Clostridium difficile and Candida sp. are both important causes of healthcare-associated infections [Reference Lessa1, Reference Magill2]. Risk factors are similar for Candida infections and C. difficile infections (CDI), and include broad spectrum antibiotic (BSA) use and prolonged intensive-care unit (ICU) stays [Reference Bignardi3, Reference Wey4]. CDI itself and CDI treatment with certain antibiotics may lead to disruption of the indigenous microbiota, predisposing patients to overgrowth of endogenous pathogens such as Candida sp. and dissemination into the blood through the compromised mucosal barrier present during CDI [Reference Nerandzic5, Reference Guastalegname6]. The purpose of our study was to describe the prevalence and characteristics of adults developing candidaemia following CDI and evaluate the factors associated with developing candidaemia following CDI.

The CDCs Emerging Infections Program conducts population-based surveillance for CDI and candidaemia in the Atlanta area (population 3·8 million). The methods of CDI and candidaemia surveillance have been described previously [Reference Lessa1, Reference Cleveland7]; in brief, both surveillance systems rely on active laboratory-based case-finding in residents of the catchment area, followed by medical record abstraction for identified cases.

A CDI case was defined as a positive C. difficile toxin or molecular assay in a surveillance area resident aged ⩾18 years and a candidaemia case was defined as a Candida sp.-positive blood culture collected from a surveillance area resident aged ⩾18 years. Infection with CDI and candidaemia was defined as candidaemia occurring in the 120 days after CDI during 1 September 2009–31 December 2013. Cases with both infections were identified by merging the two surveillance databases by patient's name and date of birth.

To calculate the prevalence of candidaemia in adults with CDI, we divided the number with co-infection by the number with CDI in a given year. We used the Cochran–Armitage test for trend. We conducted a matched case-control study to identify factors associated with candidaemia following CDI. Controls were defined as CDI cases without candidaemia who had survived at least 120 days post-CDI episode. For each case with both candidaemia and CDI, we identified up to three controls matched by age group (18–44, 45–64, ⩾65 years) and location of CDI disease onset [healthcare facility onset (HCFO), community-onset healthcare-facility associated, community associated]. Not all cases had three controls available because only 10% of the HCFO CDI cases had a full medical record review based on surveillance methodology [Reference Lessa1]. We used conditional logistic regression analysis to identify factors associated with candidaemia following CDI. In all analyses, the level of significance was set at α = 0·05.

Of 13 615 adults with CDI, we identified 113 (0·8%) patients who developed candidaemia in the 120 days after CDI. The prevalence of candidaemia following CDI declined from 1·4% in 2010 to 0·6% in 2013 (P < 0·001), and varied by age group, with the highest prevalence in cases aged 45–64 years (1·1%) compared to the other two age groups (18–44 years: 0·8%; ⩾65 years: 0·7%; P = 0·036).

Of the case-patients, median age was 62 years [interquartile range (IQR) 53–70 years], 54% were female, 54% were black, and 77% had CDI onset in a healthcare facility (Table 1). Diabetes (46·9%) was common. Vancomycin, β-lactam/β-lactamase inhibitors, and fluoroquinolones were the most common antibiotics administered to case-patients in the 12 weeks before CDI onset. Fifty (44·3%) case-patients experienced severe CDI. For CDI treatment, 40·7% of case-patients received metronidazole only, 10·6% received vancomycin only, and 39·8% received vancomycin + metronidazole; no case-patients received fidaxomicin.

Table 1. Demographic and clinical characteristics of case-patients with candidaemia following Clostridium difficile infection (CDI) and controls (CDI infection only), Atlanta metropolitan area, 2009–2013

OR, Odds ratio; CI, confidence interval.

† Controls were matched on age group (18–44, 45–64, ⩾65 years) and location at the time of CDI onset. Note that not all cases had 1:3 matching; HCFO cases in age groups 18–44 and 45–64 years had only one or two controls each because of lack of availability of controls in these groups.

‡ Refers to 12 weeks before CDI onset.

§ Severe CDI = having one or more of the following: ileus, toxic megacolon, pseudomembranous colitis (within 5 days before or after initial C. difficile + stool, or a white blood cell count >15 000 cells/μl within 1 day before or after initial C. difficile + stool.

* Significant P value.

The median time between CDI and candidaemia diagnosis was 19 days (IQR 8–45 days). Eighty-eight percent of case-patients had a central venous catheter in place in the 2 days before candidaemia diagnosis. Candida albicans was the most frequent species recovered (37·2%). Thirty-three (29·2%) case-patients died within 30 days of candidaemia.

We identified 257 controls for 113 cases. Compared to controls, cases had higher odds of being black [matched odds ratio (mOR) 1·87, 95% confidence interval (CI) 1·16–3·05], having diabetes (mOR 1·66, 95% CI 1·01–2·77) or inflammatory bowel disease (mOR 5·07, 95% CI 1·29–23·67), and having a higher Charlson comorbidity index score (mOR 1·99, 95% CI 1·13–3·63) for score of ⩾2 vs. <2). In the 12 weeks before CDI, cases also had higher odds of having received proton pump inhibitors (mOR 2·10, 95% CI 1·28–3·53), or BSAs (see Table 1 for individual mOR). Cases were also more likely to have severe CDI (mOR 2·10, 95% CI 1·29–3·42), colectomy (mOR 13·26, 95% CI 1·54–114·08), ICU admission (mOR 2·31, 95% CI 1·00–5·39), and to receive CDI treatment with vancomycin + metronidazole compared to metronidazole alone (mOR 2·08, 95% CI 1·18–3·74).

Using data from two active population- and laboratory-based surveillance systems, we evaluated over 10 000 CDI patients for development of candidaemia and identified a total of 113 persons with both infections. We found that candidaemia following CDI was rare (<1%). The decline observed in the prevalence of CDI/candidaemia is similar to the decline in trends in overall candidaemia rates in the Atlanta area [Reference Cleveland7]. Distribution of Candida sp. causing infection in CDI patients was also similar to that reported in the general population [Reference Cleveland7]. Some factors, such as black race, diabetes, BSA use, and ICU admission, which have been reported in the past as risk factors for candidaemia [Reference Wey4, Reference Fanfair8], continue to be important in patients with preceding CDI. However, this study also identified new potential risk factors for candidaemia in CDI patients, including severe CDI and CDI treatment with both vancomycin + metronidazole.

Treatment with vancomycin + metronidazole was associated with higher odds of developing candidaemia after CDI, compared to metronidazole alone. While treatment with both antibiotics could be a marker of severity of CDI illness [Reference Cohen9], it may also be independently associated with development of candidaemia as more disruption of mucosal integrity and disruption of gut microbiota, including suppression of anaerobes and overgrowth of Candida has been reported with vancomycin treatment [Reference Nerandzic5, Reference Lewis10]. Although we do not report on the relationship between daily dose of vancomycin used to treat CDI and risk of subsequent candidaemia in this study, previous research has shown that higher daily doses of vancomycin (⩾500 mg/day) was an important risk factor for development of bloodstream infections, including candidaemia, following CDI [Reference Falcone11]. Poor outcomes, including recurrent CDI and worse histopathology have also been noted in mice treated with vancomycin compared to untreated mice [Reference Warren12]. There is the potential for even greater disruption of the normal microbiome with exposure to two CDI antibiotics. Severe CDI may contribute to an increased risk of candidaemia due to severe mucosal damage, which may facilitate translocation of Candida, particularly in the setting of Candida overgrowth occurring during CDI treatment. Severe CDI may also be a marker for patients who require ICU admission, surgical intervention, or placement of a central line, all of which are known to be risk factors for candidaemia. Our findings are consistent with a recent smaller study from Italy in which severe CDI was also found to be associated with candidaemia [Reference Russo13].

This study has several limitations. Although we were not able to conduct a multivariable analysis because data on antibiotic use after CDI diagnosis, other than those used for CDI treatment, and presence of a central line, both important risk factors for candidaemia [Reference Wey4], were not available for all CDI patients, we were able to find some associations that will be important for evaluation and confirmation in future studies. We were unable to identify three controls per case due to surveillance methodology, which may have limited our ability to find significant associations. The effect of fidaxomicin use on development of candidaemia could not be evaluated since no cases received fidaxomicin.

The prevalence of candidaemia in patients with recent CDI is low. However, clinicians should be vigilant for candidaemia in CDI patients who have previously recognized risk factors for candidaemia, and those with severe CDI or CDI treatment with certain two-drug regimens. Further research is needed to evaluate independent risk factors for candidaemia following CDI.