The relationship between hemoglobin and study outcomes in fully adjusted models. All models are adjusted for age, sex, race, study, prior cardiovascular disease, diabetes, smoking, systolic blood pressure, LVH and eGFR. Plots are generated using restricted cubic splines with 4 knots generated in S-Plus. Hgb, hemoglobin in g/dL. Hatch marks represent the relative proportion of individuals at a given hemoglobin level. Overall effect indicates whether hemoglobin is a statistically significant risk factor for the outcome, while the linearity test indicates if the relationship between hemoglobin and the outcome is linear; for composite and mortality outcomes (p<0.001), the relationship is non-linear.