sign in sign up
<<
Home , Glycated hemoglobin, Hypoalbuminemia

EndocrineOr i g i nJournal a l 2014 Advance Publication doi: 10.1507/endocrj. EJ13-0450

A newer conversion equation for the correlation between HbA1c and glycated albumin

Kaori Inoue1), 4), Tetsuro Tsujimoto1), Ritsuko Yamamoto-Honda1), Atsushi Goto2), 4), Miyako Kishimoto1), Hiroshi Noto1), Hiroshi Kajio1), Seiichi Doi3), Sumio Miyazaki3), Yasuo Terauchi4) and Mitsuhiko Noda2)

1) Department of Endocrinology, , and , Center Hospital, National Center for Global Health and Medicine, Tokyo 162-8655, Japan 2)Department of Diabetes Research, National Center for Global Health and Medicine, Tokyo 162-8655, Japan 3)Department of Clinical Laboratory, Center Hospital, National Center for Global Health and Medicine, Tokyo 162-8655, Japan 4)Department of Endocrinology and Metabolism, Yokohama City University Graduate School of Medicine, Yokohama 236-0004, Japan

Abstract. Glycated (HbA1c) and glycated albumin (GA) are frequently used as glycemic control markers. These markers are influenced by either altered hemoglobin metabolism or albumin metabolism. We investigated the correlation between HbA1c and GA by collecting only data that had not been affected by the turnover of either HbA1c or GA and proposed a novel equation for accurately estimating the extrapolated HbA1c (eHbA1c) value based on the GA value. Data sets for a total of 2461 occasions were obtained from 731 patients (including non-diabetes patients) whose HbA1c and GA values were simultaneously measured. Data sets obtained from patients undergoing hemodialysis, patients with hematological malignancies, pregnancy, chronic diseases, hyperthyroidism, steroid treatment or a transfusion during the past 3 months, or patients without albumin, hemoglobin, eGFR, or urinary measurements and data sets with an eGFR of less than 30 mL/min/1.73 m2, a hemoglobin level of less than 10 mg/dL, an albumin level of below 3.0 g/mL, or a urinary protein level of 3+ were excluded. Finally, we selected 284 data sets. We then analyzed these data sets, performed a scatter plot to examine the correlation between HbA1c and GA, and established an equation describing the resulting correlation. Based on all the data points, the resulting equation was HbA1c = 0.216 × GA + 2.978 [R² = 0.5882, P < 0.001].

Key words: , Glycated albumin

To evaluate glycemic control, glycated levels over the past few months. The metabolic are often used as glycemic control markers, rather turnover of albumin is faster than hemoglobin, with a than measuring the actual glucose levels using meth- lifespan of approximately 17 to 23 days. Accordingly, ods such as self- of blood glucose (SMBG) GA is used as an index of short-term glycemic con- and continuous glucose monitoring (CGM). Among trol [2]. For example, the GA : HbA1c ratio has been the various glycated proteins, glycated hemoglobin suggested to be a better marker of glycemic variability (HbA1c) and glycated albumin (GA) are frequently than HbA1c in , especially in fulminant used as glycemic control markers. HbA1c is used as type 1 diabetes [3]. Importantly, a few past studies have the gold standard index of glycemic control in clinical suggested that HbA1c is closely associated with the practice for diabetes treatment [1]. It has been reported fasting plasma glucose level, while GA is more closely that these markers are closely associated with the dia- associated with the postprandial plasma glucose level, betic complications. Since the lifespan of erythrocytes compared with the HbA1c level [4-5]. is approximately 120 days, HbA1c reflects the plasma Although these glycemic control markers are well correlated with blood glucose levels, HbA1c is influ- Submitted Oct. 27, 2013; Accepted Feb. 21, 2014 as EJ13-0450 Released online in J-STAGE as advance publication Mar. 28, 2014 enced by alterations in hemoglobin metabolism and Correspondence to: Mitsuhiko Noda, Department of Diabetes GA is influenced by alterations in albumin metabolism. Research, National Center for Global Health and Medicine, 1-21-1 In clinical practice, conditions such as , chronic Toyama, Shinjuku-ku, Tokyo 162-8655, Japan. renal failure, hypersplenism, chronic liver diseases, E-mail: [email protected] hyperthyroidism, , and pregnancy ©The Japan Endocrine Society

Endocrine Journal Advance Publication 2 Inoue et al. need to be considered when interpreting HbA1c or GA We excluded the patients whose previous HbA1c values [6-13]. However, past studies included the par- values were missing, or their HbA1c levels were ticipants who were suffering from those diseases, the changeable, we selected 550 data sets from the patients selection errors might be caused. as a result of the exclusion. Next, we excluded data sets In the present study, we intended to establish a lin- obtained from patients undergoing hemodialysis and ear regression equation describing the GA value with- patients with hematological malignancies, pregnancy, out altered albumin metabolism versus the HbA1c chronic liver diseases, hyperthyroidism, steroid treat- value without altered hemoglobin metabolism to cal- ment, or a blood transfusion during the past 3 months. culate an extrapolated HbA1c (eHbA1c) value for As a result, 368 data sets remained. Next, we excluded the accurate evaluation of glycemic control. Such an data sets without albumin, hemoglobin, eGFR, or uri- equation would enable quick decisions to be made in nary protein measurements and data sets with an eGFR clinical practice regarding diabetes treatment based of less than 30 mL/min/1.73 m2, a hemoglobin level on a given GA value, instead of measuring HbA1c, in of less than 10 mg/dL, an albumin level of below 3.0 patients whose blood control was not stable, change- g/mL, or a urinary protein level of 3+. Finally, we able within the short-term, or with altered hemoglobin selected 284 data sets (Fig. 1). metabolism. Many studies have reported the correla- This study was approved by the institutional ethi- tion between HbA1c and GA, but few studies have dis- cal committee of the National Center for Global Health cussed this correlation in detail. Thus, we investigated and Medicine and was performed in accordance with the correlation between HbA1c and GA by collecting the Declaration of Helsinki. only data that had not been affected by the turnover of either HbA1c or GA and proposed a novel equation for Statistical analysis accurately estimating eHbA1c based on the GA value. We performed the statistical analyses using Stata/IC 11. Data on the patient characteristics are shown as the Materials and Methods mean ± SE. A total of 284 data sets were used to per- form a regression analysis between HbA1c and GA. We retrospectively analyzed the medical charts We have therefore used the bootstrapping method [15] of patients attending the National Center for Global to assess internal validity of the performance of the pre- Health and Medicine (Tokyo, Japan) during 2011 and diction model. We applied 1,000 resampling procedure selected data sets for a total of 2461 occasions were with replacement to obtain the bootstrap bias-corrected obtained from 731 patients (including non-diabetes confidence intervals and the bias estimates -(= aver patients) whose HbA1c and GA values were simulta- age of bootstrapped estimates – estimate in the origi- neously measured. If these values were measured in the nal sample). We also performed stratification according patients on more than one occasion, we selected the to sex and the change in HbA1c as of the most recent data set containing the smallest HbA1c value. visit (decreased, no change, or increased). Then, for HbA1c was measured using high-performance liq- 145 patients in whom body mass index (BMI) data was uid (HPLC) (ARKRAY ADAMS- available, we also stratified the patients according to A1C HA-8160; Kyoto, Japan) and was corrected their BMI (<22 kg/m2, ≥22 and <25 kg/m2, ≥25 kg/m2) to the National Glycohemoglobin Standardization and performed a regression analysis. To evaluate the Program (NGSP) values [14]. GA was measured using statistical interaction, we incubated the product inter- an enzymatic method with albumin-specific protei- action terms in the regression models. nase, ketoamine oxidase, and an albumin assay reagent (Lucica GA-L; Asahi Kasei Pharma Co., Tokyo, Japan) Results with the use of an autoanalyzer (Hitachi 770; Hitachi Instruments Service Co., Tokyo, Japan). Each patient The 284 individuals whose data were analyzed con- was assessed for clinical features such as age, sex, sisted of 201 men (62.5 ± 0.9 years) and 83 women (65.8 height, body weight, body mass index, blood and ± 1.6 years), as shown in Table 1. The mean HbA1c was sample data, history and duration of diabetes mellitus, 7.5% ± 0.1% (men) and 7.4% ± 0.2% (women), and the medications, and complications based on the data con- mean GA was 20.9% ± 0.3% (men) and 20.9% ± 0.7% tained in the medical records. (women). Of the 201 men, 10 individuals had no history

Endocrine Journal Advance Publication Endocrine Journal Advance Publication Novel equation to calculate eHbA1c 3

Data sets for a total of 2461 occasions were obtained from 731 patients whose HbA1c and GA values were simultaneously measured.

Excluded The patients whose previous HbA1c values were missing, or their HbA1c levels were changeable. 550 patients Excluded Hemodialysis (12*) Hematological malignancies (111*) Pregnancy (2*) Chronic liver diseases (19*) Hyperthyroidism (6*) Steroid treatment (82*) Excluded 368 patients Blood transfusion during the past Data of Alb, Hb, eGFR, urinary protein 3months (39*) were missing. eGFR<30 (44*) Hb<10 (16*) Alb≤3.0 or urinary protein3+ (37*) 284 patients *partially overlapped Fig. 1 Flow diagram for study. Data sets for a total of 2461 occasions were obtained from 731 patients (including non-diabetes patients) whose HbA1c and GA values were simultaneously measured. If these values were measured in the patients on more than one occasion, we selected the data set containing the smallest HbA1c value. From among 2461 data sets, we selected 731 data sets from 731 patients. We excluded the patients whose previous HbA1c values were missing, or their HbA1c levels were changeable, we selected 550 data sets from the patients as a result of the exclusion. Next, we excluded data sets obtained from patients undergoing hemodialysis or patients with hematological malignancies, pregnancy, chronic liver diseases, hyperthyroidism, steroid treatment or a blood transfusion during the past 3 months. As a result, 368 data sets remained. Next, we excluded data sets without albumin, hemoglobin, eGFR, or urinary protein measurements, data sets with an eGFR of less than 30 mL/min/1.73 m2, a hemoglobin level of less than 10 mg/dL, an albumin level of below 3.0 g/mL, or a urinary protein level of 3+. Finally, 284 data sets were analyzed.

Table 1 Characteristics of 284 patients Characteristics men (n = 201) women (n = 83) Age (years) 62.5±0.9 65.8±1.6 HbA1c (%) 7.5±0.1 7.4±0.2 GA (%) 20.9±0.3 20.9±0.7 non-diabetic/type1DM/type2DM/other (person) 10/7/180/4 12/8/63/0 diet/OHAs*/**/GLP-1** (person) 30/143/52/4 16/57/22/2 Mean±SE SE, standard error; OHAs, oral hypoglycemic agents; *, OHAs alone or its combination therapy with insulin or GLP-1; **, insulin alone or its combination therapy with OHAs or GLP-1 combination therapy with OHAs

of diabetes, 7 patients had type 1 diabetes, 180 patients pies. Of the 284 patients, BMI values were available for had , and 4 patients had some other type a total of 145 patients. The mean BMI was 24.9 ± 0.4 of diabetes. Of the 83 women, 12 individuals had no kg/m2 (108 men) and 23.3 ± 0.6 kg/m2 (37 women). , 8 patients had type 1 diabetes, and We performed a scatter plot to examine the correlation 63 patients had type 2 diabetes. Regarding the medical between HbA1c and GA using data from 284 patients, treatment, the diabetic men were treated using diet ther- and established an equation describing the resulting cor- apy (n = 30), oral anti-hypoglycemic agents (n = 143), relation (Fig. 2). Based on all the data points, the result- insulin (n = 52), GLP-1 (n = 4), or combination thera- ing equation was as follows: HbA1c = 0.216 × GA + pies. The diabetic women were treated using diet ther- 2.978 [R² = 0.5882, P < 0.001]. The bootstrap bias-cor- apy (n = 16), oral anti-hypoglycemic agents (n = 57), rected 95% confidence intervals (bias estimates) were insulin (n = 22), GLP-1 (n = 2), or combination thera- from 0.193 to 0.238 (0.001) for the slope and from

Endocrine Journal Advance Publication Endocrine Journal Advance Publication 4 Inoue et al.

Fig. 2 Correlation between HbA1c and GA according to sex. We performed a scatter plot to examine the correlation between HbA1c and GA using data from 284 patients, and established an equation describing the resulting correlation. The equation was as follows: HbA1c = 0.216 × GA + 2.978 [R² = 0.5882, P < 0.001]. In a stratified analysis according to sex, no interaction existed.

Fig. 3 Correlation between HbA1c and GA according to the change in the HbA1c value. A stratified analysis was performed according to the change in HbA1c as of the most recent visit (decreased, no change, or increased). We performed a scatter plot to examine the correlation between HbA1c and GA. In a stratified analysis according to the change in HbA1c, no interaction was found.

2.547 to 3.458 (-0.0143) for the intercept. No interaction no interaction was found in a stratified analysis accord- existed when the data was stratified according to sex. ing to the change in HbA1c (Fig. 3). Next, we stratified the analysis according to the Similarly, BMI data for 108 men and 37 women were change in HbA1c as of the most recent visit (decreased, stratified into 3 groups: <22 kg/m2, ≥22 and <25 kg/ no change, or increased). We performed a scatter plot to m2, ≥25 kg/m2. We performed a scatter plot to exam- examine the correlation between HbA1c and GA, but ine the correlation between HbA1c and GA. No inter-

Endocrine Journal Advance Publication Endocrine Journal Advance Publication Novel equation to calculate eHbA1c 5

Fig. 4 Correlation between HbA1c and GA according to BMI. One hundred eight men and 37 women whose BMI values were known were stratified according to BMI as follows: <22 kg/m2, ≥22 and <25 kg/m2, ≥25 kg/m2. We performed a scatter plot to examine the correlation. In a stratified analysis according to BMI, no interaction existed.

action was found in a stratified analysis according to that glycemic control markers are closely associated BMI (Fig. 4). with the diabetic complications [22]. In addition, we analyzed the correlation by restrict- Thus, we investigated the correlation between ing the analysis to the 243 patients with type 2 diabe- HbA1c and GA by collecting only data that had not tes. The resulting equation was HbA1c = 0.211×GA + been affected by the turnover of either HbA1c or GA 3.185 [R2 = 0.5307, P < 0.001]. and proposed a novel equation for accurately estimat- ing eHbA1c based on the GA value. Discussion In this study, we analyzed the correlation between HbA1c and GA using clinical data from 284 patients Our objective for the treatment of diabetes is to pre- whose HbA1c values were stable, who were not preg- vent the incidence or the progression of the micro- nant, and who were free from diseases affecting hemo- vascular and macrovascular complications that are or albumin metabolism. As a result, the follow- specific for diabetes. DCCT (Diabetes Control and ing equation was established: eHbA1c = 0.216 × GA + Complications Trial), UKPDS (United Kingdom 2.978. Prospective Diabetes Study), and the Kumamoto Study When the patients were stratified into 3 groups suggested that better glycemic control was associ- according to the change in HbA1c as of the most recent ated with a lower risk of microvascular complications visit (decreased, no change, or increased), no interac- [16-18]. Further, the Funagata Study and the DECODE tion was found. This result suggests that the glycemic (Diabetes Epidemiology: Collaborative analysis of control of the subjects in the present study was suffi- Diagnostic criteria in Europe) study suggested that ciently stable. blood glucose levels at 2 hours after the 75g OGTT The high correlation between HbA1c and GA is well (oral ) were more strongly asso- known [23]. Our results are consistent with Tahara’s ciated with the risk of (CVD) findings. Tahara examined the correlation using a lin- than fasting glucose levels [19-20]. Recently, the EDIC ear regression analysis among 154 patients with type 2 study (Epidemiology of Diabetes Interventions and diabetes. Their regression equation was HbA1c (JDS) = Complications), an observational follow-up study per- 0.204 × GA + 2.59 [24], which is similar to our equa- formed since the end of the DCCT and in an extended tion. Although we enrolled patients with various types DCCT cohort, was reported [21]. It has been reported of diabetes as well as non-diabetic patients in our study,

Endocrine Journal Advance Publication Endocrine Journal Advance Publication 6 Inoue et al. most (243 out of 284) of the patients had type 2 dia- turnover of albumin metabolism [12]. Hyperthyroidism betes. This may explain the similarity of our findings and steroid treatment, in addition to nephropathy, are with those of Tahara. In addition, we analyzed the cor- known to lower the GA values because of accelerated relation by restricting the analysis to the 243 patients albumin synthesis. Thyroid is also known to with type 2 diabetes. The resulting equation was HbA1c promote albumin metabolism. A study showed that the = 0.211×GA + 3.185 [R2 = 0.5307, P < 0.001], which serum GA level was reduced in patients with thyrotox- further supports the similarity of our results with those icosis, but no apparent change in HbA1c was seen. In of Tahara. The differences in the intercepts between the addition, GA had significant inverse correlations with two studies might have arisen because of the difference the free T3 and free T4 levels, as well as a sig- between the JDS value and the NGSP value. Moreover, nificant positive correlation with the serum TSH level by restricting the analysis to the 15 patients with type 1 [13]. Additionally, the BMI is known not to affect diabetes, the resulting equation was HbA1c = 0.195 × GA the HbA1c values, while a negative correlation exists + 3.173 [R2 = 0.9070, P < 0.001]. Non-diabetic were 22 between the BMI and GA. A previous study showed patients, the correlation between HbA1c and GA was that in obese children, a significant positive correla- not so strong. Therefore, we could not evaluate its equa- tion was seen between HbA1c and BMI, but a signif- tion model in non-diabetics in this study. icant negative correlation was seen between GA and In clinical practice for diabetes treatment, patients suf- BMI [25]. Similarly, in adult diabetic patients, a signifi- fering from other diseases that could affect the HbA1c cant negative correlation between the BMI and the GA data, even if they do not affect the glucose level, are fre- level was seen. By contrast, no correlation between the quently encountered. The HbA1c value can affect the BMI and the HbA1c level was seen [26]. While the rea- lifespan of erythrocytes, while an aberrant GA value is sons for these relations remain unknown, one possible possible if albumin turnover is changeable. Accordingly, explanation is that obesity increases albumin turnover. in patients with conditions such as anemia, chronic renal Furthermore, chronic inflammatory reactions might failure, hypersplenism, chronic liver diseases, hyperthy- also increase albumin turnover. In our study, however, roidism, or hypoalbuminemia, the relationship between a stratified analysis according to the BMI showed no HbA1c and GA may be affected. Thus, a careful selec- interaction between these parameters, although the rea- tion of study participants is important to estimate a reli- son for the lack of an interaction was not clear. able correlation between HbA1c and GA. Our study had certain limitations. First, we retrospec- In patients with , because the tively selected patients in whom simultaneous HbA1c lifespan of the erythrocytes is shortened, the HbA1c and GA measurements had been obtained. Thus, a selec- values are lower relative to the plasma glucose level tion bias may exist. Second, as the data were collected [6]. On the other hand, the HbA1c values are higher from a single hospital and the GA values were not stan- in patients with iron deficiency anemia, and false high dardized, the present results might not be directly appli- HbA1c values are observed in iron-deficient states cable to other hospitals. Although the bootstrap con- without anemia [7]. During pregnancy, the HbA1c val- fidence intervals and the bias estimates supported the ues are higher during the third trimester because of iron internal validity of our findings, external validation is deficiency, whereas the GA is not affected. Therefore, needed before applying to other populations. Third, we GA may be a more suitable marker for monitoring gly- enrolled patients with various types of diabetes as well cemic control during pregnancy [8-9]. In chronic liver as non-diabetic patients in our study. In the future, we diseases, such as chronic and liver , are going to investigate the correlation between HbA1c hypersplenism lowers the HbA1c values because of the and GA by sorting out the types of diabetes. shortened lifespan of the erythrocytes, whereas it raises In conclusion, we propose an equation for calculat- the GA values because of reduced albumin synthesis ing eHbA1c to evaluate the glycemic control of patients and the prolonged half-life of serum albumin [10]. In with altered hemoglobin metabolism. cases with chronic renal failure, renal anemia lowers the HbA1c values because the lifespan of the erythro- Acknowledgments cytes is shortened [11]. However, in patients with dia- betic nephropathy presenting with marked proteinu- This work was funded by health sciences research ria, the GA values are lower because of the increased grants (comprehensive research on life-style related

Endocrine Journal Advance Publication Endocrine Journal Advance Publication Novel equation to calculate eHbA1c 7

diseases including cardiovascular diseases and diabe- Disclosure tes mellitus H22-019 and H25-016) from the Ministry of Hearth, Labour and Welfare of Japan. None of the authors have any potential conflicts of interest associated with this research.

References

1. International Expert Committee (2009) International alysis. Int 73: 1062-1068. Expert Committee report on the role of the A1C assay 12. Okada T, Nakao T, Matsumoto H, Nagaoka Y, Tomaru in the diagnosis of diabetes. Diabetes Care 32: 1327- R, et al. (2011) Influence of on glycated 1334. albumin values in diabetic patients with chronic kidney 2. Tahara Y, Shima S (1995) Kinetics of HbA1c, glycated disease. Intern Med 50: 23-29. albumin, and and analysis of their weight 13. Koga M, Murai J, Saito H, Matsumoto S, Kasayama functions against preceding plasma glucose level. S (2009) Effects of thyroid hormone on serum gly- Diabetes Care 18: 440-447. cated albumin levels: Study on non-diabetic subjects. 3. Matsumoto H, Murase-Mishiba Y, Yamamoto N, Diabetes Res Clin Pract 84: 163-167. Sugitatsu-Nakatsukasa S, Shibasaki S, et al. (2012) 14. Kashiwagi A, Kasuga M, Araki E, Oka Y, Hanafusa Glycated albumin to glycated hemoglobin ratio is a sen- T, et al. (2012) International clinical harmonization of sitive indicator of blood glucose variability in patients glycated hemoglobin in Japan: From Japan Diabetes with fulminant type 1 diabetes. Intern Med 51: 1315- Society to National Glycohemoglobin Standardization 1321. Program values. Diabetol Int 3: 8-10. 4. Yoshiuchi K, Matsuhisa M, Katakami N, Nakatani Y, 15. Efron B (1982) The Jackknife, the Bootstrap and Other Sakamoto K, et al. (2008) Glycated albumin is a bet- Resampling Plans. Philadelphia: Society for Industrial ter indicator for glucose excursion than glycated hemo- and Applied Mathematics globin in type 1 and type 2 diabetes. Endocr J 55: 503- 16. The Diabetes Control and Complications Trial research 507. group (1993) The effect of intensive treatment of dia- 5. Sakuma N, Omura M, Oda E, Saito T (2011) Converse betes on the development and progression of long-term contributions of fasting and postprandial glucose to complications in insulin-dependent diabetes mellitus. N HbA1c and glycated albumin. Diabetol Int 2: 162-171. Engl J Med 329: 977-986. 6. Panzer S, Kronik G, Lechner K, Bettelheim P, Neumann 17. UK Prospective Diabetes Study (UKPDS) group (1998) E, et al. (1982) Glycosylated (GHb): an Intensive blood-glucose control with sulphonylureas index of red cell survival. Blood 59: 1348-1350. or insulin compared with conventional treatment and 7. Kim C, Bullard KM, Herman WH, Beckles GL (2010) risk of complications in patients with type 2 diabetes Association between iron deficiency and A1c levels (UKPDS33). Lancet 352: 837-853. among adults without diabetes in the national health 18. Okubo Y, Kishikawa H, Araki E, Miyata T, Isami S, et and nutrition examination survey, 1999-2006. Diabetes al. (1995) Intensive insulin therapy prevents the pro- Care 33: 780-785. gression of diabetic microvascular complications in 8. Hashimoto K, Noguchi S, Morimoto Y, Hamada S, Japanese patients with non-insulin-dependent diabe- Wasada K, et al. (2008) A1C but not serum glycated tes mellitus: a randomized prospective 6-year study. albumin is elevated in late pregnancy owing to iron defi- Diabetes Res Clin Pract 28: 103-117. ciency. Diabetes Care 31: 1945-1948. 19. Tominaga M, Eguchi H, Manaka H, Igarashi K, Kato T, 9. Hashimoto K, Osugi T, Noguchi S, Morimoto Y, Wasada et al. (1999) Impaired glucose tolerance is a risk factor K, et al. (2010) A1C but not serum glycated albumin is for cardiovascular disease, but not impaired fasting glu- elevated because of iron deficiency in late pregnancy in cose. The Funagata Diabetes Study. Diabetes Care 22: diabetic women. Diabetes Care 33: 509-511. 920-924. 10. Koga M, Kasayama S, Kanehara H, Bando Y (2008) 20. The DECODE study group on behalf of the European CLD (chronic liver diseases)-HbA1c as a suitable indi- Diabetes Epidemiology group (2004) Prediction of cator for estimation of mean plasma glucose in patients the risk of cardiovascular mortality using a score that with . Diabetes Res Clin Pract 81: includes glucose as a risk factor. The DECODE study. 258-262. Diabetologia 47: 2118-2128. 11. Peacock TP, Shihabi ZK, Bleyer AJ, Dolbare EL, Byers 21. Nathan DM, McGee P, Steffes MW, Lachin JM and the JR, et al. (2008) Comparison of glycated albumin and DCCT/EDIC research group (2014) Relationship of hemoglobin A1c levels in diabetic subjects on hemodi- glycated albumin to blood glucose and glycated hemo-

Endocrine Journal Advance Publication Endocrine Journal Advance Publication 8 Inoue et al.

globin (HbA1c) values and to , nephropathy between levels of HbA1c and glycated albumin by liner and cardiovascular outcomes in the DCCT/EDIC study. regression analysis using a measurement error model. Diabetes 63: 282-290. Diabetes Res Clin Pract 84: 224-229. 22. Li JP, Lin L, Wei FS, Qi Z, Rui YZ, et al. (2007) 25. Nishimura R, Kanda A, Sano H, Matsudaira T, Miyashita Increased serum glycated albumin level is associated Y, et al. (2006) Glycated albumin is low in obese, non- with the presence and severity of coronary artery dis- diabetic children. Diabetes Res Clin Pract 71: 334- ease in type 2 diabetic patients. Circ J 71: 1067-1073. 338. 23. Nathan DM, Steffes MW, Sun W, Rynders GP, Lachin 26. Koga M, Matsumoto S, Saito H, Kasayama S (2006) JM (2011) Determining stability of stored samples ret- Body mass index negatively influences glycated albu- rospectively: the validation of glycated albumin. Clin min, but not glycated hemoglobin, in diabetic patients. Chem 57: 286-290. Endocr J 53: 387-391. 24. Tahara Y (2009) Analysis of the method for conversion

Endocrine Journal Advance Publication