Clostridioides difficile is a gram-positive spore-forming anaerobic bacterium that is the main etiology of antibiotic-associated diarrhea and one of the most common healthcare-associated pathogens. Clinical manifestations range from asymptomatic colonization to fulminant colitis and death. Most commonly, C. difficile infection (CDI) causes acute-onset watery, profuse diarrhea. The burden of disease is furthered by infection recurrence. Reference Gnocchi, Gagliardi and Gismondi1–Reference Khanna, Baddour and Huskins3 Studies in adults have shown that after a first episode of recurrent CDI (rCDI), 45%–65% of patients will continue to have recurrences that may last for several years, mostly due to poor therapeutic response of rCDI and persistent gut dysbiosis. Reference Gnocchi, Gagliardi and Gismondi1–Reference McFarland, Ozen and Dinleyici4 Thus, rCDI leads to substantially increased medical costs due to additional hospital stays, longer periods of additional precautions, longer courses of antibiotic therapy, and increased morbidity.
Although less common in children than adults, the incidence of pediatric CDI has been on the rise both in hospitals and in the community for >10 years. Reference Gnocchi, Gagliardi and Gismondi1,Reference Nicholson, Thomsen and Slaughter5 A small number of studies have reported the recurrence rate of CDI in children to be between 12% and 30%. Reference McFarland, Ozen and Dinleyici4–Reference Khanna, Baddour and Huskins8 Although risk factors are well established in adults, less is known about what predisposes the pediatric population to rCDI. Recent studies have identified factors such as age, comorbid diseases including inflammatory bowel disease (IBD) and malignancy, prior hospitalizations, tracheostomy tube dependence, recent use of acid-blocking agents, and previous antibiotic exposure as being associated with an increased risk of rCDI in children. Reference Gnocchi, Gagliardi and Gismondi1,Reference Nicholson, Thomsen and Slaughter5,Reference Razik, Rumman and Bahreini9–Reference Kociolek, Palac and Patel11
Therapies for rCDI vary from antibiotic treatment with vancomycin to fecal microbiota transplantation, depending on severity and number of recurrences. Reference Gnocchi, Gagliardi and Gismondi1,Reference Eyre, Walker and Wyllie10,Reference McDonald, Gerding and Johnson12 Evidence for superior effectiveness between vancomycin and metronidazole for the prevention of rCDI in children is not available, and current Canadian treatment guidelines have not established optimal management of multiple recurrences.
The primary objective of this study was to determine the incidence of rCDI in the Canadian pediatric population by risk group, notably immunocompromised patients and those with IBD. The secondary objective was to identify risk factors for rCDI among these at-risk groups, including treatment type at the time of the primary CDI.
Methods
Study design
Through the Pediatric Investigators Collaborative Network on Infections in Canada (PICNIC), we identified tertiary-care pediatric centers interested in participating: The Montreal Children’s Hospital of the McGill University Health Centre (MCH), the Centre Hospitalier Universitaire Sainte-Justine (CHUSJ), and the Children’s Hospital of Eastern Ontario (CHEO).
In this retrospective cohort study, we identified all patients aged 1–18 years with a primary episode of laboratory-confirmed CDI who were hospitalized or evaluated in outpatient clinics or emergency departments (ED) at participating sites between April 1, 2012, and March 31, 2017.
Definitions were based on the Canadian Nosocomial Infection Surveillance Program (CNISP) of the Public Health Agency of Canada (PHAC). 13 The definition of CDI included diarrhea (defined as 3 or more watery stools in 24 hours) and a positive C. difficile laboratory confirmation by nucleic acid amplification test (NAAT), toxigenic culture, or enzyme-linked immunosorbent assay (ELISA). Primary CDI was defined as either the first recorded episode of CDI or an episode of CDI occurring at least 8 weeks after the prior positive diagnostic test date. Reference McDonald, Gerding and Johnson12 Exclusion criteria included a nonprimary episode of CDI (last episode 8 weeks or less prior) and an identified alternate cause of diarrhea according to research team or physician diagnosis, such as concomitant use of laxatives or stool softeners. Children aged <12 months were excluded due to the high prevalence of asymptomatic colonization in this age group. Recurrence, or rCDI, was defined as an episode of CDI occurring ≤8 weeks from diagnostic test date of the primary episode, despite successful treatment and complete resolution of symptoms of the previous infection. Reference McDonald, Gerding and Johnson12
Adverse outcomes recorded were treatment failure, toxic megacolon, colonic perforation, surgical intervention due to CDI, and death caused by CDI. Treatment failure was defined as no clinical improvement after 7 days of appropriate antibiotic treatment. All other adverse outcomes are diagnoses that were determined to be attributed to CDI by the primary medical teams responsible at the time of infection.
We defined immunodeficiency as patients with any 1 or more of the following non–mutually exclusive comorbidities: malignancy, primary or secondary immunodeficiency, and a bone-marrow or a solid-organ transplant.
Data collection
Individuals with positive C. difficile results were identified through laboratory information systems, and their medical records were reviewed to assess eligibility. Charts of patients meeting inclusion criteria were reviewed, and data were extracted by each site’s respective trained research team. Medical progress notes, nursing notes, laboratory results, and medication administration records were reviewed to extract demographic information and variables of interest onto a standardized case report form (CRF). At the end of data collection, deidentified data were sent to the CHUSJ (the coordinating site) for analysis.
Data analysis
Our study had defined an alternate cause of diarrhea to be an exclusion criterion. Although both teams at CHEO and CHUSJ did not remove patients with concomitant IBD from the study under this exclusion criteria, MCH did. This resulted in an internal validity error due to differences in data collection methods. To correct this, we performed 2 separate analyses: the first (primary analysis) included data from all 3 sites but excluded all IBD patients. The second served as a sensitivity analysis using data from CHEO and CHUSJ only, including IBD patients.
The dependent variable was rCDI and the independent variables were risk factors recorded. Continuous variables were presented as medians with their interquartile range, and categorical variables were presented as frequencies. Incidence rates of rCDI in different at-risk pediatric groups were compared using χ2 and Fisher exact tests. Multivariate logistic regression was used to determine and quantify risk factors significantly associated with rCDI with 95% confidence intervals. A P value <.05 was considered statistically significant, and area under the curve (AUC) was used to evaluate model fit. All analyses were performed using R version 3.5.2 software (R Foundation for Statistical Computing, Vienna, Austria).
Results
We identified 588 patients who were diagnosed with CDI between April 1, 2012, and March 31, 2017, and 187 patients did not meet the inclusion criteria. In our primary analysis, 115 patients with IBD were excluded. Of 286 included participants, 37 (12.9%) had an episode of rCDI, henceforth referred to as cases. The incidence proportions were 12.8% (19 of 148) at CHUSJ, 14.9% (7 of 47) at CHEO, and 12.1% (11 of 91) at MCH. The incidence rate of rCDI in all hospitalized patients in this primary analysis was estimated at 2.6 cases of rCDI per 1,000 hospital days at risk (95% CI, 1.7–3.9). All 3 hospitals exclusively used NAAT-based testing for CDI diagnosis.
Overall, very few adverse outcomes were recorded: 8 patients with nonrecurrent CDI (3.2%, 8 of 249) and 2 rCDI cases (5.4%, 2 of 37) had treatment failure of their primary episode, and 1 death was attributed to CDI in a patient who did not have rCDI (Table 1).
Table 1. Basic Characteristics of Study Population of All 3 Study Centers, Excluding Those With Inflammatory Bowel Disease
Note. SD, standard deviation; CDI, Clostridioides difficile infection; rCDI, recurrent CDI; CHUSJ, Centre Hospitalier Universitaire Sainte-Justine; CHEO, Children’s Hospital of Eastern Ontario; MCH, McGill University Health Center; HA, healthcare associated; GERD, gastric esophageal reflux disease.
a Units unless otherwise specified.
b comorbidities: malignancy, primary or secondary immunodeficiency, a bone-marrow transplant, or a solid organ transplant.
c Secondary immunodeficiency: patients aged <1 year or more immunosuppressive therapies.
d HA associated with current admission: CDI symptoms occur 72 h or more after admission.
e HA associated with previous admission: CDI symptoms occur <72 h after the current admission AND the patient had been previously hospitalized and discharged within the previous 4 weeks OR the patient presented with CDI symptoms to the ED or an outpatient location AND the patient had been previously hospitalized and discharged within the previous 4 weeks.
f Healthcare exposure: 2 or more visits at any of the following locations: chemotherapy or radiation therapy session, dialysis, day surgery, day hospital, transfusion clinic, interventional radiology or emergency department OR had a single visit to the emergency department for 24 hours or more.
g HA associated with previous health care exposure: CDI symptoms occur <72 h after the current admission AND the patient had a previous healthcare exposure within the previous 4 weeks OR The patient presents with CDI symptoms to the ED or an outpatient location AND the patient had a previous healthcare exposure within the previous 4 weeks.
h Community associated: CDI symptoms occur <72 h after admission, with no history of hospitalization or any other healthcare exposure within the previous 12 weeks OR presented with CDI symptoms to the ED or an outpatient location with no history of hospitalization or any other healthcare exposure within the previous 12 weeks.
i Indeterminate: Criteria not met for either Healthcare associated related to the current hospitalization, Healthcare associated related to a previous admission, Healthcare associated related to a previous health care exposure, or Community associated.
j Treatment failure: No clinical improvement after 7 d of appropriate antibiotic treatment.
In our sensitivity analysis of patients from CHEO and CHUSJ only, patients with IBD were included, and 126 patients from MCH were excluded. Among the 275 included participants, 40 (14.5%) had rCDI. Incidence rates of rCDI in hospitalized patients in our sensitivity analysis was 5.8 cases of rCDI per 1,000 hospital days at risk (95% CI, 3.9–8.4).
The overall incidence proportion of rCDI was 12.9% (37 of 286). Of 184 patients hospitalized at the time of their primary CDI, 25 (13.6%) developed rCDI. Of 102 nonhospitalized patients (ie, those seen in the emergency department or in outpatient setting), 12 (11.8%) developed rCDI (P = .72). Incidence proportion of rCDI was significantly higher in immunocompromised patients (17.5%, 26 of 153) compared to immunocompetent patients (8.3%, 11 of 133; P = .03) (Fig. 1). Moreover, the incidence proportions of rCDI in patients treated with metronidazole for the primary episode was 13.4% (23 of 172), compared to 5.4% (2 of 37) in those treated with vancomycin (P = .31).
Figure 1. Recurrent CDI (rCDI) incidence proportions amongst immunocompromised and nonimmunocompromised patients from all 3 study centers based on hospitalization status at the time of primary CDI. Hospitalized patients were either hospitalized due to CDI diagnosis or already hospitalized at the time of CDI diagnosis. Nonhospitalized patients were seen in the emergency department or an outpatient clinic. Immunocompromised patients are those with 1 or more of the following factors: malignancy, primary or secondary immunodeficiency, a bone marrow transplant, or a solid-organ transplant.
In our sensitivity analysis, the overall incidence proportion of rCDI was 14.5% (40 of 275). Of patients seen in the emergency room or in an outpatient setting at the time of their primary CDI, 14.5% (11 of 76) developed rCDI, as did 14.6% (29 of 199) of those hospitalized (P = 1). Unlike in the primary analysis, there was no statistically significant difference in rCDI incidence proportions between immunocompromised (17.9%, 27 of 151) nor immunocompetent patients (10.5%, 13 of 124; P = .09), but the trend remained the same. Among all patients included in the sensitivity analysis, incidence proportions of rCDI were 18.7% (23 of 123) in immunocompromised patients with no IBD, 16.7% (6 of 36) in patients with IBD and no history of immunosuppression, 14.3% (4 of 28) in immunocompromised patients with concomitant IBD, and 8.0% (7 of 88) in patients with neither (P = .18).
In our primary analysis, patients with a history of immunodeficiency of any kind had 2.31 times the odds of developing rCDI compared to those with no immune suppression, independent of their age and whether they were initially treated with vancomycin monotherapy rather than metronidazole monotherapy (95% CI, 1.12–5.09; P = .03). Moreover, treatment with only vancomycin did not show statistically significant protection from rCDI when adjusted for immunocompromised status and age (OR, 0.33; 95% CI, 0.05–1.15; P = .14) when compared to patients from any of the other recorded treatment categories.
Similar to our primary analysis, our sensitivity analysis revealed that patients with a history of immune suppression had 1.85 times the odds of developing rCDI compared to those with no immune suppression, independent of their age, IBD diagnosis, and whether they were treated with vancomycin exclusively for their primary episode rather than metronidazole (95% CI, 0.92–3.91; P = .09). Again, treatment of the primary episode with vancomycin monotherapy was not significantly protective for rCDI in this model when adjusted for immunocompromised status, age, and history of IBD (OR, 0.46; 95% CI, 0.07–1.66; P = .31) compared to patients from any of the other treatment categories.
Additionally, in our models, IBD patients who were not immunocompromised had similar odds of developing rCDI to those of patients without IBD, independent of age or treatment with vancomycin monotherapy (Table 2).
Table 2. Multivariate Logistic Regression for Comorbidity Risk Analysis
Note. OR, odds ratio; CI, confidence interval; IBD, inflammatory bowel disease. Includes all patients with immunodeficiency and/or IBD, hospitalized and nonhospitalized, from CHUSJ and CHEO only. Reference groups were (1) patients not immunocompromised with IBD and patients not immunocompromised at all (2) patients with IBD and immunocompromised or patients without IBD at all, and (3) patients immunocompromised without IBD, patients with IBD not immunocompromised, patients neither immunocompromised nor with IBD.
Discussion
In our Canadian retrospective cohort study of pediatric patients with primary CDI, 12.9%–14.5% of patients developed rCDI. This incidence proportion range is similar to published incidences of rCDI of 12%–30% in pediatric populations. Reference McFarland, Ozen and Dinleyici4–Reference Khanna, Baddour and Huskins8 In our study, incidence proportions in the immunocompromised patient population were significantly higher. Although previous studies reported higher rCDI rates in patients with malignancy, Reference Nicholson, Thomsen and Slaughter5,Reference Willis, Nicholson and Esbenshade14 we detected higher incidence proportions in patients with immunodeficiency of any kind: patients with malignancy, primary or secondary immunodeficiency (ie, induced by medication), bone marrow transplant, or solid organ transplant. We also identified higher incidence proportions among IBD patients with or without immunosuppression than in those with neither condition, although these differences were not statistically significant.
Given that IBD patients often receive immunosuppressive medications to treat IBD, we also investigated IBD as a risk factor for recurrence independently of immunodeficiency. IBD patients in this study who were not immunocompromised did not show higher odds of rCDI than patients without IBD. This model, however, lacked statistical significance, likely due to its small sample size. Given the body of evidence that has shown higher CDI recurrence rates in adult and pediatric IBD populations, Reference Razik, Rumman and Bahreini9,Reference Cho, Spencer and Hirten15,Reference Kelsen, Kim and Latta16 it remains unclear whether IBD is a risk factor for CDI recurrence independently of concomitant immunodeficiency. Future studies with a larger sample size should investigate this factor.
A key component of our analysis was that of treatment type as a risk factor for recurrence. Several studies in adults demonstrated higher efficacy and clinical response for vancomycin compared with metronidazole for treatment of primary CDI, both in nonsevere and severe cases. Reference Stevens, Nelson and Schwab-Daugherty17–Reference Johnson, Louie and Gerding19 However, prospective data on treatment superiority between the 2 antibiotics in children are lacking. Although vancomycin is now first-line therapy in adults, most pediatric guidelines still allow a choice of either metronidazole or vancomycin for treatment of a mild-to-moderate primary CDI episode. Reference McDonald, Gerding and Johnson12 Moreover, neither vancomycin nor metronidazole treatment of primary CDI has been associated with higher risk of recurrence, regardless of disease severity in adults. Reference Stevens, Nelson and Schwab-Daugherty17 Our study did not show statistically significant protection from rCDI when the primary episode was treated with vancomycin, independently of immunocompromised status or IBD diagnosis. However, these results may have been due to small sample size, which made the clinical significance difficult to estimate.
Research has shown significantly increased efficacy of alternative therapies such as fidaxomicin and fecal microbiota transplants (FMTs) in reducing recurrence. A 2019 network meta-analysis reported significantly reduced recurrence rates and higher sustained cure rates (clinical cure without any recurrence) in patients treated with fidaxomicin versus vancomycin. Reference Okumura, Fukushima and Taieb20 With recent FDA and Health Canada approval for use in children, studies in pediatrics continue to emerge. In the SUNSHINE trial, a prospective, multicenter, randomized, investigator-blind, phase 3 parallel-group trial, rCDI rates before the end of the study were significantly lower among children who received fidaxomicin compared to those treated with vancomycin. Reference Wolf, Kalocsai and Fortuny21 This study, however, excluded IBD patients. Fecal microbiota transplantation (FMT) has also proven to be a highly effective treatment for rCDI in 2 major randomized clinical trials in adults. Reference McFarland, Ozen and Dinleyici4,Reference Kociolek and Gerding22,Reference Van Nood, Vrieze and Nieuwdorp23 Although observational studies of FMT efficacy for treatment of rCDI begin to emerge in pediatrics, Reference Nicholson, Mitchell and Alexander24–Reference Aldrich, Argo, Koehler and Olivero27 controlled clinical trials have not been performed in the pediatric population to date, and there is little data regarding efficacy of rCDI prevention. Neither fidaxomicin nor FMTs were used in our study, as their administration in pediatrics in Canada remains scarce. Further prospective research into these treatments in pediatric populations at-risk of recurrence may help to identify a therapy that is protective against rCDI.
This study had several limitations. Small sample size of rCDI cases likely played an important role in the sensitivity of our analyses. Moreover, only NAAT was used by all 3 hospitals for the diagnosis of primary CDI or rCDI. Given that this method detects the presence of C. difficile toxin genes, it does not distinguish between true infection and colonization in a patient who may have an alternate cause of diarrhea, especially in children aged 1–3 years. Reference Gnocchi, Gagliardi and Gismondi1,Reference McDonald, Gerding and Johnson12 Although we excluded patients with alternate causes of diarrhea diagnosed at the time of symptom onset, it would be interesting to increase validity by use of more specific methods that detect actual toxin production, such as toxigenic culture or ELISA. Reference Gnocchi, Gagliardi and Gismondi1,Reference McDonald, Gerding and Johnson12
Overall, we detected a non-negligible incidence of rCDI in this pediatric cohort, with higher risk seen in patients with underlying immunodeficiency and in those with IBD. Although the pathogenesis of rCDI is multifaceted, insufficient IgM and IgG titers against C. difficile toxin A has been associated with higher risk of recurrence among adult patients. Reference Kyne, Warny and Qamar28–Reference Gilbert, Leslie and Putler31 Additional studies should measure such immune markers in pediatric patients with and without immunodeficiency.
Although our results on the added benefit of vancomycin as first-line therapy are not statistically significant, they merit further investigation with larger study populations, given that current pediatric guidelines for the treatment of primary episodes of CDI do not yet recommend vancomycin over metronidazole. Moreover, therapies such as fidaxomicin and FMT have shown substantial efficacy in reducing rCDI in randomized and observational studies in adults. Future prospective studies should seek to elucidate their efficacy in at-risk pediatric populations to try to reduce the burden of rCDI in these groups.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/ice.2023.174
Acknowledgements
We thank the nurses at IPAC Canada (Infection Prevention and Control).
Financial support
No financial support was provided relevant to this article.
Conflicts of interest
C.Q. is the Canada Research Chair, Tier 1 in infection prevention from hospital to the community. All other authors report no conflicts of interest relevant to this article.
