Impaired Glucose and Insulin Homeostasis in Moderate-Severe CKD

Abstract

Kidney disease leads to clinically relevant disturbances in glucose and insulin homeostasis, but the pathophysiology in moderate-severe CKD remains incompletely defined. In a cross-sectional study of 59 participants with nondiabetic CKD (mean eGFR =37.6 ml/min per 1.73 m²) and 39 healthy control subjects, we quantified insulin sensitivity, clearance, and secretion and glucose tolerance using hyperinsulinemic-euglycemic clamp and intravenous and oral glucose tolerance tests. Participants with CKD had lower insulin sensitivity than participants without CKD (mean[SD] 3.9[2.0] versus 5.0 [2.0] mg/min per µU/ml; P<0.01). Insulin clearance correlated with insulin sensitivity (r=0.72; P<0.001) and was also lower in participants with CKD than controls (876 [226] versus 998 [212] ml/min; P<0.01). Adjustment for physical activity, diet, fat mass, and fatfree mass in addition to demographics and smoking partially attenuated associations of CKD with insulin sensitivity (adjusted difference, −0.7; 95% confidence interval, −1.4 to 0.0 mg/min per µU/ml) and insulin clearance (adjusted difference, −85; 95% confidence interval, −160 to −10 ml/min). Among participants with CKD, eGFR did not significantly correlate with insulin sensitivity or clearance. Insulin secretion and glucose tolerance did not differ significantly between groups, but 65% of participants with CKD had impaired glucose tolerance. In conclusion, moderate-severe CKD associated with reductions in insulin sensitivity and clearance that are explained, in part, by differences in lifestyle and body composition. We did not observe a CKD-specific deficit in insulin secretion, but the combination of insulin resistance and inadequate augmentation of insulin secretion led to a high prevalence of impaired glucose tolerance.

CKD leads to clinically relevant disturbances in glucose and insulin homeostasis. In the absence of diabetes, patients with ESRD have long been noted to have a high prevalence of glucose intolerance.^1,2 Decreased insulin sensitivity (i.e., insulin resistance) because of diminished postreceptor insulin signaling in muscle is an important contributor to glucose intolerance in ESRD.^3,4 In addition, patients with diabetes have been noted for >60 years to require lower doses of therapeutic insulin at low levels of GFR.^5,6 More recently, mild-moderate CKD has been associated with increased rates of hypoglycemia,⁷ particularly with intensive diabetes therapy.⁸ Decreased renal insulin clearance has been cited as a likely cause of these observations.

A number of important questions regarding glucose and insulin homeostasis in CKD remain to be answered. First, the effects of moderate-severe CKD on insulin sensitivity are not certain. Published studies of moderate-severe CKD have most commonly used estimates of insulin sensitivity on the basis of fasting insulin or insulin measured during an oral glucose tolerance test (OGTT), which may be biased in CKD because of changes in insulin clearance and may not accurately reflect changes in insulin sensitivity localized to muscle.^9,10 Furthermore, published studies have not accounted for important confounders and mediators, such as physical activity, diet, and body composition, or have found inconsistent associations of GFR with insulin sensitivity.^11–13 Second, effects of moderate-severe CKD on insulin clearance have not been quantified, hindering our understanding and treatment of hyperglycemia in this setting. Third, it is not clear whether insulin secretion is appropriately augmented to compensate for insulin resistance and maintain normal glucose tolerance in moderate-severe CKD.

We designed the Study of Glucose and Insulin in Renal Disease (SUGAR) to address these gaps in knowledge. Specifically, we studied nondiabetic human subjects with moderate-severe CKD and matched healthy control subjects with neither diabetes nor CKD using gold standard measurements of glucose and insulin homeostasis. We hypothesized that nondiabetic moderate-severe CKD is associated with significant reductions in insulin sensitivity and clearance and that insulin secretion would be inadequate to fully compensate for decreased insulin sensitivity, thus leading to glucose intolerance.

Results

Participant Characteristics

Of 98 participants in SUGAR, 59 had CKD (defined as eGFR<60 ml/min per 1.73 m²), and 39 did not (eGFR≥60 ml/min per 1.73 m²; defining control subjects). Among participants with CKD, mean age was 63.6 years old, 51% were women, and race was self-reported as black for 22% and Asian or Pacific Islander for 8% (Table 1). Chart review revealed that hypertension was most often cited as the likely cause of CKD, but few participants had undergone kidney biopsies (Supplemental Table 1). Mean eGFR for participants with CKD was 37.6 ml/min per 1.73 m², with 31%, 37%, and 32% falling in the ranges 8 to <30, 30 to <45, and 45 to <60 ml/min per 1.73 m², respectively. Albumin excretion rate (AER) was ≥30 mg/24 h for 53% of participants with CKD, including 20% with AER≥300 mg/24 h. Mean eGFR for controls was 88.8 ml/min per 1.73 m², and only one control had AER>30 mg/24 h. Compared with participants with CKD, controls were less likely to be treated with antihypertensive medications (33% versus 90%, respectively), were less likely to self-report physician–diagnosed cardiovascular disease (5% versus 32%, respectively), were more physically active, and had greater lean body mass and lower fat mass.

Table 1. - Characteristics of participants in SUGAR

Characteristic	eGFR<60 ml/min per 1.73 m2	eGFR≥60 ml/min per 1.73 m2
N	59	39
Demographics
Age, yr	63.6 (13.9)	61.0 (12.4)
Women	30 (51)	17 (44)
Race
White	41 (69)	34 (87)
Black	13 (22)	4 (10)
Asian/Pacific Islander	5 (8)	1 (3)
Medical history and lifestyle
Cardiovascular disease	19 (32)	2 (5)
Current smoking	10 (17)	3 (8)
Median physical activity, HAP score	78.0 (73.0–82.0)	85.0 (77.0–90.5)
Dietary intake
Energy, kcal	1758.2 (540.8)	2047.9 (556.4)
Fat, g	70.6 (26.4)	81.8 (32.1)
Carbohydrates, g	209.2 (79.0)	243.1 (78.3)
Protein, g	71.0 (26.1)	79.6 (24.5)
Glycemic index	60.1 (5.5)	58.3 (4.2)
Medication use
Any antihypertensive medication	53 (90)	13 (33)
RAS antagonists	38 (64)	8 (21)
Diuretics	27 (46)	2 (5)
β-Blockers	23 (39)	3 (8)
Calcium-channel blockers	27 (46)	3 (8)
Physical characteristics
Height, cm	170.4 (10.3)	172.9 (10.7)
Weight, kg	88.1 (19.8)	82.6 (20.6)
Body mass index, kg/m2	30.2 (6.0)	27.5 (6.3)
Fatfree mass, kg	53.7 (11.7)	56.1 (13.1)
Fat mass, kg	31.9 (11.6)	27.7 (14.0)
Systolic BP, mmHg	134.5 (15.2)	122.9 (13.1)
Diastolic BP, mmHg	80.6 (9.4)	77.0 (10.0)
Laboratory data
Median serum creatinine, mg/dl	1.7 (1.5–2.1)	0.9 (0.8–1.0)
Median serum cystatin C, mg/L	1.6 (1.4–2.0)	0.9 (0.7–1.0)
eGFR, ml/min per 1.73 m2	37.6 (12.5)	88.8 (17.1)
Median urine AER, mg/24 h	39.2 (14.3–225.1)	5.7 (3.5–8.5)
Aspartate transaminase, IU/L	22.0 (18.5–27.5)	23.0 (21.0–27.0)
Alanine transaminase, IU/L	19.0 (15.0–24.0)	19.0 (17.0–22.0)
Hemoglobin, g/dl	13.2 (1.4)	13.9 (1.5)
Median parathyroid hormone, pg/ml	70.0 (51.0–96.0)	45.0 (35.0–62.0)

Data are means (SDs) for continuous variables, N (percentages) for categorical variables, and medians (interquartile ranges) where noted. Some data were missing for fatfree mass and fat mass (n=5), AER (n=3), and dietary intakes (n=5). HAP, human activity profile; RAS, renin-angiotensin system.

Insulin Sensitivity

The insulin sensitivity index (SI) measured using the hyperinsulinemic-euglycemic clamp was lower in participants with versus without CKD (mean =3.9 versus 5.0 mg/min per microunits per milliliter; P<0.01) (Table 2). One participant was an outlier, with an eGFR of 9.5 ml/min per 1.73 m² and an SI of 13.0 mg/min per microunits per milliliter (Figure 1). Participants with CKD had lower insulin sensitivity than control subjects, regardless of whether control subjects had hypertension and were within strata defined by plasma glucose concentration 2 hours after oral glucose administration (Supplemental Figures 1 and 2). Among participants with CKD, there was no clear relationship of eGFR with insulin sensitivity with or without excluding the outlier (β excluding outlier =0.08 mg/min per microunits per milliliter per 10 ml/min per 1.73 m² eGFR; 95% confidence interval [95% CI], −0.07 to 0.23; P=0.30). Of recorded clinical characteristics, physical activity (r=0.25; P=0.01) and fat mass (r=−0.20; P=0.05) correlated most strongly with SI, white participants tended to have higher SI, and use of β-blockers was associated with lower insulin sensitivity (Supplemental Tables 2–4). Compared with normal eGFR, CKD was associated with a 0.9-mg/min per microunits per milliliter lower SI (95% CI, 0.1 to 1.6 lower; P=0.03) adjusting for age, sex, and race. Additional adjustment for a number of clinical characteristics modestly altered the estimated magnitude of association (Table 3). In a final parsimonious multivariable model, CKD was associated with a 0.7-mg/min per microunits per milliliter lower SI (95% CI, 0 to 1.4 lower; P=0.06). Excluding the outlier with eGFR=9.5 ml/min per 1.73 m², CKD was associated with a 0.8-mg/min per microunits per milliliter lower SI in the final multivariable model (95% CI, 0.1 to 1.5 lower; P=0.02) (Supplemental Table 5). AER was not associated with SI (Supplemental Figure 3).

Table 2. - Measurements of glucose and insulin homeostasis in the SUGAR

Measurement	eGFR<60 ml/min per 1.73 m2, n=59	eGFR≥60 ml/min per 1.73 m2, n=39	P Value
Fasting measurements
Fasting glucose, mg/dl	102.7 (8.8)	100.1 (10.1)	0.19
Fasting insulin, μU/ml	10.1 (5.3)	7.3 (5.2)	0.01
Hemoglobin A1c, %	5.6 (0.4)	5.5 (0.4)	0.56
Hyperinsulinemic-euglycemic clamp measurements
Steady-state insulin, μU/ml	190.0 (39.5)	162.0 (34.2)	<0.001
Glucose disposal rate, mg/min	657.1 (199.2)	739.1 (212.1)	0.06
Insulin sensitivity, mg/min per microunit per milliliter	3.9 (2.0)	5.0 (2.0)	<0.01
Insulin clearance, ml/min	876.1 (225.7)	998.3 (212.1)	<0.01
IVGTT measurements
Acute insulin response, μU × min/ml	339.7 (241.3–529.4)	332.5 (197.5–546.7)	0.84
OGTT measurementsa
Matsuda index	4.2 (2.5)	6.6 (5.0)	<0.01
Insulinogenic index, μU/ml per milligram per deciliter	0.8 (0.5–1.4)	0.7 (0.4–1.0)	0.67
Glucose AUC, (mg × min)/ml	19,407 (17,512–21,680)	18,667 (16,930–21,661)	0.50
2-h plasma glucose, mg/dl, N (%)			0.30
<140	20 (35)	19 (49)
140–199	31 (54)	15 (38)
≥200	6 (11)	5 (13)

Clamp insulin sensitivity and the Matsuda index were the primary and secondary measures of insulin sensitivity, respectively. Acute insulin response, quantified as the incremental plasma insulin AUC during minutes 2–10 of the IVGTT glucose, and insulinogenic index, quantified as the difference in plasma insulin divided by the difference in plasma glucose from baseline to 30 minutes of the OGTT, were the primary measures of insulin secretion. Glucose AUC and 2-hour plasma glucose during the OGTT were the primary measures of glucose tolerance.

^aOGTTs were not completed for two participants with eGFR<60 ml/min per 1.73 m². Cell contents are means (SDs), except for acute insulin response and insulinogenic index (medians [interquartile ranges]) and 2-hour plasma glucose values (N [percentages]).

Figure 1.: Estimated GFR<60 mL/min/1.73 m² was associated with reduced insulin sensitivity and clearance but not insulin secretion or glucose tolerance. Distributions of insulin sensitivity (measured by hyperinsulinemic-euglycemic clamp), insulin clearance (measured by hyperinsulinemic-euglycemic clamp), insulin secretion (acute insulin response to intravenous glucose), and glucose tolerance (glucose AUC during the OGTT) are summarized using boxplots and scatterplots.

Table 3. - Associations of CKD with measures of glucose and insulin homeostasis

Covariate Adjustments	Insulin Sensitivity		Insulin Clearance		Insulin Secretion		Glucose Tolerance
	Difference (95% CI), mg/min per microunit per milliliter	P Value	Difference (95% CI), ml/min	P Value	Percentage Difference (95% CI)	P Value	Difference (95% CI), (mg × min)/ml	P Value
None (unadjusted)	−1.1 (−1.9 to −0.3)	<0.01	−122 (−209 to −35)	<0.01	3 (−24 to 40)	0.84	487 (−929 to 1904)	0.50
Age, race, and sex	−0.9 (−1.6 to −0.1)	0.03	−85 (−165 to −4)	0.04	−3 (−30 to 35)	0.86	375 (−1177 to 1926)	0.64
Height and weight	−0.7 (−1.4 to 0.1)	0.09	−93 (−172 to −15)	0.02	−10 (−36 to 26)	0.54	163 (−1352 to 1679)	0.83
Fat mass	−0.7 (−1.4 to 0.1)	0.07	−95 (−177 to −13)	0.02	−11 (−37 to 27)	0.52	218 (−1335 to 1772)	0.78
Fatfree mass	−0.9 (−1.7 to −0.1)	0.02	−79 (−158 to 0)	0.05	0 (−27 to 36)	0.99	337 (−1211 to 1886)	0.67
Physical activity	−0.8 (−1.6 to 0.1)	0.08	−93 (−182 to −3)	0.04	−4 (−32 to 34)	0.79	155 (−1426 to 1736)	0.85
Macronutrient intake	−0.9 (−1.7 to −0.1)	0.03	−86 (−165 to −7)	0.03	−4 (−31 to 33)	0.81	406 (−1103 to 1914)	0.60
Glycemic index	−0.8 (−1.5 to 0)	0.05	−76 (−154 to 3)	0.06	−5 (−32 to 32)	0.75	372 (−1191 to 1934)	0.64
Smoking	−1 (−1.7 to −0.3)	<0.01	−97 (−172 to −22)	0.01	0 (−28 to 39)	0.98	249 (−1310 to 1808)	0.76
CVD	−0.8 (−1.6 to 0)	0.05	−90 (−174 to −6)	0.04	3 (−26 to 43)	0.85	61 (−1526 to 1648)	0.94
Fully adjusted model	−0.7 (−1.4 to 0)	0.06	−85 (−160 to −10)	0.03	−5 (−33 to 34)	0.78	−63 (−1660 to 1534)	0.94

Cell contents are the differences associated with CKD (versus control subjects), with 95% CIs and P values. Individual covariates were added one at a time (two at one time for height and weight) to the base model that included age, sex, and race. The fully adjusted model adjusts for age, sex, race, fat mass, fatfree mass, physical activity, glycemic index, and smoking. Insulin sensitivity and clearance were measured during the hyperinsulinemic-euglycemic clamp. Insulin secretion was measured as the acute insulin response to intravenous glucose (i.e., the incremental plasma insulin AUC during minutes 2–10 of the IVGTT). Glucose tolerance was measured as glucose AUC during the OGTT. CVD, cardiovascular disease.

Insulin Clearance

Insulin clearance measured during the clamp correlated strongly with SI (r=0.72; P<0.001). This relationship was similar among participants with and without CKD (Figure 2). Insulin clearance also correlated moderately with glucose disposal rate during the clamp (r=0.47; P<0.001) and insulin sensitivity estimated from the OGTT using the Matsuda index (r=0.31; P=0.002). Insulin clearance was lower in participants with versus without CKD (mean =876 versus 998 ml/min) (Table 2). Among participants with CKD, the correlation of eGFR with insulin clearance was positive but not statistically significant (β=14-ml/min clearance per 10 ml/min per 1.73 m² eGFR; 95% CI, −0.6 to 28.5 ml/min; P=0.06) (Figure 1). Of recorded clinical characteristics, fatfree mass correlated most strongly with insulin clearance (r=0.46; P<0.001) (Supplemental Tables 2 and 3). In a final parsimonious multivariable model, CKD was associated with an 85-ml/min lower insulin clearance (95% CI, 10 to 160 ml/min lower; P=0.03) (Table 3). AER was not associated with insulin clearance (Supplemental Figure 2).

Figure 2.: Insulin sensitivity and insulin clearance were highly correlated. Scatterplots show linear regression lines stratified by CKD status.

Insulin Secretion

We evaluated the acute insulin response to intravenous glucose and the insulinogenic index (also known as the early insulin response to oral glucose). During the intravenous glucose tolerance test (IVGTT), the first–phase insulin response (the incremental plasma insulin area under the curve [AUC] from 2 to 10 minutes) did not differ comparing participants with versus without CKD (Figure 3, Tables 2 and 3). During the OGTT, the insulinogenic index (the difference in plasma insulin divided by the difference in plasma glucose from baseline to 30 minutes) also did not differ (Supplemental Table 6). Results were similar with additional adjustment for insulin SI. Plasma insulin concentrations beyond 10 minutes of the IVGTT and 30 minutes of the OGTT were higher among participants with CKD (Figure 3).

Figure 3.: Insulin secretion (quantified as the acute insulin response to intravenous or oral glucose) and glucose tolerance (measured during the OGTT) did not differ among participants with and without CKD. Data points and error bars are means and 95% CIs, respectively. (A) Plasma insulin during the IVGTT; acute insulin response is calculated as incremental AUC from minutes 2–10. (B) Plasma insulin during the OGTT. (C) Plasma glucose during the OGTT.

Glucose Tolerance

Glucose tolerance measured by OGTT was frequently abnormal among participants with and without CKD (Table 2). Among participants with CKD, 54% had impaired glucose tolerance (2-hour plasma glucose =140–199 mg/dl), and an additional 11% met criteria for occult diabetes (2-hour plasma glucose ≥200 mg/dl). Neither glucose tolerance categories nor glucose AUC differed significantly comparing participants with and without CKD (Figure 3, Tables 2 and 3).

Discussion

In this study, moderate-severe nondiabetic CKD was associated with reduced insulin sensitivity and insulin clearance measured using the gold standard hyperinsulinemic-euglycemic clamp method. With adjustment for lifestyle factors and body composition, magnitudes of association were attenuated, and the association of CKD with insulin sensitivity was of borderline statistical significance. Among participants with CKD, we observed wide ranges of insulin sensitivity and clearance. Neither insulin sensitivity nor insulin clearance correlated significantly with eGFR. However, insulin sensitivity and clearance correlated strongly with each other, suggesting a physiologic link between these processes. Adiposity and lean body mass were also strongly correlated with insulin sensitivity and clearance. Insulin secretion did not differ comparing participants with and without CKD but was sufficient to maintain normal glucose tolerance in only 35% of participants with CKD.

In the published literature, the relationship of kidney function across its spectrum with insulin sensitivity is not clear.¹⁰ DeFronzo et al.³ showed that insulin sensitivity measured by clamp was substantially reduced in patients with ESRD compared with healthy control subjects.³ On the other end of the GFR spectrum, in community-based cohorts with mean eGFRs in the normal range, lower eGFR has been generally associated with higher fasting insulin concentration and lower OGTT-derived estimates of insulin sensitivity.^14–16 In addition, in a population of 70-year-old Swedish men not selected for CKD, eGFR and insulin sensitivity measured by clamp were directly correlated.¹⁷ Fewer studies have examined insulin sensitivity in moderate-severe CKD. Fliser et al.¹¹ studied 50 patients with IgA GN or polycystic kidney disease and 16 healthy controls using the frequently sampled IVGTT. CKD was associated with reduced insulin sensitivity, but there was no correlation of GFR (measured as inulin clearance) with insulin sensitivity. Similarly, in a study of 95 participants with stage 3 or 4 CKD, Trirogoff et al.¹³ reported that adiposity but not eGFR was associated with insulin resistance estimated using fasting insulin concentration. In contrast, among 29 Japanese patients with moderate-severe CKD and 10 healthy controls, Kobayashi et al.¹² reported a direct correlation of eGFR with glucose disposal rate measured by clamp.

Our results, generated using gold standard methods in a diverse clinic–based population, support the notion that insulin sensitivity is generally reduced across the spectrum of CKD, consistent with published studies. Within moderate-severe CKD, we did not detect as association of eGFR with insulin sensitivity. We cannot rule out the possibility that such an association exists, because our power was limited by sample size. Nonetheless, our results suggest that factors other than GFR are the major determinants of reduced insulin sensitivity (i.e., insulin resistance) in moderate-severe CKD, congruent with the studies by Fliser et al.¹¹ and Trirogoff et al.¹³ Alternative candidate contributors to insulin resistance include elements of kidney function beyond GFR (e.g., ability to secrete small molecule uremic toxins); CKD complications, such as impaired mineral metabolism and acid-base homeostasis; medications; physical activity; diet; and body composition.

In our study, the association of moderate-severe CKD with reduced insulin sensitivity was partially attenuated with adjustment for lifestyle factors and body composition. Although some prior studies of CKD and insulin sensitivity have measured body composition¹³ or lifestyle factors,¹⁴ none accounted for both using high-quality measurements and measuring insulin sensitivity directly. Our results highlight differences in physical activity, diet, and adiposity as important confounders or mediators of insulin resistance in CKD. Most of our participants with CKD had hypertension, a known correlate of insulin resistance, as expected. However, participants with CKD had lower insulin sensitivity than control subjects with or without hypertension, suggesting that hypertension does not explain the association of CKD with reduced insulin sensitivity. Still, use of specific antihypertensive medications may contribute to insulin resistance in CKD. For example, β-blockers were used more commonly among participants with versus without CKD and associated with reduced insulin sensitivity in our study. Overall, our data suggest that insulin resistance in moderate-severe CKD is because of a combination of factors that are common to many conditions (e.g., physical inactivity, diet, and adiposity), differences in medication use, and factors that are more specific to CKD.

We measured whole–body insulin clearance directly during the clamp. The rate of insulin clearance was high (876 or 998 ml/min with or without CKD, respectively), consistent with the known short circulating half-life of insulin, which regulates blood glucose and other aspects of metabolism on a minute-to-minute basis. In an insulin dose-response study covering a wide range of insulin concentrations, the splanchnic bed (including the liver) accounted for approximately 70% of total insulin clearance, with the kidneys accounting for approximately 20% of insulin clearance.¹⁸ In that study and others using renal vein catheterization, net renal insulin clearance has been estimated to be approximately 200 ml/min in healthy subjects, with two thirds attributable to glomerular filtration and tubular reabsorption and one third attributable to tubular uptake through peritubular capillaries.^18–20 Insulin taken into renal tubules via either route is then catabolized by insulin-degrading enzyme.^21,22 In our study, participants with versus without CKD had a difference in mean eGFR of 51 ml/min per 1.73 m² and an adjusted difference in mean insulin clearance of 77 ml/min. This mean difference in insulin clearance is consistent with the mean difference in GFR and a proportional loss of insulin extraction through peritubular capillaries.

Among participants with CKD, there was a wide range of insulin clearance. Surprisingly, this variation was not significantly correlated with eGFR. Instead, insulin clearance was strongly correlated with insulin sensitivity and lean body mass. The observed correlation of insulin clearance with clamp SI (r=0.72) could be artificially high, because steady–state insulin concentration is included in the denominator of each, but the association was confirmed using glucose disposal rate or the Matsuda index in place of SI. Strong correlations of insulin clearance and sensitivity have been reported in the settings of obesity, glucose intolerance, and cirrhosis as well as in dogs.^23–27 It has been suggested that decreased insulin clearance compensates for decreased insulin sensitivity or vice versa.^23–28 However, experimental models suggest that the mechanisms of insulin clearance and action may be physiologically linked. In hepatocytes, the binding of insulin to its receptor is the first step for initiation of intracellular insulin signaling cascade and for endocytosis and degradation of insulin by insulin-degrading enzyme.^22,29 Our data provide support from an in vivo human experiment that insulin sensitivity and clearance are coupled. The fairly strong observed correlation of insulin clearance with fatfree mass further suggests that muscle may play a larger role in insulin clearance than generally acknowledged, at least during administration of exogenous insulin.

By design, our study excluded people with diabetes, for whom the effect of CKD on insulin clearance could differ. With this caveat, our results call into question whether decreasing insulin requirements among patients with diabetes and advanced CKD can be explained by decreased renal insulin clearance alone.^5,6,30 Differences in mean whole–body insulin clearance by CKD status were modest in our study, even at the lowest eGFRs. Our data suggest that nonrenal insulin clearance may decrease in patients with advanced CKD who develop progressive insulin resistance. If muscle is, indeed, an important site of insulin clearance, muscle wasting or impaired muscle function could potentially contribute to decreased insulin requirements. Impaired glucose production caused by malnutrition, reduced renal gluconeogenesis, or an impaired counter-regulatory response may also modify insulin requirements and contribute to hypoglycemia in advanced CKD.

In health, the islet β-cell increases insulin secretion to compensate for insulin resistance and maintain glucose tolerance.³¹ In our study, insulin secretion (measured as the early plasma insulin response to intravenous or oral glucose) and glucose tolerance (measured by OGTT) did not differ significantly comparing participants with and without CKD. Nonetheless, insulin secretion was sufficient to maintain normal glucose tolerance for only 35% of participants with CKD. Our results suggest that the combination of insulin resistance and an inability to adequately augment insulin secretion leads to a high prevalence of glucose intolerance in moderate-severe CKD.

Published data on CKD and insulin secretion are sparse. Among nondiabetic older adults studied by OGTT, moderate CKD was associated with insulin resistance and also, appropriately augmented insulin secretion, yielding no significant differences in glucose tolerance or risk of incident diabetes over longitudinal follow-up.¹⁴ In contrast, among patients on hemodialysis who were nondiabetic, Idorn et al.³² reported that a diminished incretin effect and severe fasting hyperglucagonemia contributed to impaired gastrointestinal–mediated glucose disposal compared with that in normal control subjects. It is possible that a defect in insulin secretion develops with progressive loss of renal function. Additional work is required to determine whether CKD-specific factors affect insulin secretion and glucose tolerance across stages of CKD.

Strengths of our study include the enrollment of a diverse population with the full range of moderate to severe CKD, the use of gold standard methods to accurately and comprehensively measure glucose and insulin metabolism, and the collection of high–quality covariate data. With these strengths, our study makes a number of important contributions to the literature, including the evaluation of stages 3 and 4 CKD, which has high public health relevance because of its high prevalence and high cardiovascular risk; the simultaneous evaluation of insulin sensitivity and clearance, which were strongly correlated, suggesting a new basis for hypoglycemia in advanced CKD; the complementary evaluation of insulin secretion and glucose tolerance to more comprehensively characterize abnormal glucose metabolism in CKD; and the thorough evaluation of lifestyle factors and body composition, which seem to explain a substantial portion of the insulin resistance seen in this population. Limitations include the inability to assess specific sites or mechanisms of insulin sensitivity or clearance; the relatively small study size, which reduces power to detect small associations independent of numerous covariates; and the cross-sectional nature of the study design, which precludes causal inference and the evaluation of health outcomes over time. In addition, enrollment of our study necessarily involved some degree of participant self-selection. Therefore, differences in the degree or type of self-selection by CKD control category may have introduced bias, and our results may not be generalizable to all patients with moderate-severe CKD.

In conclusion, moderate-severe nondiabetic CKD is associated with reduced insulin sensitivity and clearance, which were highly correlated in our population. Decreased physical activity, diet, and body composition (greater adiposity with reduced lean mass) seem to confound or mediate these associations in part. Insulin secretion and glucose tolerance did not differ significantly comparing participants with and without CKD, but insulin secretion was sufficient to maintain normal glucose tolerance for only 35% of participants with CKD. Additional studies are needed to determine whether CKD-specific factors contribute to changes in glucose and insulin homeostasis and whether insulin resistance and glucose intolerance contribute to adverse clinical health outcomes in CKD.

Concise Methods

Study Population

The SUGAR is a cross-sectional study of glucose and insulin metabolism in moderate-severe nondiabetic CKD. From 2011 to 2014, we recruited participants from nephrology and primary care clinics associated with the University of Washington and neighboring institutions in Seattle, Washington. Potentially eligible patients were approached at clinic visits or by mail, and study brochures were placed in nephrology clinics for self-referral. Two preexisting observational studies enrolled participants with and without CKD from the same patient base, and participants in these studies were also screened and contacted for SUGAR. All interested and potentially eligible individuals were invited to attend a screening visit, at which eligibility was assessed and written informed consent was obtained (Supplemental Figure 4).

From this population, we enrolled participants with moderate-severe CKD defined as eGFR<60 ml/min per 1.73 m². We then recruited control subjects with eGFRs≥60 ml/min per 1.73 m² and spot urine albumin-to-creatinine ratios <30 mg/g, targeting the distributions of age, sex, and race of enrolled participants with CKD. Eligibility was determined at the screening visit, at which eGFR was calculated from serum creatinine measured at a clinical laboratory. Exclusion criteria for both groups included age <18 years old, a clinical diagnosis of diabetes, maintenance dialysis or fistula in place, history of kidney transplantation, use of medications known to reduce insulin sensitivity (including corticosteroids and immunosuppressants), fasting serum glucose ≥126 mg/dl, and hemoglobin <10 g/dl.

Of 157 participants providing informed consent, 34 did not qualify for the SUGAR, and 24 others did not attend the first study visit (Supplemental Figure 1). One participant developed fasting hyperglycemia between screening and study visits and was excluded after participation but before analyses, leaving a final analytic sample of 98 participants.

eGFR and Albuminuria

Serum creatinine and cystatin C (Gentian) were measured in fasting serum collected immediately before the clamp using a Beckman DxC Automated Chemistry Analyzer (Beckman Coulter, Inc., Brea CA). Creatinine and cystatin C concentrations are traceable to isotope dilution mass spectrometry and ERM-DA471 from the International Federation of Clinical Chemistry, respectively. Interassay coefficients of variation were 1.5%–3.0%. GFR was estimated from creatinine and cystatin C concentrations using the Chronic Kidney Disease Epidemiology Collaboration formula.³³ AER was measured using 24-hour urine samples. Urine albumin was measured using a turbidimetric method on a Beckman DxC Automated Chemistry Analyzer (Beckman Coulter, Inc.; interassay coefficient of variation =0.8%–1.7%). The eGFR calculated from creatinine and cystatin C was used for all analyses. With this measurement, compared with screening, four participants recruited as patients with CKD were reclassified as controls (three had normal AER, one had AER=52 mg/d, and none had a documented cause of CKD), and one control (with AER 39 mg/d) was reclassified as having CKD.

Measurements of Glucose and Insulin Homeostasis

Insulin clearance and sensitivity were measured using the hyperinsulinemic-euglycemic clamp technique adapted from the method by DeFronzo and coworkers.^2,3,34 Because the optimal insulin infusion rate in our target population was not known, we first performed a dose-ranging study among five participants with moderate-severe CKD. After a 2-hour infusion of deuterated glucose, each participant completed two or three successive periods, during which insulin was infused at 40, 80, and 400 mU/m² per minute (Supplemental Material). Endogenous glucose production was suppressed among only two of five participants at 40 mU/m² per minute, but in all five participants at 80 mU/m² per minute, glucose disposal rate increased linearly with attained insulin concentration through 80 mU/m² per minute; glucose disposal rate was submaximal at 80 mU/m² per minute (Figure 4). We, therefore, chose to infuse insulin at 80 mU/m² per minute for the main SUGAR.

Figure 4.: The insulin infusion rate used during the hyperinsulinemic-euglycemic clamp was determined using an insulin dose-ranging study. Five participants completed hyperinsulinemic-euglycemic clamps with multiple insulin infusion rates (40 and 80 mU/m² per minute, with n=3 also receiving 400 mU/m² per minute) on a single day. (A) Endogenous glucose production and (B) glucose disposal rate are plotted for each participant by insulin infusion rate or attained insulin concentration, respectively, with lines connecting data points for each individual. Dotted lines in B reflect the break in x axis for insulin concentrations attained at 80 versus 400 mU/m² per minute. Endogenous glucose production, determined using infusion of deuterated glucose, was suppressed for two of five participants at 40 mU/m² per minute, five of five participants at 80 mU/m² per minute, and three of three participants at 400 mU/m² per minute (data points for highest dose not shown). Glucose disposal rate, a measure of insulin sensitivity, increased fairly linearly from fasting to 40 and 80 mU/m² per minute. For three participants receiving insulin at 400 mU/m² per minute, glucose disposal rate was higher at 400 versus 80 mU/m² per minute.

Each participant was admitted to the University of Washington Clinical Research Center after an overnight fast. Intravenous catheters were placed in peripheral veins in each upper extremity and kept patent with a slow infusion of normal saline. One arm was warmed to allow for the sampling of arterialized blood. Three fasting plasma samples were drawn 5 minutes apart. The IVGTT was then initiated with an infusion of unlabeled 20% dextrose (11.4 g/m² over 60 seconds), after which plasma was frequently sampled for insulin and glucose concentrations. Thirty minutes after commencing the dextrose infusion, an insulin infusion was initiated as a prime (160 mU/m² per minute for 5 minutes) followed by a constant rate (80 mU/m² per minute). A variable rate infusion of unlabeled 20% dextrose was administered to maintain blood glucose (measured every 5 minutes) at approximately 90 mg/dl. Beginning 120–150 minutes after initiation of the insulin infusion, the dextrose infusion rate was held constant for 30 minutes, over which time three steady–state plasma samples were obtained 15 minutes apart. Plasma concentrations of insulin (two–site immune–enzymometric assay; Tosoh 2000 Autoanalyzer) and glucose (glucose hexokinase method; Roche Module P Chemistry Autoanalyzer; Roche, Basel, Switzerland) as well as the dextrose concentration of the infusate were measured at the Northwest Lipid Research Laboratories (Seattle, WA).

The glucose disposal rate was calculated as the glucose infusion rate during the last 30 minutes of the clamp adjusted for the drift in plasma glucose concentration using the Steels nonsteady–state equations.³⁴ Insulin sensitivity (SI) was calculated as (glucose disposal rate × concentration of infused glucose)/(insulin concentration at steady state − fasting insulin concentration). The inclusion of insulin concentration in the denominator of SI accounts for variability in achieved insulin concentrations and is supported by the linear relationship between insulin concentration and glucose disposal rate within individuals in our dose-ranging study (Figure 4). We adjusted for body size parameters rather than incorporating lean mass into SI to evaluate the relationships of different body size metrics with SI. Insulin clearance was calculated as the insulin infusion rate divided by steady–state insulin concentration. Acute insulin response was calculated as incremental insulin AUC 2–10 minutes after the IVGTT dextrose infusion.

A standard 75-g OGTT was performed approximately 1 week after the IVGTT and clamp. Plasma glucose and insulin concentrations measured 0, 30, 60, 90, and 120 minutes after glucose ingestion were used to calculate the Matsuda index, an index of insulin sensitivity.³⁵ The insulinogenic index, a measure of the early insulin response to oral glucose, was calculated as the difference in insulin concentration divided by the difference in glucose concentration from fasting to 30 minutes.³⁶ Glucose AUC was calculated as a measure of glucose tolerance.³⁷

Covariates

Demographics and medical history were reported by participants. Prevalent cardiovascular disease was defined as a physician diagnosis of myocardial infarction, stroke, resuscitated cardiac arrest, or heart failure or a history of coronary or cerebral revascularization. Medications were ascertained by the inventory method. The Human Activity Profile maximum activity score was used to quantify physical activity.³⁸ The adjusted activity score was highly correlated with the maximum activity score (r=0.91), and similar results were observed when analyses used the adjusted activity score. Food intake was recorded using 3 days of prospective food diaries analyzed with Nutrition Data System for Research software. Body composition was measured by DXA (GE Lunar or Prodigy and iDXA; EnCore Software versions 12.3 and 14.1; GE Healthcare, Waukesha, WI).

Statistical Analyses

All analyses were performed using R 3.1.2 (R Foundation for Statistical Computing, Vienna, Austria; http://www.R-project.org/). Boxplots and scatterplots displayed distributions of dependent variables by eGFR. Differences by CKD status were tested using the t test (for continuous variables; assuming unequal variance) or the chi-squared test (for categorical variables). Univariate relationships of continuous variables were evaluated using Pearson correlation. Multivariable relationships were evaluated using linear regression with Huber–White SEMs.³⁹ Approximately 5% of participants were missing data on diet or body composition. For regression analyses, these values were multiply imputed using chained equations, and imputations were combined using the rules of Rubin.⁴⁰

Disclosures

I.H.d.B. received research support from Abbvie and consulted for Amgen, Inc. (Thousand Oaks, CA), Bayer HealthCare (Whippany, NJ), Boehringer Ingelheim (Mannheim, Germany), and Janssen Biotech (Horsham, PA).

S.E.K. received research support from Eli Lilly and consulted for Astra Zeneca, Boehringer Ingelheim, GlaxoSmithKline, Intarica Therapuetics, Janssen, Merck, Novo Nordisk, and Receptos. B.K. received an honorarium from Keryx Biopharmaceuticals.