Prognostic Value of Serial Lactate Dehydrogenase Measurements for Determining Early Mortality in ICU Patients: A Retrospective Cohort Study

Background: This study investigated whether lactate dehydrogenase (LDH) levels measured at ICU admission predict early in-hospital mortality among critically ill medical patients in a single-center retrospective cohort study conducted in Turkey. Specifically, we aimed to (i) determine an optimal LDH threshold; (ii) examine the temporal trajectory of discriminatory performance over 72 h; and (iii) assess LDH as an independent predictor beyond established severity scores. Methods: In this single-center retrospective cohort study, 681 adults admitted to a medical ICU between January 2015 and January 2025 were analyzed. Serial LDH measurements were obtained at 0, 24, 48, and 72 h after ICU admission. This study was approved by the Institutional Ethics Committee (Decision No.: 2025/99). ROC analysis was performed under a predefined sensitivity constraint (≥0.70), and time-to-event outcomes were examined using Kaplan–Meier methods and Cox proportional hazards regression. Determinants of maximum LDH were assessed using a GLM with Gamma distribution and log link. Results: The 28-day mortality rate was 39.1%. ROC analysis identified an optimal 24-h LDH cut-off of approximately 275 U/L (AUC = 0.650; sensitivity = 0.70). Discriminatory performance improved progressively over time (AUC of 0.632 at baseline to 0.690 at 72 h), suggesting that serial measurements may capture evolving prognostic information more effectively than single-time-point measurements. Kaplan–Meier analyses demonstrated a stepwise decline in survival with increasing LDH across all categorization approaches (all log-rank p < 0.001). In multivariable Cox models, log-transformed maximum LDH within the first 72 h was the strongest independent predictor of mortality (HR = 2.2–2.6; p < 0.001), demonstrating larger effect sizes than APACHE II, SOFA, and age in fully adjusted models. GLM analysis indicated that male sex was associated with approximately 33% lower expected maximum LDH. Conclusions: Admission LDH is an independently predictive and readily obtainable prognostic biomarker for early in-hospital mortality in critically ill medical patients. LDH may complement established ICU risk assessment tools, and its integration into clinical workflows as a triage adjunct warrants further evaluation in prospective multicenter studies.