Annals of Clinical & Laboratory Science 37:241-247 (2007)
© 2007 Association of Clinical Scientists
Clostridium difficile Infection in an Urban Medical Center: Five-year Analysis of Infection Rates among Adult Admissions and Association with the Use of Proton Pump Inhibitors
Suresh Jayatilaka1,2,
Rada Shakov1,2,
Rodney Eddi3,
Gentiana Bakaj4,
Walid J. Baddoura1,2 and
Vincent A. DeBari2
1 Division of Gastroenterology and 2 Department of Internal Medicine, School of Graduate Medical Education, Seton Hall University, South Orange, New Jersey; 3 Department of Internal Medicine, Mount Sinai School of Medicine, St. Joseph’s Regional Medical Center, Paterson, New Jersey; and 4 St. George’s University School of Medicine, Grenada, West Indies
Address correspondence to Vincent A. DeBari, Ph.D., St Michael’s Medical Center, 268 Martin Luther King Blvd., Newark, NJ 07102, USA; tel 973 877 2813; fax 973 754 2799; e-mail debarivi{at}shu.edu.
 |
Abstract
|
|---|
C. difficile-associated diarrhea (CDAD) has become a major cause of morbidity in hospitalized patients. In this study of five-year (2001–2005, inclusive) trends of incidence of CDAD among adults in an inner-city medical center, the overall annual incidence increased from 5.08 to 8.42 cases/103 admissions (p = 0.0005). Age distribution remained fairly constant for 2001–2004 but decreased significantly in 2005 (p = 0.005); no significant change was observed for gender. During the five-year period, we observed a decline in the use of histamine type 2 receptor antagonists (H2A) with a concomitant increase in the use of proton pump inhibitors (PPI) as a prophylactic measure to prevent stress ulcers. The usage of PPI correlated exactly (rs = 1.0; p = 0.017) with the increase in CDAD incidence. A case (n = 122)-control (n = 244) study for the final year was conducted, examining the association of PPI and H2A with CDAD. After controlling for the effect of antibiotic use, PPI either pre- or during admission was associated with CDAD (odds ratio, OR (adjusted) = 2.75, 95% CI = 1.68 to 4.52; p = 0.0001); the association with H2A was not significant (OR (adjusted) = 0.95, 95% CI = 0.39 to 2.34; p = 0.9153). If only first-time use during hospital stay is considered, PPI were also strongly associated with CDAD (OR (adjusted) = 1.88, 95% CI: 1.07 to 3.31; p = 0.0283) and H2A were not associated with CDAD (OR (adjusted) = 0.73, 95%CI: 0.26 to 2.06; p = 0.5520). These data suggest that the widespread prescription of PPI for stress ulcer prophylaxis in acute care facilities may contribute to the increased incidence of CDAD.
Keywords: Clostridium difficile, infectious diarrhea, stress ulcer prophylaxis, proton pump inhibitors
|
Introduction
|
|---|
Within the past decade, Clostridium difficile, a Gram-positive, spore forming bacillus, has emerged as a major causative agent for nosocomial diarrhea [1,2]. The incidence of C. difficile-associated diarrhea (CDAD) has been studied in a number of settings and has been found to range widely, with estimates between 1 and 30 per thousand admissions [3]. Recent outbreaks of a particularly virulent form of the organism have prompted increased surveillance for CDAD [4,5].
Most epidemiologic studies have sought to identify potential risk factors for CDAD, such as primary diagnosis [5] or therapeutic modalities, primarily antibiotics [6]. For the most part, these investigations have analyzed data collected over relatively brief periods of time. We sought to determine the change in incidence of CDAD in a large, urban, tertiary-level, acute care facility over the 5-year period 2001–2005, with a specific focus on the potential association of CDAD with the use of gastric acid-reducing drugs (GARD), sub-classified into type 2 histamine receptor antagonists (H2A) and proton pump inhibitors (PPI). We also conducted an observational investigation (case-control model) for the final year (2005) using age- and sex-matched controls. This paper reports the results of these studies.
|
Materials and Methods
|
|---|
Setting.
This study was conducted in a 750-bed, tertiary care teaching hospital in an urban community (Paterson, NJ) in the northeastern United States. This institution represents one of several facilities within the St. Joseph’s Healthcare System including another acute-care facility, a long-term geriatric care facility, and a children’s hospital. All records from which the data were obtained were from adult patients at St. Joseph’s Regional Medical Center. The St. Joseph’s Institutional Review Board classified this study as exempt.
Protocol 1. Incidence and correlation with GARD usage.
Data were obtained from computer-based admission and discharge records of all adult inpatients over the period from 1 January 2001 to 31 December 2005, tabulated on an annual basis. Only patients 
18 yr of age on the day of admission were included; no upper age limit was established. Incidence was calculated based on the assumption that each case was a new case of CDAD. Screening of discharge summaries was based on relevant diagnostic (ICD-9) codes.
A preliminary survey of potential pharmacologic risk factors was conducted using the pharmacy department’s database and quantified as the number of patients treated with all GARD, sub-classified into H2A and PPI. Usage of these agents (as number of patients treated) was tested for correlation with CDAD incidence.
Protocol 2. Case-control study of year 2005 data.
For the final year of the incidence period (2005), we had identified, by ICD-9 coding, 130 cases of presumptive CDAD. Of these, we were able to retrieve appropriate clinical data on 122 cases, all of which were considered confirmed by the presence of a positive test for C. difficile toxin A or B and associated episodes of diarrhea. For each of the cases, we sequentially enrolled two controls, admitted during the same time period, and age- and sex-matched to the CDAD cases. Data on usage of GARD were gathered from review of the patients’ charts.
The cases and controls were further stratified into those who were treated with GARD only after admission (PPI use "post-admission" or H2A use "post-admission") and GARD prior to or after admission (designated as "pre- or during admission"). We also assessed the use of antibiotics, classifying them into two groups, all antibiotics (excluding vancomycin and metronidazole, both of which are used to treat CDAD) and fluroquinolones only, since they are an established risk factor for CDAD [6].
Analytical method.
Fecal analysis for C. difficile toxins A and B was performed using an ELISA test kit (TechLab, Blacksburg, VA) [7].
Statistical methods.
Changes in dichotomous categorical data (annual incidence and gender distribution) were evaluated for statistical significance by the Chi-square test for trend. Because frequency distributions of age data were found not to be normally distributed by D’Agostino-Pearson omnibus normality test, age differences over the period 2001–2005 were evaluated by a non-parametric method, the Kruskal-Wallis test (utilizing Dunn’s test, post hoc, to ascertain group-wise differences). Correlation of drug use with incidence was conducted using Spearman’s non-parametric method, assuming no normality of distribution for these data over the 5 yr of the study. For univariate associations analyzed in the case-control study, case/control vs risk factor present/absent was cross-tabulated in contingency tables for which statistical significance was evaluated by Fisher’s exact test. Multivariate analysis to determine the confounding effect of antibiotic use was performed by logistic regression. Ages were compared by the Mann-Whitney test when not normally distributed and by the t-test for the independently assorted data (unpaired t-test) when they were normally distributed. All continuous data are expressed as medians and interquartile ranges (IQR). Because of the retrospective nature of the study, odds ratios (OR) and 95% confidence intervals of the OR (rather than relative risk) were calculated. All p values were calculated on a two-sided basis with 
set at 0.05. Prism software (GraphPad Corp., San Diego, CA), run on a Windows/PC platform, was used for all calculations except logistic regression for which an on-line software routine (http://www.statpages.org/logistic.html), described previously [8], was used.
|
Results
|
|---|
Incidence of CDAD and correlation with GARD usage during the study period.
Over the study period, a total of 76,908 admissions were observed. The number of annual admissions remained fairly constant with a range of 15,133 in the yr with the fewest admissions (2003) to 15,683 in 2001, the yr with the most admissions, representing <4% change in admissions over the study period. Cases of CDAD increased from 80 in 2001 to 130 in 2005, with a slight decrease to 91 cases in 2003, yielding an overall change in incidence over the period from 5.08 to 8.42 cases/103 admissions (Fig. 1
), a statistically significant increase (p = 0.0005).
During the period 2001–2004, the ratio of males to females with CDAD in the study group declined from 0.90 to 0.56, but rose slightly in 2005 to 0.65 (Table 1). This trend was not statistically significant (p = 0.149). Table 1 also provides data on differences in age distribution (because the age data are not normally distributed, these data are expressed as medians, with the inter-quartile ranges also provided). Age distributions remained fairly constant for the period 2001–2004, with median values in the mid 70s. However, there was a significant decline (p by Kruskal-Wallis was 0.005) for the final year of the study to a median of 67.5 (p < 0.05 vs all other years, Dunn’s test).
Among the GARD, a decline was observed with H2A over the period and this was reflected in the correlation of GARD usage with CDAD for which a value of rs = –0.90 was obtained. Although this correlation was not statistically significant (p = 0.083), it should be noted that the data represent only 5 yr, necessitating a "perfect" correlation, ie, rs = ±1.0, to achieve a value of p < 0.05. The usage of PPI, however, did increase substantially over the study period and correlated strongly with CDAD incidence (rs = 1.00; p = 0.017).
Association of CDAD with GARD usage for year-five (2005).
Flowcharts show the GARD use data for the cases (Fig. 2A) and controls (Fig. 2B). Fig. 2A illustrates the distribution of GARD use among patients with CDAD before admission and changes made to their regimen after admission. Fig. 2B illustrates those data for controls. Baseline characteristics of age, gender, and antibiotic usage are compared for the overall group and for subsets and are shown in Table 2. Differences between the cases and controls were not significant for any comparison. However, antibiotic use was highly significant for the overall group and for the PPI subgroups though not significantly different for the H2A groups. For this reason, Table 3 provides data on odds ratios (OR) for both univariate analysis of GARD use (Table 3A) and for a multivariate comparison in which antibiotic use was considered a confounding variable (Table 3B).
Of 122 patients with CDAD, 53 (43%) were on GARD before admission (44 on PPI and 9 on H2A). Of 244 controls, 69 (28%) were on GARD before admission (63 on PPI and 6 on H2A). In order to determine the association between GARD use and CDAD, 2 separate univariate analyses were performed. These are summarized in Table 3A
. First, we analyzed the association between CDAD and PPI use at any time before or during admission. Eighty-four patients with CDAD (69%) were on PPI before or after admission, compared to 112 controls (46%). Nine patients with CDAD (7.4%) were treated with H2A, compared to 17 controls (6.7%). The association between GARD use and CDAD was analyzed for both PPI and H2A. Use of PPI either pre- or during admission was associated with CDAD (odds ratio, OR = 2.61, 95% CI = 1.65 to 4.12; p = 0.0001); the association with H2A was not significant (OR = 1.06, 95% CI = 0.46 to 2.46; p = 1.00).
A second analysis was performed on cases and controls who were naïve to GARD prior to admission. Thirty-four patients with CDAD (28%) were started on PPI after admission. For controls, 46 patients (19%) received PPI for the first time in hospital. Six patients with CDAD (5%) with no previous history of GARD use were started on H2A after admission. This number for controls was 14 (5.7%). If only first-time use during hospital stay is considered, PPI were also strongly associated with CDAD (OR = 2.57, 95% CI: 1.45 to 4.55; p = 0.0016), and H2A, again, were not associated with CDAD (OR = 0.86, 95%CI: 0.32 to 2.30; p = 1.00).
Table 3B provides adjusted odds ratios for GARD use controlled for antibiotic therapy. Considering all antibiotic use as a confounding variable, PPI use is still associated, significantly, with CDAD (adjusted OR, pre- or during admission = 2.75, p = 0.0001; PPI use, post-admission, OR = 1.88, p = 0.028). Considering only fluoroquinolones, PPI use is also associated significantly with CDAD (adjusted OR, pre- or during admission = 2.56, p = 0.0001; PPI use, post-admission, OR = 1.82, p = 0.030). The effect of H2A remains insignificant for all situations examined.
|
Discussion
|
|---|
C. difficile infection in patients can cause an infectious diarrhea or pseudomembranous colitis, due to an inflammatory response to toxin-induced cytokines in the colon [9]. The severity of CDAD varies from mild to life-threatening, depending on the toxigenic potential of the causative strain [10]. Two toxins, designated toxin A and toxin B, are elaborated by the organism; strains of C. difficile that lack these are not believed to cause disease [11].
Among the iatrogenic risk factors for CDAD, antibiotic therapy has received the greatest attention [2,3,12]. However, recent evidence suggests that GARD may also be associated with CDAD. Among GARD, the PPI have become prominent and are widely prescribed as prophylactic agents for stress-induced ulcers [13], replacing H2A, a group of GARD with a lower therapeutic potential than PPI. The replacement of H2A with PPI in recent years led to the suggestion that there might be an increase in CDAD in healthcare facilities and, thus, in the present investigation [14,15].
The setting in which this study was conducted is representative of large, urban, North American hospitals. In addition to a diverse racial and ethnic inner-city population, as a major, university-affiliated tertiary medical center, it has a substantial patient population from the surrounding, generally upper-middle class, northern New Jersey suburbs.
Clearly, the statistically significant trend representing a nearly 66% increase in CDAD incidence over the study period cannot be solely attributed to the use of PPIs despite the observation of a strong correlation. In order to explore the association between GARD and CDAD we used an observational model, the case-control study, to test the hypothesis that PPI use, but not H2A use, is associated with CDAD. Our findings strongly suggest that patients treated with PPI either prior to or upon admission have a significantly greater likelihood of developing CDAD. However, the use of H2A either prior to or during admission is not associated with CDAD.
It is reasonable to speculate that the profound increase in gastric pH brought about by virtually complete inhibition of the proton pump results in a favorable environment for C. difficile growth. The H2A, on the other hand, are less effective in increasing gastric pH and thus provide a less suitable environment for the organism.
The results of this study have implications for so called "GI prophylaxis" in acute care facilities, ie the use of GARD for prevention of stress ulcers, a common practice for general medical patients [16]. In our institution, upwards of 85% of all patients admitted are treated with GARD. It is not unlikely that use of H2A might accomplish the desired goal of prevention of stress ulcers in hospitalized patients without having the patients incur the presumed risk of CDAD that is associated with the use of PPI.
We recognize the limitations inherent in this study. Case-control studies provide only observational evidence with regard to risk. Nevertheless the number of cases that we studied is substantial and based on the size of the study we consider our data to strongly support the hypothesis that PPI, but not H2A, are associated with CDAD. Another issue that might be considered a limitation is that there was scant endoscopic proof of characteristic lesions associated with C. difficile infection. However, we consider episodes of diarrhea in the presence of a positive test for C. difficile toxin to be strong presumptive evidence of CDAD. We also recognize that our study is based on data from a single institution. However the size and demographics of the institution in which the study was performed provide support for our conclusions.
|
Acknowledgement
|
|---|
Portions of this work were presented at the annual meeting of the Association of Clinical Scientists in Hershey, PA, on 16 to 20 May 2007.
|
References
|
|---|
- Poutanen SM, Simor AE. Clostridium difficile-associated diarrhea in adults. CMAJ 2004;171:51–58.[Abstract/Free Full Text]
- Schroeder MS. Clostridium difficile-associated diarrhea. Am Fam Physician 2005;71:921–928.[Medline]
- Samore MH, DeGirolami PC, Tlucko A, Lichtenberg DA, Melvin ZA, Karchmer AW. Clostridium difficile colonization and diarrhea at a tertiary care hospital. Clin Infect Dis 1994;18:181–187.[Medline]
- McDonald LC, Killgore GE, Thompson A, Owens RC Jr., Kazakova SV, Sambol SP, Johnson S, Gerding DN. An epidemic, toxin gene-variant strain of Clostridium difficile. NEJM 2005;353:2433–2441[Abstract/Free Full Text]
- Martirosian G, Szczesny A, Cohen SH, Silva J Jr. Analysis of Clostridium difficile-associated diarrhea among patients hospitalized in a tertiary care academic hospital. Diagn Microbiol Infect Dis 2005;52:153–155.[Medline]
- Pepin J, Saheb N, Coulombe MA, Alary ME, Corriveau MP, Authier S, Leblanc M, Rivard G, Bettez M, Primeau V, Nguyen M, Jacob CE, Lanthier L. Emergence of fluroquinolones as the predominant risk factor for Clostridium difficile-associated diarrhea: A cohort study during an epidemic in Quebec. Clin Infect Dis 2005;41:1254–1260.[Medline]
- Laughon, BE, Viscidi, RP, Gdovin SL, Yolken RH, Bartlett JG. Enzyme immunoassays for detection of Clostridium difficile toxins A and B in fecal specimens. J Infect Dis 1984 149:781–788.[Medline]
- DeBari, VA. Computation of decision levels from differentiated logistic regression probability curves. Ann Clin Lab Sci 2006;36;194–200.[Abstract/Free Full Text]
- Kuijper EJ, Coignard B, Tull P. Emergence of Clostridium difficile-associated disease in North America and Europe. Clin Microbiol Infect 2006;12 Suppl 6:2–18.[Medline]
- Nath S, Thornley J, Kelly M, Kucera B, On SL, Holmes B, Costas M. A sustained oubreak of Clostridium difficile in a general hospital: Persistence of a toxigenic clone in four units. Infect Control Hosp Epidemiol 1994;15:382–389.[Medline]
- DeBarbeyrac B, Guinet R, Quentin C, Cantet P, Bebear C. Clostridium difficile and its cytotoxin in diarrhoeic stools of hospitalized patients. Ann Biol Clin 1989;47:67–70.[Medline]
- Settle CD, Wilcox MH. Antibiotic-induced Clostridium difficile infection. Aliment Pharmacol Ther 1996;10:835–841.[Medline]
- Harty RF, Ancha HB. Stress ulcer bleeding. Curr Treat Options Gastroenterol 2006;9:157–166.[Medline]
- Cunningham R, Dale B, Undy B, Gaunt N. Proton pump inhibitors as a risk factor for Clostridium difficile diarrhoea. J Hosp Infect 2003;54: 243–245.[Medline]
- Dial S, Delaney JA, Schneider V, Suissa S. Proton pump inhibitor use and risk of community-acquired Clostridium difficile-associated disease defined by prescription for oral vancomycin therapy. CMAJ 2006;175:745–748.[Abstract/Free Full Text]
- Janicki T, Stewart S. Stress-ulcer prophylaxis for general medical patients: a review of the evidence. J Hosp Med 2007;2:86–92.[Medline]
Top
Abstract
Introduction
