Comment on the Paper: Hospitalization and Mortality in Hemodialysis Patients: Association with Hemoglobin Variability
tivariate logistic regression analysis using hypertension and coronary artery disease as covariates. However, those covariates do not meet the criteria for confounder’s definition since they are on a causal path between the predictor and the outcome [2] , which means the higher hemoglobin, the higher blood pressure [3] , and the higher hemoglobin, the worse coronary artery disease [4] . Above all, the authors entered in the model all covariates and observed what would come up as the result. Thus, they were testing a prognostic and not an etiological model, so there was no room for the inferences made, but only for the probabilities for the outcomes specified [5] . Dear Editor, I read with great interest the paper published in your journal by Handelman et al. [1] . The authors concluded that the number of excursions below the range appeared to have the most influence on mortality, whereas time below range, area below range, and measures of amplitude have the most effect on hospitalization. They also specified that this variability is linked to outcomes of hospitalization and mortality, with the most impact coming from measures below the target range. However, the observational study design that the authors applied cannot account for such conclusions because it was not a randomized controlled trial to infer causality. The authors performed the epidemiological study instead and applied mulPublished online: February 6, 2014