Chapter 41. Antibiotic-Associated Diarrhea and Clostridium difficile-Associated Disease

Antibiotic-associated diarrhea (AAD) or unexplained diarrhea occurring with the administration of antibiotics is a common complication of antimicrobial therapy, for which several underlying mechanisms have been proposed: (1) disturbance of the normal intestinal flora, (2) allergic and/or toxic effects of the drug on the intestinal mucosa, (3) pharmacologic effects on motility, and (4) overgrowth of toxin-producing Clostridium difficile.C. difficile–associated disease (CDAD) is an infection of the colon that develops almost exclusively in the setting of antimicrobial use and is associated with a wide spectrum of symptoms, from mild diarrhea to pseudomembranous colitis, and in some cases toxic megacolon and death. Pseudomembranous colitis, characterized by severe inflammation of the inner lining of the colon with the formation of pseudomembranous material, is usually caused by C. difficile infection. Other toxin-producing pathogens have been associated with AAD, like Staphylococcus aureus,1Clostridium perfringens,1 and Klebsiella oxytoca,2 but they remain infrequent.

C. difficile is found in various natural habitats and in feces of mammals. Patients acquire C. difficile from the hospital environment or from stools of colonized or infected patients, via the hands of medical or nursing staff. In the first days after birth, neonates rarely exhibit C. difficile in their stools.3,4 However, during the first few weeks of life they rapidly become colonized. In a study of preterm infants, colonization rate was 15% at 7 days, increasing to 33% at 14–21 days; toxin B was detected in 90% of stool specimens of those colonized.5 Other researchers reported colonization rates of 31% at 10 days, 71% between 10 and 19 days, 85% between 20 and 29 days, and 100% after 30 days of life.4 Between 25% and 100% of infants aged 6–12 months are colonized3,4,6,7; in this older population, a lower proportion (15–38%) of those colonized have toxigenic strains.7–9 The duration of the carrier state is unknown. Caesarean section, duration of hospitalization, exposure to anti-infectives, underlying comorbidities and exclusive formula-feeding are risk factors for C. difficile colonization.7,9,10

In the general population, AAD occurs in 5–32% of patients between initiation of therapy and up to 2 months after the end of treatment.11–14 Among 650 outpatient children receiving oral antibiotics, 11% developed AAD, beginning a mean of 5.3 (±3.5) days after the start of antimicrobial treatment and lasting a mean of 4.0 (±3.0) days. Younger children (<2 years; 18%) are more likely to experience AAD than their older counterparts (≥2 years; 3%). Certain classes of antibiotics are associated with an increased risk of AAD, the most frequent culprit being amoxicillin/clavulanate (which causes diarrhea in 23% of recipients).15

In adults, toxigenic C. difficile is the most frequent cause of nosocomial AAD and is implicated in 10–30% of cases.16 In children, this relationship is much weaker. Young children, who often are asymptomatic carriers of toxigenic C. difficile strains, are also highly sensitive to the nonspecific diarrheal effects of some antibiotics. Consequently, reports of C. difficile toxin-positive stools in young children with AAD (found in 9–13% of such cases17,18) might overestimate the number of genuine CDAD cases. In a case-control study comparing age-matched toxigenic C. difficile positive and negative patients with diarrhea, no difference was found in symptoms, risk factors, or outcomes (albeit most CDAD cases were treated with oral metronidazole). Alternative etiologies were found in 50% of toxigenic C. difficile patients (viral enteric pathogens in 39%).19 Consequently, confirmed outbreaks of pediatric nosocomial CDAD (n-CDAD) have been rarely documented.20

Since the end of 2002, outbreaks due to a hypervirulent clone have been reported in the United States, Canada, the United Kingdom, France, Belgium, and the Netherlands.21 This has completely changed the epidemiology and clinical impacts of CDAD in adults. In the province of Quebec, Canada, more than 30 hospitals have been struggling with an epidemic of CDAD characterized by a high case-fatality rate (16.7%) and increased risk of recurrence following metronidazole or vancomycin treatment, mostly in patients aged 65 years and older.22–24 These outbreaks have been associated with an emerging toxinotype III ribotype NAP1/027 strain characterized by massive toxins A and B hyperproduction.25 However, little information is available on the impact of NAP1/027 on children. In a retrospective analysis of 200 CDAD cases diagnosed in a university-affiliated pediatric hospital in 2000–2003 (no characterization of strains was performed, but this analysis has taken place in a city plagued by NAP1/027 outbreaks during that period), a majority were outpatients at onset, but 23% needed to be hospitalized for the care of CDAD. Out of 110 patients treated for CDAD, 31% experienced at least 1 recurrence, none required a colectomy and only 2 died.26 The frequent need for hospitalization may reflect the higher virulence of the predominant strain within this pediatric population. In a well-defined population of Canada where NAP1/027 emerged at the end of 2002, the incidence of CDAD increased from 36 per 100,000 in 1991–1992 to 156 per 100,000 in 2003; while the incidence among people aged ≥65 years increased 10-fold, to 866 per 100,000 in 2003, the annual incidence in children remained stable, between 20 and 40 per 100,000.23 This suggests that children are much less prone than the elderly to develop symptomatic infections when exposed to NAP1/027.

Community-acquired CDAD (c-CDAD) incidence is difficult to appraise, since several reports are based on retrospective databases within which past exposure to health care services is not always clear. In 1995, a prospective nationwide study in Sweden reported an annual population incidence of 58 cases per 100,000 inhabitants, combining nosocomial and community-acquired cases. In inhabitants aged 0–19 years, the population incidence rate was 20 per 100,000. To assess the proportion of c-CDAD, they specifically searched for prior contact with health care facilities. An absence of prior hospitalization was reported in 28% of cases, and this proportion increased to 52% in children younger than 4 years.27

Antibiotics exposure within the previous 8 weeks is the strongest risk factor for CDAD, and all types of antibiotics have been implicated. In children, one outbreak was associated with lincomycin usage.20 Among adults, some antibiotic classes are more likely to cause normal flora disruption and CDAD than others, as shown in Table 41–1. Similar data are not available for children. The emergent, hypervirulent, NAP1/027 strain has been consistently associated with the use of fluoroquinolones,22,28,29 which would have limited consequences for children among whom quinolones are rarely used. Host factors associated with CDAD in children include acquired or congenital causes of colonic stasis (e.g., infant botulism, Hirschprung's disease)30–33 and imunosuppression (e.g., cancer, organ transplantation, etc.).34–36

Cystic fibrosis patients are frequently colonized with toxigenic C. difficile strains,37 presumably a consequence of frequent hospitalizations and repeated use of broad-spectrum antibiotics. However, CDAD is unusual in this population, with one case report and a case series of four patients with fulminant CDAD.38,39 Several hypotheses have been raised to explain the scarcity of CDAD in cystic fibrosis patients: unique intestinal environment due to pancreatic insufficiency, higher rate of colonization by bacteria having an inhibiting effect on C. difficile growth, and lack of mucosal receptors for C. difficile toxins.

C. difficile is a gram-positive, obligate anaerobic, endospore-forming bacillus. The ability of C. difficile to form spores under unfavorable growth conditions is the key to its transmissibility and resistance to eradication. Spores have thick walls and are metabolically inactive; thus, they withstand extremes of temperature, drying, atmospheric oxygen, and other unfavorable conditions. Sporulation allows C. difficile to resist the low-level disinfectants used for routine surface cleaning in health care facilities and at home,1 and may also be a key factor promoting endogenous relapses of CDAD after treatment. Spores survive the low pH of the stomach and germinate in the small bowel and colon. C. difficile is harmless in many individuals because its proliferation is limited by competition from the normal colonic flora. Once a patient has been exposed, the risk of developing CDAD is driven by the disruption of normal intestinal flora and host factors (impaired immune response).

In patients with abnormal colonic flora, usually due to antibiotic consumption but also due to antineoplastic agents (some of which have antimicrobial properties), both toxigenic and nontoxigenic strains can proliferate. Only toxigenic isolates cause disease. C. difficile may produce at least three different toxins: toxin A, toxin B, and binary toxin. Toxins A and B disrupt the cytoskeleton of colonic epithelial cells and destroy the epithelial layer. They also induce production of proinflammatory cytokines resulting in marked colonic mucosal inflammation. Binary toxin (actin-specific ADP-ribosyl-transferase), which is similar to the C. perfringens iota toxin, also disrupts the cytoskeleton of cells in tissue culture,40 but its effects in animal models of C. difficile infection have been equivocal.41 Binary toxin was recently identified in 6% of strains in a single center,41 but its gene is present in virtually all isolates of NAP1/027. The NAP1/027 hypervirulent clone has deletions in the tcdC gene, a putative negative regulator of toxin A and B expression, this in turn resulting in increased production of toxins A and B (16–22 times higher than with other strains) and more severe disease.25,42 The contribution of binary toxin to NAP1/027 pathogenesis remains unclear.25

There are several theories to explain why neonates do not exhibit symptoms, despite high colonization rates: immaturity of toxin receptors,43 neutralizing effects of maternal antibodies7,43 and immaturity of neonatal immune system.43,44 Serum antibodies against C. difficile toxins A and B are found in 60–70% of children older than 3 years. In a cohort of infants and children with diarrhea and C. difficile in their stools, 35% had serum IgA below normal range and 16% had low-serum IgG. They were classified as transient hypogammaglobulinemia of infancy, because they all increased their Ig levels over the following 12 months. Patients with hypogammaglobulinemia are more likely to experience a recurrence.45 A specific immune response persists throughout life and may have a protective role against disease occurrence and relapses in adults.46,47

Colonization by C. difficile may result in different clinical presentations: (1) asymptomatic carriage; (2) nonspecific colitis; (3) pseudomembranous colitis (fulminant or not); and (4) recurrent disease. Most patients present with watery diarrhea, albeit 10–15% have bloody stools. Diarrhea usually starts 3–21 days after the use of antibiotics and, prior to the emergence of NAP1/027, was frequently self-limited. Fever and colicky abdominal pains often accompany watery stools. In the most severe cases, a toxic megacolon may develop and lead to perforation and peritonitis. Leukemoid reactions are more common in patients infected with NAP1/027, and a patient with diarrhea and a leukocytosis higher than 25,000 per μL should be considered to have CDAD until proven otherwise. Among elderly adults with NAP1/027 infection, septic shock can develop in the absence of peritonitis. Complicated CDAD (defined by one or more of: toxic megacolon, perforation, shock-necessitating vasopressors, other indication for an emergency colectomy, death within 30 days of diagnosis) has been reported repeatedly in recent outbreaks associated with NAP1/027 in adults but rarely in children.23 In most cases, diarrhea improves within a few days of treatment, but recurrences are frequent and can be difficult to treat.

Differential diagnosis of diarrhea in children is broad and addressed in other sections of this book. Infectious causes of diarrhea are frequently linked to specific exposures, and can be classified as nosocomial or community-acquired, the latter with or without recent travel history (Table 41–2). Nosocomial diarrhea is a frequent nosocomial infection in children. Viral pathogens (rotavirus, adenovirus, etc.) are the primary causes of nosocomial diarrhea, and C. difficile is the most frequent bacterial etiology in industrialized countries. In the developing world, enteropathogens associated with acute diarrhea (Salmonella, Shigella, etc.) are more frequently reported. In the outpatient setting, C. difficile is an infrequent pathogen. When found, it can correspond to CDAD or merely reflect asymptomatic carriage with a concomitant coinfection causing the diarrhea.

In practice, when a patient exposed to antibiotics develops diarrhea, the diagnostic algorithm merely attempts to sort out those with CDAD from all others. In adults, diagnosis of CDAD is usually confirmed by toxin detection in stools, or in selected cases by endoscopy. In the pediatric population, diagnosis is more difficult. First, as endoscopy is more invasive and usually requires anesthesia, it is infrequently performed. Second, asymptomatic carriage of toxigenic C. difficile strains limits the clinical utility of toxin testing in infants.

Culture of C. difficile followed by toxin detection would be the most sensitive test but is cumbersome and used only in research settings. Culture is essential for strain-typing studies that are essential to identify outbreak strains and cross infections. In clinical practice, the cell culture cytotoxicity assay (CTA) is generally considered as the gold standard because of its high specificity and biologic sensitivity (5 pg of toxin B). The main drawbacks are that it is labor-intensive, has a slow turnaround time (≥48 hours), and requires tissue culture facilities and expertise.8 Therefore, an enzyme immuno-assay (EIA) for toxins A and/or B of C. difficile is frequently used for routine testing. When compared to CTA, sensitivity of EIA ranges between 65% and 85%, but its specificity ranges between 94% and 100%.8,16,48 The EIA has a rapid turnaround time (<24 hours), is technically easy-to-perform, and is cheaper than CTA. Although most toxigenic strains (∼97%) produce both toxins A and B, outbreaks associated with strains that produce toxin B only were recently reported.49,50 These toxin A negative strains have been identified in children and can represent as much as 22–48% of toxigenic strains.8,51,52 To circumvent this problem, we recommend selection of A + B EIA tests or CTA. Clinicians must be aware that false-negative results occur and should repeat the test and/or seek diagnosis by endoscopy or CT scan when the clinical suspicion of CDAD is high.

Endoscopy reveals inflamed and edematous mucosa with multiple discrete, nodular, and polypoid lesions covered with yellowish exudates (pseudomembranes) with skip areas of normal mucosa (Figure 41–1). In severe cases, the mucosa may be completely covered by pseudomembranes. Some cases show nonspecific mucosal inflammation without pseudomembrane formation. Usually the findings are so characteristic for experienced endoscopists that a biopsy is not needed. Endoscopy can be useful when stools are not available because of ileus, when a rapid diagnosis is necessary, or when the stool toxin test is negative but clinical suspicion remains.

CT scanning should not normally be needed for diagnosis and probably has a low sensitivity in mild to moderate disease. However, it can be extremely useful in severe cases. Using the criteria of colonic wall thickening >4 mm, pericolic stranding, unexplained ascites, nodularity of the colonic wall, and the accordion sign, a retrospective review of 110 patients reported a sensitivity of 52%, a specificity of 93%, a positive predictive value of 88%, and a negative predictive value of 67%.53 CT findings in children are similar to those in adults.54,55

The first steps in managing CDAD are to discontinue, whenever possible, the inciting antibiotics, and to implement rehydration. Physicians should avoid antiperistaltic agents as they may increase the risk of toxic megacolon (Figure 41–2).56 Sometimes, the diarrhea has spontaneously subsided by the time results of the toxin assay become available, and no antibiotic treatment is needed. In centers where the NAP1/027 hypervirulent clone is present, empirical treatment is started when clinical suspicion is high, pending results of the toxin assay; clinicians should follow closely patients with rapidly progressing symptoms to identify potentially life-threatening complications. In patients with very severe and protracted disease, a surgeon should be consulted to evaluate the need for an emergency colectomy, which can be life-saving.57

A treatment algorithm is shown in Figure 41–3. Traditionally, metronidazole and oral vancomycin have been the main agents used for CDAD treatment. The equivalence of these two drugs is based on two randomized controlled trials that showed no difference in outcomes but included small numbers of patients (<50 per arm).58,59 However, vancomycin is the only FDA-approved drug for treating CDAD and has favourable pharmacokinetics, with faecal levels generally exceeding 1000 mg/L.60,61 For that reason, some authorities recommend that vancomycin be used as the initial treatment of very severe cases of CDAD.62 Its use as initial treatment for CDAD markedly decreased in the mid-1990s following recommendations by the Hospital Infection Control Practices Advisory Committee to avoid using vancomycin in hospitals, to reduce the selection pressure for the emergence of vancomycin-resistant enterococci.63 Since then, metronidazole (20–40 mg/kg per day, in three to four doses, for 10–14 days) has been recommended as the first-line treatment of CDAD, with oral vancomycin (40–50 mg/kg per day, in four doses) to be used mainly if metronidazole is found to be ineffective, contraindicated or poorly tolerated.62 In a retrospective observational study that covered the period 1991–2003, after adjustment for confounding factors reflecting severity of disease and host characteristics, patients who received vancomycin as their initial treatment were less likely to develop a complicated outcome than those treated with metronidazole.23 However, after the emergence of NAP1/027 in the same hospital, vancomycin was no longer superior to metronidazole, perhaps because toxin production is so rapid that the disease follows its natural course.24 Intravenous metronidazole is an alternative for patients in whom an oral treatment is not possible, or in combination with oral vancomycin in complicated CDAD. Intravenous vancomycin is not excreted in the bowel and should never be used to treat CDAD.

Bacitracin is generally considered inferior to metronidazole and vancomycin and is no longer used. Several bacteria and yeasts (Bifidobacterium sp., Lactobacilli GG, Saccharomyces boulardii) have been proposed for prevention of AAD. A recent meta-analysis of randomized placebo-controlled trials of probiotics for AAD prevention demonstrated a lack of effect when analyzing data based on intention-to-treat analysis64 There is no evidence that probiotics reduce the risk of CDAD among patients receiving high-risk antibiotics, nor that they reduce the risk of recurrences. Administration of S. boulardii can cause fungemia, especially in immunosuppresed patients.65 Therefore, we do not recommend probiotics for CDAD treatment or prevention.

Several novel treatments for CDAD are currently investigated in phase III trials, including nitazoxanide (an agent approved for the treatment of cryptosporidiosis and giardiasis),66 rifaximin (a poorly absorbed rifamycin), and tolevamer—a toxin-binding polymer.

In most children, CDAD is a self-limited disease with diarrhea lasting for 4–5 days. Complications associated with CDAD in children are unusual but include dehydration, electrolyte imbalance, shock, hypoalbuminemia, toxic megacolon, and colon perforation. In a 13-year retrospective review of CDAD, 301 cases were identified in patients less than 18 years of age, only 4 of whom (1.3%) developed complicated CDAD (as defined above). Several unusual complications of CDAD have been reported in children: reactive arthritis,67,68 rectal prolapse,69 Henoch–Shonlein purpura,70 and hemolytic–uremic syndrome.71,72 Rare cases of extraintestinal C. difficile infections have been reported in children: bacteremia,73,74 chronic osteomyelitis in a patient with sickle cell disease, and appendicitis with a C. difficile peritonitis.75

Recurrences occur in 20–31% of cases, but are less common in children than in the elderly.26,76,77 About one fourth of children with one recurrence experience a second recurrence within 60 days.77 The first recurrence of CDAD is generally treated with the same antibiotic that was used for the initial episode. For the unfortunate patients who experience more than one recurrence, the most effective strategy is probably the administration of tapered/pulsed oral vancomycin, given qid for 10–14 days, tid for a week, bid for a week, daily for a week, and then every other day for several weeks. It is thought that this provides rapid killing of vegetative forms as the spores germinate, while allowing the normal flora to build up and eventually compete with C. difficile.78 Cholestyramine is poorly effective in patients with multiple relapses.