Feature-based clustering of hemoglobin trajectories and subsequent risk of end-stage kidney disease in stages 3-4 CKD: a landmark analysis.

Background: Anemia and protein-energy wasting (PEW) are common complications in chronic kidney disease (CKD) and are associated with disease progression. However, the prognostic significance of longitudinal hemoglobin (Hb) trajectory patterns, particularly in the context of nutritional status, remains unclear. This study aimed to identify distinct Hb trajectories in CKD stages 3-4 patients and evaluate their associations with renal outcomes, employing a dual analytical approach: exploratory analysis of concurrent eGFR decline and predictive analysis of subsequent dialysis risk using a landmark design. Methods: This retrospective landmark cohort study included 694 non-dialysis CKD patients from a Japanese medical center. All completed a 2-year landmark period with Hb measured at regular 90-day intervals (± 15 days). Feature-Based Clustering of Longitudinal Trajectories was applied to Hb measurements during the 2-year landmark period to identify trajectory groups. For exploratory association analysis, Cox proportional hazards regression assessed the relationship between Hb trajectories and concurrent ≥30% eGFR decline within the landmark period. For predictive risk stratification, competing risk Fine-Gray regression models (with death as the competing event) evaluated the risk of dialysis initiation occurring after the 2-year landmark period, with trajectory groups serving as baseline predictors. Results: Five distinct Hb trajectory groups were identified: (1) Low-Stable ( Conclusions: Distinct longitudinal Hb trajectories exist in non-dialysis CKD patients and are independently associated with renal progression risk. Maintaining higher, stable Hb levels is associated with delayed renal replacement therapy initiation, whereas rapid Hb decline identifies a high-risk subgroup requiring intensive monitoring and integrated nutritional intervention.