Longitudinal albumin and blood pressure trajectories identify mortality risk phenotypes in hemodialysis

Serum albumin and systolic blood pressure (SBP) are strong predictors of adverse outcomes in hemodialysis patients, yet clinical interpretation often relies on static measurements rather than longitudinal patterns. Functional data analysis enables characterization of dynamic biomarker trajectories and identification of patient phenotypes that may better capture underlying nutritional, inflammatory, and hemodynamic states. Here, we applied data-driven clustering methods to map albumin and SBP trajectories and to identify phenotypes associated with higher mortality risk. We analyzed 38,863 adults who initiated hemodialysis in Fresenius Kidney Care North America clinics between 2010 and 2014 and had ≥ 9 months of albumin and SBP measurements during the first year. Monthly averages were modeled as functional data. Unsupervised clustering using K-means and functional principal component analysis (FPCA) identified albumin and SBP trajectory phenotypes. Cox proportional hazards models assessed the association between trajectory clusters and all-cause mortality after the first year, adjusting for demographics, diabetes, catheter use, interdialytic weight gain, erythropoietin dose, hemoglobin, and neutrophil-to-lymphocyte ratio. K-means clustering identified three albumin phenotypes (low, medium, high), and FPCA identified three trajectories (high, rising, falling). Low or declining albumin was consistently associated with the highest mortality risk. Compared to the high-albumin group, low-albumin patients had a 2.63-fold higher adjusted mortality risk (95% CI 2.42 to 2.87), and those with falling albumin had a similarly elevated risk (HR 2.40; 95% CI 2.13 to 2.70). SBP clustering revealed distinct patterns: K-means derived low pre-dialysis SBP was associated with increased mortality (HR 1.26; 95% CI 1.16 to 1.36), whereas medium- and high-SBP clusters demonstrated more favorable outcomes. FPCA-derived falling SBP trajectories, but not rising SBP trajectories, were independently associated with higher mortality risk after adjustment (HR 1.15; 95% CI 1.04 to 1.28). Longitudinal albumin and SBP trajectories identify reproducible patient phenotypes with markedly different mortality risk. Persistently low or declining albumin was a strong predictor of death, and a low or falling SBP was associated with poorer survival. Incorporating trajectory-based assessments may improve identification of high-risk patients and support more individualized management strategies in maintenance hemodialysis.