This was a comprehensively retrospective study including 183 invasive candidiasis cases according to the latest diagnostic guideline by EORTC/MSG . Descriptive analysis, chi-squared tests, Mann-Whitney U tests and t-tests were used to describe the characteristics of the included patients and Candida species, antifungal susceptibility, status of diagnostic tests, treatment and prognosis. Logistic regression models were employed to investigate the association between risk factors/ treatment and prognosis.
The incidence of IC in this study (0.26‰) was relatively lower compared to the incidence in previous studies based on hospital settings, which was reported to be 1.9-2.4‰ in the US, 1.18‰ in Latin America, 0.92-1.19‰ in Europe and 0.13-1.22‰ in Asia [7, 12]. In addition, though the incidence of IC was increasing in several centers owing to the increasingly intensive care such as long-term administration of broad-spectrum antibiotics, the ten-year period in this study witnessed an overall decrease in the IC incidence. These might be due to the increasing awareness in this disease and the effective prevention of nosocomial infection. Several studies in Asia also reported a decrease in healthcare-associated infection including nosocomial candidemia owing to sequential multifaceted interventions such as hand hygiene, care bundles, and antimicrobial stewardships in the recent years [6, 13, 21]. In addition, 43.2% (79/183) of the episodes in this study occurred in ICU settings, similar to the global statistics, which was approximately 44.5-50.0% [1, 22-24].
Since the proven diagnose of IC based on culture takes a median of 2-3 days and has low sensitivity of 21-71% , timely empirical therapy is often required to reduce mortality. In order to predict the possible infected Candida species and timely prescribe empirical antifungal drugs targeting the suspected species, grasping the geographical distribution and clinical characteristics of each Candida species is essential. In this study, C.albicans was the most common species (45.8%) of the isolates, followed by C.parapsilosis (19.5%). This is consistent with many studies in China, including the multicenter China-SCAN study and CHIF-NET study, which all reported C.parapsilosis (19.8-27.8%) to be the second common species following C.albicans [8, 11, 13, 25, 26]. There were also single-center studies in China reporting C.tropicalis (18.0%) and C. glabrata (43.0%) as the most popular C.non-albicans [6, 7]. Grasping the local prevalence is essential as the Candida spp. distribution varies from center to center. In terms of clinical prevalence, this study found that C. non-albicans was more often to cause bloodstream infection than infections in other deep sites, which was consistent with previous findings [25, 27]. Specifically, the present study and several previous studies all found that C. parapsilosis had a preference in younger patients without neutropenia or immunosuppressed condition, while C.glabrata and C. krusei had a preference in aged patients with poorer conditions such as severe anemia, renal disease and ventilator dependence [4, 27]. C.tropicalis and C. krusei were additionally found to be associated with hematologic malignancies and immunosuppression in this study and the study of Horn et al.. Previous exposure to antifungal drugs may also influence the infected species. Prior fluconazole therapy was found to be a risk factor for infection with fluconazole non-susceptible C.non-albicans . This study further found an association between C.tropicalis and prior exposure to voriconazole, and between C.glabrata and prior exposure to fluconazole. C. krusei candidemia was also reported to be associated with prior use of antifungal agents . All this information is essential, including local Candida spp. prevalence distribution, the infected sites and medical conditions of the patients, since they can provide clues for the infected Candida species and help clinicians to make targeted prescription timely.
In terms of risk factors, abnormal colonization is regarded as the prerequisite of the subsequent invasive candidiasis. 70.3% of the patients in this study had Candida colonization before IC onset. The median time to develop IC after colonization was 13.5 days in this study and reported to be 7-25 days in non-neutropenic ICU patients [16, 28]. Thus, additional attention should be paid to those with abnormal Candida colonization, especially to those with colonization in more than two sites, heavy colonization in more than one site, and colonization in the urine, who have significantly higher risk of IC . Interestingly, in this study, 23.0% of the patients developed IC by a Candida species different from the colonized Candida species and 24.3% colonized with multi-Candida species. This phenomenon might be explained as the colonization with different Candida species contributed to the burden of fungi, which eventually caused symptomatic infections . Thus, when choosing antifungal drugs for the suspected infected species in the empirical treatment, the clinicians should be aware that the colonized Candida species may be different from the ones causing invasive infection. In addition, as Candida is a common symbiotic yeast, the positive predictive value of colonization for IC is low (2-4%), whereas the negative predictive value is high (99-100%) . Dynamic monitoring of colonization can be laboratory intensive and significantly increase the economic burden of patients . In this circumstance, Candida Score is an alternative to predict the risk of IC development, as patients with a Candida Score≥3 are prone to IC (sensitivity 81%, specificity 74%) and would benefit from empirical treatment [17, 18]. This study also found that patients with higher Candida Score had increased risk of mortality, thus should be paid more attention.
Culture from blood or sterile sties is currently the gold standard for proven diagnosis. Though a successful recovery of Candida spp. might take 2-3 days , detecting fungal growth in blood sample culture just took a median of 28 hours in this study. Still, empirical treatment should be delivered when waiting for the results of culture if IC is suspected, since timely antifungal therapy within 24 hours can significantly reduce the mortality of candidemia .
Serum 1,3-β-D-glucan (BDG) test is another widely used test to facilitate diagnosis. β-D-glucan is a pan-marker for various fungal infections. Previous studies have demonstrated its high sensitivity (64.8%-89% at a threshold of 70-80pg/ml) yet low specificity (56.7%-60% at a threshold of 80pg/ml) in the IC diagnosis [30-32]. Though its sensitivity and specificity may vary due to the difference in the threshold chosen and the performance of the diagnostic kits, the results in this study was still disappointing. At the threshold of 100pg/ml, only 30% of the cases were tested positive for BDG within seven days of disease onset. Further investigation during the follow-up showed 29.7% of the false negative results were due to insufficient increase compared to the threshold, and 14.8% were due to the lag in increase. Some studies previously suggested to use the negative results of BDG to exclude IC diagnosis and allow discontinuation of the empirical therapy [30, 33, 34]. However, this might delay the preemptive therapy as BDG in this study reacted slowly (averagely 3 days later) and rose insufficiently, which might cause false-negativity. Using the decrease in BDG levels to evaluate the treatment response, as suggested in previous studies , is also unreliable, since many factors can cause false positivity in BDG, such as albumin administration . Thus, the results in this study reminded clinicians that the result of BDG, either positive or negative, should not be overly relied on, and the diagnostic culture should be combined to determine the diagnosis. Despite these limitations, ten patients in this study received preemptive therapy according to BDG results, indicating BDG could still respond rapidly and guide timely preemptive therapy in certain circumstances. To further improve diagnostic value, many other markers have been suggested to be combined with BDG, including traditional inflammatory markers (e.g. hsCRP and platelet count)  and other fungal antigen (mannan or anti-mannan immunoglobulin M) .
In this study, the antifungal resistance/non-WT mainly existed in fluconazole (14.4%) and voriconazole (9.6%), similar to the worldwide data, in which fluconazole resistance in all Candida spp. ranged from 8.1% to 39.3%, followed by voriconazole ranging from 5.4% to 14.0%, Echinocandins ranging from 1.4% to 2.1% and Amphotericin B ranging from 0% to 1% [6, 7, 11, 25, 26]. Specifically, fluconazole resistance was prevalent in C.tropicalis (37.0%) and C.glabrata (24.1%) in this study, relatively more common than those reported in previous Chinese studies (3.1-11.6% for C.tropicalis and 12.0-18.5% for C.glabrata, respectively) [6, 11, 13, 14, 25]. Voriconazole non-WT C.glabrata isolates in this study (58.6%) was also more common compared to previous studies (0-17.8%) [2, 4, 11, 14, 25, 37, 38]. This might indicate a rise in the resistance to azoles in C.glabrata and C.tropicalis in the recent years, which was also observed in the previous studies [14, 25]. These findings indicate that clinicians should be cautious to use fluconazole/ voriconazole when C.glabrata and C.tropicalis are suspected to be the causative agents. Though the overall resistance to Echinocandins is uncommon, C.glabrata (10.3%) and C.tropicalis (7.4%) could develop resistance to Echinocandins in this study. Furthermore, this study found that previous exposure to fluconazole/ voriconazole could mutually increase the risk of developing resistance to voriconazole/ fluconazole. Exposure to caspofungin could also increase the risk of micafungin resistance. Further tests also confirmed the existence of cross-resistance between fluconazole and voriconazole, and between caspofungin and micafungin. All of these indicate the need of performing a drug-susceptibility test no matter the type of antifungal drug, azoles or Echinocandins, is used, especially when the patient has previous antifungal drug exposure. Amphotericin B is an alternative when multi-resistance is evident, since this study together with most previous studies reported non-resistance to Amphotericin B in all the five common Candida species [6, 11, 14, 26].
In terms of treatment, apart from the systematic antifungal therapy, there are two essential interventions contributing to an effective infection control: source control and timely administration of systematic antifungal therapy. In the present study, 16 candidemia patients did not receive any systematic antifungal treatment. Five of them recovered after changing CVC or hemodialysis tubes. One acute abdomen induced candidemic patient recovered after surgery. Eight non-candidemic IC patients recovered after surgery or drainage without systematic antifungal therapy. All of these cases demonstrated the importance of effective source control. Previous studies also demonstrated that failure or delay in source control significantly increased the mortality risk (OR: 6.78-77.40) [10, 26]. Timely systematic antifungal drug administration is also essential which can significantly reduce mortality . Empirical treatment was strongly recommended in ICU hospitalized patients suspected with IC . In this study, more than half of the patients received antifungal treatment within 24 hours, which might contribute to the lower mortality. However, we failed to detect any significant association between prophylaxis/ empirical treatment and better survival. This might be due to insufficient dosage administrated in the prophylaxis/ empirical treatment. 53.5% (8/15) of those who received prophylaxis initiated with fluconazole/ voriconazole ≤200mg/d, while the recommended dosage was 400mg/d . Also, 61.9% (13/21) received empirical therapy with fluconazole/ voriconazole less than the recommended 400mg/d.
Currently, Echinocandins are recommended as the first-line therapy in both empirical and targeted therapy of IC, and fluconazole is an alternative only recommended in patients with hemodynamic stability and without fluconazole resistance [1, 39]. However, in this study, the majority of patients received fluconazole as initial therapy and main therapy, even though 28.0% (59/211) of the isolates had exposure to fluconazole before isolation. After adjustment in the multivariable regression models, initiating antifungal therapy with fluconazole also had a better survival outcome compared to micafungin. This might be due to the small proportion of C.krusei and C.glabrata infections included in this study, and the relatively low resistance/non-WT in the isolates. This indicated that fluconazole could still be a cost-effective choice as long as the resistance was not evident.
The attributable mortality rate of IC could exceed 70% while the majority reported it to be 10%-38% [1, 3-6]. Regardless of the evidence of statistical significance, most studies, including the present study, reported lower mortality rates in C.parapsilosis [1, 27] and higher mortality rates in C.glabrata [6, 13, 27]. Particularly, one C.parapsilosis infected patient enrolled in this study presented with three episodes of IC with the longest duration of 143 days, indicating the relatively weak virulence of C.parapsilosis. In terms of baseline medical condition, the majority of studies reported advanced age, neutropenia, severe sepsis, respiratory and renal failure, and iatrogenic factors including ICU hospitalization, mechanical ventilation and dialysis were associated with increased mortality [3, 8, 13, 26, 38], while appropriate empiric antifungal therapy administrated within five days and proven catheter-related candidemia were found to be protective factors against early mortality . In this study, we also confirmed the CVC induced IC to be a protective factor, and further found hematologic malignancies, digestive tract perforation, heart failure, long-term administration of broad-spectrum antibiotics and TPN to be risk factors for increased mortality. Due to these differences in the baseline conditions of the included population, infected sites and Candida species, time to initiate the therapy and the choice of antifungal drugs, the specific mortality in each study differed a lot. The crude and attributable 30-day mortality rate in this study was 24.5% and 18.9% respectively, lower than most previous studies. This might be due to the large proportion of C.parapsilosis infections, more children patients included, and timely systematic antifungal drug administration in this study. In addition, the median time to death after diagnosis was reported to be 14-15 days [3, 13], which was consistent with this study.
There were several limitations in this study. First, as a retrospective study, many tests investigated were not conducted or monitored regularly, including colonization, 1-3-β-D-glucan and culture re-examination which might influence the determination of infective condition and bias the analytic results. The mortality outcomes in 13.1% (24/183) of the patients were missing which might underestimate the mortality rate as many patients with missing outcomes requested a discharge regardless of a poor prognosis. Second, due to data availability, only IC patients were included. The lack of a control group without invasive candidiasis with similar demographic characteristics limited the ability of this study to explore the true risk factors of developing IC. Third, as a single-center study, the collected data was limited even though the ten-year data was gathered. The small sample size might reduce the power of this study to detect a significance when conducting statistical analysis. Thus, a large-scale multi-center prospective study is required to further investigate the characteristics of invasive candidiasis.