Decrease in Frequency of Liquid Stool in Enterally Fed Critically Ill Patients Given the Multispecies Probiotic VSL#3: A Pilot Trial

This single-site, double-blind, randomized, placebo-controlled pilot study was conducted at Launceston General Hospital, a 330-bed, university-affiliated hospital in Tasmania, Australia. The hospital is a major gastrointestinal surgery referral center and accepts patients with major trauma, with the exception of patients with neurotrauma.

The study was done in the 6-bed general ICU between July 2005 and October 2006. The ICU admits approximately 500 patients per year, and about 60% to 75% receive mechanical ventilation. The unit accommodates both high-acuity patients treated with mechanical ventilation who require complex organ support and lower acuity high-dependency patients who require perioperative care. Patients with all forms of acute illnesses, except postoperative cardiothoracic and acute neurosurgical problems, are treated. The ICU is run as a “closed unit”; admission and treatment rights are limited to staff intensivists. Care was provided 24 h/d by an intensivist in conjunction with full-time junior medical staff (residents or registrars). This study was approved by the Human Research Ethics Committee (Tasmania) Network and the Griffith University Human Research Ethics Committee and was registered with the Australian Clinical Trials Registry. Written informed consent was obtained from patients or their next of kin.

The primary outcome was frequency of liquid stool (mean number of episodes per patient per day), because liquid stool is the type most closely associated with the term diarrhea. Secondary outcomes included the frequency (mean number of episodes per patient per day) of combined liquid and loose (unformed) stool and the weight (grams per day) of liquid stool and of liquid and loose (unformed) stool (per patient per day).

This study was intended as a pilot for a larger multicentered study. Calculation of the sample size indicated that 150 patients per group would be required to detect a 30% reduction in episodes of liquid stools and that 64 patients per group would be required to detect a 40% reduction with a 80% power if the episode rate is 1.01 per day (SD, 0.75) as observed in the first 15 patients in the pilot study.

All adult patients (18 years or older) who required enteral nutrition delivered through a feeding tube were considered for inclusion in the study if the attending intensivist expected that artificial nutritional support would be required for more than 72 hours and no contraindication to use of the gastrointestinal tract was present.

Exclusion criteria included expected death within 24 hours of admission, established treatment limitations, an expected interhospital transfer within 72 hours of admission, a specific surgical request not to use the gastrointestinal tract (eg, enteric anastomosis, imminent bowel resection or endoscopy), ischemic bowel, enteric fistula, bowel obstruction, and exacerbated inflammatory bowel disease. Additionally, patients were excluded if their treatment included established or recently commenced (<48 hours) parenteral nutrition and they had been started on enteral tube feeding before ICU admission. Because of the confounding effect on stool weight and consistency, patients were also excluded if their initial illness was gastrointestinal hemorrhage or diarrhea or was complicated by these 2 entities. In patients with no contraindication, enteral nutrition was started within 24 hours of ICU admission.

For randomization of patients to the 2 groups, a hospital pharmacist not involved in recruitment, patient care, or data collection used Microsoft Excel’s pseudorandom number generator. Assignments to the intervention or placebo groups were recorded in sequentially numbered, sealed opaque envelopes. All ICU staff, investigators, study patients, and the patients’ families had no knowledge of group allocation.

The type of enteral nutrition administered was standardized; all patients received Isosource or Renal or Diabetic Resource (Novartis, Melbourne, Australia) formulas if clinically indicated. All enteral nutrition was administered via a nasoorogastric or nasojejunostomy tube according to the ICU “closed” enteral feeding protocol. Enteral feedings were started at 20 mL/h and were increased 20 mL/h every 4 hours until the target dosing rate was achieved. In instances in which gastric residual volumes exceeded 150 mL, intravenous prokinetic agents were introduced with feeding and were continued until individual target rates for enteral feeding were achieved. The target feeding rate was individualized for each patient and was based on weight and adjusted for metabolic demands; typical target rates were 25 to 35 cal/kg per day and 0.8 to 1.5 g protein per kilogram per day.

Both the control and the intervention groups received a nutritional supplement (Sustagen, Mead Johnson Co, Melbourne, Australia); the intervention group also received VSL#3. All enteral formulas and the nutritional supplement were free of fiber and prebiotic additives. The nutritional supplement was used to reconstitute the live freeze-dried VSL#3 product in a liquid form so that both preparations (placebo and intervention) appeared identical and had a similar viscosity. VSL#3 (VSL Pharmaceuticals, Gaithersburg, Maryland) contains live freeze-dried lactic acid bacteria, and each dose contains approximately 450 billion live lactic acid bacteria in defined ratios of lyophilized Bifidobacterium breve, Bifidobacterium longum (>10 × 10⁹/g), Bifidobacterium infantis (>10 × 10⁹/g), L acidophilus, Lactobacillus plantarum, Lactobacillus casei, L bulgaricus, Streptococcus thermophilus (>100 × 10⁹/g). The exact composition of the bacteria of the VSL#3 product is not specified by the distributor (Orphan Australia Pty Ltd, Dandenong, Victoria, Australia), and the bacterial count of the strains listed without numbers may vary between batches. The probiotic preparation was prepared in the pharmacy. One VSL#3 sachet was combined with 50 mL of vanilla-flavored nutritional supplement (Sustagen) and mixed into a suspension. The placebo suspension of the supplement was indistinguishable from the VSL#3 in terms of color, consistency, and taste. The mixture of VSL#3 and supplement was tested to ensure it maintained its viscosity, osmolality, and taste for up to 24 hours while refrigerated.

The intervention and placebo preparations were made up daily and used to fill 50-mL opaque syringes. Each day, 2 syringes identified by study number, date, and preparation time were delivered to the ICU and stored in the refrigerator. The probiotic preparation of VSL#3 or the placebo was administered to each study patient twice per day, approximately 12 hours apart, as a bolus. Administration was recorded in each patient’s medication record by attending nurses. After each administration, the feeding tube was flushed with 30 mL of sterile water to maintain patency of the tube.

All fecal output was examined and was recorded by using the validated King’s Stool Chart.⁴² The chart uses a scoring system that involves characteristics of fecal frequency, weight, and consistency based on 3 categories of fecal weight and 4 categories of fecal consistency, with the aim of standardizing fecal output.⁴² A comprehensive staff training program including the use of the stool chart was undertaken.

The study continued after the cessation of tube feeding; patients in whom tube feeding was stopped consumed the intervention or placebo preparation orally. Follow-up continued for 21 days after transfer from the ICU to the general unit or until discharge from the hospital.

Demographic and clinical data were also collected. All patients had daily biochemistry tests, including levels of serum albumin and total protein, and hematology profiles, including coagulation studies, as part of routine clinical management. Stool cultures and assays for C difficile toxins A and B were performed routinely for all patients who had diarrhea.

Analysis was on an intention-to-treat basis. Demographic descriptors of the study population in the intervention and placebo groups were compared by using general linear modeling for continuous variables and logistic regression for categorical variables to determine randomization effectiveness for known or potential confounding variables. The absolute (per patient per day) and relative (incidence rate ratio [IRR]) incidence rates of episodes of liquid stools and of liquid or loose stools in each group were estimated by using repeated-measures Poisson regression. The daily absolute weight of stool and mean differences for liquid and liquid and loose (unformed) stools were estimated by using repeated-measures general linear modeling. Results are expressed as absolute and relative effect sizes with 95% confidence intervals (95% CI) of these estimates and P values for the comparisons. The analyses were also performed with adjustments for potential confounding variables; variables in the models were selected by using backward stepwise Poisson regression and general linear modeling, respectively, from a list of antibiotics (β-lactams, clindamycin, macrolides, ciprofloxacin, tetracyclines, cotrimoxazole, gentamicin, vancomycin, rifampicin, metronidazole, antifungals, and antivirals), age, sex, Acute Physiology and Chronic Health Evaluation II score and Simplified Acute Physiology Score II, and serum albumin levels, with forced inclusion of treatment group and variables with significance in models of other stool consistencies. Statistical analyses were performed by using STATA Statistics/Data Analysis, Version 10.1 (StataCorp, College Station, Texas).

Of 457 patients evaluated for potential enrollment, 407 did not fulfill the inclusion criteria, and 5 who did fulfill the criteria refused consent. A total of 45 patients were randomly assigned to the intervention (n = 20) or control (n = 25) group (see FigureF1). We had no reported or detected deviations from the study protocol, and all patients had follow-up. The pilot study was stopped after 15 months because of slower than expected recruitment and because a planned a priori interim analysis after the enrolment of 30 patients indicated a significant effect on the primary outcome (mean number of episodes of liquid and combined liquid and loose stool per patient per day).

The demographic and clinical characteristics of the 2 groups did not differ significantly (Table 1). Patients were in the study for mean of 11.9 days (SD, 5.6) and received enteral nutrition for a mean of 8.5 days (SD, 5.4). The 2 groups did not differ significantly in type of feeding tubes; 2 of 20 (10%) in the intervention group and 4 of 25 (16%) in the control group had nasojejunal tubes; the remaining patients had nasogastric tubes.

Compared with the control group, the intervention group had a reduction in the frequency of liquid stools (IRR, 0.50; 95% CI, 0.27 to 0.93; P = .03) and of combined liquid and loose stools (unformed) (Table 2). The 2 groups did not differ significantly in stool weight (Table 2). When these analyses were adjusted for potentially confounding variables, the variance of the results was reduced, and the clinically significant differences in the effect of VSL#3 on measures of stool frequency and consistency were also statistically significant (Table 3).

All patients received intravenous antibiotics of some form during the study period. We found significant positive associations between frequency of loose stool and certain antibiotics: β-lactams (IRR, 2.04; P = .001), clindamycin (IRR, 3.41; P < .001), macrolides (erythromycin or azithromycin; IRR, 2.62; P < .001), and antifungals (IRR, 2.09; P = .003). We found negative associations for serum level of albumin (IRR, 0.72; P = .004) and use of jejunal feeding tubes (IRR, 0.78; P = .02; Table 4). No association was found between metoclopramide use and the frequency of liquid stools (IRR, 1.06; 95% CI, 0.64 to 1.75; P = .82) or liquid stool weight (mean difference, 0 g; 95% CI, −45 to 42; P = .96).

A total of 8 of the 45 patients died: 5 in the intervention group and 3 in the placebo group. Survival analysis indicated no significant difference in those mortality rates (hazard ratio, 1.22 for VSL#3 vs placebo; 95% CI, 0.31 to 4.83; P = .78; adjusted for age, use of nasojejunal tube, serum albumin level less than 25 g/L, and alternative feeding type).

In this small double-blind, randomized, placebo-controlled pilot study, use of the probiotic VSL#3 was associated with a 50% reduction in the frequency (rate per day) of liquid stool in ICU patients expected to receive enteral nutrition for at least 72 hours. This outcome is similar to the results of another study³⁰ on diarrhea associated with enteral feeding with a comparable sample and method. The outcome is also comparable with the findings of other RCT studies in different groups of patients.^23,40 

We intended to recruit 150 patients, but recruitment (n = 45) was much slower than expected; because of changes in surgical coverage, fewer surgical patients were available who required enteral feeding. On the basis of our results, a sample size of 45 patients had an 80% power to detect a 47% reduction in frequency of liquid stools, and 99 patients in each group would be required to detect a 30% reduction in the frequency of liquid stools.

Although we found that the frequency of liquid stool decreased significantly, the absolute volume of stool in both groups was not particularly large. Some clinicians may not classify the amounts as diarrhea, a situation that highlights difficulties with current definitions of diarrhea. For example, the World Health Organization⁴³ defines diarrhea as 3 or more loose or watery stools in a 24-hour period; no amount is specified. No single standard definition of the term diarrhea exists; Lebak et al⁴⁴ found more than 33 different definitions in the literature. Criteria included in definitions of diarrhea include characteristics (loose/watery, formed/semiformed/nonformed) and frequency (>3 stools/day, >4 stools/day of stools passed with in a specific time frame [<24 hours]).² Other definitions include the previously mentioned criteria plus the volume of stool passed per episode (>200 g/d of liquid stool) and combinations of any of the criteria described here. A strength of our study is that we used a validated tool to quantify the fecal output of all patients by measuring the 3 dimensions of stool output: frequency, consistency, and weight of stool.

All patients were routinely tested for C difficile. Although reported rates of infection caused by C difficile are as high as 20% in enterally fed patients,³³ no episodes of C difficile infection occurred in either group of patients during the study. However, because C difficile infection can develop up to 4 weeks after the cessation of antimicrobial therapy, some undetected positive associations between C difficile and diarrhea may have occurred. The apparent absence of C difficile infection in this study is comparable to the finding in a recent RCT²³ in which a probiotic was administered to reduce the incidence of diarrhea associated with use of antibiotics and infection by C difficile. In that study,²³ the intervention group had no episodes of C difficile infection, whereas 17% of the placebo group did.

During the study, prokinetics were prescribed for patients who were experiencing high gastric residual volumes of enteral feeding. Use of metoclopramide was not associated with the frequency and weight of liquid stools. Erythromycin was associated with increased frequency and weight of liquid stools, but we could not determine whether this finding was due to the prokinetic effect of the drug or to its antibiotic action. Nasojejunal placement of the feeding tube and serum level of albumin (<25 g/L) were associated with increased frequency and weight of liquid stool as expected,^33,36 but because of our small sample numbers, we could not analyze this interaction.

Any broad-spectrum antibiotic, such as cephalosporins, penicillin, and clindamycin,⁴⁵ can predispose patients to diarrhea. Most patients in our study received intravenous antibiotics. We found significant positive associations between certain antibiotics (β-lactams, clindamycin, erythromycin, and antifungals) and the frequency and weight of liquid and liquid and loose (unformed) stools. The multivariate analysis indicated that those antibiotics were associated with increased likelihood of episodes of liquid and loose stool and that the effect of VSL#3 in reducing that likelihood was independent of such confounding effects. Although most of our patients received intravenous antibiotics in the ICU, we did not collect data on use of oral antibiotics, a step that might have been useful. Most likely oral antibiotics produce higher local gastrointestinal concentrations of antibiotic than that achieved by intravenous administration. Thus, compared with intravenous antibiotics, oral antibiotics might attenuate the beneficial effects of probiotics.

Many probiotic preparations are available, with different single or multiple species. The potential of improved synergy or the ability of multistrain probiotics to work together to enhance intestinal health may provide a rational basis for future research and potential clinical use,⁴⁶ although any therapeutic hypotheses would need to be tested in RCTs. The large numbers of potential single and multiple combinations of species and dosages create a daunting volume of empirical trial data that needs collecting.

The probiotic we used, VSL#3, is a commercially available, high-potency preparation containing 8 different bacterial species that have been effective in the management of serious intestinal disorders in patients with gastrointestinal disease.^17,–21 VSL#3 survives gastric fluid, bile, and pancreatic secretions and can effectively colonize the gastrointestinal tract, regulate intestinal immunomodulary responses, and restore healthy levels of resident intestinal bacteria.^18,21,47 Although we found that VSL#3 reduced the number of episodes of liquid stools in ICU patients receiving enteral feeding, we cannot extrapolate and suggest which species in the preparation might be more efficacious than another species or whether the combined species might have worked synergistically to reduce the numbers of liquid stools. The reduction in the episodes of liquid stool in our study is similar to the results of a recent study²³ in which a multistrain probiotic drink was used to reduce the incidence of antibiotic-associated diarrhea.

Use of a high-potency multi-species probiotic in ICU patients receiving enteral feedings who have diarrhea is extremely limited. Thus, our findings are a proof of concept: VSL#3 can alter the risk of diarrhea in enterally fed ICU patients. The choice of probiotic preparation most likely is an important consideration for future research. For example, some preparations may actually contain inactive organisms, whereas others may have organisms that do not survive transit through the gut. Multiple RCTs with VSL#3 should be followed by similar studies on different preparations to determine the relative effectiveness of each preparation in specific health conditions.

Our study has limitations that may narrow the generalizability of the findings. The study was undertaken at a single site, with a limited case mix, and some important categories of critically ill patients were excluded from the sample, in particular neurosurgical patients. Only about 10% of ICU admissions were eligible, producing a small sample size. Despite the study limitations, particularly the high severity-of-illness scores of the patients in our sample, and given the potential beneficial effects of probiotics on diarrhea, further investigation is required. Although we attempted to classify our patient groups (eg, operative vs nonoperative), because of the small sample size, potential subgroups of patients who might benefit more than other patients could not be identified. Another limitation was use of a nutritional supplement as the placebo and as the preparation used to reconstitute the freeze-dried VSL#3 to a liquid form. Although the 2 main types of enteral nutrition used to feed patients (Diabetic Resource and Isosource) were relatively iso-osmolar (300–490 mOsm), retrospectively, we discovered that the supplement is hyperosmolar (671 mOsm) and therefore may be a confounder because of its high osmolarity and potential to cause diarrhea.

Although our data suggest that VSL#3 reduced the frequency of liquid stool in critically ill patients, the study was not designed or large enough to assess the potential detrimental effects attributable to VSL#3 (eg, secondary sepsis or impaired function of the gastrointestinal tract). The heterogeneous underlying illnesses of the patients further limited our ability to assess mortality. Although we found no significant differences in mortality between the groups, the wide confidence intervals of the hazard ratio estimates preclude any firm conclusions about the safety of the use of the preparation. We did not monitor directly for adverse effects, because monitoring would have required vastly larger numbers of patients to exclude clinically relevant but uncommon adverse effects. The lack of monitoring limits the scope of our study. Potential risks include superinfection or breakthrough invasion into the blood with individual probiotic strains; these problems would be most likely in patients who are immunocompromised^48,49 or have severe acute pancreatitis.⁵⁰ In our study, all patients with liquid or soft stools had cultures of blood, urine, and sputum and assays for C difficile infection 3 times per week as routine monitoring for sources of infection. We found no cultures positive for individual probiotic strains.

Finally, measurement of the outcome relied on direct observation of stool by nursing staff. Although staff members were trained for this task, and periodic compliance checks were undertaken to ensure adherence, some episodes of diarrhea could have been missed or recorded incorrectly.

Patients with diarrhea who were given a probiotic preparation had a 50% reduction in the number of episodes of liquid stool per day and smaller, but still clinically significant, reductions in the number of episodes of liquid and loose (unformed) stool and in stool weight. Diarrhea remains an important problem for both patients and health care providers. Probiotics used in the correct context may be an appropriate response to the problem of diarrhea, and further investigation is warranted. The balance of beneficial and adverse effects in patients who might receive this treatment can be reliably estimated only by doing much larger scale RCTs and meta-analysis of multiple RCTs.

This study was undertaken at Launceston General Hospital as part of a Griffith University postgraduate degree. We thank the nursing, medical, and laboratory personnel; clinical pharmacists; and the Ramsay Library technicians at Launceston General Hospital for their assistance. Thank you to Kevin Whelan, coauthor of the King’s Stool Chart, for permission to use the validated measurement tool, which can be accessed at www.kcl.ac.uk/stoolchart.