Hemoglobin Concentration and Cognitive Impairment in the Renal REasons for Geographic And Racial Differences in Stroke (REGARDS) Study

Abstract

Background.

There is growing interest in determining the degree of anemia, which is clinically significant. The goal of this study was to determine the association between hemoglobin concentration and cognitive impairment in a large sample of U.S. adults.

Methods.

We used cross-sectional data from 19,701 adults participating in the REasons for Geographic And Racial Differences in Stroke study. Cognitive impairment was defined as a score of 4 or less on the six-item screener. Hemoglobin was analyzed in 1 g/dL increments relative to the World Health Organization (WHO) threshold (<13 g/dL for men and <12 g/dL for women).

Results.

The mean hemoglobin concentration was 13.7 ± 1.5 g/dL. The prevalence of cognitive impairment increased from 4.3% among individuals with a hemoglobin >3 g/dL above the WHO threshold to 16.8% for those with a hemoglobin ≥2 g/dL below the WHO threshold. After adjustment for demographics, chronic health conditions, health status, and inflammation, the association between reduced hemoglobin and cognitive impairment was attenuated and no longer significant, including among those with hemoglobin ≥2 g/dL below the WHO threshold (odds ratio 1.39, 95% confidence interval = 0.94–2.04). A test for linear trend was of borderline significance (p value = .06). For 94% of the sample within 2 g/dL of the WHO threshold, there was no relationship between hemoglobin concentration and the odds of cognitive impairment. The associations did not differ by sex and race.

Conclusions.

Within a large sample of community-dwelling adults, there was no significant association between hemoglobin concentration and cognitive impairment after multivariable adjustment.

ANEMIA is a common condition among older individuals, present in up to 30% of community-dwelling elderly, depending on the definition used and the population studied (1,2). Accumulating evidence suggests that anemia may be an important risk factor for cognitive impairment in the elderly and those with chronic disease. For example, anemia has been linked to cognitive impairment in hospitalized elderly (3) and in persons with end-stage renal disease (4,5), heart failure (6), and malignancy (7).

There is less known about the association between anemia or hemoglobin concentration and cognitive impairment in the general population. Previous studies that were limited to women, or to a single geographic region, did not account for important confounders such as kidney function, or excluded participants with severe anemia (8–14). In addition, the applicability of the World Health Organization (WHO) definition for anemia, a hemoglobin <13.0 g/dL among men or <12.0 g/dL among women, in elderly and non-White populations has been recently questioned because even modest decreases in hemoglobin concentration among those not classified as “anemic” by WHO criteria are associated with increased morbidity and mortality (15–17). In order to clarify these associations, we used a large biracial cohort to examine the independent association between hemoglobin concentration and anemia status with cognitive impairment prevalence. We wished to assess whether hemoglobin concentration and anemia status were associated with cognitive impairment independent of sex, race, and other confounders.

METHODS

Participants

The REasons for Geographic And Racial Differences in Stroke (REGARDS) study is a cohort study of black and white individuals aged 45 years and older designed to identify factors that contribute to the excess stroke mortality in the southeastern United States and among blacks. Cognitive function is an important secondary outcome in REGARDS. The sampling methods of the REGARDS study have been described in detail elsewhere (18). Briefly, REGARDS participants were recruited from a commercially available nationwide list of more than 250 million individuals and 120 million households (Genesys, Inc., Fort Washington, PA). The cohort is 42% black, and males constituted 38% of blacks and 50% of whites. Exclusion criteria included race other than black or white, active treatment for cancer, medical conditions preventing long-term participation, severe cognitive impairment judged by the telephone interviewer, residence in or inclusion on a waiting list for a nursing home, or inability to communicate in English. For these analyses, we included all participants enrolled between December 18, 2003 and November 1, 2007 (24,465 participants). Among these participants, 4,652 individuals (15.5%) had missing hemoglobin values, 7 individuals had a hemoglobin <6 or >20 g/dL, 10 individuals had missing cognitive data, 38 individuals had missing serum creatinine values, and 57 individuals (0.1%) had self-reported kidney failure or an estimated glomerular filtration rate <10 mL/min/1.73 m², leaving 19,701 participants in the analytic cohort.

Data

The data used in these analyses were obtained from a baseline telephone interview and a subsequent in-home examination conducted approximately 2–4 weeks after interview by a health professional (18). During the telephone interview, demographic characteristics, health conditions, and use of antihypertensive or diabetes medications were obtained. During the subsequent in-home examination, anthropometric measurements and phlebotomy were obtained. Blood pressure was measured twice with the participant seated and the average of the measurements employed. Blood was drawn in the fasting state with samples shipped to the central laboratory at the University of Vermont for determination of serum markers, including hemoglobin, glucose, total cholesterol, C-reactive protein (CRP), and creatinine. The hemogram was performed the day after sample collection by automated cell counting on a Beckman Coulter LH 755 Hematology Workcell (Beckman Coulter, Inc. Fullerton, CA). The interassay coefficient of variation for hemoglobin was ±3.0%. Glucose, cholesterol, and creatinine were measured by colorimetric reflectance spectrophotometry using the Ortho Vitros Clinical Chemistry System 950 IRC (Johnson & Johnson Clinical Diagnostics, Rochester, NY). Serum creatinine measurements were calibrated to a method traceable to creatinine determined by isotope dilution mass spectrometry as previously described (19), and these values were then employed to estimate glomerular filtration rate using the four-variable Modification of Diet in Renal Disease Study equation (20). CRP was measured in batches during enrollment utilizing a validated high-sensitivity particle-enhanced immunonephelometric assay on the BNII nephelometer (N High Sensitivity CRP; Dade Behring Inc, Deerfield, IL) (21).

Starting on December 18, 2003, a six-item cognitive screening examination was incorporated into the REGARDS baseline telephone interview and administered to all participants enrolled on or after that date. Designed for either in-person or telephone administration, the six-item screener (22) is a validated test of global cognitive function that includes recall and orientation items derived from the widely used Mini-Mental State Examination (23). Scores on the six-item screener range from 0 to 6; a score of four or less has a sensitivity of 74.2%–84.0% and a specificity of 80.2%–85.3% in community and clinical samples for a diagnosis of cognitive impairment (22).

Definitions

We evaluated hemoglobin concentration using two different methods. First, we categorized participants as anemic or nonanemic, defined as a hemoglobin <13.0 g/dL among men or <12.0 g/dL among women, according to WHO criteria. We also categorized participants into 1.0 g/dL increments above and below the WHO anemia threshold to assess whether there were linear or threshold associations between hemoglobin concentration and cognitive function relative to the WHO threshold (24).

Prevalent cerebrovascular disease was defined as self-report of stroke. Prevalent coronary heart disease was defined as electrocardiogram evidence of a myocardial infarction or self-report of a myocardial infarction, coronary artery bypass surgery, coronary angioplasty, or coronary stenting. Hypertension was defined as a systolic blood pressure >140 mm Hg, a diastolic blood pressure >90 mm Hg, the use of antihypertensive medications, or self-reported current treatment for hypertension. Diabetes was defined as fasting glucose ≥126 mg/dL, nonfasting glucose ≥200 mg/dL, or self-reported current treatment for diabetes. Elevated cholesterol was defined as a total cholesterol ≥240 mg/dL or self-reported current treatment for elevated cholesterol. Estimated glomerular filtration rate was categorized as <30, 30–59, and ≥60 mL/min/1.73 m², according to National Kidney Foundation recommendations (25). CRP was analyzed as a log-transformed continuous variable. Alcohol and tobacco use were categorized as current, past, or never. Depressive symptoms were assessed with the Center for Epidemiological Studies-Depression four-item scale and depression defined as a score ≥4 (26). Self-reported health status was categorized as poor versus good or better. Geographic region was categorized as Stroke Belt, Stroke Buckle, or other, as previously defined (18).

Analysis

We used t tests, analysis of variance, Mann–Whitney, Kruskal–Wallis, and chi-square as appropriate to test differences between participants with and without anemia. We used logistic regression models to examine the association between hemoglobin concentration and cognitive impairment. First, we determined the association, expressed as the odds ratio (OR) and 95% confidence interval, between hemoglobin concentration and cognitive impairment. Next, we adjusted for age, race, sex, education, and geographic region. In the final model, we adjusted for chronic health conditions, lifestyle factors, depressive symptoms, self-reported health status, estimated glomerular filtration rate categories, and log CRP level in addition to demographic factors. We tested for a linear trend across hemoglobin categories by entering the categorical variable for hemoglobin into the model as an ordinal term. We used interaction terms and conducted stratified analyses to explore the association between potential effect modifiers, such as sex, race, and age in the main multivariable models. All analyses were performed with SAS v 9.1 (Cary, NC). A p value <.05 was considered statistically significant.

RESULTS

The participants included in the analytic cohort had a mean (SD) age of 63.9 (9.7) years, 37.7% were men, and 40.1% were black (Table 1). The mean six-item screener score (SD) was 5.6 (0.7), (ie, on average, less than one item was missed). There were 1,353 individuals (6.9%) who missed two or more items meeting the definition for cognitive impairment.

Table 1.

Characteristics of Participants With and Without Anemia

	All Participants	Anemia	No Anemia	p Value
	(N = 19,701)	(N = 3,087)	(N = 16,614)
Demographics
Age (years)*	63.9 (9.7)	66.0 (10.4)	63.5 (9.5)	<.001
Male sex (%)	37.7	33.3	38.5	<.001
Black race (%)	40.1	66.7	35.5	<.001
Stroke Belt or Buckle (%)	59.3	62.6	58.7	<.001
Education (%)
Less than high school	11.2	18.4	9.8	<.001
High school	26.2	28.7	25.7
Post high school	27.4	26.1	27.7
Professional	35.3	26.9	36.8
Comorbidity
Poor self-reported health (%)	3.4	6.9	2.8	<.001
Prior stroke (%)	5.4	9.0	4.8	<.001
Coronary heart disease (%)	22.0	26.9	21.4	<.001
Diabetes (%)	20.0	37.7	16.7	<.001
Hypertension (%)	57.7	72.4	55.0	<.001
Elevated cholesterol (%)	43.1	41.2	43.5	.02
Current alcohol use (%)	51.9	37.3	54.6	<.001
Current tobacco use (%)	14.7	10.2	15.5	<.001
Depressive symptoms (%)	8.2	10.2	7.8	<.001
Laboratory measurements
Hemoglobin (g/dL)	13.7 (1.5)	11.4 (0.9)	14.0 (1.2)	<.001
Estimated GFR (%)
≥60 mL/min/1.73 m2	90.0	75.6	92.7	<.001
30–59 mL/min/1.73 m2	9.2	21.1	7.0
10–29 mL/min/1.73 m2	0.8	3.3	0.3
C-reactive protein†	2.2 (3.3)	3.0 (3.7)	2.0 (3.3)	<.001

Notes: GFR = glomerular filtration rate.

^*Mean (standard deviation).

^†Geometric mean (SD).

Anemia, defined according to WHO criteria, was present in 15.6% of participants and was more common among women versus men (16.8% vs 13.8%, p < .001) and more common among blacks versus whites (25.3% vs 9.2%, p < .001). In unadjusted analyses, anemia was associated with an increased prevalence of cognitive impairment (OR = 1.91, 95% confidence interval = 1.68–2.17); however, this association was attenuated and no longer significant after adjustment for demographic factors, kidney function, diabetes, hypertension, hyperlipidemia, coronary heart disease, stroke, depressive symptoms, tobacco and alcohol use, health status, and inflammation (OR = 1.05, 95% confidence interval = 0.90–1.22).

The distribution of hemoglobin concentration by sex and race is shown in Figures 1 and 2. The prevalence of cognitive impairment ranged from 16.8% for participants with a hemoglobin concentration ≥2 g/dL below the WHO anemia threshold to 4.3% among those with a hemoglobin concentration >3 g/dL above the WHO anemia threshold (Table 2). Individuals with a hemoglobin concentration ≥2 g/dL below the WHO anemia threshold had an increased odds for cognitive impairment (OR = 1.76, 95% confidence interval = 1.25–2.50) after adjustment for demographic characteristics. However, after additional adjustment for kidney function, diabetes, hypertension, hyperlipidemia, coronary heart disease, stroke, depressive symptoms, tobacco and alcohol use, health status, and inflammation, there was no longer an independent association between reduced hemoglobin concentration and cognitive impairment (Table 2). A test for linear trend across hemoglobin categories was of borderline statistical significance (p value .06).

Figure 1.

Hemoglobin distribution by sex among REasons for Geographic And Racial Differences in Stroke participants. For hemoglobin categories of 6–10, 10–11, 11–12, 12–13, 13–14, 14–15, 15–16, 16–17, and 17–20 g/dL, there were N = 160, 454, 1,449, 3,377, 4,113, 2,149, 503, 61, and 12 women and N = 37, 89, 245, 653, 1,463, 2,316, 1,759, 698, and 171 men.

Figure 2.

Hemoglobin distribution by race among REasons for Geographic And Racial Differences in Stroke participants. For hemoglobin categories of 6–10, 10–11, 11–12, 12–13, 13–14, 14–15, 15–16, 16–17, and 17–20 g/dL, there were N = 134, 379, 1,130, 2,188, 2,180, 1,232, 506, 119, and 32 Blacks and N = 59, 165, 566, 1,853, 3,387, 3,233, 1,758, 637, and 142 Whites.

Table 2.

Association of Hemoglobin Level With Cognitive Impairment

	Hemoglobin Concentration Relative to WHO Threshold for Anemia
Cognitive Impairment	≥2.0 g/dL below	1.0–1.9 g/dL below	<1.0 g/dL below	≤1.0 g/dL above	1.1–2.0 g/dL above	2.1–3.0 g/dL above	>3.0 g/dL above	p Value for Linear Trend
	N = 285	N = 703	N = 2,099	N = 4,818	N = 6,452	N = 3,909	N = 1,435
N (%)	48 (16.8)	84 (12.0)	208 (9.9)	374 (7.7)	395 (6.2)	182 (4.7)	62 (4.3)
Unadjusted OR (95% CI)	3.09 (2.23–4.29)	2.07 (1.16–2.66)	1.68 (1.41–2.00)	1.28 (1.10–1.46)	1.00 (Referent)	0.74 (0.62–0.89)	0.69 (0.52–0.91)	—
Adjusted OR* (95% CI)	1.76 (1.25–2.50)	1.22 (0.94–1.59)	1.06 (0.88–1.28)	1.02 (0.87–1.18)	1.00 (Referent)	0.86 (0.71–1.03)	0.82 (0.62–1.09)	—
Adjusted OR† (95% CI)	1.39 (0.94–2.04)	1.00 (0.75–1.32)	0.97 (0.79–1.17)	0.98 (0.83–1.14)	1.00 (Referent)	0.86 (0.71–1.04)	0.80 (0.59–1.07)	.06

Notes: Results expressed as prevalence (%), OR, and 95% CI. The referent category for these analyses is 13.1–14.0 g/dL among women and 14.1–15.0 g/dL among men. CI = confidence interval; OR = odds ratio; WHO = World Health Organization.

^*Adjusted for age, race, sex, education, and region.

^†Adjusted for age, race, sex, education, region, kidney function, diabetes, hypertension, hyperlipidemia, coronary heart disease, stroke, depressive symptoms, tobacco and alcohol use, health status, and log C-reactive protein level.

Among women, a hemoglobin concentration ≥2 g/dL below the WHO threshold (6–10 g/dL) was significantly associated with an increased odds for cognitive impairment (OR = 1.87, 95% 1.13–3.09), whereas among men, this association was not significant. However, the p value for interaction by sex was not significant (Figure 3A and B). Similarly, there was no evidence of effect modification by race (Figure 3C and D) or median age (p value for interaction terms not significant).

Figure 3.

Adjusted odds ratio (OR) for cognitive impairment among men (A), women (B), whites (C), and blacks (D). The referent category is 13.1–14.0 g/dL among women and 14.1–15.0 g/dL among men. Bars represent 95% confidence intervals. Note: Odds ratios are adjusted for age, race (in sex stratified models), sex (in race stratified models), education, region, kidney function, diabetes, hypertension, hyperlipidemia, coronary heart disease, stroke, depressive symptoms, tobacco and alcohol use, health status, and log C-reactive protein level.

DISCUSSION

There is growing interest in the consequences of anemia; especially in determining the hemoglobin concentration at which cognitive and physical function are optimized and mortality reduced. Based on prior research in clinical populations, we hypothesized that modest reductions in hemoglobin concentration not classified as anemia by the WHO definition would be associated with cognitive impairment in a large biracial sample of community-dwelling U.S. adults. However, we could not confirm a significant association between the WHO definition of anemia or hemoglobin concentration and cognitive impairment. Although we found a trend for a twofold increase in the odds of cognitive impairment as hemoglobin concentration declined from 17–20 g/dL to 6–10 g/dL, this finding was of borderline statistical significance after adjustment for multiple confounding factors. More than 94% of REGARDS participants had hemoglobin concentrations within 2 g/dL of the WHO threshold for anemia, and for these individuals, there was no significant increased odds for cognitive impairment after multivariable adjustment.

Although there is evidence linking specific causes of anemia with impaired cognitive function (eg, B12 deficiency), it is not known whether anemia per se is causally implicated in the development of cognitive impairment. Based on results of several epidemiological studies, it has been speculated that cerebral hypoxia or reduced aerobic capacity as a consequence of chronic anemia may contribute to cognitive decline (27,28). For example, among 1,744 adults over the age of 70 years, Denny and colleagues found that anemia was associated with poorer baseline cognitive function and a higher risk for decline over 8 years (12). In a smaller cohort of elderly women, two studies reported that anemia was associated with poorer executive function at baseline and significant but relatively modest declines in memory over 9 years (11,14). Other observational studies have reported improvements in cognitive function among patients with anemia of chronic disease treated with erythropoiesis-stimulating agents (4,29,30). Alternatively, anemia may simply be a marker of conditions associated with dementia and cognitive impairment, such as inflammation, frailty, and declining health status (31,32), and this may explain why we failed to find an association between anemia and cognitive impairment after adjustment for several of these factors. It is possible that we adjusted for factors in the causal pathway between anemia and cognitive impairment, though it is interesting to note that much of the association was attenuated by adjustment for demographic factors alone.

Little is known about the hemoglobin threshold at which the prevalence of cognitive impairment and the risk of cognitive decline may be increased. We found essentially no significant association between hemoglobin concentrations within 2 g/dL of the WHO threshold and cognitive impairment. There was a modest increase in the odds of impairment below this threshold and a modest decrease in the odds of impairment above this threshold that did not reach statistical significance after adjustment. In a cross-sectional study of 13,000 hospitalized elderly, Zamboni and colleagues found that hemoglobin concentrations below the WHO threshold were associated with poorer cognitive function (3). Similar to our cross-sectional findings, in a prospective study of 1,435 Swedish elderly, Atti and colleagues found that WHO anemia was associated with a nonsignificant 1.3-fold increased risk for dementia (9). The association became statistically significant when a more stringent definition of anemia (based on a hemoglobin less than fifth percentile for sex) was applied. Conversely, when a less stringent definition of anemia was applied (a hemoglobin <25th percentile for sex), the association was not significant. These findings contrast with previous studies of mortality and physical function, where even mild reductions in hemoglobin concentration in nonanemic participants have been shown to be deleterious (16,17). Although this difference may simply reflect larger measurement error of cognitive function versus physical function, it may also suggest that cardiac and skeletal muscle are relatively more sensitive to the effects of anemia as compared with the central nervous system. Because the number of at-risk individuals increases considerably if higher hemoglobin cut points are used to define anemia, these potential differences underscore the need to carefully delineate the degree of anemia that is physiologically and clinically significant.

There are limited data regarding sex and race differences in the consequences of anemia on cognitive function. Denny and colleagues reported that WHO anemia was associated with an increased risk for cognitive decline among women but not among men, wherea the risk for cognitive decline was similar among whites and blacks (12). In the current study, we found a statistically significant association between severe reductions in hemoglobin (6–10 g/dL) and cognitive impairment among women, possibly indicating a threshold effect of hemoglobin with cognitive function. Similar trends were seen among blacks. However, tests for interaction were not significant; thus, this finding should be interpreted with caution. Additional studies are needed to clarify whether anemia may have different consequences for cognitive function among women and blacks.

Limitations of the current study include the cross-sectional design and the lack of information on specific causes of anemia, such as iron, folate, or B12 deficiency. We also lacked information on medications, such as nutritional supplements and erythropoiesis-stimulating agents, which may affect hemoglobin concentration, cognitive function, or both. The six-item screener is a relatively insensitive measure of cognitive function, which may have biased our results toward the null. Use of more sensitive measures of cognitive function to capture individuals with mild cognitive impairment and, in particular, assessment of executive function may strengthen the associations reported here. Measurement error of hemoglobin may have also biased our results toward the null, although it should be noted that the coefficient of variation of hemoglobin in REGARDS was similar or better than other large epidemiological studies. Finally, REGARDS oversampled individuals living in the Stroke Belt and was not intended to be a representative sample of the U.S. population, although there is no reason to believe that these findings would not be generalizable to community-dwelling adults living outside the Stroke Belt.

In summary, in a large group of community-dwelling U.S. adults, 94% had hemoglobin concentrations that were not associated with an increased odds for cognitive impairment after multivariable adjustment.

FUNDING

Cooperative agreement U01 NS041588 from the National Institute of Neurological Disorders and Stroke, National Institutes of Health, Department of Health and Human Service, K23 AG028952 from the National Institute of Aging (M.K.T.), and research grants from Amgen Corporation (M.K.T. and D.G.W.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute of Neurological Disorders and Stroke or the National Institutes of Health. Representatives of the funding agency have been involved in the review of the manuscript but not directly involved in the collection, management, analysis, or interpretation of the data. Amgen did not have any role in the design and conduct of the study, the collection, management, analysis, and interpretation of the data or the preparation or approval of the manuscript. The manuscript was sent to Amgen for review prior to submission for publication.