Invasive candidiasis in intensive care units in China: a multicentre prospective observational study

Abstract

Objectives

To describe the epidemiology, microbiology and management of invasive Candida infection (ICI) in intensive care units (ICUs) in China.

Methods

A multicentre, prospective, observational study in 67 hospital ICUs across China. Patients were ≥18 years old with clinical signs of infection and at least one of the following diagnostic criteria: (i) histopathological, cytopathological or microscopic confirmation of yeast cells from a normally sterile site; (ii) at least one peripheral blood culture positive for Candida; and (iii) positive Candida culture from a normally sterile site. The China-SCAN study is registered with ClinicalTrials.gov (NCT T01253954).

Results

ICI incidence was 0.32% (306 patients/96 060 ICU admissions) and median time between ICU admission and diagnosis was 10.0 days. Candida albicans was the most prevalent single isolate (41.8% of patients), although non-albicans species accounted for the majority of infections. Diagnostic confirmation was based solely on at least one positive blood culture in 290 (94.8%) cases. Treatment was initiated after diagnostic confirmation in 166/268 (61.9%) patients. Triazoles (62.7%) and echinocandins (34.2%) were the most commonly used antifungal agents; first-line therapy was typically fluconazole (37.7%). The median duration of antifungal therapy was 14 days. The mortality rate was 36.6% (112/306); the median time between diagnosis and death was 14.5 days. Mortality was higher in older individuals, those with solid tumours, those with recent invasive mechanical ventilation and those with a higher sequential organ failure assessment score at diagnostic confirmation. Susceptibility to first-line antifungals was associated with lower mortality than dose-dependent susceptibility or complete resistance (P = 0.008).

Conclusions

More infections were caused by non-albicans than Candida albicans strains. The majority of patients were treated only after diagnostic confirmation, rather than empirically. First-line antifungal susceptibility was associated with lower mortality.

invasive Candida infection, Candida albicans, ICUs, antifungal

Introduction

Candida species are among the most common fungal pathogens in hospitalized patients, particularly in intensive care units (ICUs), where invasive Candida infection (ICI) is associated with mortality rates of 35%–80% in immunocompromised and other seriously ill patients.^1–4 Although firm incidence data are limited due to diagnostic difficulties,⁵ ICI is the third most common cause of infection in ICUs worldwide, accounting for 17% of infections.⁶ The 2011 global epidemiological EPIC II study, which examined data from 76 countries, reported ICU Candida bloodstream infections in 6.9/1000 patients.² Invasive candidiasis in ICUs is an increasing concern due to the high associated mortality and resource utilization.⁷ A typical episode of candidaemia in the USA costs US$25 000–55 000,² while in China, the mean hospitalization cost for patients with invasive fungal infection (IFI) in 2004/2005 was ∼US$17 000, significantly higher than that for patients without IFI (US$8500; P = 0.001), despite similar mean daily costs (∼US$500).⁸

There are significant geographical differences in the epidemiology of Candida infection. In ICUs in the USA, Candida albicans accounts for 50%–70% of Candida infections.^2,9 Non-albicans species are more common in parts of Asia, South America and southern Europe,³ although, as a heterogeneous group with varying epidemiology, virulence and antifungal susceptibilities, prevalence varies widely between geographical areas. Knowledge of regional differences is therefore an important guide to local prescribing practices and health policy, and has far-reaching clinical implications due to the different patterns of disease severity and mortality associated with different Candida species.^2,10 For example, although C. albicans has been the most commonly isolated strain in hospital patients during the past 20 years, there is a trend towards increasing rates of infection with non-albicans strains.¹Candida glabrata is now responsible for 15%–20% of Candida infections in most countries, a shift believed to reflect the increasing use of azole antifungals and changing patterns of resistance to widely used antifungal agents.^11,12

As pharmaceutical and technological advances and more aggressive treatment practices allow people to live longer with serious illness, more patients are treated in ICUs and the incidence of nosocomial fungal infections, including candidiasis, is increasing.^3,11,13,14 Data on the epidemiology and management of IFI in critically ill patients in China are limited and definitions of IFI vary, complicating comparison between studies. However, earlier diagnostic testing and timely initiation of adequate antifungal therapy in hospitalized patients,^15–17 including treatment of those in the ICU,^18,19 are known to lower mortality rates and should be prioritized.

Most studies of Chinese data published to date are retrospective, single-centre studies in which ICI is not clearly defined and diagnostic techniques are not clearly described or which fail to comply with recognized standards. We conducted a prospective, multicentre study in ICUs in China (The China Survey of Candidiasis or China-SCAN) between November 2009 and April 2011, with the aim of describing the epidemiology, microbiology, management and outcomes of ICIs in this critical care setting. These data should be used to develop clinical practice recommendations to improve the management and outcomes of such cases.

Methods

Study design and patients

The China-SCAN study was a prospective, observational study in 67 closed ICUs in general hospitals distributed throughout China, conducted between November 2009 and April 2011. Eligible ICUs had managed at least three cases of proven, probable or possible IFI, defined according to national guidelines,²⁰ in the previous year and had ≥10 ICU beds; physicians caring for patients with ICIs in the ICU served as study investigators. Patient inclusion criteria were age ≥18 years with clinical signs of infection (in the opinion of the treating physician) and at least one of the following diagnostic criteria: (i) histopathological, cytopathological or direct microscopic confirmation of yeast cells in a specimen obtained by needle aspiration or biopsy from a normally sterile site (other than mucous membranes); (ii) at least one peripheral blood culture positive for Candida; and (iii) positive Candida culture from a sample obtained by sterile technique from a normally sterile site (e.g. cerebrospinal, pleural, peritoneal or peritoneal abscess fluid). Patients were excluded if they had a positive blood culture from a catheter but negative peripheral blood cultures, positive culture from a normally sterile site that could not be confirmed as uncontaminated (following placement of a drainage tube, culture from a peritoneal abscess obtained by paracentesis or following placement of a drainage tube, ascitic fluid sample obtained via drainage tube following surgery for intestinal perforation or anastomosis), ICI diagnosed before ICU admission or non-ICI (e.g. candiduria). All eligible patients were included. Patient management was at the discretion of the attending physician. Patients were followed until discharge or death, whichever occurred sooner. The study was approved by the Ethics Committee of Zhongda Hospital of Southeast University, the lead investigation site. Other participating hospitals accepted the central ethics committee review or conducted a further, independent, ethics review, according to their own institutional policy (20 hospitals). The study complied with the Declaration of Helsinki regarding ethical principles of human subjects research and the relevant ethical requirements of the International Conference on Harmonisation/Good Clinical Practice guidance and national regulations. All patients provided written informed consent. The China-SCAN study is registered with ClinicalTrials.gov (NCT T01253954).

Data collection and management

A case report form (CRF) was completed for each patient by his/her physician as soon as possible after confirmed diagnosis of Candida infection, defined as the date on which the first positive specimen was obtained. The CRF included the following: ICI risk factors; disease severity indicators [acute physiology and chronic health evaluation II (APACHE II) and sequential organ failure assessment (SOFA) scores]; clinical manifestations of candidiasis; diagnostic techniques and outcomes; timing of first-line antifungal treatment; number, type and duration of antifungal therapy and basis for their selection; and clinical and microbiological outcomes. Complete clinical efficacy was defined as elimination of all signs and symptoms of ICI, and accompanying radiographic resolution; partial efficacy was defined as improvement rather than complete elimination. Microbiological elimination was defined as negative culture from the original infection site. Patients were considered to have relapsed or become reinfected when a second course of antimicrobial therapy was required within 14 days of complete or partial clinical resolution, or when repeat culture from the original site was positive for Candida. Each centre provided demographic and clinical information, including age, gender, date of ICU admission and discharge, and survival status at discharge, for all ICU patients admitted each month, to facilitate data analysis.

All data from the CRFs were entered in to a proprietary Microsoft SQL server-based data management system with audit trails validated according to General Principles of Software Validation; Final Guidance for Industry and FDA Staff.²¹ Data monitoring and analysis, including confirmation of eligibility and adherence to the study protocol, were conducted by the independent contract research organization and approved by the principal investigator. Descriptive statistics, such as the mean, standard deviation (SD), median and range, were used for continuous variables; frequency and percentage were used to describe categorical variables. Groups were compared using Student's t-test, analysis of variance, Wilcoxon rank-sum test, χ² test or Fisher's test, as appropriate. Risk factors for mortality were analysed using univariate analysis followed by multivariate regression. A total of 31 variables that were considered potential risk factors for mortality were included in the univariate analysis, including age, gender, concomitant disease, diagnosis at admission, illness severity (APACHE II and SOFA scores), recent invasive procedures, susceptibility to first-line antifungal agents, management of catheters after confirmed ICI diagnosis and use of immunopotentiation therapy after ICI diagnosis. Each risk factor with a P value <0.15 on univariate analysis was examined further using multivariate logistic regression in which clinical significance and interaction between variables were taken into account. Application of the Hosmer–Lemeshow goodness-of-fit test resulted in a P value of 0.165, indicating an acceptable fit. Factors with P ≤ 0.05 in the final analysis were considered statistically significant.

Fungal samples were transferred to a central laboratory (Research Centre for Medical Mycology, Peking University First Hospital, Beijing, China) for in vitro susceptibility testing of all isolates, using the CLSI M27-A3 microbroth dilution method, against amphotericin B, fluconazole, voriconazole, itraconazole and caspofungin. The breakpoints for antifungal susceptibility were those defined by the CLSI for fluconazole [susceptible ≤8 mg/L; susceptible dose-dependent (SDD) 16–32 mg/L; and resistant ≥64 mg/L], voriconazole (susceptible ≤1 mg/L; SDD 2 mg/L; and resistant ≥4 mg/L) and caspofungin (CLSI echinocandin breakpoint = 2 mg/L).^22,23 The breakpoint for itraconazole in CLSI M27-A3 is for Candida species isolated from mucosal infections; the MICs of this drug for isolates in this study were not defined. The breakpoint for amphotericin B susceptibility was ≤2 mg/L, based on the literature.^24,25 Comprehensive methods and results of the antifungal susceptibility testing will be described in greater detail in a separate publication.

Results

Incidence and patient characteristics

A total of 96 060 ICU patients from 67 centres throughout China were screened. Of these, 306 (0.32% or 32 per 1000 ICU admissions) were diagnosed with ICI. The majority of patients (62.1%) were transferred to the ICU from another department within the same hospital. The median time between hospital admission and confirmed diagnosis of ICI was 15.5 days (range: 0–1611 days) and between ICU admission and diagnosis was 10.0 days (0–330 days); 80.7% of ICI cases occurred >48 h after ICU admission.

The baseline characteristics of the patients with ICI are presented in Table 1. The most common underlying diseases were type 2 diabetes, chronic cardiac dysfunction and solid tumours. The mean (SD) APACHE II score at admission was 27.2 (7.2). One-quarter of patients had organ failure at admission. Based on a definition of severe sepsis as the presence of known or suspected infection in association with systemic inflammatory response syndrome and organ dysfunction, 167 (54.6%) patients had severe sepsis at diagnosis according to data drawn from the CRFs. Among these patients, 86/167 (51.5%) had septic shock, 28.1% of all ICI patients (86/306). Endophthalmitis and meningitis were reported in five patients (1.6%) and two patients (0.7%), respectively. Immune suppression was evident in <6% of patients.

Table 1.

Baseline characteristics of 306 patients diagnosed with ICI in 67 hospital ICUs in China

Category	Value
Age (years), mean (SD)	61.5 (20.0)
Gender, male/female	210/96
Body weight (kg), mean (SD)	62.8 (10.6)
Concomitant disease, n (%)
type 2 diabetes	66 (21.6)
chronic cardiac dysfunction	64 (20.9)
solid tumour	57 (18.6)
chronic obstructive pulmonary disease	35 (11.4)
chronic renal insufficiencya	33 (10.8)
chronic hepatic insufficiencyb	16 (5.2)
haematological malignancy	3 (1.0)
type 1 diabetes	1 (0.3)
Route of hospital admission, n (%)
transfer from within hospital	190 (62.1)
directly to ICU from outpatient or emergency department	62 (20.3)
transfer from ICU of different hospital	54 (17.6)
Type of ICU, n (%)
general	255 (83.3)
surgical	34 (11.1)
medical	7 (2.3)
emergency	8 (2.6)
neurological	2 (0.7)
Organ failure at ICU admission, n (%)
single	59 (19.3)
multiple	18 (5.9)
Nature of injury at ICU admission, n (%)
surgical	58 (19.0)
traumatic	37 (12.1)
Illness severity at ICU admission, mean (SD)
APACHE II score	27.2 (7.2)
SOFA score	11.1 (3.4)

APACHE II is a disease severity classification system applied within 24 h of admission to an ICU. An integer score from 0 to 71 is calculated and higher scores indicate more severe disease and higher risk of death.

SOFA is another ICU scoring system based on organ function or failure rate. The score is based on scores for the respiratory, cardiovascular, hepatic, coagulation, renal and neurological systems.

^aAll patients receiving long-term haemodialysis.

^bAs in APACHE II: biopsy-proven cirrhosis and documented portal hypertension; episodes of past upper gastrointestinal bleeding attributed to portal hypertension; or prior episodes of hepatic failure/encephalopathy/coma.

Approximately three-quarters of patients diagnosed with ICI had a recent history of invasive mechanical ventilation, urethral or central venous catheterization or antibiotic therapy (median duration: 10 days) (Table 2). The management of indwelling catheters varied: central venous and drainage catheters in situ at diagnosis were removed after diagnosis as part of infection management in 82/209 (39.2%) and 14/88 (15.9%) patients, respectively. Other risk factors present in 30%–40% of patients during the 2 weeks prior to study entry were surgery, gastrointestinal dysfunction, total parenteral nutrition and presence of a drainage tube. Of the patients who had had surgery, the majority had undergone abdominal procedures. The median time from surgery to confirmed ICI diagnosis was 9.0 days.

Table 2.

Predisposing factors for fungal infection present within 2 weeks (3 months for immunosuppressants) prior to study entry, or at diagnosis, among 306 patients with ICI in ICUs in China; percentages reflect the proportion of the whole study population (n = 306)

Risk factor	n (%)
Immune compromised
immunosuppression	17 (5.6)
immunosuppressant therapya	15 (4.9)
HIV infection	2 (0.7)
neutropenia	5 (1.6)
Gastrointestinal dysfunctionb	181 (59.2)
Invasive procedures
catheterizationc	279 (91.2)
central venous	253 (82.7)
urethral	232 (75.8)
drainage tube	111 (36.3)
indwelling arterial	54 (17.6)
invasive mechanical ventilation	236 (77.1)
total parenteral nutrition	130 (42.5)
surgery	117 (38.2)
abdominal	77 (25.2)
Antibiotic therapy	241 (78.8)
cephalosporins	119 (38.9)
carbapenems	94 (30.7)
penicillins	62 (20.3)
glycopeptides	52 (17.0)
quinolones	37 (12.1)
two-drug combinations	156 (51.0)
monotherapy	85 (27.8)

^aSteroid therapy: >0.5 mg/kg/day of prednisone over 1 month (n = 8), cancer chemotherapy (n = 7), post-solid organ transplant immunosuppression (n = 2), allogeneic bone marrow transplantation or allogeneic haematopoietic stem cell transplantation (n = 1) or tumour necrosis factor therapy (n = 3).

^bHaemorrhage, food intolerance, perforation, surgery, acalculous cholecystitis or intra-abdominal hypertension.

^cIncludes patients who were catheterized within 2 weeks of the first positive sample but in whom the catheter was removed before diagnosis.

Microbiology

A total of 389 Candida isolates from 244 patients were identified and analysed in the central laboratory. Although all 306 patients had ICI diagnosis confirmed in local laboratories, not all hospitals sent isolates to the central laboratory for confirmation (for reasons including individual hospital policy and suboptimal storage or handling of isolates), so the dataset is incomplete in this regard. C. albicans was the most prevalent species, isolated in 102/244 (41.8%) patients [mixed infection with other Candida species in 4/244 (1.6%)]; other common isolates were Candida parapsilosis [58/244 (23.8%); mixed infection in 3/244 (1.2%)], Candida tropicalis [43/244 (17.6%); mixed infection in 2/244 (0.8%)] and C. glabrata [30/244 (12.3%); mixed infection in 3/244 (1.2%)]. Further Candida species were identified in <2% of samples.

More than 90% of C. albicans samples were susceptible to fluconazole (145/156; 92.9%), voriconazole (154/156; 98.7%) or caspofungin (156/156; 100%), but only 0.6% (1/156) showed susceptibility to itraconazole. Indeed, resistance to itraconazole was widespread, with rates as high as 96.0% (48/50) among C. glabrata, which also exhibited fluconazole resistance in 18.0% (9/50) of samples. Voriconazole resistance was identified in 7/389 (1.8%) Candida isolates. No resistance to caspofungin was identified in any Candida strain.

Diagnosis and management of ICIs

The most common diagnostic procedure was the culture of peripheral blood, ascitic fluid or pleural effusion, undertaken in 100%, 16.6% and 6.6% of patients, respectively. Based on CRF documentation, the median time from sample collection to positive report of fungal culture was 5 days (Q1–Q3: 3–6 days). Diagnostic confirmation was based solely on at least one positive blood culture in 290 (94.8%) cases, on positive fluid culture from a normally sterile site (cerebral spinal fluid, ascitic fluid or pleural fluid) in 12 (3.9%) cases and on candidaemia combined with positive culture from a normally sterile site in 3 (1.0%) cases. Diagnosis was confirmed by histopathology in one patient (0.3%).

A total of 268/306 (87.6%) patients received antifungal therapy, with a mean delay of 0.3 days between specimen collection and treatment initiation. Among the 38 patients who did not receive antifungal therapy, reasons included discharge from the ICU (n = 15) and death before, or soon after, diagnostic confirmation (n = 13). Ten patients withdrew from treatment or chose to forego comprehensive treatment due to imminent death.

A total of 58 (21.6%) patients started treatment before the first diagnostic specimen was collected. Antifungal treatment was initiated prior to diagnostic confirmation of ICI (as empirical therapy) in 74/268 (27.6%) patients and after confirmation in 166 (61.9%) patients. Antifungal therapy was started following indirect microbiological evidence of infection, e.g. β-glucan test, or positive culture from bronchoalveolar lavage sample, alongside clinical evidence, in 28 patients (10.4%).

First-line treatment comprised a single agent in 264/268 (98.5%) patients, most commonly fluconazole (101/268; 37.7%). Other widely used first-line agents were caspofungin (64/268; 23.9%) and voriconazole (49/268; 18.3%). Physicians indicated that their selection of first-line antifungal agent was based on the APACHE II score in 161/268 (60.1%) patients and on the SOFA score in 118/268 (44.0%), and took into account renal function and potential for renal toxicity in 137/268 (51.1%) and need to avoid drug interactions in 81/268 (30.2%). Other contributory factors were physicians' previous treatment experience (37/268; 13.8%) and other considerations, including economic factors and results of in vitro susceptibility testing (n ∼15; 5.6% for all). The most commonly used class of antifungal overall (throughout the treatment of IFI) was the triazoles (288/459; 62.7%), followed by the echinocandins (157/459; 34.2%). The median duration of antifungal therapy was 14 days (range: 0–104 days). Antifungal therapy was modified at least once in almost half of all patients (132/268; 49.3%) and 11.6% of patients (31/268) had treatment adjusted three or more times. The most common reasons for the first treatment adjustment were as follows: new availability of microbiological data, such as fungal culture or susceptibility testing results, following empirical treatment (115/132; 87.1%); lack of efficacy with the current agent (64/132; 48.5%); and, in fewer patients, clinical improvement necessitating a dose reduction or change to oral antifungals (21/132; 15.9%).

Treatment outcomes and prognostic factors

The majority of evaluated patients (193/266; 72.6%) experienced clinical improvement, although complete clinical remission occurred in only 94/266 (35.3%) patients. Treatment was clinically ineffective in 73/266 (27.4%). A total of 136/268 (50.7%) patients experienced microbiological eradication of Candida. Four patients (4/193; 2.1%), all with candidaemia, relapsed or were reinfected after clinical improvement, with a median time of 6.5 days.

In total, 60 patients withdrew from or decided to forego comprehensive treatment, of whom 29 died in hospital and 31 were discharged at their request. The overall mortality rate was 36.6% (112/306), although this increased to 40.7% (112/275) when based only on patients who died in hospital or completed treatment and were discharged at their physician's discretion; the median time from ICI diagnosis to death was 14.5 days. Analysis of the risk factors for mortality revealed a higher risk with increasing age and higher SOFA score at the time of diagnostic confirmation, and in individuals with solid tumours or recent invasive mechanical ventilation (Table 3). There was no difference in antifungal treatment duration between those who died and those who survived (median duration 12.5 (1–99) versus 16 (1–105) days, respectively; P = 0.12).

Table 3.

Risk factors for death (univariate analysis) among 306 patients with ICI in ICUs in China; values in bold are statistically significant at the 5% level

Category	Death (n = 112)	No death (n = 194)	Comparison (P value)
Age (years), mean (SD)	66.3 (20.6)	58.8 (19.2)	0.001
Gender, n (%)			0.445
male	80 (71.4)	130 (67.0)
female	32 (28.6)	64 (33.0)
Body weight (kg), mean (SD)	61.8 (10.6)	63.3 (10.6)	0.341
Concomitant disease, n (%)
type 1 or type 2 diabetes	31 (27.7)	37 (19.1)	0.063
chronic cardiac dysfunction	29 (25.9)	35 (18.0)	0.104
solid tumour	28 (25.0)	29 (14.9)	0.030
chronic obstructive pulmonary disease	14 (12.5)	21 (10.8)	0.710
chronic renal insufficiencya	17 (15.2)	16 (8.2)	0.060
chronic hepatic insufficiencyb	7 (6.3)	9 (4.6)	0.597
haematological malignancy	1 (0.9)	2 (1.0)	0.906
Organ failure at ICU admission, n (%)			0.241
single	25 (22.3)	34 (17.5)
multiple	9 (8.0)	9 (4.6)
none	78 (69.6)	151 (77.8)
Nature of surgical injury at ICU admission, n (%)			0.084
surgical	16 (14.3)	42 (21.6)
traumatic	10 (8.9)	27 (13.9)
none	86 (76.8)	125 (64.4)
Illness severity at diagnosis, mean (SD)
APACHE II score	28.2 (6.9)	26.6 (7.3)	0.059
SOFA score	11.6 (3.1)	10.8 (3.6)	0.071
Illness severity at ICU admission, mean (SD)
APACHE II score	27.7 (7.0)	26.7 (7.1)	0.212
SOFA score	11.6 (3.2)	10.2 (3.6)	0.001
Invasive procedures within 2 weeks prior to diagnosis, n (%)
catheterization
central venousc	98 (87.5)	155 (79.9)	0.090
urethral	91 (81.3)	141 (72.7)	0.092
drainage tubed	40 (35.7)	71 (36.6)	0.877
indwelling arterial	21 (18.8)	33 (17.0)	0.701
invasive mechanical ventilation	97 (86.6)	139 (71.6)	0.003
total parenteral nutrition	53 (47.3)	77 (39.7)	0.230
surgery	40 (35.7)	77 (39.7)	0.542
Management of central venous catheter, n (%)			0.091
no catheter	27 (24.1)	70 (36.1)
removed after first positive sample obtained	52 (46.4)	75 (38.7)
not removed after first positive sample obtained	33 (29.5)	49 (25.3)
Management of drainage catheter, n (%)			0.253
no catheter	77 (68.8)	141 (72.7)
removed after first positive sample obtained	3 (2.7)	11 (5.7)
not removed after first positive sample obtained	32 (28.6)	42 (21.6)
Immunopotentiation therapy after ICI diagnosis, n (%)	53 (47.3)	74 (38.1)	0.117

^aAll patients receiving long-term haemodialysis.

^bAs in APACHE II: biopsy-proven cirrhosis and documented portal hypertension; episodes of past upper gastrointestinal bleeding attributed to portal hypertension; or prior episodes of hepatic failure/encephalopathy/coma.

^cIncluding internal jugular, subclavian and femoral vein catheters present within 2 weeks of first positive sample and still in situ at diagnosis.

^dIncluding thoracic, abdominal and intraventricular catheters present within 2 weeks of first positive sample and still in situ at diagnosis.

Patients who responded to first-line antifungal therapy (i.e. isolates were susceptible on susceptibility testing) had significantly lower mortality (56/179; 31.3%) than those who demonstrated partial (9/16; 56.3%) or complete (9/18; 50.0%) resistance (P = 0.04). On multivariate logistic regression, risk factors for death included older age, higher SOFA score at time of diagnosis and failure to remove the drainage catheter after the first positive sample was obtained (Table 4). Susceptibility to first-line therapy was associated with lower mortality than resistance or susceptible dose dependence to first-line therapy.

Table 4.

Risk factors for death (multivariate analysis) among 306 patients with ICI in ICUs in China; Hosmer–Lemeshow goodness-of-fit test: χ² 11.69; P = 0.16

Factors	Estimate	Standard error	χ2	P value	OR	95% CI
Intercept	−4.496	1.516	8.801	0.003	—	—
Age	0.020	0.009	4.779	0.029	1.020	1.002; 1.039
SOFA score at diagnosis	0.156	0.055	8.009	0.005	1.169	1.049; 1.303
Susceptibility to first-line therapy (susceptibility versus SDD/resistance)	−1.135	0.429	6.992	0.008	0.321	0.139; 0.745
Drainage
no drainage catheter versus catheter removed after first positive sample	1.961	1.129	3.018	0.082	7.107	0.778; 64.948
failure to remove catheter after first positive sample versus catheter removed	2.402	1.178	4.157	0.042	11.050	1.098; 111.248
Central venous catheter
no central venous catheter versus catheter removed after first positive sample	−0.763	0.424	3.235	0.072	0.466	0.203; 1.071
failure to remove central venous catheter after first positive sample versus catheter removed	−0.648	0.431	2.257	0.133	0.523	0.225; 1.218
Immunopotentiation therapy (yes versus no)	0.547	0.338	2.614	0.106	1.728	0.890; 3.353

Compared with general ICU patients, those with an ICI were hospitalized for longer (median: 25 versus 3 days; P < 0.0001), were more likely to discontinue or forego ICI treatment (19.6% versus 11.7%; P = 0.0001) and had a significantly higher mortality rate (36.6% versus 8.9%; P < 0.0001).

Patients who were susceptible to first-line antifungal therapy were more likely to experience microbiological elimination of Candida species and complete clinical resolution than those who were resistant [microbiological elimination: 52.0% (93/179) versus 22.2% (4/18); P < 0.05; complete clinical resolution: 34.6% (62/179) versus 11.1% (2/18); P < 0.05].

Compared with patients with C. albicans infection, those infected with non-albicans strains were more likely to require adjustment of first-line antifungal therapy (52.7% versus 30.6%; P = 0.002), stayed longer in hospital (median: 44 versus 32 days; P = 0.002) and the ICU (median: 29 versus 18 days; P = 0.004), and showed a non-statistically significant trend towards higher mortality (38.4% versus 29.6%; P = 0.17). Susceptibility to first-line antifungal treatment was similar (∼83%) among albicans and non-albicans strains.

Discussion

Data on ICI in Chinese ICUs are scarce and this is, to our knowledge, the largest prospective study of ICI in China, and possibly anywhere. The incidence of ICI in ICUs across China was 0.32%, a rate that accords with global incidence rates of 0.03%–0.5% in hospital-based studies.³ Ours may be an underestimate of the true incidence, as diagnosis was based solely on positive blood culture, i.e. candidaemia, in 94.8% of cases. This may reflect the difficulty of more invasive sampling techniques in critically ill patients, but nonetheless breaches guidance recommending multisite sampling.²⁶ Diagnostic sampling occurred after starting antifungal therapy in some cases and we cannot exclude insufficient fungal culture time, despite providing recommendations (≥7 days at 28–30°C) to all centres. Strict entry criteria to exclude possible contamination may also have contributed to underestimation of the true incidence. Because we did not collect all variables related to severe sepsis or septic shock, this rate might also be a slight underestimate.

In accordance with previous studies,²⁷ mortality was significantly higher and ICU stay longer in patients with ICI than in the general ICU population. The mortality rate of 36.6% and the higher rate of 40.7%, based on patients who died in hospital or completed treatment and were discharged at their physician's discretion, are similar to reported rates of 30%–60% among ICU patients²⁸ and in two small retrospective studies in China.^14,29

Although C. albicans is the most prevalent Candida species in the ICI literature,³⁰ non-albicans species are increasing in frequency as the albicans species decreases, both globally³¹ and in China.³² The reasons for this shift may include the increasing use of relatively low-cost azoles, surgery and an ageing population, consistent with the success of aggressive treatment in seriously ill patients. In our study, which concurs with that of Bassetti et al.³³ in Italy, C. albicans was the most frequently isolated Candida species, identified in 41.8% of patients, although non-albicans species, particularly C. parapsilosis and C. glabrata, were more commonly isolated overall. Bassetti et al.'s finding that limiting fluconazole use can reduce non-albicans candidaemia should spur reconsideration of the use of fluconazole for IFI prophylaxis.

Interestingly, our findings and those of Bassetti et al.³³ are not typical of those from other studies in European countries^34–37 or China, where C. albicans remained predominant in recent studies. Aside from two small Chinese ICU studies that identified C. albicans in ∼45% of ICI patients,^38,39 the majority of recent IFI studies in China report C. albicans rates of 55%–60%,^8,40,41 although these studies were less rigorous than ours in confirming ICI⁴¹ and in excluding potentially contaminated samples^38,40 as well as being retrospective.^39–41 Importantly, in our study, infection with non-albicans species was associated with longer hospitalization and ICU stay and a trend towards higher mortality than C. albicans infection, highlighting the importance of rational prescribing to ensure efficacy against non-albicans strains.

Susceptibility to first-line treatment was associated with superior Candida elimination, more effective clinical resolution and a lower mortality rate. However, most patients (62%) received antifungal therapy only after diagnostic confirmation. The resulting 35% clinical remission rate and high proportion of clinical inefficacy (27%) underscore the need for rapid, more accurate identification and earlier empirical treatment of ICI, as well as effective first-line antifungal agents prescribed in accordance with local prevalence and susceptibility patterns.

The limitations of this study include possible underestimation of the ICI incidence, as described above. Because not all isolates were sent to the central laboratory, estimates of Candida species distribution may be inaccurate. Differences in clinical practice patterns across centres probably influenced the diagnosis and management of ICI and may have contributed to the underdetection of cases. Analysis of the clinical outcomes according to the antifungal drug dose and of the characteristics of patients with ineffective response or relapse would have been valuable.

Several studies have demonstrated a relationship between earlier antifungal therapy and improved clinical outcomes,^15,17 particularly in severe sepsis.⁴² Accurate analysis of this relationship was beyond the scope of this study, due to inconsistent diagnostic procedures (timing, procedure, reporting of lab tests, etc.) and timing of antifungal therapy across centres. Based on our findings, we recommend the following: greater focus on early, multisite diagnostic sampling, which should be performed before starting treatment; timely removal of drainage catheters after the first positive sample; and prescribing in accordance with local epidemiological patterns. Older patients and those with solid tumours, organ failure and recent invasive mechanical ventilation should be closely monitored and aggressively treated in cases of suspected ICI.

Conclusions

ICI is associated with high mortality and non-albicans species accounted for the majority of cases in this large, prospective study in ICUs in China. The majority of patients were treated only after diagnostic confirmation of ICI, rather than empirically. In this study, susceptibility to first-line antifungal therapy was associated with lower mortality, highlighting the need for earlier, more accurate diagnostic testing of patients with suspected ICIs and earlier use of antifungals that can reliably eradicate Candida to reduce mortality.

Members of the China-SCAN Team

Bin Zang and Ting Luan, Shengjing Hospital, Affiliated to China Medical University; Yan Kang and Yan Zhou, West China Hospital, Sichuan University; Wei Cui and Ana Li, The Second Affiliated Hospital of Zhejiang University School of Medicine; Bingyu Qin, Huangzhang Shao, Henan Provincial People's Hospital; Yingzhi Qin and Xinjing Gao, Tianjin Third Central Hospital; Tiehe Qin and Shouhong Wang, Guangdong General Hospital; Qiang Fang and Xia Zheng, The First Affiliated Hospital of Medical School of Zhejiang University; Dongpo Jiang and Shanmu Ai, Daping Hospital; Weiqin Li and Rao Nie, Jinling Hospital; Qin Gu and Ming Chen, Drum Tower Hospital, Affiliated to Nanjing University Medicine School; Hongsheng Zhao and Shu Lu, Affiliated Hospital of Nantong University; Haibo Qiu, Fengmei Guo and Yi Yang, Zhongda Hospital, Southeast University; Jiandong Lin and Guo Lin, The First Affiliated Hospital of Fujian Medical University; Duming Zhu and Yijun Zheng, Zhongshan Hospital, Fudan University; Jian Liu and Bing Li, The First Affiliated Hospital of Lanzhou University; Rongqing Sun and Yuexia Li, The First Affiliated Hospital of Zhengzhou University; Difen Wang, Xianquan Liang, Yuanyi Liu and Fang Wu, Affiliated Hospital of Guiyang Medical College—Surgical/Emergency ICU; Maoqin Li and Yanjun Xu, Xuzhou Central Hospital; Dawei Liu and Wei Tu, Peking Union Medical College Hospital; Jianhuan Zhu and Jihui Ye, Ningbo First Hospital; Lijun Liu and Shiquan Yu, The Second Affiliated Hospital of Soochow University; Xiaojun Yang and Xigang Ma, General Hospital of Ningxia Medical University; Ruiqiang Zheng and Jiangquan Yu, Jiangsu Subei People's Hospital; Mingyan Zhao and Songlin Yang, The First Affiliated Hospital of Harbin Medical University; Jianguo Li and Bo Hu, Zhongnan Hospital of Wuhan University; Xiangdong Guan and Lifen Li, The First Affiliated Hospital, SUN YAT-SEN University; Dawei Wu and Binfeng Du, Qilu Hospital of Shandong University; Chuanyun Qian and Rong Liu, First Affiliated Hospital of Kunming Medical University; Jun Jin, Ying Zhan and Xinjing Yang, The First Affiliated Hospital of Soochow University—Emergency/Central ICU; Jianxin Zhou and Qiang Wang, Beijing Tiantan Hospital Affiliated to Capital Medical University; Yuan Xu and Hua Zhou, Beijing Tongren Hospital, Capital Medical University; Xiaochun Ma and Renyu Ding, The First Hospital of China Medical University; Xiaoqin Zhao, Zhanhong Tang, Yongyi Meng and Yiping Pan, The First Affiliated Hospital of Guangxi Medical University—Internal/Surgery; Xifang Lin and Jie Chen, The First Affiliated Hospital of Wenzhou Medical College; Xue Wang and Qindong Shi, First Affiliated Hospital of Medical College of Xi'an Jiaotong University; Qing Song and Hongjun Kang, Chinese PLA General Hospital; Kaijiang Yu and Junbo Zheng, The Second Affiliated Hospital of Harbin Medical University; Weixing Zhang and Lei Huang, Peking University, Shengzhen Hospital; Dechang Chen and Hu Wang, Shanghai Changzheng Hospital; Yuhang Ai and Zhiyong Liu, Xiangya Hospital Central-South University; Quan Cao, Suming Zhou, Shaohua Liu and Jing Zhou, The First Affiliated Hospital with Nanjing Medical University; Xianyao Wan and Jiuzhi Zhang, First Affiliated Hospital of Dalian Medical University; Zhenjie Hu and Guijun Zhu, Hebei Medical University Fourth Hospital; Yongqiang Wang and Jianfu Li, Tianjin First Center Hospital; Yimin Li and Xiaoqing Liu, The First Affiliated Hospital of Guangzhou Medical University; Zhenyang He and Xiaohong Xie, Hainan Provincial People's Hospital; Jinyan Xing and Xinghua Yang, The Affiliated Hospital of Medical College Qingdao University; Jie Yan and Hongyan Xu, Wuxi People's Hospital; Xiangyou Yu and Hua Zhong, The First Teaching Hospital of Xinjiang Medical University; Xinwei Mu and Yongming Chen, Nanjing First Hospital; Yongjie Yin and Haibo Liu, The Second Hospital of Jilin University; Shiying Yuan and Zhaohui Fu, Wuhan Union Hospital, Tongji Medical College of Huazhong University of Science and Technology; Shusheng Li and Wei Zhu, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology; Jing Yan and Jin Chen, Zhejiang Hospital; Chunting Wang and Jinjiao Jiang, Shandong Provincial Hospital; Yaoqing Tang and Lei Li, Rui Jin Hospital, Shanghai Jiao Tong University School of Medicine; Jianqiong Xiong and Lei Zhang, Southwest Hospital; Xiujiang Li and Xingang Liu, The First Hospital of Jilin University; Lixin Zhou and Tieou Yu, First People's Hospital of Foshan; Hang Xu, The First Affiliated Hospital of Medical College, Shihezi University; Guangning Tang, Shihezi People's Hospital; Siqing Ma, Qinghai Provincial People's Hospital; Fachun Zhou and Dong Wan, The First Affiliated Hospital of Chongqing Medical University; Ruoyu Li, Wei Liu and Zhe Wan, Research Center for Medical Mycology, Peking University First Hospital, Peking University.

Funding

This work was supported by Merck Sharp & Dohme China, who sponsored the study. Medical writing and editing assistance were supported by Merck & Co., Inc., Whitehouse Station, NJ, USA.

Transparency declarations

F.G. has received grant funding from Pfizer. Y.Y. has received honoraria/speaking fees from Pfizer and Merck Sharp & Dohme China. H.Q. is a speaker for Pfizer and Merck Sharp & Dohme China, and has received research grants from Pfizer, Merck Sharp & Dohme China and Xian-Janssen. The remaining authors have no conflicts of interest to disclose.

The funders participated in the design of the study, but had no role in study management, monitoring, data management, statistical analysis or development of this article.

We accept direct responsibility for this paper.

Medical writing and editing assistance were provided by Esther Nathanson and Liz Bullock of Watermeadow Medical.

Author contributions

H.Q., F.G., Y.Y., Q.F., D.L., X.G., K.Y., Y.T. and R.L. designed the study. F.G., H.Q. and Y.Y. contributed to the manuscript writing. All authors were involved in patient recruitment, served as study investigators at their hospital site and reviewed the final manuscript.

Acknowledgements

Part of this work was presented as a poster at the Thirty-first International Symposium of Intensive Care and Emergency Medicine, Brussels, 2011 (Poster no. P239).

We are grateful to Esther Nathanson and Liz Bullock, of Watermeadow Medical, for developing a first draft from an agreed outline, collating comments and providing editorial assistance.

We would like to thank the patients and investigators who participated in this study. We also acknowledge the investigators at each study site, without whom this study would not have been possible (see below; names in italics are investigators not listed in the author list).

Author notes

†

Members of the China-SCAN Team are listed in the Acknowledgements section.