Mechanisms of hemoglobin cycling in anemic end-stage renal disease patients treated with erythropoiesis-stimulating agents

Abstract Background and Aims A considerable fraction of renal anemia patients treated with erythropoiesis-stimulating agents (ESAs) experience hemoglobin ‘cycling’ periods during which hemoglobin levels periodically over- and undershoot a specified target range [1]. This is an undesired condition, which also necessitates frequent adjustment of the drug dose. The causes of hemoglobin cycling are not understood on a fundamental level. Through a combination of biomedical simulations and quantitative data analysis, we aimed at identifying mechanisms of hemoglobin cycling and the physiological and treatment-related factors contributing to cycling. Method We developed a biomedical modeling and simulation scheme that captures the essential features of ESA therapy involving a minimal feedback model for the delayed hemoglobin response after ESA administration and a stereotypic ESA dosing algorithm [2]. In this model, hemoglobin cycling emerged as a consequence of the delayed feedback between a patient's physiological response to ESA administrations and the resulting dose adjustments, which mutually maintain themselves in a causal loop. Guided by this model, we developed a set of statistical indicators that can detect this type of self-sustained hemoglobin cycling in clinical time series, including the Pearson cross correlation between hemoglobin levels and given ESA doses and hemoglobin standard deviation. Results A model analysis showed that physiological and treatment-related delays in hemoglobin response and ESA dose adjustment and administration were a major driver of hemoglobin cycling. Such delays are caused, e.g., by the cell cycle and maturation times of erythroid progenitors in response to ESA administration on one hand and restrictions in ESA dose changes and typical administration intervals on the other hand. In clinical time series, this was reflected by a systematic delay between bursts of ESA administrations and hemoglobin cycles. We quantified 1134 such datasets from hemodialysis patients and found that more persistent cycling was associated with negative correlation coefficients between hemoglobin levels and ESA doses. Using model simulations with similar behavior, we found that many patient- and therapy-related factors affect the hemoglobin dynamics in several (sometimes competing) ways, obfuscating their overall effect. For example, a larger RBC lifespan generally both entails higher hemoglobin levels but also prolongs the effect of ESA misdosings. However, model simulations suggested that longer RBC lifespans generally have a favorable effect in requiring less ESA on average and preventing hemoglobin cycling. Moreover, the model showed that for a given patient, ESA-specific properties such as its half-life may have a ‘sweet spot’ range of values within which hemoglobin cycling was suppressed while still showing a desired effect on the hemoglobin dynamics. Furthermore, the choice of the hemoglobin target window had an impact on hemoglobin stability. If the window was chosen too small, more frequent dose adjustments could initiate and maintain a cycling behavior. Conclusion Hemoglobin cycling in renal anemia patients often emerges from a complex interplay of multiple physiological and ESA treatment-related factors, with delays between ESA administration, hemoglobin response and subsequent ESA dose adjustment being a key driver of cycling. Biomedical modeling and simulation can aid in systematically exploring these interplays and find improved dosing strategies.