1.1%). Overall, 143 (4.9%) participants experienced a SAE that resulted in death, 10 (0.3%) SAEs that were considered to be life-threatening, 447 (15.2%) SAEs that prolonged hospitalisation, four (0.1%) SAEs that resulted in persistent or significant disability or incapacity and 11 (0.4%) experienced other SAEs that were considered to be significant medical events (see Appendix 9, Table 32). The proportion of patients in each SAE severity category, the frequency of individual SAEs within each category and the proportion of participants experiencing one or more SAE were similar in the two study arms. (see Appendix 9, Table 32).

Following the occurrence of a SAE, the patient’s clinical team withdrew the IMP from 14 (0.5%) participants and discontinued them temporarily in 90 (3.1%) participants, with a similar proportion in each study arm (see Appendix 9, Table 33). A common reason for discontinuing the IMPs was to reduce the number of medications for the patient rather than any concern regarding the safety of the bacterial organisms.

Economic analysis

Cost-effectiveness and uncertainty

Base-case analysis showed only a small total health-care cost difference between the probiotic and placebo arms (see Table 25). This was mainly due to the relatively small implementation cost of the probiotic and the marginal cost savings for diarrhoea episodes in the probiotic arm. The cost difference resulted in an ICER of £22,701 per QALY at 1 year with a probability of the intervention being cost-effective at a £20,000 willingness-to-pay threshold of 48%. The CEAC depicts the probability of the intervention being cost-effective at different willingness-to-pay thresholds. (Figure 4).

FIGURE 4.

Cost-effectiveness acceptability curve, base case analysis: total health-care cost.

If the implementation costs of the probiotics only are taken into account, without consideration of any downstream effects, the ICER increases to £189,662 per QALY at 1 year with a probability of cost-effectiveness at £30,000 of 2% (Figure 5). Thus, based on a £30,000 willingness-to-pay threshold and implementation costs, probiotics are not cost-effective.

FIGURE 5.

Cost-effectiveness acceptability curve, base case analysis: probiotics implementation cost.

Results of the cost–consequences analysis are reported in Table 28. As overall differences in costs and clinical outcomes between the two arms were small, the clinical effectiveness and cost-effectiveness of probiotics in the prevention of AAD in this study can be considered limited. Even though probiotics appeared cost-effective in the cost–utility analysis based on total health-care costs, no significant budgetary impact can be anticipated. This is due to the small differences in total cost between the probiotic and placebo arms and the lack of statistical significance in the primary outcomes. Subgroup analysis was not undertaken, as the covariate analysis did not identify any specific population that clearly benefited from receiving the probiotic. Cost per case of diarrhoea averted was not analysed as the study did not demonstrate a difference in diarrhoea frequency between the two groups.

TABLE 28 - Clinical effectiveness and cost-effectiveness outcomes: cost–consequences analysisa

SD, standard deviation.

Values in brackets represent 95% CI unless indicated otherwise.

Impact variable	Probiotic	Placebo	Difference	p-value
Costs impact
Implementation cost per patient (£)	73.02 (70.20 to 75.83)	0.00 (0.00 to 0.00)	73.02 (70.20 to 75.83)	0.01
Total health-care cost per patient (£)	8020.11 (7622.31 to 8417.90)	8011.37 (7600.53 to 8422.22)	8.74 (−4.32 to 21.78)	0.98
Cost of AAD episode (£)	1742.15 (1438.39 to 2045.92)	2220.38 (1696.52 to 2744.23)	−478.23 (−1192.34 to 235.89)	0.12
QoL impact
EQ-5D scores at baseline	0.51 (SD 0.341)	0.51 (SD 0.332)	0.00	–
EQ-5D changes baseline – 4 weeks	0.08 (SD 0.343)	0.08 (SD 0.334)	0.00 (−0.03 to 0.03)	0.74
EQ-5D changes baseline – 8 weeks	0.12 (SD 0.348)	0.11 (SD 0.352)	0.01 (−0.01 to 0.03)	0.45
Health impact
Number of AAD cases per arm	159/1470 (10.8%)	153/1471 (10.4%)	6 (0.4%)	0.72
Number of CDD cases per arm	12/1470 (0.8%)	17/1471 (1.2%)	−5 (−0.4%)	0.35

Sensitivity analysis

Changes in the parameters included in the microcosting of a diarrhoea episode and changes in the cost of an inpatient day (the average cost per day amounted by a patient while in hospital) did not result in significant changes to the difference in overall cost between the probiotic and the placebo arms (Table 29). Furthermore, a decrease or increase in staff time for probiotic administration did not significantly change the cost-effectiveness results (Table 30). Considering probiotic implementation costs only, a reduction in staff time by 50% resulted in an ICER of £107,818 per QALY and a probability of cost-effectiveness at £30,000 of 16%, whereas doubling of staff time increased the ICER to £353,402 per QALY and was associated with a probability of cost-effectiveness at £30,000 of < 1%. The CEACs for these results can be found in Figures 6 and 7.

TABLE 29 - Sensitivity analysis: changes in mean health-care cost per patient following parameter change within defined ranges

Parameter changed	Mean health-care cost per patient
	Probiotic (95% CI)	Placebo (95% CI)	Difference	p-value
Costing of diarrhoea (£)
All diarrhoea costs lower value	7871.70 (7483.57 to 8259.83)	7843.83 (7445.22 to 8242.44)	27.87	0.92
All diarrhoea costs upper value	8095.37 (7694.56 to 8496.19)	8111.94 (7691.43 to 8532.46)	−16.57	0.96
Other costs (£)
Hospital inpatient day lower value	6716.00 (6389.38 to 7042.61)	6699.56 (6359.39 to 7039.73)	16.14	0.95
Hospital inpatient day upper value	9298.27 (8833.97 to 9762.58)	9312.79 (8831.04 to 9794.54)	−14.52	0.97

TABLE 30 - Sensitivity analysis: changes of ICER based on probiotic implementation cost following parameter change within defined ranges

Probiotic implementation cost (£)	Cost per patient (95% CI)	ICER	Probability cost-effective
Mean staff time per dose lower value	41.51 (40.11 to 42.92)	107,818 per QALY	0.046
Mean staff time per dose upper value	136.06 (130.44 to 141.68)	353,402 per QALY	0.00

FIGURE 6.

Cost-effectiveness acceptability curve, sensitivity analysis: probiotic implementation cost when staff cost halved.

FIGURE 7.

Cost-effectiveness acceptability curve, sensitivity analysis: probiotic implementation cost when staff cost doubled.

Summary of cost-effectiveness results

• Cost and duration of hospital stays, cost of diarrhoea, cost of antibiotics and total health-care cost per patient were very similar between the probiotic and the placebo arms. No statistically significant cost differences were found between the two study arms.

• Incremental total health-care cost of participants who suffered from AAD was £4531.36 (95% CI £3439.80 to £5622.92), which was significantly higher than for non-diarrhoea patients, independent of study arm. This was mainly due to increased length of hospital stay and additional diarrhoea-associated costs.

• Duration of hospital stay (initial stay and readmissions), staff time for antibiotic administration and diarrhoea-associated costs were identified as main components of the total health-care costs across both study arms.

• Between baseline and 8 weeks, mean QoL increase was 0.01 QALY higher in the probiotic than in the placebo arm; however, this difference was not statistically significant.

• The ICER associated with probiotic use at 1 year was estimated at £22,701 per QALY gained when total health-care costs were considered and £189,662 per QALY gained considering probiotic implementation costs only. However, the similarity in total cost, number of diarrhoea cases and patient QoL between the two trial arms limits the relevance of the ICERs.

• One-way sensitivity analyses did not show any significant effect on difference in total health-care costs between the trial arms and the overall conclusion of the cost-effectiveness assessment.

Key findings

In an analysis according to treatment allocated, we did not find adequate evidence that a preparation of four strains of live lactobacilli and bifidobacteria was clinically effective in preventing or ameliorating AAD in 2941 inpatients aged between 65.0 and 107.5 years. Overall, AAD occurred in 312 (10.6%) participants, 159 (10.8%) participants in the probiotic arm and 153 (10.4%) participants in the placebo arm (RR 1.04; 95% CI 0.84 to 1.28; p = 0.72). CDD was an uncommon cause of AAD, occurring in 29 (1.0%) participants. CDD occurred less frequently in the probiotic (12 participants; 0.82%) arm than in the placebo arm (17 participants; 1.2%). However, this difference was not statistically significant (RR 0.71; 95% CI 0.34 to 1.47; p = 0.35). Analysis adjusting for potential risk factors did not significantly change the findings [OR for AAD 1.0 (95% CI 0.78 to 1.27); OR for CDD 0.65 (95% CI 0.29 to 1.47)]. Secondary outcome measures including the severity of diarrhoea, frequency and duration of gastrointestinal symptoms, duration of hospital stay, and QoL were also generally similar in the two study arms. Although an episode of AAD increased total health-care costs by £4531.36 (95% CI £3439.80 to £5622.92) independent of study arm, this was not reflected in a difference in total health-care cost between the probiotic and placebo arms. Cost-effectiveness analysis based on probiotic implementation costs showed that the probiotic was not cost-effective with an ICER of £189,662 per QALY. This result was robust to changes in the key parameters.

Comparison with other studies/reviews

Strengths and weaknesses of probiotic lactobacilli and bifidobacteria in antibiotic-associated diarrhoea and Clostridium difficile diarrhoea in the elderly

Clinical implications

• The high-dose, multistrain preparation of lactobacilli and bifidobacteria evaluated in our study is unlikely to benefit unselected older inpatients exposed to antibiotics.

• Clinical judgement regarding the benefits and risks of novel interventions to prevent AAD needs to take account of the impact of other preventative measures, such as antibiotic stewardship, on disease frequency.

• The administration of additional medications to vulnerable older people, many of whom are already taking multiple medications, may not be well tolerated in practice.

• The clinical effectiveness of our preparation in preventing CDD was unclear. However, even if it is effective, the falling prevalence of CDD needs more patients to take the probiotic to prevent a single case.

• The probiotic preparation was not associated with SAEs in our study. However, surveillance for potentially uncommon adverse events is required in future studies.

Research implications

• A better understanding of the multiple potential mechanisms underlying AAD and CDD, how these may vary in specific populations and the strain-specific effects of probiotics is needed before further clinical trials of specific probiotic preparations are undertaken. Further research to identify populations at increased risk of AAD and CDD is needed to facilitate the future evaluation of probiotic interventions.

• The design of studies to evaluate the efficacy of alternative probiotics in the prevention of CDD needs to consider the effect of other measures that have reduced the frequency of CDI in some health-care institutions.

• Further research into the effect of probiotics on patient QoL will be necessary to better determine patient benefit and cost-effectiveness.

Contributions of authors

Professor Stephen J Allen wrote the original protocol, was the chief investigator, oversaw the study, wrote the initial draft of the main report and contributed to the final report.

Ms Kathie Wareham was a co-applicant and refined the protocol, was the trial manager and contributed to the final report.

Dr Duolao Wang was a co-applicant and refined the protocol, wrote the statistical analysis plan and undertook the statistical analysis and contributed to the final report.

Mrs Caroline Bradley was a co-applicant and refined the protocol and contributed to the final report.

Dr Bernadette Sewell designed the economic analysis, wrote the initial draft of the economics analysis and contributed to the final report.

Ms Hayley Hutchings wrote the initial draft of the QoL analysis and contributed to the final report.

Dr Wyn Harris was a co-applicant and refined the protocol and contributed to the final report.

Dr Anjan Dhar was a co-applicant and refined the protocol and contributed to the final report.

Dr Helga Brown was a co-applicant and refined the protocol and contributed to the final report.

Dr Alwyn Foden was a co-applicant and refined the protocol and contributed to the final report.

Professor Mike B Gravenor was a co-applicant and refined the protocol, contributed data analysis and interpretation of the statistics and contributed to the final report.

Professor Dietrich Mack was a co-applicant and refined the protocol and contributed to the final report.

Professor Ceri J Phillips was a co-applicant and refined the protocol, designed the economic analysis and contributed to the final report.

Trial Steering Committee members

Professor Stephen Bain (Chairperson), Chairperson in Medicine (Diabetes), College of Medicine, Swansea University; Diabetes Lead Clinician, ABMUHB.

Dr John Sloss, Consultant Microbiologist, Darlington Memorial Hospital, Hollyhurst Road, Darlington, County Durham.

Mr Graham Tanner, Royal British Legion welfare case worker.

Data monitoring and Ethics Committee members

Professor John Williams (Chairperson), Professor of Health Services Research, College of Medicine, Swansea University and Consultant Gastroenterologist, ABMUHB.

Dr Duolao Wang; Senior Lecturer in Medical Statistics, Department of Medical Statistics, London School of Hygiene and Tropical Medicine, London, UK. Independent Statistician from study start to January 2012.

Professor Kerenza Hood, Director, South East Wales Trials Unit, Institute of Translation, Innovation, Methodology and Engagement, Cardiff University School of Medicine, Neuadd Meirionnydd, Heath Park, Cardiff, UK. Independent Statistician from January 2012.

Dr John Brazier, Anaerobe Reference Laboratory, Cardiff, UK, retired.

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. 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.

Patient information sheet

Patient consent form

Patient consent form (PDF download)

Patient advocate information sheet

Patient advocate assent form

Patient advocate assent form (PDF download)

Case report form

Case report form (PDF download)

Daily follow-up log

Daily follow-up log (PDF download)

Weekly follow-up log

Weekly follow-up log (PDF download)

Additional antibiotics form

Additional antibiotics form (PDF download)

Severe adverse events form

Severe adverse events form (PDF download)

Diarrhoea sheet

Diarrhoea sheet (PDF download)

Classification of severity of Clostridium difficile diarrhoea

Classification of severity of Clostridium difficile diarrhoea (PDF download)

Notes

(a) The patient or next of kin is approached for consent in the afternoon if verbal and written information about the trial is provided in the morning, or the following day if provided in the afternoon; and (b) either 21 capsules of probiotic or placebo.

Mild

• normal white blood cell count (WCC)

• stool frequency less than three per day and

• stool consistency type 5–7 on Bristol Stool Form Scale. 38,39

Moderate

• WCC raised but < 15 × 10⁹/l and

• stool frequency three to five per day. 39

Severe

• WCC raised but > 15 × 10⁹/l or

• an acute rising serum creatinine (> 50% increase above baseline) or

• temperature > 38.5 °C or

• evidence of severe colitis (based on clinical examination or imaging). 39

Life-threatening

• hypotension or

• partial or complete ileus or

• toxic megacolon or

• radiological evidence of severe disease. 39