Objectives: To examine the associations between hemoglobin (Hb) concentration and (1) 5-year all-cause mortality and (2) serum erythropoietin (EPO), as the basis for the identification of data-driven thresholds, and to assess the clinical relevance of mildly low Hb.
Design: Prospective study.
Setting: Population based.
Participants: Community-dwelling women aged 65 and older with moderate-to-severe disability--Women's Health and Aging Study I, Baltimore, Maryland, 1992-2000.
Methods: Proportional hazards regression was used to model the relationship between baseline Hb (available for 686 subjects) and time to death. A generalized linear model was used to assess the cross-sectional association between Hb and EPO in 641 subjects.
Results: A curvilinear slope of steady mortality decrease up to the Hb threshold of 13.9 g/dL was observed. Hb of 11 g/dL was independently associated with greater mortality than the World Health Organization (WHO) low-normal cutoff of Hb of 12 g/dL (hazard ratio (HR)=1.2, 95% confidence interval (CI)=1.1-1.4), whereas Hb of 14 g/dL was linked to 24% lower mortality (HR=0.76, 95% CI=0.63-0.92), after comprehensive adjustment for major health status and disease-burden indicators. A curvilinear, statistically significant slope of steady EPO decrease with increasing Hb up to the threshold of 14.3 g/dL was consistently observed.
Conclusion: The meaningfully lower mortality risk with higher Hb levels provides empirical evidence against the notion that Hb currently perceived as mildly low is clinically benign. Furthermore, the mortality risk gradient observed even within the WHO normal Hb range suggests that Hb levels higher than what is currently recommended might offer clinical advantage. The relationship between Hb and EPO provided supporting physiological evidence for this hypothesis.