Si-Ho Kim, Jiho Park, Seok Jun Mun, Hee-Won Moon, Bomi Kim, Sung Un Shin, Miri Hyun, Hyun Ah Kim, Joon-Sik Choi, Yu Mi Wi, Jae-Hoon Ko, Hae Suk Cheong, Sook-In Jung, Kyong Ran Peck, KSID (Korean Society of Infectious Diseases) candida study group
Four clinically distinct candidemia subphenotypes were identified and were associated with progressively increasing mortality, suggesting the potential prognostic relevance of subphenotype-based stratification. External validation is required to confirm their reproducibility and clinical relevance.
OBJECTIVES: Candidemia is a heterogeneous clinical syndrome that may not be adequately characterized using conventional analyses based on individual variables. We aimed to identify clinically relevant subphenotypes of candidemia using a data-driven approach.
METHODS: We conducted a multicenter retrospective cohort study across eight university-affiliated medical centers in South Korea between 2019 and 2023. Only the first episode of candidemia was included. Latent class analysis (LCA) was performed using demographic and clinical variables, including age, comorbidities, Candida species, and presumed source of candidemia. Associations between subphenotypes and mortality were evaluated using adjusted Cox proportional hazards models, and robustness was assessed by leave-one-center-out analyses.
RESULTS: Among 2,279 patients with candidemia, LCA identified four subphenotypes. Class 1 was enriched for solid cancer, Class 2 for older age, hypertension, and diabetes, Class 3 for liver disease, end-stage renal disease, and solid organ transplantation, and Class 4 for hematologic malignancy. Classes 3 and 4 also had longer intervals from hospital admission to candidemia and higher proportions of non-albicans Candida. Thirty-day Kaplan-Meier survival probabilities progressively declined across subphenotypes (60.2%, 56.9%, 49.3%, and 44.7%, respectively). Compared with Class 1, Classes 3 and 4 were independently associated with increased mortality (adjusted hazard ratios 1.44 [95% CI, 1.12-1.85] and 1.76 [95% CI, 1.23-2.53], respectively). Findings were consistent in leave-one-center-out analyses.
CONCLUSIONS: Four clinically distinct candidemia subphenotypes were identified and were associated with progressively increasing mortality, suggesting the potential prognostic relevance of subphenotype-based stratification. External validation is required to confirm their reproducibility and clinical relevance.