Total short-term variability in biomarkers of hyperglycemia in older adults

There is growing interest in the use of nontraditional short-term biomarkers of hyperglycemia (fructosamine, glycated albumin and 1,5-anhydroglucitol [1,5-AG]) to complement fasting glucose and hemoglobin A1c [HbA1c] for prognosis and management of diabetes.(1) The within-person variability in HbA1c and fasting glucose have been previously characterized,(2–4) but little is known about the variability of these alternative biomarkers of hyperglycemia in the general population. We quantified the 6-week total variability in fructosamine, glycated albumin and 1,5-AG in older adults with and without diabetes, and compared it to that of fasting glucose and HbA1c.

We included 153 participants from the Atherosclerosis Risk in Communities (ARIC) Study(5) who attended the initial visit 5 exam (2011-13) and returned for a second visit, scheduled 4-8 weeks later; had complete data; fasted ≥8 hours at both visits; and did not have outlying values. Institutional review boards approved all procedures, and all study participants provided written informed consent.

Serum fructosamine, glycated albumin and 1,5-AG were measured using the Roche Cobas 6000 (Roche Diagnostics Corporation). Fructosamine was measured using a colorimetric method (Roche Diagnostics Corporation). Glycated albumin (Asahi Kasei Lucica GA-L) and 1,5-AG (GlycoMark) were measured using enzymatic methods. Glucose was measured in plasma using the Beckman Olympus 480 analyzer (Beckman Coulter, Inc.) using a hexokinase method. HbA1c was measured in whole blood using the using a high performance liquid chromatography method (Tosoh G7 analyzer, Tosoh Medics, Inc.), standardized to the Diabetes Control and Complications Trial (DCCT) assay. The inter-assay CVs were 3.2% for fructosamine at a mean concentration of 220.3 μmol/L, 4.4% for glycated albumin at 0.45 g/dL, 0.9% for 1,5-AG at 18.0 μg/mL, 2.7% for glucose at 121.7 mg/dL, and 1.9% for HbA1c at 5.36%.

Analyses were conducted separately in persons with and without diagnosed diabetes. For each biomarker, we calculated means by diabetes status: the mean at the original exam, the mean at the second exam, and the mean difference (second – original). To partition the total variance of the repeated measurements into the between-subject variance (σ_BS²) and within-subject variance (σ_WS²), we used linear mixed effects models with each biomarker as the dependent variable and the participant as a random effect. We calculated the between-person coefficient of variation $\left({\mathrm{CV}}_{\mathrm{G}}\right):\left[\right(\sqrt{{{\sigma }_{\mathit{BS}}}^2})∕\mu ]\ast 100$, where μ is the mean of all values (both original and second measurements). Similarly, we calculated the within-person coefficient of variation $\left({\mathrm{CV}}_{\mathrm{W}}\right):\left[\right(\sqrt{{{\sigma }_{\mathit{WS}}}^2})∕\mu ]\ast 100$. The CV_W is a function of the within-person biological coefficient of variation (CV_I) and the analytical coefficient of variation (CV_A) (or each method’s CV reported by the laboratory): ${\mathit{CV}}_W=\sqrt{{{\mathit{CV}}_A}^2+{{\mathit{CV}}_I}^2}$. We then calculated the index of individuality: $\left(\sqrt{{{\mathit{CV}}_A}^2+{{\mathit{CV}}_I}^2}\right)∕{\mathit{CV}}_G$ or equivalently, CV_W/CV_G. We also calculated the intraclass correlation coefficient (ICC): ${{\sigma }_{\mathit{BS}}}^2∕({{\sigma }_{\mathit{BS}}}^2+{{\sigma }_{\mathit{WS}}}^2)$. All statistical analyses were conducted using Stata, version 13.0 (StataCorp, College Station, Texas, USA).

The mean age was 76 years, 33% had a diagnosis of diabetes, 39% were male, and 74% were white. The mean time between the original and second examinations was 45 days (SD, 16 days) (range of 23-102 days). Estimates of total within-person variability (CV_W) for fructosamine, glycated albumin, and 1,5-AG were: 3.4%, 2.7%, and 2.9%, respectively, in persons without diabetes, and 3.7%, 3.8%, and 5.7% in persons with diabetes (Table 1). CV_ws for fasting glucose and HbA1c were similar to those previously reported (Table 1).(2,3) The point estimates for the CV_Ws were greater in persons with diagnosed diabetes—60% of whom reported taking glucose-lowering medication(s)—as compared to those without diagnosed diabetes, although the confidence intervals for fructosamine and glycated albumin in persons with and without diabetes were overlapping. The results in persons with diabetes may partially reflect the influence of diabetes management on glycemic variability. The CV_ws of fructosamine, glycated albumin, and 1,5-AG were between those of fasting glucose and HbA1c. Of the nontraditional biomarkers, the ICC was lowest for fructosamine, and highest for 1,5-AG (Table 1). Patterns of Spearman’s rank correlation coefficients were similar to those of ICCs (Table 1). All biomarkers had a favorably low index of individuality, with 1,5-AG having the lowest (Table 1). Results for HbA1c and fasting glucose were similar to previously reported findings.(2–4)

Table 1. Total variability in biomarkers of hyperglycemia in older adults with and without diabetes, the Atherosclerosis Risk in Communities Study, 2011-13, N=153.

	Original exam Mean (SD)	Second exam Mean (SD)	Difference (Second- Original) Mean (SD)	CVW (95% CI)†	ICC (95% CI)†	r (95% CI)	Index of Individuality (95% CI)†
No Diagnosed Diabetes (N=103)
Fructosamine, μmol/L	241.5 (22.9)	239.4 (21.0)	−2.10 (11.3)	3.4% (2.9, 3.8)	0.86 (0.83, 0.89)	0.83 (0.76, 0.88)	0.40 (0.34, 0.46)
Glycated albumin, %	13.8 (1.5)	13.7 (1.7)	−0.04 (0.5)	2.7% (2.3, 3.0)	0.95 (0.94, 0.96)	0.91 (0.88, 0.94)	0.24 (0.20, 0.27)
1,5-AG, μg/mL	17.5 (6.0)	17.6 (6.1)	0.04 (0.7)	2.9% (2.7, 3.2)	0.99 (0.99, 0.99)	0.99 (0.99, 0.99)	0.09 (0.08, 0.09)
Fasting glucose, mg/dL	104.7 (17.6)	104.5 (15.9)	−0.18 (7.9)	5.3% (4.6, 6.0)	0.89 (0.85, 0.93)	0.72 (0.61, 0.80)	0.35 (0.27, 0.44)
HbA1c, %	5.7 (0.4)	5.7 (0.4)	−0.01 (0.1)	1.5% (1.3, 1.7)	0.95 (0.95, 0.96)	0.95 (0.92, 0.96)	0.22 (0.19, 0.25)
Diagnosed Diabetes (N=50)
Fructosamine, μmol/L	263.2 (39.8)	261.8 (38.7)	−1.46 (13.6)	3.7% (3.0, 4.3)	0.94 (0.92, 0.96)	0.84 (0.73, 0.90)	0.25 (0.20, 0.31)
Glycated albumin, %	15.5 (2.8)	15.7 (2.9)	0.2 (0.8)	3.8% (2.9, 4.7)	0.95 (0.94, 0.97)	0.91 (0.85, 0.95)	0.22 (0.17, 0.27)
1,5-AG, μg/mL	15.1 (6.5)	15.1 (6.5)	0.04 (1.2)	5.7% (4.2, 7.2)	0.98 (0.98, 0.99)	0.98 (0.97, 0.99)	0.13 (0.10, 0.17)
Fasting glucose, mg/dL	125.5 (29.8)	125.3 (32.7)	−0.14 (17.1)	9.6% (7.3, 11.8)	0.85 (0.80, 0.90)	0.84 (0.73, 0.90)	0.42 (0.31, 0.53)
HbA1c, %	6.3 (0.8)	6.3 (0.8)	0.04 (0.2)	2.0% (1.5, 2.5)	0.98 (0.97, 0.98)	0.96 (0.92, 0.98)	0.16 (0.12, 0.19)

Abbreviations: CI, confidence interval; CV_W, within-person coefficient of variation; ICC, intraclass correlation coefficient; r, Spearman’s rank correlation coefficient; SD, standard deviation

^†95% CIs were bootstrapped using 200 replications

Estimates of total variability of these biomarkers were intermediate between fasting glucose and HbA1c, consistent with their biology.(1) Notably, 1,5-AG had very high correlations and estimates of reliability, which may be partially attributed to the relatively wide range of values for this biomarker. All biomarkers of hyperglycemia had a low index of individuality, which suggests that changes in biomarker levels over time within an individual primarily reflects biological change. Limitations of our study included having only two measurements for each participant, and using internal laboratory quality control materials instead of running duplicates to obtain the analytical CV. In summary, we found that nontraditional biomarkers of hyperglycemia are consistent over approximately 6 weeks, and had lower within-person variability than fasting glucose. These data should help inform the use of these biomarkers in research and clinical settings.