Risk factors for Recurrent Clostridioides Difficile Infection in Children Free

We conducted a 13-year review (January 1, 2008—December 31, 2021) of CDI on the basis of International Classification of Diseases 9th and 10th Revision codes among children cared for at a major urban quaternary hospital. Data were included for pediatric patients aged 2 to 21 years who met all the following CDI criteria: (1) diarrhea, which was defined as 3 or more loose stools per day as reported by the patient, family member, or health care provider and recorded in the medical record; (2) presence of C. difficile toxin confirmed by either enzyme immunoassay (EIA) or polymerase chain reaction (PCR) and (3) treatment of CDI. Stool testing consisted of a screening EIA to detect glutamate dehydrogenase (GDH) (Dynex Technologies, Chantilly, VA or Alere, Waltham, MA), which detects all C. difficile strains, regardless of toxin production. Toxin production was detected using an EIA for toxin A/B (Alere). Samples with discrepant GDH and toxin A/B results were tested by using a PCR for the toxin B gene (Cepheid, Sunnyvale, CA). The same definition of CDI was used for both initial and recurrent episodes. We reviewed charts for the following information: demographics, number of CDI episodes, route of contraction (ie, community versus nosocomial), disease severity, treatment of each episode, past medical conditions, immune status, markers of disease severity (ie, white blood cell, albumin and creatinine), and recurrence of symptoms. Antibiotic usage (yes or no) and acid reduction therapy (proton pump inhibitors or H2 blockers) within 30 days of the initial episode of CDI, and for 1 year after the initial episode were recorded. C. difficile–specific therapy for each episode of CDI was determined and further classified as metronidazole alone, vancomycin alone, vancomycin combined with metronidazole, vancomycin with taper, nitazoxanide, or fidaxomicin. The percentage of children receiving vancomycin relative to those treated with either vancomycin or metronidazole was determined for each year. All children with a diagnosis of CDI were included, regardless of whether they received their care on an inpatient or outpatient basis. The study was approved by the institutional review board.

Patients were classified as having either a single episode (sCDI) or >1 episode (recurrent) (rCDI). Treatment failure was defined as the recurrence of disease occurring within 4 weeks to 1 year after initial treatment. Cure was defined as no further episodes during that period. Children with rCDI were further classified as follows: (1) early recurrence, a repeat CDI episode ≥4 weeks and ≤2 months from the preceding CDI episode; (2) late recurrence, a repeat CDI episode >2 months and <1 year from the preceding CDI episode. Recurrences occurring 1 year or more after initial infection were defined as cure with recurrence (cwrCDI). These episodes were noted but not included in the primary analysis. CDI severity was classified based on a review of laboratory features within 48 hours of admission, including the following: white blood cell: ≤2 or >15 k/μL, albumin: ≤3 g/dL, and creatinine: ≥0.5 mg/dL. Patients with either none, 1, or 2 or more of these laboratory features were classified as having mild, moderate, or severe disease, respectively.¹²  The onset of disease was classified based on timing of symptom development using the National Healthcare Safety Network guidelines as follows: community onset (≤3 days before hospitalization), hospital onset (≥4 days of hospitalization), and community onset-hospital acquired (≤3 days before hospitalization but within 4 weeks of hospitalization).¹³ 

Preexisting medical conditions (if any) were noted and further classified on the basis of immune status. Patients were categorized as immunocompromised if they suffered from any of the following conditions: inflammatory bowel disease, organ transplantation (solid-organ and stem cell), malignancy, HIV infection, or were receiving other immunosuppressive therapies, including biologicals.

Descriptive analyses were performed to calculate summary statistics for sociodemographic and clinical variables, medications that were administered, and the overall outcome achieved (single episode cure versus need for 1 or more repeated treatments). For the analysis of the effects of treatment of recurrence, children not receiving metronidazole of vancomycin as primary therapy (n = 8) were excluded. We assessed bivariate associations of sociodemographic and clinical variables with severity and with overall outcome using χ-square tests. Multivariable logistic regression was then performed to analyze risk factors for recurrence among children with an underlying chronic disease. For this analysis, we included variables that were significantly associated with recurrence on bivariate analysis. We report unadjusted and adjusted odds ratios (ORs) and 95% confidence intervals (CIs) for the bivariate and multivariable analyses, respectively. IBM SPSS Statistics version 27 was used for all analyses (IBM SPSS Statistics for Windows, Version 27, Armonk, NY IBM Corp). The number needed to treat (NNT) was calculated using ClinCalc.com (www.clincalc.com/stats/NNT.aspx)

Using DSM codes, 468 patients with a CDI diagnosis were identified, whereas 165 patients did not meet study inclusion criteria and were excluded leaving 303 patients for analysis. Reasons for exclusion included: <2 years of age, >21 years old, not experiencing diarrhea, did not have a positive test, and did not receive treatment (Fig 1). Approximately 50% of the children (150 of 303, 49.5%) identified as Hispanic; and 27.4% (83 of 303) identified as Black (see Table 1), which is consistent with the population served by our hospital. Many children (163 of 303, 53.8%) had an underlying immunocompromising condition, with the most common being malignancy followed by inflammatory bowel disease. Another 28.1% (85 of 303) had a chronic condition that was not associated with decreased immunity, the most common condition involving the central nervous system (Supplemental Table 3). Recent antibiotic usage (199 of 303, 65.7%) and acid reduction therapy (146 of 303, 48.3%) were common among children with CDI. Cephalosporins were the most commonly used antibiotics, whereas H2 blockers were the most common form of acid reduction therapy (Supplemental Figs 4 and 5). Mild, moderate, and severe disease occurred in 171 of 303 (56.4%), 117 of 303 (38.6%), and 15 of 303 (4.9%) of patients, respectively. Most patients with CDI were inpatients, with this cohort accounting for 255 of 303 (84.2%) of children with an initial CDI diagnosis and for 43 of 68 (63.2%) of children with recurrent disease.

Out of 303 patients with CDI, 210 patients received metronidazole as initial therapy, while 85 received vancomycin (Fig 1). The remaining 8 patients were treated with a variety of regimens: vancomycin taper, vancomycin combined with metronidazole, nitazoxanide, and fidaxomicin. The annual percentage of children receiving vancomycin for an initial episode of CDI ranged from (0%–66.7%) with the highest percentage of vancomycin usage in the last 3 years of this study, ranging from 37.5% to 66.7% (Supplemental Figure 4).

Two-hundred and thirty-five children (77.6%) had sCDI, while 68 (22.4%) experienced rCDI (Table 1). There was no difference in age or ethnicity between children with sCDI and rCDI. However, rCDI was more common among females (Table 1). All children with rCDI had an underlying condition, including 51 (75%) who were immunocompromised and 17 (25%) who had a chronic condition not related to immunosuppression. The most common immunosuppressive conditions (in decreasing order) were malignancy 78 of 163 (47.9%), inflammatory bowel disease 39 of 163 (23.9%), and solid organ transplant recipient (SOT) 21 of 163, (12.5%). The most common chronic conditions not associated with immunosuppression involved the following systems: neurologic 8 (47.1%), gastrointestinal 3 (17.6%), genitourinary 2 (11.8%), and endocrine 2 (11.8%). Six patients had cwrCDI; 2 were immunocompromised, and the remaining 4 had a chronic condition not associated with immunosuppression. Among children with rCDI, the median number of total episodes (including the initial episode) was 2 (range 2–5) with a mean of 1.50±0.72 repeat episodes. Of the 68 patients with rCDI, 25 (36.8%) had early recurrence and 43 (63.2%) had late recurrence. The mean times to recurrence for early and late disease were 49.8±11.4 and 162.2±83.2 days respectively (Fig 2).

rCDI was significantly more common among patients initially treated with metronidazole (54 of 210, 25.7%,) compared to vancomycin (12 of 85, 14.1%; P = .030, OR 2.11, 95% CI 1.06–4.02, NNT 8.6) (Fig 3A). A similar pattern was present when only children with chronic conditions were analyzed (Fig 3B, NNT 6.9). Importantly, there were no differences in the proportion of children who were immunocompromised or had a chronic condition among children receiving vancomycin or metronidazole as initial therapy (P = .88). Likewise, there was no difference in disease severity among patients receiving either vancomycin or metronidazole for initial treatment (P = .30). To understand the contribution of time to the risk of recurrence, several analyses were done. No difference in recurrence rates by year among this cohort were observed (P = .13). Furthermore, no difference in recurrence rates were noted pre- and post- 2017, the year of the most recent IDSA guidelines were present (P = .15).¹¹ 

In addition, the effects of ongoing antibiotic and acid reduction therapies were analyzed with respect to the risk of late recurrence. Among the entire cohort, antibiotic usage between 2 months and 1 year after an initial CDI episode was more common for those with a late recurrence compared to those with sCDI, 26 of 41 (63.4%) versus 104 of 229 (45.4%), respectively (P = .042). However, among children with a chronic condition (ie, those children that actually experienced recurrences), there was no difference in antibiotic usage between sCDI and late recurrence groups (96 of 180, 53.3% versus 26 of 43, 60.4%, respectively, P = .5.). Likewise, among children with a chronic condition, no difference in the frequency of acid reduction therapy was noted for patients with sCDI compared to those with late recurrence, 92 of 180 (51.1%) versus 20 of 43 (46.5%), respectively (P = .87).

On bivariate analysis, female sex (P = 0.001, OR 2.65, 95% CI: 1.44–4.70) and initial treatment with metronidazole (P = 0.026, OR 2.19, 95% CI: 1.11–4.28) were the only variables associated with an increased risk of any recurrence (early or late) among children with a chronic disease. Moreover, these associations remained significant in logistic regression analysis with both variables included (P values .002 and .04, respectively)

Regardless of the initial regimen, treatment failure rates for the first rCDI episode were very high for vancomycin and metronidazole, 9 of 26 (34.6%) and 16 of 38 (42.1%), respectively (P = 1.0) (Table 2). Of the 54 patients who had recurrence after initial treatment with metronidazole, 36 were retreated with metronidazole, of which 22 (61.1%) achieved cure and 14 (39.9%) experienced another CDI episode. Vancomycin was used to treat the other 15 patients who experienced an initial metronidazole failure; 10 of these patients (66.6%) experienced cure, whereas the other 5 (33.3%) experienced recurrence. Of the 12 patients who had recurrence after initial vancomycin therapy, 1 was treated with metronidazole, and this patient had a subsequent late recurrence. Eleven patients were retreated with oral vancomycin; 7 of these patients achieved cure, whereas 2 had early and 2 had late recurrences. The failure rates for vancomycin taper strategy, nitazoxanide, and fidaxomicin were low, but the total number of cases was too small to analyze. Of the 6 patients who experienced cwrCDI, all were initially treated with metronidazole, 5 were retreated with metronidazole, and 1 with vancomycin with their repeat episodes.

Findings from our study demonstrate that pediatric CDI is associated with relatively high rates of mild disease and recurrence, as previously reported.^2,3,14  We, like others, have found that rCDI disproportionately affects children with chronic underlying diseases.^8,9  In our study, rCDI occurred exclusively in children with a chronic condition. This observation has significant implications for our understanding of rCDI pathogenesis and its treatment. Many children with rCDI were immunocompromised, consistent with the hypothesis that the immune system, including antibody against toxin B, plays an important role in preventing recurrence.¹⁵  However, not all children were immunocompromised. Chronic neurologic, genitourinary, and gastrointestinal problems (Supplemental Table 3) were also common among patients with rCDI, leading us to hypothesize that intestinal dysmotility in conjunction with antibiotic-induced dysbiosis contributes to the development of rCDI. Intestinal dysmotility is difficult to measure and, as a result, its role in rCDI is less appreciated. However, it has been suggested that dysmotility leads to recurrence by slowing the elimination of C. difficile from the intestinal tract though additional study is needed to confirm this theory.^16,17  A female predominance in children with rCDI was also present in our study and has been previously been reported in some but not all studies, though the basis for this remains unclear^18–20 

Based on CDC criteria, rCDI is typically defined as a repeat episode occurring between 2 weeks and 2 months after initial infection. In contrast, repeat episodes occurring 2 months after initial infection are defined as new infections (https://www.cdc.gov/hai/eip/cdiff-tracking.html). These are functional definitions to help track the acquisition of disease within the hospital. Despite this definition, several studies in adults have demonstrated considerable overlap between relapse and reinfection with respect to time from initial infection based on molecular identification of the infecting strains.^21–23  A study of 102 adult patients, found that 88% of recurrent episodes within 2 months of infection were due to relapse, whereas 65% of episodes after 2 months were because of relapse.²¹  In a recent pediatric study, ∼72% of infections occurring after 2 months represented relapsed disease.²⁴  Given these findings, we defined both early and late repeat episodes as recurrences, acknowledging that a precise distinction based on pathophysiology is impossible without molecular data. At the same time, this approach allows us to assess the effects of initial treatment on long-term outcomes, which is important regardless of the underlying disease mechanism.

Among children with an underlying chronic disease, initial metronidazole use was associated with rCDI both on bivariate and multivariate analysis (NNT for vancomycin 6.9). This finding conflicts with current treatment guidelines, in which metronidazole remains a drug of choice for initial treatment of CDI in all children, regardless of the presence of a chronic condition.¹¹  These guidelines are based in part on the observation that children tend to get less severe CDI in general and that metronidazole and vancomycin have comparable efficacy and risk of recurrence in the treatment of mild CDI.^25–27  Nonetheless, most studies have been confined to adults and define rCDI as occurring within 8 to 12 weeks of initial episode, potentially missing a significant number of late repeat episodes that were likely because of recurrence. Interestingly, in a study looking at mild CDI in adults, metronidazole was associated with a greater risk of new infection (defined as a new episode after 8 weeks) compared to vancomycin.²⁸  There is less pediatric data regarding specific C. difficile treatment and the risk of recurrence, and the existing data are conflicting.^3,20,29,30 

Recent observations have indicated other advantages to the use of vancomycin over metronidazole for the treatment of CDI. In a retrospective review of 196 children hospitalized with CDI, vancomycin demonstrated more rapid improvement in symptoms when compared to metronidazole, although no differences in rates of recurrence were observed.³⁰  In other adult studies, vancomycin was superior to metronidazole in eradicating C. difficile from stool cultures, improving symptomatology, and preventing spread within the hospital.^31,32  Consistent with these findings, metronidazole is no longer considered a drug of choice for treating CDI in adults, regardless of disease severity.¹¹  We hypothesize that vancomycin promotes more rapid clearance of C. difficile from the intestinal microbiome of children with CDI, lowering their risk of relapse later. This effect would be especially important in hosts with deficient immune responses and chronic conditions. It is also possible that vancomycin and metronidazole had differential long-term effects on the microbiome (in terms of variability or other protective organisms) resulting in different risks for later infection.

There was a remarkably high treatment failure rate for children with rCDI in our study. Unfortunately, treatment options for pediatric rCDI are more limited and have been less extensively studied when compared to adults. While fidaxomicin has recently been approved for use in children, more data on this agent,^33,34  bezlotoxumab, and fecal microbiota transplantation are needed to optimize the treatment of rCDI in children.^35,36  We note that the gastrointestinal microbiome of children is significantly different from adults, and it is not clear that the treatment used for adults can be assumed to be optimal for children.^37,38 

Our study is limited by its retrospective nature, which may have introduced bias related to which treatment regimen individual patients received. Nonetheless, we found no differences in the overall types of patients initially treated with metronidazole versus vancomycin, nor in the severity of initial disease among treatment groups. In contrast to previous studies, we included within our definition of recurrent disease episodes >2 months after the initial episode. This definition and the observational nature of this study may have allowed for unrecognized confounders between episodes (ie, antibiotic exposures) and in turn affect recurrence rates between the groups, which is best addressed in a prospective study. However, we did not find differences in the amount of antibiotic usage for patients with late recurrence on the basis of whether they had received vancomycin or metronidazole initially. We did notice an apparent increase in vancomycin usage during the last few years of our study and it is possible that unrecognized factors over this time (eg, strain differences) contributed to our results. Finally, our study was conducted at a single site, potentially limiting the generalizability of its findings.

Despite these limitations, this review highlights the relationship between rCDI and chronic disease in children. The basis for this relationship, including the specific mechanisms, remains to be determined. Our findings suggest that vancomycin may be superior in preventing recurrences when compared with metronidazole in children. This finding may be particularly important in planning the initial treatment of CDI in children with underlying conditions at risk for recurrent disease and for whom effective treatment options for recurrence are limited. Future, prospective studies are needed to confirm our findings. We plan to further investigate the role of initial CDI treatment including newer treatment choices, specifically fidaxomicin on long-term outcomes by examining effects on C. difficile persistence as well as associated alterations in the gastrointestinal microbiome.

We thank Carlos Cruz for his critical help in data analysis of medical records and Dr Hnin Khine for her kind help in reviewing the manuscript.