OBJECTIVE

Lowering hemoglobin A1c to <7% reduces the risk of microvascular complications of diabetes, but the importance of maintaining this target in diabetes patients with kidney failure is unclear. We evaluated the relationship between A1c levels and mortality in an international prospective cohort study of hemodialysis patients.

RESEARCH DESIGN AND METHODS

Included were 9,201 hemodialysis patients from 12 countries (Dialysis Outcomes and Practice Patterns Study 3 and 4, 2006–2010) with type 1 or type 2 diabetes and at least one A1c measurement during the first 8 months after study entry. Associations between A1c and mortality were assessed with Cox regression, adjusting for potential confounders.

RESULTS

The association between A1c and mortality was U-shaped. Compared with an A1c of 7–7.9%, the hazard ratios (95% CI) for A1c levels were 1.35 (1.09–1.67) for <5%, 1.18 (1.01–1.37) for 5–5.9%, 1.21 (1.05–1.41) for 6–6.9%, 1.16 (0.94–1.43) for 8–8.9%, and 1.38 (1.11–1.71) for ≥9.0%, after adjustment for age, sex, race, BMI, serum albumin, years of dialysis, serum creatinine, 12 comorbid conditions, insulin use, hemoglobin, LDL cholesterol, country, and study phase. Diabetes medications were prescribed for 35% of patients with A1c <6% and not prescribed for 29% of those with A1c ≥9%.

CONCLUSIONS

A1c levels strongly predicted mortality in hemodialysis patients with type 1 or type 2 diabetes. Mortality increased as A1c moved further from 7–7.9%; thus, target A1c in hemodialysis patients may encompass values higher than those recommended by current guidelines. Modifying glucose-lowering medicines for dialysis patients to target A1c levels within this range may be a modifiable practice to improve outcomes.

Diabetes is present in more than 66% of U.S. hemodialysis patients and is a major contributor to the increased morbidity and mortality in this population (1). Optimal management of glycemia in diabetic hemodialysis patients, however, is uncertain. Although hemoglobin A1c is the standard measure of glycemic control in diabetes, its interpretation in dialysis patients may be compromised by reduced red cell life span and the use of exogenous erythropoietin (2,3). Moreover, published findings on the association between A1c and clinical outcomes in diabetic hemodialysis patients are conflicting (46), and current guidelines for the management of these patients are based primarily on data from nondialysis patients (7,8). We reported previously that an A1c level >7.3% was associated with increased mortality in the Japan Dialysis Outcomes and Practice Pattern Study (9). Whether findings in the Japanese diabetic hemodialysis patients are relevant in other ethnic groups is uncertain, however, and suggests the need for larger multinational studies to evaluate appropriate A1c levels in the hemodialysis population.

In this study, we estimate the effect of glycemic control, based on A1c level, on all-cause mortality in the Dialysis Outcomes and Practice Patterns Study (DOPPS). DOPPS is a prospective cohort study of randomly selected in-center hemodialysis patients from a representative sample of facilities within each of 12 countries (Australia, Belgium, Canada, France, Germany, Italy, Japan, New Zealand, Spain, Sweden, the U.K., and the U.S.).

Study population

Across the DOPPS phases, the study design and methodology were structured with the following key elements: 1) random selection of dialysis units stratified by type of facility and geographic region in each country; 2) collection of demographic data, diabetes as cause of end-stage renal disease (ESRD), and mortality data for all hemodialysis patients in each study unit; 3) collection of detailed patient data from a random selection of 20 to 40 patients within each dialysis unit at study entry and at 4-month intervals; 4) collection of kidney disease quality of life information; and 5) collection of facility practice information, determined from questionnaires completed annually by the dialysis unit’s medical director (medical directors survey) and by the unit’s nurse manager.

The methodologies of sample selection and data collection were substantially similar in phases 3 and 4, with possible differences in operational aspects, such as the proportion of data collected electronically. Many of the selected facilities in phase 3 continued participation in phase 4, and there was overlap in the patient populations but not in the follow-up periods.

The DOPPS 3 (2005–2008) and DOPPS 4 (2008–2011) populations included 12,954 patients with type 1 or type 2 diabetes. Patients were coded as having diabetes if they had a diagnosis of type 1 or type 2 diabetes, if they had received medications for diabetes (insulin or oral) before enrollment, if they had diabetic gastroparesis, or if they were marked as “diabetic” on the patient census, which is a complete listing of patients dialyzed in the participating facility. There were 28,458 patients in the DOPPS phases 3 and 4 sample, of whom 12,954 had diabetes. An additional 296 patients were excluded because of missing dates for age, study entry, or study exit. The population for this study consisted of 9,201 patients with type 1 or type 2 diabetes after exclusion of patients with no A1c measurements during the first 8 months of DOPPS follow-up.

Covariates

The primary analyses were performed with the first available A1c value in those with at least one measurement during the first 8 months of data collection. In our data collection design, laboratory values and medication use were reported at study entry and by interval summary forms collected every 4 months afterward. Covariate data were collected at study entry on patient age, sex, race (Black vs. other races), BMI, albumin, years of dialysis, creatinine, 12 comorbid conditions, insulin use, oral diabetes medication use, LDL cholesterol, hemoglobin, country, and study phase.

Missing data for covariates (potential confounders), described in the footnote of Table 1, were accounted for by using multiple imputation on the population of patients with at least one valid A1c measurement, as implemented by the IVEware program (10), and analyzed with the MIAnalyze procedure in SAS STAT 9.2 (11). Missing A1c levels were not imputed, because patients whose physicians did not order measurement of A1c levels are likely to differ in important ways from patients whose physicians did order the test, invalidating the assumptions behind missing data imputation. Analytical results derived from the multiply imputed data were compared with results obtained with other methods for dealing with missing data (including complete case analyses and with missing data indicators with single-value imputation). The differences were trivial (most of the hazard ratio [HR] estimates were within 0.03; the largest difference was 0.08) and did not affect our conclusions. Further sensitivity testing of the effects of missing data for variables missing in more than 10% of the patients (BMI, weight loss, and cholesterol) consisted of imputation without these variables and without controlling for these variables and of complete case analyses excluding these variables. In general, the relationships between A1c and mortality were similar among all three methods.

Table 1

Clinical and demographic characteristics by number and A1c category

Clinical and demographic characteristics by number and A1c category
Clinical and demographic characteristics by number and A1c category

Certain models evaluated patients with and without indicators of poor nutrition, or with and without recent diabetes treatment. For these models, patients were identified as having poor nutritional status if they had any one or more of the following factors (n = 2,037): BMI <19 kg/m2, weight loss during the first 8 months of DOPPS follow-up at a rate equivalent to 10% of body weight per year; albumin <3.0 mg/dL, or cachexia. Patients were identified as recently treated for diabetes if at the start of follow-up they were receiving oral diabetes medicines (n = 768), insulin (n = 2,685), or both (n = 293). No diabetes medications were recorded for 3,477 patients, and 1,978 patients had insufficient drug information for coding before imputation.

Analyses

Standard descriptive statistics were used to characterize the DOPPS patients with type 1 or type 2 diabetes. Follow-up started at A1c measurement and ended at the time of death (outcome event), 7 days after transfer from the facility, or as of the date of most recent data availability (December 2011 or earlier), whichever came first.

The effect of A1c on all-cause mortality was examined by Cox proportional hazards analyses. All models were adjusted for patient age, sex, race (Black vs. other races), BMI, hemoglobin, albumin, years of dialysis, creatinine (because this is a marker of dietary protein intake and muscle mass and is a predictor of clinical outcome in ESRD patients), 12 comorbid conditions, insulin use, LDL cholesterol, country, and study phase.

Tests of the interaction between A1c and either nutritional status or diabetes treatment involved likelihood ratio tests of each covariate multiplied by the six indicator variables of A1c categories, although effect estimates and confidence limits were produced with separate models for each population (i.e., with and without diabetes treatment, or with and without indicators of poor nutritional status). The assumption of proportional hazards for the A1c categories was evaluated by visual inspection of the log(−log[survival]) versus log(time) plot and by testing the interaction between these categories and log(time), yielding P = 0.03. Although the relationship was slightly stronger during earlier periods of follow-up for mortality, there was reasonable adherence to proportional hazards.

Sensitivity analyses

Sensitivity analyses were performed by comparing the results obtained through the approach described here with models, with mean A1c during the first 8 months of DOPPS follow-up as the main predictor and restricting to 6,669 patients with at least two A1c measurements during that baseline period. These analyses were also adjusted for weight loss during the 8-month period. Because follow-up for this latter analysis started after the baseline period, the duration of follow-up was on average about 4 months shorter than in the main analysis. Serum glucose concentration was also examined, but it had a much weaker relation with mortality, presumably because serum glucose levels are more easily influenced by short-term factors.

Another sensitivity analysis tested the impact of dialysis information on the adjusted relationship between A1c and mortality. The dialysis information included Kt/V (dialyzer clearance of urea × dialysis time/volume of distribution of urea), vascular access type, number of sessions per week, and dialysis duration.

Clinical and demographic characteristics

The first available A1c ranged from 3.1 to 19.2%, whereas the mean A1c across 8 months ranged from 3.4 to 15.0%. Table 1 shows clinical and demographic characteristics of the patients with type 1 or type 2 diabetes according to the number of A1c measurements made during the initial 8-month baseline period, as well as the first available A1c level among those with at least one A1c measurement during that period. Patients with more A1c measurements during the first 8 months were treated more frequently with insulin or oral medicine than those with fewer A1c measurements and differed with respect to a number of other demographic and comorbid conditions. In addition, patients in whom A1c levels were not measured had a slightly lower prevalence of peripheral vascular disease and were more often Black. Among patients with at least one measurement of A1c, those with higher initial A1c values were also treated more frequently with insulin than were those with lower values; on average they were younger, more likely to be Black, higher in BMI, and were likely to have initiated dialysis more recently.

A1c and mortality

Among patients with at least one measurement of A1c, average follow-up was 1.4 years (1st–99th percentile, 8 days–3.3 years). The death rate was 0.16/year (1,983 deaths /12,513 patient-years). Among these patients (the main analysis), the adjusted mortality rate was lowest at A1c levels of 7–7.9% (Fig. 1A). A log likelihood test for nonlinearity showed an overall P value of 0.01 for the combined contribution of the squared and cubic A1c terms to the linear model between A1c and mortality. The relation was similar in shape in an unadjusted analysis, though slightly steeper at lower A1c levels and less steep at higher levels.

Figure 1

A: Risk of mortality by initial A1c, adjusted for age, sex, race, BMI, years of dialysis, albumin, creatinine, 10 comorbid conditions, insulin use, hemoglobin, HDL cholesterol, country, and study phase. B: Risk of mortality by mean A1c, adjusted for age, sex, race, BMI, years of dialysis, albumin, creatinine, 10 comorbid conditions, insulin use, hemoglobin, HDL cholesterol, country, and study phase. (A high-quality color representation of this figure is available in the online issue.)

Figure 1

A: Risk of mortality by initial A1c, adjusted for age, sex, race, BMI, years of dialysis, albumin, creatinine, 10 comorbid conditions, insulin use, hemoglobin, HDL cholesterol, country, and study phase. B: Risk of mortality by mean A1c, adjusted for age, sex, race, BMI, years of dialysis, albumin, creatinine, 10 comorbid conditions, insulin use, hemoglobin, HDL cholesterol, country, and study phase. (A high-quality color representation of this figure is available in the online issue.)

Close modal

Sensitivity analyses

Our sensitivity analysis used the mean A1c during the first 8 months of DOPPS follow-up, and this showed a death rate of 0.20 deaths/year (1,337 deaths/6,679 patient-years). When examined in a Cox model, the mortality rates were similar among patients with zero, one, or two measures of A1c during the initial 8 months of follow-up when restricted to patients who survived through all 8 months (HR for 0 vs. 2, 1.01; 95% CI 0.89–1.15; , 1 vs. 2, 1.09; , 0.95–1.25), with adjustment for phase and country. The conclusions reached with mean A1c were not different from those based on the initial A1c. Most of the HR estimates based on the mean A1c were within 0.12 of the estimates made from the first available A1c in the adjusted model. The results of the adjusted models that used first available A1c level were substantially similar to the results from models that used mean A1c levels (Fig. 1B). Another sensitivity analysis, excluding BMI, weight loss, and cholesterol and performing a complete case analysis on the remaining variables, yielded similar effect estimates, with the largest difference in the HR being at A1c <5%; namely 1.41 (1.13–1.77) in the complete case model instead of 1.30 (1.05–1.61) in the imputed data. A final sensitivity analysis, adding dialysis information (Kt/V, vascular access type, duration, and sessions per week) resulted in a model very similar to the one shown in Fig. 1A. The largest change in an A1c category’s HR was only 0.019, and there were no appreciable differences in statistical significance of any of the categories.

Impact of nutritional status

In the examination of whether nutritional status modified the relationship between A1c and mortality, markers of poor nutritional status were more common in A1c categories below 6% (Table 2). Figure 2 shows the estimated effects of A1c on mortality by nutritional status (good vs. poor). The shapes of the estimated dose-response associations differed noticeably (P = 0.12 for the likelihood ratio test of the interaction) between patients with and without indicators of poor nutritional status (Fig. 2). For patients without those indicators, there was little association with mortality for patients with A1c <9%, but the rate increased for patients with A1c ≥9%. In contrast, the association for patients with indicators of poor nutrition demonstrated higher mortality rates for patients with A1c <7% and ≥8%.

Table 2

Counts of patients with indicators of poor nutrition by mean A1c status

Counts of patients with indicators of poor nutrition by mean A1c status
Counts of patients with indicators of poor nutrition by mean A1c status
Figure 2

Risk of mortality by initial A1c among patients with and without indicators of poor nutrition, adjusted for age, sex, race, BMI, years of dialysis, albumin, creatinine, 10 comorbid conditions, insulin use, hemoglobin, HDL cholesterol, country, and study phase. (A high-quality color representation of this figure is available in the online issue.)

Figure 2

Risk of mortality by initial A1c among patients with and without indicators of poor nutrition, adjusted for age, sex, race, BMI, years of dialysis, albumin, creatinine, 10 comorbid conditions, insulin use, hemoglobin, HDL cholesterol, country, and study phase. (A high-quality color representation of this figure is available in the online issue.)

Close modal

Diabetes treatment

We also examined whether diabetes treatment modified the estimated effect of A1c on mortality. The pattern of dose-response associations between patients who were treated with oral diabetes medicines or insulin and those patients who were not treated with these medicines were similar; however, the mortality rate among patients with low A1c levels was somewhat higher among patients receiving oral medicine or insulin. The P value for the overall log-likelihood test of interaction between diabetes treatment and A1c category was 0.97.

As expected, the proportion of patients being prescribed oral diabetes medicines or insulin during the baseline period was positively associated with A1c level. The percentage of patients on any diabetes medicine was 22% for patients with A1c levels below 5% and 35% for patients with A1c levels below 6%. This percentage rose to 71% for patients with A1c levels at 9% or above, leaving 29% of these patients untreated.

A1c levels strongly predict all-cause mortality in hemodialysis patients with type 1 or type 2 diabetes. In the current study, mortality was lowest at A1c levels of 7–7.9% and increased progressively for either lower or higher A1c levels. The relationship between low A1c and mortality appeared to be even stronger in patients with indicators of poor nutritional status, including low serum albumin, low BMI, or presence of cachexia. These findings suggest that optimal A1c levels among hemodialysis patients with diabetes may need to be less stringent than levels recommended for patients with diabetes who do not have advanced chronic kidney disease (CKD). Careful attention to the use of diabetes medicines, which our data indicate are frequently prescribed to hemodialysis patients with A1c <6% and frequently not prescribed to those with A1c levels ≥9%, is a readily modifiable practice that may improve clinical outcomes.

Our findings differ from previous studies that did not show a relationship between A1c levels and mortality. A recent study did not show a relationship between A1c level and patient survival, whereas glycated albumin levels were more predictive of patient outcomes (12). That study was limited by a relatively small sample size (444 subjects), however, and thus by a low power for detecting A1c effects on clinical outcomes. Elevated A1c levels were also found not to be associated with mortality in a retrospective cohort study of maintenance hemodialysis patients in Canada (13). Differences in findings between these studies and ours may relate to variations in patient case mix, other confounding factors, or differences in duration of follow-up. Indeed, although a study based on data from a large dialysis organization did not identify an association between A1c level and survival during a 12-month follow-up period (6), updated analyses of the same study population were consistent with our findings, with increased mortality rates observed at extremes of A1c levels (5). Our findings are also consistent with those previously published on data from another large dialysis organization in the U.S., in which a similar increased mortality rate was seen at both low and high A1c levels (4,14). Our findings differ somewhat from both those studies in that we found the lowest mortality rate for patients with A1c levels between 7 and 7.9%, whereas Kalantar-Zadeh et al. (4) found the lowest rate, after adjustment for malnutrition-inflammation complex syndrome, for patients with A1c levels between 5 and 5.9%. The Molnar et al. (14) abstract showed approximately constant rates throughout the 5–7.9% range after adjusting for malnutrition-inflammation complex syndrome when using a baseline measure and the lowest rate to be in the 7–7.9% range when using the time-averaged measure. Analyses in Japanese populations also revealed increased mortality with high A1c levels (9,15). Thus our findings and those presented by others suggest that intensive glycemic control (A1c <6.0% or perhaps <5%) may not be optimal in the ESRD) population.

The National Kidney Foundation (NKF) released the Kidney Disease Outcomes Quality Initiative (KDOQI) clinical practice guidelines for the management of diabetes in CKD in 2007 (8). These guidelines recommend that the target A1c for persons with diabetes and CKD be set at <7%, the same as for diabetes patients without CKD. Given the limited published literature on this topic among patients with advanced CKD, the NKF workgroup based their recommendation primarily on data obtained from diabetes patients with CKD stages 1 and 2. New evidence from both clinical trials and observational studies published since release of these guidelines point to the need for a higher A1c target in some patient groups (4,6). For instance, the Action to Control Cardiovascular Risk in Diabetes (ACCORD) study, conducted in the non-ESRD population, showed that intensive therapy to normalize A1c levels was associated with increased mortality, a result similar to the findings of this study, and did not result in significant reduction in cardiovascular events (16). In response to the new evidence, the NKF is updating the KDOQI guidelines. The American Diabetes Association also updated its 2012 Clinical Practice Recommendations by strengthening the evidence for its recommendation that higher A1c goals may be appropriate in some patients with diabetes (7).

It should be noted that caveats exist in evaluating A1c measures in the ESRD population. First, a reduced life span of erythrocytes, as is common in dialysis patients, may result in lower A1c levels than for non-ESRD diabetes patients with the same degree of glycemic control (2). Furthermore, treatment with exogenous erythropoietin results in an increased proportion of reticulocytes in the circulation, which may be associated with less time for hemoglobin glycosylation to occur (3). Indeed, recent work suggests the potential for glycated albumin, which, unlike A1c, is not influenced by changes in erythrocyte survival or erythropoiesis-stimulating agent dose, as a measure of glucose control in the ESRD population (12,17). In addition, there are concerns that HbA1c may not reflect glycemic control in the short term because of its prolonged half-life, whereas glycated albumin may potentially reflect short-term changes in plasma glucose (17). Further study is required to evaluate the use of glycated albumin (18,19), however, especially in light of a recent study that suggests alteration in albumin quantitation among hemodialysis patients because of increased oxidative stress (20). Despite differences in the use of A1c for measuring glycemic control in diabetes patients with and without ESRD, the current KDOQI guidelines proposed similar standards for diabetes management as set by the American Diabetes Association (21), with an A1c target below 7%.

Our findings clearly suggest the continued importance of periodic A1c measurement, because A1c level is strongly associated with mortality. The potential target A1c level appears to differ from that in the general population, however, in that a higher range should be considered for ESRD dialysis patients. Our findings of a higher target range or a less intensive target may potentially be explained by greater fluctuation in glucose levels in the hemodialysis diabetes population (22). It is possible that the net catabolic balance (23) combined with the frequent occurrence of poor nutritional status of patients on hemodialysis (24) may require some degree of liberalization of target glucose levels. Various competing factors in the chronic dialysis setting may also alter the net balance of glucose control, such as changes in tissue sensitivity to insulin, existence of metabolic acidosis, variations in dextrose concentration in dialysate solution, all of which may all have varying effects on glycemic control (22). Patients with ESRD also have reduced clearance of insulin and certain oral hypoglycemic drugs used to treat diabetes. Prolonged circulation of these agents may therefore precipitate hypoglycemic episodes.

Certain limitations should be considered in the interpretation of our findings. For example, unmeasured confounders may have biased our estimates of the A1c association with mortality. We also observed a significant percentage of patients with diabetes who had no recorded measurement of A1c levels, and generalization of our findings to these patients may not be appropriate. In addition, there may be a potential for selection bias if the presence or absence of an A1c measurement during the 8-month period is associated with both the level of A1c and mortality. It is encouraging to note, however, that our effect estimates were similar when we restricted the study population to patients who had at least two A1c measurements during the first 8 months of DOPPS follow-up.

Our analyses also suggest that 78% of patients with A1c<5% were not receiving glucose-lowering agents. A potential explanation for the low A1c levels among these untreated patients may relate to poor nutritional status, which is highly prevalent among ESRD patients (23). Another possible explanation for the low A1c despite lack of treatment is the concept of “burnout” of diabetes with onset of ESRD, in which some observations have suggested spontaneous decreases of hemoglobin A1c levels among ESRD patients (25). Whereas there is a possibility that patients with A1c <5% and who were not receiving glucose-lowering agents may actually have been mislabeled as having diabetes, it is unlikely that any such misclassification is selective or biased toward those with low A1c, and any such random misclassification would only have biased our findings toward the null hypothesis.

In summary, our findings of a strong association of both high and low A1c levels with elevated mortality suggest the importance of A1c measurement in the management of patients with diabetes undergoing chronic hemodialysis. This analysis supports accumulating evidence that a target range hemoglobin A1c may be indicated in dialysis patients, rather than upper limit cut-point of <7%, as noted in previous practice guidelines. Clinical trials comparing target goals are warranted. The DOPPS is an international population, and the consistency of our findings with those limited to U.S.-specific analyses is reassuring. Finally, opportunities for improved use of hypoglycemic agents are suggested by these international data.

The DOPPS is administered by Arbor Research Collaborative for Health and is supported by scientific research grants from Amgen (since 1996), Kyowa Hakko Kirin (since 1999, in Japan), Sanofi Renal (since 2009), Abbott (since 2009), Baxter (since 2011), and Vifor Fresenius Renal Pharma (since 2011), without restrictions on publications. No other potential conflicts of interest relevant to this article were reported.

S.P.B.R. wrote the manuscript. K.P.M. researched data, wrote the manuscript, contributed to discussion, and reviewed and edited the manuscript. J.R.T. researched data. R.G.N., H.M., B.W.G., M.I., S.H.J., R.L.P., and F.K.P. contributed to discussion and reviewed and edited the manuscript. B.M.R. reviewed and edited the manuscript. R.V. researched data, contributed to discussion, and reviewed and edited the manuscript. S.P.B.R. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

The authors acknowledge the editorial assistance of Heather Van Doren, senior medical editor with Arbor Research Collaborative for Health.

1.
Foley
RN
,
Collins
AJ
.
End-stage renal disease in the United States: an update from the United States Renal Data System
.
J Am Soc Nephrol
2007
;
18
:
2644
2648
[PubMed]
2.
Little
RR
,
Tennill
AL
,
Rohlfing
C
, et al
.
Can glycohemoglobin be used to assess glycemic control in patients with chronic renal failure?
Clin Chem
2002
;
48
:
784
786
[PubMed]
3.
Nakao
T
,
Matsumoto
H
,
Okada
T
, et al
.
Influence of erythropoietin treatment on hemoglobin A1c levels in patients with chronic renal failure on hemodialysis
.
Intern Med
1998
;
37
:
826
830
[PubMed]
4.
Kalantar-Zadeh
K
,
Kopple
JD
,
Regidor
DL
, et al
.
A1C and survival in maintenance hemodialysis patients
.
Diabetes Care
2007
;
30
:
1049
1055
[PubMed]
5.
Williams
ME
.
Diabetic CKD/ESRD 2010: a progress report?
Semin Dial
2010
;
23
:
129
133
[PubMed]
6.
Williams
ME
,
Lacson
E
 Jr
,
Teng
M
,
Ofsthun
N
,
Lazarus
JM
.
Hemodialyzed type I and type II diabetic patients in the US: Characteristics, glycemic control, and survival
.
Kidney Int
2006
;
70
:
1503
1509
[PubMed]
7.
American Diabetes Association (ADA). Clinical Practice Recommendations, 2012 Available at: http://professional.diabetes.org/CPR_search.aspx. 2012
8.
KDOQI.
KDOQI Clinical Practice Guidelines Clinical Practice Recommendations for Diabetes and Chronic Kidney Disease
.
Am J Kidney Dis
2007
;
49
(2 Suppl. 2):
S12
S154
9.
Hayashino
Y
,
Fukuhara
S
,
Akiba
T
, et al
.
Diabetes, glycaemic control and mortality risk in patients on haemodialysis: the Japan Dialysis Outcomes and Practice Pattern Study
.
Diabetologia
2007
;
50
:
1170
1177
[PubMed]
10.
Raghunathan TE, Solenberger PW, Van Hoewyk J. IVEware: Imputation and variance estimation software. Ann Arbor, MI, Survey Methodology Program, Survey Research Center, Institute for Social Research, University of Michigan. 2002
11.
SAS/STAT Software. Version 9.2 of the SAS System for Windows [computer program]. Cary, NC, SAS, 2011
12.
Freedman
BI
,
Andries
L
,
Shihabi
ZK
, et al
.
Glycated albumin and risk of death and hospitalizations in diabetic dialysis patients
.
Clin J Am Soc Nephrol
2011
;
6
:
1635
1643
[PubMed]
13.
Shurraw
S
,
Majumdar
SR
,
Thadhani
R
,
Wiebe
N
,
Tonelli
M
Alberta Kidney Disease Network
.
Glycemic control and the risk of death in 1,484 patients receiving maintenance hemodialysis
.
Am J Kidney Dis
2010
;
55
:
875
884
[PubMed]
14.
Molnar MZ, Ricks JL, Kovesdy CP, et al. Glycemic control and mortality in hemodialysis patients with diabetes mellitus: a six-year cohort study [Abstract TH-OR085]. J Am Soc Nephrol 2011;22:21A
15.
Oomichi
T
,
Emoto
M
,
Tabata
T
, et al
.
Impact of glycemic control on survival of diabetic patients on chronic regular hemodialysis: a 7-year observational study
.
Diabetes Care
2006
;
29
:
1496
1500
[PubMed]
16.
Action to Control Cardiovascular Risk in Diabetes Study Group
Gerstein
HC
,
Miller
ME
, et al
.
Effects of intensive glucose lowering in type 2 diabetes
.
N Engl J Med
2008
;
358
:
2545
2559
[PubMed]
17.
Inaba
M
,
Okuno
S
,
Kumeda
Y
, et al
Osaka CKD Expert Research Group
.
Glycated albumin is a better glycemic indicator than glycated hemoglobin values in hemodialysis patients with diabetes: effect of anemia and erythropoietin injection
.
J Am Soc Nephrol
2007
;
18
:
896
903
[PubMed]
18.
Mehrotra
R
,
Duong
U
,
Jiwakanon
S
, et al
.
Serum albumin as a predictor of mortality in peritoneal dialysis: comparisons with hemodialysis
.
Am J Kidney Dis
2011
;
58
:
418
428
[PubMed]
19.
Vos
FE
,
Schollum
JB
,
Walker
RJ
.
Glycated albumin is the preferred marker for assessing glycemic control in advanced chronic kidney disease
.
NDT Plus
2011
;
4
:
368
375
20.
Michelis
R
,
Kristal
B
,
Snitkovsky
T
,
Sela
S
.
Oxidative modifications impair albumin quantification
.
Biochem Biophys Res Commun
2010
;
401
:
137
142
[PubMed]
21.
Goldstein
DE
,
Little
RR
,
Lorenz
RA
,
Malone
JI
,
Nathan
DM
,
Peterson
CM
American Diabetes Association
.
Tests of glycemia in diabetes
.
Diabetes Care
2004
;
27
(
Suppl. 1
):
S91
S93
[PubMed]
22.
Shrishrimal
K
,
Hart
P
,
Michota
F
.
Managing diabetes in hemodialysis patients: observations and recommendations
.
Cleve Clin J Med
2009
;
76
:
649
655
[PubMed]
23.
Lim
VS
,
Kopple
JD
.
Protein metabolism in patients with chronic renal failure: role of uremia and dialysis
.
Kidney Int
2000
;
58
:
1
10
[PubMed]
24.
Combe
C
,
McCullough
KP
,
Asano
Y
,
Ginsberg
N
,
Maroni
BJ
,
Pifer
TB
.
Kidney Disease Outcomes Quality Initiative (K/DOQI) and the Dialysis Outcomes and Practice Patterns Study (DOPPS): nutrition guidelines, indicators, and practices
.
Am J Kidney Dis
2004
;
44
(
Suppl. 2
):
39
46
[PubMed]
25.
Kovesdy
CP
,
Park
JC
,
Kalantar-Zadeh
K
.
Glycemic control and burnt-out diabetes in ESRD
.
Semin Dial
2010
;
23
:
148
156
[PubMed]