Re-evaluation of glycated hemoglobin and glycated albumin with continuous glucose monitoring system as markers of glycemia in patients with liver cirrhosis.

Liver cirrhosis (LC) is frequently accompanied by glucose intolerance. The present study was designed to determine whether glycated hemoglobin A1c (HbA1c) and glycated albumin (GA) were predictive markers of glycemia, as determined by a continuous glucose monitoring system (CGMS), in patients with LC. A total of 30 patients with LC, including 3, 19, 5, 2 and 1 with LC due to hepatitis B virus, hepatitis C virus, non-alcoholic steatohepatitis, alcohol and unknown causes, respectively, were assessed by CGMS. The average, maximum and minimum blood glucose (BG) levels were measured by CGMS, and correlated with HbA1c and GA. The average, maximum and minimum BG in these individuals were 142±38.7, 209.3±65.7 and 85.1±25.4 mg/dl, respectively. HbA1c was significantly correlated with average BG (r=0.447, P=0.015) and maximum BG (r=0.523, P=0.004). In addition, GA was significantly correlated with average BG (r=0.687, P<0.001) and maximum BG (r=0.648, P<0.001). Neither HbA1c nor GA was significantly correlated with minimum BG. Correlation analysis yielded formulas by which HbA1c and GA were predictive of average BG in individuals with LC: Average BG=19.2 × HbA1c (%) + 36.5 and average BG=6.6 × GA (%) + 13.0, respectively. In conclusion, HbA1c and GA showed significant correlations with average and maximum BG, as determined by CGMS. The derived formulas allow for estimates of average BG based on HbA1c and GA, and may contribute to the control of glycemia in patients with LC.

Introduction

The liver performs numerous functions associated with glucose metabolism. Glucose uptake by the liver depends on circulating blood glucose (BG) concentration and contributes to the maintenance of glucose homeostasis. Glucose uptake by the liver is decreased in patients with liver cirrhosis (LC) because of a portal-systemic shunt and a decrease in viable hepatocytes, resulting in post-prandial hyperglycemia (1). Indeed, 80% of patients with LC also exhibit abnormal glucose tolerance and 25% have been diagnosed with diabetes (2). Glycated hemoglobin A1c (HbA1c), which reflects average plasma glucose concentrations over the preceding 1–2 months, is generally used as a marker of recent control of plasma glucose (3). HbA1c concentration is also regarded as a treatment marker in patients with diabetes. Indeed, guidelines formulated by an international expert committee composed of members of the European Association for the Study of Diabetes and the International Diabetes Federation and the American Diabetes Association have set a target HbA1c as 7%, as higher levels are associated with increased risks of cardiovascular disease and diabetic nephropathy, neuropathy and retinopathy (4).

Glycated albumin (GA) is another indicator of glucose metabolism. Due to fact that the half-life of albumin (ALB) (17 days) is shorter compared with that of hemoglobin (30 days), GA is a better marker of short-term BG levels (5). As glycosylation of ALB takes around 9 min, and is faster than that of HbA1c, GA is regarded as a more suitable marker of average glucose level in patients with greater fluctuations of glucose, including patients with acute and transient increases in postprandial BG level and night time hypoglycemia (6).

HbA1c or GA measurements have limitations in particular diseases, including chronic liver diseases and LC. Hypersplenism in patients with LC results in a shorter half-life of erythrocytes, resulting in the underestimation of HbA1c (7). In addition, since ALB concentrations are lower in patients with LC, a compensatory mechanism can extend the half-life of ALB in these patients, resulting in an overestimation of GA (7,8). Therefore, HbA1c and GA have been regarded as inadequate indicators of average BG concentrations in patients with LC (7). However, these results derived from studies in which patients performed 7–8 self-monitoring blood glucose (SMBG) tests per day, with average glucose levels determined from individual, discontinuous glucose concentrations. Therefore, it remains unclear whether HbA1c and GA are inappropriate indicators of average glucose levels in patients with LC.

Continuous glucose monitoring systems (CGMS) continuously measure glucose concentrations from glucose-oxidase reactions in the interstitial space and sensors placed in subcutaneous tissue. Glucose concentrations in the interstitial space are converted to BG levels based on four daily calibrations with SMBG. Sensors in CGMS measure glucose concentration every 10 sec and record average values every 5 min, resulting in more accurate average BG levels over 24 h (9). Significant positive correlations between HbA1c and average glucose levels, as determined by CGMS, have been observed in patients with diabetes (10). To date, however, correlations between HbA1c, GA and CGMS-determined average glucose levels remain to be evaluated in patients with LC.

The present study evaluated whether HbA1c and GA correlated with CGMS-determined average plasma glucose level, and assessed whether HbA1c and GA can be predictors of glucose metabolism in patients with LC.

Materials and methods

Patients

Patients diagnosed with LC at Saga Medical School Hospital (Saga, Japan) between 2011 and 2013 were included in the present study. The diagnosis of LC was based on liver biopsy findings and/or platelet counts <5×104/µl. Each patient underwent a general medical check-up, including physical and physiological examinations, and a screening blood test. Patients with decompensated LC, defined as a Child-Pugh score ≥10; patients with liver neoplasm; and patients using any agents for the treatment of diabetes were excluded. Patients with severe anemia (hemoglobin, <8 g/dl) and with proteinuria (positive urine protein in qualitative test) were also excluded. Finally, a total of 30 patients (16 males and 14 females) with LC were enrolled.

All subjects provided written informed consent for the use of their data. The study design was approved by the Institutional Review Board of Saga University Hospital (no. 2009-09-09). The study was performed in conformity with the ethical guidelines of the 7th revision of the Declaration of Helsinki (October, 2008).

Physical examination, serum biochemistry and liver histology

Body mass index was calculated as the weight (kg) divided by height (m2). Venous blood samples were obtained from all patients following a 12 h overnight fast, and blood cell counts, prothrombin time (PT), and concentrations of ALB, total-bilirubin (T-BIL), aspartate aminotransferase, γ-glutamyl trans-peptidase (GGT), hemoglobin A1c (HbA1c) and GA were measured using standard techniques. One day after admission, HbA1c was determined by high-performance liquid chromatography (Arcray Inc., Kyoto, Japan) and serum GA was measured enzymatically using an ALB-specific protease, ketoamine oxidase, and an ALB assay reagent (Lucica GA-L; Asahi Kasei Pharma, Tokyo, Japan).

Liver biopsy specimens were fixed in 10% formalin, embedded in paraffin, sectioned and stained with hematoxylin and eosin, and Azan for histological evaluation. A single experienced pathologist who was unaware of the clinical conditions of the patients evaluated all liver biopsy specimens. Liver histology of patients infected with hepatitis C virus (HCV) and hepatitis B virus (HBV) were evaluated according to the METAVIR scoring system (11), with LC diagnosed as METAVIR stage 4 (F4). Liver fibrosis in patients with non-alcoholic steatohepatitis (NASH) and alcoholic steatohepatitis was evaluated using Brunt's classification, with LC diagnosed as Brunt's stage 4 (12).

CGMS

Patients were equipped with a CGMS device (Medtronic miniMed, Northridge, CA, USA) and monitored for 72 h. Each CGMS device was calibrated with SMBG four times per day. After the 72 h monitoring period, all recorded data were downloaded onto a personal computer. Glucose profiles and glucose excursion parameters were evaluated with MiniMedSolutions software version 3.0 (MiniMed, Symar, CA, USA). Parameters analyzed included average, maximum and minimum BG concentrations, and the standard deviation of glucose concentration.

Predictive average BG with HbA1c and GA

Predictive average BG was calculated from HbA1c using the conversion formulas for patients with type 2 diabetes (10) and the conversion formula between HbA1c and GA (13): Average BG (mg/dl)=28.7 × HbA1c (%) - 46.7 and average BG (mg/dl)=6.2 × GA (%) + 38.8. After converting HbA1c to GA using the conversion formula, the correlation between GA and average BG was confirmed.

Statistical analysis

Continuous variables were reported as the mean ± standard deviation and categorical variables as frequencies. Simple correlation analyses were performed using the Spearman correlation coefficient. The average BG obtained by CGMS and estimated average BG were compared using Student's t-test. All analyses were performed using EZR (Saitama Medical Center, Jichi Medical University, Saitama, Japan), a graphical user interface for R (The R Foundation for Statistical Computing, Vienna, Austria) (14). P<0.05 was considered to indicate a statistically significant difference.

Results

Patient characteristics

The demographic and clinical characteristics of the 30 patients and their glycemic parameters obtained with CGMS are summarized in Table I. Since all patients had LC, their mean PLT counts (9.93±4.98 × 104/µl), PT (79.3±14.8%) and Alb concentration (3.24±0.54 mg/dl) were abnormally low. Their mean HbA1c and GA was 5.54±1.12% and 19.6±4.98%, respectively. HbA1c was >6.5% in 5 patients and GA was >20 mg/dl in 11 patients. CGMS was successfully performed in all patients, and the average, maximum and minimum BGs were obtained for 72 h. The average BG was >126 mg/dl in 19 patients and the maximum BG was >200 mg/dl in 13 patients. The mean minimum BG was 85.1±25.4 mg/dl, with 9 patients having a minimum BG <70 mg/dl and were considered hypoglycemic. Child-Pugh scores ranged between 5 and 8. LC was most frequently caused by HCV infection, observed in 19 patients.

Table I.

Baseline demographic and clinical characteristics of the 30 patients.

Characteristic	Demographics
Males, n (%)	16 (53.3)
Age, years	70.1±10.7
BMI	24.4±4.42
WBC (/µl)	4,346.7±1,515.8
Hb (g/dl)	12.4±2.02
PLT (x104 /µl)	9.93±4.98
PT (%)	79.3±14.8
ALB (g/dl)	3.24±0.54
T-BIL (mg/dl)	1.14±0.5
AST (U/l)	58.8±36.0
ALT (U/l)	32.6±38.7
GGT (U/l)	53.6±39.6
FPG (mg/dl)	107.9±25.1
Insulin (units)	17.3±22.2
HbA1c (%)	5.54±1.12
GA (%)	19.6±4.98
1,5-AG (µg/ml)	18.8±9.8
Glycemic parameters obtained from CGMS
Average BG (mg/dl)	142±38.7
Maximum BG (mg/dl)	209.3±65.7
Minimum BG (mg/dl)	85.1±25.4
Child-Pugh score, n (%)
5	10 (33.3)
6	8 (26.7)
7	7 (23.3)
8	5 (16.7)
Etiology, n (%)
HBV	3 (10)
HCV	19 (63.3)
NASH	5 (16.7)
Alcohol	2 (6.7)
Unknown	1 (3.3)

BMI, body mass index; WBC, white blood cell; Hb, hemoglobin; PLT, platelets; PT, prothrombin time; ALB, albumin; T-BIL, total bilirubin; AST, asparatate aminotransferase; ALT, alanine aminotransferase; GGT, γ-guanosine triphosphate cyclohydrolase; GA, glycoalbumin; 1,5-AG, 1,5-anhydroglucitol; FPG, fasting plasma glucose; HBV, hepatitis B virus; HCV, hepatitis C virus; NASH, non-alcoholic steatohepatitis; BG, blood glucose; CGMS, continuous glucose monitoring system.

Diagnostic ability of HbA1c and fasting plasma glucose (FPG) for hyperglycemia

HbA1c level and FPG level are commonly used for a diagnosis of diabetes. In order to investigate the diagnostic ability of HbA1c and FPG in the patients with LC, the present study analyzed the frequency of the patients who potentially fulfilled the diagnostic criteria of diabetes (HbA1c ≥6.5% and/or FPG ≥126 mg/dl), according to the average BG measured by CGMS (Table II). As expected, only 9.1% of the patients with average BG ≥140 mg/dl, 11.1% of the patients with average BG ≥150 mg/dl and 0% of the patients with average BG ≥ 200 mg/dl met the diagnostic criteria of diabetes (HbA1c ≥6.5% and FPG ≥126 mg/dl).

Table II.

HbA1c and fasting plasma glucose level in the liver cirrhosis patients with hyperglycemia determined by CGMS.

	Average blood glucose (CGMS)
Characteristic	≥140 mg/dl (n=11, %)	≥150 mg/dl (n=9, %)	≥200 mg/d (n=3, %)
HbA1c ≥6.5	5 (45.5)	5 (55.6)	2 (66.7)
FPG ≥126	3 (27.2)	3 (33.3)	1 (33.3)
HbA1c ≥6.5 and FPG ≥126	1 (9.1)	1 (11.1)	0 (0)
HbA1c ≥6.5 or FPG ≥126	7 (63.6)	7 (77.8)	3 (100)

CGMS, continuous glucose monitoring system; Hb, hemoglobin; FPG, fasting plasma glucose.

Correlation between glycemic parameters and CGMS parameters

The present study evaluated the correlations of HbA1c, GA and 1,5-anhydroglucitol (1,5-AG) concentrations with CGMS parameters (Table III). HbA1c concentration was significantly correlated with the average BG (r=0.45, P=0.015), maximum BG (r=0.52, P=0.004) and the standard deviation of BG (r=0.49, P=0.008). GA was also significantly correlated with average BG (r=0.69, P<0.01), maximum BG (r=0.65, P<0.01) and the standard deviation of BG (r=0.73, P<0.01). FPGs also exhibited significant correlations with average BG (r=0.55, P=0.002), maximum BG (r=0.51, P=0.004) and the standard deviation of BG (r=0.43, P=0.018). No glycemic marker exhibited a significant correlation with minimum BG.

Table III.

Correlation between glycemic parameters and continuous glucose monitoring system in patients with LC.

	Average BG		Maximum BG		Minimum BG		Standard deviation
Characteristic	r-value	P-value	r-value	P-value	r-value	P-value	r-value	P-value
HbA1c	0.447	0.015	0.523	0.004	0.143	0.460	0.485	0.008
GA	0.687	<0.001	0.648	<0.001	0.071	0.725	0.732	<0.001
1,5-AG	−0.403	0.041	−0.588	0.002	−0.101	0.624	−0.534	0.005
FPG	0.545	0.002	0.511	0.004	−0.014	0.941	0.428	0.018

BG, blood glucose; GA, glycoalbumin; 1,5-AG, 1,5-anhydroglucitol; FPG, fasting plasma glucose.

Based on the correlations of the distributions of HbA1c and GA relative to average BG, the average BG could be predicted from HbA1c and GA concentrations: Average BG=19.2 × HbA1c (%)+36.5 (Fig. 1A) and average BG=6.6 × GA (%)+13.0 (Fig. 1B). Two outliers were identified in the correlation distribution (patients 16 and 18) for HbA1c and average BG, which deviated from the predicted formula and lay outside the 95% confidential interval. As expected, both patients exhibited anemia; hemoglobin concentrations were 9.9 g/dl and 9.7 g/dl in patients 16 and 18, respectively. In the correlation distribution of GA and average BG, however, these two patients were not outliers and lay within the 95% confidential interval (Fig. 1B).

Figure 1.

Correlations between average BG and (A) HbA1c (r=0.447, P=0.015) and (B) GA (r=0.687, P<0.001) concentrations, and between maximum BG and (C) HbA1c (r=0.523, P=0.004) and (D) GA (r=0.648, P<0.001) concentrations. Open dots indicate each individual, black lines show the correlation regression line and gray lines indicate 95% confidential intervals. The two black circles in (A) indicate outliers with severe anemia. Hb concentrations were 9.9 g/dl in patient 16 and 9.7 g/dl in p-atient 18. BG, blood glucose; HbA1c, glycated hemoglobin; GA, glycated albumin; Hb, hemoglobin.

The correlations of HbA1c and GA with maximum BG are shown in Fig. 1C and D, respectively. Both HbA1c (r=0.523, P=0.004) and GA (r=0.648, P=0.001) concentrations revealed significant positive correlations with maximum BG. One outlier in the correlation between HbA1c and average BG (patient 18) was also an outlier in the correlation between HbA1c and maximum BG (Fig. 1C), however, not in the correlation between GA and maximum BG (Fig. 1D).

Comparison between average BG obtained from CGMS and prediction formula

The present study also compared the average BG calculated from HbA1c and GA using the formulas derived from patients with type 2 diabetes (10,13) and the average BG obtained from CGMS. The regression line between HbA1c and average BG, obtained from CGMS, was above the regression line between HbA1c and average BG calculated using the formula derived from patients with type 2 diabetes (Fig. 2A). By contrast, the regression line between GA and average BG obtained from CGMS was below the regression line obtained between GA and average BG calculated using the above formula (Fig. 2B). These findings suggested that HbA1c significantly underestimated (112.3±32.2 mg/dl) and GA significantly overestimated (160.3±30.9 mg/dl) average BG in patients with LC relative to the BG determined with CGMS (142.0±38.7 mg/dl). Differences between the average BG obtained with CGMS and the average BG calculated from HbA1c and GA are shown in Table IV. It was revealed that the modified formulas were better able to predict average BG from HbA1c and GA in patients with LC.

Figure 2.

(A) Regression analysis of HbA1c relative to the average BG, as measured by CGMS (black line) and according to the formula derived from patients with type 2 diabetes (10): Average BG=28.7 × HbA1c-46.7 (dotted line). Open dots represent each individual. (B) Regression analysis of GA relative to BG, as measured by CGMS (black line) and according to the formula derived from patients with type 2 diabetes (10,13): Average BG (mg/dl)=6.2 × GA (%)+38.8 (dotted line). Open dots represent each individual. HbA1c, glycated hemoglobin; BG, blood glucose; CGMS, continuous glucose monitoring system; GA, glycated albumin.

Table IV.

Comparison between average BG measured on CGMS and calculated from HbA1c and GA.

CGMS measured	Factor used in prediction	Prediction formula	Modified prediction formula
142±38.7	HbA1c	112.3±32.2[a]	142.8±21.6
	GA	160.3±30.9[a]	141.8±32.7

P<0.001 vs. average BG measured by CGMS. BG, blood glucose; CGMS, continuous glucose monitoring system; GA, glycoalbumin; Hb, hemoglobin.

Discussion

CGMS can monitor glucose metabolism continuously and more precisely than SMBG. Indeed, CGMS in the 30 patients included in the present study revealed that 19 patients had average BG >126 mg/dl and 13 exhibited a maximum BG >200 mg/dl. By contrast, only three patients had fasting glucose >126 mg/dl, only four had HbA1c >6.5% and 11 had GA >20%, indicating that CGMS is more sensitive compared with these other, fixed in detecting disorders of glucose metabolism in patients with LC. Using CGMS to evaluate glycemic parameters, the present study tested our hypothesis that HbA1c and GA may predict abnormal glucose tolerance in patients with LC. As expected, HbA1c and GA correlated significantly with average BG, as determined by CGMS, with GA showing a more significant correlation with average BG compared with other glycemic parameters, including HbA1c, 1,5-AG and FPG. In addition, anemia, which caused outliers in the correlation between HbA1c and average BG, had no effect on the correlation between GA and average BG.

Despite these significant correlations of GA and HbA1c with average BG measured on CGMS, there were differences between the latter and average BG calculated from formulas based on HbA1c and GA. Specifically, the formula based on HbA1c tended to underestimate and the formula based on GA tended to overestimate average BG relative to that determined by CGMS. These formulas, however, were derived from patients with type 2 diabetes and may be appropriate for patients with LC, in whom the half-life of erythrocytes is shorter and the half-life of ALB longer compared with that in non-LC patients (10). Since CGMS is impractical in all patients with LC, due to its invasiveness, costs and limited quality of life during monitoring, formulas are required to more accurately calculate average BG from HbA1c and GA concentrations. Based on the CGMS data, the present study determined more accurate formulas for calculating average BG from HbA1c [average BG=19.2 × HbA1c (%)+36.5] and GA [average BG=6.6 × GA (%)+13.0] concentrations.

Viable hepatocytes store glucose as glycogen and prevent rapid increase of postprandial BG level. A reduction in the number of viable hepatocytes, as in LC, reduces glycogen capacity in the liver, increasing circulating glucose concentrations and postprandial hyperglycemia. Gluconeogenesis in the liver is also reduced, resulting in hypoglycemia while fasting. Hypoglycemia with lack of glycogen causes the catabolism of fat and skeletal muscle and can lead to sarcopenia (15,16).

Since cirrhotic patients have the shunt from the portal vein to the systemic circulation, glucose and insulin in the portal vein flow into the systemic circulation, bypassing liver cells. This leads to hyperglycemia and hyperinsulinemia after meals. Hyperinsulinemia downregulates the expression of insulin receptors in peripheral tissues, inducing insulin resistance (17–22). Therefore, glucose homeostasis in patients with LC differs greatly from that in healthy individuals. Specific management of glucose tolerance is therefore required in individuals with LC.

Abnormal glucose tolerance has been reported to affect the pathogenesis and prognosis of LC. Diabetes increases the risk of hepatocellular failure and the mortality rate in patients with LC (23–25). Diabetes in LC also increases the risk of complications of LC, including hepatic enteropathy, spontaneous bacterial peritonitis and rupture of esophageal varices (26–29). Diabetes and glucose intolerance in individuals with LC must therefore be controlled more carefully. The findings presented in the present study may assist with improving the pathogenesis and prognosis of LC.

Previous reports have investigated glucose metabolism in chronic liver diseases with CGMS (30–32). Kawaguchi et al revealed that 50% of the patients with compensate LC exhibited nocturnal hypoglycemia concomitant with higher serum-free fatty acid level compared with the patients without nocturnal hypoglycemia (30). In the patients with biopsy-proven NASH, it was determined by CGMS that median glucose levels, standard deviation of glucose levels and maximum glucose levels were significantly higher in the patients with advanced liver fibrosis (31). CGMS enables us to analyze a specific and unusual glucose homeostasis of chronic liver disease, as well as diabetes and various disease states (32).

The limitations of the present study included the performance of CGMS in patients while hospitalized. Measurements of HbA1c and GA indicated glycemic control in these patients prior to hospitalization, indicating that patient lifestyle differed greatly prior to and following hospitalization, an important limitation when comparing glycemic markers with CGMS parameters.

Another limitation was the differences in etiology of LC in this patient cohort, particularly since HCV and NASH cause impaired glucose tolerance without liver fibrosis. Due to the limited number of patients in the present study, it was difficult to compare differences in glucose intolerance among the subgroups of patients with different etiologies of LC.

In conclusion, HbA1c and GA exhibited good correlation with the average BG and maximum BG, as evaluated by CGMS. The modified prediction formulas developed revealed improved accuracy compared with previous formulas in estimating average BG from HbA1c and GA in patients with LC. The use of these formulas may contribute to control of glycemia in patients with LC.