Determinants of mortality in non-neutropenic ICU patients with candidaemia

The Australian Candidaemia Study involved a three-year prospective nationwide surveillance of all episodes of candidaemia in Australia from August 2001 to July 2004 as reported elsewhere [11]. Fifty of 52 Australian public and private microbiology laboratories participated in the study. Clinical information on each episode was collected on a standardised data form at day 5 and day 30 following the first isolation of Candida species from blood. Data included patient demographics, major concomitant conditions, risk factors occurring within the preceding 30 days (such as surgical and other invasive interventions, vascular access devices, and receipt of total parenteral nutrition, haemodialysis, immunosuppressive therapies and antimicrobial agents), source of candidaemia, clinical signs of sepsis, complications, results of diagnostic studies (including serum creatinine at days 1 and 5), antifungal therapy and clinical outcomes at 30 days. Candida isolates were forwarded to reference laboratories for phenotypic and genotypic species identification (performed in the National Mycology Reference Laboratory, Women's and Children's Hospital, Adelaide and the Molecular Mycology Reference Laboratory, Westmead Hospital, Sydney, respectively), and susceptibility testing using Clinical Standards Laboratory Institutes methodology [12] (performed in the National Mycology Reference Laboratory). Approval for the study was obtained from the Human Research Ethics Committees of all participating institutions. Informed written consent was obtained from patients that were included in the study.

The definition of ICU acquisition of candidaemia was the occurrence of the first positive blood culture growing Candida species at 48 hours or more following ICU admission or 48 hours or less following ICU discharge. Paediatric, neonatal or neutropenic (absolute neutrophil count ≤ 1 × 10⁹ neutrophils/L) patients were excluded. ICUs included hospital wards or units providing invasive ventilatory and/or intensive haemodynamic support; high dependency units and coronary care units were excluded. Risk factors over the 30 days prior to onset of candidaemia were assessed. Vascular access device-related candidaemia required the isolation of the same Candida species from both blood and catheter tip. Relapses were defined as recurrent positive blood cultures with the same species within 30 days of the original positive blood culture after an initial clinical and microbiological response.

Clinical data were analyzed using SPSS (Version 16.0, SPSS, Chicago, IL, USA). Incidences were calculated using ICU admission data from participating ICUs covering the study period. Univariate analyses were performed using the Student's t test (continuous variables) and chi-squared or Fisher's exact tests (categorical variables). Assessment of factors associated with mortality were performed using multivariate Cox regression models with hospital mortality as the dependant variable, censored for hospital discharge, using the backwards selection method after initially including all biologically-plausible variables, and those with an unadjusted association of P < 0.2. A P < 0.05 was set as the limit for acceptance or removal of variables. For all analyses, the fluconazole dose was adjusted for renal impairment [13] and the fluconazole minimum inhibitory concentration (MIC) and fluconazole dose:MIC ratio were log₁₀ transformed. Survival analyses were performed on two separate patient cohorts: the entire ICU cohort and the subset of patients in whom fluconazole was the sole initial antifungal therapy during the initial 72 hours of therapy (but exclusion of patients with breakthrough infection, defined as occurrence of candidaemia more than 72 hours prior to collection of the first positive blood culture).

Over the three-year study period, there were 183 episodes of ICU-acquired candidaemia in 183 patients from 38 ICUs. The mean age ± standard deviation was 58.6 ± 18.6 years and 57% were male. Most patients had undergone a recent surgical procedure (67%), had received recent antimicrobial therapy (97%) and were ventilated at the time of candidaemia diagnosis (79%). The median time from ICU admission to development of candidaemia was eight days (interquartile range (IQR), 5 to 15 days; range, 2 to 86 days). Almost three-quarters of episodes (74%) occurred in tertiary-referral hospital ICUs. The overall incidence of ICU-acquired candidaemia calculated from 19 ICUs (of the 22 ICUs reporting at least three candidaemia episodes) ranged from 0.53 to 6.46 per 1000 ICU admissions (mean, 2.06 per 1000 admissions; 95% confidence interval, 1.73 to 2.44).

Candida albicans accounted for 62% (111/178) of episodes, Candida glabrata for 18% (32), Candida parapsilosis for 8% (14), Candida tropicalis for 6% (10), Candida krusei for 4% (7), Candida dubliniensis for 1% (2), and other Candida species accounted for 3% (5) of episodes. There were two mixed infections (one C. albicans/C. glabrata and one C. glabrata/unidentified Candida species). Antifungal susceptibility results were available for 174 isolates (Table 1). Based on Clinical Standards Laboratory Institutes MIC breakpoints [14‐16], all isolates were susceptible to amphotericin B, 167 (96%) to flucytosine, 136 (78%) to fluconazole, 115 (66%) to itraconazole and 172 (99%) to voriconazole. Of the 32 C. glabrata isolates, two (6%) were susceptible to fluconazole, 23 (72%) were susceptible-dose dependent and seven (22%) were resistant. Two of the fluconazole-resistant isolates were also resistant to voriconazole. All seven C. krusei isolates were susceptible to voriconazole. All of the other 135 Candida isolates were fluconazole susceptible with the exception of one isolate of C. albicans that demonstrated dose-dependent susceptibility. The MIC₉₀ for caspofungin was 0.25 μg/mL (n = 54: Table 1).

Manifestations of sepsis [17] were common both at diagnosis of candidaemia (84%) and at day 5 (76%). The source of candidaemia was attributed to an intravascular device in 35%, an intra-abdominal source in 10%, the urinary tract in 3%, other sources in 5% and an unknown source in 47%.

Ophthalmological manifestations consistent with intraocular candidiasis were demonstrated in six of 48 (13%) patients who underwent ocular examination. Other infective complications of candidaemia included nine episodes of metastatic renal infections, three cases of endocarditis (all graded 'possible' infection by Duke's criteria [18]) and two patients with hepatosplenic candidiasis (documented by computed tomography scan and post-mortem examination). Relapses occurred in 24 of 183 (13%) episodes. Neither metastatic infective foci nor relapses were associated with specific Candida species or any underlying co-morbidity.

Antifungal therapy was initiated in 156 (85%) patients: of these, fluconazole in 76%, amphotericin B deoxycholate in 12%, a lipid formulation of amphotericin B in 4%, caspofungin in 4%, and voriconazole or posaconazole in 3%. There was considerable variation in the time to initiation of antifungal therapy: in 21% it was initiated within 24 hours of drawing the first positive blood culture, in 14% between 24 and 48 hours, in 29% between 48 and 72 hours, and in 35% greater than 72 hours.

Other processes of care were also assessed. Among patients surviving five days, follow-up blood cultures were obtained in 68% and central venous catheters removed within five days in 80%. Among patients surviving 30 days, an ocular examination had been performed in 36% (90% of which were performed by an ophthalmologist).

Among the entire ICU cohort with outcome data (n = 173), the crude in-hospital 30-day mortality was 56%, with median time to death after drawing the first positive blood culture of seven days (IQR, 2 to 12 days). Variables associated with an increased risk of death by multivariate Cox regression analyses are presented in Table 2: those independently associated by multivariate analysis included host-related factors (increasing age, mechanical ventilation at time of candidaemia and management in the ICU for reasons other than multitrauma) and non-receipt of systemic antifungal therapy. Several other host-related factors (total parenteral nutrition, receipt, haemodialysis receipt and presence/non-removal of vascular access devices) were also associated with mortality on univariate – but not multivariate – analysis. Of note was the lack of association with mortality for different Candida species or for delays in initiation of antifungal therapy.

Among the subset of patients who received initial fluconazole therapy for the first 72 hours (n = 93), crude mortality was 52%. Variables independently associated with increased risk of death by multivariate Cox regression analysis included increasing age and haemodialysis receipt (Table 3). Although there was a non-significant trend between time to initiation of fluconazole and mortality (Table 4), this was not significant by multivariate analysis. Organism-related (Candida species and fluconazole MIC), pharmacokinetic-related (renal-adjusted fluconazole dose) or pharmacodynamic-related (fluconazole dose:MIC ratio) factors were not associated with mortality.