Risk factors for persistent candidemia and prognostic implications: Results from the ECMM Candida III study

Candidemia is a severe complication in critically ill and immunocompromised patients, and is associated with high morbidity and mortality. While early fungal clearance may improve outcomes, the association between follow-up blood culture (FUBCs) results and clinical outcomes remains insufficiently explored. This sub-analysis of the ECMM Candida III study investigates predictors of persistent candidemia and the impact of positive FUBC results on clinical outcomes.

The multicenter ECMM Candida III study enrolled adults with culture-proven candidemia from 60 European centers (2018-2019). This sub-analysis included patients with at least one FUBC result reported (n = 258; 40.8%). Statistical analysis used SPSS 29 and R. Binary logistic regression was used to identify predictors of persistent candidemia. To assess mortality risk factors, Cox proportional hazards regression models were constructed.

Of 258 patients, 52 (20.2%) had persistent candidemia based on positive FUBCs (median duration of candidemia 6 days). Utilization of echinocandins as first line treatment was less frequent (61.5% vs. 78.2%; p=0.014) in those with positive FUBCs. Mortality was significantly higher in the FUBC-positive group (50% vs. 32%; p=0.016). In the multivariable logistic regression model, lower EQUAL Candida Scores, reflecting reduced adherence to guideline-recommended management, were independently associated with persistent candidemia (OR 0.003, 95% CI 0.0002-0.07; p<0.001). Univariable Cox regression identified persistent candidemia ≥5 days (HR 2.16; 95% CI 1.33-3.53; p= 0.002) as a significant predictor of mortality. In the multivariable Cox regression model, intensive care unit (ICU) admission (HR 1.59, 95% CI 1.02-2.50; p= 0.039) and persistent candidemia ≥5 days (HR 2.06, 95% CI 1.26-3.37; p= 0.004) remained independent predictors of mortality.

Persistent candidemia was predicted by poor adherence to treatment guidelines, as shown by low EQUAL Candida Scores, particularly due to the lack of initial echinocandin use. After controlling immortal time bias, persistent candidemia ≥5 days and ICU admission remained independent predictors of mortality in the multivariable model.