Longitudinal trajectory characteristics of the erythropoietin resistance index and their association with all-cause mortality in maintenance hemodialysis patients

Anemia is common in kidney failure, and most patients depend on erythropoiesis-stimulating agents (ESAs). We grouped maintenance hemodialysis (MHD) patients by how their erythropoietin resistance index (ERI) changed over time, and asked whether these patterns predicted death. We reviewed records of MHD patients treated at our center between 2021 and 2025. K-means clustering was applied to repeated ERI measurements to define trajectories. Logistic regression tested whether trajectory group, continuous ERI, and ERI quartile were related to all-cause mortality, the primary outcome. We also used spline models, subgroup and sensitivity analyses, and examined whether ERI improved prediction. A total of 150 patients were included, and 110 had enough repeated ERI data for trajectory analysis; 26 of them died (23.6%). Baseline ERI quartiles differed in dialysis vintage, dry weight, hemoglobin, and sex. In adjusted models, none of the ERI measures—continuous, quartile, or trajectory—predicted mortality. Dialysis vintage, sex, and hemoglobin were stronger predictors, and adding ERI barely changed discrimination. ERI varied widely between patients and tracked the severity of anemia and overall clinical state. It did not, however, show an independent association with death that we could detect. With a retrospective design and limited precision, our results are inconclusive about prognostic value rather than evidence that none exists.