Epidemiology, risk factors and outcomes of Candida albicans vs. non-albicans candidaemia in adult patients in Northeast China

This study aimed to evaluate the clinical characteristics, risk factors and outcomes of adult patients with candidaemia caused by C. albicans vs. non-albicans Candida spp. (NAC). All adult hospitalised cases of candidaemia (2012–2017) at a tertiary hospital in Shenyang were included in the retrospective study, and a total of 180 episodes were analysed. C. parapsilosis was the most frequently isolated species (38.3%), followed by C. albicans (35.6%), C. glabrata (13.9%), C. tropicalis (10%) and others (2.2%). As initial antifungal therapy, 75.0%, 3.9%, 5.6% and 2.2% of patients received fluconazole, caspofungin, micafungin and voriconazole, respectively. Multivariate analyses revealed that total parenteral nutrition was associated with an increased risk of NAC bloodstream infections (BSI) (OR 2.535, 95% CI (1.066–6.026)) vs. C. albicans BSI. Additionally, the presence of a urinary catheter was associated with an increased risk of C. albicans BSI (OR 2.295 (1.129–4.666)) vs. NAC BSI. Moreover, ICU stay (OR 4.013 (1.476–10.906)), renal failure (OR 3.24 (1.084–9.683)), thrombocytopaenia (OR 7.171 (2.152–23.892)) and C. albicans (OR 3.629 (1.352–9.743)) were independent risk factors for candidaemia-related 30-day mortality, while recent cancer surgery was associated with reduced mortality risk (OR 26.479 (2.550–274.918)). All these factors may provide useful information to select initial empirical antifungal agents.


Introduction
Candida is an important causative organism of bloodstream infections (BSIs). Over the last two decades, candidaemia has been reported as the fourth and seventh most common healthcare-associated BSI in US and European studies, respectively [1,2]. In a recent multicentre point-prevalence survey, Candida species emerged as the most common bloodstream pathogen and accounted for up to 22% of healthcare-associated BSIs [3]. Additionally, candidaemia remains associated with high-mortality rates, prolonged hospital stays and increased healthcare costs [1,[4][5][6][7]. Overall, mortality rates among patients range from 19.6% to 67% worldwide [1,6,[8][9][10][11][12] and the major risk factors for candidaemia include receipt of parenteral nutrition, exposure to broad spectrum antibiotics, presence of central venous catheter (CVC), prior surgery and ICU stay [13,14]. Even though C. albicans overall accounts for the majority of Candida spp. causing candidaemia, the proportion of non-albicans Candida (NAC) spp. is rising and has even exceeded that of C. albicans in some recent studies from Europe, Asia and America [8][9][10][15][16][17].
In general, the most frequently isolated NAC spp. from candidaemia cases include C. glabrata, C. parapsilosis, C. tropicalis and C. krusei [5,10,16,17]. Reduced in vitro susceptibility to antifungal agents has been observed among several NAC spp., which may present a therapeutic challenge. C. glabrata and C. krusei tend to be less susceptible to azole agents than other Candida spp., while C. parapsilosis shows the highest minimum inhibitory concentrations (MICs) to the echinocandins than other Candida spp. [18,19]. Several retrospective studies suggest that delayed initiation of antifungal therapy after the onset of candidaemia is associated with increased mortality, which highlights the importance of early appropriate antifungal therapy [20]. The local epidemiology and variable antifungal susceptibility profiles of different Candida spp. are critical for the selection of antifungal therapy prior to culture and susceptibility data being available.
However, risk factors related to NAC BSI, and the distribution and antifungal susceptibility of Candida spp. from BSI differ geographically, whereas data on risk factors, antifungal susceptibility and outcomes in C. albicans and NAC BSIs remain scarce in Northeast China. Therefore, this retrospective study was performed to compare clinical characteristics and outcomes between C. albicans and NAC candidaemia, analyse prognostic factors and determine risk factors related to C. albicans or NAC at a tertiary hospital in Shenyang.

Subjects and study design
This retrospective observational study was undertaken from January 2012 to October 2017 at a tertiary grade A comprehensive hospital in Shenyang, China. The setting is a China Medical University-affiliated teaching hospital with over 6000 beds currently. All adult hospitalised cases of candidaemia were included in the study. The following data were collected from medical records: age, sex, major underlying diseases and predisposing factors, surgical and invasive procedures, complications, blood tests, isolated Candida spp., antifungal susceptibility, antifungal therapy, duration of hospitalisation and final outcome. The study was approved by the Medical Ethics Committee in Shengjing Hospital of China Medical University (reference number 2018PS506K).

Microbiological methods
Blood samples were drawn under sterile conditions and processed using a BD BACTEC 9120 or BD BACTEC FX400 (Becton Dickinson Diagnostic Instrument Systems, Sparks, USA) blood culture system. After the identification of Candida isolates utilising the VITEK 2 YST (bioMérieux, Durham, USA) card, antifungal susceptibility testing of fluconazole, itraconazole, voriconazole and amphotericin B was determined using a ATB FUNGUS 3 (BIOMERIEUX) strip following the manufacturer's instructions. The interpretation of antifungal susceptibility of fluconazole, itraconazole, voriconazole and amphotericin B against C. albicans, C. glabrata, C. parapsilosis, C. tropicalis and C. krusei was performed by applying clinical breakpoints (CBPs) defined by the Clinical Laboratory Standards Institute (CLSI) [21] or European Committee on Antimicrobial Susceptibility Testing (EUCAST) [22]. MIC CBPs for echinocandins and 5-fluorocytosine by the CLSI or EUCAST were not available at the time, and therefore are not reported in the current study.

Clinical definitions
An episode of candidaemia was defined as the isolation of a Candida spp. recovered from the first blood culture in a patient with accompanying clinical signs and symptoms. Subsequent positive cultures from the same patient were considered as a new episode if the interval between the two episodes was more than 30 days. The onset of candidaemia was taken as the day the first positive blood culture for Candida spp. was drawn from the patient. Except for recent surgery, defined as y within 3 months, the predisposing factors occurred within 30 days prior to the onset of candidaemia. Elderly patients were defined as those 65 years of age or older. Laboratory test results were obtained at the onset of candidaemia. Renal failure was defined as a serum creatinine above 104 µmol/l. Anaemia was defined as a haemoglobin level below 110 g/l for women and 130 g/l for men. Hypoalbuminaemia was defined as albumin <30 g/l and hyponatraemia as a serum sodium concentration <135 mmol/l. Leukopaenia was defined as a peripheral white blood cell count less than 4 × 10 9 /l and thrombocytopaenia as a platelet count <100 × 10 9 /l. Antifungal therapy was defined as empirical if started before susceptibility test results were obtained and as targeted thereafter. The outcome was registered after 30 days from the onset of candidaemia.

Statistical analysis
Continuous variables were expressed as the mean ± standard deviation in normally distributed data or as the median and interquartile range (25th-75th percentiles) in data with non-normal distributions, respectively. Categorical variables were reported as number (%). Continuous variables were compared using the Student's t-test or Mann-Whitney U-test according to the normality of distribution. Categorical variables were compared using the χ 2 test or χ 2 continuity correction test. Variables with P-values of ⩽0.10 including age, sex, underlying diseases and predisposing factors were entered into the multivariable logistic regression model to identify factors associated with NAC vs. C. albicans candidaemia. All variables with P-values of ⩽0.10 in the univariate analysis of factors associated with mortality were included in the multivariate logistic regression model to identify independent factors associated with mortality. Two-sided P-values of <0.05 were considered statistically significant. Statistical analyses were carried out with SPSS (IBM SPSS Statistics 21.0).

Results
During the study period, 188 episodes of candidaemia in 187 adult patients were initially identified. Two patients were excluded from further analysis because they had a single positive blood culture for Candida spp. without signs and symptoms of infection. One blood culture was considered to be contaminated due to simultaneous isolation of Staphylococcus epidermidis with C. glabrata. Five patients were lost to follow-up within 30 days from the onset of candidaemia and were therefore excluded. Thus, the remaining 180 episodes of candidaemia from 179 adult patients were included in the study. Over the study period, 74 patients developed candidaemia in the first 3-year period (2012-2014) (74/ 2845, 26.0/1000), which did not evidently differ from cases per hospitalisation-days in the second 3-year period (2015-2017) (106/3769, 28.1/1000) (P = 0.611).
Multivariate analysis (Table 3) confirmed that TPN was significantly associated with an increased risk of NAC candidaemia (odds ratio (OR) 2.535; 95% confidence interval (CI) 1.066-6.026; P = 0.035), whereas the presence of a urinary catheter was distinctly associated with an increased risk of C. albicans infection (OR 2.295; CI 1.129-4.666; P = 0.022). Current and former smokers were more associated with NAC candidaemia (OR 1.824; CI 0.885-3.756; P = 0.103), but the trend did not reach statistical significance.

Discussion
The distribution of Candida spp. associated with candidaemia varies greatly in different regions of the world. In recent years, some studies [8][9][10][15][16][17] have observed a shift from the more antifungal susceptible type species C. albicans towards NAC spp., and this is corroborated by the findings of this 6-year study where NAC isolates accounted for the majority (64.4%) of all adult candidaemia episodes. Likewise, we found C. parapsilosis to be the most prevalent NAC species (38.3%) which accords with a report from Mexico of 37.9% in 398 patients of all ages during a 3-year surveillance program from five hospitals [23]. Moreover, there are several other reports of C. parapsilosis as the most frequent NAC species (20-25%) from blood cultures in Spain [5,19], Italy [15], Latin America [24] and Shanghai, China [9,25]. The second most common NAC spp. in the current study was C. glabrata (13.9%), which has also been commonly recorded in studies from the USA [10], the UK [13], Greece [26], Beijing, China [17] and Taiwan [27]. In contrast, C. tropicalis ranked as the third most common in our series (10%) but was the most frequent NAC isolate in other studies in China, notably Nanjing [28], Chongqing [8], Shanghai [29], Shandong [30] and Taiwan [31]. The reasons for the varied species distribution and frequency remain unclear but patient demographics and underlying medical conditions are most likely contributory factors [16,19]. The higher frequency of C. parapsilosis BSI recorded here in surgical patients clearly warrants further investigation.
Several possible risk factors have been previously identified for the development of NAC candidaemia and include prior antifungal exposure, malignancy, immunosuppressive therapy, abdominal surgery [32], presence of CVCs [26], artificial surgical implants [17], head trauma and bacterial sepsis [29]. Our finding of an association of an indwelling urinary catheter and a higher risk of BSI caused by C. albicans vs. NAC is consistent with other studies [33]. Multivariate analysis also confirmed that TPN was associated with an increased risk of NAC spp., compared with C. albicans spp., but this is in contrast with Chow et al. [34] who linked TPN with a decreased risk of BSI caused by NAC spp., vs. C. albicans. In the current study, C. parapsilosis accounted for 59.5%, of all NAC isolates whereas in Chow's series [34], C. glabrata  (49%) was more than twice as frequent as C. tropicalis (19%) or C. parapsilosis (18%), which may explain the different findings. Frequent usage of TPN has been linked with C. parapsilosis candidaemia compared with C. albicans, as well as a negative association with C. tropicalis [35]. Therefore, we speculate that the possible reason for divergent views may be the variable proportion of Candida spp. in the local epidemiological setting. Smoking increases susceptibility to a wide range of bacterial and viral infections [36] as well as oral candidiasis in HIV infected adults [37]. Based on pooled data from contemporary cohort studies in the USA, the risk of death from infections in current smokers is more than twice that of individuals who have never smoked [38]. However, data on invasive candidiasis and candidaemia in adult smokers remain scarce. Here, we explored the role of smoking in NAC and C. albicans BSI. Although, by multivariate analysis, current and former smoking was not independently associated with the occurrence of NAC BSI, it was likely to be a predictive factor for NAC candidaemia as according to our data current and former smokers had an increased risk for NAC BSI, but no significant relationship was found between smoking and death due to candidaemia. More epidemiological, clinical and mechanistic approaches are needed to study further the impact of smoking in such patients.  C. albicans blood isolates showed significantly greater susceptibility to fluconazole (89.1%) than NAC isolates (62.1%) (P < 0.001) by applying the species-specific new CBPs. In contrast to the overall high-fluconazole susceptibility rate of C. parapsilosis (88.4%) and C. tropicalis (61.1%) recorded here, some studies from other cities in China have reported relatively reduced fluconazole susceptibility in their corresponding Candida spp. For example, a survey from Nanjing reported that only 74.2% of C. albicans, 57.7% of C. parapsilosis, 9.1% of C. glabrata and 31.6% of C. tropicalis were susceptible to fluconazole [28]. Based on the CBPs, all 25 isolates of C. glabrata in this study exhibited SDD to fluconazole, which supports the recommendations from the updated guideline for the management of candidiasis that transition to higher-dose fluconazole should be considered for patients with fluconazole-susceptible C. glabrata isolates [14]. We identified only two patients with C. krusei, one of which was resistant to voriconazole; each was effectively treated with micafungin and caspofungin, respectively.
Treatment of candidaemia is increasingly problematic owing to accumulated resistance of Candida isolates to antifungal agents, especially fluconazole worldwide [24,39]. Considering safety and efficacy aspects of treatment, echinocandins are now recommended especially in the early treatment of candidaemia, by both USA and European guidelines [14,39]. In the current study, fluconazole was the most frequently used antifungal agent for primary therapy, followed by echinocandins and voriconazole, although amphotericin B exhibited excellent in vitro activity overall against Candida spp. The plausible explanations for the prevalent use of fluconazole are the species distribution pattern of isolates, their high rates of susceptibility of C. albicans and C. parapsilosis to the agents, and its safety and tolerability compared with amphotericin B, as well as a significantly lower cost. Due to the lack of echinocandin susceptibility testing in the present study, these agents were used as initial empirical therapy for critically ill patients or those considered likely to have a fluconazole-resistant Candida spp. based on the recommendations of the USA and European guidelines. [14,39] Several studies have reported discrepant results between outcomes of candidaemia and C. albicans spp. in comparison with NAC spp. Differences in mortality were not statistically significant in patient cohorts in the USA (58% vs. 57%) [34], Beijing (34.7% vs. 38.6%) [17] and Shanghai (37.3% vs. 27.9%) [29]. Here, we found that C. albicans BSI was an independent risk factor associated with the mortality rate compared with NAC BSI (OR 3.629; CI 1.352-9.743; P = 0.011), which mirrors data from Shandong and Nanjing in China [8,30]. By contrast, in Greece, Dimopoulos et al., [26] reported NAC candidaemia to be associated with higher mortality than C. albicans in nonimmunosuppressed, non-neutropenic patients after ICU admission (OR 6.7; 95% CI 1.2-37.7; P = 0.03), which could be attributed to inappropriate or delayed therapy owing to the slower growth of NAC isolates on primary culture; patients with NAC candidaemia may also have been more critically ill than those with C. albicans.
Mortality rates of candidaemia have been attributed to the relative virulence of different Candida spp., a failure of host-defence mechanisms, patient's underlying diseases and complications, inappropriate or delayed treatment and other factors [5,20,40]. In animal models, C. parapsilosis and C. krusei exhibit less virulence than C. albicans, C. tropicalis and C. glabrata [40]. A meta review of dozens of studies concluded in terms of both overall and attributable mortality that C. tropicalis and C. glabrata are associated with the highest mortality (40 ± 70%), followed by C. krusei and C. albicans (20 ± 40%), the lowest being C. parapsilosis [40]. Accordingly, the predominance of C. parapsilosis in our study may have partly contributed to the observed lower mortality in patients with NAC.
Apart from C. albicans, the other variables strongly associated with mortality were ICU stay, renal failure and thrombocytopaenia, which suggest that the more severely ill patients had a worse prognosis. Additionally, almost all clearance of candidaemia in survivors who did not receive antifungal treatment occurred in surgical patients following the removal of an indwelling CVC or drainage catheter. Our results indicated that recent cancer surgery was associated with a higher probability of survival but an explanation for this remains unclear. Cancer patients who were judged reasonably able to tolerate surgery, often lacked distant metastases and had less disease severity and fewer complications. Such host factors may therefore be contributory to a reduced mortality rate for recent cancer surgery, and may aid clinicians better to judge the prognosis of these patients with candidaemia.
Our study has several main limitations. Firstly, owing to its retrospective cohort design, factors such as the management of CVC, the use of appropriate antifungal treatments, immunosuppressive, cancer and glucocorticoid therapy, and underlying chronic lung and heart disease, were not evaluated owing to incomplete medical data acquisition. Secondly, in vitro echinocandin susceptibility testing of Candida isolates was not performed, and thirdly, the relatively small sample size may impact on the confidence intervals and analysis of risk factors. Lastly, the study was limited to a single centre and thus, the results may not be applicable to other settings.
In summary, over the last 6 years, NAC was predominant among Candida isolates from adult candidaemia at Shengjing Hospital in Shenyang. TPN was associated with an increased risk of developing NAC candidaemia compared with C. albicans and patients with a urinary catheter were clearly at an increased risk of BSI due to C. albicans. Four independent risk factors for candidaemia-related death were identified, namely, ICU stay, thrombocytopaenia, the isolation of C. albicans as a significant predictor of survival and recent cancer surgery. Epidemiology and Infection 7