Glycemic Monitoring and Management in Advanced Chronic Kidney Disease

Abstract

Glucose and insulin metabolism in patients with diabetes are profoundly altered by advanced chronic kidney disease (CKD). Risk of hypoglycemia is increased by failure of kidney gluconeogenesis, impaired insulin clearance by the kidney, defective insulin degradation due to uremia, increased erythrocyte glucose uptake during hemodialysis, impaired counterregulatory hormone responses (cortisol, growth hormone), nutritional deprivation, and variability of exposure to oral antihyperglycemic agents and exogenous insulin. Patients with end-stage kidney disease frequently experience wide glycemic excursions, with common occurrences of both hypoglycemia and hyperglycemia. Assessment of glycemia by glycated hemoglobin (HbA_1c) is hampered by a variety of CKD-associated conditions that can bias the measure either to the low or high range. Alternative glycemic biomarkers, such as glycated albumin or fructosamine, are not fully validated. Therefore, HbA_1c remains the preferred glycemic biomarker despite its limitations. Based on observational data for associations with mortality and risks of hypoglycemia with intensive glycemic control regimens in advanced CKD, an HbA_1c range of 7% to 8% appears to be the most favorable. Emerging data on the use of continuous glucose monitoring in this population suggest promise for more precise monitoring and treatment adjustments to permit fine-tuning of glycemic management in patients with diabetes and advanced CKD.

Graphical Abstract

CKD, end-stage renal disease, diabetes mellitus, hypoglycemia, hemoglobin A1c, CGM, glucose and insulin metabolism, insulin therapeutic use, hypoglycemic agents therapeutic use, practice guidelines

Essential points

• Glucose and insulin metabolism are profoundly altered by advanced chronic kidney disease

• Risk of hypoglycemia is increased by several factors, including failure of kidney gluconeogenesis, impaired insulin clearance by the kidney, defective insulin degradation due to uremia, increased erythrocyte glucose uptake during hemodialysis, impaired counterregulatory hormone responses, nutritional deprivation, and variability of exposure to antihyperglycemic agents and exogenous insulin effect

• Patients with end-stage kidney disease frequently experience wide glycemic exposures, with common occurrences of both hypoglycemia and hyperglycemia

• Assessment of glycemia by hemoglobin A_1c is hampered by a variety of chronic kidney disease–associated conditions that can bias the measure either to the low or high range

• Except for documented interferences from icodextrin metabolites and nonglucose sugars in peritoneal dialysis solutions with the use of glucose dehydrogenase-based–pyrroloquinoline-quinone/glucose oxidase–based glucose meters, the evidence is very limited in the end-stage kidney disease population

• Emerging data on the use continuous glucose monitoring in this population suggest the potential for more precise monitoring and treatment adjustments in patients with diabetes and advanced chronic kidney disease

Unequivocal evidence from landmark studies have shown that uncontrolled diabetes is associated with short- and long-term complications, including chronic kidney disease (CKD) (1, 2). With the increasing prevalence of type 2 diabetes, diabetes has become the leading cause of CKD and is responsible for approximately half of all cases of end-stage kidney disease (ESKD) worldwide (3, 4). In 2016 there were more than 726 000 patients with ESKD in the United States, with 66% to 86% of them having a diagnosis of diabetes depending on age and race/ethnicity (5). In patients with type 2 diabetes, studies have reported a higher prevalence of CKD among those older than 65 years compared with younger patients (58% vs 25%) and in African Americans compared with non-Latino whites (43% vs 38%) (6). Diabetes is more prevalent in ESKD than in earlier CKD stages because diabetes may co-occur with other causes of CKD or develop after ESKD, particularly in kidney transplant recipients.

Treatment of ESKD by hemodialysis is associated with poor long-term survival, with an adjusted survival after 3 and 5 years of hemodialysis initiation of only 55% and 40%, respectively. Patients with diabetes have the worst adjusted-survival rates among the ESKD population (4). Notably, patients with diabetes and advanced CKD, including CKD stage 4 and 5 or with an estimated glomerular filtration rate (eGFR) of less than 30 mL/min/1.73 m², also have a high risk of death and similar complications to those with ESKD on dialysis (7). Thus, there is a compelling need for better strategies to improve glycemic outcomes in this high-risk population.

Several landmark studies including the Diabetes Control and Complications Trial and the follow-up Epidemiology of Diabetes Interventions and Complications studies reported that intensive vs standard glycemic control reduces microvascular complications in patients with type 1 diabetes (8, 9). In patients with type 2 diabetes, the United Kingdom Prospective Diabetes Study (7) reported that intensive glycemic treatment reduced microvascular complications compared to standard treatment (7, 10, 11). Recently, newer antidiabetic agents like sodium-glucose loop transporter-2 inhibitors and glucagon-like peptide receptor-1 agonists have shown significant benefits on the development and progression of CKD in addition to glycemic control (12-14). Yet, there is no clear consensus on optimal glycemic management strategies for patients with advanced CKD (4). For this high-risk population, the present review provides a comprehensive discussion of 1) glucose and insulin metabolism; 2) risks of hyperglycemia and hypoglycemia; 3) glycemic goals and use of traditional monitoring methods; and 4) potential advantages of using continuous glucose monitoring (CGM) for glycemic monitoring.

Glucose and Insulin Metabolism in Normal and Diabetic Conditions

Normal physiology

Euglycemia is tightly maintained in humans by the rate of “glucose appearance” into the circulation—from intestinal absorption, glycogenolysis and gluconeogenesis—and the rate of glucose disappearance. In a simplistic bihormonal model, glucagon mediates the rate of glucose appearance and insulin mediates glucose removal (disappearance) into the musculoskeletal and fat tissue and by suppression of hepatic glucose production (Fig. 1A). During the fed state, glucose appearance is primarily influenced by the gastric emptying time. During fasting, endogenous glucose production is mediated mainly by hepatic glycogenolysis during the first 8 to 12 hours, followed by gluconeogenesis (15-18). The kidney also plays a role in endogenous glucose production, but mostly during prolonged starvation periods.

Figure 1.

A, Insulin and glucose metabolism by the liver, fat and muscle tissues, and kidney with normal renal function. B, Insulin and glucose metabolism by the kidney, fat and muscle tissues, and pancreas in early chronic kidney disease (CKD) resulting in higher risk of hyperglycemia. C, Insulin and glucose metabolism by the kidney, fat and muscle tissues, and pancreas in advanced CKD and hemodialysis (HD) resulting in higher risk of hypoglycemia. RBC, red blood cells

Pathophysiology in diabetic conditions

Under physiologic conditions, endogenous insulin is continuously produced at a rate equivalent to 0.5 to 1 units per hour—termed basal insulin secretion and representing approximately 48% to 52% of the total daily insulin production. After food challenges—typically after breakfast, lunch, and dinner—insulin secretion is increased 3 to 10 times over about a 4-hour postprandial period, then returns to the basal rate. These patterns are mediated by a pulsatile but continuous insulin secretory mechanism during a 24-hour period. Blood glucose mediates these insulin pulses, secreted in a biphasic manner, with a first rise within the 3 to 5 minutes that lasts up to 10 minutes (first phase), followed by a slower and extended phase of 60 to 120 minutes (second phase) (15-17). In obese patients, the pattern of basal and postmeal insulin secretion rates, as well as the secretory pulses, are maintained but at a considerably higher rate (5-6 times), remaining elevated after meals and not fully returning to baseline (19-21). Patients with type 1 diabetes typically lack clinically meaningful insulin secretion (22, 23), and those with type 2 diabetes show a blunted glucose-mediated insulin secretion pulse of approximately 70% of normal (24). First-phase insulin secretion is lost not only in patients with type 2 diabetes at diagnosis, but also in patients with prediabetes (18).

Glucose and Insulin Metabolism in Diabetes with Advanced Chronic Kidney Disease

The interplay between the kidney, glucose, and insulin is complex, with multiple interactions. During the fed state, the kidney uptake of glucose accounts for up to 20% of all glucose removed from the circulation; but during prolonged fasting states, it can produce up to 20% to 25% of blood glucose via gluconeogenesis (25-27). Endogenous insulin clearance is mediated by the liver, up to 40% to 50%, with the remainder entering the systemic circulation. Systemic insulin reaching the kidney is filtered by the glomerulus (up to 60%-65%) with reuptake into proximal tubular cells. Insulin also is transported from postglomerular peritubular vessels to proximal tubular cells (up to 35%). At this location, insulin then undergoes degradation, resulting in approximately 1% excreted in the urine (28, 29). Notably, the kidney is responsible for a larger portion of metabolism, up to 80%, of exogenous insulin because it does not go via first-pass metabolism in the liver (15-18, 28-30).

Pathophysiological Mechanisms in Early and Advanced Chronic Kidney Disease

Glucose metabolism in CKD is mediated by multiple mechanisms: 1) impaired glucose disposal by muscle and peripheral tissues due to uremia; 2) reduced insulin removal by the damaged kidney; 3) persistent mild inflammatory state; and 4) oversecretion of counterregulatory hormones (28, 31) (Fig. 1B and C). In addition, it has been proposed that patients with advanced CKD are predisposed to postprandial hyperglycemia due to impaired osmotic diuresis and increased muscle insulin resistance (32). Other metabolic abnormalities observed in CKD such as vitamin D deficiency, obesity, metabolic acidosis, and accumulation of “uremic toxins” might also contribute to the increased insulin resistance and the development of acquired defects in the insulin-receptor signaling pathway (31, 33, 34).

Following a biphasic course over time, patients with early CKD stages may be exposed to a higher insulin resistant state, increasing insulin needs in patients with type 1 diabetes or requiring initiation of insulin in patients with type 2 diabetes. However, patients with ESKD are prone to hypoglycemia, with decreased insulin clearance. An interesting phenomenon called “burn-out diabetes” is well described among some patients (~15%-30%) with ESKD (GFR < 20 mL/min/1.73 m²) and type 2 diabetes. These patients have been previously treated with insulin or other antihyperglycemic agents, and as CKD advances to ESKD they need less or no medications for glycemic control (31). Several factors have been proposed, including prolonged half-life of endogenous and exogenous insulin, lessened insulin resistance resulting from removal of uremic toxins by dialysis, decreased gluconeogenesis, and poor nutritional status (35). Biesenbach et al studied insulin requirements among patients with type 1 diabetes and insulin-treated patients with type 2 diabetes progressing from early CKD to ESKD and reported a decrease of total daily insulin dose of 40% in patients with type 1 diabetes and approximately 50% among patients with type 2 diabetes, regardless of residual insulin secretion (36).

Patients with diabetes on hemodialysis commonly experience hypoglycemic episodes that requires adjustment or discontinuation of insulin and oral antihyperglycemic agents. Several hypoglycemic mechanisms have been recognized: 1) decreased gluconeogenesis and impaired insulin clearance by the kidney; 3) reduced insulin degradation by the kidney, liver, and muscle due to uremia; 4) increased erythrocyte glucose uptake during hemodialysis; 5) impaired counterregulatory hormone responses (cortisol, growth hormone); 6) nutritional deprivation; and 7) variability of exposure to antihyperglycemic agents and exogenous insulin effect due to the hemodialysis procedure or frequent treatment/doses changes by patients or medical staff (4, 28, 29, 31, 37). It has been proposed that the main determinant of plasma glucose levels after hemodialysis is the glucose concentration of the dialysate (38). Previously used dialysate solutions with high glucose concentrations (up to 1600 mg/dL) led to hyperglycemia. With advances in ultrafiltration techniques, the use of glucose-free dialysates was applied more frequently. However, hypoglycemia became problematic with the use of these solutions. As a result, dialysate solutions with a glucose concentration of 100 to 200 mg/dL are the current standard and produce less hypoglycemia (38, 39).

Reduced plasma insulin removal during hemodialysis and variable insulin secretion from pancreatic β cells also may increase the risk of hypoglycemia (31). Insulin removal occurs by diffusion or convection across dialyzer membranes because of its low molecular weight and the existence of a concentration gradient. Insulin clearance also depends on the type of dialyzer membrane: greater with polysulfone membranes and lower with a polyester-polymer alloy (40-42). Although the primacy of the removal mechanism is unclear, it involves diffusive and convective transport processes as well as electrostatic interactions between dialyzer membranes and insulin (43). Insulin secretion from pancreatic β cells is determined by dialysis-induced changes in plasma glucose levels and residual ability to respond to these changes (42). Therefore, patients with reduced β-cell function and insulin secretion will have greater reduction in insulin levels and consequently increased risk of hyperglycemia after hemodialysis treatment, particularly if a polysulfone membrane is used. These patients may need higher doses of exogenous insulin and/or other antihyperglycemic agents to achieve adequate glycemic control. Conversely, patients with preserved β-cell function and insulin secretion may have a higher risk of hypoglycemia with hemodialysis, particularly after large hyperglycemia excursions (31). Attention to preventing dialysis-associated hyperglycemia helps to avoid administration of additional glucose-lowering agents as a strategy to prevent interdialysis and postdialysis hypoglycemia. In sum, finding a balance in timing and dosing of antihyperglycemic agents is paramount to optimize glycemic treatment in diabetic patients treated by hemodialysis.

Peritoneal dialysis is an alternate mode of dialysis that helps to protect from hypoglycemia by absorption of dextrose from the dialysis solution (concentrations of 1.5%, 2.5%, and 4.25%). Additionally, insulin can be put into the dialysis solution, thus providing a relatively constant amount of basal insulin via the portal circulation to reduce glycemic excursions. However, the increase in calories from absorption of dextrose also promotes weight gain and higher insulin requirements.

Recently with the use of continuous glucose monitoring, patients with and without diabetes on hemodialysis were found to be exposed to wide glycemic excursions, also known as increased glycemic variability (GV) (37, 44). A unifying hypothesis for diabetes-related vascular complications is that hyperglycemia increases the formation of advanced glycation end products and reactive oxygen species in endothelial cells, podocytes, and mesangial cells, resulting in a chronic proinflammatory state (45). Recently, a few studies also have shown that GV may increase oxidative stress and inflammatory mediators (45-48). Furthermore, there is evidence that patients with CKD and diabetes have increased levels of oxidative stress and inflammatory mediators, which may contribute to excess mortality (49-54). Proinflammatory biomarkers such as C-reactive protein, interleukin-6, and serum amyloid A have been linked to increased mortality in patients with ESKD and diabetic kidney disease (49, 53, 54). Soluble urokinase plasminogen activator receptor has been proposed as another proinflammatory biomarker for cardiovascular events, vascular and kidney damage, and mortality in advanced CKD (54, 55).

Clinical Implications of Hyperglycemia in Patients With Advanced Chronic Kidney Disease

The United Kingdom Prospective Diabetes Study, the largest study conducted of glycemic management in patients with type 2 diabetes, showed that intensive control can reduce risk of microvascular complications, including progression to “nephropathy,” now termed diabetic kidney disease (1, 2, 7, 11, 54-56). However, more contemporary trials, like the ADVANCE, VADT, and ACCORD trials, showed that among participants with long-term type 2 diabetes, including some with early-stage CKD, intensive glycemic control produced either no cardiovascular benefit or even higher mortality (57, 58 , 59). Of note, the higher mortality was seen only in the group randomly assigned to receive intensive control in the ACCORD trial, but later meta-analyses of these trials suggested a benefit of intensive glycemic control on cardiovascular outcomes (60). However, as most trials did not include patients with advanced CKD, their optimal long-term glycemic control goal, as measured by HbA_1c, remains uncertain.

In 2007, Kalantar-Zadeh and colleagues reported that HbA_1c levels greater than or equal to 10% were incrementally associated with a higher death risk in comparison with HbA_1c levels in the 5% to 6% range among 23 618 patients with ESKD on hemodialysis. The adjusted all-cause and cardiovascular death hazard ratios (HRs) for HbA_1c greater than or equal to 10% were 1.41 (95% CI, 1.25-1.60) and 1.73 (95% CI, 1.44-2.08), respectively (P < .001) (61). Drechsler et al also evaluated the impact of glycemic control in 1255 patients with diabetes on hemodialysis and reported that those with an HbA_1c greater than 8% had a greater than 2-fold higher risk of sudden death compared with those with an HbA_1c less than or equal to 6% (HR 2.14; 95% CI, 1.33%-3.44%). A trend for higher risks of stroke and deaths resulting from heart failure was observed, whereas myocardial infarction risk was not increased (62). Similarly, the Dialysis Outcomes and Practice Pattern Study observed a U-shaped association between HbA_1c levels and risk of death, with higher mortality for levels less than 6% and 9% or higher, in 9201 hemodialysis patients with type 1 diabetes or type 2 diabetes (63). A study by Ricks et al of 54 757 hemodialysis patients treated at DaVita dialysis centers corroborated these data (64). Nevertheless, not all reports have concurred with these observations. For example, in a study by Williams and colleagues of 24 875 patients with diabetes on hemodialysis treated at Fresenius Medical Care dialysis centers, there was no association between HbA_1c level and mortality in unadjusted analysis after 1 year (65). However, in a follow-up study with adjustment for confounders (malnutrition, inflammation, anemia, comorbidities) and using a design with time-varying HbA_1c levels and follow-up extended to 3 years, levels of HbA_1c less than 6.5% or greater than 11% were actually associated with an increased risk of mortality (66). Finally, a rigorous meta-analysis of 10 studies (83 684 patients, type 1 diabetes and type 2 diabetes) calculated that, overall, hemodialysis patients with baseline and mean HbA_1c greater than 8.5% had an increased risk-adjusted mortality of 14% (HR 1.14; 95% CI, 1.09%-1.19%) and 29% (HR 1.29; 95% CI, 1.25%-1.35%), compared to HbA_1c 6.5% or 7.4%, respectively (Fig. 2) (67).

Figure 2.

Association of mean hemoglobin A_1c and adjusted all-cause mortality risk in patients with diabetes on hemodialysis: results of a meta-analysis of 10 studies (n = 83 684 patients).

Patients with diabetes treated by maintenance hemodialysis with severe hyperglycemia and diabetic ketoacidosis (DKA) often present with mild hyperglycemic symptoms (68). Notably, volume overload is more common than volume depletion because osmotic diuresis does not occur (68, 69). In a recent study, we reported that among patients with DKA, those with ESKD on hemodialysis presented with more severe hyperglycemia on admission (mean glucose 804.5 ± 362.6 mg/dL vs 472.5 ± 137.7 mg/dL), despite having lower HbA_1c (mean 9.6% ± 2.1 vs 12.0% ± 2.5) compared to patients with preserved kidney function (P < .001 for both comparisons). The rates of hypoglycemia less than 70 mg/dL (34% vs 14%, P = .002), volume overload (28% vs 3%, P < .001), and need for mechanical ventilation (24% vs 3%, P = < .001) also were substantially higher in the ESKD group during the hospitalization. After adjusting for multiple covariates, patients with DKA and ESKD have higher odds of hypoglycemia (odds ratio [OR] 3.3; 95% CI, 1.51-7.21; P = .003) and volume overload (OR 4.22; 95% CI, 1.37-13.05, P = .01] compared to patients with DKA and preserved kidney function (70).

In small studies using CGM technology, it has been shown that the use of peritoneal dialysis, containing 2.27% to 3.8% glucose concentrations, result in severe and sustained hyperglycemia compared to the use of more physiologic solutions with 1.36% glucose concentration or nonglucose-containing solutions (71, 72).

Clinical Implications of Hypoglycemia in Patients with Advanced Chronic Kidney Disease

Hypoglycemia is a consequence of antihyperglycemic therapy associated with high risk of morbidity and mortality (73, 74). Large, randomized clinical trials have questioned the benefits of intensive glycemic control for patients with long-term type 2 diabetes and multiple comorbidities (56, 58, 75). Chu and colleagues recently reported that up to 19.2% of participants with diabetes who develop ESKD have an episode of hypoglycemia the year prior to initiation of dialysis, with higher incidence in hemodialysis (10.5%) compared to peritoneal dialysis (7.6%) during the first year after dialysis initiation (76). Moen et al also reported that the frequency of hypoglycemia is higher in hemodialysis patients with diabetes compared to patients with no diabetes (77). In a single-center study of ambulatory patients with diabetes on maintenance hemodialysis, the prevalence of hypoglycemia ranged from 46% to 52% (78). Among hospitalized patients with diabetes and ESKD, Gianchandani et al reported a prevalence of hypoglycemia less than 70 mg/dL, less than 54 mg/dL, and less than 40 mg/dL of 51%, 28%, and 10%, respectively. Multiple episodes of hypoglycemia were noted in 35% of these patients (79).

Hypoglycemic episodes are associated with a higher risk of recurrent hypoglycemia and mortality after initiation of dialysis (76). Recently, Rhee and colleagues analyzed the Veteran Affairs database to study 20 156 veterans with diabetes and predialysis CKD transitioning to dialysis over 1 to 2 years. One or more hypoglycemia-related hospitalizations occurred before initiating dialysis in 5.9%. Independent risk factors for hypoglycemia-related hospitalization were Hispanic ethnicity, heart failure, cerebrovascular disease, high HbA_1c, and use of insulin (76, 80). Furthermore, hypoglycemia-related hospitalizations before transition to hemodialysis were strongly associated with higher mortality after transition to dialysis (80).

Given the large comorbidity burden of patients with ESKD, it is unclear whether hypoglycemia-related morbidity and mortality is causal or a marker of overall disease severity and burden (81). In addition, hypoglycemia is associated with increased risk of cardiac arrhythmias (82), stroke (83), seizures (84), and sudden cardiac death (84, 85). In a descriptive study of admissions by Haviv et al among 1545 patients with ESKD and with and without diabetes, 3.6% were admitted for hypoglycemia. The most commonly identified causes were drug-induced hypoglycemia (46%), followed by sepsis (39%) and severe malnutrition (7%) (81). Importantly, high GV commonly occurs in tandem with hypoglycemia among patients with type 2 diabetes (86-88). Frequent hypoglycemic events are associated with high GV values (89), and reducing hypoglycemia strongly correlates with decreased GV (90-92). In addition, high GV is linked to increased risk of cardiovascular events and death (93) and all-cause mortality (94).

Glycemic Treatment Goals

Overall, the available evidence suggests that among patients with diabetes and advanced CKD, the target HbA_1c levels may be different from those recommended by current guidelines for other patients. Although targeting HbA_1c levels of less than 7% is associated with greater survival in patients with lower comorbidity burden and adequate nutritional status, lower HbA_1c levels are associated with an increased risk of death in those with comorbidities and malnutrition.

The National Kidney Foundation–Kidney Disease Outcomes Quality Initiative (NKF-KDOQI) guidelines recommend an HbA_1c target of approximately 7% for most patients with CKD. However, a personalized approach with less strict glycemic targets (HbA_1c 7%-8%) is endorsed by NKF-KDOQI and other diabetes guidelines for patients with advanced CKD because of shorter-life expectancy, high comorbidity burden, or high risk of hypoglycemia (3, 95). These recommendations were largely driven by the high risk of iatrogenic hypoglycemia from treatment with antihyperglycemic agents. However, newer agents (eg, incretin therapies) with a lower risk of hypoglycemia have not been extensively studied in this group. In addition, HbA_1c has less reliability in the setting of advanced CKD (Fig. 3), as discussed further later (4, 96). The goal for glycemic targets to optimize clinical outcomes in these patients is unknown and an important topic for research (3, 4, 96).

Figure 3.

Limitations of glycemic biomarkers (hemoglobin A_1c [HbA_1c], fructosamine, and glycated albumin) in patients with advanced chronic kidney disease (CKD)

Glycemic Monitoring by Traditional Methods in Advanced Chronic Kidney Disease

Hemoglobin A_1c, fructosamine, and glycated albumin

The NKF-KDOQI and Kidney Disease Improving Global Outcomes clinical practice guidelines recommend assessment of glycemic control with HbA_1c, in combination with home blood glucose monitoring, as a cornerstone of diabetes management in patients with ESKD (3). However, numerous ESKD-related factors may adversely affect the value of HbA_1c. High values have been reported in patients with elevated blood urea nitrogen and metabolic acidosis due to formation of carbamylated hemoglobin, which cannot be distinguished from glycated hemoglobin in certain assays (97, 98). In contrast, low HbA_1c levels are frequently observed in the presence of anemia, use of erythropoietin-stimulating agents, reduced erythrocyte lifespan from uremia, and erythrocyte lysis during the hemodialysis procedure. Owing to these limitations, the use of alternative glycemic markers such as fructosamine and glycated albumin have been developed (99-101) (Fig. 3). These biomarkers reflect glycemia in a briefer time frame (2-4 weeks) than HbA_1c because of their shorter survival in blood. Glycated albumin predicts all-cause and cardiovascular mortality in patients treated by chronic hemodialysis (102). However, the glycated albumin assay is biased to the low for glycemic monitoring by hypoalbuminemia, a common condition in patients with CKD due to protein losses in the urine, malnutrition, or peritoneal dialysis. Conversely, fructosamine has an assay bias to the high by hypoalbuminemia. Fukami et al have developed and validated the use of serum albumin–adjusted glycated albumin as an indicator of glycemic excursions markers—namely SD and maximum plasma glucose—in individuals with advanced CKD (101). Nevertheless, glycated albumin and fructosamine have insufficient accuracy for overall glycemic control, limited availability, and have not been adequately validated in populations with advanced CKD (4, 100).

Self-monitored blood glucose by capillary glucose meter

In most instances, clinicians and patients with ESKD make treatment decisions, particularly insulin doses, based on point-of-care (POC) blood glucose concentrations. However, POC blood glucose is limited by poor sample stability, and several factors could affect glucose concentration including anemia, acute illness, medications, or interfering substances. Methods for glucose measurement are based on 3 enzymes: 1) glucose oxidase-based (GO), 2) hexokinase-based (HK), or 3) glucose dehydrogenase-based (GDH) (103-105). GDH can be attached to a coenzyme, including pyrroloquinoline-quinone (GDH-PQQ), GDH nicotine adenine dinucleotide (GDH-NAD), or glucose dehydrogenase flavin adenine dinucleotide (GDH-FAD). Electrochemical enzymatic sensors are subject to interferences from substances that may electrochemically react with the sensor’s electrode or have cross-reactivity with the enzyme (eg, high levels of acetaminophen, ascorbic acid, icodextrin, maltose, triglycerides, uric acid, or abatacept) (103-106).

Serious hypoglycemic events have been reported with the use of GDH-PQQ–based glucose meters. GDH-PQQ methodology cannot distinguish between glucose and other sugars. Some of these reports were among patients using peritoneal dialysis solutions containing icodextrin. In these patients, falsely elevated glucose readings were reported by the POC meter, but the true glucose concentration was much lower (104). Other methods, such as HK, GDH-NAD, or GDH-FAD do not have such interference with nonglucose sugars (106). Because some reports also showed interference with GO meters, glucose meters using GDH-PQQ and GO should not be used in patients with ESKD on dialysis or using other interfering medications (eg, some immunoglobulins, abatacept, parenteral maltose/galactose/xylose solutions) (103-105).

Several other examples of interference may cause false glucose results (103-105). For instance, low hematocrit (< 35%) may result in falsely high glucoses in the glucose meter using the GO technique, but many contemporary meters correct for this. Conversely, high acetaminophen plasma levels (> 8 mg/dl) may result in falsely high blood glucose readings. Hypoxia (partial pressure of oxygen < 45 mmHg) or oxygen therapy (partial pressure of oxygen > 150 mmHg) may cause falsely high and low glucose in GO-based meters, respectively. High levels of triglycerides, uric acid (> 20 mg/dL), or bilirubin may cause pseudohypoglycemia. Except for the interferences found from the use of peritoneal solutions in the GDH-PQQ/GO–based meter, the evidence is very limited in the ESKD population. Therefore, the performance of glucose meters in individual patients should be carefully assessed in these clinical situations.

More important, the POC blood glucose approach fails to detect asymptomatic and nocturnal hypoglycemia and fails to provide a complete glycemic profile throughout the day, particularly during hemodialysis sessions (37). Hence, there is a critical need to find a standardized long-term outcome measure for glycemic control in patients treated by dialysis, but also a glycemic measure that will allow patients and clinicians to make rapid therapeutic decisions to prevent hypoglycemia in real time (4, 100). Tables 1 and 2 show recommendations for long-term glycemic monitoring and management in these patients.

Table 1.

Long-term glycemic monitoring and management in advanced chronic kidney disease

Population	HbA1c	Frequency	Glycemic indexing by CGMI	Insulin requirements
CKD stages 1 to 5 including kidney transplant	Yes	Twice per year Up to 4 times per year if not achieving target or change in therapy	Correlate interstitial glucose with HbA1c for individual patients	Lower 25% to 30% basal insulin dose for patients with T1D and CKD 3 (129)
CKD stage 5 on dialysis	No	Not applicable	Consider	Lower 50% TDD for patients with type 2 with CKD-V (36) Lower total daily insulin dose by 35% to 40% for patients with T1D and CKD V (36) Lower (25%) basal insulin dose for pre-HD days (37)

CKD stages are based on estimated glomerular filtration ratios as follows: stage I: less than 90, stage II: 60 to 89, stage III: less than 60, stage IV: less than 30, stage V: less than 15 mL/min/1.73 m²).

Abbreviations: HbA_1c, hemoglobin A_1c; CKD, chronic kidney disease; CGMI, CGM Glucose Management Indicator; TDD, total daily insulin dose; HD, hemodialysis; T1D, type 1 diabetes; T2D, type 2 diabetes.

Table 2.

Antihyperglycemic agents dosing and considerations in advanced chronic kidney disease (labels reviewed as of April 1, 2020)

Medication	Metabolism and excretion	Labeling dosing recommendations by GFR (mL/min/1.73 m2)	Dose in ESKD and/or dialysis
Biguanides
Metformin	Kidney	No dose adjustment if eGFR > 45 mL/min/1.73 m2 Do not start and reduce dose if already on therapy and eGFR 30 to 45 mL/min/1.73 m2 Discontinue if eGFR < 30 mL/min/1.73 m2	Contraindicated because of risk of lactic acidosis
Second-generation sulfonylureas
Glipizide	Liver Excretion of < 10% of unchanged drug in urine	No dose adjustment if eGFR > 50 mL/min/1.73 m2	No adjustment, but conservative initial dose (eg, 2.5 mg daily) recommended Use with caution long-acting formulations because of risk of hypoglycemia
Glimepiride	Liver Excretion in urine 60% of drug	Consider alternative if eGFR < 15 mL/min/1.73 m2	Start lower dose of glimepiride (eg, 1 mg daily), caution recommended because of risk of hypoglycemia
Glyburide	Kidney Excretion of 50% of drug in urine	Avoid use	Contraindicated
Meglitinides
Nateglinide	Liver Excretion of 75% to 80% of drug in urine	No dose adjustment if eGFR > 30 mL/min/1.73 m2	Initiate conservatively at 60 mg with meals if eGFR < 30 mL/min/1.73 m2
Repaglinide	Liver Minimal excretion of parent drug in urine	No dose adjustment if eGFR > 30 mL/min/1.73 m2	Initiate conservatively at 0.5 mg with meals if eGFR < 30 mL/min/1.73 m2
DDPIV inhibitors
Sitagliptin	Kidney Excretion of 87% of unchanged drug in urine	100 mg daily if eGFR > 50 mL/min/1.73 m2 50 mg daily if eGFR 30 to 50 mL/min/1.73 m2 25 mg daily if eGFR < 30 mL/min/1.73 m2	Maximum dose of 25 mg daily
Saxagliptin	Liver/Kidney Excretion of 60% unchanged drug or active metabolite in urine	No dose adjustment if eGFR ≥ 45 mL/min/1.73 m2 Dose of 2.5 mg daily if eGFR ≤ 45 mL/min/1.73 m2	Maximum dose of 2.5 mg daily
Linagliptin	Liver Excretion of < 5% to 7% of drug in urine	No dose adjustment	No dose adjustment
Alogliptin	Kidney Excretion of 60% to 71% of unchanged drug in urine	25 mg daily if eGFR > 60 mL/min/1.73 m2 12.5 mg daily if eGFR 30 to 60 mL/min/1.73 m2 6.25 mg daily if eGFR < 30 mL/min/1.73 m2	6.25 mg daily
GLP1 RA agonists
Exenatide	Proteolytic degradation following glomerular filtration Excretion of majority of dose in the urine	No dose adjustment if eGFR > 50 mL/min/1.73 m2 Caution when initiating or escalating doses if eGFR 30 to 50 mL/min/1.73 m2 Not recommended with eGFR < 30 mL/min/1.73 m2	Contraindicated
Lixisenatide	Proteolytic degradation and glomerular filtration	No dose adjustment required for eGFR 60 to 89 mL/min/1.73 m2 No dose adjustment required for eGFR 30 to 59 mL/min/1.73 m2, but monitor patients for side effects and changes in kidney function Clinical experience is limited with eGFR 15 to 29 mL/min/1.73m2; monitor patients for side effects and changes in kidney function	Avoid if eGFR < 15 mL/min/1.73 m2
Albiglutide	Proteolytic degradation	No dose adjustment required for eGFR 15 to 89 mL/min/1.73 m2 Use caution with initiation or escalating doses and monitor for gastrointestinal reactions in patients with CKD	Not recommended
Liraglutide	Proteolytic degradation (not specific organ as a major route of elimination) Intact drug not detected in urine	No dose adjustment Post-marketing studies showed increased risk of gastrointestinal effects with higher doses Monitor for gastrointestinal reactions in patients with CKD	No dose adjustment Postmarketing studies showed increased risk of gastrointestinal effects with higher doses
Dulaglutide	Proteolytic catabolism	No dose adjustment	No dose adjustment Monitor eGFR in patients with CKD reporting severe adverse gastrointestinal reactions
Semaglutide injectable	Proteolytic cleavage of peptide backbone and sequential beta-oxidation of fatty acid sidechain Excretion of 3% of unchanged drug in urine	No dose adjustment Monitor eGFR function when initiating or escalating doses or in patients with adverse gastrointestinal reactions	No dose adjustment No clinically relevant change in semaglutide pharmacokinetics
Semaglutide oral		No dose adjustment Monitor eGFR when initiating or escalating doses or in patients with adverse gastrointestinal reactions	No dose adjustment No clinically relevant change in semaglutide pharmacokinetics
SGLT2 inhibitors			Expected not to be effective for glycemic control in advanced CKD
Canagliflozin	Liver Excretion of < 1% of unchanged drug in urine	No dose adjustment if eGFR ≥ 60 mL/min/1.73 m2 100 mg daily if eGFR 45 to 59 mL/min/1.73 m2 Avoid use and discontinue in patients with eGFR persistently < 45 mL/min/1.73 m2	Contraindicated
Dapagliflozin	Liver Excretion of < 2% of unchanged drug in urine	Avoid initiating if eGFR < 60 mL/min/1.73 m2 Not recommended with eGFR 30 to 60 mL/min/1.73 m2	Contraindicated with eGFR < 30 mL/min/1.73 m2 Unknown effect of hemodialysis
Empagliflozin	Liver Excretion of25% to 50% of unchanged drug in urine	No dose adjustment required if eGFR ≥ 45 mL/min/1.73 m2	Avoid use and discontinue in patients with eGFR persistently < 45 mL/min/1.73 m2
Ertugliflozin	Liver Excretion of1% of unchanged drug in urine	No dosage adjustment or increased monitoring needed in patients with mild CKD Safety and efficacy in mild-to-moderate CKD not established	Contraindicated
Alpha glucosidase inhibitors
Acarbose	Intestinal	Avoid if eGFR < 30 mL/min/1.73 m2	Contraindicated
Miglitol	Intestinal	Avoid if eGFR < 25 mL/min/1.73 m2	Contraindicated
Thiazolidinediones
Pioglitazone	Liver Excretion of negligible amount of unchanged drug in urine	No dose adjustment	No dose adjustment recommended Caution with use given fluid retention and adverse effects on bone metabolism
Amylin analog
Pramlintide	Kidney	No dose adjustments for eGFR > 20 to <50 mL/min/1.73 m2	No studies performed

Abbreviations: CKD, chronic kidney disease; DDPIV, dipeptidyl peptidase-4; eGFR, estimated glomerular filtration rate; ESKD, end-stage kidney disease; GLP1 RA, glucagon-like peptide receptor-1; SGLT2, sodium-glucose loop transporter-2.

Glycemic Monitoring by Continuous Glucose Monitoring in Advanced Chronic Kidney Disease

Pilot studies with continuous glucose monitoring in advanced chronic kidney disease

With the advent of CGM, hypoglycemia and GV are now known to be common, even in patients with well-controlled type 2 diabetes (86-88). Several studies have reported more frequent hypoglycemia in patients on hemodialysis compared with those not on dialysis (Table 3). Small pilot studies using CGM have recently found specific glycemic patterns for patients undergoing dialysis. For example, in patients without diabetes, the mean CGM glucose concentration was lower during hemodialysis hours than nonhemodialysis hours, with a nadir at the third hour after dialysis initiation. The mean CGM glucose was 128 (± 20) mg/dL the day before, 93 (± 8) mg/dL the day of hemodialysis, and 105 (± 13) mg/dL the day after (44). Similarly, it was reported that in patients with either type 1 diabetes or type 2 diabetes, there was a tendency toward lower CGM glucose levels during hemodialysis session hours, with the lowest glucose level occurring approximately 12 hours after hemodialysis. However, glucose levels tended to be higher on days off hemodialysis (107, 108). Sobngwi and colleagues performed a study in patients with type 2 diabetes on hemodialysis, using euglycemic clamp studies the days before and after hemodialysis. Because meal intake is also an important variable, the authors provided a fixed caloric intake (average 2200 calories), with 3 standardized meals and 2 snacks per day, and patients fasted for 8 to 24 hours before the index dialysis session. Insulin requirements were compared for different periods, including intradialysis and interdialysis times (37). The total daily dose to achieve euglycemia was 23.6 ± 7.7 vs 19.9 ± 4.9 units per day before and after hemodialysis, respectively, reflecting improved insulin sensitivity after a dialysis session. Notably, the basal insulin need was 25% lower during the posthemodialysis period.

Table 3.

Summary of studies using continuous glucose monitoring technology in hemodialysis patients with advanced chronic kidney disease

No.	Study author and year	Population	CGM duration	Results
1	Képénékian, 2014 (130)	DM2, HD, n = 28	2.25 d	Retrospective glycemic patterns reduced HbA1c, but no hypoglycemia
2	Gai, 2014 (131)	DM, HD, n = 12	6 d	Retrospective CGM use Lower glucose during HD, with nadir at third hour of HD Glycemic peak after HD occurring at 2.5 h after
3	Joubert, 2015 (108)	DM1, DM2, n = 15	5 d	Lower CGM glucose during HD period Retrospective use of CGM improved HbA1c and mean CGM
4	Vos, 2012 (99)	DM2, CKD no HD, n = 25	2 d	Good correlation of mean CGM and glycated albumin, poor with HbA1c
5	Mirani, 2010 (132)	DM2, HD, n = 12	3 d	High glycemic variability after HD
6	Jung, 2010 (133)	DM2, HD, n = 9	6 d	Hypoglycemia during and after HD
7	Sobngwi, 2010 (44)	No DM, HD, n = 14	4 d	Lower CGM glucose during HD, lowest glucose after 3 h
8	Riveline, 2009 (134)	DM2, HD, n = 19	4 d	Poor correlation of mean CGM and HbA1c
9	Chantrel, 2009 (135)	DM1, DM2, HD, n = 33	3 d	Frequent hypoglycemia during HD, with higher CGM glucose during early morning day after
10	Kazempour, 2009 (107)	DM2, HD, n = 17	2 d	Frequent hypoglycemic events, mostly after HD
11	Jhaverani, 2018 (136),	DM2, HD, n = 10	2 d (12 h on HD and 12 h no HD)	Mean glucose lower on HD day, particularly during 4 h of HD period
12	Sobngwi, 2010 (37)	DM, HD, n = 10	2 d (24 h before HD, 4 h on HD and 24 h post-HD)	Reduction of 25% in basal insulin requirements day after dialysis, compared to day before. No changes in bolus insulin requirements. Overall decrease of 15% of total daily insulin dose postdialysis
13	Presswala, 2019 (123)	DM2, CKD, no HD, n = 80	14 d (FreeStyle Libre Pro)	Robust correlation between CGM-derived average glucose and HbA1c and fructosamine Fructosamine not accurate for eGFR) < 30 mL/min; HbA1c not affected
14	Yajima, 2019 (124)	DM2, HD, n = 13	14 d (FreeStyle Libre Pro)	Overall MARD 19.5% MARD 31.9% for glucose < 70 mg/dL Only 49% and 51% of glucoses fell into zones A and B of EGA

Abbreviations: CGM, continuous glucose monitoring; DM2, type 2 diabetes; EGA, error grid analyses; eGFR, estimated glomerular filtration rate; HbA_1c, hemoglobin A_1c; HD, hemodialysis; MARD, mean amplitude relative difference.

Role of continuous glucose monitoring in advanced chronic kidney disease—moving beyond hemoglobin A_1c

CGM use has the advantage of providing better assessment of glycemic patterns and insulin needs among diabetic patients with advanced CKD. CGM has the potential to become a new standard of care for assessment of glycemic control in diabetic patients treated by maintenance hemodialysis given the well-known limitations of HbA_1c and other glycemic biomarkers (100). However, there are limited studies in the CKD population, particularly using novel factory-calibrated sensors. With the expansion of CGM studies and growing clinical use, there has been an evolution of “new/nontraditional glucose metrics,” such as 1) time in target range; 2) time in hyperglycemia range; 3) time in hypoglycemia range; and 4) GV. In the clinical realm, HbA_1c targets remains commonly used because of strong predictive capacity for diabetic complications. Additionally, there is headway to “move beyond HbA_1c” (109), and a recent international consensus conference provided guidance on clinical targets for CGM-derived glucose metrics (110, 111). Although there is consensus on the adoption of such novel CGM-derived glucose metrics, no studies have yet been conducted among patients on maintenance dialysis. Of note, a recent study by Beck et al showed a strong association of time in target range with retinopathy and microalbuminuria among patients with type 1 diabetes in the Diabetes Control and Complications Trial cohort (112).

The 24-hour glucose profile provided by CGM allows patients and health care providers to recognize glucose patterns, including responses to meals, medications, acute illness, or other stressors. Factory-calibrated CGM devices reduce the burden of diabetes care by reducing use of finger sticks for blood glucose monitoring. Moreover, CGM systems provide a benefit of recognizing declining glucose levels before occurrence of hypoglycemia and enable closed-loop insulin programs (“artificial pancreas”) by adjusting the insulin infusion rate to prevent hypoglycemia (eg, Tandem Basal and Control IQ [113, 114], Medtronic G670 [115, 116]).

CGM data can be used to generate a CGM index (glucose management indicator or GMI), which is a proxy for long-term glycemia in conjunction with the HbA_1c measurement in individual patients, allowing adjustment of glycemic goals accordingly (110, 111). CGM-estimated HbA_1c (eA_1c) was a term previously reported on CGM reports and derived from the CGM measured mean glucose. However, it was well recognized that CGM-eA1c and laboratory-measured HbA_1c did not correlate in clinical practice and in research studies. A formula (GMI [%] = 3.31 + 0.02392 × [mean glucose in mg/dL]) to calculate GMI was developed and validated by Beck and colleagues using modern CGM technology (117). After that, a multidisciplinary team of diabetologists, patients, and laboratory experts recommended the use of GMI in CGM reports, instead of eA_1c (118). An online calculator is provided by the Jaeb Center at http://www.agpreport.org/agp/links. GMI may be useful for patients with advanced CKD, including those treated with dialysis, for whom reliability of HbA_1c is low. It should be noted that the assay bias of HbA_1c relative to GMI could potentially change over time within patients, particularly when there are changes in clinical characteristics that affect red blood cell turnover or protein glycation.

Clinical experience, caveats, and potential interferences with continuous glucose monitors

Mostly used in research, the Yellow Spring Instrument 2300 STAT is considered the gold-standard reference for glucose monitoring assessment by regulatory agencies. The biosensor uses a GO-based membrane for oxidizing glucose to gluconolactone and hydrogen peroxide. The hydrogen peroxide oxidizes at the platinum anode of the electrochemical probe, producing an electron flow—hence also considered an amperometric method—proportional to the glucose concentration in the sample. Laboratory-based methods mostly use HK methods, less commonly GO or GDH. The HK method—also considered a spectrometric/photometric approach—relies on the conversion of glucose to glucose-6-phophate initially, then glucose-6-phophate conversion to 6-phosphogluconate with the formation of nicotinamide-adenine-dinucleotide. Nicotinamide-adenine-dinucleotide absorbs light, and the light absorption is proportional to the glucose concentration in the sample. Similar to the GO-based technique, amperometric GDH is specific for beta-D-glucose with little interference.

Among commercially available real-time CGM systems (119), 2 are factory calibrated and do not require finger stick blood glucose calibrations, namely: Abbott FreeStyle Libre (Abbott Diabetes Care) and Dexcom G6 (Dexcom, Inc). Other systems require finger stick blood glucose measurements for calibrations: Medtronic Guardian 3 and Enlite 2 (Medtronic, Inc), Dexcom G5, and Eversense (Senseonics, Inc)—the latter being the only implantable CGM. For patients on dialysis, with a large comorbidity burden and taking insulin, CGMs that do not require finger sticks for blood glucose measurements are expected to improve quality of life and decrease the burden of diabetes care.

Nonimplantable CGMs (Guardian, Dexcom, Abbott FreeStyle Libre) are attached to the skin of the upper extremities or abdominal area and use transcutaneous sensing, via a small filament wire inserted into the subcutaneous tissue, to measure glucose from the interstitial fluid (120). Reports of infections, skin reactions, or open wounds are very rare, with proven safety from several reports. However, there are several interfering factors that need to be considered for individual patients, such as anemia, hypoxia, or oxygen administration, high uric acid levels, and high doses of aspirin or acetaminophen (Table 4). Data regarding other clinical situations common (as shown in Table 4) in patients with advanced CKD are lacking, which is another important topic for research that will inform use of CGM in this population.

Table 4.

Interference and potential effects of certain substances on continuous glucose monitoring of glucoses

CGM system	Methodology	Interference	No interference	Potential interference based on methodology (106, 137)
Medtronic Guardian 3 (122, 126)	GO	Acetaminophen Ethanol/Wine Albuterol Lisinopril Atenolol Atorvastatin Ascorbic acid	Hydrochlorothiazide, losartan	Anemia Ascorbic acid Hypoxia hypothermia hypotension Mannitol Uric acid (> 20 mg/dL) Polycythemia oxygen therapyHypertriglyceridemia Bilirubin > 54 mg/dL
Dexcom G6 (127)	GO + Perm-selective membrane coating		Acetaminophen (up to 1000 mg every 6 h)
FreeStyle Libre 14 d (122, 138)	GO + Redox sensing membrane	Ascorbic acid Salicylic acid	Acetaminophen, methyldopa, tolbutamide
Senseonics Eversense (128)	Nonenzymatic electrochemical fluorescent-based polymer	Tetracycline Mannitol	Acetaminophen Ascorbic acid Amoxicillin Creatinine Heparin Levofloxacin Urea Galactose Maltose Lactose Xylose Sugar alcohols (eg, sweeteners) At therapeutic doses: salicylic acid lactate L-DOPA Piroxicam

Abbreviation: CGM, continuous glucose monitoring; GO, glucose oxidase-based.

The Abbott FreeStyle Libre, a factory-calibrated sensor, uses a wired GO-based enzyme technology with sensing membrane. In validating studies, salicylic acid and ascorbic acid were shown to affect sensor accuracy, probably by reacting with the sensor’s electrode (120, 121). However, the use of wired-enzyme technology with sensing membrane, working at very low voltage, was shown not to be susceptible to acetaminophen (122), a commonly used medication with known interference in electrochemical enzymatic-based sensors; and to reduce susceptibility to in vivo oxygen changes (both low or high), and decrease response to electrochemical interferent substances at therapeutic ranges, such as acetaminophen, and uric acid levels (121). There are no available studies assessing whether icodextrin, a component of peritoneal dialysis, interferes with the FreeStyle Libre sensor.

An observational study of 80 patients with type 2 diabetes and CKD stages 3 to 5 showed good correlation between the average glucose concentration measured by the FreeStyle Libre Pro CGM and serum HbA_1c and fructosamine. However, fructosamine was not accurate for eGFR less than 30 mL/min/1.73 m², whereas HbA_1c was not affected by different levels of eGFR from 7 to 45 mL/min/1.73 m² (123). In addition, a small observational study of 13 patients with type 2 diabetes on hemodialysis showed an overall mean amplitude relative difference of 19.5%, as compared to capillary blood glucose. However, the mean amplitude relative difference was 31.9% for glucose less than 70 mg/dL, and only 49% and 51% of glucose fell into zones A and B of error grid analyses (124). Thus, until more evidence becomes available, GO-based (eg, FreeStyle Libre) sensors should not be used in patients receiving peritoneal dialysis—similarly as described previously for glucose meters—and with caution in hemodialysis (eg, to assess patterns rather than focusing on specific glucose values) given data indicating lower accuracy of this method compared to others.

The Dexcom system uses an “advanced” (proprietary) version of electrochemical GO-based technology for glucose measurement. Since the development of the Dexcom G4 version, the sensor wire diameter was decreased, aiming to minimize trauma and to improved accuracy over time. In addition, the GO-based system was modified to reduce the impact from low oxygen states and to operate at a high signal-to-noise ratio, improving the performance at low glucose levels (125). In the pilot study by Basu et al testing the susceptibility of previous Dexcom and Medtronic CGM versions (Dexcom G4 Platinum and Medtronic Guardian Soft-Sensor), it was shown that several substances may interfere with reading of interstitial glucose by CGM sensors: acetaminophen, ethanol, albuterol, lisinopril, atenolol, and atorvastatin, and wine (122). In later studies Basu and colleagues confirmed that ingestion of 1 g of acetaminophen resulted in CGM measurements ranging from 85 to 400 mg/dL, whereas plasma glucose was maintained at 90 mg/dL (126). In the latest version, Dexcom G6, the manufacturers aimed to reduce the susceptibility to pharmacologic interferences by applying a perm-selective membrane coating to the sensor surface that inhibits the diffusion of potential pharmacologic spurious signal into the interior of the sensor (127). Consequently, a study of 66 patients with type 1 diabetes and type 2 diabetes confirmed that doses of 1 g of acetaminophen orally did not affect the accuracy, with a mean interference of 3.1 (SD ± 4.8) mg/dL from the comparator (127).

Implantable sensors (eg, Eversense) are less likely to be suitable for patients with ESKD treated by hemodialysis given the presence of vascular accesses (eg, catheters, fistulas) in the upper extremities, where the sensor is implanted. Eversense sensors use a nonenzymatic, electrochemical fluorescent-based, glucose-indicating polymer to measure glucose in the interstitial tissue (128). Hence, these sensors are less likely to have interference from substances that react at the electrode or that may cross-react with the enzyme (as detailed previously). However, substances that bind to the glucose-indicating polymer or that may absorb fluoresce light within the sensor’s spectrum may produce false readings. These substances include lactate, L-DOPA, piroxicam, pralidoxime iodide, salicylic acid, tetracycline, ribose, and mannitol. However, in studies performed during the validation of the sensors, only mannitol, sorbitol, and tetracycline showed a bias that exceeded limits of interference at therapeutic drug concentrations (128). Despite extensive data validating the improved accuracy of these sensors in different clinical scenarios, there are no studies evaluating interference from commonly used substances in the advanced CKD population.

Conclusion

Glucose and insulin metabolism in patients with diabetes are profoundly altered by advanced CKD. Risk of hypoglycemia is exacerbated by failure of gluconeogenesis in the kidney as well as reduced clearance of many antihyperglycemic agents, particularly insulin. Resistance to insulin is concurrently amplified by the uremic state. Patients with ESKD frequently experience wide glycemic exposures, with common occurrences of both hypoglycemia and hyperglycemia. Assessment of glycemia by HbA_1c is hampered by a variety of CKD-associated conditions that can bias the measure either to the low or high ranges. Alternative glycemic biomarkers, such as glycated albumin or fructosamine, are even less reliable than HbA_1c. Therefore, HbA_1c remains the preferred glycemic biomarker despite its limitations. Based on observational data for associations with mortality and risks of hypoglycemia with intensive glycemic control regimens in advanced CKD, an HbA_1c range of 7% to 8% appears to be most favorable. Except for documented interferences from icodextrin metabolites and nonglucose sugars in peritoneal dialysis solutions with the use of GDH-PQQ/GO-based glucose meters, the evidence is very limited in the ESKD population. The advent of CGM offers promise for more precise monitoring and treatment adjustments to permit fine-tuning of glycemic management in people with diabetes and advanced CKD.

Abbreviations

Abbreviations
 
• CGM

  continuous glucose monitoring

 
• CKD

  chronic kidney disease

 
• DKA

  diabetic ketoacidosis

 
• eA1c

  continuous glucose monitoring estimated hemoglobin A1c

 
• eGFR

  estimated glomerular filtration rate

 
• ESKD

  end-stage kidney disease

 
• FAD

  flavin adenine dinucleotide

 
• GDH

  glucose dehydrogenase–based

 
• GMI

  glucose management indicator

 
• GO

  glucose oxidase–based

 
• GV

  glycemic variability

 
• HbA1c

  hemoglobin A_1c

 
• HK

  hexokinase-based

 
• HR

  hazard ratio

 
• NAD

  nicotine adenine dinucleotide

 
• NKF-KDOQI

  National Kidney Foundation–Kidney Disease Outcomes Quality Initiative

 
• POC

  point of care

 
• PQQ

  pyrroloquinoline-quinone

Acknowledgments

References for this review were identified through searches of PubMed for articles published from January 1, 1990, to June 30, 2019, by use of the terms “renal failure,” “dialysis,” “renal replacement therapy,” “advanced renal failure”, “glycemic monitoring,” and “glucose control” in combination with the term “diabetes.” Relevant articles were identified through searches in the authors’ personal files. Articles resulting from these searches and relevant references cited in those articles were reviewed. Articles published in English were included.

Financial Support: This work was supported by grants from the National Institutes of Health (NIH)/National Institute of Diabetes and Digestive and Kidney Diseases (Grants P30DK11102 and 1K23DK123384-01 to R.J.G.); the NIH/National Center for Advancing Translational Sciences (Grant UL1 TR002378 from the Clinical and Translational Science Award program to G.E.U.) and the National Institute of Health and National Center for Research Resources (Grant 1P30DK111024-01 to G.E.U.); and the NIH (Grants 4UL1TR00426-10, 1U2CDK114886-01, 5UM1DK100846-03, 2U01DK10086-07, 1U54DK083912, and 2UC4DK101108-02 to K.R.T.) and the Centers for Disease Control and Prevention (Grant 75D301-19-Q-69877 to K.R.T.).

Additional Information

Disclosure Summary: R.J.G. has received research support to Emory University for investigator-initiated studies from Novo Nordisk, and consulting fees from Abbott Diabetes Care, Sanofi, Valeritas, Eli Lilly, and Novo Nordisk. R.W.B. has received consulting fees, paid to his institution, from Insulet, Bigfoot, and Lilly, grant support and supplies, provided to his institution, from Tandem and Dexcom, and supplies from Ascensia and Roche. G.E.U. has received research support to Emory University for investigator-initiated studies from Merck, Novo Nordisk, Dexcom Inc, and Sanofi. K.R.T. has received consulting fees from Eli Lilly and Company, Boehringer Ingelheim, Astra Zeneca, Gilead, Goldfinch Bio, Bayer, and Novo Nordisk. M.F.S. has nothing to disclose. All authors have submitted the International Committee of Medical Journal Editors Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Data Availability: Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.