Epidemiology and prognostic factors of nosocomial candidemia in Northeast Brazil: A six-year retrospective study

Candidemia has been considered a persistent public health problem with great impact on hospital costs and high mortality. We aimed to evaluate the epidemiology and prognostic factors of candidemia in a tertiary hospital in Northeast Brazil from January 2011 to December 2016. Demographic and clinical data of patients were retrospectively obtained from medical records and antifungal susceptibility profiling was performed using the broth microdilution method. A total of 68 episodes of candidemia were evaluated. We found an average incidence of 2.23 episodes /1000 admissions and a 30-day mortality rate of 55.9%. The most prevalent species were Candida albicans (35.3%), Candida tropicalis (27.4%), Candida parapsilosis (21.6%) and Candida glabrata (11.8%). Higher mortality rates were observed in cases of candidemia due to C. albicans (61.1%) and C. glabrata (100%), especially when compared to C. parapsilosis (27.3%). Univariate analysis revealed some variables which significantly increased the probability of death: older age (P = 0.022; odds ratio [OR] = 1.041), severe sepsis (P < 0.001; OR = 8.571), septic shock (P = 0.035; OR = 3.792), hypotension (P = 0.003; OR = 9.120), neutrophilia (P = 0.046; OR = 3.080), thrombocytopenia (P = 0.002; OR = 6.800), mechanical ventilation (P = 0.009; OR = 8.167) and greater number of surgeries (P = 0.037; OR = 1.920). Multivariate analysis showed that older age (P = 0.040; OR = 1.055), severe sepsis (P = 0.009; OR = 9.872) and hypotension (P = 0.031; OR = 21.042) were independently associated with worse prognosis. There was no resistance to amphotericin B, micafungin or itraconazole and a low rate of resistance to fluconazole (5.1%). However, 20.5% of the Candida isolates were susceptible dose-dependent (SDD) to fluconazole and 7.7% to itraconazole. In conclusion, our results could assist in the adoption of strategies to stratify patients at higher risk for developing candidemia and worse prognosis, in addition to improve antifungal management.

Introduction clinical management and outcome (survival or death). Vital signs were classified according to the parameters established in the literature [15,16] together with the medical interpretation in the patients' records. Classification of blood cell counts were based on the reference ranges defined locally by the hospital laboratory. Sepsis, severe sepsis and septic shock were defined according to Angus and van der Poll [17]. Crude mortality rate was calculated at 7 and 30 days from candidemia onset. The following antifungal dosages were considered adequate: fluconazole (FLU) 400 mg/day, amphotericin B deoxycholate (AMB) 0.5-1.0 mg/kg/day, amphotericin B lipid complex (ABLC) 3.0-5.0 mg/kg/day, caspofungin (CPF) 50 mg/day, micafungin (MCF) 100 mg/day, anidulafungin (ADF) 100 mg/day [18].

Ethics
This study was approved by the Local Research Ethics Committee ("Comitê de Ética em Pesquisa da Liga Norte Riograndense Contra o Câncer") under the protocol number 042/042/ 2012. Written patient consent was not required because of the observational nature of the study.

Laboratory procedures
Blood samples were processed using the Bact/Alert system (BioMérieux, France). All positive cultures were inoculated onto the surface of Sheep Blood Agar and incubated at 30˚C for 48-96 h. Yeast growth was confirmed by Gram staining and the initial identification was performed at the referred hospital with the Vitek 2 Compact YST system (BioMérieux, France), according with manufacturer's instructions. The strains were sent to the Laboratory of Medical and Molecular Mycology, Department of Clinical and Toxicological Analysis, Federal University of Rio Grande do Norte and a confirmation in identification was performed according to classical methods [19]. Of note, accurate identification was also performed using MALDI-TOF [20] when necessary. Unfortunately, some strains were not identified at the species level due to limitations of the initial screening performed at the hospital microbiology laboratory and lack of viability/or availability of some strains for further analysis. Antifungal susceptibility to amphotericin B (Sigma Chemical Corporation, St. Louis, MO, USA), fluconazole (Pfizer Incorporated, New York, NY, USA), itraconazole (Pfizer Incorporated, New York, NY, USA) and micafungin (Merck, Rahway, NJ, USA) was performed using the broth microdilution method according to the Clinical and Laboratory Standards Institute (CLSI) document M27-A3 [21]. The reference strains C. parapsilosis ATCC 22019 and C. krusei ATCC 6258 were used as quality controls. Minimum inhibitory concentration (MIC) values were interpreted according to the current clinical breakpoints suggested by CLSI for the most common species of Candida [22,23].

Statistical analysis
Continuous variables were expressed as mean ± standard deviation (SD) and compared using Student t test or Mann-Whitney test. Categorical variables were expressed as frequencies and percentages and compared using Chi-square (X 2 ) test or Fisher's exact test, as appropriate. Logistic regression analysis was performed with variables that presented P � 0.1 in the comparisons of groups to identify possible risk factors associated with mortality at 30 days after candidemia. Variables of clinical relevance and with sample size �60 found to be significant in the univariate analysis were included in a multivariate logistic model. All tests were 2-tailed, and a P-value <0.05 was determined to represent statistical significance. Statistical analyses were performed using the Statistical Package for the Social Sciences (SPSS) software, version 20 (IBM SPSS, Chicago, IL, USA).

Results
A total of 87 patients out of 37,768 admitted to the study hospital between 2011 and 2016 had at least one episode of candidemia. However, 19 individuals were excluded (16 patients who had candidemia before 48 h of hospitalization and 3 patients with no medical records). The mean incidence of candidemia cases was 2.23/1000 admissions, ranging from 1.03 to 3.02 throughout each different year of study, with a trend to increase from 2014-2016 (Fig 1).
Tables 1 to 4 present the main demographic and clinical characteristics of all the patients included in the present study, classified according to the outcome of candidemia after 30 days (survival or death).  The main predisposing factors found were the previous use of antibacterial agents (66/68; 97.1%), the presence of CVC (54/68; 79.4%), corticosteroid therapy (38/68; 55.9%) and surgery (38/68; 55.9%; Table 3).
Of the 68 patients included in the study, 19 came from another hospital at admission (27.9%), 41 were female (60.3%) with a mean age of 56.0 ± 15.5 years (Table 1), mean hospital length of stay (LOS) of 63.9 ± 50.5 days and mean time between admission and development of candidemia of 35.6 ± 32.2 days.
The predominant age class ranged from 61 to 70 years (18/68; 26.5%; Fig 2B). It is worth mentioning that 45.6% (31/68) of the patients were elderly (aged between 61 and 90 years); while only one 12-year-old child was enrolled in the study since our hospital did not have a pediatric ward ( Fig 2B).
The most prevalent underlying conditions were cardiovascular disease (49/68; 72.1%), diabetes mellitus ( The use of antifungal agents before the onset of candidemia was observed in 16.2% (11/68) of the patients (Table 4), of which 63.6% (7/11) used fluconazole and 63.6% (7/11) used antifungal drugs for very short periods (1 to 5 days). The previous exposure to antifungals did not influence the isolation of NCAC species (P = 0.451).
The relationship between these variables and the outcome was confirmed in the univariate logistic regression, except for sepsis and Candida species isolated (Table 5). Some characteristics of the patients significantly increased the probability of death: older patients (P = 0.022; OR = 1.041), severe sepsis (P < 0.001; OR = 8.571), septic shock (P = 0.035; OR = 3.792), hypotension vs. hypertension (P = 0.003; OR = 9.120), neutrophilia (P = 0.046; OR = 3.080), thrombocytopenia (P = 0.002; OR = 6.800), MV on candidemia onset (P = 0.009; OR = 8.167) and greater number of surgeries (P = 0.037; OR = 1.920; Table 5). Multivariate analysis included age, severe sepsis, septic shock, use of MV and blood pressure on candidemia onset (Table 6). Age (P = 0.040; OR = 1.055), severe sepsis (P = 0.009;  OR = 9.872) and hypotension vs. hypertension (P = 0.031; OR = 21.042) were independently associated with higher probability of death (Table 6). It is worth mentioning that the probability of death increased about 10-fold in patients who had severe sepsis and 21-fold in patients with hypotension compared to those who had hypertension at the onset of candidemia. Table 7 shows the results of the in vitro activity of 4 systemically active antifungal agents against BSI isolates of Candida spp. All isolates tested were susceptible to amphotericin B and micafungin, while a few of them were resistant (2/39; 5.1%) and susceptible dose-dependent (SDD; 8/39; 20.5%) to fluconazole and SDD to itraconazole (3/39; 7.7%). There were two strains (an isolate of C. albicans and another of C. tropicalis) SDD to both fluconazole and itraconazole (2/39; 5.1%).
The distribution of Candida species observed in our study is consistent with other studies conducted in Brazil and Latin America, showing a relatively lower prevalence of C. albicans (although it is still the most prevalent species) and a higher prevalence of C. parapsilosis and C. tropicalis among the NCAC species (alternating between second and third places), and C. glabrata as the fourth most prevalent species [24,26,38,39]; whereas in the US and several other European countries C. glabrata appears generally as the second most prevalent species [8].
C. albicans and C. glabrata were the species most associated with mortality, especially when compared to C. parapsilosis. Other studies have also found a correlation between C. albicans and C. glabrata with higher mortality, as well as lower mortality rates in cases of candidemia due to C. parapsilosis [11,24,31,32,35].
Another important finding of our study was the high frequency of fluconazole use, being the first choice in most cases. A recent guideline for the management of candidiasis recommended an echinocandin as initial therapy for candidemia [9], however this class of antifungal drugs is not yet very accessible due to its high cost [40]. Amphotericin B deoxycholate was the second most commonly used antifungal drug, however its lipid formulations are preferable because of its high toxicity [41], except in some specific cases [9].
Comparing our results using univariate and multivariate logistic regression analysis with the existing literature, we found other studies that have demonstrated an association between age, clinical condition (sepsis, septic shock, APACHE score) and mechanical ventilation with a higher mortality risk in patients with candidemia [30,32,35,42,43]. It is important to mention that hypotension is one of the criteria for the definition of septic shock [17] and it is also evaluated in the APACHE score, therefore its association with worse prognosis found in our study was expected; although we highlighted that its association as an independent risk factor had not yet been described.
Despite the low rate of antifungal resistance found in our study, there was a higher proportion of strains susceptible dose-dependent to fluconazole (20.5%), mainly among C. glabrata isolates (80%), consistent with the widely known lower susceptibility of C. glabrata to fluconazole [11]. These results indicate a greater probability of therapeutic failure if fluconazole is used, especially in cases of C. glabrata BSI.  Epidemiology and prognostic factors of nosocomial candidemia Finally, our antifungal susceptibility profile corroborates with other studies conducted in Brazil and Latin America in general, where Candida spp. resistance to echinocandins and amphotericin B remains rare [39,44].
In conclusion, we observed a high incidence of candidemia, displaying a tendency to increase over the 6 years of the study, as well as a high mortality rate, proving a nosocomial problem that deserves attention. We believe that our study contributed to the knowledge of the local epidemiology of candidemia and could be used to assist in the adoption of strategies to stratify patients at higher risk for developing candidemia and worse prognosis in low income regions of the globe, in addition to improve antifungal management (prophylaxis, empirical and definitive therapy) which has not been shown to be effective in the study hospital. We emphasize that this is the first study in Northeast Brazil that has made such a deep analysis in this regard, despite our limitations, mainly due to the nature of the study (retrospective and single center).