Glycated Albumin Levels in Patients with Type 2 Diabetes Increase Relative to HbA1c with Time

We recently reported that glycated albumin (GA) is increased in subjects with longer duration of diabetes and with decreased insulin secretory function. Based on this, we investigated whether GA increases with time relative to glycated hemoglobin (HbA1c) and the association between GA and beta-cell function. We analyzed 340 type 2 diabetes patients whose serum GA and HbA1c levels had been repeatedly measured over 4 years. We assessed the pattern of changes with time in glycemic indices (GA, HbA1c, and GA/HbA1c ratio) and their relationship with beta-cell function. In all patients, glycemic indices decreased and maintained low levels around 15 and 27 months. However, from 39 months to 51 months, GA significantly increased but HbA1c tended to increase without statistical significance. We defined ΔGA/HbA1c as the difference between the nadir point (at 15 to 27 months) and the end point (at 39 to 51 months) and found that ΔGA/HbA1c was positively correlated with diabetes duration and negatively related to beta-cell function. In multivariable linear regression analyses, ΔGA/HbA1c was independently associated with diabetes duration. In conclusion, this study demonstrated that serum GA levels increase relative to HbA1c levels with time.


Introduction
Glucose monitoring is essential for the appropriate care and treatment of patients with diabetes in order to avoid diabetic complications and hypoglycemia. An accurate measure of glucose level allows physicians and patients to make optimal decisions about food, physical activity, and medications [1]. Of the glycemic indices, the American Diabetes Association recommends glycated hemoglobin (HbA 1c ) testing in all diabetic patients as an initial assessment and then as a part of continuing care [2]. This recommendation is derived from clinical data that shows that HbA 1c reflects average glycemic status over 2-3 months and predicts diabetic complications [3,4]. Although HbA 1c provides useful information, it might be inadequate in clinical situations such as anemia, renal insufficiency, and gestational diabetes. Glycated albumin (GA) has been gaining popularity as an indicator in several physiologic and pathologic conditions [5] because it provides more information than the gold standard HbA 1c . In line with this trend, we have demonstrated the clinical relevance of GA in type 2 diabetes mellitus (T2D) with insulin secretory dysfunction rather than insulin resistance [6], fluctuating or poorly controlled glycemic excursions [7], and progressing atherosclerosis [8].
In the natural course of T2D, however, beta-cell function decreases as duration of diabetes increases [9]. Moreover, glycemic excursions worsen due to decreased beta-cell function [10]. In a recent cross-sectional study, we reported that the levels of GA/HbA 1c were significantly elevated in subjects with long diabetic duration, largely attributed to the inverse relationships between GA and pancreatic beta-cell secretory indices [11], and suggested that clinicians should be careful in interpreting GA as only an indicator of glycemic control in T2D cases of longer duration. However, no longitudinal studies investigating the change in GA and HbA 1c over time in patients with T2D have been published.
In this longitudinal observational study, we investigated the changing pattern of glycemic indices such as GA, HbA 1c , and GA/HbA 1c over 4 years in order to determine whether GA increases more with time relative to HbA 1c in subjects with T2D. We also investigated which clinical and biochemical parameters are associated with changes in the GA/HbA 1c ratio.

Subjects and Data Collection.
In this longitudinal observational study, we recruited patients with T2D who had enrolled in previous studies [6,7] between May 2009 and June 2011 and who were followed up in June 2014. Using electronic medical records, we reviewed and rechecked demographic and clinical data for age, gender, metabolic parameters, and duration of diabetes. The diabetic duration was defined from the date the patients were first diagnosed with diabetes by blood tests or by patient recall from interviews.
To investigate the changes in glycemic indices with time, we tried to include patients whose duration of diabetes was less than 5 years. Patients were included if they were (1) aged ≥20 years, (2) had repeated laboratory data for both HbA 1c and GA up to the final follow-up point, and (3) had undergone a baseline standardized liquid meal test (Ensure, Meiji Dairies Corporation, Tokyo, Japan; 500 kcal, 17.5 g fat (31.5%), 68.5 g carbohydrate (54.5%), and 17.5 g protein (14.0%)) after an overnight fast. Patients were excluded if they had any medical conditions that could alter HbA 1c or GA levels such as liver cirrhosis or chronic kidney diseases (estimated glomerular filtration rate (GFR) by chronic kidney disease epidemiology collaboration formula <60 mL/min/1.73 m 2 ), pregnancy, or hematologic disorders or if they were being treated with steroids.
The protocol of this study was approved by the Institutional Review Board at Severance Hospital (IRB numbers 4-2009-0656, 4-2012-0398, and 4-2014-0507). Written informed consent for this study was not required by the Institutional Review Board because researchers only accessed the database for analysis purposes, and personal information was not used.

Laboratory
Measurements. The baseline glycemic indices (GA, HbA 1c , and GA/HbA 1c ) were defined as the values measured at enrollment. Subsequently, serum GA and HbA 1c were measured every 3 or 6 months. The end point glycemic indices of each subject were measured between 39 and 51 months. For glucose and C-peptide analyses, blood samples were collected at 0 and 90 min (basal and stimulated values) as part of the standardized liquid meal test. Pancreatic betacell functions in the context of ambient insulin secretory function were assessed using the following indices: (1) PCGR (stimulated C-peptide level/stimulated glucose level × 100), (2) C-peptide increment (ΔC-peptide = stimulated Cpeptide − basal C-peptide), and (3) C-peptide-genic index [CGI = (stimulated C-peptide − basal C-peptide)/(stimulated glucose − basal glucose)]. Measurement techniques included the hexokinase method for glucose and high-performance liquid chromatography using Variant II Turbo (Bio-Rad Laboratories, Hercules, CA) for HbA 1c . Serum GA was analyzed by an enzymatic method using an albumin-specific proteinase, ketoamine oxidase, an albumin assay reagent (LUCICA GA-L; Asahi Kasei Pharma Co., Tokyo, Japan), and a Hitachi 7699 P module autoanalyzer (Hitachi Instruments Service, Tokyo, Japan). GA values were calculated from the ratio of GA to total serum albumin and expressed as a percentage. Serum C-peptide levels were measured in duplicate using an immunoradiometric assay method (Beckman Coulter, Fullerton, CA).

Statistical Analysis.
All continuous variables were presented as mean ± standard deviation (SD) or median (quartiles) or as mean ± standard error (SE) for variables on the graphs. Categorical variables were described as N (%). Differences were analyzed using Student's t-test for the continuous variables and the chi-square test for categorical variables.
Repeated measured analysis of variance (ANOVA) with Bonferroni correction and paired t-test were used to determine the significance of differences in glycemic indices according to duration of diabetes. We compared GA, HbA 1c , and GA/HbA 1c levels at baseline and 3, 15, 27, and 39 to 51 months after enrollment in all patients who had glycemic values available at that time point. Because all glycemic indices reached their lowest level between 15 and 27 months (arbitrarily defined nadir point), we defined ΔGA/HbA 1c as the difference in GA/HbA 1c between the end point (39 to 51 months) and nadir point (15 to 27 months). One-way ANOVA with Tukey correction was used to compare the differences of duration of diabetes and PCGR according to the tertiles of ΔGA/HbA 1c ratio. Multivariable linear regression analysis was performed to determine the independent relationship of the studied variables including duration of diabetes associated with ΔGA/HbA 1c increase. Statistical analyses were performed using PASW Statistics version 18.0 for Windows (SPSS Inc., Chicago, IL, USA).

Associations between ΔGA/HbA 1c and Clinical and Biochemical Parameters.
Since the GA/HbA 1c ratio significantly increased from the nadir point to the final follow-up point, which was designated as ΔGA/HbA 1c (Figure 1(e)), we tried to determine the clinical and biochemical parameters that are associated with ΔGA/HbA 1c (Table 2 = 0.007) was more strongly associated with ΔGA/HbA 1c than ΔC-peptide (STD = −0.139, = 0.011).

ΔGA/HbA 1c Was Independently Associated with Duration of Diabetes.
Multivariable linear regression models were applied to determine the clinical and laboratory variables associated with ΔGA/HbA 1c (Table 3). We focused on certain parameters that can directly or indirectly reflect the insulin secretory function, such as PCGR, duration of diabetes, and medication history of DPP-IV inhibitor which can effectively reduce postprandial glucose. After adjustment for clinically important variables such as age, sex, BMI, waist circumference, and estimated GFR in model 1, history of DPP-IV inhibitor use was negatively associated with ΔGA/HbA 1c (STD = −0.111, = 0.049). After additional inclusion of PCGR in model 2, PCGR showed significant correlation with ΔGA/HbA 1c (STD = −0.161, = 0.009), but history of DPP-IV inhibitor use lost its significance. In model 3, duration of diabetes was further adjusted and the significant correlation of PCGR with ΔGA/HbA 1c disappeared (STD = −0.111, = 0.080). However, duration of diabetes was still independently associated with ΔGA/HbA 1c (STD = 0.172, = 0.005). Moreover, this association remained significant even after adjustment for glycemic status of subjects (inclusion of mean GA in model 4 and mean HbA 1c in model 5, resp.).
Additionally, we conducted multiple linear regression analyses to determine variables associated with PCGR at baseline (Supplementary Table 1

Discussion
Evidence has accumulated on the clinical relevance of GA as a glycemic index. However, the optimal use of GA as a glucose monitoring tool has not been fully investigated. Based on a previous cross-sectional study that showed that GA values are significantly influenced by the duration of T2D in cases where beta-cell function gradually decreases with time, we hypothesized that the ratio of GA to HbA 1c might not be constant over time. In this study of more than 4 years, we assessed glycemic excursion by measuring HbA 1c and GA and investigated discrepancy between two glycemic indices according to multiple time points. This study has three main findings: first, we found an initial sharp decrease in these glycemic indices, followed by maintenance at a low level, and then a gradual increase. Unlike for GA, the HbA 1c increase was statistically insignificant. Second, the change in GA/HbA 1c ratios, defined as the difference between the nadir point and the end point, was independently associated with baseline duration of diabetes. Third, impaired betacell function accounted for the association between longer duration of diabetes and increase in GA relative to HbA 1c , as well as the increase in the GA/HbA 1c ratio.
Because HbA 1c is formed via a nonenzymatic glycation process of hemoglobin in erythrocytes [12], medical conditions such as pregnancy, hemolytic anemia, chronic kidney disease, or end stage renal disease with dialysis could alter HbA 1c levels. In those cases, GA may be a more reliable marker than HbA 1c [5]. In contrast to HbA 1c formation, which requires intracellular glucose and protein metabolism, GA is formed directly via an extracellular nonenzymatic glycation process in plasma. However, medical conditions associated with albumin metabolism such as obesity, hyperthyroidism, and nephrotic syndrome, as well as glucocorticoid treatment [5], are known to affect GA levels. To avoid complications, we did not include patients with liver cirrhosis, chronic kidney diseases, pregnancy, and hematologic disorders or those who were being treated with steroid therapy.
With respect to the clinical relevance of the GA/HbA 1c ratio, it is known that the ratio is significantly correlated with insulin secretory beta-cell function but not with insulin resistance [6]. Recent study also showed that lower insulin secretory capacity predicted increased levels of GA/HbA 1c ratio in subjects with T2D [13]. Moreover, the GA/HbA 1c ratio in patients with T1D and T2D more accurately reflected glucose excursion [7,[14][15][16] and diabetic vasculopathy [8,17] than HbA 1c alone. The GA/HbA 1c ratio was significantly 7 higher in T2D patients treated with insulin than in those treated with either diet or oral hypoglycemic agents [7,18]. This observation might explain why history of insulin use is associated with either significant hyperglycemia or decreased beta-cell function. Our study also showed that ΔGA/HbA 1c between end point and nadir point is significantly associated with decreased insulin secretory function-related clinical and laboratory variables such as baseline and mean GA, mean HbA 1c PCGR, ΔC-peptide, and diabetic duration ( Table 2). Of the assessed glycemic indices, baseline HbA 1c did not predict the changes in the GA/HbA 1c ratio. With respect to the effect of insulin secretory factors on GA values, a recent cross-sectional study reported that GA levels significantly increased more in patients with longer duration of T2D and impaired beta-cell function measured by ΔCpeptide regardless of HbA 1c levels [11]. Consistent with this finding, our longitudinal study also showed that patients with higher levels of ΔGA/HbA 1c had longer duration of diabetes and lower levels of PCGR ( Figure 2). Furthermore, PCGR representing beta-cell function was associated with diabetic duration and insulin use at baseline and mean GA but not with mean HbA 1c . Based on these findings, we could infer that patients with T2D of longer duration and with higher GA/HbA 1c are more likely to have impaired beta-cell function and need insulin.
Our study had several strengths. First, this study is a longitudinal study with a long follow-up period of more than 4 years, which allowed us to investigate the changes in GA and HbA 1c levels over time. Second, about 80% of participants had a relatively short duration of diabetes (≤5 years) at enrollment. Lastly, we conducted mixed meal tests to obtain basal and stimulated C-peptide levels, which were then used to calculate PCGR as a measure of beta-cell function. That allowed for standardization of the stimulation calories and glucose content. Because it can be easily calculated and is a reliable indicator of beta-cell function, the PCGR is being used more frequently to help determine the optimal antidiabetic drug treatment [19,20]. In our study, PCGR levels were strongly associated with ΔC-peptide ( = 0.808, < 0.001) which strongly predicted beta-cell function (Supplementary Figure 1). In multivariable linear regression analyses, PCGR was also associated with ΔGA/HbA 1c . However, because the duration of diabetes strongly affects ΔGA/HbA 1c , after adjusting for duration of diabetes, the association between ΔGA/HbA 1c and PCGR disappeared (Table 3).
This study has the following limitations. First, we did not measure beta-cell function or glucose levels during followup period or at the end point. Thus, we did not prove that the difference between GA and HbA 1c is caused by a decline in beta-cell function during the follow-up period. Second, since this is a retrospective study, the follow-up period varied among the participants. Third, because we did not assess changes in medication, we could not adjust for its effects.

Conclusions
We conclude that both impaired beta-cell function and longer duration of diabetes are associated with an increase in GA relative to HbA 1c and an increase in the GA/HbA 1c ratio. The GA/HbA 1c ratio was significantly correlated with insulin secretory beta-cell function and increased as duration of diabetes increased. In this regard, clinicians should be extra careful when interpreting GA and GA/HbA 1c ratio values in subjects with longer duration of diabetes. Further welldesigned prospective studies enrolling larger populations are warranted.