Probiotics to reduce antibiotic administration in care home residents aged 65 years and older: the PRINCESS RCT

Article history

The research reported in this issue of the journal was funded by the EME programme as project number 13/95/10. The contractual start date was in September 2015. The final report began editorial review in November 2019 and was accepted for publication in June 2020. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The EME editors and production house have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the final report document. However, they do not accept liability for damages or losses arising from material published in this report.

Copyright statement

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

2021 Butler et al.

Rigorous research is needed to underpin improved care for care home residents

As the population ages, care homes are becoming an increasingly common and important sector for delivering health and social care. Novel interventions that aim to enhance care and quality of life should not be taken up in routine care without efficacy first being established. Thus far, there have been few randomised controlled trials (RCTs) of interventions that aim to improve the quality of care for care home residents (CHRs). As a result, far more is known about effective treatments in hospitals than about what works most effectively to improve care and the experience for older people living in care homes. 2

Because immunity wanes with increasing age and frailty, CHRs are more prone to infections and account for a high volume of antibiotic use. Antibiotic use drives antimicrobial resistance (AMR). Regular probiotics, so-called ‘good bacteria’, have been proposed as a feasible and safe intervention that may reduce the incidence of infections and, thus, antibiotic use. However, regular probiotics for reducing infections and antibiotic use have so far not been evaluated in a rigorous, long-term RCT among CHRs. Furthermore, there are gaps in our understanding of how probiotics might work. Therefore, we evaluated the efficacy and possible mechanism of probiotics in CHRs to prevent infections and thus reduce antibiotic use.

What are probiotics and how might they work?

Probiotics are defined by the World Health Organization as ‘live microorganisms which when administered in adequate amounts, confer a health benefit to the host’. 3 They are present in certain products available in supermarkets as foodstuffs and in formulations used for specific therapeutic purposes. They may prevent infection by blocking pathogenic colonisation and enhancing gut–immune interaction, with influence on mucosal and systemic immunity, leading to enhanced natural killer (NK) cell activity and vaccine response in older people. 4 The gastrointestinal tract may be a major reservoir for antimicrobial-resistant bacteria. This is important because many urinary tract infections (UTIs) are thought to arise from auto-inoculation with organisms from the gut. 5 As people age, the microbiological diversity in their gut reduces because they are less mobile and adopt different eating patterns and their immune systems become less active. By increasing the diversity of microorganisms in the gut, taking probiotics regularly may reduce the prevalence of pathogenic bacteria and stimulate the immune system, adding to overall resilience and general well-being as well as reducing infections and antibiotic use. Ageing is independently associated with reduced immune response to infections. 6

Are probiotics safe and feasible for care home residents?

Although probiotics carry theoretical risks of causing infection beyond the gut and transferring antibiotic-resistant genes, there have been no reports of bacteraemia or fungaemia attributable to the probiotics when used in trials,7–9 and only a very few case reports in patients with immunodeficiency or severe ulceration of the colonic mucosa. 10 Gastrointestinal side effects and rash are generally no more common in patients on probiotics than in those on placebo. 11 Systematic reviews found no serious adverse effects in participants of trials of probiotics for antibiotic-associated diarrhoea (AAD). Some older people already use them regularly, despite an inadequate evidence base supporting their effect on common infectious diseases. Adverse events (AEs) in trials of probiotics have generally not been well reported. 12

However, there is emerging evidence that probiotics after antibiotic use may delay the reconstitution of the faecal and mucosal microbiome, because soluble factors secreted from the probiotic species could inhibit human microbiome growth. 13 Inhibition by probiotics of reconstitution of the human microbiome may be a mechanism that explains a surprise finding from one retrospective hospital-based cohort study of 5209 patients for whom prior probiotic exposure was an independent predictor of Clostridium difficile infection [odds ratio (OR) 1.39, 95% confidence interval (CI) 1.08 to 1.80]. 13

Probiotics are feasible to administer to CHRs in the course of routine care: Carlsson et al. 14 confirmed the feasibility of serving a probiotic intervention for 6 months to people with dementia in care homes. The authors found that the intervention was easy to serve alongside usual diet, there were few side effects and staff were able to complete the processes and measures. In contrast to antibiotic use, long-term probiotic use does not result in resistance in commensal gut organisms. 15

Existing trials and the research gap

Small trials found benefits from probiotic colonisation by multidrug-resistant Gram-negative organisms in critically ill patients16 and by vancomycin-resistant enterococci (VRE) in renal dialysis patients. 17 A Cochrane review18 of probiotics to prevent acute upper respiratory tract infections (URTIs) included 13 RCTs, 10 of which were meta-analysed (including a total of 3720 participants). Probiotics reduced episodes of acute URTI (OR 0.53, 95% CI 0.37 to 0.76), duration of episodes (mean difference –1.89 days, 95% CI –2.03 to –1.75 days; p < 0.001, low-quality evidence) and antibiotic prescribing (OR 0.65, 95% CI 0.45 to 0.94). Side effects of probiotics were minor. The review noted poor allocation concealment, and insufficient heterogeneity in several studies, and recommended that future RCTs should include older people, as few were included in the trials reviewed.

King et al. 19 systematically reviewed studies of probiotics to reduce antibiotic use for common acute infections. The authors included 17 RCTs, all conducted in infants and children. The mean duration of probiotic supplementation ranged from 4 days to 9 months. Children who received probiotics had a lower risk of being prescribed antibiotics than those receiving placebo (pooled relative risk 0.71, 95% CI 0.54 to 0.94); however, trial quality was rated as being variable. The authors recommended evaluating probiotics to reduce antibiotic use in other population groups.

Regarding older adults, Turchet et al. 20 performed a single-centre pilot study of otherwise well ‘free-living’ older people, randomised to receive a probiotic containing Lactobacillus casei or placebo for 3 weeks. The authors found no differences in frequency of URTI in this short study, but did find a reduction in the severity and duration of URTIs. Guillemard et al. 21 randomised otherwise well people living in care homes in France to a probiotic containing L. casei or placebo for 3 months. Probiotics decreased the duration of common infectious diseases, especially URTIs. Makino et al. 22 in Japan considered whether or not the intake of yogurt fermented with L. delbrueckii over 12 weeks had an effect on the common cold, and found that the risk was 2.6 times lower (OR 0.39) and that there was increased NK cell activity in the yoghurt group. However, randomisation and concealment may not have been adequate in this study. Van Puyenbroeck et al. 23 found no difference in the duration of respiratory symptoms or the probability of respiratory symptoms in ‘healthy older people’ in nursing homes in Antwerp from a probiotic containing L. casei Shirota for 176 days. However, people with relevant medical conditions and those with cognitive deficits were excluded, and these people may be most likely to benefit meaningfully from regular probiotics. Antibiotic prescription rates were not reported. The large number of missing self-complete diary data generated analytical challenges.

Thus, overall, the evidence base supporting a recommendation for frailer elderly people either to consume or to not consume probiotics is insufficient to robustly guide care. New trials that are properly designed in terms of allocation and concealment, that focus on antibiotic use for all-cause infections in CHRs (including those without capacity and who are most frail) and with reliable, frequent external ascertainment and with longer-term follow-up are therefore warranted.

Rationale for the PRINCESS trial

Infections are a common cause of suffering for and increased resource use by CHRs. Even so-called ‘minor infections’ such as UTIs and URTIs can have an important negative impact on the health, well-being and dignity of older frail people. The 2010 Adult Social Care Statistics reported that there were 229,900 people in residential care in England, with numbers predicted to steadily increase. 24 The 2011 census reported that there were 291,000 people over the age of 65 years resident in the care home population in England and Wales. 24

A year-long observational study (n = 274) identified 609 infections that led to an antibiotic prescription incidence of 2.16 antibiotic prescriptions per resident year (95% CI 1.90 to 2.46). The most common indications were respiratory tract infections (RTIs) (47% of prescriptions), UTIs (29%) and skin infections (18%). 25 Common infectious diseases in CHRs led to suffering, loss of dignity, hospitalisation, general practitioner (GP) visits and death. A health needs assessment of 240 CHRs in South Tyneside Primary Care Trust (now called South Tyneside Clinical Commissioning Group) identified 167 hospital admissions accounting for 1595 bed-days over 1 year. 26 On average, at least four beds in the acute trust were used for CHRs all year round, costing around £400,000. About 25% of admissions were due to infection. Common infectious diseases in CHRs increase GP and care home burden and costs, affecting opportunities for other aspects of care. Antibiotic prescribing leads to health-care-associated infections, such as C. difficile infection, and drives AMR. Infection with an antimicrobial-resistant organism is commonly associated with recent consumption of antibiotics, even after controlling for age, comorbidity and other risk factors. 27 Antimicrobial-resistant infections are more serious, last longer and are costlier to manage. 28,29 Probiotics have been found to be effective in preventing AAD,9,11,30 although a recent large trial31 found no benefit. However, there are no adequate data on probiotics for the prevention of all-cause common infections and antibiotic prescribing in CHRs.

Health-care-associated infections include common infectious diseases and cause significant debility, hospital admissions and the death of CHRs, burdening both the health service and care home staff. Among CHRs in Norway, the risk of deterioration in general physical condition was twice as high those who developed health-care-associated infections than in matched controls who did not have health-care-associated infection, and the former were nine times more likely to be admitted to hospital. 32

Carriage of antimicrobial-resistant bowel organisms increases the chances of an antimicrobial-resistant UTI. Antimicrobial-resistant Gram-negative septicaemia and antimicrobial-resistant UTI are on the increase in the community,33 especially among older people. 34 Care homes are a reservoir for antimicrobial-resistant organisms that cycle between the community and hospitals. 35,36

There have been calls for ‘diligently planned large-scale randomised and blinded trials, preferably devoid of commercial interests’ with end points being objectively assessed. 37

Rationale for the probiotics selected for evaluation in the PRINCESS trial

The effects of probiotics are thought to vary by strain because of differing resistance to gastric acid and bile, ability to colonise mucosa and susceptibility to antibiotics. The probiotic combination selected for this study comprised two organisms, Lactobacillus rhamnosus GG and Bifidobacterium animalis subsp. lactis BB-12 Lactobacillus rhamnosus GG has been extensively studied in a wide variety of populations including older individuals. 38 This strain has been found to be safe. Although the effect of this probiotic strain on infections in CHRs or older people has not yet been assessed, relevant evidence proving efficacy in other populations supports its use in this trial. A meta-analysis of RCTs involving 1805 children found that Lactobacillus rhamnosus GG was associated with a reduced risk of otitis media, URTIs and antibiotic treatments. 39 This meta-analysis included a double-blind, placebo-controlled RCT of 742 hospitalised children that found that Lactobacillus rhamnosus GG reduced the risk of gastrointestinal infections (GIs) [risk ratio (RR) 0.40, 95% CI 0.25 to 0.70] and RTIs (RR 0.38, 95% CI 0.18 to 0.85), and the duration of these infections. 40 A RCT of 281 children in day care found that Lactobacillus rhamnosus GG reduced the risk of RTIs over 3 months (RR 0.66, 95% CI 0.52 to 8.82, number needed to treat = 5) and reduced the time with a RTI. 41 A double-blind RCT found that Lactobacillus rhamnosus GG acts as an immune adjuvant to influenza vaccination, as measured by levels of protective antibodies to the H3N2 flu strain. This study stressed the need for future studies of probiotics as immune adjuvants, focusing on groups known to have a poor response to influenza vaccination. 42 When administered orally as lozenges, a combination of probiotics has been found to have a beneficial effect on oral and dental health, reducing plaque and gingival inflammation without affecting oral microbiota. 43

Many studies include more than one probiotic strain in the intervention, and, in several studies, Lactobacillus rhamnosus GG and Bifidobacterium animalis subsp. lactis BB-12 have been included in the same probiotic formulation. 44,45 A randomised placebo-controlled trial of 231 college students taking a combination probiotic including Bifidobacterium animalis subsp. lactis BB-12 for 12 weeks found that the duration and severity of URTIs were reduced by the active probiotic combination. 46 A randomised, placebo-controlled, double-blind, parallel-group dose–response study investigated the impact of 4-week commercial yoghurt consumption supplemented with Bifidobacterium animalis subsp. lactis BB-12. The probiotic strain remained active during gut transit and was associated with an increase in beneficial bacteria and a reduction in potentially pathogenic bacteria in faeces. 47 In a double-blind, placebo-controlled study, 109 1-month-old infants receiving Bifidobacterium animalis subsp. lactis BB-12 were reported to have experienced fewer respiratory infections (65% vs. 94%, RR 0.69, 95% CI 0.53 to 0.89; p = 0.014) than the control infants. 48 A randomised, double-blind, placebo-controlled study of infant formula supplemented with the probiotics Lactobacillus rhamnosus GG and Bifidobacterium animalis subsp. lactis BB-12 or placebo was administered daily until the age of 12 months. Those infants receiving the active probiotic had fewer episodes of otitis media, received fewer courses of antibiotics in the course of routine care and had fewer recurrent RTIs. 49 A multispecies probiotic supplementation that included Lactobacillus rhamnosus GG and Bifidobacterium animalis subsp. lactis BB-12 reduced symptoms of irritable bowel syndrome and stabilised the bowel microbiota. 50 A randomised, double-blind, placebo-controlled trial in 12 wards in two nursing homes in Finland involved 209 CHRs who received (1) 10⁹ colony-forming units (CFUs) per day of B. longum strains or (2) 10⁹ CFU per day of Bifidobacterium animalis subsp. lactis BB-12 or (3) placebo for 7 months. There was 85% compliance, and the groups receiving active products had more frequent bowel movements than the placebo group. 51 In a 7-week double-blind crossover study, 36 patients were randomised to receiving yoghurt with Bifidobacterium animalis subsp. lactis BB-12 or placebo. Bifidobacterium animalis subsp. lactis BB-12 was safe. During the Bifidobacterium animalis subsp. lactis BB-12 intake period, the defecation frequency increased compared with the no-intake period for the group, and comfort after defecation improved significantly. 52

Thus, neither proposed probiotic strain has ever been evaluated for the purposes intended in the PRINCESS (Probiotics to Reduce Infections iN CarE home reSidentS) trial and there is evidence from other population groups that these strains have the effect of reducing frequency, severity and duration of infections and stabilising bowel flora. Many infections in older people result in auto-inoculation from the bowel. It is possible that the two strains complement each other in terms of efficacy as a result of different modes of action.

Objectives of the PRINCESS trial

The primary objective of the PRINCESS trial was to evaluate the effect of a dose of daily oral probiotics on cumulative systemic antibiotic administration days for all-cause, acute infections.

Secondary objectives included determining the effect of daily probiotic intake on subcategories of infections, incidence and duration of diarrhoea in CHRs who are being treated with oral antibiotics, incidence and duration of infections, prevalence of C. difficile infection, health utility, well-being, hospitalisation, mortality, gastrointestinal colonisation with antimicrobial-resistant bacteria, and Candida from oral samples.

Mechanistic immunology outcomes aimed to explore the effect of daily probiotic combination intake on influenza vaccine response, cytokine and chemokines, monocyte and neutrophils, probiotic in stool samples, and influence of baseline vitamin D on estimates of AMR.

Trial design

The PRINCESS trial was an individually randomised, double-blind placebo-controlled trial of the clinical efficacy and safety of a combination of two probiotic organisms given daily to CHRs in reducing cumulative antibiotic administration days (CAAD). Trial participants were CHRs from 23 care homes in Wales and England. Following enrolment, and on completion of a baseline assessment, participants were randomised to receive a capsule containing either the probiotic combination (intervention) or a formulation of maltodextrin, microcrystalline cellulose, magnesium stearate and silicon dioxide (placebo control) to be taken once daily for 52 weeks.

An internal pilot was conducted in four care homes (two in Wales and two in England) between December 2016 and February 2017. The aims of the internal pilot included assessing levels of interest expressed by care homes in implementing the trial, estimating the likely CHR recruitment rate, establishing that it would be possible to ascertain adherence to the study product (SP) (probiotic or placebo) and measuring study outcomes for most of the participants. The internal pilot also allowed us to develop a mechanism for capturing antibiotic use while participants were away from the care home or in hospital. A hospital discharge case report form (CRF) was developed for collection of hospital data after the internal pilot and was used in most cases. No other trial procedures were changed as a result of the internal pilot phase.

The trial protocol has been published1 and amendments made to the trial protocol during the conduct of the trial are listed in Appendix 1.

Setting and participants

Eligibility

Trial interventions

The active and placebo capsules were manufactured by Chr. Hansen A/S (Hørsholm, Denmark) and supplied to the trial team at no cost to the study. Boxes of active SP and boxes of placebo SP were sent from Chr. Hansen directly to Production Services and Clinical Trials, Nottingham University Hospitals NHS Trust, for labelling with trial information and pack identifiers. The boxes were then transported to the Centre for Trials Research, Cardiff University (CTR), for long-term storage in a temperature-monitored fridge. Because the SP was stable at room temperature for 2 years, some packs were stored at room temperature at the CTR before posting to study sites.

During the trial, owing to expiry dates, three batches of SP were sent from Chr. Hansen to Nottingham University Hospitals NHS Trust for labelling (the last two batches were from the same manufacture run).

The SP capsules contained either Lactobacillus rhamnosus GG and Bifidobacterium animalis subsp. lactis BB-12 or placebo (containing maltodextrin, microcrystalline cellulose, magnesium stearate and silicon dioxide). The dates of manufacture of SP were December 2015 (total cell count per capsule 1.6 × 10¹⁰) and March 2017 (total cell count per capsule 1.36 × 10¹⁰).

There were 31 capsules per tube. SP capsules could be swallowed whole, opened and sprinkled onto warm or cold (but not hot) food or dissolved in warm or cold (but not hot) liquid. Before any use of the probiotics, a sample probiotic and placebo were labelled and sent for analysis. Each capsule was dissolved in 10 ml of sterile water then further diluted 10^–3 and 10^–6. All three solutions (50 µl) were spiral plated onto two plates each of Columbia blood agar (CBA) and fastidious anaerobe agar (FAA) (both from Thermo Fisher Scientific Oxoid Ltd, Basingstoke, UK) and incubated at 35 ± 1 °C aerobically and in carbon dioxide (CBA plates) anaerobically (FAA plates) for 24–48 hours. All colony types were identified using matrix-assisted laser desorption/ionisation time of flight (MALDI-TOF) mass spectrometry and individual counts were made.

Data collection and randomisation

Data were collected at the care home, using information gathered from the participant, the participant’s relatives, the care home staff, hospital discharge summaries and the participant’s care home records. Data were collected either on paper CRFs or electronically via a Structured Query Language (SQL) online database, which was hosted on a secure password-protected server at the CTR. All data collected on paper were later entered electronically to the database either by the RNs remotely or at the CTR. Data were added to the database using a secure tablet, a laptop or a PC. A trial-specific Data Management Plan was developed outlining in detail the procedures that were put in place to ensure that high-quality data were produced for statistical analysis. The database had in-built validation rules and ranges checks, with incomplete fields and data outliers flagged at the time of entry. Missing data and validation errors were queried periodically with RNs, as necessary. A list of self-evident corrections was iterated during the trial to enable correction of unambiguous data entry errors. Data collected on paper held or returned to the central monitoring site (i.e. CTR) were subject to quality control checks against data held on the database (10% of data collected on paper). All data, including sensitive and personal data, were handled in accordance with the Data Protection Act 199860 and then the General Data Protection Regulation 2016. 61

On completion of data cleaning, the data were extracted from the database and provided to the statistician for analysis. As per Cardiff University’s procedures, data will be retained for 15 years following trial closure.

Participant tracking through the trial was assisted through database-generated automated e-mails, which were sent when data collection for a particular time point was complete. A summary of data collection by time point is provided in Table 1.

Follow-up data collection

All baseline and outcome assessments were conducted by people blind to study allocation. Participants were visited each week by the study RNs from the date of randomisation for data collection.

Biological samples

During the enrolment process, consent (or consultee advice) to collect blood, saliva and stool samples was sought. Stool samples were cultured to assess accuracy of randomisation and confirm adherence to the probiotic combination. Two mechanistic substudies requiring blood samples were embedded within the main trial: an immunology substudy and an influenza vaccine substudy. Separate consent was sought for the influenza vaccine substudy (on the main consent form). All samples were voluntary, and the participant could still take part in the trial if they did not wish to have any samples taken.

Blood samples were collected during the baseline review. Participants who were part of the immunology substudy were asked for a repeat blood sample at the 12-month (or final) review. Blood samples for the influenza substudy were scheduled around the participant’s influenza vaccine. Saliva and stool samples were collected during the baseline review and at the 3- and 12-month (or truncated) reviews. Samples taken during the baseline review were taken before randomisation and SP initiation.

Results from sample investigations were not used for routine health monitoring.

Outcome measures

A table of all objectives and outcome measures is shown in Table 2.

Primary outcome

The trial primary outcome measure was CAAD for all-cause infections. This was ascertained from the total number of days of systemic antibiotic administration as recorded in care home medical records and discharge summaries if the participant was admitted to hospital, collected retrospectively by the RNs during weekly visits to CHRs.

Secondary outcomes

Secondary outcomes were investigated, under the following headings.

Tertiary outcomes/aims

• To determine if the level of serum vitamin D correlated with colonisation of antimicrobial-resistant bacteria in faecal isolates.

• Qualitative substudy to understand how the PRINCESS trial was implemented in order to identify opportunities for improving processes that might be relevant for future, similar studies.

Collection of adverse events

An extensive review of published literature and product dossiers provided by Chr. Hansen A/S, the SP manufacturer, led to the conclusion that probiotics, including the strains used in the PRINCESS trial, are generally well tolerated by elderly people and the risks to CHRs associated with the probiotics used in the PRINCESS trial were very low. A review by Doron and Snydman70 summarised the risks. Infections (e.g. septicaemia) caused by the probiotic bacteria had not been reported in trials, and theoretical risks of gene transfer of antibiotic resistance and overstimulation of the immune response had not been reported. Although deleterious metabolic activities (e.g. bowel ischaemia and D-lactic acidosis in patients with pancreatitis) had been reported, the mode of probiotic delivery was different from that in the PRINCESS trial. Therefore, if any of the above events were to occur in the PRINCESS trial, they would have been unexpected.

Minor gastrointestinal symptoms, such as abdominal cramping, nausea, soft stools, flatulence (passing wind) and taste disturbance, had been reported. It is also possible that a minor allergic reaction could have occurred. If these events were to have occurred, they would be classed as expected AEs, but would be classed as unexpected if they fulfilled the definition of serious. Therefore, the patient information sheet stated that side effects from probiotics are very rare and have never been serious, but that flatulence and bloating have been reported to occur in a minority of people.

The study population was expected to have a vast number of health events in the normal course of their care at their stage of life. Given the potential frailty of the trial population and the high incidence of hospitalisation and death in the course of routine care, we aimed to not cloud any true emerging safety profile by collection of unrelated data; the AE reporting procedure was developed to reflect this. AEs that were both (1) not serious and (2) unlikely to be related to the SP or study procedures (e.g. when a participant had a minor fall that did not require hospitalisation) were not recorded.

Non-serious AEs that were, or could be, related to either the SP or the study procedures were recorded on the weekly record form and weekly further information CRF. The AE was described in a free-text box on the CRF.

Serious adverse events (SAEs) were recorded on the weekly record in the further information section of the CRF. This required the RN who was completing the form to consider if any SAEs were related to the study intervention or study procedures and subsequently indicated this using the causality classification (definitely, probably, possibly, unlikely, not related). All SAEs were discussed (in person, by telephone or by e-mail) by the RN with a second reviewer (another PRINCESS trial RN or a PRINCESS trial clinician involved in the study and trained to do this task) to confirm the causality classification (definitely, probably, possibly, unlikely, not related). Details of the second reviewer, date and classification were recorded on the PRINCESS trial weekly record in the further information section of the CRF. Where there was a difference in classification between the two reviewers, the highest category of causality (most likely to be related) was selected. If the main and second reviewer classified the AE as definitely or probably related to study participation or study intervention, a serious adverse reaction (SAR) form was completed. This SAR form was then passed to a clinical reviewer, usually the trial’s chief investigator, who then allocated expectedness. If the SAE was classified as unexpected (and definitely or probably related to study procedure or study intervention), this would require further reporting to the sponsor, the REC and the Independent Data Monitoring Committee. Unblinding would occur if it was classified as related to the study intervention.

If the CHR died, the cause of death was sought. If the CHR was admitted to hospital, confirmation of infection and any routine microbiology results were sought. SAEs that were deemed to be not related to the SP or procedures did not require any further action and were managed in accordance with the care home’s routine procedures. These SAEs were reported annually in the safety report to the Independent Data Monitoring Committee.

Statistical methods

Recruitment

We screened 1309 CHRs, of whom 332 were consented into the study. Consultee involvement was required for 220 residents (66.3% of those consented). Baseline data collection was completed for 318 CHRs (seven died and a further seven withdrew between consent and completion of baseline data collection), and a further eight were not randomised (following completion of baseline data collection but prior to randomisation, two CHRs died and six withdrew completely from the study). This is reported in the Consolidated Standards of Reporting Trials (CONSORT) trial flow diagram (Figure 1).

FIGURE 1.

A CONSORT flow diagram. a, Participants withdrew for clinical reasons; b, probiotic arm: four participants withdrew for clinical reasons, two moved care home and in one case the reason for withdrawal was unknown. Placebo arm: five withdrew for clinical reasons, one withdrew as a result of a participant/family request and one moved care home; c, completion of the second follow-up assessments occurred between 144 and 364 days post randomisation and depended on the date of randomisation. The median second review day is the same in both arms (placebo arm: median 319, IQR 219–356; probiotic arm: median 321, IQR 244.5–355.5). Fifty-three participants in each arm had at least one follow-up truncated; and d, probiotic arm: one participant withdrew for clinical reasons, two reasons were not known and one participant had moved care home. Placebo arm: five participants withdrew for clinical reasons, in one case the reason for withdrawal was not known and one participant had moved care home.

Baseline characteristics

Baseline stool characteristics

In total, 159 baseline stool samples were received at the laboratory (51.3% of randomised participants), 76 from participants allocated to placebo (49% of participants in this group) and 83 from those allocated to probiotic combination (53.5% of participants in this group). Owing to the method of collecting stool samples in a care home, the RNs were not always able to be present at the collection of samples; therefore, the ‘Universal Container with spoon’ was often left in the care home for the care home staff to collect a sample at a convenient time. This may have contributed to the lower collection rate.

All but one of the stool samples received was processed (one specimen leaked in transit), and there was growth on the non-selective growth medium from all the samples processed. A total of 402 bacterial isolates were cultured from 158 stool samples (i.e. some stool samples contained more than one type of bacteria, e.g. E. coli and Klebsiella pneumoniae), with 189 cultured from 75 samples from participants in the placebo arm and 213 cultured from 83 samples from participants in the probiotic combination arm. The mean number of species cultured was 2.5 (SD 0.88) per sample in the placebo arm and 2.6 (SD 0.93) per sample in the probiotic combination arm. A total of 29 species were cultured (Figure 2).

FIGURE 2.

Bacterial species cultured from stool samples at baseline.

Enterobacterales isolates were cultured from 157 samples (99.4%), and resistance to at least one of the tested antimicrobials was found in 119 of these (75.8%), 57 (77.0%) from participants in the placebo arm and 62 (74.7%) from participants in the probiotic combination arm. AMR in E. coli isolated from baseline samples can be seen in Figure 3.

FIGURE 3.

Antimicrobial susceptibility of E. coli cultured from stool samples at baseline (n = 154).

Three samples were confirmed positive for VRE at baseline.

Sixty-eight samples were found to contain bacteria that grew on ESBL-selective media, suggesting resistance to third-generation cephalosporins. Thirty-one of these were confirmed to harbour an ESBL, with the rest harbouring a native AmpC enzyme conferring third-generation cephalosporin resistance. All ESBL-positive bacteria were identified as E. coli. These enzymes are predominantly carried on plasmids and so are able to transfer to different species; when causing infections, they precipitate treatment with a carbapenem.

Seven samples contained isolates that grew on CRE plates, suggesting carbapenem resistance, mainly Pseudomonas aeruginosa; however, no transferable resistance determinant was confirmed (i.e. carbapenemase).

Primary outcome analysis

Care home residents randomised to probiotic combination had a mean CAAD of 12.9 days (SE 1.49 days) and CHRs randomised to placebo had a mean CAAD of 12.0 days (SE 1.50 days). The adjusted CAAD rate in those allocated to probiotic combination compared with placebo was 1.13 days (adjusted IRR 1.13, 95% CI 0.79 to 1.63; p = 0.495) (Table 7).

Secondary outcomes analyses

There was evidence that CHRs allocated to probiotic combination were administered more antibiotics for lower respiratory tract infections (LRTIs) (adjusted IRR 1.43, 95% CI 1.05 to 1.93; p = 0.022). We found no evidence of differences between trial arms in antibiotic use for UTIs, URTIs and skin infections (Table 8). No evidence of between-arm differences was found for duration of infection, the number of infection days (Table 9, and see Appendix 5, Figures 19 and 20) or any of our prespecified diarrhoea outcomes (see Appendix 5, Table 32, and Appendix 5, Figure 21).

There was a difference between trial arms in self-reported generic well-being/capability measured using the ICECAP-O scale at the 3-month follow-up, with CHRs allocated to the probiotic combination arm reporting an ICECAP-O value 0.06 points lower (i.e. worse well-being) than those allocated to the placebo arm (adjusted mean difference –0.06 points, 95% CI –0.11 to –0.001 points; p = 0.047). There was no evidence of any between-arm differences for other self- and proxy-reported well-being and quality-of-life outcomes (Tables 10 and 11).

Table 12 describes the findings for hospital stay and death. There was no evidence of any difference between trial arms in terms of CHRs being hospitalised at least once during the post-randomisation study period (adjusted OR 1.25, 95% CI 0.74 to 2.11; p = 0.407), number of hospital stays (adjusted IRR 1.17, 95% CI 0.72 to 1.90; p = 0.530), cumulative number of hospital days (adjusted IRR 1.00, 95% CI 0.43 to 2.29; p = 0.997), death (adjusted OR 1.03, 95% CI 0.59 to 1.80; p = 0.904) or time from randomisation to death (hazard ratio 0.99, 95% CI 0.67 to 1.46; p = 0.958).

Secondary microbiology outcomes analyses

Stool analysis

At 3 months post randomisation, 108 stool samples were received at the laboratory (34.8% of randomised participants), 52 from CHRs allocated to placebo (33.5% of participants in this group) and 56 from those allocated to the probiotic combination (36.1% of participants in this group). All samples were processed and there was growth on the UTI selective agar plate in all instances.

A total of 288 isolates were cultured from 108 stool samples: 139 isolates were cultured from 52 samples from participants in the placebo arm and 149 isolates were cultured from 56 samples from participants in the probiotic combination arm. The mean number of isolates cultured per sample was 2.67 in the placebo arm and 2.66 in the probiotic arm. The five most common named isolates that were cultured, accounting for 208 (72.2%), were E. coli, P. aeruginosa, Enterococcus faecalis, E. faecium and K. pneumoniae. In total, 34 different organisms were cultured (Figure 4).

FIGURE 4.

Organisms cultured from stool samples at 3 months post randomisation.

Enterobacterales isolates were cultured from 107 samples (99.1%), and resistance to at least one of the tested antimicrobials was found in 76 of these (71.0%), 39 from participants in the placebo arm (75.0% of participants in this group) and 37 from participants in the probiotic combination arm (67.3% of participants in this group). AMR in E. coli at 3 months can be seen in Figure 5.

FIGURE 5.

Antimicrobial susceptibility of E. coli cultured from stool samples at 3 months (n = 106).

Three samples were confirmed positive for the presence of VRE at 3 months.

At 3 months, 45 samples were found to contain isolates (mainly E. coli and K. pneumoniae) with resistance to third-generation cephalosporins, 16 of which harboured an ESBL-producing bacterium.

Six samples contained carbapenem-resistant isolates, mainly P. aeruginosa; however, no transferable resistance determinant was confirmed (i.e. carbapenemase gene).

At the second follow-up time point, 67 stool samples were received at the laboratory (21.6% of randomised participants), 30 from CHRs allocated to placebo (19.4% of participants in this group) and 37 from those allocated to probiotic combination (23.9% of participants in this group). All but one of the samples were processed (one specimen was too scant) and there was growth on the UTI selective agar plate in all samples processed.

A total of 180 isolates were cultured from 66 stool samples: 75 isolates were cultured from 29 samples from CHRs in the placebo arm and 105 isolates were cultured from 37 samples from CHRs in the probiotic combination arm. The mean number of isolates per sample was 2.59 in the placebo arm and 2.84 in the probiotic arm. The five most common named isolates that were cultured, accounting for 120 (66.7%) isolates were E. coli, E. faecalis, P. aeruginosa, E. faecium and Proteus mirabilis. In total, 30 species were cultured (Figure 6).

FIGURE 6.

Organisms cultured from stool samples at the second follow-up time point.

Enterobacterales were cultured from 65 samples (98.5%), of which 60 were tested for resistance to one or more antimicrobials. Antimicrobial-resistant Enterobacterales isolates were found in 42 (70.0%) samples, of which 19 (70.4%) were from CHRs allocated to placebo and 23 (69.7%) were from those allocated to the probiotic combination. AMR in E. coli isolates at the second follow-up time point can be seen in Figure 7.

FIGURE 7.

Antimicrobial susceptibility of E. coli cultured from stool samples at the second follow-up time point (n = 58).

Similar to baseline and 3 months, three samples were confirmed positive for VRE at the second follow-up time point. Only one patient was positive for VRE at baseline, and at 3 and 12 months.

There was no evidence of differences between the trial arms in the presence of Enterobacterales isolates resistant to at least one of the tested antimicrobials at either 3 months or the second follow-up time point (Table 13).

TABLE 13 - Between-arm differences for stool microbiology outcomes

Ratio is probiotic arm/placebo arm. Adjusted for CHR sex. Clustering of CHRs within care homes was accounted for by fitting a two-level logistic regression model (analysis at 3 months was based on 107 participants within 17 care homes; and the analysis at the second follow-up time point was based on 60 participants within 13 care homes). Baseline Enterobacterales were not included in the models as this resulted in > 10% of participants being removed from the analysis as a result of missing data.

Outcome	Time point	Trial arm						Adjusted ORa	95% CI		p-value
		Placebo			Probiotic
		Number of CHRs	Frequency	%	Number of CHRs	Frequency	%		Lower limit	Upper limit
Enterobacterales in stools	3 months	52	52	100.0	55	56	98.2	Not analysable
	Second follow-up	29	29	100.0	36	37	97.3	Not analysable
Enterobacterales in stools resistant to at least one of the tested antimicrobials	3 months	39	52	75.0	37	55	67.3	0.61	0.24	1.56	0.303
	Second follow-up	19	27	70.0	23	33	69.7	0.76	0.20	2.89	0.683
VRE in stools	3 months	0	0	0.0	3	3	100.0	Not analysable
	Second follow-up	0	0	0.0	3	3	100.0	Not analysable

At 3 months post randomisation, 7 out of 107 samples tested (6.5%) were found to contain C. difficile, the majority from CHRs who were randomised to the probiotic combination [6/55 (10.9%) vs. 1/52 (1.9%) in the placebo trial arm; adjusted OR 6.51, 95% CI 0.75 to 56.57; p = 0.090]. At the second follow-up, 2 out of 64 (3.1%) of the samples tested contained C. difficile. Both samples were from CHRs randomised to the probiotic combination.

Stool culture for study probiotic combination organisms

Of the 125 samples that resulted in growth on the Lactobacillus selection agar (125/158, 79.1%), 58 were from CHRs randomised to placebo. The number of samples found to contain L. rhamnosus 3 months post randomisation was higher in the probiotic combination group than in the placebo group [47/56 (83.9%) vs. 19/52 (36.5%); adjusted OR 9.19, 95% CI 3.51 to 24.07; p < 0.001]. This finding persisted at the second follow-up time point [27/37 (73.0%) vs. 9/29 (31.0%); adjusted OR 6.41, 95% CI 2.14 to 19.20; p = 0.001].

Similarly, B. animalis subsp. lactis was cultured more frequently from samples from CHRs randomised to the probiotic combination than from those randomised to placebo at both 3 months [29/56 (51.8%) vs. 2/52 (3.8%); adjusted OR 26.90, 95% CI 5.94 to 121.66; p < 0.001] and the second follow-up time point [21/37 (56.8%) vs. 2/29 (6.9%); adjusted OR 21.96, 95% CI 2.97 to 162.43; p = 0.002]. Study probiotic combination organisms were cultured from stool samples from those CHRs randomised to the probiotic combination intervention in all study sites and were checked against random allocation codes after the database lock.

At both 3 months and the second follow-up time point, the mean colony counts (‘weight of growth’) for both L. rhamnosus and B. animalis subsp. lactis were higher in CHRs allocated to the probiotic combination arm than in those allocated to placebo (Tables 14 and 15). Colonisation with these probiotic organisms is not uncommon; therefore, we expected to see some of the placebo group with these organisms, but the numbers of patients in the placebo trial arm with positive colonisation were small.

TABLE 14 - Presence of L. rhamnosus GG in stools by trial arm

The denominators are based on all samples (bacterial counts = 0 for those with no growth). The denominators for baseline were: placebo = 75, probiotic = 83. The denominators for 3 months were as follows: placebo = 52, probiotic = 56. The denominators for the second follow-up were as follows: placebo = 29, probiotic = 37.

L. rhamnosus GG in stoolsa	Trial arm		Total
	Placebo	Probiotic
Time point
Baseline
Samples with no growth, n (%)	56/75 (74.7)	55/83 (66.3)	111/158 (70.3)
Bacterial counts (CFU/ml), mean (SD)	14,427 (52,666)	20,020 (77,864)	17,365 (66,946)
Bacterial counts (CFU/ml), median (IQR)	0 (0–10)	0 (0–270)	0 (0–180)
3 months
Samples with no growth, n (%)	33/52 (63.5)	9/56 (16.1)	42/108 (38.9)
Bacterial counts (CFU/ml), mean (SD)	46,729 (2,772,633.46)	704,478 (3,048,311)	387,784 (2,218,564)
Bacterial counts (CFU/ml), median (IQR)	0 (0–2100)	5550 (240–130,000)	0 (680–26,000)
Second follow-up
Samples with no growth, n (%)	20/29 (69.0)	10/37 (27.0)	30/66 (45.5)
Bacterial counts (CFU/ml), mean (SD)	13,970 (43,139)	151,527 (526,728)	91,086 (39,899)
Bacterial counts (CFU/ml), median (IQR)	0 (0–100)	29,000 (0–100,000)	240 (0–40,000)

TABLE 15 - Presence of B. animalis subsp. lactis BB-12 in stools by trial arm

The denominators are based on all samples (bacterial counts = 0 for those with no growth). The denominators for baseline were: placebo = 75, probiotic = 83 The denominators for 3 months were as follows: placebo = 52, probiotic = 56. The denominators for the second follow-up were as follows: placebo = 29, probiotic = 37.

B. animalis subsp. lactis in stoolsa	Trial arm		Total
	Placebo	Probiotic
Time point
Baseline
Samples with no growth, n (%)	71/75 (97.7)	80/83 (96.4)	151/158 (95.6)
Bacterial counts (CFU/ml), mean (SD)	17,800 (106,559)	2922 (18,773)	9984 (74,777)
Bacterial counts (CFU/ml), median (IQR)	0 (0–0)	0 (0–0)	0 (0–0)
3 months
Samples with no growth, n (%)	50/52 (96.2)	27/56 (48.2)	77/108 (71.3)
Bacterial counts (CFU/ml), mean (SD)	28,846 (170,738)	1,717,750 (5,109,484)	904,574 (3,761,923)
Bacterial counts (CFU/ml), median (IQR)	0 (0–0)	5000 (0–415,000)	0 (0–27,500)
Second follow-up
Sample with no growth, n (%)	27/29 (93.1)	16/37 (43.2)	43/66 (65.2)
Bacterial counts (CFU/ml), mean (SD)	362 (1856)	214,824 (444,510)	120,591 (347,763)
Bacterial counts (CFU/ml), median (IQR)	0 (0–0)	2500 (0–180,000)	0 (0–80,000)

Saliva analysis

At 3 months post randomisation, 218 saliva samples were received at the laboratory (from 70.3% of randomised CHRs), 105 from participants allocated to placebo (67.7% of participants in this group) and 113 from those allocated to the probiotic combination (72.9% of participants in this group). All but two of the saliva samples received were processed (one snapped and consent was not given for the other).

A total of 235 isolates were cultured from 168 saliva samples: 119 isolates from cultured from 80 samples from participants in the placebo arm and 116 isolates were cultured from 88 samples from participants in the probiotic combination arm. The five most common named organisms that were cultured, accounting for 229 (97.4%) isolates, were C. albicans, C. glabrata, C. dubliniensis, C. tropicalis and C. parapsilosis. In total, nine different organisms were cultured (Figure 8).

FIGURE 8.

Candida species detected from saliva samples of CHRs at 3 months post randomisation.

At the second follow-up time point, 161 saliva samples were received at the laboratory (from 51.9% of randomised participants), 76 from CHRs allocated to placebo (49.0% of participants in this group) and 85 from those allocated to the probiotic combination (54.8% of participants in this group). All saliva samples received were processed.

A total of 179 organisms were cultured from 127 saliva samples: 119 organisms were cultured from 57 samples from CHRs in the placebo arm and 116 organisms were cultured from 70 samples from CHRs in the probiotic combination arm. The four most common named organisms that were cultured, accounting for 172 (96.1%) of isolates, were C. albicans, C. glabrata, C. dubliniensis and C. tropicalis. In total, nine different organisms were cultured (see Appendix 5, Figure 22).

There was no evidence of any between-arm differences in terms of presence or amount of candidiasis from oral samples at either follow-up time point (Tables 16 and 17).

TABLE 16 - Between-arm differences for the presence of candidiasis from oral samples

Ratio is probiotic arm/placebo arm. Adjusted for CHR sex. The clustering of CHRs within care homes is accounted for by fitting a two-level logistic regression model (the analysis at 3 months was based on 199 participants in 22 care homes; and the analysis at the second follow-up time point was based on 150 participants in 20 care homes).

Outcome	Trial arm						Adjusted ORa	95% CI	p-value
	Placebo			Probiotic
	Number of CHRs	Frequency	%	Number of CHRs	Frequency	%
Presence of candidiasis from oral samples
3 months	105	80	76.2	113	88	77.9	1.23	0.54 to 2.83	0.623
Second follow-up	76	57	75.0	85	70	82.4	1.27	0.50 to 3.21	0.620

TABLE 17 - Between-arm differences for the levels of Candida from oral samples

+/–, scanty growth; +, light growth; ++, moderate growth; +++, heavy growth.

Ratio is probiotic arm/placebo arm. Adjusted for CHR sex. The clustering of organisms within CHRs within care homes is accounted for by fitting a three-level ordinal regression model (the analysis at 3 months was based on 174 organisms in 138 participants in 22 care homes; and the analysis at the second follow-up time point was based on 131 organisms in 103 participants in 20 care homes).

Outcome	Trial arm								Adjusted ORa	95% CI	p-value
	Placebo				Probiotic
	Number of CHRs	Weight	Number of organisms	%	Number of CHRs	Weight	Number of organisms	%
Amount of candidiasis present in saliva
3 months	80	(–/+)	20	16.8	88	(–/+)	21	18.1	0.66	0.20 to 2.17	0.489
		(+)	19	16.0		(+)	26	22.4
		(++)	38	31.9		(++)	20	17.2
		(+++)	42	35.3		(+++)	49	42.2
Second follow-up	57	(–/+)	14	17.3	70	(–/+)	15	15.3	0.50	0.12 to 2.16	0.356
		(+)	11	13.6		(+)	11	11.2
		(++)	23	28.4		(++)	31	31.6
		(+++)	33	40.7		(+++)	41	41.8

Adverse events

Data about AEs were collected during the study period following randomisation. During this period, 283 events in 120 CHRs were recorded. A total of 60.8% of events were hospitalisations (172/283); in 22.9% of events, CHRs had died (65/283) (Table 24). Three AEs were trial related and occurred in the placebo arm: one participant experienced an increase in difficulty swallowing and the speech and language team recommended that the SP should be discontinued because of a choking risk; one participant experienced frequent episodes of diarrhoea that the care home staff felt had been worse since he started the SP; and one participant reported feeling bloated.

EuroQol-5 Dimensions

Self- and proxy-reported EQ-5D-5L data were collected at baseline, 3 months and the second follow-up. The individual domains, split by time point and trial arms, are provided in Appendix 5 (see Tables 34–36 and Figures 23–25). The findings appear to be largely consistent across the domains of the EQ-5D-5L.

Background

Common chronic conditions are linked to an altered immune functionality, which comprises excessive inflammation and declined immunoregulatory and immunostimulatory responses. These changes have been described as biomarkers of ageing, but tend to be heterogeneous between individuals. 72,73 The reduced intestinal microbial diversity in many older people compared with healthy young individuals, as well as healthy old individuals, indicates that gut microbiota might be a target to decrease the progressive immune decline74 and, subsequently, the rate of infections in CHRs. Immunophenotypes provide useful information about functionality of B cells, which have been linked to a decreased antibody production and, therefore, poor response towards influenza vaccination.

Altering the gut microbiota may have the potential to reduce frailty in the aged because it may modulate different aspects of innate75,76 and acquired immunity,77,78 including inflammation. It is therefore relevant to examine strategies tailored towards modifications of gut bacteria in CHRs to improve gut microbial diversity in order to enhance the immune response to exposure to pathogens. In terms of quantifying the immune system, there is a clear consensus around the parameters of the FBC and reference values suggestive of ongoing clinical conditions such as infection. With regard to more detailed immune phenotyping, there are guidelines and a general agreement that age-related alterations include a decreased de novo generation of naive T and B cells, and accumulation of memory cells. 9 In addition, there are many individual immune components and responses that can be measured,11,12 usually by sampling blood, although it is acknowledged that none of these individual immune assessments represent the breadth of the immune response. It is recognised that response to vaccination represents an integrated readout of the ability of the immune system to respond to challenge in vivo. 11,12

Research on the effects of probiotics in CHRs is currently lacking and, hence, we aimed to perform detailed mechanistic analysis of participants’ immune responses to daily probiotics versus placebo in the PRINCESS trial. These findings have been described using FBCs, blood immunophenotypes and various ex vivo immune measures in the trial population. The null hypotheses for our mechanistic analysis were that there would be no difference between groups in FBCs, immunophenotypes, influenza vaccine response, ex vivo cytokine response to TLR agonists, and haematology and biochemistry. The alternative hypothesis was that there would be a difference in these outcomes between the two trial groups.

Aims

The objectives of the mechanistic analyses were to determine the effect of daily probiotic intake on:

• influenza vaccine response

• immunology measures (ex vivo responses to immune challenge)

• haematology and biochemistry measures (FBC) –

  • neutrophils, lymphocytes, monocytes, white blood cells (WBCs), eosinophils, basophils identified through a full blood cell count

  • lymphocyte and monocyte subsets identified through immune cell phenotyping.

Methods

The trial design and randomisation procedures were described in Chapter 2. The mechanistic analyses included all participants from the main PRINCESS trial who consented to provide blood samples at baseline and at the end of the trial. We aimed to obtain 150 samples at baseline with a view to achieving between 70 and 100 pairs of samples (allowing for attrition and participant refusal, moving out of care home, etc.). Participants from the main PRINCESS trial were characterised at baseline, and at the end of the study samples were paired according to data availability at both baseline and the end of the intervention period. Influenza vaccine efficacy analysis aimed to include all participants who consented to provide blood samples (or whose consultees had provided agreement on their behalf) and had been taking the SP for at least 2 months prior to receiving the seasonal influenza vaccination.

The schedule of study procedures is reiterated below:

• screening – screening was conducted prior to recruitment to assess participant eligibility and obtain consent

• baseline – a baseline blood sample was taken following consent but prior to randomisation

• second follow-up – blood sample collected at least 6 and up to 12 months post randomisation.

Owing to slower than anticipated recruitment, we truncated the second face-to-face follow-up for some participants. The minimum period until the second face-to-face follow-up was 6 months post randomisation. The time at which this second follow-up visit took place was accounted for in any analysis involving this time point. The majority of analyses were conducted for all participants with available data provided they initiated SP.

The influenza vaccine substudy analysis was conducted for those who consented to blood sample collection (or whose consultees had provided agreement on their behalf), had had a blood sample collected on the day of their routine seasonal influenza vaccination (or up to 10 days prior to this) and another collected 4 weeks (28 days) after they had received their vaccination.

Participants who were enrolled in this study received the most recent recommended seasonal influenza vaccine for the year 2017. The vaccines for the 2017–18 northern hemisphere consisted of the strains A/Michigan/45/2015 (H1N1)pdm09-like virus, A/Hong Kong/4801/2014 (H3N2)-like virus and B/Brisbane/60/2008-like virus. A quadrivalent product was to be used in this study; therefore, the vaccine included B/Phuket/3073/2013.

The blood sample handling, processing and analysis methods are detailed elsewhere. 79

Results

A CONSORT flow diagram (Figure 9) depicts the progress of the immunology trial participants through to the end of trial sample collection (obtained for 60 participants). Thirty participants were randomised to the daily oral probiotic combination group and 30 participants to the placebo group. Preliminary analyses testing the influence of delayed processing on the parameters examined in blood indicated the maximum time frame in which these parameters can be analysed for each variable under assessment. To control this, the time of sample collection directly from the participant was recorded and posted with relevant information. This time was noted when the sample was delivered to the specialised laboratory, and analyses were carried out taking account of the time delays. Immune mediators measured in plasma and in supernatants from whole-cell cultures were analysed up to 24 hours after sample collection. Other than those parameters, the great majority of parameters could be analysed up to 72 hours after sample collection, with the exception of specific variables within the FBC, immune phenotypes and phagocytic function, in which analyses had to be conducted within 48 hours after collection (see Figure 9).

FIGURE 9.

A CONSORT flow diagram for the immunology substudy.

Where paired samples for a patient were not available because one or both of the samples arrived at the specialised laboratory outside the predetermined time, the patient was not included in the data set for any analyses affected by the time delay. The analyses conducted are shown in Figure 10.

FIGURE 10.

Analyses conducted in the immunology substudy. ImMed, immune mediators; IMPh, immunophenotypes; PhFx, phagocytic function.

Paired samples, representing sample availability at both baseline and post intervention for a patient, were available for analysis for 30 patients.

Characterisation of participants

A total of 184 participants were randomised and recruited to the immunology substudy at baseline. The mean age was 85.4 years (SD ±7.45 years). The general characteristics of study participants at baseline are shown in Table 25 and frailty characteristics are depicted in Figure 11.

TABLE 25 - Characteristics of participants recruited at baseline

Variable	Mean	SE	SD
Age (years)	85.4	0.6	7.5
Time at care home (years)	1.9	0.2	2.2
Height (cm)	1.6	0.0	0.1
Weight at study baseline (kg)	66.6	1.4	15.7
Middle upper arm circumference (cm)	27.3	0.3	3.9

FIGURE 11.

Relative distribution of participants according to frailty status. The scale is based on the frailty score, ranked as follows: 1 = fittest category for their age (active and energetic); 2 = well (absent symptomatology of disease but less active); 3 = managing well (medical problems under control but not active); 4 = vulnerable (symptoms that limit activities); 5 = mildly frail (impairment of daily activities); 6 = moderately frail (progressive impairment and decline in activities); and 7 = severely frail (completely dependent cognitively or physically; not terminally ill).

Anti-vaccine antibody response (vaccine strain-specific response)

A total of 39 participants were recruited into the anti-vaccine antibody response work during the 2017/18 vaccination schedule (19 participants were randomised to daily oral probiotic and 20 participants were randomised to placebo).

Antibody titres at baseline (pre-influenza vaccination)

Influenza vaccine-specific antibodies were measured in serum. Protection against the viruses is analysed through two outcomes: seroprotection and seroconversion. Seroprotection is defined as an antibody titre ≥ 40 haemagglutination units. At baseline (i.e. pre-influenza vaccination), a large proportion of the population was already seroprotected. The antibody titres for the strain A/Michigan/45/2015 indicated that 41% of the population were seroprotected before the commencement of the study. The proportion of the population seroprotected against the strain A/Hong Kong/4801/2014 was 76.9%. The quadrivalent vaccine included the strains B/Brisbane/60/2008 and B/Phuket/3073/2013, for which the percentages of seroprotection were 94.9% and 84.6%, respectively. Seroprotection status before receiving the quadrivalent version of the influenza vaccine is reported in Figure 12.

FIGURE 12.

Pre-vaccination seroprotection status for quadrivalent influenza vaccine in elderly participants in the PRINCESS immunology substudy.

Titres status and seroprotection post vaccination

No statistical differences were observed in the anti-influenza titres when the analyses from the placebo and probiotic combination groups were compared (see Appendix 5, Figures 26–29).

As expected, there was an increase in the proportion of participants seroprotected, as measured by the presence of anti-vaccine antibody titres in the post-vaccination period. Descriptively, it was observed that the proportion of participants who were seroprotected regardless of allocation of SP (either placebo or probiotic combination) increased. There was an increase in the percentage of seroprotection of 28.2% for the strain A/Michigan/2015, 12.8% for the strain A/Hong Kong/2014, 2.5% for the strain B/Brisbane/2008 and 12.5% for the strain B/Phuket/2013, as described by the frequency of cases presented in Table 29.

TABLE 29 - Seroprotection status in the period pre and post vaccination

Quadrivalent influenza vaccine strain	Seroprotection status	Time period
		Pre vaccination (n)	Post vaccination (n)
HAI1_A/Michigan/2015	No	23	12
	Yes	16	27
	Total	39
HAI1_A/Hong Kong/2014	No	9	4
	Yes	30	35
	Total	39
HAI1_B/Brisbane/2008	No	2	1
	Yes	37	38
	Total	39
HAI1_B/Phuket/2013	No	6	1
	Yes	33	38
	Total	39

Subsequently, a detailed analysis was conducted of the proportion of seroprotected participants in the post-vaccine period according to group (either placebo or probiotic combination). It was found that the probiotic did not exert any statistically significant effect on the antibody response to the virus strains used in the quadrivalent influenza vaccine. It was observed that the percentage of positive seroprotection was higher, but not significantly so, in the probiotic combination group than in the placebo group for the strains A/Michigan/2015 and A/Hong Kong/2014. Figure 13 illustrates the percentage of seroprotection by group.

FIGURE 13.

Anti-vaccine antibody seroprotective response according to strain of analysis and allocation group (placebo or probiotic combination).

Subjects were defined as experiencing seroconversion when their post-vaccine sample demonstrated a fourfold increase in antibody titres compared with their pre-vaccine sample. The results of the seroconversion assessment are shown in Figure 14 and are presented as titres of antibodies in the post-vaccine period, analysed per strain composing the quadrivalent influenza vaccine. It was found that the probiotic intervention did not exert a significant effect on the seroconversion rate compared with the placebo combination group.

FIGURE 14.

Anti-vaccine antibody seroconversion response according to strain of analysis and allocation group (placebo or probiotic combination).

Discussion

There was no evidence to suggest that the effect of a probiotic combination consisting of Lactobacillus rhamnosus GG and Bifidobacterium animalis subsp. lactis BB-12 on blood immune cell numbers or subtypes, or on responses to seasonal influenza vaccination, was different from that of placebo. These results are in keeping with the clinical outcomes of the trial, which showed no evidence of an effect of the probiotics on antibiotic consumption or on the duration or number of infections. Systematic reviews have indicated positive effects of other strains of probiotics in healthy older people14 and, hence, our findings may not be applicable to other populations and probiotic species. Limitations include high losses in the follow-up period, which significantly reduced available data and, hence, statistical power. The relatively raised mortality rate, which is comparable to other studies in older participants, contributed to data loss. 80 Moreover, this multicentre study was exposed to other variables affecting the sample size, including that the clinical settings (care homes) were a considerable distance from the laboratory where immunological and inflammatory analyses would take place. This meant that the time between sample collection and arrival at the laboratory was variable and, when considerable delays occurred, it was not possible to carry out experiments on the sample. This significantly affected the data available for immune mediators, cytokine production and chemokines assessed in this study.

Aim

The aim of the qualitative evaluation was to achieve a deep understanding of how the PRINCESS trial was implemented in order to identify aspects that went well and opportunities for improving processes that might be relevant for future, similar studies.

Methods

Results

Semistructured interviews were carried out with nine RNs who implemented the trial in care homes across England and south Wales, seven CHMs from care homes that participated in the trial from across England and south Wales and two relatives who had a parent that participated in the trial. The interviews were carried out between February and May 2019. The interviews lasted between 13 to 71 minutes. Regrettably, one of the audio-recordings of an interview with a CHM was corrupted and could not be used in the analysis; therefore, only six of the seven CHM interviews were included in the analysis. Of the six interviews with CHMs included in the analysis, three were from larger (≥ 50 beds) homes providing residential and nursing care and three were from moderate-sized (20–40 beds) homes providing residential care. Two of the homes (both residential care homes) were privately owned, standalone companies. The remaining four homes were all part of privately owned care home groups. Of the three homes providing residential and nursing care, two were identified as higher recruiters (≥ 12 participants recruited), with one identified as a lower recruiter (fewer than five participants). Two of the three homes providing residential care were moderate recruiters (between 5 and 12 participants) and one was a higher recruiter. Care home size did not appear to be associated with numbers of participants recruited.

The key themes identified through the framework analysis are summarised in Table 30.

To provide context, quotations are captioned by interviewee group [i.e. RN, CHM or relative of a trial participant (relative)].

Qualitative evaluation summary

Research nurses, CHMs and relatives of CHRs perceived the trial intervention as very low risk, with important potential benefits. The data presented here echo the findings of previous research, that is that the acceptability of research studies (including interventional trials) with CHRs is dependent on the research being perceived as low risk and absolutely relevant to the care home population. 56 The PRINCESS trial was implemented in a setting where CHRs were closely monitored by the care staff and RNs, which made the involvement of vulnerable older people acceptable. From the perspective of CHRs’ relatives, any intervention that had the potential to reduce the frequency and impact of infections in their loved one was highly desirable. Despite there being an evidence gap concerning the effectiveness of the probiotics used in the PRINCESS trial to prevent all-cause common infections and antibiotic prescribing in this population,1 there was an expectation from some interviewees that participating in the trial would provide those randomised to the active SP with access to a beneficial substance. 86 Therefore, the potential impact of the intervention was likely to have played a key role in the acceptability of the trial with this population. 86

Despite the perceived low risk of the intervention, the most frequently discussed barrier, particularly regarding those without capacity, was taking the SP capsule alongside their regular medication. This feedback was not surprising, given the level of polypharmacy in CHRs with complex needs. 87–89 There was a need to discuss this particular aspect of the trial with the CHR, relatives and staff at the care home during the screening process. There were some concerns (from CHRs and relatives) that the addition of the SP may have a negative impact on health or well-being. An unintentional consequence of this may have been that CHRs with less complex health needs were identified as being suitable to take part in the trial. RNs and CHMs reported that, once CHRs with capacity enrolled on the trial, they generally did not have any issues taking the SP with their other medication.

Implementing the recruitment process was far more administratively complex when CHRs lacked the capacity to provide consent. Engaging with consultees by letter was difficult to monitor effectively and RNs felt more comfortable if they were able to speak face to face with the relative about the trial. Posters and group meetings with relatives were effective methods of engaging those who attended their relative on a more regular basis.

The consultee model of asking a relative to provide advice on behalf of the CHRs was viewed as more acceptable than approaching a professional (i.e. a staff member from the care home) to act as a nominated consultee; it was important that the consultee had known the CHRs before they lost capacity. This is somewhat in keeping with previous exploratory studies around the acceptability of utilising the advice of personal consultees for CHRs who have lost capacity. 56 ‘Gatekeeping’ by practitioners in control of recruitment (via the rejection of one of the approved approaches to recruitment) is a common research phenomenon, particularly when accessing more vulnerable populations. 90–92 This inauspicious view of the nominated consultee enrolment route does, however, leave those CHRs without family unable to participate in research, and risks further underserving an already under-represented population. 93

Interviewees observed that relatives acting as potential consultees utilised a ‘best interest’ perspective (e.g. they felt that participation might unsettle their relative) to inform their decision to give advice for or against participation, or felt bound by a moral imperative for or against participation. 94 The latter basis for decision-making was evidenced when consultees spoke about the type of person their relative was before they lost capacity, indicating, for example, that they disliked taking any unnecessary medication (aligned with advising against trial participation) or that ‘this was right up Dad’s street’ (aligned with advising for participation). The Mental Capacity Act 200558 requires consultees’ decisions to be based on the person’s wishes and feelings about participation; however, decision-making is more complex, as reported by the consultees in this study.

The ability to build relationships and gain entry to care homes, and thus undertake research activities, was heavily facilitated by the leadership, motivation and presence of the CHM. From a research perspective, CHMs are often described as ‘gatekeepers’ of the care home environment. The trial held a great deal of value from the perspectives of the CHMs and, thus, RNs were valued and welcomed into the care home setting. In addition, there was some evidence that consultation between the CHM and care home staff about the trial, and communication around expectations of their required involvement, promoted entry, RN–staff relationships and trial progress. 57 However, despite high levels of CHM and staff motivation, and a good working relationship between the CHMs and RNs, it could be difficult to maintain or sustain research activities when the RNs were absent from the home for longer periods of time (e.g. staff on annual leave or leaving their role). Therefore, it is likely that RN support, and regular care home engagement and presence, is still required when undertaking trials like PRINCESS in care homes. 95

Although not reported by all RNs interviewed, there was a perception by some that the presence and involvement of nursing staff (i.e. staff with qualifications in nursing recognised by the Nursing and Midwifery Council in the UK) in the home better facilitated the gathering of valid information about the CHR and the reliable undertaking of particular tasks (e.g. recording SP adherence, questionnaire completion). Nurses working in the care home context were also contrasted with ‘carers’ (i.e. staff trained in providing CHRs’ personal care) with regard to their perceived confidence, how busy they were and how much they valued the importance of the research. This contrast was not particularly mentioned by CHMs or relatives of participating CHRs. It is possible that there were some qualitative differences between the homes where nursing staff were more involved in the research than carers; however, without an interhome comparison of data quality and quantity, it is difficult to determine if this is indeed one of the many variables that can predict good-quality data collection. 95 Literature around how carers, nursing staff and research professionals work together to undertake research in long-term care facilities is still in its infancy, particularly from the perspective of care assistants, so there is little with which to compare this study’s findings. However, variables such as previous care home experience, relationships with CHRs, interstaff relationships and staff inclusion do appear to play a part in work-related identities and relationships, and may influence the implementation of research activities in care homes. 96,97

Conclusions

To our knowledge, this is the first qualitative exploration of a double-blind, randomised, placebo-controlled trial carried out in the care home context. The acceptability of the trial, and the readiness to enrol CHRs without capacity, was justified by the low risk and vigilance of the care home setting. The implementation of a placebo-controlled trial in care homes was facilitated by a present and available CHM or clinical lead who demonstrated strong leadership by taking responsibility for trial activities and demonstrated a robust appreciation of the compound benefits that the trial could have for the residents of their home. Nevertheless, conducting this placebo-controlled trial with the care home population undoubtedly required the sustained engagement and activity of trained researchers to set up and maintain research activities in the homes. PRINCESS trial researchers were trained health professionals and had a good understanding and awareness of how care homes function. Their professional identity and experience in a caring profession facilitated the development of good care home staff–researcher relationships, which was key in the effective implementation of the trial in the care home setting.

Summary of clinical efficacy findings

The PRINCESS trial found that administration of a daily dose of the probiotics Lactobacillus rhamnosus GG and Bifidobacterium animalis subsp. lactis BB-12 did not reduce CAAD for all-cause, acute infections in CHRs. All secondary outcomes were consistent with the main finding in that we found no evidence of a beneficial effect of the probiotic combination we studied on antibiotic use overall or on the main categories of common infections that affect the population we studied.

Similarly, the intervention did not favourably affect duration of infections and hospitalisations, death, AAD, immune function measurements or health status, capability and quality of life. Our study population consisted of CHRs in the UK with an average age of > 85 years; two-thirds of participants were women and two-thirds of participants lacked mental capacity to consent. Approximately 21% of participants died during follow-up. We found that those randomised to the probiotic combination arm were administered antibiotics on 10% more days than those in the placebo arm. A 10% difference was prespecified as a clinically meaningful difference, but was not statistically significant, because the power of the study was lower than anticipated as a result of fewer events overall than predicted in our power calculations.

Mechanistic findings

In the mechanistic substudies, we found no evidence that the probiotic combination we studied had a significant effect on blood immune cell numbers or subtypes, or on responses to seasonal influenza vaccination.

Qualitative evaluation

The qualitative interviews found that conducting a randomised placebo-controlled trial in a care home was acceptable to care home staff, family members of CHRs and RNs. The implementation of the trial was facilitated by engaged CHMs; however, the sustained involvement of trained researchers to set up and maintain research activities in the homes was required. The relationship between the RNs and care homes was seen as key to the effective implementation of the trial in the care home setting.

Strengths and limitations

Studies of probiotics often produce contradictory findings and have been criticised for poor design, selective reporting and selective inclusion in systematic reviews, poorly described and verified outcomes, inadequate reporting of harms, and reliance on surrogate outcomes. 98

The PRINCESS trial was a double-blind, placebo-controlled trial that recruited successfully to its revised power target.

Randomisation using minimisation according to care home and CHR sex resulted in comparable intervention groups in terms of medical history (including previous antibiotic use); demographic, haematological and microbiological parameters; health status; capability; and quality-of-life measures (i.e. very similar). To maximise the generalisability of the findings of this trial to the wider care home population, an inclusive approach to recruitment was adopted. It is important that populations included in clinical trials should adequately represent the population relevant to the proposed clinical practice94,99 and, thus, adults lacking capacity to provide consent to participate in clinical trials for themselves should be included in relevant health research, unless there are good scientific reasons justifying their exclusion. 53

There is no established minimal clinically important difference for ICECAP-O score. However, a study100 has previously compared the measure in an older, post-acute patient population with the measure in the general population and found a mean difference on the ICECAP-O of 0.04 points. Although this cannot be used to confirm an important difference, our difference of 0.06 points (in the direction of worse well-being for participants allocated to probiotic) was greater, and may be indicative of a distinctive difference between arms.

The probiotic combination was quality assured by the manufacturer and probiotic and placebo were cultured after labelling to ensure correct labelling. SP pack identification numbers were allocated using randomly permuted blocks.

The SP was given to participants by care home staff using existing, routine care home medication administration processes. Administration of SP was recorded on the standard MAR sheets, and the responsible person reported how the SP was given, and whether it was fully, partially or not consumed. Relevant documentation was signed as correct by the responsible care home staff member. We captured daily SP usage data for a total of well over 70,000 days, with > 90% of the SP recorded as having been taken at the full dose. We are able, therefore, to verify adherence by a large proportion of study participants, which is likely to have been higher than a study of ‘free-living’ older people.

Stool samples pre and post probiotic from participants who volunteered for this additional aspect of the study confirmed that the probiotic combination was correctly produced, labelled, taken by participants and entered the gut.

The probiotic and placebo were provided to the study by the manufacturer as an unconditional grant, and the team delivering this publicly funded study was fully independent of the manufacturers of the probiotic combination or free from any other potentially relevant financial conflicts.

Studies of probiotics in older people have suffered from potential bias from poor ascertainment of outcomes. 23 The PRINCESS trial was able to guard against this by ensuring that participants were visited each week by a registered nurse blind to the participant’s allocation. Outcome data were prospectively obtained from the MAR sheets, care home clinical records, hospital records, observation of the participant and discussion with the participant (where possible) and the participants’ friends and family and care home staff. Weekly participant diary data were available for 97.4% of participants randomised to placebo and 98.7% of participants randomised to the probiotic combination. There were missing days for only 1.3% of the days initially scheduled for data capture using the study weekly diary.

Analysis was conducted only once the database had been locked. Analysis was done according to a statistical analysis plan signed off by an Independent Data Monitoring Committee and checked by a statistician independent of the study team. The allocation code was broken only once that statistical analysis report was finalised.

However, study progress and participant recruitment were slower than anticipated, owing mainly to longer than anticipated contracting processes, and difficulty in engaging with consultees. The latter problem was addressed by obtaining approval for a broader range of consultees who could be consulted on behalf of CHRs. We had planned to follow up all participating CHRs for a full year whenever possible, but these delays meant that follow-up had to be truncated for some. A total of 36.5% of participants were followed up for a full year as planned. However, as expected, about one-fifth of study participants died during follow-up and we were not able to follow up a minority (34.2%) of participants for a full year because of study delays. Nevertheless, every participant could have been followed up for at least 6 months. Fewer events, overall, than predicted in our power calculations meant that the study was underpowered to detect a statistically significant clinically meaningful difference in the primary outcome.

Our findings apply only to the probiotic combination studies for CAAD in CHR. Probiotics and probiotic combinations are likely to differ in their efficacy by setting, study population and outcomes measured.

Interpretation of results

These findings show that daily oral administration of the probiotics Lactobacillus rhamnosus GG and Bifidobacterium animalis subsp. lactis BB-12 does not reduce CAAD for all-cause acute infections in this population, or duration of infections, and does not have a measurable benefit on quality of life or a range of infection-related secondary outcomes. Although evidence of no benefit is a robust finding across all the outcomes that we measured, there was a trend towards harm with many of the outcomes, with better evidence for some analyses; for example, more people receiving probiotics experienced LRTIs, and CHRs randomised to the probiotics arm experienced a greater burden of AAD in one measure. Although such findings should be treated with caution, given multiple testing, it is noteworthy that findings in 21 out of 33 secondary outcomes trend away from benefit of the probiotic combination we studied.

Implications for health and social care

Based on these findings, CHRs should not be advised to consume a combination of the probiotics Lactobacillus rhamnosus GG and Bifidobacterium animalis subsp. lactis BB-12 each day to reduce antibiotic administration. It is important to highlight that the conclusions are relevant only for these specific strains of probiotic; we cannot conclude that no probiotic would be helpful in this population. We are not able to make recommendations about other probiotics or probiotic combinations because certain effects of probiotics may be strain specific. Neither are we able to indicate that these findings are applicable to other populations in different settings. It is also important to highlight that probiotic supplementation may vary according to immune status and age; we cannot conclude that a probiotic would be helpful in an enriched subgroup of the population (e.g. those who are immunocompromised, those aged > 80 years or those with the highest level of frailty).

Future research implications

Global sales of probiotics have been valued at over $40B and are projected to reach over $64B by 2023. 101 Some of these probiotics are marketed on the basis of health claims (although no such claims are currently permitted in the European Union), so rigorous research is needed to support evidence-based decision-making about probiotic use. AMR is a public health threat of global significance, and there is widespread interest in interventions that can replace or reduce antibiotic use, such as through the prevention of infections. 102

A systematic review19 of probiotics to reduce antibiotic use for common infections included 17 RCTs, all conducted in infants and children, and covered 13 different probiotic formulations, all comprising single or combination Lactobacillus and Bifidobacterium strains. The review found that probiotics reduced the risk of antibiotic prescription relative to placebo. Mean duration of probiotic supplementation ranged from 4 days to 9 months. The review found significant statistical heterogeneity among study results and variable study quality, with many studies being criticised for poor reporting of outcomes. 19 A further systematic review103 that included 20 RCTs explored the effect of probiotics, specifically Lactobacillus and Bifidobacterium strains, on the duration of acute respiratory infections in otherwise healthy children and adults, and found that these probiotics reduced the duration of respiratory illness. 103

Despite these systematic reviews of trials finding that probiotics had a favourable impact on antibiotic use in children, and on the duration of infections in children and healthy adults, we found evidence of no benefit from the probiotics we studied in a frail, care home population. Our findings are congruent with a more recent meta-analysis on the effectiveness of probiotics in preventing infections in older people, which included 15 studies covering 5916 participants with a mean age of 75.21 years. The authors concluded that the quality of this evidence was poor, that it did not support the use of probiotics for reducing infections in older adults, that safety outcomes were similar between probiotics and placebo and that more research was needed. 104 However, our study does not suffer from many of the potential biases of the studies included in the reviews, or in the study by Van Puyenbroeck et al. ;23 their trial included largely healthy older people, and it found no benefit in terms of antibiotic use or duration of common infections from the probiotic L. casei Shirota, but there was a high risk of ascertainment bias because of missing data.

Regarding mechanisms and potential harms, recent research has suggested that certain probiotics may delay return of the host gut microbiome to its normal state. 105 A retrospective, single-centre study found probiotics to be a risk factor for C. difficile infection in hospitalised patients. 13 Our trial and associated mechanistic findings could not rule out potential harm from this probiotic combination; however, SAEs were similar in our study groups.

Our mechanistic substudy results are in keeping with the clinical outcomes of the trial. Systematic reviews have indicated positive effects of other strains of probiotics in healthy elderly subjects106 and, hence, our findings may not be applicable to other populations and probiotic species.

Because probiotics are a feasible and cheap potential intervention, further rigorous efficacy, mechanisms and effectiveness trials of other probiotics in other population groups and settings may be indicated regarding antibiotic use and susceptibility to and recovery from common infections. Potential harms should be carefully studied in such trials.

Acknowledgement of contributions

Participants

Finally, and most importantly, we would like to thank all those who participated in the trial, and their consultees (if applicable), without whom this trial would not have been possible.

Contributions of authors

Professor Christopher C Butler (https://orcid.org/0000-0002-0102-3453) was the chief investigator for the study. He contributed to the conception and overall design of the study, obtaining funding and the implementation and interpretation of the work. He led the drafting of the report.

Dr Eleri Owen-Jones (https://orcid.org/0000-0003-0850-4724) (Research Associate) was the trial manager, contributing to the trial design and implementation, acquisition of data and the writing of the Abstract and Chapters 1 and 2 of the report. She also co-ordinated the compilation, formatting, proofreading and final approval of the report.

Mandy Lau (https://orcid.org/0000-0001-5894-570X) (Research Associate) was the trial statistician, conducting the statistical data analysis and contributing to the writing of Chapters 2 and 3, critical review of the report and final approval of the report.

Dr David Gillespie (https://orcid.org/0000-0002-6934-2928) (Deputy Director of Infection, Inflammation and Immunity Trials, Senior Research Fellow) was a co-investigator and contributed to the trial design, conducting the statistical data analysis, the writing of Chapters 2 and 3, critical review of the report and final approval of the report.

Dr Mark Lown (https://orcid.org/0000-0001-8309-568X) (GP and Clinical Lecturer) was a co-investigator and contributed to the overall trial design, interpretation of the findings, the writing of the mechanistic sections of the report, critical review of the report and final approval of the report.

Professor Philip C Calder (https://orcid.org/0000-0002-6038-710X) (Professor) was a co-investigator and contributed to the overall trial design, interpretation of the findings, the writing of the mechanistic sections of the report, critical review of the report and final approval of the report.

Helen Stanton (https://orcid.org/0000-0003-0197-3667) (Research Assistant) was the data manager and led the design and implementation of the qualitative substudy and contributed to the acquisition and interpretation of data, the writing of Chapters 2 and 5 of the report, critical review of the report and final approval of the report.

Dr Mandy Wootton (https://orcid.org/0000-0002-2227-3355) (Consultant Scientist/Operational Manager, Specialist Antimicrobial Chemotherapy Unit) was a co-investigator and lead microbiologist, and contributed to the overall trial design, acquisition of data, interpretation of the findings, the writing of the microbiology sections of the report, critical review of the report and final approval of the report.

Vivian Castro Herrera (https://orcid.org/0000-0003-0422-5496) (PhD Student) was a co-investigator and contributed to the laboratory work, interpretation of the findings, the writing of the mechanistic sections of the report, critical review of the report and final approval of the report.

Professor Antony Bayer (https://orcid.org/0000-0002-7514-248X) (Professor of Geriatric Medicine) was a co-investigator and contributed to the overall trial design, interpretation of the findings, critical review of the report and final approval of the report.

Dr Jane Davies (https://orcid.org/0000-0002-9409-8605) (Research Associate) was a RN and contributed towards acquisition of data and interpretation of the findings, critical review of the report and final approval of the report.

Alison Edwards (https://orcid.org/0000-0002-9603-2618) (Research Associate) was a RN and contributed towards acquisition of data and interpretation of the findings, critical review of the report and final approval of the report.

Mina Davoudianfar (https://orcid.org/0000-0002-2930-9437) (Clinical Trial Manager) was the trial manager and contributed to the trial design and implementation, acquisition of data and final approval of the report.

Heather Rutter (https://orcid.org/0000-0002-7297-2234) (Senior RN) was a RN and contributed to acquisition of data and interpretation of the findings, critical review of the report and final approval of the report.

Professor Kerenza Hood (https://orcid.org/0000-0002-5268-8631) (Professor of Trials and Director of Centre for Trials Research) was a co-investigator and contributed to the conception of the study and the overall trial design, supervised the statistical analysis and contributed to the critical review of the report and final approval of the report.

Professor Michael Moore (https://orcid.org/0000-0002-5127-4509) (Professor of Primary Care Research) was a co-investigator and contributed to the overall trial design, interpretation of the findings, critical review of the report and final approval of the report.

Professor Paul Little (https://orcid.org/0000-0003-3664-1873) (Professor of Primary Care Research) was a co-investigator and contributed to the overall trial design, interpretation of the findings, critical review of the report and final approval of the report.

Dr Victoria Shepherd (https://orcid.org/0000-0002-7687-0817) (Research Associate) was a co-investigator and contributed to the overall trial design, interpretation of the findings, critical review of the report and final approval of the report. She also co-ordinated the compilation, formatting, proofreading and final approval of the report.

Dr Rachel Lowe (https://orcid.org/0000-0002-7306-0085) (Research Fellow) was the Senior Trial Manager and contributed to the trial design and implementation, acquisition of data, critical review of the report and final approval of the report.

Elizabeth A Miles (https://orcid.org/0000-0002-8643-0655) (Lecturer) was a co-investigator and contributed to the overall trial design, interpretation of the findings, the writing of the mechanistic sections of the report, critical review of the report and final approval of the report.

Julia Townson (https://orcid.org/0000-0001-8679-3619) (Research Fellow) was the Senior Trial Manager and contributed to the trial design and implementation, acquisition of data, critical review of the report and final approval of the report.

FD Richard Hobbs (https://orcid.org/0000-0001-7976-7172) (Professor and Head of Primary Care Health Sciences) was a co-investigator and contributed to interpretation of the findings, critical review of the report and final approval of the report.

Professor Nick A Francis (https://orcid.org/0000-0001-8939-7312) (GP and Professor of Primary Care Research) was a co-investigator and contributed to obtaining funding, the overall trial design, interpretation of the findings, critical review of the report and final approval of the report submission.

Publication

Owen-Jones E, Lowe R, Lown M, Gillespie D, Addison K, Bayer T, et al. Protocol for a double-blind placebo-controlled trial to evaluate the efficacy of probiotics in reducing antibiotics for infection in care home residents: the Probiotics to Reduce Infections iN CarE home reSidentS (PRINCESS) trial. BMJ Open 2019;9:e027513.

Data-sharing statement

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

Patient data

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

Disclaimers

This report presents independent research. 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, the MRC, NETSCC, the EME 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 EME programme or the Department of Health and Social Care.

Introduction

(Completed written consent should have been sent to Helen Stanton prior to conducting the interview.)

• I would like to ask you about your experiences of carrying out the PRINCESS trial in the care home you work in, and to ask you what it has been like to be a part of the PRINCESS trial.

• This interview will be audio-recorded. The recording will be treated with the strictest confidentiality and may be listened to by the research team but by no-one else. The recording will not be labelled with your name, and any written record or report derived from it will be fully anonymised.

• Are there any questions you would like to ask me before we start?

Carrying out the PRINCESS trial

In the PRINCESS trial, we asked [care home management]/[care home team members] such as yourself to carry out various activities for the trial with residents, consultees and the PRINCESS researchers who came in to your care home.

1. To get us started, could you talk me through how your care home became involved in the PRINCESS trial?

   [Prompts: Management only: What did you think when the researchers at Cardiff or Oxford invited your care home to take part? What made you feel this trial was right for your care home? Was there any aspect of the trial/how you were approached that confirmed that you were willing for the home to participate?]

2. What were your roles and responsibilities on the trial?

   [Prompts: Were you involved more heavily at some stages of the trial (e.g. set-up, recruitment, follow-up)? Did you take charge of any particular tasks associated with the trial (e.g. paperwork, study product, recruitment, management and oversight)?]

3. What were your experiences of your initial contact with the researchers/nurses on the trial?

   [Prompts: Did you meet with the PRINCESS researchers/nurses before or during trial set-up? How did you (and the care home team) adapt to carrying out PRINCESS trial tasks in the care home? How did you (and the care home team) and the residents adapt to the PRINCESS researchers/nurses coming in to the care home to carry out PRINCESS trial tasks? How did you build a rapport with the PRINCESS researchers/nurses?]

4. If not stated previously in the interview: Were you involved in the recruitment process itself?

   • If yes, proceed to next question.

   • If no, skip to question 7.

5. What was your role in the recruitment process as a whole?

   [Prompts: How did you acquire information about the resident that would affect their suitability for the trial? Were there any participants who were eligible on paper, but you did not recommend approaching or recruiting? If so, why? How did you work with relatives (i.e. consultees) to recruit residents? How did you work with PRINCESS researchers/nurses to recruit residents? Did you undertake assessments of mental capacity for the purposes of recruitment? How did you find this? What was the most challenging aspect of recruiting residents? Did you have concerns about any part of the process?]

6. How did you feel about the consultee process?

   [Prompts: Were you involved in identifying consultees? How did you feel about older, perhaps more vulnerable individuals being recruited to a randomised controlled trial such as PRINCESS? Did you (or any of the care home team) act as nominated consultees for any residents? Did you contact/utilise any other resources/agencies for advice or information around decision-making where residents lacked capacity?]

The PRINCESS trial was a placebo-controlled trial, where the study product given to participating residents was either an inactive capsule or a capsule containing a probiotic. If the resident was initially allocated the probiotic their capsules would contain probiotic throughout the trial, and if they were initially allocated the inactive capsule (the placebo) they would continue taking placebo throughout.

1. If not stated previously in the interview: Were you involved in the handling or distribution of the study product?

   • If yes, proceed to next question.

   • If no, skip to question 11.

2. The study product could be taken as a capsule, in liquid or sprinkled on food. How did this go?

   [Prompts: Were there any issues with residents taking the study product? Did you have any concerns about the methods of ingestion?]

3. Were there any factors which you think may have affected the residents taking the study product daily?

   [Prompts: Were there any barriers to ensuring that residents received their study product? Do you think the routines at your care home lend themselves to giving out study product?]

4. How did you know whether the resident had taken the study product?

   [Prompts: Did you record ingestion on MAR charts? Did you ever have to recall from memory about study product consumption, dose and method of ingestion? To what extent were you confident that residents received the correct study product that they were allocated?]

Working within the care home environment

We asked the researchers/nurses employed to work on the PRINCESS trial to record study product adherence, as well as incidences of infections, antibiotics, diarrhoea and hospitalisations on weekly record forms.

1. If not stated previously in the interview: Were you also involved in the data collection for the trial?

   • If yes, proceed to next question.

   • If no, skip to question 13.

2. How did you go about acquiring the information you needed to complete the data collection forms?

   [Prompts: Please explain how you collected information about the residents? Did you find care home records easy to locate? How easy was it to derive information from written care home notes and MAR sheets? To what extent did you speak with residents or consultees about data collection?]

3. To what extent did you work with the PRINCESS researchers/nurses to collect information with/about the resident?

   [Prompts: What was your working relationship like with researchers/nurses in the care home? Were you able to help researchers/nurses collect information that they needed about the resident? Did you feel you had time to discuss participating residents with them? Were care home staff knowledgeable about the residents taking part? Did researchers/nurses ask you to take responsibility for any particular trial activities (e.g. taking samples)? How did you communicate about tasks?]

4. More generally, what was it like hosting trial activities in your care home?

   [Prompts: Were there any barriers/challenges to carrying out research activities (or helping the PRINCESS researchers/nurses to carry these out) in the care home environment? How did you attempt to overcome these? Going forward what do you think would make carrying out research in care homes easier?]

Working with vulnerable older people and consultees

The PRINCESS trial participants were older, perhaps very vulnerable residents. Those who lacked the mental capacity to provide consent to be part of the trial could take part if consultee advice/agreement was provided.

1. How did you feel about carrying out PRINCESS trial activities with vulnerable older adults in the care home environment?

   [Prompts: Did you have concerns about any aspect of the trial/carrying out research activities? Did you experience any challenges/barriers to assisting with the research activities?]

2. How do you think the residents and consultees felt about taking part in the trial?

   [Prompts: Did residents/consultees seem engaged/interested in the research? Did anything about the trial or trial activities cause any particular worry to the residents, consultees/relatives? (or anyone else?)]

At baseline, 3-month and 12-month time points researchers were also asked to collect responses to questionnaires (ICECAP and EQ-5D-5L) that measure health and capability outcomes.

1. If not stated previously in the interview: Did you carry out the ICECAP and EQ-5D-5L questionnaires with the resident? Did you ever act as a proxy on behalf of the resident and complete the questionnaires as you felt the resident would’ve answered if they were able?

   • If yes, proceed to next question.

   • If no, skip to question 20.

2. What was your experience of asking residents to answer these questions?

   [Prompts: Did you feel comfortable asking these types of questions? How relevant did you feel these questions were to older people or to residents lacking capacity?]

3. How did you feel putting yourself in the resident’s shoes and responding to the questions as you think they might have done?

Practical aspects of the research and training

I’d like to explore some of practical, hands-on aspects of the trial and the training provided.

1. Did you have enough room to store trial materials (e.g. trial site files, sample containers) and study product at the care home?

2. (If not already mentioned) How did you find taking and sending samples at the three follow-up time points (baseline, 3 months and 12 months)?

   [Prompts: What type(s) of samples were you involved in collecting? What was your typical procedure for taking, storing and sending samples? Were there any barriers/challenges to collecting samples? Did you complete the forms to go with the samples? How did you find this?]

3. What training (if any) did you receive for the PRINCESS trial?

   [Prompts: How was the training delivered? Was it useful? Did you have enough information? Is there any other information you would have liked? How much time did it take to complete? Was this OK? What would you change about the training?]

Trial intervention, probiotics and infections

So far, we’ve been talking specifically about the PRINCESS trial. I’d like to ask you a few more general questions about your views on probiotics and managing infections in older people in care homes.

1. What was your overall opinion of the intervention that was being trialled (i.e. probiotics)?

   [Prompts: Did you have an opinion on probiotics before your home became involved in the trial? Did you have any safety (or other) concerns about use of probiotics in the older population? Has it changed the way you view probiotic supplements?]

2. What do you think are the main challenges with managing infections in older people in care homes?

3. In terms of health-care services for older people in care homes (including those lacking capacity to make decisions for themselves), what do you think would be the most important thing to improve?

Conducting research in care homes/research process

Almost at the end of the interview; we’re interested in your overall experience of being involved in the PRINCESS trial and your thoughts on conducting research in care homes.

1. Could you tell me what it has been like to take part in the PRINCESS trial?

   [Prompts: Were there any problems? What were the best things about taking part? What were the worst things about taking part?]

2. Before you were involved in the PRINCESS trial did you have any prior experience hosting or carrying out research in your care home or in other care homes you have worked in?

3. Lastly, what are your views about care home research in general?

4. Do you have any other comments?

Thank you very much for taking part in this interview.

Introduction

(Completed written consent should have been sent to Helen Stanton prior to conducting the interview.)

• I would like to ask you about your experiences of carrying out the PRINCESS trial in care homes, and to ask you what it has been like to be a part of the PRINCESS trial.

• This interview will be audio-recorded. The recording will be treated with the strictest confidentiality and may be listened to by the research team but by no-one else. The recording will not be labelled with your name, and any written record or report derived from it will be fully anonymised.

• Are there any questions you would like to ask me before we start?

Carrying out the PRINCESS trial

In the PRINCESS trial, we asked research professionals such as yourself to carry out various research activities with residents, consultees and the care home staff in the care home.

1. To get us started, could you talk me through your involvement in the PRINCESS trial, and what your roles and responsibilities were on the trial?

   [Prompts: Did you have a particular interest in the study, or were you allocated the role as part of your overall work responsibilities? Did you work on the trial at any particular stage of the trial (e.g. set-up, recruitment, follow-up)? Did you take charge of any particular tasks associated with the trial (e.g. collecting samples, follow-ups, study product, recruitment, management and oversight)?]

2. What were your experiences of your initial contact with the care home?

   [Prompts: Were you involved in visiting the care homes before or during trial set-up? What reception did you get from management? How did you adapt to carrying out PRINCESS trial tasks in the care home? How do you think the care home staff and residents adapted to you being in the care home to carry out PRINCESS trial tasks? How did you build a rapport with the care home staff and residents?]

3. If not stated previously in the interview: Were you involved in the recruitment process itself?

   • If yes, proceed to next question.

   • If no, skip to question 7.

4. How did you assess resident eligibility for the trial?

   [Prompts: How easy was it to acquire information about the resident that would affect their suitability for the trial? Were there any participants who were eligible on paper, but you did not approach or recruit? If so, why?]

5. How did you approach the task of recruiting residents in to the trial?

   [Prompts: Talk me through the stages that you went through to recruit residents with or without capacity? How did you work with relatives (i.e. consultees) to recruit residents? How did you work with care home staff to recruit residents? Did you undertake assessments of mental capacity for the purposes of recruitment? How did you find this? What was the most challenging aspect of recruiting residents? Did you have concerns about any part of the process?]

6. How did you feel about the consultee process?

   [Prompts: How did you feel about recruiting older, perhaps more vulnerable individuals in to a randomised controlled trial such as PRINCESS? How did you identify consultees? Did you have to approach care home staff to act as nominated consultees for any residents? Did you contact/utilise any other resources/agencies for advice or information around decision making where residents lack capacity?]

The PRINCESS trial was a placebo-controlled trial, where the study product given to participating residents was either an inactive capsule or a capsule containing a probiotic.

1. If not stated previously in the interview: Were you involved in the handling or distribution of the study product?

   • If yes, proceed to next question.

   • If no, skip to question 11.

2. The study product could be taken as a capsule, in liquid, or sprinkled on food. How did this go?

   [Prompts: Were there any issues with residents taking the study product? Did you have any concerns about the methods of ingestion?]

3. Were there any factors which you think may have affected study product adherence?

   [Prompts: Were there any barriers to ensuring that residents received their study product? Did the care home routines lend themselves to giving out study product?]

4. How did you monitor adherence to study product?

   [Prompts: Did you look at MAR charts? How well were these completed? Did you ever have to use verbal reports about study product consumption, dose and method of ingestion? To what extent were you confident that residents received the correct study product that they were allocated?]

Working within the care home environment

We asked researchers to record study product adherence, as well as incidences of infections, antibiotics, diarrhoea and hospitalisations on weekly record forms.

1. If not stated previously in the interview: Were you involved in the data collection for the trial?

   • If yes, proceed to next question.

   • If no, skip to question 14.

2. How did you go about acquiring the information you needed to complete the data collection forms?

   [Prompts: please explain how you sourced information about the residents? Did you find care home records easy to locate? How easy was it to derive information from written care home notes and MAR sheets? Were care home staff helpful/knowledgeable about the residents taking part? To what extent did you speak with residents about their data?]

3. To what extent did you work with the care home staff to collect information with/about the resident?

   [Prompts: What was your working relationship like with staff in the care home? Were care home staff helpful in assisting you to collect information about the resident? Did you feel care home staff had time to discuss participating residents with you? Were care home staff knowledgeable about the residents taking part? Did you delegate any particular trial activities to care home staff (e.g. taking samples)? How did you communicate about tasks?]

4. More generally, what was working in the care home like?

   [Prompts: Were there any barriers/challenges to carrying out research activities in the care home environment? How did you attempt to overcome these? Going forward what do you think would make carrying out research in care homes easier?]

Working with vulnerable older people and consultees

The PRINCESS trial participants were older, perhaps very vulnerable residents. Those who lacked the mental capacity to provide consent to be part of the trial could take part if consultee advice/agreement was provided.

1. How did you feel about carrying out PRINCESS trial activities with vulnerable older adults in the care home environment?

   [Prompts: Did you have concerns about any aspect of the trial/carrying out research activities? Did you experience any challenges/barriers to carrying out the research activities?]

2. How do you think the residents and consultees felt about taking part in the trial?

   [Prompts: Did residents/consultees seem engaged/interested in the research? Did anything about the trial or trial activities cause any particular worry to the residents, consultees/relatives? (or anyone else?)]

At baseline, 3-month and 12-month time points researchers were also asked to collect responses to questionnaires (ICECAP and EQ-5D-5L) that measure health and capability outcomes.

1. If not stated previously in the interview: Did you carry out the ICECAP and EQ-5D-5L questionnaires with the resident or a proxy acting on behalf of the resident?

   • If yes, proceed to next question.

   • If no, skip to question 19.

2. What was your experience of asking residents and proxies to answer these questions?

   [Prompts: Did you feel comfortable asking these types of questions? How relevant did you feel these questions were to older people or to residents lacking capacity?]

Practical aspects of the research and training

I’d like to explore some of practical, hands-on aspects of the trial and the training provided.

1. Did you have enough room to store study materials (e.g. trial site files, sample containers) and study product at the care home?

2. How did you find the data collection materials and using the trial database?

   [Prompts: Did you find any of the information difficult/awkward to collect? What would you change about the data collection? How did you find using an online database, versus completing paper forms?]

3. How did you find taking and sending samples at the three follow-up time points?

4. What did you think of the training provided for the PRINCESS trial?

   [Prompts: How was the training delivered? Was it useful? Did you have enough information? Is there any other information you would have liked? How much time did it take to complete? Was this OK? What would you would change about the training?]

Trial intervention, probiotics and infections

So far, we’ve been talking specifically about the PRINCESS trial. I’d like to ask you a few more general questions about your views on probiotics and managing infections in older people in care homes.

1. What was your overall opinion of the intervention that was being trialled (i.e. probiotics)?

   [Prompts: Did you have an opinion on probiotics before you became part of the study? Did you have any safety (or other) concerns about use of probiotics in the older population? Has it changed the way you view probiotic supplements?]

2. What do you think are the main challenges with managing infections in older people in care homes?

3. In terms of health-care services for older people in care homes (including those lacking capacity to make decisions for themselves), what do you think would be the most important thing to improve?

Conducting research in care homes/research process

Almost at the end of the interview; we’re interested in your overall experience of taking part in the PRINCESS trial and your thoughts on conducting research in care homes.

1. Could you tell me what it has been like to take part in the PRINCESS trial?

   [Prompts: Were there any problems? What were the best things about taking part? What were the worst things about taking part?]

2. Before you took part in the PRINCESS trial did you have any prior experience with conducting research in care homes?

3. Lastly, what are your views about care home research in general?

4. Do you have any other comments?

Thank you very much for taking part in this interview.

Introduction

(Completed written consent should have been sent to Helen Stanton prior to conducting the interview)

• I would like to ask you about your experiences of [taking part in]/[providing advice on behalf of your relative for] the PRINCESS trial.

• This interview will be audio-recorded. The recording will be treated with the strictest confidentiality and may be listened to by the research team but by no-one else. The recording will not be labelled with your name and any written record or report derived from it will be fully anonymised.

• Are there any questions you would like to ask me before we start?

Carrying out the PRINCESS trial

In the PRINCESS trial, we asked [people such as yourself]/[older people living in care homes] to take part in the research that would help us determine if a probiotic supplement can help reduce infections and, ultimately, reduce the consumption of antibiotics.

1. To get us started, could you talk me through how [you]/[your relative] became involved in the PRINCESS trial?

   {Prompts: What did you think when [you]/[your relative] were invited to take part? What made you feel this trial was right for [you]/[your relative] to take part in?}

2. What happened when [you]/[your relative] were approached to take part in the trial?

   {Prompts: What were your experiences of your initial contact with the researchers/nurses who worked on the trial? What was it like having other people you weren’t familiar with coming in to the home and discussing the trial with [you]/[you on behalf of your relative]?}

   [For consultees only] How did you feel about providing advice about the trial on behalf of your relative?

   [Prompts: How did you feel about your relative, whom you might consider vulnerable being recruited in to a randomised controlled trial such as PRINCESS? Were there any aspects of this process you found particularly challenging? Did you speak to anyone else about the study, or consult anyone else for advice, about whether it was a good idea for your relative to take part in PRINCESS?]

3. How were channels of communication about the trial?

   [Prompts: Were updates and information about aspects of the trial forthcoming? Did you feel like you were included in plans (e.g. visits from researchers) and decisions about the trial?]

PRINCESS was a placebo-controlled trial, where the study product given to participating residents was either an inactive capsule, or a capsule containing a probiotic. If the resident was initially allocated the probiotic their capsules would contain probiotic throughout the trial, and if they were initially allocated the inactive capsule (the placebo) they would continue taking placebo throughout.

1. How did you feel about [your relative] taking the study product?

   {Prompts: Did [you]/[they] experience any issues taking the study product? Did you feel you had enough information about the study product and the potential effects it could have?}

2. The study product could be taken as a capsule, in liquid, or sprinkled on food. Did you have any concerns about the method in which it was taken?

3. Were there any factors which you think may have affected [you]/[your relative] taking the study product daily?

   {Prompts: Were there any barriers to ensuring that residents received their study product? Do you think the routines at [your]/[your relative’s] home lend themselves to giving out study product?}

4. Did you feel [you]/[your relative] were always receiving the correct study product for [you]/[them]?

Working with the care home team and PRINCESS researchers/nurses

We asked the researchers/nurses employed to work on the PRINCESS trial to collect data about [you]/[your relative]. This included whether [you]/[your relative] had taken study product each day, and whether [you]/[your relative] had shown signs of infection, taken antibiotics, had diarrhoea, or had been admitted to hospital.

1. To what extent did the PRINCESS researchers/nurses speak with you to collect the information needed for the trial?

2. What was your impression of the way the PRINCESS researchers/nurses carried out the trial in [your]/[your relative’s] home?

   {Prompts: Did you see the PRINCESS researchers/nurses often at [your]/[the] home? Do you think they enjoyed coming in to [your]/[the] home? Did you get an opportunity to chat with the researchers/nurses when they visited? What kind of relationship did you have with the PRINCESS researchers/nurses?}

3. How do you think the staff at [your]/[the] home found hosting and carrying out the PRINCESS trial with [you]/[your relative] and others at [your]/[the] home?

   {Prompts: What impact do you think having the trial at [your]/[the] home had on the staff and residents at [your]/[the] home? Do you think doing the trial at [your]/[the] home changed anything about the way routines usually run? If yes, what and why do you think this was?}

Practical aspects of the research

We asked for permission to collect samples from [you]/[your relative] at the beginning of the trial, and at 3 months and 12 months after initially joining the study.

1. (If not already mentioned) Did [you]/[your relative] have samples taken at the three follow-up time points (baseline, 3 months and 12 months)?

   {Prompts: What type(s) of samples did [you]/[your relative] have taken? Did you have any concerns about any aspect of the sample-taking? Is there anything that could have been improved regarding this part of the trial?}

Trial intervention, probiotics and infections

So far, we’ve been talking specifically about PRINCESS trial. I’d like to ask you a few more general questions about your views on probiotics and managing infections.

1. What was your overall opinion of the intervention that was being trialled (i.e. probiotics)?

   [Prompts: Did you have an opinion on probiotics before your home became involved in the trial? Did you have any safety (or other) concerns about use of probiotics in the older population? Has it changed the way you view probiotic supplements?]

2. [For consultees only] What do you think are the main challenges with managing infections in older people in care homes?

3. [For consultees only] In terms of health-care services for older people in care homes (including those lacking capacity to make decisions for themselves), what do you think would be the most important thing to improve?

Conducting research in care homes/research process

Almost at the end of the interview; we’re interested in your overall experience of being involved in the PRINCESS trial as a [participant]/[consultee on behalf of your relative].

1. [As a participant] Could you tell me what it has been like to take part in the PRINCESS trial? [As a consultee] How do you think your relative has found taking part in the PRINCESS trial?

   [Prompts: Were there any problems? What were the best things about taking part? What were the worst things about taking part?]

2. Before you were involved in PRINCESS [as a participant]/[in an advisory capacity on behalf of your relative] had you taken part in research before?

3. Lastly, what are your thoughts about doing research in this environment?

4. Do you have any other comments?

Thank you very much for taking part in this interview.