Candidemia is related to high mortality with figures that oscillate between 40% and 60% according to different series. As well as with higher mortality, fungemia is associated with longer average hospital stays and an increase in the economic cost per process. In recent years, a higher frequency has been noticed in the prevalence study of nosocomial infection in Spanish hospitals (EPINE), reaching figures in some centres of even 50% out of the total of episodes of hospital bacteraemia, thus accounting for the sixth microorganism in frequency.28 On the other hand, we have witnessed several epidemiological and clinical changes in the characteristics and risk factors of the patients, together with increasing resistance to fluconazole in relation to the higher representativity of non-albicans Candida in the total of episodes of candidemia.17,19

In view of these clinical problems, the aim of our study is to compare the changes in the epidemiological and clinical characteristics of 2 cohorts of patients with nosocomial candidemia (NC) in 2 different periods.

The Hospital Universitario Virgen de la Arrixaca, in Murcia (Spain), is a 944-bed facility, 611 of which belong to the general hospital. Its area of care encompasses a population of approximately 450,000 individuals. It is, in addition, a reference hospital in certain specialties such as neurosurgery, burn trauma, cardiovascular surgery and organ transplantation.

All episodes of bloodstream infection due to Candida species during the period from January 2002 to May 2005 (P2) were included and compared to our previous cohort of patients corresponding to the period 1993–1998 (P1).7

The isolation and detection of the species of Candida were done following the microbiological protocols and in accordance with the international regulations of the Centres for Desease Control and Prevention.30 During this study, blood cultures were collected as part of normal clinical practice and incubated for 5 days on the BACTALERT system. Yeasts were identified using the Vitek 2 YST card from bioMerieux (Durham, USA).

The study of susceptibility to amphotericin B and fluconazole was done by using the Yeast One method (TREK Diagnosis Systems, Cleveland, USA) following the NCCLS criteria.14 Thus, Candida spp. was considered to be resistant to fluconazole in the case of MIC>64μg/mL, of intermediate susceptibility or of dose-dependent resistance if MIC was comprised between 16 and 32μg/mL and susceptible if MIC<16μg/mL. Resistance to amphotericin B was defined if MIC>1μg/mL. The quality control was carried out using Candida parapsilosis (ATCC 22019).

Candidemia was considered to be of community acquisition whenever Candida spp. were isolated within the first 72h of hospital admission, provided the patient had not been hospitalised in the previous month. Candidemia was considered to be of nosocomial acquisition when a positive blood culture was obtained after the first 72h of hospitalisation in patients who showed no symptoms or signs of infection at the time of hospital admission. Likewise, when a positive culture was obtained within the first 72h of hospitalisation in a patient who had been hospitalised during the previous month, infection was understood to be of nosocomial acquisition. The retrospective review of the clinical records was done following a study protocol in which epidemiological, clinical and microbiological variables were collected.

Patients were assessed with regard to the prognosis of their underlying disease according to the criteria of McCabe and Jackson.12 Their condition was then classified as “rapidly fatal” when death was expected to take place in days or weeks, “ultimately fatal” when death was likely to take place in some months or years, and “nonfatal” when death was not predictable. The severity of the patient's condition at presentation was assessed according to Winston et al.31 as follows: “critical” when the patient's clinical condition was rapidly deteriorating and the probability of death during the first 24h was high; “poor” when the clinical condition was deteriorating and death was probable but not imminent; “fair” when the clinical condition was deteriorating but death was not probable; and “good” when the clinical situation did not change during the first 24h and death was not probable.

In the assessment of neoplasia, we considered patients with active solid organ cancer or with haematological malignancies.

In order to identify prognostic factors, and in accordance with previous works,22 all clinical and epidemiological characteristics, all complications and the type of antifungal treatment were examined in relation to the final outcome of patients, with “recovery” understood to be the disappearance of all active signs and symptoms of infection.13 When the physician considered the death of the patient to be related to the infection, it was designated as “death related to candidemia.” If the physician considered that death had occurred after recovery from the infection and was related to the underlying disease or other medical or surgical complications, it was classified as “death not related to candidemia.” The death of the patient was taken into consideration when it took place during admission to hospital, in accordance with previous studies.22

Among the epidemiological factors the presence of central venous and urinary catheters or nasogastric tube, the treatment with parenteral nutrition and the previous use of antibiotics were included. As for the latter, the type of agent and number of cycles received were taken into account, a cycle being considered when the patient had received antibiotic treatment with a minimum duration of 7 days in the previous 6 weeks before NC. The previous use of antifungal agents was defined as the administration of this type of drugs in the previous 6 weeks before candidemia. Origin of NC, complications and outcome were analysed according to standard criteria.3

The antifungal used was considered active when, according to microbiological data, it presented in vitro activity (MIC that allowed to define the isolate as susceptible) against the Candida sp. isolated. We considered the initial empiric antifungal treatment to be appropriate when it exhibited in vitro activity against the corresponding Candida sp. isolated, when it was used in a correct dosage, and thus its use did not have to be modified after the antibiogram results were obtained. The efficacy of the treatment was considered to be assessable after at least 5 days of administration. The duration of the treatment was adjusted to the guidelines, requiring a minimum of 14 days to be considered as complete and up to 30–40 days in the case of development of complications.15

“Early” treatment was defined as the empiric treatment initiated facing the presence of fever and before having any microbiological information or in the first 48h after receiving the notification of positive blood culture for yeasts. On the other hand, treatment initiated after 48h of receiving the microbiological information was estimated as “late” treatment.

The removal of the venous catheter was considered as appropriate when it was done in the first 48h after having received the microbiological information.

Data were analysed using the statistics program SPSS 15.0 (SPSS Software, Chicago, USA). A descriptive study was performed for the clinical and epidemiological characteristics as well as for the prognostic factors of patients with NC. The relation or association between pairs of qualitative variables was determined through analyses of contingency tables by means of Pearson's χ2 test, complemented by an analysis of residues with the aim of determining the directional dependence. In the case of quantitative variables, means have been compared using Student's t test. The difference was considered significant at p<0.05. The 2 cohorts of patients (P1 and P2) were statistically compared.

During P2, the complete medical records of 107 patients with NC were available, out of which 79 were men and 28 women, with an average age of 61 years (range 17–87) (52 years—range 18–81 in P1; p<0,05); out of which 53 died (49.5%), against a mortality rate of 51% (41 out of 80) in P1. Annual incidence of the period of study was of 15.8 cases per 10,000 admitted patients while it was 9/10,000 patients/year in the historical series of P1.

In P2, C. albicans was isolated in 49 cases (45.8%) against 53 (66.2%) in P1 (p<0,05). In P2, 9 (8.4%) isolates of Candida were resistant to fluconazole (C. glabrata (8) and C. krusei (1)) and 8 (7.6%) presented intermediate or dose-dependent susceptibility (C. glabrata (6) and C. parapsilosis (2)) and there was only a blood culture whose isolate (C. lusitaniae) was resistant to amphotericin B. In the first study all the isolates were susceptible to fluconazole25 except for the 2 isolates of C. krusei.

Among the predisposing factors for the development of candidemia in P2 the most outstanding were as follows: presence of central catheter (90.6%), parenteral nutrition (90.6%), urinary catheter (90.6%), previous use of 2 or more cycles of broad-spectrum antibiotics (80.4%), previous surgery (78.5%), mechanical ventilation (75.7%) and previous transfusions (70%).

As underlying diseases we highlight diabetes mellitus (25%) and neurological pathology of vascular origin (8.4%) as more prevalent in P2.

Acute severity of illness at onset in P2 was critical in 49.6% (only in 28.7% in P1). Complications turned up in 86 cases in P2 (80.4%) as against 65% in P1. The most statistically significant ones were respiratory distress (42.5%), renal failure (43%) and disseminated intravascular coagulation (DIC) (13%).

The comparative study of the clinical epidemiological pattern of the patients corresponding to P1 and P2 is detailed in tabla 1.

Table 1. Clinical and epidemiological characteristics of nosocomial candidemia in P1 versus P2.

In our comparative study we found an increase of the incidence of candidemia in the second period, a greater clinical seriousness at onset, an increase of the percentage of candidemias caused by non-albicans Candida, and a higher frequency of patients with diabetes mellitus and neurological pathology of vascular origin, without all these differences producing significant changes in their outcome (similar mortality, 50%).

The increase of incidence from 9 to 15.8 per 10,000 admitted patients represents a phenomenon equally noticed in other studies of the literature; thus, in US series it oscillates between 19 and 24 cases per 10,000 hospital admissions.19

On the other hand, although in different series of the literature2,9,13,17,19C. albicans continues to be the most frequent isolate in nosocomial candidemias (50–60%), different studies4,5,11,25,26 have already shown a higher incidence of non-albicans Candida than of C. albicans, as it happens in our cohorts (66.2% versus 45.8%). This phenomenon can be related to the demographic variables, the different groups of patients studied in different hospitals and countries and the previous use of antibiotics (fluoroquinolones) and antifungals (fluconazole). Thus, C. krusei has been related to the presence of acute leukemia as underlying disease, the previous use of fluoroquinolones and fluconazole, although we have not been able to clearly prove this latter aspect.6,24,25 It is also remarkable the increase in the different series, as it happens in ours, of the frequency of isolations of C. parapsilosis in relation with the antecedent of parenteral nutrition and in other studies also in the paediatric population, a cohort that was not included in our study.1,6

In the second cohort we detected one case of candidemia by C. lusitaniae in a critically ill haematological patient who had been hospitalised for a long time, carried central venous catheter with parenteral nutrition, had received various cycles of broad-spectrum antibiotics and prophylaxis with fluconazole and who, as expected, showed bad outcome29 and resistance to amphotericin B. The small number of candidemias by diverse Candida sp. (2–3%) is similar to the result found in longer cohorts (5%).21 These cases are mainly noticed in oncological or haematological patients who carry central venous catheters or devices for the administration of chemotherapy.

In relation with the predisposing factors, we did not find differences between our results and those appearing in the medical literature with regard to seriousness of illness at onset, sources of infection and complications developed by patients with NC.19

As for the percentage of resistance of non-albicans Candida against fluconazole, there are differences among the series corresponding to different countries and hospitals, in such a way that in some cohorts20,23,27 it does not exceed 5%. In our series we found significant resistance of C. glabrata to fluconazole (8%) and 7.6% of the isolates showed intermediate or dose-dependent susceptibility. These results are similar to those of other authors.9,17,18,19 As previously mentioned, this fact has been related to the higher frequency of non-albicans Candida and to the increase of fluconazole consumption, mainly in hospitals with a high number of oncological or haematological patients and bone marrow transplants who are administered prophylaxis with fluconazole. In this sense, another azolic agent, voriconazole, shows a higher in vitro10 activity against C. glabrata than fluconazole. Peman et al.16 collected information from over 27,000 isolates of yeasts and found that 5683 isolates of C. glabrata had a voriconazole MIC90<1μg/mL against a value >32μg/mL in the case of fluconazole. C. glabrata crossed resistance to azoles is variable and can increase with the previous use of fluconazole at subtherapeutic doses, by the induction of genetically codified expulsion pumps. Thus, in isolates with fluconazole MIC>64μg/mL the voriconazole MIC rises and, though 13% of the isolates resistant to fluconazole keep the susceptibility against voriconazole, it would be wise to limit its use as empiric treatment in these patients, specially when there is the antecedent of previous use of fluconazole.

Mortality in the different series13,17,19 of patients with NC oscillates between 40% and 75%. In our experience it has remained at similar levels in both periods of study (around 50%), which highlights the great severity of these infections.8

In conclusion, the percentage of NC by C. non-albicans has increased throughout time and mortality has not varied. Therefore, we consider that constant monitorization of patients who develop candidemia is essential, in order to be able to detect the epidemiological changes and resistance patterns in each hospital, as well as to identify the risk factors that allow early detecting of these infections and thus improving the prognosis of the patients.

Authors have nothing to declare.

Corresponding author.