A Population-Based Study of Hemoglobin, Race, and Mortality in Elderly Persons

Abstract

Background. Anemia is associated with increased mortality risk. The impact of mildly low hemoglobin concentration (Hb) on risk for mortality remains unclear, especially among blacks. We examined the racial differences between Hb and mortality.

Methods. This was a population-based study conducted from 1993 through 2006, in a geographically defined community of Chicago, Illinois. A stratified, random sample of 1806 participants 65 years old or older and 50% black, who were participating in the Chicago Health Aging Project and underwent clinical evaluation. Mortality was ascertained using the National Death Index. Cox proportional hazard models were used to assess the independent relation of Hb to mortality risk.

Results. The proportion of participants with anemia by World Health Organization (WHO) criteria (Hb < 13.0 g/dL for men and < 12.0 g/dL for women) was 39% among blacks, and 17% among whites. Blacks had lower mean Hb (12.6 ± 1.5 g/dL) than did whites (13.5 ± 1.5 g/dL). In multivariable analysis, anemia was associated with increased mortality risk in blacks (hazard ratio [HR], 1.90; 95% confidence interval [CI], 1.43–2.53) and in whites (HR, 1.85; 95% CI, 1.32–2.59). Among blacks, Hb 0–0.9 g/dL below the anemia threshold is associated with increased mortality risk compared to Hb 0–0.9 g/dL above the anemia cutoff (HR, 1.84; 95% CI, 1.21–2.79), Hb 1.1–2.0 g/dL above the anemia cutoff (HR, 1.35; 95% CI, 0.88-2.05) and Hb 2.1–3.0 g/dL above the anemia cutoff (HR, 2.24; 95% CI, 1.12–4.47). The terms for interaction between black ethnicity/race and anemia suggested that blacks did not have a statistically significant difference in mortality risk compared to whites. Subgroup analyses of interaction terms suggested that Hb 0.1–1.0 g/dL above anemia cutoff group, blacks may have lower mortality risk compared to whites in the mildly low normal ranges of Hb (p =.02).

Conclusion. Both anemia by WHO criteria and mild reductions in Hb were related to increased risk of mortality in older blacks and whites.

Hemoglobin, Anemia, Race, Mortality, Population-based study

ANEMIA is a common condition in the elderly population and is associated with increased mortality (1–3). Anemia is more common among non-Hispanic blacks (4,5), with the National Health and Nutrition Examination Survey (NHANES) (4) showing a threefold greater anemia prevalence among blacks as compared to non-Hispanic whites. This ethnic/racial difference has also been reported in younger adults with lower hemoglobin concentrations (Hb) than whites even after considering disease status, behavioral risk factors, nutritional intake, and iron status (6,7). These results have led to (a) speculation that blacks may have a lower set point for Hb than whites (8) and (b) uncertainty as to whether the presence of anemia or lower Hb has adverse clinical consequences in blacks similar to those in whites (9,10). To address this issue, we examined the relationships of anemia status and Hb level threshold to all-cause mortality in a biracial population of older blacks and whites.

Methods

Design and Participants

The Chicago Health and Aging Project (CHAP) study aims to identify risk factors for Alzheimer's disease. It is a prospective, epidemiological study of a geographically defined, urban biracial community population in three adjacent south side neighborhoods in Chicago: Morgan Park, Washington Heights, and Beverly with a reasonable distribution of socioeconomic characteristics within each ethnic/racial group. More details of the study design have been previously published (11,12).

Of the 7813 age-eligible residents identified through census of the community areas, 6158 (78.9%) were enrolled for the baseline population interview. Original data collection started in 1993, and data collection occurred in cycles, each lasting 3 years, with each cycle ending as the succeeding cycle began. As of the third cycle in 2000, CHAP began enrolling community residents who had turned 65 since the inception of the study. CHAP participants were randomly selected at each cycle and stratified by age, sex, race, and global cognitive function, to undergo a detailed clinical evaluation. Details of the stratified randomization process have been previously published (11). The randomly selected participants form the basis of the present study (n = 1806). Interviews and physical examinations were conducted with standardized instruments that assess health history, physical and cognitive function, and serological evaluations of all 1806 participants. Written informed consent was obtained, and the study was approved by the Institutional Review Board at Rush University Medical Center.

Data

Information on vital status was obtained from informants at regular follow-up contact and through newspaper obituaries. All deaths are further verified through matching with the National Death Index (NDI). The last date of NDI ascertainment was December 2003. Blood samples were uniformly collected using sterile technique by phlebotomists and nurses. Hb was measured from blood samples using a Beckman-Coulter (Wood Dale, IL) LH750 automated processor with Coulter reagents. Anemia was defined using the World Health Organization (WHO) criteria, as an Hb < 13 g/dL for men or < 12 g/dL for women (13).

Demographic variables used in the analyses included age, sex, race, education, and income. Medical conditions, including hypertension, diabetes mellitus, stroke, cardiovascular disease, hip fracture, cancer, Parkinson's disease, and thyroid disease, were identified by self-report. Global cognitive function was measured by averaging performance across the Mini-Mental State Examination (14), East Boston Memory Test (15), and Symbol Digit Modalities Test (16). Physical function was assessed using Katz Index of Activities of Daily Living (17). Cigarette smoking (ever smoked in life) and overall health status were assessed based on a series of questions derived from the Established Population for Epidemiological Studies of the Elderly (EPESE) project (18). Depressive symptoms were measured using a modified version of Center for the Epidemiological Study of Depression (CESD; 19). Blood tests included mean cell volume (MCV), serum creatinine and cholesterol. Glomerular filtration rate (GFR) was calculated using the Modification of Diet in Renal Disease (MDRD) equation (20). Body mass index was calculated by dividing the measured weight in kilograms by the square of the measured height in meters.

Analytic Approach

Hb level was analyzed in two ways: as a dichotomous variable of anemia versus non-anemia using the WHO criteria and as a categorical variable of Hb groups above or below the anemia cutoff. Cox proportional hazards models (21) were used to test the association between baseline Hb and mortality during follow-up, adjusting for other covariates in blacks and whites separately. All models included age, sex, race, and global cognitive function. Hb was added to the primary model to test the association to mortality risk. Additional models further tested the association of Hb with mortality after adding other potential confounders. In the first step, income and medical conditions were added to the original model. In the second step, physical function and depressive symptomatology were added to the model. In our final fully adjusted model, we also added smoking status, body mass index, GFR, MCV, cholesterol, and self-reported health status to our model. Biological measurements were included as continuous variables. We then tested if the association between Hb and mortality risk varied between blacks and whites for the whole sample by including the interaction of race and Hb.

All analyses were done in SAS (22), and the figures were drawn in S-plus (SAS Institute Inc., Cary, NC). Model assumptions were found to be adequately met and assessed graphically and analytically.

Results

Baseline Characteristics

The mean age of the 1806 participants was 79.9 years (standard deviation [SD] = 6.6 years), with approximately 50% blacks and 50% whites (Table 1). The proportion of participants with anemia according to WHO criteria was 28%; 39% among blacks, and 17% among whites. The mean (SD) Hb was 13.1 (1.6) g/dL for the entire sample. Whites had higher mean Hb values than blacks [13.5 (1.5) vs 12.6 (1.5) g/dL].

Hb and Mortality

Participants were followed for a mean of 3.9 years (median 3.5 years). A total of 530 deaths (29%) were recorded. In our fully adjusted model, anemia was associated with a statistically significant 75% increased mortality risk for blacks (HR, 1.90; 95% CI, 1.43–2.53) (Table 2) and for whites (HR, 1.85; 95% CI, 1.32–2.59). In analyses stratified by ethnicity/race, among the group with GFR < 60 mL/min per 1.73m², increased risk of mortality with anemia status was observed both among blacks (HR, 1.65; 95% CI, 1.07–2.54) and among whites (HR, 2.23; 95% CI, 1.46–3.41). Among the group with GFR ≥ 60 mL/min per 1.73 m², increased risk of mortality with anemia was also observed in blacks (HR, 2.14; 95% CI, 1.46–3.13) and in whites (HR, 1.87; 95% CI, 1.11–3.16).

Threshold for the Relationship of Hb to Mortality by Ethnicity/Race

To examine the Hb threshold that minimized mortality risk, we categorized Hb into six groups below and above the WHO anemia criteria (Hb > 1 g/dL below, Hb 0–0.9 g/dL below, Hb 0.1–1.0 g/dL above, Hb 1–1.9 g/dL above [reference group], Hb 2–2.9 g/dL above, and Hb > 3 g/dL above). In fully adjusted analyses stratified by ethnicity/race, among whites, the point estimates for risk of mortality reached statistical significance for all groups having Hb below that of the referent group (Table 2). Among blacks, Hb > 1 g/dL below the referent group (i.e., Hb < 12.0 g/dL for men or < 11.0 g/dL for women) had increased mortality risk (HR, 1.95; 95% CI, 1.24–3.06). Similarly, Hb 0–0.9 g/dL below the reference group showed a statistically nonsignificant trend for higher mortality (HR, 1.35; 95% CI, 0.88–2.05) compared to the reference group (Hb 1.0–1.9 g/dL above) (Figure 1). We then performed further analyses to examine the mortality risk for participants having Hb 0–0.9 g/dL below the WHO anemia cutoff. When compared to participants having a Hb 0.1–0.9 g/dL above the anemia threshold, Hb 0–0.9 g/dL below the anemia cutoff was associated with an 84% increased risk for mortality among blacks (HR, 1.84; 95% CI, 1.21–2.79). Furthermore, Hb 0–0.9 g/dL below the anemia cutoff increased the risk for mortality among blacks (HR, 2.23; 95% CI, 1.12–4.47) compared to an Hb 2.1–3.0 g/dL above the anemia cutoff.

Interaction Between Anemia Status and Hb Concentration and Mortality

Although a higher point value for risk of death with anemia status was found among blacks compared to whites (1.90 vs 1.85 in the fully adjusted models), the statistical interaction between race and anemia status was not statistically significant (p =.88), suggesting that the increased mortality of anemia did not differ strongly by race. In additional analyses, the interaction term for black ethnicity/race with the Hb–mortality relationship was not statistically significant except for Hb 0.1–1.0 g/dL above the anemia cutoff (Table 3), suggesting that whites had significantly higher risk for mortality than blacks within that hemoglobin group.

Discussion

In this stratified random sample from a population-based cohort of older people, we found that anemia is associated with an estimated 90% increased mortality in blacks and an 85% increased mortality in whites. Racial differences did not appear to substantially influence the strong association between anemia and increased mortality in this population. Among blacks, Hb levels just below the WHO anemia threshold of Hb < 13.0 g/dL for men or < 12.0 g/dL for women significantly increased mortality compared to Hb just above the anemia cutoff.

As with previous studies (4,23), we found a lower average Hb among black older adults. Using the WHO criteria, Denny and colleagues (9) in the EPESE reported anemia prevalence of 34% in blacks and 14% in whites, where as Guralnik and colleagues in the NHANES analysis reported 28% in blacks and 9% in whites. Racial differences in Hb may be in part explained by comorbidity, but Hb has been reported to be lower in blacks than in whites even after taking into account disease status, behavioral risk factors, nutritional intake, and iron status (6,7). One study indicated that the trait of a high prevalence of α-thalassemia among blacks may account for lower Hb, at least among non-elderly individuals (23). These results have led to speculation that blacks may have a lower set point for Hb than whites (8).

Irrespective of our findings, the clinical significance of the lower Hb and anemia status in older blacks remain poorly defined. Denny and colleagues (9) in the Duke EPESE reported on mortality and outcomes among 1744 participants, of whom 54% were black. They found an increased unadjusted (relative risk [RR] = 1.7; CI, 1.4–2.1) and adjusted (RR = 1.4; CI, 1.2–1.8) mortality risk among blacks with anemia. The mortality impact of anemia did not differ between blacks and whites. In contrast, Patel and colleagues (10) in the Health Aging and Body Composition (ABC) Study recently reported that anemia was not associated with higher mortality in blacks (HR = 1.15; 95% CI, 0.77–1.72)] in contrast to the detrimental impact observed in whites. However, the Health ABC Study differs from our study in two important ways. First, CHAP is a population-based study of persons older than 65 years living in the same geographically defined communities regardless of their functional status, whereas the Health ABC Study included a healthier cohort who were well functioning at baseline. Second, the age range of study participants differed: 65 years or older for CHAP and 70–79 years at baseline for Health ABC. Whether these differences account for differences in findings is unclear.

This study builds on prior work by extending the population-based finding of the relationship of Hb to mortality risk in blacks and whites. Our results showed that anemia defined by WHO criteria carries increased mortality risk in blacks and whites. Among blacks, even Hb just below the WHO anemia threshold (Hb 0–0.9 g/dL below) significantly increased mortality risk when compared to Hb slightly above the anemia cutoff (Hb 0.1–1 g/dL above). Among whites, Hb below and slightly above (Hb 0.1–1.0 g/dL) the WHO anemia threshold carries increased mortality risk.

Whether blacks have significantly different mortality risk compared to whites is still unresolved. The point estimate for risk of mortality for anemia by WHO was nominally higher among blacks (HR = 1.90) than among whites (HR = 1.85). The statistical interaction term of the Hb–mortality relationship with black ethnicity/race was nonsignificant (p =.88), suggesting that increased mortality risk of anemia did not differ between blacks and whites. Additionally, when we categorized Hb above and below the anemia cutoff in subgroup analyses, our findings suggested that, for Hb 0.1–1 g/dL above the anemia cutoff, but not for any other Hb groups, blacks had significantly lower mortality than whites (p =.02). Overall, these findings argue against using an Hb threshold below the WHO criteria to define anemia in blacks when considering anemia as a prognostic marker for mortality. However, our findings support considering a higher Hb threshold to define the prognostic impact of anemia in whites. Further study of the impact of low Hb by race on mortality and other health-related measures is warranted.

Despite the findings from this and other observational studies of older participants showing that lower Hb has a strong and independent association with increased mortality, the benefits and risks of anemia correction remain largely unexplored for older anemic adults. Paradoxically, recent data (24) show that achieving higher Hb (>13 g/dL) through pharmacologic treatment with erythropoietin-stimulating proteins may increase mortality, at least for anemia due to chronic kidney disease. Only clinical trials can answer whether anemia treatment affords a benefit.

This study has several limitations. First, unrecognized confounding factors may exist. The comprehensive adjustment performed and the small changes in multivariate analyses in the elevated risk of mortality due to lower Hb diminish this possibility. A further limitation is inability to differentiate the mortality risks by anemia etiologies. Last, our study did not examine repeated measures of Hb over time, and cannot address the relationship of change in Hb to subsequent mortality.

Our study also has substantial strengths. It is a large, population-based study with comprehensive and uniform measurement of potentially relevant confounders. Furthermore, the biracial design permits direct contrasting of two racial groups within the study.

Our findings emphasize the importance not only of anemia status, but that Hb just below anemia cutoff is significantly related to overall mortality in both blacks and whites. We believe that our results support the WHO threshold of Hb < 13.0 g/dL for men or < 12.0 g/dL as reasonable anemia criteria for older blacks, and validate other studies showing that a higher Hb threshold may be considered for whites. Further research is needed to examine the relation of Hb to other important geriatric outcomes in racially diverse populations.

Decision Editor: Luigi Ferrucci, MD, PhD

Figure 1.

Hazard ratio (HR) for mortality by hemoglobin (Hb) concentration groups and race

Table 1.

Selected Characteristics of the Study Population.

Characteristic	Black N = 897	White N = 909	Total (N = 1806)
Age, mean (SD), y	78.9 (6.4)	80.9 (6.6)	79.9 (6.6)
Female gender, %	533 (59.4)	521 (57.3)	1054 (58.4)
Education, mean (SD), y	10.9 (3.5)	13.8 (3.4)	12.4 (3.7)
Income categories, mean (SD)	3.9 (2.0)	5.8 (2.6)	5.0 (2.5)
Smoker, %	440 (49.1)	425 (46.8)	865 (47.9)
Medical conditions
Coronary artery disease (%)	75 (8.4)	75 (8.3)	150 (8.3)
Stroke (%)	68 (7.6)	76 (8.4)	144 (7.9)
Cancer (%)	95 (10.6)	113 (12.4)	208 (11.5)
Thyroid disease (%)	27 (3.0)	37 (4.1)	64 (3.5)
Hypertension (%)	503 (56.1)	330 (36.3)	833 (46.1)
Diabetes mellitus (%)	160 (17.8)	105 (11.6)	265 (14.7)
Parkinson's disease (%)	6 (0.7)	9 (0.9)	15 (0.8)
Hip fracture (%)	23 (2.6)	32 (3.5)	55 (3.0)
Global Cognition, mean (SD)	−0.238 (0.792)	0.184 (0.796)	−0.025 (0.821)
MMSE, mean (SD)	24.7 (5.3)	26.5 (4.7)	25.6 (5.1)
Katz ADL disability > 1 (%)	171 (19.1%)	144 (15.8%)	315 (17.4%)
CESD, mean (SD)	1.9 (2.1)	1.3 (1.8)	1.6 (2.0)
Self-Reported Health Status (poor or fair) (%)	342 (38.1%)	171 (18.8%)	513 (28.4%)
BMI, mean (SD), kg/m2	27.3 (5.7)	25.5 (4.9)	26.4 (5.4)
GFR, mean (SD), mL/min per 1.73 m2	65.9 (21.8)	61.5 (17.4)	63.7 (19.8)
MCV, mean (SD), fL	87.9 (6.4)	91.7 (5.0)	89.8 (6.3)
Anemia: WHO criteria (%)	350 (39.0%)	158 (17.4%)	508 (28.1%)
Hb, mean, g/dL (SD)	12.6 (1.5)	13.5 (1.5)	13.1 (1.6)
Death, No. all-cause (%)	248 (27.6%)	282 (31.0%)	530 (29%)
Time to death, mean (SD), y	3.7 (2.8)	4.0 (2.8)	3.9 (2.8)

Note: SD = standard deviation; ADL = activities of daily living; MMSE = Mini-Mental State Examination; CESD = Center for Epidemiological Study of Depression scale; BMI = body mass index; GFR = glomerular filtration rate; MCV = mean cell volume; Hb = hemoglobin concentration; WHO = World Health Organization.

Table 2.

Multivariate-Adjusted Hazard Ratio for Mortality by Race.

	Blacks	Whites
	Hazard Ratio (95% Confidence Interval)
Anemia Status (World Health Organization [WHO])
Model A	1.96 (1.52–2.53)	2.29 (1.74–3.00)
Model B	1.90 (1.43–2.53)	1.85 (1.32–2.59)
Hemoglobin (Hb) concentration (reference group: Hb 1.1–2 g/dL above WHO anemia cutoff)
> 1 g/dL below
Model A	2.15 (1.47–3.16)	2.66 (1.74–4.06)
Model B	1.95 (1.24–3.06)	2.17 (1.28–3.65)
0–0.9 g/dL below
Model A	1.41 (0.96–2.08)	2.37 (1.63–3.43)
Model B	1.35 (0.88–2.05)	2.14 (1.39–3.30)
0.1–1 g/dL above
Model A	0.92 (0.62–1.37)	1.37 (0.99–1.88)
Model B	0.86 (0.56–1.32)	1.62 (1.13–2.32)
2.1–3 g/dL above
Model A	0.76 (0.40–1.45)	1.01 (0.69–1.47)
Model B	0.69 (0.36–1.34)	1.08 (0.71–1.66)
> 3 g/dL above
Model A	0.76 (0.18–3.14)	1.16 (0.71–1.92)
Model B	0.81 (0.19–3.41)	1.41 (0.84–2.39)

Note: For the models, we adjusted for the following covariates for total cohort (for black and white cohorts, race has been removed from the models): A (primary model): age, sex, education, race, global cognition; B (fully adjusted model): age, sex, education, race, global cognition, income, coronary artery disease, diabetes, hypertension, stroke, cancer, hip fracture, Katz Activities of Daily Living, Center for Epidemiological Study of Depression scale, smoking status, self-reported health status (poor, fair), body mass index, glomerular filtration rate, serum cholesterol, mean cell volume.

Table 3.

Interaction Terms Between Race and Hemoglobin (Hb) Concentration.

	Coefficient	Hazard Ratio	95% CI
Anemia Status (World Health Organization)
Black Participants × Anemia Status
Model A 0.29	−0.19	0.82	0.58–1.18
Model B 0.88	0.03	1.03	0.69–1.55
Hb Concentration (Categorical Groupings)
Black Participants × Hb Groups
Hb > 1 g/dL below	−0.15	0.86	0.45–1.63
Hb 0–0.9 g/dL below	−0.50	0.60	0.34–1.09
Hb 0.1–1 g/dL above	−0.64	0.52	0.30–0.91*
Hb 1.1–2 g/dL above	(Reference)	1.0
Hb 2.1–3 g/dL above	−0.56	0.57	0.26–1.24
Hb > 3 g/dL above	−0.53	0.59	0.13–2.69

Notes: Hazard ratio was adjusted for age, sex, race, education, global cognition, income, coronary artery disease, diabetes, hypertension, stroke, cancer, hip fracture, Katz Activities of Daily Living, Center for Epidemiological Study of Depression scale, smoking, self reported health status (poor, fair), body mass index, glomerular filtration rate, serum cholesterol, mean cell volume.

*p =.023.

CI = confidence interval.

This work was supported by National Institute on Aging grant AG11101.

We thank Ann Marie Lane for community development and oversight of project coordination, Michelle Bos, Holly Hadden, Flavio LaMorticella, and Jennifer Tarpey for coordination of the study. We also thank Todd Beck for statistical programming and George Dombrowski for data management support.

Dr. Dong had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Drs. Dong, Tang, Mendes de Leon, and Evans were responsible for the conception and design as well as analysis and interpretation of data. All authors were involved in the drafting of the manuscript, critical revision of the manuscript, and statistical analysis of the manuscript. Drs. Artz and Shah were responsible for the conception and design, analysis and interpretation of data, critical revision of the manuscript, and statistical analyses of the manuscript.

Dr. Artz has served on the advisory board and received research support from Amgen, Thousand Oaks, CA.