Hemoglobin and All-Cause Mortality: Results from the Baltimore Longitudinal Study of Aging (BLSA).

Whether mild-to-moderate hemoglobin (Hb) reduction is associated with increased mortality independent of underlying diseases is still debated. Two epidemiological studies addressing this issue in high risk populations (disabled older women, and persons 85 years or older) reported conflicting results. The BLSA offers the opportunity to investigate this association in an adult generally healthy population. The BLSA is a large cohort study on aging started in 1958. The participants are volunteers living in the Baltimore-Washington area (USA), well educated and self-describing as generally healthy persons. Participants are examined every 1-2 years. In the present analysis, all participants evaluated at least one occasion for Hb were included. Subjects were not excluded for any pre-existing conditions or specific outcomes. Hb levels were measured in 2979 participants recruited from 1958 to 2002 (1793 men and 1186 women, 78.7% Caucasians, age range: 18–96 years). Mean age at initial evaluation was 52.3 (±17.4) years for women and 52.1 (±17.9) for men, and the average time from baseline to death/censoring was 17.6 years for men and 12.4 years for women. Crude mortality rates were 21/1000 person years for men and 9.7/1000 person years for women. After adjusting for age, a U-shaped relationship (obtained fitting a Cox model with a Poisson regression approach according Therneau & Grambsch, 2000) was observed between Hb concentration and mortality with nadir at Hb=12.8g/dL in women and 15.0g in men. In the same model, significant excess of mortality was associated with Hb ≤ 11.2g/dL (95%CI 9.8g/dL–12.3 g/dL) and ≥ 14.0 g/dL (95%CI 13.5g/dL–14.9g/dL) in women, and ≤ 13.4 g/dL (95% CI 12.5g/dL–14.4g/dL) and ≥ 16.2g/dL (95% CI 15.4g/dL–17.2g/dL) in men. After adjusting for demographics, anthropometrics, smoking status, radiation exposure, and co-morbid diseases (heart disease, stroke, pulmonary disease, cancer, diabetes, hypertension), men with Hb ≤ 13.4 g/dL (mean: 12.7 g/dL, SD: 0.6) were at 30% higher risk (RR: 1.31, 95%CI: 1.08–1.58) compared to the reference group (Hb: 13.4–16.2). Accordingly, women with Hb ≤ 11.2 g/dL (mean: 10.2 g/dL, SD: 1.0) showed 100% greater risk compared to the reference group (Hb: 11.2–14.0), however this association was not statistically significant (RR: 2.01, 95% CI: 0.88–4.63). When men with Hb levels <10g/dL and women with <9 g/dL were excluded the results did not substantially change. Our findings showed that in the general population moderately-low Hb levels per se were associated with increased mortality risk. The impact of anemia correction on mortality in the elderly remains to be determined.