Anemia Treatment, Hemoglobin Variability, and Clinical Events in Patients With Nondialysis-Dependent CKD in Japan

BACKGROUND
Anemia management in patients with non-dialysis-dependent chronic kidney disease has attracted attention with the introduction of novel therapeutic agents; however, few studies have provided comprehensive epidemiologic information.


METHODS
A retrospective cohort study was conducted in adult patients with stage ≥3a non-dialysis-dependent chronic kidney disease and hemoglobin <11 g/dL (January 2013-November 2021; N=26,626) to assess longitudinal treatment patterns, hemoglobin, and iron parameters (ferritin and transferrin saturation) for anemia management. Time-dependent Cox proportional hazard models were applied to assess the risk of clinical events, including death, cardiovascular events, dialysis introduction, and red-blood-cell transfusion, associated with temporal fluctuation patterns of hemoglobin levels.


RESULTS
The cumulative incidence of anemia treatment initiation within 12 months was 37.1%, including erythropoiesis-stimulating agents 26.5%, iron oral 16.8%, iron intravenous 5.1%, and hypoxia-inducible factor prolyl hydroxylase inhibitor 0.2%. The mean (±standard deviation) hemoglobin levels were improved from 9.9±1.2 g/dL to 10.9±1.6 g/dL at 12 months. Despite erythropoiesis-stimulating agents or hypoxia-inducible factor prolyl hydroxylase inhibitor therapy, 30.1% of patients remained hemoglobin <10 g/dL. The risks of premature death, cardiovascular events, dialysis introduction, and red-blood-cell transfusion were significantly higher in groups with consistently low hemoglobin or low-amplitude hemoglobin fluctuation around the lower limit of target hemoglobin range than in patients with target hemoglobin range (p <0.05). Likewise, significantly higher risks for dialysis introduction and red-blood-cell transfusion were associated with high-amplitude hemoglobin fluctuation across target hemoglobin range were observed.


CONCLUSIONS
The findings underscore the importance of stable hemoglobin control within the target range to reduce the mortality and morbidity risks in patients with non-dialysis-dependent chronic kidney disease while highlighting the suboptimal and heterogeneous treatment of anemia in clinical practice.


Time-dependent Cox proportional models to assess the risks of adverse clinical events associated with hemoglobin fluctuation
The risks of adverse clinical events can be affected not only by the baseline characteristics of patients, but also by parameters continuously changing over the follow up period.Therefore, the classical Cox hazard model may not be suitable for assessing the risks of clinical events associated with transitioning hemoglobin fluctuation.We applied the Anderson and Gill model 1 for the analysis to evaluate the risk of transitioning hemoglobin (Hb) fluctuation patterns associated with the risk of clinical events.This model treats the observational period of each patient as an independent factor.The observational periods of each patient were divided into every 6-month period in which the time-dependent covariates were collected.In addition, the occurrence of clinical events was also assessed in each 6-month period.
The time-dependent covariates were modeled to evaluate the risk associations between the timedependent covariates and the clinical events observed in the period following the assessment period in which the time-dependent covariates were collected (Fig. ).Time-independent covariates were collected for the baseline period, based on the information collected for 6 months prior to the index date.The analysis was performed in the overall patient population from the index date to the end of the follow-up period.

Fig. Assessment windows in the time-dependent Cox proportional hazard models
The models were adjusted by Hb fluctuation patterns, use of an erythropoiesis stimulating agent, iron oral (including dose), iron intravenous, hypoxia-inducible factor prolyl hydroxylase domain enzyme inhibitor, red-blood-cell transfusion, and ferritin category (ferritin <100 ng/mL or ≥100 ng/mL) as time-dependent covariates, and estimated glomerular filtration rate, albumin, c-reactive protein, age, sex, cardiovascular disease, diabetes mellitus, heart failure, and etiology of kidney disease (hypertension, glomerulonephritis, renovascular disease, polycystic kidney disease, and auto-immune disease) as time-independent covariates.The Hb fluctuation patterns were categorized into six groups based on the previous literature 2,3 : within the target Hb range (target); consistently below the target (Low); consistently above the target (high); low-amplitude fluctuation around the upper limit of the target (LAH); low-amplitude fluctuation around the lower limit of the target (LAL); and, high amplitude fluctuation across the target range (HA).4][5] The sensitivity analysis was performed using the target Hb range of 11-12 g/dL.
The hazard ratios of clinical events for each Hb fluctuation group compared to target Hb group were calculated with point estimates and 95% confidence intervals.

Reporting checklist
This article was written following the STROBE (STrengthening the Reporting of OBservational studies in Epidemiology) statement, 6 details of which are further elaborated in Table S3.CKD, chronic kidney disease; ICD-10, international classification of disease, 10 th revision; eGFR, estimated glomerular filtration rate Table S2.Definitions of adverse clinical events collected in the study.

Clinical outcomes Definitions
Observed number of events in the overall patient population, n (%) All-cause death All death information and death record extracted from the electronical medical records in the follow-up period 5,991 (22.5)

Figure S1 .Figure S2 .
Figure S1.Cumulative incidence curves of anemia treatment initiation in subgroups categorized by TSAT, ferritin, or CKD stages.

Figure S3 .Figure S4
Figure S3.Changes in ferritin levels in the overall patient population during the follow up period.

Table S3 . STROBE Checklist
Interpretation 20 Give a cautious overall interpretation of results considering objectives, limitations, multiplicity of analyses, results from similar studies, and other relevant evidence Generalisability 21 Discuss the generalisability (external validity) of the study results Other information Funding 22 Give the source of funding and the role of the funders for the present study and, if applicable, for the original study on which the present article is based 1, 12