OUP user menu

Gastric acid suppression does not promote clostridial diarrhoea in the elderly

S. SHAH, A. LEWIS, D. LEOPOLD, F. DUNSTAN, K. WOODHOUSE
DOI: http://dx.doi.org/10.1093/qjmed/93.3.175 175-181 First published online: 1 March 2000

Abstract

Gastric acid prevents bacterial colonization of the stomach and suppression of its secretion might predispose to Clostridium difficile (CD) diarrhoea. We retrospectively studied elderly patients admitted to medical wards of an acute hospital to determine whether the incidence of CD diarrhoea was greater among those previously treated with gastric acid suppressants. From records of stool CD toxin tests undertaken in 1995 and 1996, we found 126 cases with positive results, and selected 126 controls with negative results. Information about pre‐morbid illness, predisposing factors for CD and medication received in the preceding 16 weeks was obtained from case‐notes. A greater number of CD positive cases had received antibiotics such as Cefuroxime, ciprofloxacin or macrolides with or without metronidazole, were more severely disabled, required assistance for feeding, or had hypoalbuminaemia before the onset of diarrhoea. A greater number of controls had received lactulose, suggesting either that its laxative effect resembled CD infection prompting frequent stool tests, or that it offered protection against CD in this group. Both groups were similar for the use of proton‐pump inhibitors or H2‐receptor antagonists, suggesting that susceptible elderly patients are not more likely to develop CD diarrhoea after receiving gastric acid suppression therapy.

Introduction

Gastric acid has bactericidal and bacteriostatic properties that prevent colonization of the proximal gut by oropharyngeal or faecal micro‐organisms.1 Previous experiments on healthy volunteers given oral sodium bicarbonate confirmed that neutralization of their gastric contents facilitated the multiplication of ingested strains of bacteria such as Vibrio cholera,2,,3 Shigella4 and Campylobacter.5 Suppression of gastric acid by H2‐receptor antagonists or proton pump inhibitors has been associated with an increase in intragastric bacterial counts6 and small bowel colonization.7 Bacterial colonization of the stomach is dependent upon the degree of reduction in gastric acidity8 and hence, omeprazole therapy has been noted to result in higher gastric colony counts compared with cimetidine.9 Bacterial diarrhoea occurs more often after treatment with gastric acid anti‐secretory agents.10 Anecdotal reports of pseudomembranous enterocolitis occurring after eradication of Helicobacter pylori infection using combination therapy with proton pump inhibitors and broad spectrum antibiotics11 suggest the possibility that gastric acid suppression may be a predisposing factor for Clostridium difficile infection.

Diarrhoea due to Clostridium difficile infection occurs more often in frail elderly patients who have been hospitalized for prolonged periods. Gut superinfection with this pathogen occurs in these susceptible patients particularly after having received broad‐spectrum antibiotics that destroy their normal protective gut microflora. Faecal cytotoxin detection is considered to be the optimal routine laboratory technique for confirming the diagnosis of Clostridium difficile diarrhoea.12,,13 Infection is transmitted by the oro‐enteral route. The ubiquitous distribution of the organism, and the presence of asymptomatic carriers, facilitates the spread of infection, particularly in hospital wards where elderly patients are nursed. A higher prevalence rate of asymptomatic carriage has been observed among elderly patients admitted to long‐stay wards in comparison to those admitted to acute short‐stay wards.14

Reduction in gastric acidity by gastric acid anti‐secretory agents such as H2 receptor antagonists or proton‐pump inhibitors may encourage gastric colonization and infection by Clostridium difficile in susceptible elderly patients. A study conducted in 1991 in Minnesota by Walker et al.15 noted a significant association between the use of H2‐receptor antagonists and the prevalence of Clostridium difficile infection in elderly residents of two long‐stay hospitals, but a study conducted in the previous year by Simor et al.16 in Toronto did not suggest a similar association. More recently, therapy with sucralfate, an anti‐ulcer agent that protects the gastric mucosa without suppressing gastric acid, was not noted to be a risk factor for this illness in hospitalized adults.17

No study has yet examined whether there is any association between gastric acid anti‐secretory therapy and the incidence of Clostridium difficile diarrhoea in acutely ill elderly patients admitted to short‐stay wards. We therefore undertook a study in order to determine whether the incidence of acute Clostridium difficile diarrhoea was increased in acutely hospitalized elderly patients who had received therapy with gastric acid anti‐secretory agents.

Methods

Study design and location

A retrospective study was conducted on patients admitted to the acute general (internal) medical and elderly care wards of Morriston Hospital, a busy district general hospital in Swansea, South Wales.

Patients

Patients above the age of 65 years who developed a diarrhoeal illness after admission to any of the designated acute general medical or elderly care wards and underwent stool testing for Clostridium difficile toxin between 01/01/95 and 31/12/96, were identified from the records of all stool Clostridium toxin tests undertaken by the Microbiology and Public Health Departments during that period of time. Stool Clostridium difficile toxin tests were performed in duplicate using faecal extract to identify a cytopathic effect on Vero tissue culture cells, with one test‐tube being neutralized by a specific antibody. Of these, 126 consecutive cases with positive stool Clostridium toxin test results were included in the study group, and 126 consecutive cases with negative stool Clostridium toxin test results, of a similar age, who had been admitted and nursed at the same time on similar wards, were included in the control group.

Exclusion criteria

Younger patients and those admitted to surgical wards, intensive care or renal dialysis units and patients with previous Clostridium difficile infection were not included in the study.

Measurements

The following data were recorded for all subjects in both study and control groups from clinical information available in their medical and nursing case notes: demographic details such as age, gender, date and place of admission; independent risk factors for developing infection such as premorbid functional status before the onset of diarrhoea (including degree of disability and dependence on nursing care, presence of incontinence and requiring assistance for feeding); date of onset of the diarrhoeal illness (defined as an acute change in the pattern of bowel movement with the passage of fluid stool on more than two occasions per day for at least 2 consecutive days), and pre‐morbid parameters of chronic illness including either (a) chronic anaemia (defined as haemoglobin being below 12 g/dl on two or more occasions when tested at least 10 days prior to the onset of diarrhoea), (b) hyperglycaemia (defined as random blood glucose values being >7.5 mmol/l when tested on two or more consecutive occasions at least 10 days prior to the onset of diarrhoea), and (c) low serum albumin (defined as being <35 g/l when tested on two or more occasions at least 10 days prior to the onset of diarrhoea); regular medication received in the 16 weeks before the onset of diarrhoea included various classes of antibiotics such as penicillins, third‐generation cephalosporins (cefuroxime), macrolides (erythromycin or clarithromycin), aminoglycosides, etc. in addition to other drugs known to affect gastro‐intestinal acidity or motility such as antacids, sedatives, opiates, non‐steroidal anti‐inflammatory agents, aspirin, steroids, anticholinergics, diuretics, bronchodilators, anti‐hypertensive agents, laxatives and lactulose; gastric acid suppression therapy in the preceding 16 weeks prior to the onset of diarrhoea including dose and duration of treatment of the different types of proton‐pump inhibitors or H2‐receptor antagonists. The time window of 16 weeks was chosen in order not to miss any cases of drug‐associated Clostridium difficile diarrhoea. Cases of Clostridium difficile diarrhoea have been seen up to 2 months after antibiotic treatment;18 the effect of gastric acid suppression therapy if any was unknown.

Statistical analysis

Initially, using χ2 and χ2 trend tests, univariate analysis was undertaken on the variables. This identified apparent risk factors, and the effect of severity of these factors, for the development of Clostridium difficile diarrhoea. Subsequently the risk factors identified by this analysis were entered into a stepwise multivariate logistic regression programme in order to determine which variables remained as independent predictors of the development of Clostridium difficile diarrhoea.

Results

Table 1 shows a univariate comparison of the demographic characteristics and premorbid states of health of all patients in the study and control groups. Both groups were similar for age (mean 82 years±8, range 65–96 years) and gender (two‐thirds were female). There was no significant difference between the groups for the presence of urinary or double incontinence, premorbid parameters of illness such as anaemia (mild or severe) or hyperglycaemia (glucose intolerance or diabetes). By univariate analysis it appeared that the degree of disability (p<0.013), requirements for assisted or tube feeding (p<0.005) and low serum albumin (p<0.01) were significantly associated with Clostridium difficile infection. The significance of association was strongly related to increasing severity for each of these parameters as shown by the trend test.

Table 2 refers to the various types of medication other than antibiotics received in the preceding 16 weeks before the onset of diarrhoea. There was no difference between the two groups for the premorbid use of drugs such as antacids, opiates, non‐steroidal anti‐inflammatory agents, aspirin, steroids, diuretics, anticholinergics, antihypertensive agents, bronchodilators and laxatives, although a significant proportion of patients in the control group had received lactulose as a form of an aperient (p<0.038, odds ratio 0.43 in the negative direction).

Table 3 refers to the use of antibiotics in the preceding 16 weeks before the onset of diarrhoea. There was no significance between the groups for the use of benzyl penicillin, flucloxacillin, amoxycyllin or augmentin, vancomycin, aminoglycosides or trimethoprim. Cefuroxime was noted to be the single most important predictor of Clostridium difficile infection (p<0.0004). Other agents that had a significant association with the illness included macrolide antibiotics such as erythromycin or clarithromycin (p<0.002) and ciprofloxacin (p<0.008). In the majority of these cases, broad‐spectrum antibiotics had been prescribed within the preceding 4 to 6 weeks of the onset of diarrhoea. The use of metronidazole in the study group appeared to be also significant by univariate analysis (p<0.036).

Table 4 refers to the use of gastric acid suppression therapy in the study and control groups. There was no significant difference between the number of patients in either groups who were prescribed therapeutic or maintenance doses of H2‐receptor antagonists or proton‐pump inhibitors in the preceding 16 weeks prior to the onset of acute Clostridium difficile diarrhoea (p<0.35).

Table 5 shows results of the multivariate logistic regression analysis performed using a stepwise variable selection method. Variables selected as making independent contributions to the risk were use of cefuroxime, macrolides such as erythromycin or clarithromycin, ciprofloxacin and lactulose. Use of the first three increased the risk, whilst that of lactulose decreased the risk. Other variables, such as severe degrees of disability, dependence on assistance for feeding, and low serum albumin no longer made independent contributions to explaining the risk, largely because of inter‐relationships between the covariates, since these factors were observed to be associated with the use of cefuroxime.

Details of the inter‐relationship between assisted feeding and cefuroxime use; and hypoalbuminaemia and cefuroxime use are given in Tables 6 and 7, respectively.

View this table:
Table 1

Comparison of characteristics of all patients in study and control groups

CharacteristicsClostridium%Clostridium%p
+ve study−ve control(χ2 trend
groupgrouptest)
(n=126) (n=126)
Women78628971
Mean age (years)81.182.4
Disability 0.013*
Independently mobile2276
Walking with assistance31254132
Chair‐bound33263427
Bed‐bound60484435
Mode of feeding 0.005*
Independent47376854
Requiring assistance61484939
Fed by NGT/PEG/IV151254
Incontinence 0.97
None50405241
Only urinary incontinence14111411
Urinary and faecal incontinence62426048
Anaemia 0.30
None76606955
Mild (Hb>10 g% <13 g%)21172117
Severe (Hb <10 g%)29233628
Glucose intolerance 0.93
None82657761
Impaired (random blood glucose 7.5–10 mmol/l)29233830
Diabetes (random blood glucose >10 mmol/l)1512119
Hypoalbuminaemia 0.01*
None25204334
Mild (serum albumin >33 g/l <35 g/l)151265
Severe (serum albumin <33 g/l)86687761
  • * Significant at p<0.05 by univariate analysis.

View this table:
Table 2

Medication in the preceding 16 weeks before the onset of diarrhoea

DrugClostridium +ve%Clostridium −ve%p
study groupcontrol group(χ2 test)
(n=126) (n=126)
NSAIDS211727210.42
Aspirin362924190.10
Steroids372931250.48
Anti‐hypertensives514045360.52
Bronchodilators413233260.33
Diuretics796380641.0
Opiates534254431.0
Sedatives473736290.18
Anticholinergics655254430.20
Laxatives362943340.41
Lactulose403257450.038*
  • * Significant at p<0.05 by univariate analysis.

View this table:
Table 3

Antibiotics in the preceding 16 weeks before the onset of diarrhoea

AntibioticClostridium +ve%Clostridium −ve%p
study groupcontrol group(χ2 test)
(n=126) (n=126)
Benzylpenicillin86861.0
Amoxycillin/augmentin614846360.07
Flucloxacillin231814110.15
Aminoglycosides22000.48
Vancomycin22221.0
Trimethoprim181412100.33
Cefuroxime745930240.0004*
Erythro/clarithromycin383017140.002*
Ciprofloxacin403221170.008*
Metronidazole262113100.036*
  • * Significant at p<0.05 by univariate analysis.

View this table:
Table 4

Gastric acid suppression therapy in the preceding 16 weeks before the onset of diarrhoea

DrugClostridium +ve%Clostridium −ve%p
study groupcontrol group
(n=126) (n=126)
H2‐receptor antagonist
Maintenance dose3243
Therapeutic dose19151814
Proton‐pump inhibitor
Maintenance dose3243
Therapeutic dose21172318
Total463649380.35

View this table:
Table 5

Independent predictors for Clostridium difficile diarrhoea: Multivariate logistic regression analysis

FactorpOdds ratioDirection
Cefuroxime<0.0015.02Positive
Erythromycin/clarithromycin0.0033.01Positive
Ciprofloxacin0.0072.57Positive
Lactulose0.0070.43Negative

View this table:
Table 6

Association between use of cefuroxime and requirements for assisted feeding in both groups

Mode of feedingCefuroxime usedCefuroxime not usedTotal
(n=104)(n=138)(n=252)
Requiring assistance to feed orally5357110
Nasogastric tube feeding123
Peg feeding134
Intravenous hydration only15520
Total requiring assistance for feeding70 (67.3%)67 (45.3%)137
Total able to feed independently34 (32.7%)81 (54.7%)115

View this table:
Table 7

Association between use of cefuroxime and low serum albumin levels in both groups

Serum albuminCefuroxime usedCefuroxime not usedTotal
(n=104)(n=148)(n=252)
Low85 (81.7%)99 (66.9%)184
Normal194968

Discussion

The results of our study confirm that suppression of gastric acid secretion does not increase the incidence of symptomatic acute antibiotic‐associated Clostridium difficile diarrhoea. Previous observations of Walker15 and Simor16 suggest an strong association between the use of drugs that suppress gastric acid and the prevalence of Clostridium difficile asymptomatic carriage among residents of long‐term care facilities. Gastric acid suppression may cause bacterial diarrhoea due to other organisms such as salmonellae, campylobacters and shigellae10 but not Clostridium difficile. The most important predisposing factor for Clostridium difficile infection is the use of broad‐spectrum antibiotics such as cefuroxime, ciprofloxacin and the macrolide group of antibiotics (erythromycin or clarithromycin), a well‐known fact that was confirmed by our study.

Results of previous studies suggest that faecal incontinence15 and disability and dependence on nursing care15,,16 are predisposing factors for Clostridium difficile infection. The results of our study comparing disease‐positive patients (study group) and disease‐free subjects (control group) did not reveal any significant association between urinary or faecal incontinence and Clostridium difficile infection. When further analysed by multivariate logistic regression, these factors did not appear to be strong independent predictors of infection when the use of broad‐spectrum antibiotics was taken into account. There was an association between the use of broad‐spectrum antibiotics and severe disability or dependence on nursing care, but the latter did not add significantly to the predictive power of the use of broad‐spectrum antibiotics by itself.

The significance of assisted oral or tube feeding being a stronger predictor for infection, previously noted by Walker et al.15 was confirmed by univariate analysis in our study (p<0.005, Table 1). Patients requiring assistance for feeding are more susceptible to oro‐enteral spread of infection that may have been transmitted by food handlers and nasogastric or PEG tubes. The number of patients that were fed by PEG tubes in our study were too small in both groups to be statistically significant and larger numbers are needed in order to confirm the increased risk of infection associated with this form of feeding. However, by multivariate analysis, requirement of assisted feeding was not noted to be an independent risk factor for infection because it appeared that a higher proportion of patients in the study group who were too ill to feed themselves were also treated with broad‐spectrum antibiotics such as cefuroxime (Table 6) that subsequently led to the development of Clostridium difficile diarrhoea. Therefore, contrary to the observations of others,15,,16 our study did not confirm that severe disability or the need for assisted feeding were important predictors for Clostridium difficile diarrhoea when the use of broad‐spectrum antibiotics was taken into account.

Simor et al.16 also noted that diabetes was an important risk factor for infection, but we did not observe a similar association between Clostridium difficile diarrhoea and impaired glucose tolerance (p<0.93, Table 1) in our study.

A low serum albumin was noted to be of significance (p<0.01) as a predictor of Clostridium difficile diarrhoea by univariate analysis (Table 1). However, multivariate logistic regression analysis showed that a higher proportion of patients with lowered levels of serum albumin in the study group had received broad‐spectrum antibiotic therapy prior to the onset of diarrhoea, presumably for a severe form of infective illness, which could explain the reasons for a low serum albumin in this group (Table 7). Lowering of serum albumin is a well‐recognized non‐specific response to acute inflammation or infection. An additional point to consider is that a higher proportion of patients in this group were unable to feed independently and their reliance on assisted forms of feeding may have resulted in undernutrition as indicated by a lowered serum albumin. Therefore, a lowered serum albumin did not add any predictive power to the use of broad‐spectrum antibiotics for the development of Clostridium difficile diarrhoea in our study.

Previous studies have implicated drugs that impair gut motility as risk factors for infection. We did not observe any significant association with the use of opiates, anticholinergics, laxatives, sedatives, non‐narcotic hypnotic agents and others that affect gut motility (Table 2). An interesting observation was that lactulose had been frequently prescribed as an aperient in the disease‐free control group. Treatment with lactulose may have resulted in a diarrhoeal illness resembling Clostridium difficile infection, thereby prompting frequent stool sampling for stool Clostridium toxin on these patients; alternatively one can suggest that a change in bowel acidity due to lactulose therapy may be protective against Clostridium difficile infection. Further detailed studies are required in order to confirm this observation, which was not within the scope of the present study.

The use of broad‐spectrum antibiotics has been invariably linked with Clostridium difficile, and third‐generation cephalosporins, and macrolides (erythromycin and clarithromycin) have all been previously implicated, a finding that was confirmed by our study (Table 3). The significance of the use of metronidazole (p<0.036) in the study group which had been prescribed intravenously and in conjunction with other broad spectrum antibiotics, confirms that when used in combined broad‐spectrum antibiotic regimens, metronidazole does not offer any prophylactic benefit in preventing symptomatic Clostridium difficile diarrhoea, an interesting fact, particularly in the context of its usefulness when administered orally for the treatment of symptomatic infection. A significantly higher proportion of cases in the study group had received treatment with cefuroxime (p<0.0004), which confirmed that this antibiotic was noted to be the single most important predictor for infection and also helped to define our study group to be representative of a cohort of patients that were most susceptible to infection with Clostridium difficile during their admission to hospital.

Our study found that severe disability, requirements for assisted feeding and low serum albumin had important associations with Clostridium difficile diarrhoea. However, the most important independent predictor for infection was the use of broad‐spectrum antibiotics such as cefuroxime, macrolide agents and ciprofloxacin. The inter‐relationship between these covariates suggests that any severe and serious premorbid illness associated with systemic infection that requires the use of broad‐spectrum antibiotics, lowers the level of serum albumin, or causes severe disability that requires extensive nursing care and the need for assisted feeding predisposes to Clostridium difficile diarrhoea.

Although our study group of patients with Clostridium difficile faecal‐toxin‐positive diarrhoea were noted to have all the associated risk factors for developing Clostridium difficile diarrhoea, particularly the use of cefuroxime and other broad‐spectrum antibiotics, when compared to disease‐free subjects in the control group, there was no significant difference between the groups for the use of gastric acid anti‐secretory agents. The main hypothesis that gastric acid suppression may predispose to infection with Clostridium difficile has not been confirmed in our study of elderly patients hospitalized in acute general medical and elderly care wards. The results of both univariate and multivariate logistic regression analysis failed to show any significant association between the study and control groups for the use of H2‐receptor antagonists or proton‐pump inhibitors in either maintenance or therapeutic doses in the preceding 16 weeks.

In conclusion, the use of a broad‐spectrum antibiotic such as cefuroxime in the presence of any severe debilitating illness that prevents patients from feeding themselves or lowers their serum albumin levels results is associated with an increased risk of developing Clostridium difficile diarrhoea. Treatment with gastric acid anti‐secretory agents does not appear to increase the risk of developing Clostridium difficile diarrhoea, even though increased incidences of other types of bacterial diarrhoea10 have been associated with this form of therapy.

Footnotes

  • Address correspondence to Professor K.W. Woodhouse, Department of Geriatric Medicine, Academic Centre, Llandough Hospital, Penarth, Vale of Glamorgan CF64 2XX

References

View Abstract