Longitudinal trajectories of nutrition-related biomarkers and mortality risk in maintenance hemodialysis patients: a joint modeling analysis.

BACKGROUND

Malnutrition and nutrition-related abnormalities are common in people with kidney failure, particularly among patients receiving maintenance hemodialysis, and are associated with adverse clinical outcomes. Nutritional status and related biological processes are dynamic; however, the prognostic relevance of their longitudinal changes remains incompletely understood.

OBJECTIVE

To investigate the associations between longitudinal trajectories of nutrition-related biomarkers and all-cause mortality in patients receiving maintenance hemodialysis using joint modeling.

METHODS

This retrospective longitudinal cohort study included 345 maintenance hemodialysis patients, among whom 99 deaths occurred during follow-up. Longitudinal changes in nutrition-related biomarkers were analyzed in relation to all-cause mortality using joint models that integrate repeated biomarker measurements with time-to-event outcomes. Dynamic prediction was evaluated using a fixed 12-month prediction horizon.

RESULTS

In adjusted univariable joint models, several longitudinal biomarkers were associated with all-cause mortality. In the primary multivariable joint model incorporating longitudinal C-reactive protein (CRP) and serum iron, the current value of CRP was positively associated with mortality risk (HR, 1.030; 95% credible interval [CrI], 1.001-1.058), as was the current value of serum iron (HR, 1.073; 95% CrI, 1.015-1.132). Across landmark times from 6 to 36 months, time-dependent AUC values ranged from 0.756 to 0.807 for 12-month dynamic prediction.

CONCLUSION

The current values of CRP and serum iron were positively associated with all-cause mortality in maintenance hemodialysis patients. Joint modeling of repeated biomarker measurements may help characterize dynamic nutrition-related inflammatory and metabolic risk in kidney failure.