The Effect of Saccharomyces boulardii Primary Prevention on Risk of Hospital-onset Clostridioides difficile Infection in Hospitalized Patients Administered Antibiotics Frequently Associated With C. difficile Infection

Abstract

Background

Hospital-onset Clostridioides difficile infection (HO-CDI) is a costly problem leading to readmissions, morbidity, and mortality. We evaluated the effect of a single probiotic strain, Saccharomyces boulardii, at a standardized dose on the risk of HO-CDI within hospitalized patients administered antibiotics frequently associated with HO-CDI.

Methods

This retrospective cohort study merged hospital prescribing data with HO-CDI case data. The study assessed patients hospitalized from January 2016 through March 2017 who were administered at least 1 dose of an antibiotic frequently associated with HO-CDI during hospitalization. Associations between S. boulardii administration, including timing, and HO-CDI incidence were evaluated by multivariable logistic regression.

Results

The study included 8763 patients. HO-CDI incidence was 0.66% in the overall cohort. HO-CDI incidence was 0.56% and 0.82% among patients coadministered S. boulardii with antibiotics and not coadministered S. boulardii, respectively. In adjusted analysis, patients coadministered S. boulardii had a reduced risk of HO-CDI (odds ratio [OR], 0.57 [95% confidence interval {CI}, .33–.96]; P = .04) compared to patients not coadministered S. boulardii. Patients coadministered S. boulardii within 24 hours of antibiotic start demonstrated a reduced risk of HO-CDI (OR, 0.47 [95% CI, .23–.97]; P = .04) compared to those coadministered S. boulardii after 24 hours of antibiotic start.

Conclusions

Saccharomyces boulardii administered to hospitalized patients prescribed antibiotics frequently linked with HO-CDI was associated with a reduced risk of HO-CDI.

(See the Editorial Commentary by McFarland on pages e2519–20.)

Clostridioides difficile infection (CDI) has become one of the most common healthcare-associated infections in the United States, with the incidence nearly doubling between 2001 and 2010 [1, 2]. The Centers for Disease Control and Prevention (CDC) considers CDI an urgent threat requiring prevention and monitoring [1]. Risk factors for CDI include increasing age, proton pump inhibitor (PPI) use, and most significantly, broad-spectrum antibiotics that are hypothesized to accelerate C. difficile colonization by reducing levels of beneficial bacteria that serve as barriers to infection [3]. Coadministering probiotics with antibiotics may prevent CDI development by reinforcing the barrier of good bacteria lost through antibiotic administration [4, 5].

Beyond restoring altered intestinal microflora, administering probiotics may stimulate the immune system to prevent pathogen adhesion and invasion, and clear pathogens and toxins from the intestinal tract [6]. Saccharomyces boulardii, a specific yeast-derived probiotic, may prevent CDI by inducing direct inhibitory actions against C. difficile toxins [5, 6]. Two studies demonstrate upregulation of total and specific intestinal antibodies (immunoglobulin A) to toxin A in response to S. boulardii exposure, consequently reducing C. difficile pathogenicity [7, 8]. Other evidence demonstrates that S. boulardii directly inhibits C. difficile toxin by releasing a protease to hydrolyze C. difficile toxins [6, 9].

The available literature examining the clinical outcomes of probiotics in preventing hospital-onset C. difficile infection (HO-CDI) is contentious. Three randomized clinical trials [10–12] and 1 cohort study [13] demonstrate S. boulardii’s primary or secondary prevention effects on HO-CDI. In contrast, the largest randomized trial of probiotics to date [14] and 1 cohort study of S. boulardii [15] show no protective associations. Overall these studies are small in scale and do not take into account the timing of probiotic initiation relative to antibiotic start [5, 11, 15]. A meta-analysis found that administering probiotics >2 days after initiating antibiotics significantly reduced the efficacy by half [16]. Given mixed evidence, research needs to assess the extent to which administrative timing impacts probiotics’ beneficial effects.

Given the limitations of previous studies, the Infectious Diseases Society of America (IDSA) and the Society of Healthcare Epidemiology of America (SHEA) state that “there are insufficient data at this time to recommend administration of probiotics for primary prevention of CDI” [17]. The IDSA and SHEA guideline writers cite specific limitations such as studies including patients with abnormally high rates of CDI, using inconsistent probiotic formulations across studies, or assessing patient populations at low risk for CDI [17]. These limitations underscore the need for further research. The guideline recommendations from IDSA and SHEA conclude with suggested areas of further research including: (1) What preventive measures can be taken to reduce the incidence of CDI? (2) Can administration of probiotics effectively prevent CDI?

Our primary objective was to evaluate the effect of prescribing a single probiotic formulation S. boulardii on risk of developing HO-CDI in a large cohort of hospitalized patients receiving antibiotics frequently associated with HO-CDI. Our secondary objective was to evaluate the extent to which timing of S. boulardii initiation relative to antibiotic start affects HO-CDI risk.

METHODS

This retrospective observational cohort study compares the risk of HO-CDI in hospitalized patients who only received antibiotics frequently associated with HO-CDI vs those who received antibiotics in combination with S. boulardii. The study merged C. difficile case data with medication administration records to evaluate associations between HO-CDI incidence and S. boulardii administration occurring between 1 January 2016 and 31 March 2017. The study setting is a 220-bed level 2 trauma center nonacademic hospital.

In December 2015, our institution established S. boulardii at a dose of 500 mg twice daily as the only formulary probiotic. This decision resulted from an internally performed literature review of probiotic agents by the Pharmacy and Therapeutics Committee. The committee concluded from the literature reviewed that the use of S. boulardii at a dose of 500 mg twice daily was more often associated with positive outcomes for reducing antibiotic-associated diarrhea and CDI. Corresponding with the formulary change, an electronic pop-up box was added to the physician electronic order entry system with an option to order S. boulardii. The pop-up box appeared following the entry of an antibiotic order for β-lactams, fluoroquinolones, or lincosamides. The S. boulardii order was not an automatic reflex order. The pop-up box also listed precautions for use of S. boulardii, including (1) a history of organ transplantation with current receipt of antirejection medication; (2) concomitant receipt of chemotherapy or radiation; (3) low neutrophil count; (4) diagnosis of AIDS; or (5) active gastrointestinal ulcer.

The cohort included adult patients (1) admitted to an inpatient medical unit with an average length of stay equaling ≥3 days; and (2) having barcode administration evidence for reception of at least 1 dose of an antibiotic frequently associated with HO-CDI, which was defined as clindamycin, fluoroquinolones, third- and later-generation cephalosporins, carbapenems, and penicillins [18–20]. For the purpose of this study, “antibiotic” will subsequently refer to these defined antibiotics frequently associated with HO-CDI. The unit of analysis was defined as the first hospitalization during the study period. Only the first hospitalization per unique patient was included in the analytic dataset to remove within-patient changes in S. boulardii exposure across separate hospitalizations.

Patients were classified as either having been or not having been administered S. boulardii during hospitalization, based upon having barcode administration evidence for reception of at least 1 dose of S. boulardii, or no evidence of barcode scan for administration, respectively. Saccharomyces boulardii (Florastor Daily Probiotic Supplement, Biocodex, Redwood City, California) administration consisted of two 250-mg capsules by mouth twice daily. Each 250 mg capsule contains 5 billion colony-forming units (CFUs) for a total daily administration of 20 billion CFUs. The dose of 20 billion CFUs is consistent with previously published studies assessing CDI primary prevention [21].

The study used C. difficile case data defined by and reported to the National Health Safety Network from our institution’s infection control office. HO-CDI cases were defined as positive if an unformed stool specimen tested positive for C. difficile ≥3 days after admission and >8 weeks from a previous positive specimen result, consistent with the CDC definition for HO-CDI [22–23]. Cases were laboratory confirmed by polymerase chain reaction for the gene encoding toxin B alone without a reflex algorithm. Patients required the following criteria prior to testing: (1) ≥3 loose stools within 24 hours; (2) no laxative/stool softeners within 48 hours; (3) no positive C. difficile test in the last 30 days; and (4) no negative C. difficile test in the last 7 days. Samples were required to be watery; if the stool sample received by the laboratory was formed or semiformed, the laboratory rejected the sample.

To evaluate the effect of S. boulardii administration on HO-CDI risk, we conducted our first multivariable logistic regression to estimate the risk of HO-CDI during a hospitalization conditional upon the coadministration of S. boulardii with antibiotic(s) vs antibiotic(s) without S. boulardii. To account for both potential confounders and selection bias, the model included propensity scores generated by a separate multivariable logistic model testing the likelihood of patients receiving S. boulardii conditional upon: (1) antibiotic(s) administered during hospital admission; (2) metronidazole administration during hospitalization but >48 hours prior to CDI diagnosis; (3) PPI administration at any time during hospitalization prior to CDI diagnosis; (4) intensive care unit (ICU) admission; (5) patient age of 65 years or greater; and (6) gender. Each hospitalization was assigned a propensity score used as a covariate in the final multivariate model.

To evaluate the extent to which the timing of S. boulardii initiation relative to the first dose of antibiotic affected HO-CDI risk, we conducted a second multivariable logistic regression among only the subgroup of patients receiving S. boulardii. This model tested the risk of HO-CDI conditional upon early vs late coadministration of S. boulardii with antibiotics. “Early” was defined as S. boulardii administered within 24 hours of first antibiotic dose administration; “late” was defined as S. boulardii administered ≥24 hours after the first antibiotic dose. Similar to the model run in the full cohort, propensity scores for this second analysis were generated with a multivariable logistic model testing the likelihood of patients receiving early vs late S. boulardii conditional upon the same baseline covariates included in the primary model run in the full cohort. Each hospitalization was assigned a propensity score used as a covariate in the final multivariable model. Receiver operating characteristic curves were used to calculate C-statistics to estimate the overall global fit for all multivariable logistic regressions, including those used to generate propensity scores. All statistical analyses were conducted using SPSS version 24.0 software (IBM SPSS, Chicago, Illinois).

RESULTS

A total of 8763 patients administered at least 1 dose of an antibiotic frequently associated with HO-CDI were assessed. The cohort was 39% male and averaged 64 years of age. Patients coadministered S. boulardii and antibiotics were more often male (P < .0001) and ≥65 years of age (P < .0001) compared to patients only administered antibiotics. Ceftriaxone, piperacillin-tazobactam, levofloxacin, and ciprofloxacin were the most frequently administered antibiotics, administered in 44%, 32%, 25%, and 21% of patients, respectively. Carbapenem, fluoroquinolone, and cephalosporin administration was significantly higher in patients coadministered S. boulardii compared with those not coadministered S. boulardii (P < .0001). PPIs were administered in 46% of patients. PPI administration was significantly higher in patients coadministered S. boulardii and antibiotics (50%) compared with patients not coadministered S. boulardii (41%) (P < .0001; Table 1).

The overall incidence of HO-CDI during the study period was 0.66%. Patients admitted to the ICU and patients administered PPIs demonstrated a higher incidence of and risk for HO-CDI (Tables 2 and 3). The incidence of HO-CDI was lower in patients coadministered S. boulardii and antibiotics (0.56%) compared to patients administered antibiotics without S. boulardii (0.82%). With respect to administration timing, the incidence of HO-CDI in patients receiving early S. boulardii was less than half of those patients administered antibiotics alone without S. boulardii (0.38% vs 0.82%) (Table 2).

When adjusting for possible confounders and selection bias using the propensity score, the risk for HO-CDI was significantly less in patients administered S. boulardii and antibiotics (odds ratio [OR], 0.57 [95% confidence interval {CI}, .33–.96]) (Table 4) compared to patients administered antibiotics alone. Early S. boulardii administration displayed a stronger HO-CDI preventive effect compared to late S. boulardii administration (OR, 0.47 [95% CI, .23–.97]) (Table 5).

DISCUSSION

When adjusted for potential confounders, patients administered S. boulardii in conjunction with antibiotics frequently associated with HO-CDI had a lower risk of developing HO-CDI compared with patients not administered S. boulardii in this single-center study. The protective effect was more pronounced in the early S. boulardii subgroup vs the late S. boulardii subgroup, suggesting that early administration might offer greater HO-CDI risk reduction compared to late administration. A recent meta-analysis on the use of probiotics for primary prevention of HO-CDI revealed similar findings alluding to the importance of probiotic initiation timing relative to the first dose of antibiotic [16]. The meta-analysis observed that the preventive effect was limited to probiotic initiation within 1–2 days of antibiotic start and a lack of benefit when administered outside of 2 days.

To explore this idea further, we conducted a post hoc multivariable logistic regression conditional upon the coadministration of early S. boulardii vs no S. boulardii with antibiotics. The analysis resulted in a reduced odds risk of HO-CDI for early S. boulardii vs no S. boulardii (OR, 0.44 [95% CI, .23–.84]; P = .013; C-statistic = 0.601), demonstrating a consistent stronger effect for early S. boulardii and HO-CDI risk reduction. Presumptively, an earlier initiation of a preventive therapy would have a greater effect than a preventive therapy started further from the inciting event. Here we provide data to support that assumption with S. boulardii administration relative to antibiotic initiation. The secondary analysis included a small number of events and therefore is considered a provisional finding requiring further investigation.

Our study has several limitations. First, our sample originates from a single medical center and decreases the generalizability of the findings. Second, we were unable to assess preexisting conditions, such as immunosuppression and CDI history, or patient severity status as risk factors for HO-CDI within the study or provider-level preventive practices. To mitigate this limitation, we included ICU admission as a confounder to account for patients at a higher acuity level with severe illness. Third, providers were cautioned to avoid prescribing probiotics in patients (1) with a history of organ transplantation currently receiving antirejection medication; (2) receiving chemotherapy or radiation; (3) with low neutrophil count; (4) diagnosed with AIDS; or (5) having an active gastrointestinal ulcer [24–26]. There was no system in place to ensure S. boulardii was not prescribed to a patient who met these criteria. Consequently, we cannot make conclusions regarding the probiotic effectiveness in these patient groups.

This study has several strengths. First, including propensity scores in our models helped us account for both confounding and selection bias. The propensity score helps account for observable risk factors either associated with an increase or decreased risk of HO-CDI. In our analysis, patients administered S. boulardii could be considered at higher risk for HO-CDI as they were more often ≥65 years of age, administered a PPI during hospitalization at a higher rate, and more likely to receive a third- or fourth-generation cephalosporin, carbapenem, and fluoroquinolone than non–S. boulardii recipients. Adding age, PPI, and antibiotic class utilization as covariates in the multivariable model used to derive the propensity score allows the propensity score to account for all of these factors simultaneously. Furthermore, the propensity score helps account for underlying factors driving selection bias that could not be measured in our dataset, including, but not necessarily limited to, patient history of HO-CDI or provider-level behaviors. Second, our observational study design enables us to measure real-world effectiveness based upon daily clinical encounters in contrast to clinical trials that may not reflect outcomes or patients seen in everyday practice. Evidence from real-world practice can better inform changes in clinical guidelines or formulary policies. Third, isolating our study to a single agent at a set dose improves our ability to determine the effects of uniform doses and inform specific dose recommendations. In contrast, previous studies included multiple probiotics in the same treatment group and could not tease out the effects of isolated agents or dosages. Fourth, we conducted a subgroup analysis to explore the importance of S. boulardii administration timing relative to antibiotic start and resultant HO-CDI incidence. This subgroup analysis suggests that earlier probiotic administration timing improves probiotic primary prevention efficacy. Finally, our point estimate of OR = 0.57 is similar with those found in meta-analyses by Johnston et al (Relative Risk [RR], 0.34 [95% CI, .24–.49]) [27], Goldenberg et al (RR, 0.36 [95% CI, .26–.51]) [28], Lau et al (RR, 0.40 [95% CI, .29–.53]) [29], and Shen et al (RR, 0.42 [95% CI, .30–.57]) [16]. Similar results across multiple studies further strengthens confidence in our estimates.

We observed a number needed to treat for early S. boulardii administration of 228, amounting to 17 cases of HO-CDI that may have been prevented among the 3936 patients administered S. boulardii within 24 hours of antibiotic start during the study period. Given the magnitude of benefit and the relatively low cost of probiotics compared to the high cost of treating CDIs during hospitalization and postdischarge, adding S. boulardii probiotic as a primary preventive strategy may be a valuable strategy to reduce HO-CDI. Furthermore, results from a cost-effectiveness analysis suggest that probiotic use is a cost-effective strategy in preventing CDI in hospitalized adults receiving antibiotics [30].

Conclusions

This study provides evidence to support prescribing a uniform probiotic formulation as primary prophylaxis of HO-CDI in a high-risk population, a noted weakness in the existing literature [17]. We observed that coadministration of probiotic S. boulardii at a dose of 20 billion CFUs per day with antibiotic therapy frequently associated with HO-CDI in an older hospitalized patient population significantly reduced the incidence of HO-CDI. Furthermore, initiating probiotics within 24 hours of antibiotic start appears to offer a more pronounced barrier to HO-CDI development. Our findings in conjunction with evidence from previous studies support continued exploration of probiotics for primary prevention of HO-CDI, with particular consideration of earlier probiotic administration timing.

Notes

Disclaimer. The views expressed in this publication represent those of the authors and do not necessarily represent the official views of Hospital Corporation of America (HCA) or any of its affiliated entities.

Financial support. This research was supported in part by HCA and/or an HCA-affiliated entity.

Potential conflicts of interest. The authors: No reported conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest.

References