Adequacy of Glycemic Control in Hemodialysis Patients With Diabetes

This is a cross-sectional study of prevalent hemodialysis patients with diabetes who were established on hemodialysis for >90 days in the Kingston General Hospital (KGH) ESKD program. The KGH ESKD program is the only provider of renal replacement therapy in the region of southeastern Ontario and has a total hemodialysis population of ∼240 patients. The data were collected by chart review and interview. The presence of type 1 or type 2 diabetes was confirmed by chart review. The cause of ESKD was not identified for each patient and cannot be assumed to be diabetic nephropathy in all cases. The exclusions were age <19 years, hospital admission in the previous 2 months, and a diagnosis of diabetes for <1 year.

The outcome variable was glycemic control measured by A1C. Blood samples for A1C were drawn for each subject at the beginning of one of their respective dialysis sessions between September and December 2004. A1C assays were performed at KGH using a latex immunoassay system (Cobas Integra 400 MAb; Roche Diagnostics). For three subjects, the A1C was analyzed at a peripheral laboratory utilizing turbidometric immunoinhibition. Inadequate glycemic control was defined as a level of A1C >0.07, which translates to 7% in conventional means of expression.

Other variables were type of diabetes (type 1 or 2), duration of diabetes, diabetes care physician (family doctor, general internist, endocrinologist, or nephrologist), blood glucose monitoring (yes or no), diabetes medications (diet controlled, oral hypoglycemics, and insulin), BMI (weight in kilograms divided by the square of height in meters), time on dialysis (months), and other demographic, clinical, and laboratory variables. The erythropoietin (EPO) dose (units per week) and change in the EPO dose in the previous 2 months (yes or no) were also included because EPO may influence the A1C outcome variable directly through changes in erythrocyte lifespan and hematocrit or indirectly through influence on insulin sensitivity. The “microvascular complications” variable included retinopathy, neuropathy, and gastropathy (yes or no for each microvascular complication). The “macrovascular complications” variable included coronary artery disease (angina or previous myocardial infarction), cerebrovascular disease (stroke or transient ischemic attack), and peripheral vascular disease (yes or no for each macrovascular complication). A patient was considered to have one of the micro- or macrovascular complications if it was explicitly stated as such by a physician in the patient’s chart. The presence or absence of each complication was not verified by direct examination by the investigators.

The proportion of subjects with inadequate glycemic control was calculated as the proportion of subjects with A1C >0.07. Bivariate analysis was done comparing patients with adequate and inadequate glycemic control. Categorical variables were subjected to χ² or Fisher’s exact test as appropriate. Continuous variables were subjected to an unpaired t test or the Wilcoxon’s rank-sum test as appropriate. A forward, stepwise, multiple logistic regression controlling for age and type of diabetes was done to identify potential independent predictors of A1C >0.07. The Queen’s University research ethics board approved the protocol.

Patients receiving hemodialysis (n = 114) via KGH or a satellite unit were identified as having diabetes. Thirteen patients were excluded because they had been admitted to the hospital within the 2 months prior to the start of the study, and 1 was excluded due to missing data. A total of 100 subjects were included in the analysis. The study population was composed of 91 Caucasians, 7 First-Nation individuals, and 1 East Asian. Of the patients included, 58 were male and 42 were female. In terms of how dialysis was given, 98 patients had hospital-based hemodialysis and 2 were home based. The majority of patients received hemodialysis for 4 h three times per week. Dialysis access was an arteriovenous fistula in 70 patients, central venous catheter in 29, and artificial arteriovenous graft in 1.

Of the 100 patients, 54 had an A1C >0.07 and 46 had A1C ≤0.07. There were considerably more patients using insulin than oral hypoglycemics, with 63 patients using the former as their only medication for blood glucose control. Twenty patients used oral hypoglycemics alone, and 8 used them in conjunction with insulin. Nine patients did not use any medication for control of their blood glucose. Of the 28 patients using oral hypoglycemics, 24 of them were on a sulfonylurea, 5 were on a thiazolidinedione, 2 were on a meglitinide, and 1 patient used an α-glucosidase inhibitor.

Sixty-five patients identified their family doctor as their primary physician for management of diabetes. For the remainder, 20 identified their nephrologist for this role, 10 had a general internist, and 5 had an endocrinologist.

The characteristics of subjects by A1C ≤ and >0.07 are shown in Table 1. Compared with those with A1C ≤0.07, subjects with A1C >0.07 had a higher proportion using insulin, a lower mean EPO dose, and a longer duration of diabetes. Figure 1 shows the proportions of subjects with A1C ≤ and >0.07 according to categories of diabetes duration. A higher proportion of patients with A1C >0.07 had documentation of microvascular complications (66.7 vs. 43.5%, P = 0.02), although there was no difference in macrovascular complications noted between the two groups. There were also no significant differences in age, sex, BMI, change in EPO dose, duration of dialysis, or Kt/V_urea (dialysis adequacy as measured by urea clearance). We compared the serum albumin level, a marker of inflammation, among subjects with A1C ≤ and >0.07, and there was no significant difference (A1C ≤ vs. >0.07: serum albumin 36.6 vs. 36.8 g/l, respectively, P = 0.79).

In multiple logistic regression, while controlling for age and type of diabetes, the diabetes duration (odds ratio 1.09 [95% CI 1.04–1.15], P < 0.001), EPO dose (0.90 [0.85–0.97], P < 0.01), and blood glucose monitoring (10.06 [1.03–98.74], P = 0.05) were the only significant independent predictors of A1C >0.07 for this dataset.

The majority of patients in this study (54%) had inadequate glycemic control based on the current clinical practice guidelines of both the Canadian and American Diabetes Associations, which recommend a target A1C of ≤0.07 (10,11). Even if a higher value of 0.075 were used to define adequate glycemic control, 36% of patients in this study would have had an A1C level above target.

There are unfortunately no available data on the adequacy of glycemic control in the general diabetic population of Canada; thus, it is unknown how the figures from this study compare with diabetic patients in southeastern Ontario who do not require dialysis. However, these numbers are very similar to a recent epidemiological study by Coon and Zulkowski (12), which found that 53% of patients had an A1C <0.07 in the general rural diabetic population of Montana. Their study population had important demographic similarities to ours, with the former having a mean age of 69 years and an ethnic breakdown of 97–99% Caucasian, 1–4% Native American, and 1% African American.

A recent small study suggested the initiation of hemodialysis may not be associated with a change in a patient’s A1C levels (13). This would help explain why diabetic patients on hemodialysis may not have a greater prevalence of inadequate glycemic control than the general diabetic population. Regardless, it is important to note that the rate of inadequacy was high and that these patients had a significantly higher burden of nonrenal microvascular complications (66.7 vs. 43.5%, P = 0.02). These results are consistent with the DCCT and UKPDS, which found a significant risk reduction of microvascular complications in patients on intensive glycemic control regimens compared with those on conventional regimens (the latter group having a higher median A1C in both trials) (1,2). Such complications are a direct result of prolonged exposure to elevated levels of blood glucose and can severely affect the quality of life for these patients. It should also be reiterated that elevated levels of A1C have been associated with increased mortality in diabetic patients with and without ESKD (3,14–16).

The population of diabetic patients on hemodialysis that we studied had a high prevalence of macrovascular complications, irrespective of adequacy of glycemic control (59.3 vs. 65.2%, P = 0.54). Events such as myocardial infarctions and strokes may have myriad contributory causes, a uremic milieu being just one of them. Any role that glycemic control played may have been washed out. It is difficult to compare these results with the DCCT trial, as its patient population was much younger (13–29 years). The UKPDS trial found no significant difference in macrovascular end points between patients on intensive versus conventional glycemic control regimens, although the lower incidence of myocardial infarction among patients on an intensive regimen did border on significance.

This study looked at other factors that were associated with having inadequate glycemic control. The duration of diabetes had a significant correlation with A1C >0.07 in both bivariate analysis and multiple logistic regression when controlling for age and type of diabetes. Figure 1 shows a striking graphical representation of this relationship, with the proportion of patients with inadequate control consistently rising as the duration of diabetes lengthened. This may be in part due to increasing resistance to insulin over time, making the disease more difficult to treat. Significantly more patients who were on insulin had inadequate control, but this could represent the fact that patients with type 2 diabetes tend to follow a time course whereby an attempt is made to first control the disease with diet, then with oral hypoglycemics, and lastly with insulin. The use of insulin would therefore represent a longer-standing disease in these patients. Similarly, we hypothesize that the unexpected finding of an inverse association between blood glucose monitoring and adequate glycemic control is explained by the confounding effects of duration and severity of diabetes.

The idea of maintaining tight glycemic control through an intensive regimen is itself a relatively recent phenomenon. This too may have played a role in the significance of diabetes duration; the more recently a patient was diagnosed with diabetes, the more likely he or she would follow an intensive regimen as outlined by his or her physician. Patients who were diagnosed before these recommendations may be more accustomed to a less rigid treatment regimen and therefore less likely to change. This is not to say that these patients are all older than those with a more recent diagnosis; as previously mentioned, the diabetes duration was an independent predictor of A1C >0.07 for this dataset, even when controlling for age.

The relationship between EPO dose and adequacy of glycemic control requires further elucidation. Increasing a patient’s EPO dose will cause a corresponding rise in hematocrit, and this has been shown to artificially lower the A1C in the presence of stable blood glucose levels (17). For this reason, we also documented whether there had been a change in the patients’ EPO doses and checked the slopes of three serial hemoglobin values for each individual (hemoglobin level at the time of measuring A1C and the 2 prior months). There was no significant difference in these variables between those with and without adequate glycemic control. Intuitively, one might think that having inadequate glycemic control would mean having worse disease and requiring higher doses of EPO, whereas the converse relationship was seen in this study; patients with A1C >0.07 had a much lower EPO dose on average, with a P value <0.01 in bivariate analysis. In multiple logistic regression, it too came out as an independent predictor of A1C >0.07. It may be that EPO itself is playing a role in lowering blood glucose levels. There has been some research suggesting that EPO increases insulin sensitivity (18). Lastly, there appears to be an interaction between the stage of chronic kidney disease (CKD) and the presence of diabetes in regard to hemoglobin, probably explained by endogenous EPO production. El-Achkar et al. (19) examined the association between the stage of CKD and hemoglobin in 5,380 participants (26.9% with diabetes) in the Kidney Early Evaluation Program (KEEP 2.0) study and found a higher prevalence of anemia in patients with diabetes among those with stage 3 CKD in univariate analysis and among those with stages 3, 4, and 5 CKD in multivariate analysis controlling for demographic characteristics. The authors hypothesize that the negative influence of diabetes on hemoglobin is explained by decreased EPO production due to reduced splanchnic sympathethic stimulation and reduced androgen levels. How this relates to exogenous EPO dose, insulin sensitivity, erythrocyte lifespan, hemoglobin, duration of diabetes, diabetes type, and ultimately A1C in patients with CKD and ESKD requires further investigation.

There are some limitations to this study, in addition to those inherent to retrospective analyses. The patient population included all diabetic patients on hemodialysis in southeastern Ontario. Differences in dialysis-patient populations must be taken into account when assessing its generalizability. Dialysis populations in the U.S., for example, may have a much higher representation of African-American and Hispanic patients, two ethnicities that were not represented in our study population.

As previously mentioned, the individual cause of ESKD was not identified for each patient and cannot be assumed to be diabetic nephropathy in all cases. Six of the 100 patients developed diabetes after the initiation of dialysis. It is also possible that some patients had residual renal function, which may have contributed to less morbidity in the presence of higher blood glucose levels.

This study did not look at symptomatic hypoglycemia in the patients with better glycemic control. The DCCT and UKPDS showed significantly higher incidences of symptomatic hypoglycemia in patients on an intensive glycemic control regimen, and this may represent a limitation to its implementation.

Using A1C as a measure of long-term glycemic control in this population was not a limitation. It likely does not represent as long term a measurement in patients with ESKD; there is a faster turnover of erythrocytes in these patients, with the average lifespan reduced 25–40% (20). However, it is accurate; an immunoassay was used to make these measurements, and it has been shown that between an A1C of 0.06–0.07, this type of assay relates similarly to average blood glucose among patients with ESKD and those with good renal function. A1C values above 0.075 may overestimate the level of hyperglycemia in patients with ESKD (21).

In summary, the majority of patients in this study had inadequate glycemic control, and these patients had a significantly higher burden of other microvascular complications associated with diabetes. There remains a significant opportunity to improve the quality of life of these patients in helping to prevent or delay further microvascular complications. Who should be responsible for this task is a matter of debate. In this study, the majority of patients identified their family doctor as the physician responsible for managing their diabetes. Nephrologists were identified second, followed by general internists and endocrinologists. Making comparisons among these groups would be rife with difficulty, as some patients may not even have a family doctor, and those seeing endocrinologists may represent patients with the most severe disease. However, it is interesting to note that 20% of patients identified their nephrologist as being most responsible for diabetes management. With the current difficulty many patients have in finding a family doctor, it is becoming increasingly incumbent upon nephrologists to take on this role. Studies are now under way to examine whether glycemic control can be improved by using regular hemodialysis time as an opportunity for diabetes education and for making appropriate adjustments based on patients’ blood glucose monitoring.