Persistent vs non‐persistent candidaemia in adult patients in 2007‐2016: A retrospective cohort study

Persistent candidaemia (PC) is a recognised complication of candidaemia. Our objective was to evaluate risk factors and clinical significance of PC in adult patients.

include the host's baseline characteristics (eg cancer, immunosuppressive medication, neutropenia), antifungal resistance, efficacy of therapeutic management and source control. However, the literature on PC is scarce. The most relevant problem in evaluating PC is the lack of a homogeneous definition of PC. 7 Most studies evaluating PC include adult and child populations or have been conducted only in neonatal populations. 2,[8][9][10][11] This complicates any generalisation of the results to adult patients. The definition of PC ranges from >1 day to 7 days in literature, which influences PC incidence rates. 7 The reported incidence rates of PC vary from 8% to 93%. 7 However, the reported rates are at 11%-37% when considering only studies with a definition of candidaemia duration of 3-5 days. [12][13][14][15] Although several studies have evaluated PC, the data are mostly from research designed to consider other elements than PC. [16][17][18] The aim of this study was to evaluate the underlying risk factors and clinical significance of PC, defined as blood culture positivity persisting for 5 days or longer. We compared adult patients with PC with non-persistent cases in a retrospective analysis from Southern Finland during 2007-2016.

| Study design
An observational, retrospective cohort study including adult patients with candidaemia was conducted in the hospital district Details of microbiological identification and susceptibility testing are presented in an earlier publication of this study. 19

| Definitions
An episode of candidaemia was defined as at least one blood culture positive for Candida species. If the patient had more than one episode of candidaemia, only the first episode was included in the study. Recurrent episodes were excluded (n = 24). PC was defined as an isolation of the same Candida species from any positive blood culture taken ≥5 days after the first positive blood samples were drawn. Non-PC was defined as all other candidaemia cases than persistent ones with negative blood cultures taken at least once.

| Statistical analysis
Categorical variables were summarised as counts and percentages and compared using the chi-square or Fisher's exact test, as appropriate. Continuous variables with a non-normal distribution were described as median and interquartile range (IQR) and compared using the Mann-Whitney U test. P-values ≤ .05 were considered to indicate statistical significance. Odds ratios (ORs) were calculated with 95% confidence interval (CI). Risk factors associated with PC were computed in a multivariable logistic regression analysis. The regression analysis included variables having univariate P-value < .1 and was not multicollinear. SPSS version 25 (SPSS Inc) was used for all analyses.  Table 1. Non-albicans Candida spp. caused 37.3% of PC and 35.1% of non-persistent cases without a significant difference (P = .742). However, candidaemia caused by C dubliniensis was more common among non-persistent than PC cases (PC 1.3% vs non-PC 7.9%), but the difference did not reach statistical significance (P = .065). C albicans was the leading cause (62.7%) of PC, followed by C glabrata (17.3%) and C parapsilosis (5.3%). Resistance rate for fluconazole was overall 21.3% for PC and 13.9% for non-persistent cases (Table 1). Fluconazole resistance was rare (2.2%) for Candida spp. other than C glabrata, but the rate was high among PC (76.9%) and non-persistent cases (80.0%) caused by C glabrata.

| RE SULTS
Anidulafungin susceptibility testing has been performed routinely since 2011 in our hospital district and earlier for only selected cases.
Resistance to anidulafungin was rare, seen in only 4/168 cases (2.4%). Amphotericin B resistance was also rare; a total of 2/225 cases (0.9%) were resistant to amphotericin B.
The baseline characteristics of PC vs non-persistent cases are displayed in Table 2. The median age for persistent cases was   However, no significant difference in guideline adherence was present between PC and non-PC cases (median score for PC 13.0 vs for non-PC 12.0, P = .068).
Using multivariable regression analysis (

| D ISCUSS I ON
We investigated the risk factors and clinical significance of PC in adult patients during a 10-year study period. Persistent candidaemia was found in 75 cases during the study period. Factors independently associated with PC were the presence of CVC at the onset of candidaemia, metastatic infection foci and ineffective empirical antifungal treatment. Catheter-related infection was the most common source of infection in persistent cases. CVC has been demonstrated as a risk factor for PC in other studies as well. 11,23-25 However, a recent cohort study conducted in adult patients did not find a significant association between CVC andPC. 12 In their study, CVC was a very common finding overall in both PC and non-PC groups (80% vs 72%), but the difference did not reach statistical significance.
The presence of CVC is a prominent risk factor for PC; however, the timing of CVC removal is complex. The association between early removal of CVC and PC was not significant in our study, supported by another recent study. 24 Nucci et al 26 showed that early CVC removal had no effect on persistent or recurrent candidaemia or time to mycological eradication, although earlier studies have reported an association of improved outcome and faster mycological eradication with prompt CVC removal. 27,28 Metastatic infection foci were more prominent in PC than in non-PC cases. Similarly, unexpected additional infection sites were associated with PC also in an earlier analysis. 12 This supports the assumption that an active search for metastatic infection foci is crucial during the treatment of PC. Early source control was a protective factor against PC in a subgroup analysis in our study.
Fluconazole resistance among C glabrata isolates was high in both PC and non-PC, but no significant difference emerged between the groups. Otherwise, resistance to antifungals was rare.
Ineffective empirical antifungal treatment was associated with PC in our study. Ineffective antifungal was started before the results from blood cultures were available, but all of these patients who received ineffective antifungal initially received appropriate antifungal after the culture results were disclosed. Relatively many patients (22% of PC and 11% of non-PC cases) initially received an ineffective agent, even though the median duration of the ineffective therapy was short. Half of the inappropriate treatment was prescribed at surgical departments. Treatment with fluconazole or echinocandin as an initial antifungal showed no significant difference between persistent and non-persistent cases.
Neonatal analyses have demonstrated an association between PC and non-albicans Candida species. 9,11 Our results reflected no such association in an adult population, consistent with the findings of other studies conducted in adult populations. 12,23 This indicates an overall difference in species distribution between adult and child populations. C parapsilosis is a more common cause of candidaemia in children than in adults, 29 in non-PC in our study, but the difference did not reach statistical significance. Surprisingly, the overall 30-day mortality rate was lower in PC than overall in our hospital district or in general in candidaemia. [2][3][4][5] Candidaemia is associated with severe underlying comorbidities, for example malignancies. 36 Sometimes, it is justified to focus on symptomatic treatment and refrain from control blood cultures if the patient is in a terminal phase. Patients who died within five days or had no negative blood cultures taken were excluded from our analysis. When we excluded these patients, it decreased the mortality rate in PC and non-PC groups.
Adherence to guidelines was evaluated with the Equal Candida Score. The score showed an association with mortality, but not with PC. A recent study from Taiwan reported a significant association with survival outcome as well. 37 However, other previous studies have not found a significant difference, 38-40 despite survivors having higher Equal Candida Scores than non-survivors. The patient number was smaller in these studies. The median score was 13.0 in our study.
There is still a long way to go to achieve an excellent rating according to ESCMID and IDSA guidelines.
Our study has several potential limitations. The most significant one is the retrospective nature of the analysis. Follow-up blood cultures are strongly recommended, but there were no predesigned routines for follow-up blood cultures during the study period.
Monitoring infection clearance from blood cultures is often neglected if a patient responds very well to treatment or if the patient is terminally ill. However, this study provides reliable data over a 10- year study period for adult patients.
In conclusion, the presence of CVC, metastatic infection foci and ineffective empirical antifungal were independent risk factors for PC. Active search for and treatment of metastatic infection foci, and removal of CVC are the key elements for improving patient care and preventing PC.

CO N FLI C T O F I NTE R E S T
MAH has received lecture fee from MSD and Gilead and conference invitations from Pfizer and MSD. VJA has received lecture fees from Pfizer, MSD, Astellas, Unimedic, Roche, BMS, Biogen and Gilead; has participated as PI to Varicella zoster vaccination studies; and has received a study grant for a pneumococcal vaccination study.

AUTH O R S ' CO NTR I B UTI O N
MAH and VJA conceived the study design. MAH has collected and analysed the data. Both authors drafted the manuscript and approved the final version of the manuscript.