Clinical Validation Study for Optimization of Anemia MAnagement in Hemodialysis Patients with End-Stage Kidney Disease Using the Dialysis Anemia TReatmenT Model” (SMART study)

Management of anemia in patients with ESRD is a labor-intensive process as hemoglobin (Hb) targets must be evaluated regularly, and frequent erythropoietin stimulating agent (ESA) dose adjustments are required to be maintained Hb in a narrow therapeutic range. Exceeding the Hb target or not achieving it can lead to adverse events. In this single center study, artificial intelligence and predictive algorithm was created to individualize patient dosing of ESA in patients with ESRD. We hypothesized that a computer based predictive algorithm can individualize ESA dosing leading to reduced ESA use, thereby reducing cost, while maintaining Hb in the recommended target range. 

This open-label study was approved by Intermountain Health IRB and conducted at 2 dialysis centers in Salt Lake City between July 2023 and April 2024.  Inclusion criteria were age =/> 18 years and availability of previous 150 days of Hb values and ESA doses in the EMR. Exclusions included ferritin >1000, active malignancy, PTH >1000 pg/mL, active GI bleeding, hospitalization in previous 150 days. The primary ESA used in the dialysis centers was darbepoetin. Clinical and lab data collected included hemoglobin (Hgb), ferritin, transferrin, iron saturation, Kt/V, PTH, phosphorus and albumin. Hospitalizations, gastrointestinal bleeding and dose of IV iron were tracked. Lab data were obtained every 2 weeks, and a computer-based algorithm was used to generate ESA dose recommendations during the study period of 90 days for each patient. The target range of hemoglobin was 9.0 -11.9 g/dl.  A committee consisting of the principal investigator, scientific analyst, statistician, and registered nurses from the dialysis centers reviewed the computer-generated ESA recommended doses every 2 weeks to ensure patient safety. The Medical Director had the right to override the dosing recommended by the algorithm. All statistical comparisons are unpaired two-sample Welch’s t-tests, computed in R version 4.5.0.

Forty patients were enrolled – 2 dropped out and 38 patients completed the study.  The study was analyzed as a pre-post study design, with data for each patient from 90 days before they entered the study serving as the control. There was no difference in any of the clinical or laboratory data between the study and control period and there was no difference in delivered dose of dialysis (Kt/V). There was no difference in the incidence of hospitalizations or GI bleeding during the 2 periods. The mean dose of intravenous iron sucrose was 250 mg during the control period and 209 mg during the study period (not statistically significant). There was no difference in Hb values between the control and treatment period (10.6 +/- 1.1 and 10.5 +/- 1.4, P =0.6). Total ESA administered during the control period was 18480 mcg and during treatment period was 13390 mcg (27% reduction, 95% CI, P =0.002). Figure 1

This small single center study showed that a computer-based algorithm for anemia management in the dialysis population can generate significant cost savings without jeopardizing patient safety or clinical outcomes. A larger multi-center study is required to validate these findings and determine whether predictive algorithms can be utilized successfully in the care of patients with ESRD.