Treatment with high dose of erythropoiesis-stimulating agents and mortality: analysis with a sequential Cox approach and a marginal structural model

Anemia-correction trials indicated higher mortality rates in chronic kidney disease patients assigned to higher hemoglobin targets. The safety of the high erythropoiesis-stimulating agent (ESA) doses that these patients received has therefore been questioned. However, no trial that directly compares treatment with different ESA doses has been published. We thus aimed to estimate the effect of high ESA dose on mortality in an observational cohort of dialysis patients. The Netherlands Cooperative Study on the Adequacy of Dialysis is a Dutch cohort study of incident dialysis patients in which ESA dose, comorbidities, and laboratory parameters were collected every 6 months. Mortality in patients with a high ESA dose (above median 6000 units/week) was compared with that in patients with no or low ESA dose with Cox regression analyses. To handle time-dependent confounding, a sequential Cox approach was used conditional on baseline covariates, with inverse probability of censoring weights (IPCW) for dependent censoring. Analyses were repeated with a marginal structural model (MSM) with inverse probability of treatment weights and IPCW. Hazard ratio (HR) for high ESA dose was 1.20 (95%CI 0.83-1.73) with a sequential Cox and 1.54 (95%CI 1.08-2.18) with an MSM. Truncation of weights in the MSM did not affect estimates. To compare, conventional Cox analyses indicated a baseline adjusted HR of 1.66 (95%CI 1.20-2.31). Patients treated with high ESA dose have a 1.2-1.5 increased risk of mortality. Our analyses support guidelines advising a conservative ESA dosing regimen, which carefully weighs the patients' benefits and risks