Glycated Albumin Correlates With Time-in-Range Better Than HbA_1c or Fructosamine

Abstract

Context

Intermediate-term glycemic control metrics may represent a viable alternative to continuous glucose monitoring (CGM) in patients without access to CGM.

Objective

This work aimed to compare the relationship between CGM parameters and glycated albumin (GA), glycated hemoglobin A_1c (HbA_1c), and fructosamine for 24 weeks.

Methods

We conducted exploratory comparative analyses of CGM subgroup data from a previously published 24-week prospective study of assay performance in 8 US clinics. Participants included 34 individuals with type 1 (n = 18) and type 2 diabetes (n = 16) undergoing changes to improve glycemic control (n = 22; group 1) or with stable diabetes therapy (n = 12; group 2). Main outcome measures included Pearson correlations between CGM and glycemic indices and receiver operating characteristic (ROC) analysis of glycemic index values predictive of time in range (TIR) greater than 70%.

Results

At weeks 4 and 8, GA correlations with TIR were higher than HbA_1c correlations in group 1. In group 2, GA correlations with TIR were statistically significant, whereas HbA_1c correlations were not. In both groups over the first 12 weeks, GA correlations with TIR were higher than fructosamine-TIR correlations. In the ROC analysis, GA predicted a TIR greater than 70% during weeks 2 to 24 (area under the curve >0.80); HbA_1c was predictive during weeks 12 to 24. Cutoff values for TIR greater than 70% were 17.5% (sensitivity and specificity, 0.88) for GA and 7.3% (0.86) for HbA_1c.

Conclusion

GA is the most accurate predictor of TIR over 8 weeks compared with other glycemic indices, which may assist in clinical evaluation of changes in treatment where CGM is not possible and it is too early to use HbA_1c (NCT02489773).

In conjunction with glycated hemoglobin A_1c (HbA_1c), continuous glucose monitoring (CGM) has become widely recognized as a standard-of-care tool for monitoring daily fluctuations in glycemia and informing therapeutic decision-making, especially in insulin-treated patients (1, 2). Yet even as increasing numbers of clinicians and patients rely on CGM, access to these costly devices is often limited, and some patients may not be comfortable using the technology (3-5). Traditional self-monitoring of blood glucose (SMBG) is still recommended for daily assessment of glycemic control, but finger sticks may be burdensome for patients and may not fully capture a complete picture of daily glycemic fluctuations, especially if patients test only once or twice a day (1). Intermediate-term assessments such as glycated albumin (GA) and fructosamine, which reflect glycemia over approximately 2 weeks, may bridge the gap between the 2- to 3-month glucose average measured with HbA_1c and the daily information provided by CGM or SMBG (6-8).

The Lucica Glycated Albumin–L test (Asahi Kasei Pharma) is an enzymatic assay for GA in which a ketoamine oxidase and peroxidase reaction is used to eliminate endogenous glycated amino acids and peroxide (9, 10). The remaining GA is hydrolyzed to amino acids or peptides by an albumin-specific proteinase and measured quantitatively. The GA value is presented as a ratio (mmol/mol) of glycated to total albumin concentration in the same serum sample. This approach minimizes effects of variations in albumin concentrations among individuals (10, 11).

Our group has previously published 2 studies of the Lucica GA-L compared with other glycemic indices (12, 13). This exploratory analysis builds on the prior reports by focusing on correlations between CGM and GA found in a study that originally enrolled 150 patients with either type 1 or type 2 diabetes (T1D and T2D, respectively) divided into different analysis groups based on whether their glycemic control regimen was stable or being intensified over a 6-month period. The parent study demonstrated that compared with fructosamine, GA correlates significantly better with both short-term mean blood glucose (MBG) and long-term HbA_1c (13). This post hoc analysis was conducted to determine how well GA performed as a predictor of CGM metrics, particularly time in range (TIR). Only a subset of individuals in the parent study used masked CGM, and CGM data were available for only a small subset of the original study population; we present those findings here.

Materials and Methods

Study Design

The parent study was a 24-week, prospective, multicenter, comparative study of assay performance in a clinical setting conducted at 8 sites in the United States (NCT02489773). Eligible adults aged 18 years or older with either T1D or T2D were prospectively divided into 2 groups based on therapeutic need. Group 1 included 98 patients (T1D, n = 47; T2D, n = 51) with an HbA_1c of 7.5% or greater who were prescribed a change in diabetes management. Group 2 included 52 patients (T1D, n = 26; T2D, n = 26) on a stable therapeutic regimen with an HbA_1c of less than 7.5% for whom no therapeutic changes were planned for the duration of the study. The institutional review boards at each study center approved the protocol and consent form. All patients provided written informed consent. The study design and results have been published previously (13).

Assessments

As previously described, fasting blood samples from each participant were drawn at prespecified intervals (weeks 0 [screening], 1 [study start], 2, 3, 4, 6, 8, 12, 16, 20, and 24) and tested for GA (Lucica GA-L), fructosamine uncorrected for albumin (Randox Fructosamine; Randox Crumlin), and HbA_1c (Tosoh G7/G8; Tosoh Corporation). Data from a traditional SMBG monitor (OneTouch Ultra 2 Blood Glucose Meter [LifeScan Inc]) were collected at each visit (13). In addition, a subset of participants used a masked CGM device (Dexcom G4 PLATINUM CGM System [Dexcom Inc]) beginning at enrollment (week 1) and continuing for the full 24-week study period. Participants placed the sensors themselves if they were comfortable doing so. Patients were blinded to all assessment results. Clinicians adjusted the regimens of patients in group 1 at their discretion based on HbA_1c and SMBG findings (not GA, fructosamine, or CGM).

End Points

Pearson correlations between time in, above, and below range (TIR, TAR, and TBR, respectively, measured with CGM) and other glycemic indices (GA, fructosamine, HbA_1c, and MBG determined from SMBG data) were evaluated over 24 weeks. The MBG estimation method was as described previously (13). The power of each glycemic index to detect a TIR greater than 70% was also determined using a receiver operating characteristic (ROC) analysis.

Statistical Analysis

Descriptive statistics were used to summarize all results in this study. Continuous variables were summarized by mean and SD. Categorical variables were summarized by number and percentage. To enable the comparison of changes from baseline in indices with different units of measure over time, differences between values at baseline and at each visit were converted to percentage changes from baseline.

The correlations between variables were assessed by Pearson correlation tests. The relationship between TIR and variables were studied using ROC curves and area under the curve (AUC) to assess the power of detecting TIR greater than 70%. The cutoff values were developed as the levels at which sensitivity was in agreement with specificity. All data were analyzed using StatFlex for Windows V7 (Artec).

Results

A total of 34 individuals from the original study were included in the CGM analysis; 22 participated in group 1 (T1D, n = 12; T2D, n = 10) and 12 in group 2 (T1D, n = 6; T2D, n = 6). The mean age of the entire study population was 48.4 years (Table 1). In group 1, the mean HbA_1c was 8.6% ± 0.9%, and in group 2 it was 6.8% ± 0.5%. A large proportion (35%) of the cohort were Black, and 54% were female. Other baseline characteristics were similar to those in the main study (see Table 1).

Table 1.

Participant demographics at baseline

	Group 1 (n = 22)	Group 2 (n = 12)	Overall (n = 34)
Mean age, y (SD)	47.3 (15.5)	50.3 (18.1)	48.4 (16.3)
Male, n (%)	9 (41)	6 (50)	15 (44)
Race and ethnicity, n (%)
White	13 (59)	9 (75)	22 (65)
Black	9 (41)	3 (25)	12 (35)
Hispanic ethnicity	1 (5)	3 (25)	4 (12)
Diabetes type, n (%)
Type 1	12 (55)	6 (50)	18 (53)
Type 2	10 (45)	6 (50)	16 (47)
Mean weight, kg (SD)	97.4 (23.6)	85.2 (22.7)	93.1 (23.7)
Mean BMI (SD)	33.4 (6.5)	30.8 (6.5)	32.5 (6.5)
Mean serum albumin, g/L (SD)	44 (3)	47 (2)c	45 (3)
Antihyperglycemic use
Insulin,a n (%)	16 (73)	6 (50)	22 (65)
Oral and/or noninsulin injectable agents, n (%)	9 (41)	6 (50)	15 (44)
Glycemic indices
Mean HbA1c, % (SD)	8.6 (0.9)	6.8 (0.5)d	7.9 (1.1)
mmol/mol (SD)	70 (10)	51 (5)d	63 (13)
Mean FBG, mmol/L (SD)	10.0 (3.2)	8.0 (1.0)c	9.3 (2.8)
mg/dL (SD)	180 (57)	144 (18)c	168 (50)
Mean MBG, mmol/L (SD)	9.7 (1.9)	7.7 (1.2)d	9.0 (1.9)
mg/dL (SD)	174 (35)	139 (21)d	162 (35)
Mean fructosamine, µmol/L (SD)	462 (79)	355 (54)d	424 (88)
Mean GA, mmol/mol (SD)	394 (65)	295 (37)d	359 (74)
Mean GA, % (SD)b	22.2 (3.4)	16.6 (2.1)d	20.2 (4.0)

Abbreviations: BMI, body mass index; FBG, fasting blood glucose; GA, glycated albumin; HbA_1c, glycated hemoglobin A_1c; MBG, mean blood glucose.

With or without other antihyperglycemic agents.

GA (%) = 0.05652 × GA (mmol/mol) – 0.4217 (14).

P < .05 vs group 1.

P < .01 vs group 1.

Rate of Change in Glycemic Indices Over 24 Weeks

Mean percentage changes in GA, fructosamine, MBG, and HbA_1c in each group and overall appear in Fig. 1. Similar to findings in the main trial, changes in GA over time tended to increase or decrease in parallel with MBG more than other indices. Mean TIR, TAR, and TBR from patients’ CGM devices in each group and overall are shown in Supplementary Fig. S1 (15).

Figure 1.

Mean rate of change in glycated albumin (GA), fructosamine (FRA), mean blood glucose (MBG), and glycated hemoglobin A_1c (HbA_1c) overall and in each group. Percentage changes are used so that all indices can be shown on the same scale.

Correlations Between Continuous Glucose Monitoring and Glycated Albumin, Fructosamine, and Glycated Hemoglobin A_1c

As shown in Table 2, at week 4 in the full study population, the correlation coefficients between TIR and GA, HbA_1c, fructosamine, and MBG were −0.753, −0.548, −0.609, and −0.6211, respectively, and all were statistically significant (P < .05). The coefficients for all comparisons between TIR and other indices at weeks 4, 8, and 12 were statistically significant (P < .05) overall and in group 1, but TIR-GA coefficients were higher than TIR-HbA_1c coefficients at weeks 4 and 8. The coefficients between TIR and fructosamine were lower than those between TIR and GA overall and in group 1.

Table 2.

Correlation coefficients between continuous glucose monitoring and glycated albumin, glycated hemoglobin A_1c, fructosamine, and mean blood glucose^a

	Overall (n = 34)				Group 1 (n = 22)				Group 2 (n = 12)
Wk	HbA1c	GA	FRA	MBG	HbA1c	GA	FRA	MBG	HbA1c	GA	FRA	MBG
Time in range
4	−0.548 b	−0.753 b	−0.609 b	−0.621 b	−0.523 c	−0.742 b	−0.573 b	−0.790 b	−0.243	−0.679 c	−0.682 c	−0.189
8	−0.549 b	−0.781 b	−0.651 b	−0.797 b	−0.562 c	−0.754 b	−0.647 b	−0.881 b	−0.080	−0.828 c	−0.787 b	−0.475
12	−0.748 b	−0.791 b	−0.555 b	−0.670 b	−0.754 b	−0.738 b	−0.549 c	−0.753 b	−0.486	−0.850 c	−0.575	−0.339
Time above range
4	0.634 b	0.799 b	0.596 b	0.789 b	0.572 b	0.789 b	0.609 b	0.866 b	0.468	0.672 b	0.588 c	0.521
8	0.599 b	0.786 b	0.632 b	0.914 b	0.555 c	0.780 b	0.662 b	0.932 b	0.416	0.715 c	0.650 c	0.836 b
12	0.795 b	0.781 b	0.511 b	0.806 b	0.761 b	0.757 b	0.562 b	0.795 b	0.728 c	0.692 c	0.295	0.772 b
Time below range
4	−0.268	−0.145	0.026	−0.500 b	−0.286	−0.339	−0.263	−0.441	−0.378	0.325	0.506	−0.606 c
8	−0.300	−0.225	−0.119	−0.563 b	−0.215	−0.418	−0.333	−0.575 c	−0.610 c	0.190	0.236	−0.658 c
12	−0.330	−0.173	−0.014	−0.537 b	−0.502 c	−0.548 c	−0.403	−0.640 b	−0.345	0.273	0.445	−0.641 c

Abbreviations: CGM, continuous glucose monitor; FRA, fructosamine; GA, glycated albumin; HbA_1c, glycated hemoglobin A_1c, MBG, mean blood glucose, determined with self-monitored blood glucose.

^a Boldface highlights statistically significant values.

P < .01.

P < .05.

In the group 2 comparisons, only the coefficients between TIR and GA and MBG were statistically significant at all time points. TIR-fructosamine correlations were significant at weeks 4 and 8 but not 12, and there were no significant correlations between TIR and HbA_1c at any time point in group 2. Correlations of TAR with GA, HbA_1c, fructosamine, and MBG followed a similar pattern, whereas most correlations between TBR and the other indices (except MBG) were not statistically significant.

Cutoff Values for Detection of Time in Range

An ROC analysis of AUC data showed that GA predicted a TIR greater than 70% from week 2 through week 24, whereas HbA_1c predicted a TIR greater than 70% from week 12 through week 24 (Table 3). The GA cutoff value for detecting a TIR greater than 70% during weeks 2 to 24 (AUC >0.80) was 17.5% (sensitivity and specificity, 0.88), whereas the corresponding HbA_1c value during weeks 12 to 24 was 7.3% (sensitivity and specificity, 0.86). In a separate ROC analysis, a GA of 17.2% (sensitivity and specificity, 0.75), a TIR of 71.6% (sensitivity and specificity, 0.79), and an MBG of 148.6 mg/dL (sensitivity and specificity, 0.83) were found to detect an HbA_1c greater than 7.0% (Supplementary Table S1) (15).

Table 3.

Results of receiver operating characteristic analysis for detecting time in range greater than 70%^a,b

	HbA1c, %			GA, %
Wk	AUC ± SE	Sensitivity/specificity	Cutoff	AUC ± SE	Sensitivity/specificity	Cutoff
1	0.68 ± 0.10	0.66	7.7	0.67 ± 0.11	0.50	19.3
2	0.69 ± 0.10	0.67	7.7	0.91 ± 0.06	0.83	18.2
3	0.70 ± 0.10	0.64	7.4	0.94 ± 0.06	0.91	17.0
4	0.79 ± 0.08	0.72	7.4	0.97 ± 0.03	0.92	17.4
6	0.72 ± 0.09	0.65	7.2	0.92 ± 0.05	0.88	17.6
8	0.76 ± 0.09	0.72	7.2	0.96 ± 0.03	0.89	16.9
12	0.83 ± 0.08	0.76	7.3	0.96 ± 0.03	0.92	17.4
16	0.87 ± 0.07	0.88	7.2	0.89 ± 0.06	0.82	17.6
20	0.93 ± 0.04	0.86	7.3	0.97 ± 0.02	0.90	17.4
24	0.95 ± 0.04	0.92	7.3	0.90 ± 0.06	0.81	17.8
2-24c		—	—		0.88	17.5
12-24c		0.86	7.3		—	—

Abbreviations: AUC, area under the curve; GA, glycated albumin; HbA_1c, glycated hemoglobin A_1c.

Cutoff values chosen based on equivalence of sensitivity and specificity.

Boldface highlights sensitivity/specificity greater than 0.80.

Average over 2 to 24 weeks (GA) or 12 to 24 week (HbA_1c) timespan.

Discussion

In these exploratory analyses, we expand on previous work confirming that GA is useful as a short-term measure to evaluate glycemic control (12, 13) by demonstrating a strong correlation with TIR. We confirmed our previous finding that changes in GA reflect short-term fluctuations in MBG better than HbA_1c, whereas fructosamine fluctuations did not reflect MBG as well as GA or HbA_1c in individuals undergoing treatment intensification to improve glycemic control (group 1) and in those with stable therapeutic regimens (group 2) (13). We also found that GA was more closely correlated with TIR than HbA_1c during the first 3 months of the study, regardless of whether therapy was changing or stable. In addition, GA performed better than HbA_1c in the detection of TIR greater than 70% between weeks 2 and 12, with a cutoff value for detection of TIR greater than 70% of 17.5% (sensitivity and specificity, 0.88).

GA correlations with TIR and TAR were statistically significant at all time points in both group 1 and group 2, as well as overall. Not surprisingly, HbA_1c was less well correlated (smaller magnitude coefficients and/or P values >.05) at weeks 4 and 8 across cohorts, although, as expected, at week 12 HbA_1c was well correlated both with TIR and TAR. In contrast, the correlations between fructosamine and TIR and TAR were of a lower magnitude than those between GA and TIR and TAR, and the CGM measures and were not statistically significant in group 2 at week 12. These data confirm our earlier finding that GA may be a more reliable early measure of glycemia than fructosamine, but it does not provide any insights as to why. Because GA is specific to albumin, it may be less likely than fructosamine to fluctuate in response to other serum proteins and low-molecular-weight substances (6, 16-18). Clarifying these differences requires further study.

In a study involving Japanese patients with T1D and T2D, GA was well correlated with TIR and TAR in the overall population and with TBR in patients with T1D (19). Another study from Japan likewise showed significant correlations between GA and TIR and TAR in patients with T1D and T2D (20). The Continuous Glucose Monitoring to Assess Glycemia in CKD (CANDY) study involved US patients with T2D and chronic kidney disease; in this population, changes in mean glucose measured with CGM were more strongly correlated with changes in GA than with changes in fructosamine or HbA_1c, although GA was less predictive of CGM mean glucose than HbA_1c or fructosamine. In addition to the population, this study differed from ours in that participants wore CGM for two 6-day periods separated by approximately 2 weeks rather than continuously for 24 weeks, and there were only 2 blood draws for glycemic measures, separated by approximately 3 weeks (21).

This is the first study to use an ROC approach to determine GA cutoffs predictive of TIR of greater than 70%. Our finding that GA was a more accurate predictor of TIR than HbA_1c within the first 12 weeks has important clinical implications. CGM is increasingly recommended as a complement to HbA_1c for assessing glycemic control, but not all patients have access to this costly technology, whereas others may not wish to wear or use it. GA provides clinicians with a simple alternative means of accurately evaluating the effects of treatment changes when it is too early to measure HbA_1c and CGM is not available. In addition, the ability to adjust therapeutic regimens earlier may lead to lower health care costs, including the costs of medications and/or of laboratory tests. Because GA is measured using serum, it can be measured simultaneously with other biochemical tests, whereas a separate blood draw is required for measurement of HbA_1c.

Since we completed the original study of GA, reference intervals for GA and fructosamine have been established based on data from the Atherosclerosis Risk in Communities (ARIC) study, showing that GAs of 16% and 17% are equivalent to HbA_1c values of 6.5% and 7%, respectively (22, 23). Our findings, where a GA 17.2% is a cutoff value for detecting an HbA_1c greater than 7.0%, are consistent with the ARIC study conclusions. In addition, we found that a GA of 17.5% also reflects a TIR of 70%. CGM is increasingly recommended as a complement to HbA_1c for treatment intervention and assessment of glycemic control (1, 2, 24). However, although use of the technology is expanding, it is still not accessible to many patients with diabetes, especially those with T2D (3-5). Our finding of the close correlations between GA and TIR and TAR at 1, 2, and 3 months after the study start verifies the utility of GA as an alternative assessment that can simply provide early and accurate information on glycemic control. Furthermore, the finding that a GA of 17.5% may be used to detect TIR greater than 70% could provide clinicians with useful insight into which patients might benefit from further investigation with a temporary (clinician-owned) CGM device or more intensive SMBG.

This post hoc analysis was limited to a small subset of patients from the pivotal trial, but the results presented here are consistent with those from the main study, where GA was most closely correlated with MBG measured with SMBG (13). When evaluated against CGM data, GA continued to be an accurate indicator of glucose control in people with either T1D or T2D, regardless of whether treatment regimens were being changed or staying the same. GA also was shown to accurately reflect TIR, representing an alternative complement to HbA_1c for patients without access to CGM.

Abbreviations

AUC

area under the curve

BMI

body mass index

CGM

continuous glucose monitoring

FBG

fasting blood glucose

GA

glycated albumin

HbA_1c

glycated hemoglobin A_1c

MBG

mean blood glucose

ROC

receiver operating characteristic

SMBG

self-monitoring of blood glucose

T1D

type 1 diabetes

T2D

type 2 diabetes

TAR

time above range

TBR

time below range

TIR

time in range

Funding

This work was supported and conducted by Asahi Kasei Pharma Corporation.

Author Contributions

C.V.D., T.K., and V.A.F. conceived the study and participated in analysis and interpretation of the data. C.V.D, J.R., and V.A.F. conducted the study, including acquisition, analysis, and interpretation of data. T.K. contributed to the statistical design, interpretation of data, and statistical analyses. C.V.D, J.R., T.K., and V.A.F. participated in the drafting and critical revision of the manuscript. All authors had full access to the data in the study and had final responsibility for the decision to publish. V.A.F. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Disclosures

C.V.D. has received research grants and consulting fees from Novo Nordisk and research grants from Eli Lilly, Asahi Kasei Pharma Corporation, KOWA, and the NIH. J.R. has served on scientific advisory boards and received honoraria or consulting fees from Applied Therapeutics, Boehringer Ingelheim, Eli Lilly, Hanmi, Novo Nordisk, Oramed, Sanofi, Structure Therapeutics, Terns Pharma, and Zealand; has received grants/research support from Applied Therapeutics, AstraZeneca, Boehringer Ingelheim, Eli Lilly, Hanmi, Merck, Oramed, Novartis, Novo Nordisk, Pfizer, and Sanofi; and has received honoraria for lectures sponsored by Boehringer Ingelheim, Eli Lilly, Novo Nordisk, and Sanofi. T.K. is an employee of Asahi Kasei Pharma Corporation. V.A.F. has received research grants from Bayer and Boehringer Ingelheim through his institution and directly received honoraria for consulting and lectures from Takeda, Novo Nordisk, Sanofi, Eli Lilly and Company, Astra Zeneca, Intarcia, and Asahi Kasei Pharma Corporation.

Data Availability

Original data generated and analyzed during this study are included in this published article or in the data repositories listed in “References.”

Clinical Trial Information

Trial registration number NCT02489773 (registered July 3, 2015).