FormalPara Take-home message

Early and sustained probiotic administration to adult patients requiring treatment in the intensive care is safe but ineffective in improving outcomes or reducing nosocomial infection. Whether more targeted therapy is beneficial remains uncertain.

Introduction

Critical illness requiring treatment in an intensive care unit (ICU) results in rapid and profound alterations to the gastrointestinal microflora [1, 2]. Microbiota depletion and diversity loss are associated with adverse outcomes including prolonged hospital stay, nosocomial infections and increased mortality [2,3,4]. The reintroduction of commensal bacteria using strain-specific oral probiotic therapy may mitigate these adverse effects [5, 6].

Lactobacillus (L.) plantarum 299 V is a human commensal that survives passage through the gastrointestinal tract, irrespective of gastric acidity [7]. It reduces gastrointestinal bacterial translocation, attenuates systemic inflammation in critically ill patients, and has in vitro antimicrobial activity against a wide range of potentially pathogenic species [8]. In a meta-analysis of 14 randomised trials involving adult critically ill patients, probiotic therapy decreased overall infections, a benefit most apparent in trials of L. plantarum [9]. Given that probiotics exhibit strain-specific effects, L. plantarum, as a single agent, is a strong candidate intervention to improve clinical outcomes. However, sufficient evidence to inform clinical practice is limited by trial quality and heterogeneity in the timing, dose and duration of therapy, resulting in conflicting guideline recommendations [10, 11].

The multicentre, randomised, restoration of gut microflora in critical illness trial (ROCIT) was designed to test the hypothesis that, compared with placebo, the early and sustained enteral administration of L. plantarum 299v probiotic therapy, in adult patients expected to require ongoing treatment in the ICU beyond the day after recruitment, would improve clinical outcomes including increased days alive and out of hospital to Day 60 (DAOH60).

Methods

Trial design

The investigator initiated ROCIT study was a parallel group, placebo-controlled, randomised clinical trial conducted in the ICUs of five hospitals in Perth, Western Australia (ANZCTR 12617000783325). The protocol was prospectively approved by the research ethics committee (HREC) of all participating institutions and reported prior to completion of the study (South Metropolitan Health Service Human Research Ethics Committee ref:RGS000004, St John of God Health Care Human Research Ethics Committee ref:1183) [12]. Initial HREC approval had included the provision to enrol participants who lacked capacity to provide informed consent, where prospective consent was able to be obtained from the person responsible. In June 2018, after 83 participants had been enrolled, the study management committee received an updated interpretation of local legislation from the lead HREC. This mandated that trial recruitment of the subsequent 138 participants was restricted to patients competent to provide consent prospectively. Approval to analyse and report the participants enrolled prior to this change was granted by the HREC.

Patients

Eligible patients were adults within 48 h of ICU admission and expected by the treating clinician to require ICU care beyond the calendar day after recruitment. Key exclusion criteria included immunosuppression, presence of a prosthetic heart valve or permanent pacemaker and admission to hospital from a high-level nursing or rehabilitation facility (Fig. 1). The complete exclusion criteria are provided in the Supplementary Appendix (eTable 1).

Fig. 1
figure 1

Participant flow

Full size image

Randomisation and masking

Variable-block, 1:1 randomisation, stratified by site, was generated using a web-based interface [13]. Allocation concealment was maintained using an unblinded pharmacist to assign unique, sequential numbers to each bottle of study drug. The active study drug and the placebo were prepared by a certified facility in identically packaged capsules with 60 capsules per bottle (Metagenics Australia, 741 Nudgee Road, Northgate, Qld, 4013). All members of the treating team, the study participants, research staff and outcome adjudicators were blinded to the treatment allocation. Unblinding occurred after database lock and completion of the statistical analysis.

Study treatment

The study drug was administered once daily, beginning immediately after enrolment and continued for 60 days. Study participants who were discharged from hospital prior to Day 60 were advised to continue the treatment regime until the course was complete, then return a post-discharge treatment diary and the study drug bottle to the coordinating site. The active study drug contained 2 × 1010 colony-forming units (CFUs) of L. plantarum 299v per capsule, a dose comparable or greater than that used in other studies [9]. The placebo was of identical appearance but contained only microcrystalline cellulose. Independent testing of each study drug batch at the lead site confirmed bacterial absence in the placebo capsules and > 2 × 1010 CFUs of L. plantarum 299v in the probiotic capsules. Participants were requested to refrain from initiating any probiotic treatment during the 60 days of study participation. All other aspects of care were at the discretion of the patient and clinical teams.

Outcome measures

The primary outcome was DAOH60. This is a validated composite measure for which the components of death, index hospital length of stay and the occurrence and duration of hospital readmission, plausibly, may all be improved by probiotic therapy [14]. Days spent in a rehabilitation facility or high-level nursing facility to Day 60 were considered as days in hospital and participants who died prior to Day 60 were recorded as having zero DAOH60.

Incident nosocomial infections, a secondary outcome, were assessed independently by two blinded infectious diseases specialist clinicians. These included hospital-acquired pneumonia, ventilator-associated pneumonia, Clostridioides difficile-associated diarrhoea, surgical site infection, urinary tract infection, and blood stream infection as defined by Centre for Disease Control (CDC) criteria [15]. Antibiotic-free days were collected for all days in hospital, including any readmissions to Day 60. Quality of life was assessed using the five-level EQ-5D (EQ-5D-5L) questionnaire including the EQ-5D descriptive system with five domains (mobility, self-care, usual activities, pain/discomfort and anxiety/depression) and the EQ visual analogue scale (EQ VAS) [16]. The number of missed study medication days was the sum of the missed days whilst hospitalised, plus, either the number of remaining capsules in returned bottles post-discharge or the number of omitted study diary days post-discharge, whichever was greater.

Statistical analysis

In the placebo group, a mean DAOH60 of 37 and standard deviation (SD) of nine was assumed, using previous estimates from the lead study site [12]. Based on a two-sided type I error rate of 0.05 and sample size inflation of 20% to account for rank-based testing, a further 5% each for withdrawn consent and loss to follow-up, a sample size of 220 participants was determined to have 80% power to detect a between-group difference in DAOH60 of 4 days. This difference was considered clinically meaningful to ICU consumers [12].

The primary analysis was performed on an intention-to-treat population, defined as all eligible, randomised patients. For non-parametric outcomes, significance was determined using the Wilcoxon rank-sum test, median difference was calculated using quantile regression, with the inversion method used to calculate a 95% confidence interval (CI). Fischer’s Exact test and Chi2 test were used to test association between categorical outcomes as appropriate. Analyses of pre-specified subgroups (sepsis defined according to sepsis-3 criteria [17], antibiotics at enrolment, ICU admission urgency and ICU admission type), and a post hoc septic shock subgroup, included an interaction term between assigned treatment and subgroup using a two-sided hypothesis test. The discrepancy in pre-specified subgroups reported in the trial registration and published protocol is provided in the Supplementary Appendix. A sensitivity analysis was conducted including all patients receiving study medication for ≥ 80% of days alive to Day 60. Each health state was converted into the corresponding utility index, indicating the preference of being in a health state, with utilities calculated using Australian value weights [18]. Adverse outcomes, based on clinician suspicion of an association with the intervention, were reported for all randomised patients. All serious adverse events, defined and reported according to the Note for Guidance on Clinical Safety Data Management: Definitions and Standards for Expedited Reporting (CPMP/ICH/377/95), were reported to the Data Safety Management Committee (Membership provided in the Supplementary Appendix). No interim analysis was planned or undertaken. A P value of less than 0.05 was deemed statistically significant. No correction was made for multiple comparisons. All data analyses were conducted using R version 3.5.1.

Results

Between July 2017 and December 2019, 221 patients were enrolled in the study (Fig. 1). Three patients were found to be ineligible, leaving an intention-to-treat population of 218. The primary outcome data were available for all participants. Baseline characteristics of the probiotic (n = 110) and placebo group (n = 108) participants were similar, although 81% of patients in the probiotic group compared with 62% in the placebo group received antibiotics at baseline (Table 1). The number of patients receiving study medication for ≥ 80% of days alive to Day 60 was 95 (86.4%), and 87 (80.6%) in the probiotic and placebo groups, respectively. Other measures of study treatment compliance are provided in the Supplementary Appendix (eTable 2).

Table 1 Baseline participant characteristics
Full size table

Primary outcome

The median [interquartile range (IQR)] number of DAOH60 in the probiotic group was 49.5 (IQR 37–53) and 49 (IQR 43.8–53) in the placebo group, with a between-group absolute difference of 0.0 (95% CI − 6.1 to 7.1), P = 0.55. There were no significant between-group differences observed in the components of the primary outcome, or when analysis was limited to participants with ≥ 80% compliance (Table 2). There was no significant between-group difference for the four pre-specified subgroup pairs (Table 3).

Table 2 Outcomes analysis
Full size table
Table 3 Subgroup analysis
Full size table

Secondary outcomes

Nosocomial infection occurred in 8 (7.3%) and 5 (4.6%) of the probiotic and placebo group participants, respectively, odds ratio 1.62 (95% CI 0.51–5.1), P = 0.57. No participant had more than one nosocomial infection. Other clinical outcomes including ICU and hospital mortality were similar between groups (Table 2, eTable 4). Amongst survivors, overall quality of life at Day 60, as assessed by median EQ-5D-5L VAS scores was similar in the probiotic and placebo groups, 75 (IQR 60–85) and 76 (IQR 60–90), respectively, between-group difference − 1.0 (95% CI − 14.5 to 16.3), P = 0.39. The individual components scores are provided in the Supplementary Appendix (eTable 5).

A post hoc exploratory analysis suggested significant increase in DAOH60 and antibiotic-free days in the subgroup of 24 participants with septic shock at baseline (eTable 3).

Safety

There were no serious adverse events reported amongst the 221 randomised participants, including no cases of Lactobacillus infection. There were three adverse events reported in total, one in the probiotic group and two in the placebo group (eTable 6).

Discussion

In this multicentre, randomised, placebo-controlled clinical trial, the early and sustained administration of probiotic therapy with L. plantarum 299v to adult patients admitted to the ICU did not result in a significant difference in days alive and out of hospital to Day 60. Nosocomial infection and all other pre-specified secondary outcomes were also similar between groups. The administration of L. plantarum 299v was safe. There were no serious adverse events amongst participants, including no associated Lactobacilli infections, and few reported adverse events.

A systematic review of previous clinical trials suggested a lower incidence of nosocomial infection and improved clinical outcomes amongst critically ill patients receiving probiotic therapy [9]. However, the robust design of the placebo-controlled ROCIT study, powered to detect a meaningful difference in a patient-centred outcome, provides results that are consistent with the findings of other higher quality trials [9]. The ROCIT study extends these findings by evaluating the early administration of a high dose probiotic, sustained amongst survivors until ascertainment of the primary outcome at 60 days. In addition to a lack of benefit demonstrated amongst the entire cohort, subgroup analyses based on antibiotic administration at time of enrolment, presence or absence of sepsis, ICU admission urgency and type, also failed to demonstrate benefit. Similarly, an analysis of highly compliant participants, though a post-randomisation variable that could not be determined at baseline, did not suggest that the lack of benefit could be explained by insufficient probiotic exposure. Together, these findings suggest that the widespread, untargeted administration of L. plantarum 299v to patients admitted to the ICU, although safe, is ineffective.

In a large clinical trial conducted in infants in rural India, the administration of L. plantarum decreased the risk of a composite outcome including infection and death [19]. Whilst the microbiota of adults is complex and established, in infants it is newly developing. This may explain why an untargeted approach could be successful in infants but not adults. Untargeted enteral probiotic administration to critically ill patients receiving mechanical ventilation may be of specific benefit due to the risk of aspiration of gastric contents [20]. In addition, the expected mortality of the cohort in this study was relatively low, and mechanically ventilated patients may be expected to have a higher illness severity. Although the attributable mortality of ventilator-associated pneumonia is uncertain, further study of this cohort will provide important information of the role of untargeted probiotics in another high acuity cohort [21].

Alternatively, a more targeted approach, based on a specific gastrointestinal microbiome composition may identify a population of critically ill adults who would benefit from L. plantarum administration [22]. Preliminary studies have found an association between gastrointestinal microbiota composition and adverse outcomes from critical illness [23]. To date, however, there is limited evidence that this risk is modifiable [24]. This may be explained by the substantial variation in microbiota disturbance observed between and within individuals after a uniform exposure such as broad spectrum antibiotic [25]. The post hoc analysis suggesting benefit of L. plantarum 299v amongst patients with septic shock may be a chance finding and must be considered hypothesis generating, but may be due to consistent, severe dysbiosis in this subgroup. Irrespective, developing targeted interventions will require a mechanistic and individualised understanding of the effects probiotics exert on the gastrointestinal microbiota in critical illness. Developing tools to provide information on the composition of the gastrointestinal microbiota in a clinically relevant timeframe could assist with a goal of precision restoration of gut microbiota constituents and diversity.

There are several limitations to this study. Observed days alive and at home in the placebo group were higher than estimated in the sample size calculation so that the findings may be underpowered. However, on the basis of the minimal between-group differences in all outcome measures, a false-negative finding is considered unlikely. Although the trial was initially approved to include participants who lacked capacity to provide consent, the approval to enrol incapacitated patients was rescinded after approximately one third of the participants were enrolled, and enrolment then limited to participants with capacity to provide prospective consent. Incapacitated patients may differ in important characteristics, including illness severity, and overall mortality of 5% is low for an ICU cohort. However, outcomes were similar amongst planned and unplanned ICU admissions, suggesting that identifying a cohort based on higher acuity alone is unlikely to lead to different findings. Reliance on clinical suspicion of an adverse event may increase the chance of underreporting, although lack of any outcome difference suggests that any such effect resulted in minimal impact. Finally, whether L. plantarum 299v administration targeted to a subgroup of critically ill patients, or whether untargeted administration of an alternative probiotic is beneficial, remains uncertain.

Conclusion

The early and sustained administration of probiotic therapy with L. plantarum 299v to adult patients admitted to the ICU did not result in a significant difference in days alive and at home to Day 60.