Candidemia in the Neonatal Intensive Care Unit: A Retrospective, Observational Survey and Analysis of Literature Data

We evaluated the epidemiology of Candida bloodstream infections in the neonatal intensive care unit (NICU) of an Italian university hospital during a 9-year period as a means of quantifying the burden of infection and identifying emerging trends. Clinical data were searched for in the microbiological laboratory database. For comparative purposes, we performed a review of NICU candidemia. Forty-one candidemia cases were reviewed (overall incidence, 3.0 per 100 admissions). Candida parapsilosis sensu stricto (58.5%) and C. albicans (34.1%) were the most common species recovered. A variable drift through years was observed; in 2015, 75% of the cases were caused by non-albicans species. The duration of NICU hospitalization of patients with non-albicans was significantly longer than in those with C. albicans (median days, 10 versus 12). Patients with non-albicans species were more likely to have parenteral nutrition than those with C. albicans (96.3% versus 71.4%). Candida albicans was the dominant species in Europe and America (median, 55% and 60%; resp.); non-albicans species predominate in Asia (75%). Significant geographic variation is evident among cases of candidemia in different parts of the world, recognizing the importance of epidemiological data to facilitate the treatment.


Introduction
Although blood stream infection (BSI) due to Candida species (spp.) in the neonatal intensive care unit (NICU) is less frequent than that due to Gram-positive or Gram-negative bacteria, it has higher morbidity and mortality rates. In particular, among newborns with a birth weight < 1000 g, 4-8% will develop candidemia, which has a 30% mortality in this group of patients [1]. Newborns who survive frequently have long-term neurological impairment, including cerebral palsy, blindness, hearing impairment, cognitive deficits, and periventricular leukomalacia [2]. Risk factors for neonatal candidemia include prematurity, use of central venous lines, endotracheal tubes, parenteral nutrition, broad-spectrum antibiotic administration (especially third-generation cephalosporins), prolonged hospitalization, abdominal surgery, exposure to H2 blockers, and Candida colonization. Although Candida albicans is the most prevalent yeast pathogen, BSIs caused by Candida non-albicans, particularly Candida parapsilosis complex and Candida glabrata complex, have increased in recent years [2,3].
This study aimed (i) to determine the epidemiology of Candida BSIs in the NICU of an Italian university hospital during 9 years of observation; (ii) to analyze the trend in species distribution; and (iii) to examine in vitro susceptibility to common antifungal drugs. Furthermore, for comparative purposes, a systematic review of studies concerning the distribution of Candida spp. causing candidemia in NICU patients is presented.

Study Design.
A retrospective, observational survey of all consecutive cases of candidemia was conducted at the NICU (capacity of 8 beds; level III) of a university hospital in Southern Italy, from January 1, 2007, to December 31,2015. The number of annual admissions ranged from 135 to 169, with no significant variation during the period of study. All of the neonates who had at least one positive blood culture for Candida spp. and signs or symptoms of infection were considered in this study. Only the first episode of candidemia was reported for patients with recurrent or subsequent episodes. Clinical data were searched for in the microbiological laboratory database and included sex, gestational age, birth weight, and predisposing risk factors for Candida BSI (i.e., intravascular devices, prolonged antibiotics, administration of total parental nutrition, and prolonged hospitalization).

Definitions. Extremely low birth weight (ELBW) infants
were defined as those with a birth weight ≤ 1000 g, very low birth weight (VLBW) infants were those with a birth weight <1500 g, and low birth weight infants were those with a birth weight < 2500 g. Prolonged antibiotic use was defined as >14 days of continuous administration. Late-onset sepsis (LOS) was defined as infection occurring for >48 h of life. Candidemia was considered as probably catheter-related when semiquantitative culture of the catheter tip yielded >15 colony-forming units of Candida.

Laboratory
Procedures. Blood cultures were performed using a lysis-centrifugation system (Isolator; DuPont Co., Wilmington, DE, USA). The samples were cultured on two plates of Sabouraud dextrose agar with 0.05% chloramphenicol (BioRad, Marnes-la-Coquette, France) and then incubated at 36 ∘ C (±1) and 28 ∘ C (±1). The samples were examined daily for 10 days. The isolates were identified using standard procedures (morphology on cornmeal agar plates, germ-tube production in serum, and ability to grow at 37 ∘ C and 42 ∘ C) and biochemical analysis using two methods, the Vitek 2 system and ID 32C panels (Bio-Merieux, Rome, Italy), to obtain accurate results. All strains were frozen at −70 ∘ C until further investigations [4]. Candida parapsilosis complex genotyping was performed by PCR amplification as reported previously [5,6].
The susceptibility values were interpreted taking into account the species-specific clinical breakpoints (CBPs) suggested by the Clinical Laboratory Standards Institute (CLSI) subcommittee for the most common species of Candida [7]. The epidemiological cut-off values were used to define wildtype and non-wild-type isolates if no CBPs were available from the CLSI [8,9]. Minimum inhibitory concentration (MIC) data are presented as MIC 50 (MIC causing inhibition of 50% of isolates) and MIC 90 (MIC causing inhibition of 90% of isolates).

Statistical Analysis.
The Shapiro-Wilk test was used to test the normal distribution of data. Non-normally distributed data are expressed as median and interquartile range (IQR) and were compared using the Mann-Whitney U test. Categorical data are expressed as number and percentage and were compared using 2 or Fisher's exact test. All values are two-tailed, and statistical significance was defined as < 0.05 (Social Sciences (SPSS) software 10 for Mac OS X; SPSS Inc., Chicago, IL, USA).

Literature Review.
A review of full-text articles that were published in English from January 2000 to February 2015 was performed. The MEDLINE database was used for the bibliographic research, using the following key words: "neonatal candidemia", "candidemia neonatal intensive care unit", "Candida neonatal intensive care unit, and "NICU candidemia". Additionally, the bibliographies of the selected articles were reviewed for relevant publications.
The exclusion criteria were as follows: articles that reported a period of study prior to 2000; letters, randomized, controlled trials; and studies that reported a total number of Candida BSIs less than five. The following data were collected from each selected study: geographic location, year of publication, study period, type of study, incidence, influencing factors candidemia, total number of isolated Candida spp., and relative proportion of each of the Candida spp.

Analysis of Cases in the NICU. A total of 41 infants
with Candida infection were reviewed. The overall incidence of candidemia was 3.0 per 100 NICU admissions (range, 2.2-3.0). The male : female ratio was 1.6 : 1. The cohort had a median gestational age of 30 weeks (29-31 weeks) and a median birth weight of 1110 g (900-1345 g). The majority of candidemia episodes occurred in VLBW infants (56.1%). The median duration of the total hospital stay was 11 days (8-14 days). Candidemia was catheter-related in 23 cases (56.1%). All Candida infections were classified as LOS. At the moment of candidemia, only ELBW infants were receiving antifungal prophylaxis with fluconazole (3 mg/kg/day).
Results of antifungal susceptibility are shown in Table 2. All of the strains were sensitive to tested drugs. Overall, the

Literature Review.
A total of 45 articles were selected (Tables 3 and 4). Thirty-two studies reported data from a single hospital and 27 were retrospective studies. Seventeen studies were conducted in Asia, 13 in Europe, 11 in North and South America, and 2 in South Africa. Finally, one cohort was carried out in Australia. The distribution of Candida spp. varied according to the different geographical areas. Candida albicans was the dominant species in Europe with proportions ranging from 47 to 100% [10,11,13,14,16,18,19,22,23,53] and in North and South America with proportions ranging from 40 to 69.2% [24-31, 33, 34]. Candida non-albicans species were predominant in Asia [36-40, 42, 43, 45, 47, 48], with proportions ranging from 25 to 92%, with a median of 75% ( Figure 2). In Australia, C. albicans and C. non-albicans were equally distributed (42% and 43%, resp.) [52]. For C. non-albicans, the three most prevalent species were C. parapsilosis complex, C. glabrata complex, and C. tropicalis. Generally, C. parapsilosis complex was the second most common pathogen (range, 6.2-77.8%). C. parapsilosis complex was the predominant species in some studies from Europe [12,15,17,20,21] and Asia [37,40,42,45,48]. The highest proportions of C. glabrata complex were reported in studies that were conducted in the central part of India (range, 22.2-44.4%), while the lowest proportions were observed in European countries (range, 2.5-5.9%). No cases due to C. glabrata complex were reported in South America. The highest frequency of C. tropicalis was found in South India (36.7-92%), followed by studies from South America (11.2-13.3%) and South Africa (8.8%). The lowest frequencies were observed in Europe (3.7-5%) and Australia (2%). There were no reports of C. tropicalis in North America.      Candidemia refractory to conventional antifungals was associated with prolonged antibiotic use and Candida non-albicans infection.

Robinson et al., 2012 [30]
Overall incidence: 0.45 per 100 NICU discharges. An increased time between blood culture draw and initial antifungal therapy was associated with an increased incidence of persistent candidemia.

Batista et al., 2014 [31]
Oral colonization should be considered as a risk factor for candidemia.
Hua et al., 2012 [35] Patients with C. parapsilosis had a significantly longer hospital stay than those with C. albicans sepsis.
Wu et al., 2014 [36] C. guilliermondii was associated with preterm infants and with low birth weight.
Chen et al., 2015 [37] Fluconazole prophylaxis alone was not efficacious; it had to be combined with reinforcement of management and supervision of hand hygiene to effectively prevent invasive candidiasis. C. albicans was the most prevalent species in nonpersistent candidemia. C. parapsilosis was more common among infants with persistent candidemia. Persistent candidemia was associated with an increased risk of mortality. Wu et al., 2009 [48] The most common causative microorganisms of LOS sepsis were CONS and Candida spp. C. parapsilosis was associated with a high mortality rate.  [49] Candidemia had a significantly higher rate of infectious complications, persistent bloodstream infection, and sepsis-attributable mortality than Gram-negative and Gram-positive bacteremia. Lim et al., 2012 [50] Sepsis by Gram-negative bacteria or Candida spp. presented with more severe clinical symptoms and was associated with a higher mortality rate compared with that by Gram-positive bacteria. Chen et al., 2015 [37] Decrease incidence of candidemia during the study period. Ballot et al., 2013 [51] Increased incidence of Candida non-albicans during the study period.

Discussion
This study aimed to describe the epidemiology and drug susceptibility of Candida isolates causing candidemia in a NICU of an Italian university hospital over 9 years. Our survey showed that candidemia is a common problem among critically ill neonates, with an overall incidence of 3%. This finding is higher than data reported in a literature review from Europe (1.1-1.3%) [15,17] and the North and South America (0.5-1.6%) [25,30], but lower than that reported in Asia (4-7.7%) [39,45]. This variability may reflect differences in health care practices among countries, as well as the study design adopted, including differences in the examined population. VLBW infants are known to be at a high risk of candidemia because of more aggressive and invasive therapies, such as indwelling central lines, mechanical ventilation, parenteral hyperalimentation, and longer hospital stay [1][2][3]. The majority of infected neonates have a gestational age at birth of 30 weeks or earlier and birth weight is ≤1500 g (87.8%, each one). Intravenous catheters are risk factors for Candida BSI in critically ill infants. We found that all patients had intravenous catheter placement and that candidemia was catheter-related in 56.1% of cases. This finding is not surprising because Candida spp. can adhere to platelets and fibrinogen on the surface of catheters and form biofilms that may become a reservoir for systemic spread [1][2][3].
Moreover, data regarding changes in the relative frequencies of isolated Candida spp. showed a shift toward Candida non-albicans, with a frequency higher than 50% in some NICUs. This, in part, is attributed to the increased use of azole prophylaxis and therapy [12]. However, in a recent study, where fluconazole was rarely used for prophylaxis and therapy, a high incidence of non-albicans (60.8% of all candidemia episodes) was found [20]. Similarly, our study showed a higher percentage of C. non-albicans (66%) than C. albicans and a variable drift through 9 years. In 2015, 75% of the cases were caused by non-albicans species.
Main risk factors for C. parapsilosis complex infection were the presence of indwelling vascular catheters and parenteral nutrition, both of which predispose to formation of biofilms. Morphogenesis from yeast cells to pseudohyphae is essential for biofilm formation and virulence in C. parapsilosis complex. Amino acids mediate cell differentiation, and this could explain the high incidence of this yeast in catheterized neonates who receive amino acid-rich parenteral nutrition solutions [54]. Our data highlights an association between parenteral nutrition and non-albicans spp. The high proportion of C. parapsilosis complex may explain this finding. Notably, we observed that NICU patients were more likely to develop C. parapsilosis sensu stricto (58.5%) than C. orthopsilosis (2.4%) candidemia. This finding may be explained by the greater capacity of C. parapsilosis sensu stricto to adhere to central lines compared with closely related species [55].
In agreement with other studies [13-15, 17, 18], none of the isolated strains showed resistance to fluconazole and amphotericin B. These are the antifungal drugs of choice that are used in prophylaxis and treatment of Candida BSI in neonates [56]. No fluconazole resistance may be related to the treatment policy in use at our hospital, where systemic antifungal prophylaxis with fluconazole was used only in ELBW infants. In neonates, fluconazole prophylaxis has been linked to the emergence of azole resistance [12,57].

Conclusions
Limitations of the present study are mainly related to its retrospective nature with limited follow-up data. Although all of the data were prospectively collected, some variables could not be examined because of missing data. Furthermore, we did not have data on specific characteristics of noninfected patients in our NICU. Therefore, we were not able to riskadjust our rates to compare with incidences from other reports.
Nevertheless, this study shows that C. non-albicans candidemia is increasing, despite limited use of fluconazole for prophylaxis/empiric therapy in our unit. Our results also confirm that candidemia plays an important pathogenic role in NICU patients. There is a significant variation in cases of candidemia in different geographic regions, even within the same continent. Therefore, monitoring epidemiological data to facilitate the choice of treatment is important.

Ethical Approval
The study protocol was approved by the Ethics Committee of the Azienda Ospedaliero-Universitaria Policlinico of Bari, Italy (Application no. 1321, 2007). Registered data were managed in accordance with the Italian data protection laws (privacy law).

Consent
Written informed consent was obtained from patient parents or their legal guardians.