Introduction

Antibiotic-associated diarrhea (AAD) is a common complication of antibiotic use resulting from the disruption of the normally protective intestinal microbiome. While some factors that disrupt the intestinal microbiome are well-known (such as antibiotic exposure, advanced age, diet, and health status) [1], other factors relating to geography and ethnicity are less well documented, but may have important implications for microbiome-based clinical therapies [2]. The World Health Organization (WHO) defines acute diarrhea as ≥3 loose or liquid stools per day for ≥3 days and lasting <14 days [3]. Antibiotic-associated diarrhea is defined as diarrhea occurring while on antibiotics or within 8 weeks of antibiotics [1]. The prevalence of AAD in patients receiving antibiotics can be quite high, ranging from 5-37% in adults and may be higher in children (11-40%) [1]. The onset of AAD typically occurs during antibiotic administration, but delayed onset AAD may occur in 1020% for up to 8 weeks after antibiotics have been discontinued [1]. While the severity of AAD is typically mild-moderate in severity, 16% of cases are more severe, requiring further antibiotic treatments and 10-35% may be due to Clostridioides difficile infection (CDI), which can cause healthcare-associated outbreaks. Consequences of AAD include dehydration (especially in young children), higher mortality and morbidity, increased lengths-ofstays for inpatients (4-24 days), higher risk of colectomy (1-9%), higher healthcare costs and 25-28% of those with CDI develop recurrent form of the infection [1,4].

China is the second largest consumer of antibiotics, most commonly azithromycin, clindamycin and erythromycin [5,6,7]. However, only 25-39% of antibiotics prescribed in China are given appropriately [5]. The prevalence of AAD in China ranges from 14-35%, depending upon age, hospitalization status and co-morbidities [4,8]. Efforts prevent AAD include antibiotic stewardship programs and the adjunctive use of probiotics during antibiotic use [9,10]. A survey of 138 pediatric practitioners in China found 31% gave probiotics along with antibiotics, mainly due to concerns about antibiotic resistance or development of AAD [6].

Specific types of probiotics have been recommended in clinical guidelines for the prevention of AAD, but not all probiotic strains have shown efficacy [10,11,12]. Guidelines for probiotic use to prevent AAD in children focused on Asia-Pacific populations had strong recommendations for S. boulardii CNCM I-745 and L. rhamnosus GG [13]. These two stains were also recommended for the prevention of AAD in adults in the Asia-Pacific region [14]. Indeed, the efficacy for probiotics is strain-specific and each strain or multi-strained blend must be assessed separately to determine efficacy for AAD prevention [15]. Prior meta-analyses have found Saccharomyces boulardii CNCM I-745 and Lactobacillus rhamnosus GG to have the strongest evidence for the prevention of AAD, but often have not included RCTs published in non-

English languages, due to translation problems and limited access to Chinese journals [16,17,18,19,20]. 

Saccharomyces boulardii CNCM I-745 is a strain of probiotic that has a long history of use and has demonstrated significant efficacy and safety for a variety of diseases ranging from prevention of AAD and CDI, treatment of pediatric acute gastroenteritis, to reduction of side-effects of H. pylori eradication therapies [17,20,21,22,23]. S. boulardii CNCM I-745 reaches high levels in the intestine within 2-3 days, is cleared from the colon within 5 days and can be given concurrently with antibiotics, as the yeast is not susceptible to antibiotics [24]. The ability of S. boulardii CNCM I-745 to be therapeutic is due to multiple mechanisms-of-action: direct inhibition of pathogen growth or destruction of pathogenic toxins, interference with pathogen/ toxin attachment on enteric cell surfaces, improving intestinal cell health, reduction of water secretion, immune system regulation and restoring the normally protective microbiome barrier layer [14,21,25]. S. boulardii CNCM I-745 has a remarkable safety profile in that this strain has been given to a wide variety of patient types (children, adults and the elderly, hospitalized inpatients and outpatients, patients with acute and chronic conditions) and data gathered since its introduction in Europe in the 1950s has shown few adverse reactions, most were thirst and constipation [22]. Genetic transfer of antibiotic-resistant genes has not been observed with this strain [26]. In rare cases (1/5.6 million users), immunocompromised patients or inpatients with central catheters have developed fungemia [23].

Our aim in this study is to determine its efficacy and safety of Saccharomyces boulardii CNCM I-745 for the prevention of AAD and CDI in trials done in China using a comprehensive literature search to uncover trials not previously found in nonChinese databases.

Materials and Methods

Literature search

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were followed [27]. The PRISMA checklist is provided (Supplementary Table S1). Recent recommendations for reporting probiotic meta-analyses were also followed [28]. The project and protocol were prospectively registered with PROSPERO database of systematic reviews [www.

crd.york.ac.uk/PROSPERO/] on May 25, 2023 (CRD 429831).

 The following databases were searched [PubMed, Google Scholar, and Chinese databases (Chinese Biomedical Literature (CBM) and Chinese National Knowledge Infrastructure (CNKI)] from database inception to June 2023 to identify randomized controlled trials (RCTs) of probiotic use (limited to Saccharomyces boulardii CNCM I-745) to prevent antibiotic-associated diarrhea (AAD), including Clostridioides difficile infections (CDI) using randomized controlled trials done in China. The search terms were used: (antibiotic-associated diarrhea OR C. difficile AND Saccharomyces boulardii AND China). No language restrictions were imposed and publications in Chinese were translated into English.

Secondary searches of grey literature included reference lists, authors, reviews, meeting abstracts websites and clinicaltrials.gov for unpublished trials.  A recursive search was also performed, using the bibliographies of all obtained articles. 

Study selection

Inclusion criteria included: randomized, controlled clinical trials (RCTs) in children or adults receiving new prescription of antibiotics, comparison of S. boulardii versus control interventions and published in peer-reviewed journals. Only formulations of S. boulardii CNCM I-745 fulfilling the standard definition (must be living microbe, of adequate dose and conferring a health effect on the host) were included [29]. This definition excludes dead or heat-killed microbes and prebiotics. As bacterial and fungal taxonomies shift over time, the most current strain designations are presented in this review and strain identification was confirmed with the original authors or the manufacturer whenever possible. Trials with the primary outcome of prevention of AAD or CDI were included. Studies were included only if performed in China.

Exclusion criteria included: study not done in China, nonhuman studies, early phase 1 or 2 safety or mechanism of action studies, no control group, probiotic or AAD not well-described, clinical trials for the treatment of existing diarrhea, primary outcome was for the eradication of H. pylori, other probiotic strain(s) contained in the probiotic formulation, non-China setting, reviews and duplicate reports. Cross-over trials were excluded due to the potential for effect carry-over after short wash-out periods used in these trials and direct probiotic-probiotic studies were excluded if no non-probiotic control group was included.

Data extraction

The literature was searched independently by two coauthors (LM, LT). Data from all RCTs were extracted using a standardized data extraction form (Supplementary data, Form S1) initially completed by one co-author and then each study was independently scored by the other co-author following the standard methods for systematic reviews and meta-analysis [27,30]. Any disagreements were discussed until consensus was reached. The data extracted included PICOS data: (1) population (neonate, pediatric or adults, age range and country), (2) intervention (type of probiotic or controls used, daily doses, formulation, duration and follow-up times), (3) comparisons (type of control group either placebo or open, unblinded), (4) Primary outcomes (incidence of AAD or CDI), (5) secondary outcomes (stool frequency, length of hospitalization, cost data,  safety data and clinical characterization of AAD). In addition, data on potential confounding factors were collected: type of antibiotic given, type of control, study quality, setting (inpatient or outpatient), dose and duration of probiotic. For missing data not reported in the published article, we attempted to contact the author or co-authors to obtain the missing data.

Intervention

The intervention was the yeast probiotic strain Saccharomyces boulardii CNCM I-745 (Yihuo ™ or Bioflor™, Biocodex, France). If the strain designation was not reported in the published study, the strain was verified with the manufacturer or website. The intervention may be given in oral administration as capsules or sachets. The duration of the intervention was typically given for the duration of the antibiotic(s), but may have been continued after the antibiotics are discontinued.

Outcomes

The primary outcomes include the efficacy for the reduction of AAD incidence and the reduction of CDI incidence. The outcome measures included: the number of patients developing AAD (diarrhea defined as at least two loose/watery stools in young children or ≥3 loose/watery stools in older children and adults for 48 hours, not caused by rotavirus, with an onset while on antibiotics or within 8 weeks of antibiotic cessation or CDI (AAD with positive C. difficile toxin result). Secondary outcomes included: daily stool frequency at the end of the intervention, length of hospitalization, clinical presentation of AAD (measured by duration of AAD, severity of AAD symptoms and treatment effectiveness rating), costs of medical care and frequency of adverse events. Severe AAD was defined differently in the RCTs, but typically included “diarrhea with electrolyte disorders, severe dehydration, acidosis or other symptoms of toxicity”. Outcomes may be documented by patients/parents using daily diaries or by healthcare providers if hospitalized.

Risk of bias

Each included RCT was reviewed for quality and risk of bias and scored independently by both co-authors using standard methods [31]. The risk of bias (RoB) assessed with the RoB 2.0 tool and was graded (high,  some concerns, or low) for each of five domains of bias (randomization process, deviations from intended interventions, missing outcome data, measurement of outcome, selection of reported result) and a summary table and figure of bias was generated [32].

Statistical Analysis

As recommended by experts, inclusion of studies in our meta-analysis also required at least two RCTs using a common outcome measure [28]. Statistical analysis and generation of forest plots of pooled summary estimates was performed using Stata software version 16 (Stata Corporation, College Station, Texas) with meta-analysis modules [33]. Dichotomous outcomes were assessed using relative risks (RR) and 95% confidence intervals (C.I.) and continuous outcomes were assessed using standardized mean difference (SMD) and 95% C.I. using standard methods [31]. The significance level was set at P-value < 0.05. Heterogeneity across trials was evaluated using the I² statistic, 0% indicating none and ≥50% indicating a high degree of heterogeneity across the trials [33]. Random-effects models were used for the meta-analysis if significant heterogeneity was found (I²>50%), otherwise fixed-effects models were used.  Publication bias was assessed using funnel plots and the Egger’s test [33]. Subgroup analysis was used to explore sources of heterogeneity and was assessed with the Cochrane Q test [31]. A priori subgroup analyses based on factors that might influence the magnitude of efficacy estimates were planned for the following: (a) age,  (b) daily dose S. boulardii, (c) risk of bias, (d) indication for antibiotic use (upper respiratory disease, gastrointestinal, etc.), (e) type of patient (inpatient/outpatients), (f) time of intervention initiation from onset of diarrhea, (g) rural versus urban setting, (h) route antibiotic given (IV or oral) and (i) extent of blinding. Trials were pooled in the meta-analysis if a common outcome was used and there were at least two trials within each sub-group. Sequential sensitivity analysis was done to explore the extent outcomes were dependent upon a particular trial.

Results

The literature search resulted in 361articles that were initially screened and 233 were excluded during the initial screening (Figure 1): duplicate publications (n=88), reviews (n=73), cross-over studies (n=23), pre-clinical studies (n=18) or other miscellaneous reasons (n=31). Full articles (n=128) were reviewed and 82 were excluded: treatment of existing AAD (n=18), H. pylori eradication trials (n=17), no non-probiotic controls (n=16), unclear AAD definition (n=14), direct probiotic to probiotic studies (n=12), probiotic tested was other than S. boulardii (n=2), treatment of rotaviral diarrhea (n=2) and unclear outcome data (n=1). A total of 46 RCTs published 2011-2022 were included (8,201 participants)

[34-79].  Only four RCTs were found in commonly used nonChinese databases [54,59,60,64] and the other 42 trials were found using CNKI/CMI databases. A funnel plot (Supplementary Figure S1) showed significant publication bias for the included studies Egger’s test: t= -3.15, P = 0.003) due to a lack of RCTs showing no significant efficacy of S. boulardii CNCM I-745.

Figure 1: PRISMA study inclusion flow-chart.

Study participant characteristics 

There was a mean of 178 + 125 enrolled subjects/trial (range 46-552), as shown in (Table 1). Most trials (n=44, 96%) were in pediatric subjects (aged neonates to 16 years old) and two trials were in elderly (>65 years old) patients [44,72]. All patients were hospitalized.

The types of antibiotics used varied, with most RCTs (n=23, 50%) using different types of antibiotics while four RCTs enrolled patients given only a single type of antibiotic (9%), but the types of antibiotics were not reported in 19 (41%) trials (Table 1). Most of the antibiotics were given for upper respiratory infections (38, 83%) or mixed types of infections (5, 11%) or not reported (n=3, 6%). The duration of antibiotics ranged from 5 to 21 days and most (26, 56.5%) were given only intravenously while a few RCTs (n=3, 6.5%) enrolled patients given either IV or oral antibiotics. The antibiotic route was not reported in 17 (37%) of the RCTs.

Table 1: Study population characteristics of 46 included trials for the prevention of AAD. ¹Dose/day may vary by age of patient; ² On abx + days extended post-antibiotics; ³Li J (treatment AAD arm excluded); ⁴Yu M (two study arms excluded, on for treatment of AAD by S. boulardii and another with probiotic blend). A, ampicillin; C, cephalosporin; d, days; E, ertapenum; IV, intravenous route; L, levofloxacin; LOS, during length of stay in hospital; M, macrolide; mon, months; nr, not reported in paper; No., number enrolled; OM, otitis media; On abx, given while on antibiotics; Or, oral route; Ot, other types of antibiotics; P, penicillin; Ref, reference; RTI, upper or lower respiratory tract infection; UTI, urinary tract infection; wk, weeks; y, years old.

Intervention/control characteristics

Most trials (n=44, 96%) used open controls, while two (4%) RCTs used a blinded placebo control, as shown in Table 1 [44,63]. The most common formulation for S. boulardii was as a powder (n=41 trials, 89%) or in capsules (n=1, 2%) [72], but was not reported in four trials. Many RCTs used age-dependent daily doses for S. boulardii (n=16, 35%, ranging from 150 mg to 1 g/d), while others gave a fixed daily dose, regardless of age: 250 mg/d (n=11, 24%), 300 mg/d (n=1, 2%), 500 mg/d (n=17, 37%) or 1000 mg/d (n=1, 2%). Most trials continued the intervention for the duration of the antibiotic (n= 42, 91%), three (6%) continued the intervention after antibiotics were discontinued [44,53,63] and one RCT continued for the length of the hospitalization [43]. Only 12/46 (26%) of the RCTs followed subjects post-intervention for delayed-onset AAD (follow-up times ranged from 5-63 days), while most RCTs (n=34, 74%) did not follow subjects after antibiotics were discontinued. No loss to follow-up (attrition) was reported for 40 (87%) of the trials, while six (13%) trials reported 5-26% attrition [34,35,52,54,63,68]. In 31 (67.4%) of the trials, the interventions were started within 24 hours of the antibiotic, but the initiation time was not reported in 15 (32.6%) trials.

Primary outcome: prevention AAD

Efficacy to prevent AAD was reported all included trials (n=46), as shown in (Table 2). However, the reported definitions for AAD varied (Supplementary Table S2) with 5 (11%) not reporting how AAD was defined.  Most RCTs (n=41, 89%) did report their definition of AAD in their papers, typically representing a change in stool consistency, increased frequency of watery/loose stools for longer than one day, which was associated with antibiotic use or shortly afterwards. The incidence of AAD ranged from 3/100 to 38/100 (mean 15.0 ± 7.2/100) in S. boulardii groups compared to 9/100 to 56/100 (mean 35.9 ± 10.7/100) in controls. The pooled RR for AAD was significantly lower for S. boulardii compared to controls (RR= 0.43, 95% C.I. 0.40, 0.48, P<0.0001, I²=5.5%), as shown in (Figure 2).

Table 2: Prevention of antibiotic-associated diarrhea (AAD), C. difficile infection or adverse events in 46 included trials. AAD, antibiotic-associated diarrhea, excluding rotaviral diarrhea; AE, adverse event; CDI, Clostridioides difficile infection;  nr, not reported; Ref, reference;  S., Saccharomyces; state, only statement of ‘No adverse events noted’.

Primary outcome: Efficacy to prevent CDI

Only three trials reported CDI as an outcome (Table 2) in their trials [54,55,72]. As shown in (Figure 3), the risk for CDI was significantly reduced for S. boulardii CNCM I-745 compared to controls (RR= 0.30, 95% C.I. 0.10, 0.87, P = 0.03, I²=22.9%).

Figure 3: Forest plot of efficacy for prevention of CDI:  S. boulardii compared to controls.DL= Der Simonian-Laird.

Secondary outcomes

Several secondary outcomes had sufficient data for analysis:

daily stool frequency at study end, clinical characterization of AAD (duration, severe AAD, effectiveness rating) and safety. Other a priori secondary outcomes were infrequently reported (rural versus urban and cost-effectiveness) and could not be analyzed.

Stool frequency

Only 16/46 (35%) reported the daily stool frequency during the intervention/control phase and 30 (65%) did not report this outcome measure (Table 3). Those given S. boulardii had significantly fewer mean stools/day compared to the controls (Figure 4), SMD: --1.25 stools/day, 95% C.I. -1.35, -1.16, P< 0.0001, I²=96%).

Figure 2: Forest plot of efficacy for prevention of AAD: S. boulardii compared to controls.DL, Der Simonian-Laird.

Table 3: Efficacy of S. boulardii on prevention of secondary AAD: stool frequency and duration of AAD, and stool frequency/day, AAD, antibiotic-associated diarrhea; na, not applicable; nr, not reported; Ref, reference; Sb, Saccharomyces boulardii CNCM I-745, Std dev, standard deviation.

Figure 4: Forest plot of efficacy for stool frequency/day: S. boulardii compared to controls. SMD, standardized mean difference; IV, Inverse variance.

Clinical characteristics of AAD

Three measures were used to determine if S. boulardii influenced the severity of AAD (duration of AAD, frequency of severe AAD and effectiveness rating).  Data was collected from 232 cases of AAD in S. boulardii treated patients and 454 cases of AAD in the controls.

Clinical characterization of AAD: duration of AAD.The duration of AAD (Table 3) was reported in 36 (78.3%)  of the 46 prevention trials, while 10 (22%) of the RCTs did not report this outcome.  Mean duration of AAD ranged from 0.3 to 5 days in the S. boulardii subjects and ranged 4 to 9 days in the controls. The reduction of the days of diarrhea was significantly reduced for those given S. boulardii compared to the controls (SMD: -1.29 days, 95% C.I. -1.43, -1.15 P<0.001, I²=68.8%), as shown in (Figure 5).

Figure 5: Forest plot of efficacy for reduction of AAD duration in days: S. boulardii compared to controls. SMD, standardized mean difference; IV, inverse variance.

Clinical characterization of AAD: Incidence of severe AAD. Only 17 (37%) of the RCTs reported incidence of severe AAD (Table 4) [36,38, 40, 42, 45, 46, 48, 50, 59, 61, 63, 65, 69,71,76,77,79], while 29 (63%) did not assess the severity of AAD. Of the 17 RCTs that analyzed the severity of AAD, most (70%) reported a definition of ‘severe’ AAD (Supplementary Table S2). Treatment with S. boulardii did not significantly result in less severe cases of AAD, as shown in (Supplementary Figure S2) (RR=1.00, 95% C.I. 0.57, 1.76, P = 1.0).

Table 4: Clinical characteristics of AAD in 232 cases of AAD in S. boulardii and 454 cases of AAD in controls; %, percentage;  AAD, antibiotic-associated diarrhea; n, number in numerator; N, total number in denominator, nr, not reported.