Epidemiology, Clinical Characteristics, Risk Factors, and Outcomes of Candidemia in a Large Tertiary Teaching Hospital in Western China: A Retrospective 5-Year Study from 2016 to 2020

The aim of this study was to investigate the current status of candidemia and evaluate the clinical characteristics, risk factors and outcomes among different species. We conducted a retrospective study by univariate and multivariate analysis between Candida albicans and non-albicans Candida (NAC) species in a Chinese national medical center from 2016 to 2020. Among the 259 episodes, C. albicans (38.6%) was the leading species, followed by C. tropicalis (24.3%), C. parapsilosis (20.5%), and C. glabrata (12.4%). Most C. albicans and C. parapsilosis were susceptible to nine tested antifungal agents, whereas C. tropicalis showed 30.2~65.9% resistance/non-wild-type to four azoles with great cross-resistance, indicating that fluconazole should not be used for empirical antifungal treatment. In multivariable models, the factor related to an increased risk of NAC was glucocorticoid exposure, whereas gastrointestinal hemorrhage and thoracoabdominal drainage catheters were associated with an increased risk in C. albicans. Subgroup analysis revealed leukemia and lymphoma, as well as glucocorticoid exposure, to be factors independently associated with C. tropicalis in comparison with C. albicans candidemia. No significant differences in 7-day mortality or 30-day mortality were observed between C. albicans and NAC. This study may provide useful information with respect to choosing empirical antifungal agents and exploring differences in molecular mechanisms.


Introduction
Candidemia is one of the leading causes of nosocomial bloodstream infections (BSI) and is a life-threatening invasive fungal infection associated with significant morbidity, mortality, high hospital costs, and successful clinical outcome that requires timely diagnosis and effective antifungal therapy [1,2]. Whereas the species distribution and susceptibility patterns of candidemia could vary considerably depending on geographic region and change over time, the global shift in favor of non-albicans Candida (NAC) species is of worldwide, as is the emerging and growing resistance to antifungal agents among these species [3]. Generally, C. albicans remains a leading causative agent of candidemia, but common NAC presents geographical variations, such as more familiar C. tropicalis in Asia and Latin America; more frequent C. glabrata in the USA and north/central Europe; and more prevalent C. parapsilosis in South America, southern Europe, and several parts of Asia. The top five Candida spp. account for approximately 90% of invasive candidiasis [1,3]. Given the varying patterns of susceptibility to azoles and echinocandins, changes in species distribution may drive the transformation of therapeutic and prophylactic strategies [4]. Intrinsic and acquired resistance to azoles in certain Candida spp. has posed considerable clinical challenges worldwide [5]. Echinocandins and fluconazole have been recommended as optional initial therapies for Candida infections according to clinical guidelines published by ESCMID [6] and IDSA [4].
Each Candida spp. Presents its own unique characteristics, including tissue tropism, invasive potential, virulence, and antifungal susceptibility [1,7]. To achieve improved benefits, early recognition and timely empirical antifungal treatment in populations at high risk of invasive fungal infections are essential, which require prediction of the drug resistance tendency of pathogens by taking into consideration local species distribution, antifungal susceptibility patterns, and species-related clinical features in individuals [1,4,6]. In this study, we reviewed the most recent microbial epidemiology of Candida antifungal resistance patterns by evaluating the existing circumstances of candidemia at West China Hospital of Sichuan University, a National Centre for Diagnosis and Treatment of Difficult and Critical Diseases, where approximately 4.85 million emergency and outpatient visits, as well as and 238 000 discharged patients were recorded in 2020. Furthermore, we compared C. albicans with NAC in terms of underlying conditions, possible risk factors, and clinical outcomes. We also performed a subgroup analysis of C. albicans vs. C. tropicalis to explore the key species-related variables in order to inform clinical decision making, providing interesting background on species-specific differences in terms of cellular and molecular factors.

Results
Between 2016 and 2020, a total of 667 positive blood cultures of Candida occurred in 316 patients at our hospital, as shown in Figure 1, and 57 patients were excluded. No significant variations in the proportion of Candida distribution were found, regardless of whether the mentioned patients were excluded. For the retrospective comparative study, we compared 100 patients with C. albicans candidemia with 159 patients with NAC candidemia and analyzed the differences in subgroups between C. albicans and C. tropicalis.
Antibiotics 2022, 11, x FOR PEER REVIEW 2 of 15 [1,3]. Given the varying patterns of susceptibility to azoles and echinocandins, changes in species distribution may drive the transformation of therapeutic and prophylactic strategies [4]. Intrinsic and acquired resistance to azoles in certain Candida spp. has posed considerable clinical challenges worldwide [5]. Echinocandins and fluconazole have been recommended as optional initial therapies for Candida infections according to clinical guidelines published by ESCMID [6] and IDSA [4]. Each Candida spp. Presents its own unique characteristics, including tissue tropism, invasive potential, virulence, and antifungal susceptibility [1,7]. To achieve improved benefits, early recognition and timely empirical antifungal treatment in populations at high risk of invasive fungal infections are essential, which require prediction of the drug resistance tendency of pathogens by taking into consideration local species distribution, antifungal susceptibility patterns, and species-related clinical features in individuals [1,4,6]. In this study, we reviewed the most recent microbial epidemiology of Candida antifungal resistance patterns by evaluating the existing circumstances of candidemia at West China Hospital of Sichuan University, a National Centre for Diagnosis and Treatment of Difficult and Critical Diseases, where approximately 4.85 million emergency and outpatient visits, as well as and 238 000 discharged patients were recorded in 2020. Furthermore, we compared C. albicans with NAC in terms of underlying conditions, possible risk factors, and clinical outcomes. We also performed a subgroup analysis of C. albicans vs. C. tropicalis to explore the key species-related variables in order to inform clinical decision making, providing interesting background on species-specific differences in terms of cellular and molecular factors.

Results
Between 2016 and 2020, a total of 667 positive blood cultures of Candida occurred in 316 patients at our hospital, as shown in Figure 1, and 57 patients were excluded. No significant variations in the proportion of Candida distribution were found, regardless of whether the mentioned patients were excluded. For the retrospective comparative study, we compared 100 patients with C. albicans candidemia with 159 patients with NAC candidemia and analyzed the differences in subgroups between C. albicans and C. tropicalis.

Antifungal Susceptibility Testing
The antifungal susceptibility testing (AFST) results of all isolates are presented in Table 1. Amphotericin B, flucytosine, and echinocandins demonstrated significant activity in vitro against most Candida spp. More than 96% of Candida were susceptible to three echinocandins, with the highest MIC 50 of any species being ≤ 1 µg/mL and the highest MIC 90 of any species being ≤2 µg/mL. Of the 259 Candida isolates, 175 (67.6%), 34 (13.1%), and 50 (19.3%) isolates were susceptible, susceptible dose-dependent (SDD), and resistant to fluconazole, respectively, and the main susceptible strains included C. albicans (92/100, 92.0%) and C. parapsilosis (45/53, 84.9%). The resistance rate of four azoles was the highest in C. tropicalis, with MIC 90 values of 128, 2, 8, and 1 µg/mL to fluconazole, itraconazole, voriconazole, and posaconazole, respectively. Severe azole cross resistance was observed among C. tropicalis; most fluconazole-resistant C. tropicalis were non-susceptible (29/30, 96.7%) to voriconazole, and most voriconazole-resistant C. tropicalis were resistant (25/26, 96.2%) to fluconazole. Nonetheless, more than 91% of C. albicans were susceptible to four azoles, with an MIC 90 ≤ 1 µg/mL.  Table 2 shows detailed demographic characteristics, underlying conditions and comorbidities, and clinical outcome variables. The median age was 53 (IQR: 43-66); elderly patients (age ≥ 65) accounted for 28.2% of the sample, and 31.7% of patients were female. Patients with C. tropicalis and other species of candidemia had a lower median age than those with C. albicans and C. parapsilosis. The majority of patients with candidemia were from intensive care units (ICUs) (45.6%), followed by medical wards (24.3%), surgical wards (13.1%), emergency departments (10.8%), and hematology wards (6.2%). The most common complications were gastrointestinal diseases (51.7%), lung diseases (51.4%), septic shock (32.8%), kidney diseases (28.2%), brain diseases (23.9%), liver diseases (22.8%), and solid tumors (19.3%). According to routine blood examinations (Table 3), almost all patients (93.1%) had different degrees of anemia, with insignificant differences among these species, whereas there were significant differences in platelets, white blood cells, neutrophil, and lymphocyte counts. Among patients with candidemia, the average total length of hospitalization was 32 days (IQR, 18-56) ( Table 2). Patients with C. tropicalis or other candidemia species had a longer total hospitalization than those with C. albicans or C. parapsilosis candidemia, and a shorter ICU stay was found in C. tropicalis or C. parapsilosis candidemia than C. albicans or the other candidemia; however, these differences were not statistically significant. Moreover, no difference was found in 7-day mortality, 30-day mortality, or in-hospital mortality between the Candida spp. (Table 2), which is in line with the result of the survival curve ( Figure 4).

C. albicans vs. Non-albicans Candida
Tables 4 and 5 illustrate the details of potential risk factors associated with candidemia due to C. albicans and NAC, including comorbidities, common invasive procedures, and previous drug exposure, determined by univariable analysis and multivariate logistic regression, respectively. Compared with C. albicans candidemia, the prevalence of hematological disorders was markedly more frequent in patients with C. tropicalis candidemia, whereas the proportions of invasive procedures with significant differences, such as surgery, invasive mechanical ventilation, urinary catheters, and indwelling thoracoabdominal drainage catheters, were slightly lower in the NAC group. After the multivariate analysis presented in Table 5, several potential independent risk factors for candidemia with different Candida species were identified: glucocorticoids (OR 3.076, 95% CI 1.543-6.131, p = 0.001) were associated with NAC, whereas gastrointestinal hemorrhage (OR 0.397, 95% CI 0.209-0.755, p = 0.005) and thoracoabdominal drainage catheters (OR 0.507, 95% CI 0.289-0.891, p = 0.018) were closely related to C. albicans. Results of subgroup analysis of candidemia due to C. tropicalis or C. albicans to identify the factors associated with those two infections are presented in Tables 6 and 7, respectively. Multivariate logistic regression analysis (Table 7) indicated leukemia and lymphoma (OR 10.08, 95%CI 1.127-90.133, p = 0.039), as well as glucocorticoids (OR 2.788, 95% CI 1.147-6.773, p = 0.024), as factors independently associated with C. tropicalis bloodstream infection, whereas thoracoabdominal drainage catheters (OR 0.277, 95% CI 0.131-0.588, p = 0.001) were separately connected with C. albicans candidemia according to previous logistic regression results between C. albicans vs. NAC (Table 5).

Discussion
In recent decades, an epidemiological trend shift from the dominant pathogen C. albicans to increasing incidence of NAC has been observed worldwide, although there is substantial geographic, center-to-center, and unit-to-unit variability in the relative prevalence of Candida spp. [1,3,8]. Likewise, we discovered C. albicans to be the most prevalent species, which is consistent with results from nationwide active laboratory-based surveillance in China [9]. The most common NAC identified in the current study was C. tropicalis (24.3%), which has been reported in several areas of similar latitudes [10] in contrast to reports from northern China [11], western Europe [12], and North America [13]. The distribution and frequency of Candida spp. Were influenced by not only geographic area but also the patient's underlying conditions, the antifungal drugs patients had received, local hospital-related factors, and even the local climate [14,15]. The rates of these Candida spp. were similar in most diseases, but a significantly higher rate of C. tropicalis was observed in patients with hematologic malignancies who had undergone common cancer chemotherapy leading to neutropenia, which is in agreement with results of a prior study [16]. Additionally, antifungal exposure before the onset of candidemia might be partly accountable for the migration of Candida species to NAC [16], whereas a study in Thailand demonstrated that most patients with fluconazole-resistant C. tropicalis candidemia did not have a recent azole exposure and that C. tropicalis may represent exogenous isolates acquired from the environment [17].
In general, the tested antifungal agents, in addition to azoles, appeared to have a high susceptibility rate to common Candida according to CHIF-NET surveillance, which revealed the rapid emergence of azole-resistant C. tropicalis strains in China [18]. In our hospital, C. albicans and C. parapsilosis isolates from blood displayed significantly higher susceptibility or wild-type MIC to azoles than C. tropicalis and C. glabrata according to species-specific clinical breakpoints (CBPs) [19] and epidemiological cutoff values (ECVs) [20,21]. In contrast to the high susceptibility of C. tropicalis to amphotericin B, 5-flucytosine, and echinocandins, our study revealed relatively high azole resistance of C. tropicalis to four azoles, among which 47.6%, 41.3%, 30.2%, and 65.9% of isolates were resistant or had NWT MICs to fluconazole, voriconazole, itraconazole, and posaconazole, respectively. A higher MIC 90 (128 µg/mL to fluconazole, 8 µg/mL to voriconazole) of azoles against C. tropicalis was found in our survey than in a previous investigation [18]. These important findings indicate that fluconazole and voriconazole should not be used as empirical antifungal drugs for treatment of C. tropicalis, which is in agreement with other studies [16,17]. Based on clinical practice guidelines [4,6] and our in vitro AFST data, echinocandins should be considered as the first choice for initial treatment of most episodes of candidemia and invasive candidiasis, except for central nervous system, eye, and urinary tract infections due to Candida. In addition, we observed an obvious cross resistance of C. tropicalis to fluconazole and voriconazole that may be related to different azole target Erg11p modifications or increased efflux pump activity [22]. On a larger scale, further study of the molecular mechanism of antifungal resistance, continuous antifungal resistance surveillance, development of non-cultured rapid diagnostic methods, and antifungal stewardship will be necessary to improve antifungal drug-resistant predicaments [23].
The most common individual predisposing factors for invasive candidiasis include those intrinsic to the host or the disease state, such as Candida colonization, old age, diabetes mellitus, gastrointestinal perforation, pancreatitis, sepsis, hematologic malignancy, neutropenia, transplantation, and severe immunodeficiency, as well as factors resulting from iatrogenic interventions, such as long-term and/or repeated use of broad-spectrum antibiotics, recent major surgery (particularly abdominal surgery), dialysis, parenteral nutrition (PN), use of corticosteroids, use of immunosuppressants, presence of indwelling central venous catheters, and long-term ICU stays [1,4,24]. In the current study, most patients had been previously exposed to antibiotics (>95%) and PN (>70%) without significant interspecies differences, which indicates that these may be common contributing factors for candidemia. Broad-spectrum antibiotics could confer Candida spp. a selective advantage over bacteria, causing Candida spp. overgrowth and increased gut colonization [1]. Parenteral lipid emulsion could increase Candida biofilm formation, which may explain the increased risk of candidemia in patients receiving parenteral nutrition via medical catheters [25]. Each Candida spp. presents its own unique characteristics, including tissue tropism, invasive potential, virulence, biofilm formation ability, and antifungal susceptibility [1,4,25].
Several studies have compared the characteristics of C. albicans with those of NAC candidemia, revealing differences in risk factors and outcomes [11,16,17,[26][27][28][29]. Multivariate analysis also confirmed that glucocorticoids exposure is associated with an increased risk of NAC candidemia, which is consistent with results of other studies [29]. We also found an association of gastrointestinal hemorrhage and indwelling thoracoabdominal drainage catheters with a higher risk of C. albicans candidemia, which is in agreement with results reported by Gong et al. [30], which may be attributed to C. albicans being the most frequent Candida species in the human gut mycobiome [31]. Some studies led to different, conflicting conclusions; for example, one risk factor, PN, was linked to a decreased risk of NAC candidemia by Chow et al. [28], but Zhang et al. suggested that PN was associated with an increased risk of NAC [11], and there was no significant difference among species in this cohort. The possible reason for such paradoxical conclusions may be associated with variability of Candida species distribution, as well as patients' baseline comorbidities in the local epidemiological setting.
Risk factors associated with candidemia caused by C. tropicalis or C. albicans were also compared in a subgroup analysis. The results of that analysis revealed that leukemia and lymphoma, as well as glucocorticoid exposure, are independent risk factors for C. tropicalis candidemia. A previous study reported the independent risk factors for C. tropicalis bloodstream infections as neutropenia, chronic liver disease, and male sex [17]. Our study confirmed that C. tropicalis seems to be more frequent in patients with hematological diseases (neutropenia, leukemia, and lymphoma), which is in agreement with previous studies suggesting that C. tropicalis is the most common Candida species (75.4%), rather than C. albicans (12.3%), in hematological malignancy [32], which may be linked with cytotoxic chemotherapy-induced immunosuppression.
Several investigations have recorded discrepant results of clinical outcomes between C. albicans and NAC candidemia. Although some investigations have reported that mortality was higher in patients with NAC than in those with C. albicans candidemia [11,29], insignificant differences in 7-day mortality, 30-day mortality, and in-hospital mortality were discovered among different Candida species groups in the present study (Table 2 and Figure 4). On the one hand, early mortality is linked with prompt therapeutic measures, including appropriate antifungal agents and early removal of intravascular catheters, as recommended in guidelines [4,33]; on the other hand, late mortality is associated with host factors (e.g., patient comorbid status and signs of organ dysfunction) [33]. The few significant differences in mortality observed in our study could be partly due to the choice preference for echinocandin use in our setting, as well as the similar severity of underlying conditions among patients (Table 2).
Several limitations need to be noted in our study. First, the results of a retrospective study can be exploratory and should be interpreted with caution. Second, two AFST methods were applied at different times, so we lacked data on posaconazole and echinocandins for few months. Furthermore, the study was performed at a center lacking pediatrics, obstetrics, and gynecology disciplines; thus, the results may not be applicable to other settings.

Setting, Study Design, and Data Collection
This research was conducted as a retrospective epidemiological investigation and an attendant comparative study from 2016 to 2020 at the West China Hospital of Sichuan University, a tertiary grade A academic teaching hospital with 4300 beds in Chengdu. All positive Candida blood cultures were identified from the microbiological laboratory information system. The electronic medical records of all patients were retrospectively reviewed, and the following information was collected from the hospital information system: age, sex, diagnosis at admission and discharge, comorbidities, prior use of drugs, invasive procedures, laboratory products, and clinical outcomes. The Charlson Comorbidity Index (CCI) and age-adjusted Charlson Comorbidity Index (aCCI) were calculated for each case to assess the severity of illness.

Definitions
Candidemia was defined by at least one positive blood culture for Candida spp. in patients with compatible clinical signs and symptoms of infection. Only the first case of positive Candida blood cultures was included, and patients <16 years of age, mixed candidemia or outpatient, and emergency patients with incomplete clinical information were excluded. The onset of candidemia was defined as the date when the first positive blood culture specimen was collected. Prior invasive procedures and drug exposure are defined as occurrence of the relevant event within 30 days before the onset of candidemia.

Microbiological Analysis
Blood specimens were processed using a Bact/Alert 3D automated blood culture system (bioMérieux, Marcyl'Etoile, France). Colonies of Candida isolates were identified with either matrix-assisted laser desorption ionization time-of-flight mass spectrometry (Bruker Daltoniks, Bremen, Germany) or internal transcribed spacer sequencing. All analyzed isolates had AFST performed using ATB FUNGUS three strips (bioMérieux, Marcyl'Etoile, France) from January 2016 to May 2017 and were performed using Sensititre YeastOne TM commercialized products (Trek Diagnostic Systems Ltd., East Grimstead, UK) consisting of nine antifungal agents after April 2017. C. krusei ATCC 6258 and C. parapsilosis ATCC 22019 were routinely used as quality controls. AFST was interpreted using speciesspecific CBPs defined by the Clinical Laboratory Standards Institute in the M60 [19] or ECVs in the M59 [20] and previous studies [21] to distinguish wild-type (WT) or non-wild-type (NWT) strains where CBPs were not available. The minimum inhibitory concentration (MIC) distribution of each Candida spp. was summarized to obtain MIC 50 and MIC 90 .

Statistical Analysis
Quantitative variables reported as the median and interquartile range (IQR) were analyzed with a Mann-Whitney test or Kruskal-Wallis test where applicable. Categorical variables presented as absolute numbers and relative percentages were compared between groups with a chi-square test or Fisher's exact test as appropriate. Non-colinear covariates with a p value ≤ 0.05 in the univariate analysis after deliberation based on practical clinical significance were applied in stepwise logistic regression multivariate model to identify independent factors, and the results were presented as odds ratios (OR) with their 95% confidence intervals (95% CI) and p values. The 30-day survival curves of candidemia were delineated by Kaplan-Meier survival analysis, and the difference was evaluated by the log-rank test. Linear-by-linear association analyses were performed to evaluate a changing trend in species distribution patterns over the five-year study period. All significance tests were two-tailed. P values of 0.05 or less were considered boundaries for statistical significance. All data were analyzed by IBM SPSS 26.0 (IBM, Armonk, NY, USA).

Conclusions
In conclusion, we found that C. tropicalis was the most common NAC species, with a particularly alarming resistance to azoles among candidemia in western China over the past 5 years, so empirical treatment would not recommend using azoles. NAC candidemia was more frequent than C. albicans in patients who had been exposed to glucocorticoids in contrast to patients with gastrointestinal hemorrhage and indwelling thoracoabdominal drainage catheters. Continued and careful monitoring of the clinical and mycological characteristics of candidemia is required, as well as further evaluation in clinical practice and further in-depth research.

Informed Consent Statement:
The study was a retrospective medical chart review study that complied with the criteria for a waiver of informed consent.

Data Availability Statement:
The data and material information used and analyzed in the current study are available from the corresponding author upon reasonable request.