Methods

Top Methods Results Discussion Author & Article Info References

Eligibility

Hospitalized adult patients ( 18 years of age) who began receiving nutritional support solely through tube feeding on the acute and critical care units at the Veterans Affairs Medical Center in Minneapolis (MVAMC) were eligible for our study and were matched with non-tube-fed patients. A tube-fed patient was matched with a non-tube-fed patient if the two patients resided on the same unit, were the same sex, were the same age (± 10 years), were hospitalized at the MVAMC before study entry and within 4 weeks of one another, had the same severity of illness index score [11], and had the same MVAMC nursing care score (± 1). The duration of surveillance of each non-tube-fed patient during the study was at least 4 days and was at least 80% of the duration of monitoring of the matched tube-fed patient. If a patient changed units, the unit where he or she stayed the longest was used for matching purposes. In determining the severity of illness index score, C. difficile colonization and C. difficile-associated diarrhea were not considered. Patients were recruited between March 1992 and October 1993.

Patients were excluded from participation if they had an active psychosis, had received repeated treatment with sodium polystyrene sulfonate (Kayexalate, Sanofi Pharmaceuticals, New York, New York), or had documented positive C. difficile culture or cytotoxin assay results within 1 year of the study. If a patient had a positive C. difficile culture or cytotoxin assay result at study entry, both that patient and the patient matched to that patient were excluded from further analysis (Figure 1). The MVAMC and University of Minnesota human subjects review committees approved our study.

View larger version (17K): [in this window] [in a new window]   Figure 1. Enrollment and outcomes of Clostridium difficile surveillance in acutely ill tube-fed patients and a matched cohort of non-tube-fed patients.

Patient Enrollment

We recruited 100 tube-fed patients into the study. Eight of these patients could not be matched with non-tube-fed patients, primarily because of long duration of tube feeding in an intensive or transitional care unit (range, 31 days to 77 days); 1 of the 8 could not be matched because of his young age (42 years) and duration of tube feeding in the medical intensive care unit (12 days). Data from these 8 patients were not included in the analysis (Figure 1). Nine tube-fed patients and 7 non-tube-fed patients had positive C. difficile cultures at baseline. These patients and the patients with whom they were matched were excluded from the analysis according to our exclusion criteria. The sample for analysis thus consisted of 76 tube-fed patients (74 men and 2 women) and 76 non-tube-fed patients (all men) (Figure 1).

Surveillance Procedures

We prospectively monitored tube-fed patients for C. difficile by obtaining rectal swab cultures (Baxter Diagnostics, Inc., Deerfield, Illinois) or stool specimen cultures at the following surveillance time points: before the start of tube feeding (baseline), at weekly intervals during tube feeding, and at cessation of tube feeding or before discharge from the MVAMC (if tube feeding was continued). In patients with an enterostomy, we obtained a rectal swab culture at baseline and at the end of the study for comparison with weekly cultures of enterostomy effluent. If patients who received tube feeding over the long term ( 20 days) received trials of the introduction of oral foods during tube feeding and the trials lasted for more than 2 weeks, surveillance for C. difficile was discontinued. Non-tube-fed patients were similarly monitored for C. difficile: Rectal swab cultures were obtained at the beginning of the study protocol, at weekly intervals throughout the study, and at the end of the surveillance period.

Stool frequency and consistency and the amount of enterostomy effluent were recorded daily for all patients. Stool consistency was categorized as hard and formed, soft but formed, unformed or loose, or liquid. Diarrhea was defined as at least three unformed or liquid stools per day, at least 250 mL of liquid stool per day collected in a rectal or colostomy bag, or at least 1500 mL of effluent per day collected in an ileostomy bag [4, 12]. Episodes of diarrhea induced by medical therapy [such as enemas or preoperative bowel catharsis] occurred in 23 patients and were excluded from the analysis. If a patient had diarrhea, we obtained stool specimens for C. difficile culture, alcohol shock C. difficile culture, and cytotoxin assay every 48 hours until a positive culture or cytotoxin assay result was seen or until diarrhea ceased. A patient was considered to have C. difficile-associated diarrhea when 1) he or she had a positive cytotoxin assay result or 2) a toxigenic strain of C. difficile was recovered from culture and the patient had diarrhea concurrent with or within 1 week of the positive C. difficile result (range, 0 days to 7 days). Clostridium difficile colonization was defined as a toxigenic or nontoxigenic C. difficile strain from culture in patients without diarrhea (asymptomatic colonization) or a nontoxigenic C. difficile strain from culture in patients with diarrhea (non-C. difficile diarrhea). Clinical data were recorded daily from the patients' medical records.

Laboratory Procedures

Stool, rectal swab, and enterostomy effluent specimens were cultured anaerobically for 48 hours at 35°C on prereduced cycloserine cefoxitin fructose egg yolk agar (500 µg/mL of cycloserine and 16 µg/mL of cefoxitin [Dimed, St. Paul, Minnesota]) and taurocholate cycloserine cefoxitin fructose egg yolk agar (1 mg/mL of sodium taurocholate [Sigma Chemical Co., St. Louis, Missouri]). Clostridium difficile isolates were identified by characteristic colonial structure and Gram stain appearance [13]. Diarrheal stool samples were also tested for C. difficile by the alcohol shock method, as described by Clabots and colleagues [14].

For the C. difficile cytotoxin assay, each aliquot of diarrheal stool specimen was mixed with an equal volume of phosphate-buffered saline (pH, 7.2) and inoculated onto a monolayer of HEp-2 cells. Cytopathic changes neutralized by Clostridium sordellii antitoxin indicated the presence of C. difficile cytotoxin [15]. For the last 5 months of the study, stool cytotoxicity was determined by using the Bartels cytotoxicity assay for C. difficile toxin with human foreskin fibroblasts ( 10 passages) and C. difficile antitoxin (Baxter Diagnostics, Inc.) for neutralization.

All C. difficile isolates were inoculated into chopped meat broth and incubated at 36°C for 7 days. Supernatants from these cultures were tested for cytotoxicity as described elsewhere [16]. HindIII restriction endonuclease analysis (REA) typing was done on all isolates by using a rapid DNA extraction method. Organisms were typed and grouped by comparison of restriction patterns of known REA types in a library of several hundred unique types [16].

Statistical Analysis

The proportion of surveillance days with diarrhea was calculated by dividing the number of days with diarrhea by the number of days of surveillance. Independent t-tests or Mann-Whitney U tests were used to compare differences in interval data between groups, and chi-square analyses were used to compare differences in categorical data between groups. The Mantel-Haenszel test was used to determine linear association among ordinal variables. P values less than 0.05 were considered significant.

Multivariate analyses to determine a model for C. difficile acquisition and C. difficile-associated diarrhea were done by using logistic regression analysis. We used chi-square or t-tests to screen the following reported risk factors for C. difficile acquisition or C. difficile-associated diarrhea [17, 18] for inclusion in the logistic regression models: age; diagnosis; type of unit; abdominal surgery during the study; duration of surveillance; duration of hospitalization before the study; severity of illness index score; treatment with any antibiotic; treatment with categories of antibiotics; and treatment with metoclopramide hydrochloride, antacids, H₂-receptor antagonists, or sucralfate. Other variables screened for inclusion in the regression analyses were tube feeding and type of tube-feeding formula. As a follow-up to the initial regression analyses that analyzed tube feeding in general, we did multivariate analyses of risk factors for C. difficile acquisition or C. difficile-associated diarrhea after reclassifying tube feeding according to location of the feeding-tube tip. All patients who acquired C. difficile or developed C. difficile-associated diarrhea were exposed to antibiotics; therefore, multivariate analyses were done by using individual categories of antibiotics. Variables with P values 0.1 were candidates for the multivariate analyses. Coefficients of the regression variables were tested for significance by using a Wald chi-square statistic. Final models were based on a combination of odds ratios, P values, reported risk factors, and clinical judgment. A Kaplan-Meier survival curve [19] was generated to illustrate the effect of location of the feeding-tube tip on the rate of development of C. difficile-associated diarrhea. Data were analyzed by using SAS software, version 6.10 (SAS Institute, Inc., Cary, North Carolina) and SPS software, version 6.1 (SPS Inc., Chicago, Illinois).

Results

Top Methods Results Discussion Author & Article Info References

Cohort matching resulted in equivalent groups, as shown by a lack of significant differences in any of the variables used for matching tube-fed with non-tube-fed patients (Table 1). The two female tube-fed patients were matched with male non-tube-fed patients because matches of the same sex could not be accomplished within 4 weeks. Two tube-fed patients each had a colostomy, and one non-tube-fed patient had an ileostomy. The patients' primary diagnoses were categorized as head or neck cancer (39 patients), peripheral vascular disease (33 patients), neurologic disorders (18 patients), pulmonary disorders (17 patients), gastrointestinal disorders (10 patients), cardiac disorders (9 patients), abdominal aortic aneurysms (9 patients), genitourinary or renal disorders (6 patients), and miscellaneous disorders (such as multiple myeloma or hip fracture) (11 patients).

Of the 76 tube-fed patients, 75 started tube feeding by continuous administration and 1 started it by intermittent administration. The method of tube-feeding administration was changed during the study in 34 of the 76 patients. Tube feeding was started with Osmolite (Ross Laboratories, Columbus, Ohio) in 61 patients, Osmolite HN (Ross Laboratories) in 9 patients, Impact (Sandoz Nutrition, Minneapolis, Minnesota) in 5 patients, and Vivonex (Sandoz Nutrition) in 1 patient. Thirty-four patients had nasogastric tubes, 30 had nasoduodenal tubes, 6 had percutaneous endoscopic gastrostomy tubes, and 6 had nasojejunal or jejunostomy tubes. The type of tube was changed in 17 patients during the study. At the time the study was done, all nasoduodenal and nasojejunal feeding tubes at the MVAMC were inserted by using fluoroscopy.

Acquisition of Clostridium difficile and Incidence of Clostridium difficile-Associated Diarrhea

Overall acquisition of C. difficile was significantly greater in tube-fed patients than in non-tube-fed patients (15 of 76 patients [20%] compared with 6 of 76 patients [8%]; P = 0.03) (Figure 1). More tube-fed patients than non-tube-fed patients developed C. difficile-associated diarrhea (7 of 76 patients [9%] compared with 1 of 76 patients [1%]; P = 0.03) (Figure 1). Of the 54 patients with diarrhea who were tested for C. difficile cytotoxin, 3 had a positive result and all were tube fed. Two of the 3 patients with positive results on C. difficile cytotoxin assay also had positive C. difficile cultures. No difference was seen in the percentages of tube-fed patients and non-tube-fed patients (8 of 69 patients [12%] compared with 5 of 75 patients [7%]; P > 0.05) who became colonized with C. difficile (Figure 1).

Although the incidence of diarrhea was greater in tube-fed patients than in non-tube-fed patients (38 of 76 patients [50%] compared with 16 of 76 patients [21%]; P < 0.001), the proportion of surveillance days with diarrhea was greatest in tube-fed patients after the development of C. difficile-associated diarrhea. The proportion of surveillance days with diarrhea in tube-fed patients before the development of C. difficile-associated diarrhea was not significantly different from that in tube-fed patients who did not develop C. difficile-associated diarrhea (mean ±SD, 0.28 ± 0.3 compared with 0.22 ± 0.2 [95% CI for the mean difference, –0.09 to 0.21]. The proportion of surveillance days with diarrhea in tube-fed patients after the development of C. difficile-associated diarrhea was more than three times as great as the proportion in tube-fed patients who did not develop C. difficile-associated diarrhea (mean ±SD, 0.68 ± 0.4 compared with 0.22 ± 0.2 [CI for the mean difference, 0.08 to 0.84]).

The five types of feeding tubes were reclassified according to the location of the feeding-tube tip: prepyloric only (33 patients) or postpyloric at some time during the study (43 patients). The incidence of C. difficile-associated diarrhea was greater in tube-fed patients who received postpyloric tube feeding at some time during the study (6 of 43 patients [14%]) than in those who received prepyloric tube feeding continually (1 of 33 patients [3%]) or those without a feeding tube (1 of 76 patients [1%]) (P = 0.01 for postpyloric tube feeding compared with prepyloric tube feeding compared with no tube feeding). In the 3 patients with an enterostomy, both rectal and enterostomy effluent cultures were obtained at 12 surveillance time points. All cultures yielded the same results.

Antibiotic Exposure

Similarly high percentages of tube-fed patients and non-tube-fed patients received antibiotics (71 of 76 patients [93%] compared with 65 of 76 patients [86%]; P > 0.1). Tube-fed patients and non-tube-fed patients received the same median numbers of antibiotics (2 [range, 0 to 8] and 2 [range, 0 to 7]; P > 0.1, Mann-Whitney U test). Analysis of exposure to separate categories of antibiotics showed that more tube-fed patients than non-tube-fed patients received third-generation cephalosporins (46 of 76 patients [61%] compared with 26 of 76 patients [34%]; P = 0.001) and intravenous metronidazole (38 of 76 patients [50%] compared with 8 of 76 patients [11%]; P < 0.001).

Exposure to antibiotics was similar in patients who acquired C. difficile and those who did not (21 of 21 patients [100%] compared with 115 of 131 patients [88%]; P = 0.09). All 8 patients who developed C. difficile-associated diarrhea had received antibiotics. More patients with C. difficile-associated diarrhea than patients without this diarrhea received third-generation cephalosporins (4 of 8 patients [50%] compared with 24 of 144 patients [17%]; P = 0.02) and aminoglycosides (5 of 8 patients [63%] compared with 38 of 144 patients [26%]; P = 0.03). Exposure to aminoglycosides was highly correlated with exposure to third-generation cephalosporins (P = 0.001).

Multivariate Analysis of Risk Factors for Clostridium difficile Acquisition and Clostridium difficile-Associated Diarrhea

With regard to our initial question about whether tube feeding in general is an independent risk factor for the acquisition of C. difficile, multivariate logistic regression showed that tube feeding (odds ratio [OR], 3.1 [CI, 1.1 to 8.7]; P = 0.03) and duration of surveillance (OR, 1.1 [CI, 1.004 to 1.2]; P = 0.047) were significant but that age (OR, 1.0 [CI, 0.99 to 1.1]; P = 0.07) was not. For every day of monitoring beyond the average duration of surveillance (12 days), a patient's risk for acquiring C. difficile increased by 1.1. Multivariate analysis of risk factors for the development of C. difficile-associated diarrhea showed that tube feeding (OR, 9.0 [CI, 1.02 to 79.8]; P = 0.049) but not exposure to third-generation cephalosporins (OR, 4.1 [CI, 0.9 to 18.6]; P = 0.07) or intravenous metronidazole (OR, 0.45 [CI, 0.08 to 2.5]; P > 0.2) was a risk factor for the development of C. difficile-associated diarrhea. A tube-fed patient was nine times more likely than a non-tube-fed patient to develop C. difficile-associated diarrhea.

In the final multivariate models (Table 2), when tube feeding was classified according to location of the feeding-tube tip, the significant effect of postpyloric tube feeding on the acquisition of C. difficile and the development of C. difficile-associated diarrhea was evident. A tube-fed patient with a feeding-tube tip placed distal to the pylorus was 3 times more likely to acquire C. difficile and 11 times more likely to develop C. difficile-associated diarrhea than was a non-tube-fed patient (Table 2). The relations of other variables in the models were unchanged (Table 2).

To further show the effect of tube feeding and feeding-tube tip location on the rate of development of C. difficile-associated diarrhea, we did a Kaplan-Meier survival analysis. Three groups were generated for inclusion in the hypothetical model for time free of C. difficile-associated diarrhea: Group 1 was not tube fed, group 2 received prepyloric tube feeding, and group 3 received postpyloric tube feeding at least some of the time. Group 3 (for which the feeding-tube tip was placed distal to the pylorus) developed C. difficile-associated diarrhea at the fastest rate and had the lowest proportion of patients remaining free of C. difficile-associated diarrhea (P = 0.045) (Figure 2).

View larger version (15K): [in this window] [in a new window]   Figure 2. Time free of Clostridium difficile-associated diarrhea in patients stratified according to location of feeding-tube tip. A Kaplan-Meier survival analysis showed that patients who received postpyloric tube feeding (solid line) had the fastest rate of development of C. difficile-associated diarrhea and the lowest proportion of patients remaining free of C. difficile-associated diarrhea (P = 0.045). Numbers in parentheses are numbers of patients remaining free of C. difficile-associated diarrhea. The dotted line represents non-tube-fed patients; the dashed line represents patients receiving prepyloric tube feeding.

Restriction Endonuclease Analysis Typing Results

Restriction endonuclease analysis typing was done on 80 C. difficile isolates from 20 patients and showed a diverse population of strains. Nineteen distinct REA types in 12 REA groups were identified. Eight REA groups (D, G, J, Y, AS, BM, BN, and CF) were toxigenic, and 4 (C, M, T, and AP) were nontoxigenic. In each of 2 patients, C. difficile isolates representing two different REA groups were recovered at different surveillance time points. The in vitro toxigenicity of both REA types was the same in each patient. The most striking difference in REA distribution was the presence of toxigenic group Y isolates (4 patients, 3 REA types) in tube-fed patients but not in non-tube-fed patients. The 4 tube-fed patients with group Y isolates resided on four different units.

Discussion

Top Methods Results Discussion Author & Article Info References

As far as we know, this is the first published, prospective, controlled, epidemiologic investigation of tube feeding as a risk factor for acquisition of C. difficile and development of C. difficile-associated diarrhea. Significantly more tube-fed patients (particularly those in whom feeding tubes were placed distal to the pylorus) than non-tube-fed patients acquired C. difficile and developed C. difficile-associated diarrhea. The use of antibiotics was so high in both groups of patients that we could not distinguish antibiotic use as a risk factor for C. difficile-associated diarrhea.

We propose several possible reasons for the association of C. difficile-associated diarrhea with tube feeding. First, tube-fed patients may have an increased risk for acquiring C. difficile from the hands of health care providers as a result of routine manipulation of the tube-feeding system. Clostridium difficile has frequently been recovered from the hands of personnel caring for patients infected with this pathogen; this implicates hospital staff as a source of transmission [17, 20, 21]. Studies of the sterility of tube-feeding systems have reported that manipulation of the systems is a primary cause of bacterial contamination of the systems and formulas [22-27]. Aseptic techniques, ready-to-use formulas, and closed delivery systems were found to reduce contamination of enteral feedings [28-32], and aseptic techniques were associated with a decline in rates of C. difficile-associated diarrhea [33-35]. During our surveillance period, an open tube-feeding system was used, the hang time for ready-to-use formulas was 8 hours, and tube-feeding administration sets were changed every 24 hours.

Second, contaminated formulas and delivery systems have been associated with nosocomial infection [36, 37]. In our study, this association was supported, or at least suggested, by the finding of C. difficile organisms from REA group Y only in tube-fed patients in widely scattered geographic locations in the hospital. Our REA typing data suggest sporadic transmission of and acquisition of disease by multiple strains of C. difficile, a finding noted previously at our institution and elsewhere [38, 39].

Third, tube feeding may have increased the risk for C. difficile-associated diarrhea in our study through the lack of dietary fiber in the formulas administered; this lack could have produced an intestinal environment conducive to the growth of C. difficile. In one study [40], ingestion of a fiber-free tube-feeding formula compared with a normal diet or a fiber-supplemented formula resulted in a lower concentration of short-chain fatty acids and a less acidic pH in colonic fluid. Short-chain fatty acids and acidic pH contribute to resistance to C. difficile colonization and proliferation in vitro [41, 42]. Whether a fiber-supplemented formula would have decreased or prevented C. difficile acquisition in our patient sample is uncertain, given the sample's high frequency of antibiotic use and the subsequent decreased fermentation of fiber by reduced numbers of colonic bacteria [9].

The significantly greater incidence of C. difficile acquisition and C. difficile-associated diarrhea in patients receiving postpyloric tube feeding suggests that delivery below the gastric acid barrier may facilitate the introduction and survival of C. difficile organisms. Although C. difficile spores are resistant to acid and susceptible hosts (such as patients exposed to antibiotics) may be infected with a low inoculum of this pathogen, vegetative cells are easily killed by gastric acid, reducing inocula by 99% [43]. Gastric acidity may be a relatively important host defense against C. difficile infection, as it is against infection by other enteric pathogens.

Duration of surveillance was a predictive risk factor for the acquisition of C. difficile. This result is consistent with the previous finding by Clabots and colleagues [44] that the rate at which hospitalized patients acquired C. difficile increased linearly with duration of hospital stay (approximately 8% per week). Increased duration of hospitalization has been suggested as a marker of severity of illness because it increases susceptibility to the acquisition of C. difficile [17]. Our non-tube-fed patients were matched with our tube-fed patients for duration of surveillance and hospitalization before the study, location in the intensive care unit, and severity of illness. Although the validity and reliability of Horn's severity of illness index have been established [45], the instrument may not have accounted for all differences in severity of illness between tube-fed and non-tube-fed patients.

Clinical testing for C. difficile in tube-fed patients with diarrhea was rare at the MVAMC before our study was done, probably because clinicians attributed diarrhea to tube feeding itself. Laboratory testing for C. difficile remains a controversial issue in clinical diagnosis and epidemiologic surveillance [46]. Our finding of a low incidence of positive cytotoxin assay results in patients with C. difficile-associated diarrhea who had a positive stool culture of a toxin-producing strain could be explained by a low inoculum of C. difficile in the study patients or by the low virulence of the recovered strains. We previously reported toxin production in vitro in 73% of C. difficile isolates recovered from 264 patients with diarrhea who had a positive stool culture but negative results on stool cytotoxin assay [47]. Kelly and coworkers [3] acknowledged the limitations of the cytotoxin assay in their investigation of diarrhea in tube-fed patients and concluded that using a selective culture medium for C. difficile, as we did, may have increased their ability to isolate the organism. We believe that our data support a policy of submitting stool specimens from tube-fed patients who have diarrhea for C. difficile testing, especially if C. difficile stool culturing is available.

In conclusion, our results indicate that tube feeding is a risk factor for the development of C. difficile-associated diarrhea. Administering tube feeding below the pylorus increases the risk for C. difficile acquisition and C. difficile-associated diarrhea. Use of both culture and cytotoxin assay is recommended for the optimal detection of C. difficile in tube-fed patients. Increased enforcement of barrier infection-control measures during manipulation of tube-feeding systems may prevent C. difficile infection, although no data have shown the efficacy of these practices to date.

Drs. Johnson and Gerding: Department of Medicine, Veterans Affairs Chicago Healthcare System, 333 East Huron Street, Chicago, IL 60611.

Ms. Clabots: Veterans Affairs Medical Center, Research Service 151, One Veterans Drive, Minneapolis, MN 55417.

Dr. Willard: Riverview Office Tower, Suite 600, 8009 34th Avenue South, Bloomington, MN 55425.