Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Am J Hematol. Author manuscript; available in PMC 2010 April 8.
Published in final edited form as:
Am J Hematol. 2010 January; 85(1): 57–61.
doi:  10.1002/ajh.21577
PMCID: PMC2851744

Ethnic differences in anemia among patients with diabetes mellitus: The Diabetes Study of Northern California (DISTANCE)


To examine ethnic differences in hemoglobin testing practices and to test the hypothesis that ethnicity is an independent predictor of anemia among patients with diabetes mellitus. We conducted a panel study to assess the rate of hemoglobin testing during 1999–2001 and the period prevalence and incidence of anemia among 79,985 adults with diabetes mellitus receiving care within Kaiser Permanente of Northern California. Anemia was defined as hemoglobin <13.0 g/dL in men or < 12.0 g/dL in women. Overall, 82.1% of the cohort was tested for anemia at least once during the 3-year study period. Mixed ethnicity patients were most likely to be tested, followed by whites, blacks, Latinos, and Asians (P < 0.0001). Fifteen percent of the cohort had prevalent anemia at baseline, and an additional 22% of those tested developed anemia during the study period. Anemia was more prevalent among blacks and mixed ethnicity persons compared with other racial/ethnic groups. Anemia was also more prevalent among those ≥70 years of age or with estimated glomerular filtration rate <60 ml/min/1.73 m2. In multivariable models, blacks had higher and Asians had lower odds of prevalent anemia and hazard ratios of incident anemia compared with whites. Within a large, diverse cohort with diabetes, ethnicity was predictive of anemia, even after adjustment for age, level of kidney function, and other potential confounders. Blacks with diabetes are at increased risk of anemia relative to whites. These differences may account for some of the observed ethnic disparities in diabetes complications.


Anemia is a well-known complication of chronic kidney disease and end stage renal disease requiring renal replacement therapy. Lower hemoglobin level is also associated with higher risks for left ventricular hypertrophy development and progression, heart failure [1], cognitive impairment [2], poorer quality of life, and death [1]. In patients with diabetes complicated by microvascular disease, hemoglobin level declines in parallel with declining glomerular filtration rate (GFR), while remaining relatively stable in those with normoalbuminuria and GFR > 90 ml/min/1.73 m2 [3]. Several studies have suggested that anemia occurs earlier in the course of diabetic nephropathy than in other forms of kidney disease, and anemia has been described among patients with mild to moderate chronic kidney disease (e.g., GFR 30–59 ml/min/1.73 m2) as well as among those with more severe kidney disease [4]. Hypotheses to explain a higher rate of anemia among people with diabetes include damage to renal architecture produced by chronic hyperglycemia and consequent formation of advanced glycation end products [5]; autonomic neuropathy with decreased splanchnic sympathetic stimulation of erythropoietin production [6]; a blunted erythropoietin response to anemia among erythropoietin-producing cells [7]; and systemic inflammation [8]. Prior population-based studies in the United States have shown that chronic kidney disease-associated anemia is more prevalent among blacks than whites [9]. This disparity is only partially explained by differences in socioeconomic status and access to and quality of medical care [10,11]. Thus, differences in anemia incidence among patients with diabetes could potentially explain some of the ethnic disparities observed in diabetic outcomes. However, relatively little is known about ethnic disparities of anemia among subjects with diabetes, particularly after accounting for potential differences in underlying level of kidney function. To address these knowledge gaps, we examined overall and ethnic-specific hemoglobin testing practices, and the prevalence and predictors of anemia among patients with diabetes mellitus who received care in a large integrated health care delivery system. We examined the hypothesis that ethnicity is an independent predictor of anemia among patients with diabetes.



This panel study was conducted among the membership of Kaiser Permanente of Northern California, an integrated health care delivery system providing comprehensive care for ~3.2 million health plan members (as of November 2002) in a 14-county region of Northern California, including the San Francisco Bay and Sacramento metropolitan areas. Kaiser Permanente membership is highly representative of the local underlying population, except for slightly lower representation of Kaiser membership at the extremes of income [12]. Its ethnic composition is also very similar to that of the US census enumerated population in the Bay Area Metropolitan Statistical Area (MSA): 7% African American versus 9% in the underlying MSA population, 10% versus 11% Asian/Pacific Islander, 10% versus 14% Hispanic, and 71% versus 64% Caucasian.

This study was approved by the Kaiser Permanente Institutional Review Board.

Source population

This study is one of a series included in the Diabetes Study of Northern California (DISTANCE). The goal of DISTANCE is to evaluate social disparities in care and outcomes associated with diabetes. This research is based on the Kaiser Permanente Northern California Diabetes Registry (“the registry”), a longitudinal, validated registry of clinically recognized diabetes mellitus that has been maintained continuously since its establishment in October 1993 [12]. The registry is a well-established source of data for study of the epidemiology and health services aspects of diabetes. Eligibility is based on multiple sources of case ascertainment including outpatient prescriptions for diabetes medications, glycosylated hemoglobin (HbA1c) ≥ 7%, and outpatient, emergency room and hospitalization records listing diagnostic code(s) for diabetes. The registry was 99.5% sensitive for diagnosed diabetes (compared with survey-derived self-report) as of January 2002. All automated clinical information (pharmacy, laboratory, outpatient and inpatient diagnoses and procedures, outpatient visits, and hospitalizations) is downloaded to the registry annually to provide a comprehensive, longitudinal follow-up of each registry member.

Our study sample comprised a fixed cohort of diabetes registry members who were 18 years or older as of January 1, 1999 and maintained continuous KP membership during the 3-year study period, January 1, 1999 through December 31, 2001 (n = 91,129). After excluding 5,234 members without health plan pharmacy benefits (as we do not have data on prescriptions filled in non-Kaiser pharmacies), 330 who were pregnant during the study period, 5,235 without at least one outpatient clinic visit during the study period, and 345 who had a previous kidney transplant, our study sample comprised 79,985 adults with diabetes.

Study variables

We established whether each subject obtained at least one outpatient (nonemergency department) test of hemoglobin during the study period. We included those with previously diagnosed anemia or those who received erythropoiesis-stimulating agents (EPOs), as these patients should have had at least one hemoglobin test per year to follow their response to therapy [13].

We ascertained the date during the study period on which the first hemoglobin value diagnostic of anemia was obtained. We employed the World Health Organization’s gender-specific definitions of anemia: Hemoglobin <13.0 g/dL in men and <12.0 g/dL in women [14]. Prior anemia was defined as any outpatient visit diagnosis or laboratory-defined anemia during the 2 years prior to baseline.

Study variables were derived from electronic records or from a 1994 to 1997 survey of all health plan members with known diabetes. The survey, which was conducted by mail or computer-assisted telephone interview, achieved an 83% response rate. Survey methods and characteristics of survey responders have been described previously [12]. Self-reported ethnicity (white, black, Latino, Asian, Pacific Islander, Native American, other, and mixed ethnicity) was assessed on the survey. Age at baseline was determined from administrative records. Laboratory values, including serum creatinine, estimated GFR (mL/min/1.73 m2) estimated from the abbreviated Modification of Diet in Renal Disease (MDRD) equation [=186 × (serum creatinine, mg/dL)−1.154 × (age, year)−0.203 × (0.742 if female) × (1.210 if black)] [15], and urinary albumin excretion were determined from electronic laboratory records. Microalbuminuria was defined as urinary albumin excretion ratio 30–300 mg/g creatinine on random outpatient urine and proteinuria was defined as value >300 mg/g creatinine [16]. EPO, oral diabetes medication, and insulin use were determined from electronic pharmacy records. To control for differences in comorbidity burden, we calculated an inpatient and outpatient utilization adjuster [17].

We assessed laboratory values, medication use, and utilization adjusters for the entire cohort at baseline, defined as the first quarter of observation for medication use and utilization and the first value obtained after baseline for laboratory tests. We employed this definition (Table I) to compare differences in baseline characteristics by hemoglobin testing. Models of incident anemia were limited to those patients who had a hemoglobin test. As medication exposure and laboratory parameters most relevant to associations with anemia are those obtained immediately proximal to anemia diagnosis, we assessed laboratory values and medication use, and utilization adjuster during the 3-month period prior to the lowest hemoglobin test when conducting multivariate predictive models of anemia.

Baseline Subject Characteristics, Hemoglobin Testing Rates, and Prevalence of Anemia Based on WHO Gender-Specific Criteria

Analytic methods

We calculated the prevalence of hemoglobin testing (number of subjects who had ≥1 hemoglobin test during the 3-year study period divided by the total number of subjects) overall and in subgroups stratified by ethnicity, gender, and other relevant baseline variables (e.g., baseline GFR status). Chi-squared tests were employed to examine differences in prevalence of hemoglobin testing across these subgroups.

We estimated anemia period prevalence, defined as the number of subjects with laboratory-diagnosed anemia divided by the number of subjects who had Hb testing during the study period. Chi-squared tests were employed to examine differences in anemia period prevalence by level of baseline covariates.

We employed standard descriptive statistics to examine bivariate associations between potential predictor variables and the outcome of anemia among cohort members with at least one hemoglobin test during the study period, including those with anemia diagnosis prior to baseline. We then specified multivariable logistic regression models to identify predictors of anemia. Variables that had significant associations with anemia were included in our multivariate model. Patients reporting Native American or Pacific Islander ethnicity were combined with those reporting “Other” ethnicity in logistic regression models due to small subgroup sizes. Our multivariable models included variables for race/ethnicity, age group, gender, estimated GFR, urinary albumin excretion, use of oral diabetes medications, and insulin (as measures of diabetes disease severity), EPO use, inpatient and outpatient utilization, and facility. We also specified models stratified by self-reported ethnicity to examine ethnic-specific predictors of anemia. As anemia was a common outcome, we calculated predicted probabilities of anemia from the models based on estimated least square means. Finally, we performed sensitivity analyses assessing the association of anemia with these variables assessed during the first 3 months after baseline.

We further examined predictors of anemia in Cox proportional hazards (semiparametric time-to-event) models. Patients with anemia prior to baseline were excluded from these models so as to limit confounding by prior treatment, as well as potential confounding by different prevalence of hemoglobinopathies by ethnicity. All analyses were conducted in SAS version 9.1 (SAS Institute, Cary, NC).


Study subject characteristics

Among 79,985 adults with diabetes, mean age was 60.4 ± 12.7 years, 52.5% were men, and there was substantial ethnic diversity (Table I). The frequency of estimated GFR < 60 ml/min/1.73 m2 was 25.0%, and 39.1% of the cohort had microalbuminuria or proteinuria at baseline. Overall, 15.4% of cohort members were diagnosed with anemia at or prior to cohort entry. Among the study cohort, 20.7% used insulin and 46.8% oral diabetes medications only, and 1.2% were taking EPO at baseline.

Hemoglobin testing practices

Overall, 65,696 members of the cohort (82.1%) had hemoglobin testing at least once during the 3-year study period (Table I). In bivariate analyses, hemoglobin testing was more common with female gender, older age, lower estimated GFR, and increased urinary albumin excretion. Patients using EPO or insulin at baseline also were more likely to receive hemoglobin testing. Those reporting mixed ethnicity were most likely to be tested, followed by blacks, whites, Latinos, Asians, and other ethnic groups. Those who were taking diabetes medications or EPO were also more likely to be tested (P < 0.0001 for all comparisons).

Prevalence and predictors of anemia

Among 65,969 tested subjects, the period prevalence of anemia was 34.7%. Of the 22,812 subjects with laboratory-diagnosed anemia during the study period, 8,922 had been diagnosed with anemia prior to baseline and an additional 13,890 had incident anemia. There were no substantive differences in the direction or magnitude of associations between the identified predictors of prevalent or incident anemia across patient subgroups, so we present only results for the entire tested cohort. Prevalent anemia was more common among blacks (48.5%) and persons of mixed ethnicity (41.6%) than among whites (33.6%), Latinos (31.2%), Asians (28.0%), and those of other ethnicity (22.1%; P < 0.001; Table I). As expected, the prevalence of anemia increased with age and with each decrement of GFR or increase in urinary albumin excretion. Anemia occurred in 25.1% of those with baseline estimated GFR ≥ 90 ml/min/1.73 m2 compared with 95.2% of those with estimated GFR < 15 ml/min/1.73 m2 or on dialysis. Anemia was similarly as common among those whose diabetes was treated only with diet (30.1%) or oral medications (33.1%), but more frequent among those on insulin (45.1%). As expected, anemia was more frequent among those who used EPO at baseline (83.0% vs. 34.0%) and those who had anemia diagnosed in the 2 years prior to baseline (76.9% vs. 25.7%, P < 0.0001 for all comparisons). When we repeated these analysis defining anemia prevalence as number of subjects diagnoses with anemia divided by total number in cohort (including those not tested), we found similar trends (data not shown).

In multivariable logistic regression models of prevalent anemia adjusted for gender, age, clinic site, utilization, and proximate estimated GFR, albuminuria, diabetes treatment and EPO use, odds of anemia were higher among blacks [odds ratio (OR) 2.26, 95% confidence interval (CI) 2.12–2.41], those of mixed (OR 1.37, 95% CI 1.31–1.44) or Latino (OR 1.15, 95% CI 1.07–1.24) ethnicity, compared with whites. Asians and those of other ethnicity had odds of anemia that were not significantly different than those of whites (Table II). Odds of anemia increased with age 70 years or older but there were no differences by gender. As expected, there was a graded, increased adjusted odds of prevalent anemia with estimated GFR below 60 ml/min/1.73 m2 and abnormal urinary albumin excretion, as well as with EPO use (OR 5.27, 95% CI 4.53–6.15). Insulin use (OR 1.31, 95% CI 1.25–1.39) was associated with modestly increased odds of anemia. Since anemia is not a rare outcome in this cohort, we estimated predicted probabilities using the SAS PROC GENMOD LSMEANS option. These estimates are presented along with the odds of prevalent anemia, in Table II. Excluding patients who were using EPO at baseline or those with diagnosis of anemia prior to baseline did not substantively alter our findings.

Predictors of Prevalent Anemia and Incident Anemia Among 65,696 Patients with Diabetes Mellitus who Received Hemoglobin Testing

In multivariable-adjusted (gender, age, clinic, utilization and proximate GFR, albuminuria, EPO, and diabetes medication use) models of time to incident anemia (Table II), ethnicity was a significant predictor of time to incident anemia. Compared with whites, blacks and mixed ethnicity people had higher risk of incident anemia (HR 1.62, 95% CI 1.52–1.72 and HR 1.23, 95% CI 1.17–1.28, respectively), whereas Latinos and Asians had risk similar to that of whites. As expected, increased age, decreased GFR, and increased albuminuria were associated with significantly higher risk of incident anemia among diabetes patients. The results of sensitivity analyses employing baseline rather than proximate values for laboratory and medication use revealed no substantive differences in hazard ratio estimates of time to incident anemia (data not shown).


Within a large, multiethnic cohort of persons with diabetes who have access to medical care, we found relatively high rates of hemoglobin testing (82.1% over a 3-year period). Ethnicity was a significant predictor of prevalent and incident anemia, and remained so after controlling for age, gender, estimated GFR, albuminuria, diabetes medication use, and health care utilization. Compared with whites, blacks had substantially higher prevalence and hazard of incidence of anemia, followed by those reporting mixed ethnicity and Latinos, whereas Asians and those of other ethnicity had odds of anemia (and hazard of incident anemia) not significantly different than those of whites. As expected, anemia was more prevalent with older age, reduced kidney function, documented albuminuria, and EPO use.

Findings suggest that, after controlling for renal impairment, blacks with diabetes are at increased risk of anemia relative to whites. These differences may account for some of the observed disparity in anemia prevalence in the wider population (i.e., one that includes nondiabetic patients). Denny et al. [18] and the REGARDS renal ancillary study [19] similarly found that blacks had three times the odds of anemia than whites, after controlling for age, educational attainment, GFR, and comorbidities, including diabetes. However, those studies were not limited to diabetes patients. Our group [12] and others [20] have previously demonstrated that diabetic nephropathy occurs at higher rates among Latinos, Asians, and Blacks than Whites.

Our study has several strengths [1]. We examined data on a large, well-characterized population of diabetes patients drawn from an insured population that is representative of the underlying population in ethnicity and socioeconomic status, giving our study broad applicability [2]. We were able to control for a number of key covariates that could be associated with ethnicity and anemia, including estimated GFR, diabetes medication use, a comorbidity score, anemia treatment (EPO), and gender [3]. As we capture electronic data that is uniformly available, the possibility of bias due to study enrollment is minimal [4]. Our findings were robust, even in models that excluded those with prior anemia. Exclusion of those with prior anemia would be expected to reduce potential bias due to differential prevalence of hemoglobinopathies by ethnicity.

Our study also had several limitations. We were unable to characterize the specific cause of anemia, or to collect systematic data on known variation in conditions associated with anemia across ethnic groups in the general population. For instance, the gene frequency of alpha-thalassemia deletion allele is 0.169 among African-Americans, and iron deficiency is more common among black than white women [21]. Data on hemoglobin testing were collected from patients in the course of regular medical care rather than through a systematic screening program, and 21% of diabetic patients did not have hemoglobin testing during the study period. This may introduce bias as those patients who did not have hemoglobin tested may be generally healthier than those who did have hemoglobin tested. The average hemoglobin levels we report may thus be lower than the actual levels among the entire population with diabetes. Further, ethnic differences in hemoglobin testing may introduce biases in evaluation of ethnic variation in anemia prevalence or mean hemoglobin levels. Although multiple studies among the KP membership have demonstrated that access to care is uniform across ethnicity, use of care may differ by ethnicity or other factors [22]. Additionally, there is new evidence that the MDRD equation may slightly underestimate GFR in diabetic patients with nephropathy [23], and this may result in some misclassification of the degree of renal impairment among this diabetic cohort. However, since we are only evaluating anemia relationships within each of these rather broad GFR categories, this misclassification would unlikely distort our findings. Further, the MDRD equation accounts for black–white differences, but does not account for differences between Latinos, Asians, or other ethnic groups {2074}. Finally, we do have individual-level data on income for any of the study cohort, and our data on educational attainment are incomplete. It would have been of great interest to examine whether income, educational attainment, or other measures of socioeconomic status confounded any of our findings.

Our findings add to the extant body of literature documenting higher prevalence of anemia among US blacks than US whites. Anemia is associated with increased mortality among elderly people across all ethnic groups [18]. Studies of the effect of EPO to correct anemia to normal hemoglobin levels have shown conflicting results on quality of life. Existing randomized clinical evidence does not support a therapeutic impact of these agents on the risk of cardiovascular events and death in certain populations [24,25]. The TREAT trial should help to inform the clinical benefits and risks of anemia correction with EPOs in patients with diabetic nephropathy.

Overall, future research is needed to delineate the reasons for observed variation in the development of anemia across ethnic groups with diabetes and whether the effects of therapeutic strategies to correct anemia differ by a patient’s ethnicity.


The content is solely the responsibility of the authors and do not necessarily represent the official views of the National Institutes of Health.

Contract grant sponsor: National Institute of Diabetes and Digestive and Kidney Diseases; Contract grant number: R01 DK065664.

Contract grant sponsor: Kaiser Permanente Northern California.


Conflict of interest: Nothing to report.


1. Horl WH, Ertl G. Anaemia and the heart. Eur J Clin Invest. 2005;35(Suppl 3):20–25. [PubMed]
2. Atti AR, Palmer K, Volpato S, Zuliani G, Winblad B, Fratiglioni L. Anaemia increases the risk of dementia in cognitively intact elderly. Neurobiol Aging. 2006;27:278–284. [PubMed]
3. Thomas MC, Tsalamandris C, Macisaac RJ, Jerums G. The epidemiology of hemoglobin levels in patients with type 2 diabetes. Am J Kidney Dis. 2006;48:537–545. [PubMed]
4. Kazmi WH, Kausz AT, Khan S, Abichandani R, Ruthazer R, Obrador GT, Pereira BJ. Anemia: An early complication of chronic renal insufficiency. Am J Kidney Dis. 2001;38:803–812. [PubMed]
5. Thomas MC, Forbes JM, Cooper ME. Advanced glycation end products and diabetic nephropathy. Am J Ther. 2005;12:562–572. [PubMed]
6. Spallone V, Maiello MR, Kurukulasuriya N, Barini A, Lovecchio M, Tartaglione R, Mennuni G, Menzinger G. Does autonomic neuropathy play a role in erythropoietin regulation in non-proteinuric type 2 diabetic patients? Diabet Med. 2004;21:1174–1180. [PubMed]
7. Cotroneo P, Maria RB, Todaro L, Pitocco D, Manto A, Ruotolo V, Storti S, Damiani P, Caputo S, Ghirlanda G. Blunted erythropoietin response to anemia in patients with type 1 diabetes. Diabetes Metab Res Rev. 2000;16:172–176. [PubMed]
8. Cottone S, Lorito MC, Riccobene R, Nardi E, Mule G, Buscemi S, Geraci C, Guarneri M, Arsena R, Cerasola G. Oxidative stress, inflammation and cardiovascular disease in chronic renal failure. J Nephrol. 2008;21:175–179. [PubMed]
9. Dallman PR, Yip R, Johnson C. Prevalence and causes of anemia in the United States, 1976 to 1980. Am J Clin Nutr. 1984;39:437–445. [PubMed]
10. Lea JP, Norris K, Agodoa L. The role of anemia management in improving outcomes for African-Americans with chronic kidney disease. Am J Nephrol. 2008;28:732–743. [PMC free article] [PubMed]
11. Peralta CA, Ziv E, Katz R, Reiner A, Burchard EG, Fried L, Kwok PY, Psaty B, Shlipak M. African ancestry, socioeconomic status, and kidney function in elderly African Americans: A genetic admixture analysis. J Am Soc Nephrol. 2006;17:3491–3496. [PubMed]
12. Karter AJ, Ferrara A, Liu JY, Moffet HH, Ackerson LM, Selby JV. Ethnic disparities in diabetic complications in an insured population. JAMA. 2002;287:2519–2527. [PubMed]
13. KDOQ1. Clinical Practice Guideline and Clinical Practice Recommendations for anemia in chronic kidney disease: 2007 update of hemoglobin target. Am J Kidney Dis. 2007;50:471–530. [PubMed]
14. World Health Organization. Report of a WHO Scientific Group. Geneva: WHO; 1968. Nutritional Anaemia.
15. Levey AS, Coresh J, Greene T, Stevens LA, Zhang YL, Hendriksen S, Kusek JW, van Lente F. Using standardized serum creatinine values in the modification of diet in renal disease study equation for estimating glomerular filtration rate. Ann Intern Med. 2006;145:247–254. [PubMed]
16. National Kidney Foundation. Clinical practice guidelines for chronic kidney disease: Evaluation, classification, and stratification. Am J Kidney Dis. 2002;39:S1–S266. [PubMed]
17. Schneeweiss R, Rosenblatt RA, Cherkin DC, Kirkwood CR, Hart G. Diagnosis clusters: A new tool for analyzing the content of ambulatory medical care. Med Care. 1983;21:105–122. [PubMed]
18. Denny SD, Kuchibhatla MN, Cohen HJ. Impact of anemia on mortality, cognition, and function in community-dwelling elderly. Am J Med. 2006;119:327–334. [PubMed]
19. Zakai NA, Mcclure LA, Prineas R, Howard G, Mcclellan W, Holmes CE, Newsome BB, Warnock DG, Audhya P, Cushman M. Correlates of anemia in American blacks and whites: The REGARDS Renal Ancillary Study. Am J Epidemiol. 2009;169:355–364. [PMC free article] [PubMed]
20. Young BA, Maynard C, Boyko EJ. Racial differences in diabetic nephropathy, cardiovascular disease, and mortality in a national population of veterans. Diabetes Care. 2003;26:2392–2399. [PubMed]
21. Beutler E, West C. Hematologic differences between African-Americans and whites: The roles of iron deficiency and alpha-thalassemia on hemoglobin levels and mean corpuscular volume. Blood. 2005;106:740–745. [PubMed]
22. Martin TL, Selby JV, Zhang D. Physician and patient prevention practices in NIDDM in a large urban managed-care organization. Diabetes Care. 1995;18:1124–1132. [PubMed]
23. Rossing P, Rossing K, Gaede P, Pedersen O, Parving HH. Monitoring kidney function in type 2 diabetic patients with incipient and overt diabetic nephropathy. Diabetes Care. 2006;29:1024–1030. [PubMed]
24. Drueke TB, Locatelli F, Clyne N, Eckardt KU, Macdougall IC, Tsakiris D, Burger HU, Scherhag A. Normalization of hemoglobin level in patients with chronic kidney disease and anemia. N Engl J Med. 2006;355:2071–2084. [PubMed]
25. Ritz E, Laville M, Bilous RW, O’Donoghue D, Scherhag A, Burger U, De Alvaro F. Target level for hemoglobin correction in patients with diabetes and CKD: Primary results of the Anemia Correction in Diabetes (ACORD) Study. Am J Kidney Dis. 2007;49:194–207. [PubMed]