Declining Incidence of Candidemia and the Shifting Epidemiology of Candida Resistance in Two US Metropolitan Areas, 2008–2013: Results from Population-Based Surveillance

Background Recent reports have demonstrated a decline in bacterial bloodstream infections (BSIs) following adherence to central line insertion practices; however, declines have been less evident for BSIs due to Candida species. Methods We conducted active, population-based laboratory surveillance for candidemia in metropolitan Atlanta, GA and Baltimore, MD over a 5-year period. We calculated annual candidemia incidence and antifungal drug resistance rates. Results We identified 3,848 candidemia cases from 2008–2013. Compared with 2008, candidemia incidence per 100,000 person-years decreased significantly by 2013 in both locations (GA: 14.1 to 9.5, p<0.001; MD: 30.9 to 14.4, p<0.001). A total of 3,255 cases (85%) had a central venous catheter (CVC) in place within 2 days before the BSI culture date. In both locations, the number of CVC-associated cases declined (GA: 473 to 294; MD: 384 to 151). Candida albicans (CA, 36%) and Candida glabrata (CG, 27%) were the most common species recovered. In both locations, the proportion of cases with fluconazole resistance decreased (GA: 8.0% to 7.1%, −10%; MD: 6.6% to 4.9%, −25%), while the proportion of cases with an isolate resistant to an echinocandin increased (GA: 1.2% to 2.9%, +147%; MD: 2.0% to 3.5%, +77%). Most (74%) echinocandin-resistant isolates were CG; 17 (<1%) isolates were resistant to both drug categories (multidrug resistant [MDR], 16/17 were CG). The proportion of CG cases with MDR Candida increased from 1.8% to 2.6%. Conclusions We observed a significant decline in the incidence of candidemia over a five-year period, and increases in echinocandin-resistant and MDR Candida. Efforts to strengthen infection control practices may be preventing candidemia among high-risk patients. Further surveillance for resistant Candida is warranted.


Introduction
Many healthcare-associated infections (HAIs), including some types of bloodstream infections (BSIs), are a preventable cause of morbidity and mortality in healthcare facilities. In an effort to reduce HAIs, the adoption of new healthcare policies, such as those tied to Medicare reimbursement, have incentivized reducing infection rates and produced mixed results [1][2][3][4][5][6].
BSIs caused by Candida species, also known as candidemia, remain an important public health problem [7][8][9]. Recent reports have documented significant declines in some bacterial central line associated bloodstream infections (CLABSIs) following adherence to established central line insertion practices, however, similar declines have been less evident for CLABSIs due to Candida species [10][11][12][13][14]. Candida spp. are common gastrointestinal flora and while many Candida BSIs can be attributed to the presence of a central line, some Candida spp. are also hypothesized to cause infections by translocation across the gastrointestinal tract unrelated to central line insertion practices [15,16]. Therefore, line insertion bundles developed to reduce CLABSIs alone may not be completely effective in reducing Candida BSIs.
In 2012, we reported a substantial increase over a twenty-year period in the overall population-based incidence of candidemia using population-based surveillance for candidemia in two U.S. locations [9,17]. This surveillance system captures all culture-confirmed BSIs caused by Candida spp., not just healthcare-associated CLABSIs that are tied to mandatory reporting and reimbursement policies, and provide a more complete picture of the burden of candidemia in these communities. The earlier report also noted a marked shift in the species distribution among causative organisms with a significant increase in Candida glabrata (CG), a species more likely to be resistant to azoles, the standard antifungal drug of choice.
To monitor more recent changes in the incidence of candidemia and antifungal drug resistance, we now report trends over a five-year surveillance period using population-based prospective surveillance for candidemia in metropolitan Atlanta, Georgia and Baltimore City and County, Maryland.

Surveillance population
Surveillance for candidemia was conducted among residents of Atlanta, Georgia (Fulton, DeKalb, Cobb, Gwinnett, Clayton, Douglas, Newton, and Rockdale counties; 25 hospitals, population: 3.8 million) and Baltimore City and County, Maryland (15 hospitals, population: 1. surveillance area at least 30 days apart from any other blood culture positive for Candida spp. Cases were classified as: (1) hospital-onset (HO) if the first culture was obtained >3 days after admission (with admission being day 1), (2) healthcare-associated community-onset (HACO) if the culture was obtained 3 days after admission in a patient with recent healthcare exposure, or (3) community-associated (CA) if the culture was obtained as an outpatient or 3 days of admission in a patient without documentation of recent healthcare exposure. Basic demographic and clinical information was collected on all cases. Surveillance personnel used standardized case report forms to abstract data from medical records. Laboratory records from all participating laboratories were audited monthly.
Isolate collection, identification, and antifungal susceptibility testing All available isolates were sent to CDC for species confirmation and antifungal drug susceptibility testing; testing methods have been previously described [18]. CDC-confirmed species are reported; if no isolate was received at CDC, local laboratory species identifications are reported. Because in vitro susceptibility testing may demonstrate susceptibility to one echinocandin but resistance to another that is of questionable clinical correlation [19], molecular testing was used to confirm that the majority of the CG isolates with in vitro echinocandin resistance had a mutation that has been associated with clinical failure [20].

Statistical methods and denominators
Incidence rates were calculated using year specific population estimates for Baltimore [21] and Atlanta [22] and are presented per 100,000 person-years. Age-adjusted incidence rates did not vary substantially over the five year surveillance period from crude rates, and thus we report crude annual incidence rates of candidemia.
Categorical variables were analyzed using chi-square tests or Fisher's exact tests. We used a Poisson regression model adjusting for age group to evaluate the change in overall and species specific annual incidence rates over the 5 year surveillance period. In all analyses, the level of significance was set at α = 0.05. All analyses were done using SAS software (version 9.3, SAS Institute, Inc., Cary, NC).

Human subjects
Patient data were de-identified at each site prior to access and analysis by CDC. CDC conducted ethical review of this surveillance project and deemed it a non-research activity. This activity was also evaluated individually at each location, and was either deemed a public health assessment or human subjects research, and approved by local review boards. In the Baltimore area, the Maryland Department of Health and Mental Hygiene's Institutional Review Board reviewed and approved the protocol, and the protocol was determined exempt from review by the Johns Hopkins Bloomberg School of Public Health Institutional Review Board. In the Atlanta area, the protocol was reviewed and approved by the Emory University Institutional Review Board, and the protocol was determined exempt from review by the Georgia Department of Public Health Review Board.

Hospital-specific Declines
Among 40 hospitals under surveillance, 85% reported either a decrease or no change in the frequency of cases over the surveillance period (Fig. 4). Additionally, of the eight hospitals that contributed at least 100 cases during the 5 years of surveillance, there was a substantial drop in the number of cases identified: in ATL 319 cases were identified in the first year, and 190 were identified in the last year (−40%); in BTM, 282 cases were identified in the first year and 125 were identified in the last year (−56%).

Discussion
This report from the largest U.S. population-based surveillance program for candidemia to date describes a significant decline in the overall incidence of candidemia in two metropolitan regions. This decline was seen in almost all age groups, was primarily among patients with healthcare exposure, and was especially notable among cases with a central venous catheter. Importantly, we also report an increase in echinocandin resistance as well as the emergence of multi-drug resistant Candida BSIs. In 2012, we reported an increase in candidemia incidence over a twenty year period, as seen through two periodic measurements using population-based surveillance [9], that was attributed to multifactorial changes in patient populations, including possible increases in patient populations at high risk for candidemia.
In comparison, our current report documents a subsequent decrease in incidence that is not likely to be due to changes in high-risk populations; rather, it is probable that the declines noted here are related to healthcare delivery improvements. We found that declines were predominantly among candidemia cases with healthcare exposure, but not among the small number of community-associated cases, suggesting that factors associated with healthcare delivery may be driving this decline. Our data further demonstrate declines among patients with Candida BSIs occurring in the presence of CVCs, while the number of cases without CVCs did not decline as substantially. Taken together, these data suggest that the large declines noted here may be the result of policies and practices related to catheter insertion and maintenance.
Recent policies targeted at reducing healthcare-associated infections, including state mandates for public reporting, have incentivized reducing CLABSIs [1]. Subsequent reports including data from CDC's National Health Safety Network (NHSN) have described large declines in CLABSI rates [13,23,24]. Data from our active surveillance system, which is not tied to any reimbursement or reporting mechanism, support the trends observed through systems such as NHSN and suggest these declines are robust. Anecdotally, most hospitals in our surveillance area have recently introduced policies that mandate improvements in catheter care (data not shown); further study is needed to understand if declines can be attributed to these policies.
Notably, while we documented a drop in fluconazole resistance, we also report a small but concerning increase in isolates resistant to echinocandins, and the emergence of multidrug resistant (MDR) Candida, almost all of which were the species CG. Although the decrease in fluconazole resistance is reassuring, it may be partially due to the increased use of echinocandins as primary therapy for candidemia. This shift in practice, particularly for CG infections, may be contributing to the increase in echinocandin resistance; recent reports have demonstrated the emergence of echinocandin resistant CG [17,25,26], possibly driven by wider usage of this drug class as primary therapy [27] and prophylaxis. Although the modes of action and target sites of the triazoles (i.e. fluconazole) and echinocandins are different [28], CG are the most common fluconazole-resistant species, and mutations conferring echinocandin resistance in CG can emerge quickly [29]. Microbiology laboratories should consider antifungal susceptibility testing for CG isolates, since detection of echinocandin-resistant or MDR Candida can influence therapy.
It is also concerning that 44% of MDR Candida occurred in patients without recent exposure to echinocandins. This observation suggests person-to-person transmission of this resistant phenotype is possible, as has been previously suggested [30]. Vigilant monitoring for resistance among CG species will continue to be critical, and further investigation into the possible transmission of MDR Candida will be essential for targeting prevention efforts.
This report is subject to several limitations. Although we suspect that the dramatic declines in candidemia occurring in patients with a CVC were due to improved catheter care, it is possible that declines were due to some external factor such as reductions in overall CVC usage in a particular ward. We report large declines in most hospitals under surveillance, but do not collect data on number of hospital admissions or other hospital denominator data, and are therefore unable to report if the declines are a true decrease in the risk of disease or if this is related to a shift in healthcare utilization (e.g., change in central line utilization, referral patterns, frequency of blood culture collection, etc). However, the majority of hospitals in our surveillance system demonstrated declines indicating a trend not isolated to a few institutions.
This report describes significant declines in the incidence of candidemia bloodstream infections in two major U.S. locations over a 5 year period, and an increase in echinocandin resistant CG. Continued surveillance will be important to help understand factors contributing to these declines and to monitor for the emergence of resistant Candida.