Glycated Albumin in Pristine and Non-Pristine Stored Samples in the National Health and Nutrition Examination Survey (NHANES) 1999–2004

Abstract

Background:

Glycated albumin is cleared by the Food and Drug Administration (FDA) for clinical use in diabetes care. To understand its performance in the general US population, we conducted measurements in >19 000 samples from the National Health and Nutrition Examination Survey (NHANES). Of these samples, 5.7% had previously undergone at least 2 freeze–thaw cycles and were considered “non-pristine.”

Methods:

We measured glycated albumin and albumin using the Lucica GA-L (Asahi Kasei) assay in stored serum samples from NHANES 1999–2004. Serum albumin (Roche/Beckman) was previously measured. We examined the correlations of percent glycated albumin with hemoglobin A1C (HbA_1c)and fasting glucose in the pristine and non-pristine samples. We also measured cystatin C (Siemens) and compared these to cystatin C (Dade Behring) previously obtained in a subsample.

Results:

Glycated albumin (%) was significantly lower in pristine vs non-pristine samples (13.8% vs 23.4%, P < 0.0001). The results from the Asahi Kasei albumin assay (g/dL) were highly correlated with albumin originally measured in NHANES (Pearson’s correlation coefficient, r = 0.76) but values were systematically higher (+0.25 g/dL, P < 0.0001). Cystatin C (Siemens) was similar to previous cystatin C measurements (r = 0.98) and did not differ by pristine status (P = 0.119). Glycated albumin (%) was highly correlated with HbA_1c and fasting glucose in pristine samples (r = 0.78 and r = 0.71, respectively) but not in non-pristine samples (r = 0.11 and r = 0.12, respectively).

Conclusions:

The performance of the glycated albumin assay in the pristine samples was excellent. Performance in non-pristine samples was highly problematic. Analyses of glycated albumin in NHANES 1999–2004 should be limited to pristine samples only. These results have major implications for the use of these public data.

INTRODUCTION

Glycated albumin is a short-term (2 to 4 week) marker of glycemia and is a potential alternative to hemoglobin A1C (HbA_1c). Glycated albumin was recently cleared by the FDA for clinical use in the US. To rigorously characterize the epidemiology of this emerging biomarker in the general US population, we measured glycated albumin (albumin) using a state-of-the-art method (Lucica GA-L assay, Asahi Kasei Pharma Corp) in 23,095 stored samples from the National Health and Nutrition Examination Survey (NHANES) 1999–2004. We also measured cystatin C (Siemens) in these samples. Samples were obtained from 2 repositories: the CDC Biorepository and SriSai (1, 2). The majority of samples (87.3%) were obtained from the CDC Biorepository and were pristine (i.e., had never undergone a freeze–thaw cycle); the remaining samples were obtained from the SriSai repository and were non-pristine (i.e., had undergone at least 2 freeze–thaw cycles). We measured glycated albumin and cystatin C as a part of a surplus sera project funded by the Biomarkers Consortium of the Foundation for the National Institutes of Health. All data are publicly available (2).

Our objective was to compare the glycated albumin results obtained in the pristine vs non-pristine samples. We conducted direct comparisons of serum albumin measured using the Lucica GA-L method to existing serum albumin data available from NHANES 1999–2004 (Roche Diagnostics/Beckman Coulter). We also compared glycated albumin HbA_1c and fasting glucose, which have established correlations with glycated albumin (3). To determine if the pristine vs non-pristine status influenced other biomarkers, we compared our measurements of cystatin C (Siemens) to cystatin C (Dade Behring) previously conducted in a subsample of NHANES 1999–2002 (4). Our goal was to provide recommendations for the use of the glycated albumin data in the pristine and non-pristine samples to investigators interested in using these publicly available data from our NHANES 1999–2004 stored sera study.

MATERIALS AND METHODS

Study Population

The NHANES is a large, cross-sectional survey of the civilian non-institutionalized population in the US (5). Our study population included persons who participated in NHANES 1999–2004 (3 survey cycles; see online Supplemental Methods).

Storage of Samples

We obtained surplus serum samples from the CDC Biorepository (Lawrenceville, GA) and SriSai (Frederick, MD) repositories. The CDC Biorepository (formerly, CDC and the Agency for Toxic Substances and Disease Registry Specimen Packaging, Inventory and Repository (CASPIR)) is a long-term repository for NHANES specimens; these “pristine” samples were stored immediately after collection at −80 °C or below in vapor-phase liquid nitrogen and had never undergone a freeze–thaw cycle (1). Samples from the SriSai laboratory had been stored at −80 °C and had undergone at least 2 freeze–thaw cycles (i.e., non-pristine) prior to being shipped to our laboratory.

Assay Methods

Laboratory testing was completed between 2018 and 2020 at the University of Maryland School of Medicine. Glycated albumin (albumin) was measured in serum using a complex method by Asahi Kasei Pharma (Lucica-GA-L) adapted to the Siemens Dimension Vista 1500 (Siemens Healthcare Diagnostics). The assay requires separate measurements of total albumin (bromocresol purple) and glycated albumin (enzymatic method utilizing ketoamine oxidase and an albumin-specific protease). The glycated albumin result is expressed as a percentage of total albumin: {[(glycated albumin concentration in g/dL)/(serum albumin concentration in g/dL)/1.14] × 100} + 2.9 (6). The CVs for albumin were 2.92% (low, 4.27 to 5.21 g/dL) and 5.12% (high, 4.25 to 5.19 g/dL); the CVs for glycated albumin were 4.93% (low, 0.56 to 0.758 g/dL) and 8.48% (high, 1.59 to 1.944 g/dL). We measured cystatin C using an immunoassay (Siemens Healthcare Diagnostics) on the Siemens Dimension Vista 1500. The lower and upper limits of detection were 0.23 mg/L and 8.00 mg/L, respectively. The CVs for cystatin C were 3.54% (low, 0.77 to 1.15 mg/L) and 4.36% (high, 1.61 to 2.41 mg/L).

Serum albumin, plasma glucose, and HbA_1c were measured as part of the original NHANES 1999–2004 protocols (7–9). Albumin was measured with a bromocresol purple method using the Hitachi 917 analyzer (Roche Diagnostics) in 1999–2002 and the Beckman Synchron LX20 (Beckman Coulter) in 2003–2004. Plasma glucose was measured in the subsample of morning fasting participants by a hexokinase enzymatic method using a Cobas Mira Chemistry System (Roche Diagnostic Systems). HbA_1c was measured using the Primus Automated HPLC system [models CLC330 (Primus I) or CLC385 (Primus IV)]. Cystatin C was measured in 2006 as part of a prior stored sera study in a subsample of NHANES 1999–2002 participants (4) using the Dade Behring N Latex Cystatin C assay (10).

The protocols for conduct of NHANES were approved by the National Center for Health Statistics (NCHS) ethics review board and informed consent was obtained from all participants (11). Our stored sera study was approved by the ethics review board of the NCHS.

Statistical Analyses

We compared serum albumin results in the pristine and non-pristine surplus samples (Asahi Kasei Lucica-GA-L assay) to serum albumin in the original NHANES (Roche/Beckman) using sunflower plots (a type of scatterplot that is useful for illustrating data density), Bland-Altman plots, and linear regression models. We also generated scatterplots and Pearson’s correlations for glycated albumin with HbA_1c and fasting plasma glucose, by pristine status. We conducted similar analyses in the subsample with cystatin C measurements.

RESULTS

Of the 19 067 participants in included in our study [mean (SD) age: 38.6 (22.4) years; 51% female], 94.3% of samples were received from the CDC Biorepository (pristine samples); only 5.7% of samples were non-pristine (SriSai repository; see online Supplemental Table 1).

Glycated albumin (g/dL and %) was substantially higher and more variable in non-pristine samples (Table 1). Albumin (g/dL) was modestly higher using the Asahi Kasei method vs the original albumin assay (Roche/Beckman) in the pristine samples but the two methods were highly correlated in both the pristine and non-pristine samples (Table 1; Supplemental Fig. 1).

Table 1.

Summary statistics of serum samples by pristine status, NHANES 1999–2004 (n = 19 067).^a

	Pristine	Non-pristine
Albumin and glycated albumin	n = 17 978 (94.3%)	n = 1089 (5.7%)
Asahi Kasei assay: serum albumin, g/dL	4.6 (0.52)	4.4 (0.58)
Roche/Beckman assay: serum albumin, g/dL	4.3 (0.38)	4.4 (0.40)
Difference (SD)	0.263b (0.329)	0.023 (0.424)
Pearson correlation (r)	0.77	0.68
Asahi Kasei assay: glycated albumin, g/dL	0.56 (0.15)	1.01 (0.66)
Asahi Kasei assay: percent glycated albumin, %	13.8 (3.1)	23.4 (14.8)
Cystatin C	n = 5187 (95.4%)	n = 251 (4.6%)
Siemens assay: Cystatin C, mg/L	0.89 (0.45)	0.85 (0.55)
Dade Behring: Cystatin Cc, mg/L	0.97 (0.56)	0.87 (0.59)
Difference (SD)	−0.077b (0.145)	−0.021b (0.081)
Pearson correlation (r)	0.98	0.99

^aMean (SD) unless otherwise indicated.

^bP < 0.05.

^cCalibrated: IFCC standard cystatin C = 1.12 × [(NHANES 1999–2002 reported cystatin C) − 0.12].

There were 31 potential outliers (0.16% of pristine samples and 0.18% of non-pristine samples) where the difference between albumin measurements were >3 SDs below the mean. Cystatin C in the surplus serum samples (Siemens Vista) compared to the earlier cystatin C measurements (Dade Behring) were similar in pristine (Pearson’s correlation coefficient, r = 0.98) and non-pristine (r = 0.99) samples.

Percent glycated albumin was highly correlated with HbA_1c (r = 0.78) and fasting plasma glucose (r = 0.71) in the pristine samples but not in the non-pristine samples (r < 0.13) (Fig. 1).

Fig. 1.

Sunflower plots of glycated albumin vs HbA1c and fasting plasma glucose in pristine and non-pristine samples. (A), Sunflower plot comparing glycated albumin (%) and HbA_1c (%) in pristine samples; (B), sunflower plot comparing glycated albumin (%) and HbA_1c (%) in non-pristine samples; (C), sunflower plot comparing glycated albumin (%) and fasting plasma glucose (mg/dL) in pristine samples; (D), sunflower plot comparing glycated albumin (%) and fasting plasma glucose (mg/dL) in non-pristine samples.

DISCUSSION AND CONCLUSION

Using the Asahi Kasei assay, glycated albumin measured in non-pristine stored samples that had undergone multiple freeze–thaw cycles was unexpectedly and substantially higher than in pristine samples. Serum albumin measured as part of the Asahi Kasei assay was slightly higher than the original serum albumin (Roche/Beckman) measurements in the pristine samples but highly correlated. The weak correlation between glycated albumin (%) and HbA_1c and fasting plasma glucose in the non-pristine samples strongly suggests a lack of validity of glycated albumin results in non-pristine samples. Cystatin C in the pristine and non-pristine samples were similar to prior measurements conducted in 2006.

Our results demonstrated that measurements of glycated albumin from the non-pristine NHANES 1999–2004 stored samples are not valid and should not be used in epidemiologic analyses of these data. We recommend limiting analyses of NHANES 1999–2004 glycated albumin to measurements obtained in the pristine samples. The reasonable stability of albumin and cystatin C in the pristine and non-pristine samples indicates that this issue was biomarker-specific.

Prior studies have found glycated albumin in serum to be stable under common laboratory conditions (12, 13). Sample stability studies conducted by the manufacturer demonstrated that glycated albumin measured in serum was stable for eight days at 2 to 8 °C and for at least 6 freeze–thaw cycles when stored at −80 °C (14). Prior studies have demonstrated the validity and reliability of glycated albumin measured in serum samples stored for up to 23 years (12). The poor performance of the glycated albumin measurements in the NHANES non-pristine samples is presumably due to unknown issues related to sampling handling, storage, and/or temperature. Samples not stored at the proper deep freeze temperature (i.e., −80 °C) will undergo glycation of albumin at temperatures above the freezing point of −30 °C (15). While the sample would appear frozen at −20 °C, for example, it would likely have nonfreezing water on the surface, allowing the glycation reaction to proceed and resulting in an increase in the concentration of glycated albumin. Since glycated albumin results from a nonenzymatic chemical reaction rather than a measurement of protein alone (i.e., serum albumin), the assay may be more sensitive to storage conditions and sample handling as compared to other laboratory tests such as cystatin C. Investigators measuring glycated albumin in stored samples should be aware of potential issues following multiple freeze–thaw cycles.

In summary, we found that the performance of the glycated albumin assay in the NHANES non-pristine samples was problematic. Results in pristine samples appeared highly valid, with robust correlations with HbA_1c and fasting glucose. Analyses of glycated albumin in NHANES 1999–2004 should be limited to pristine samples only. These results have major implications for the use of this large, publicly available dataset.

IMPACT STATEMENT.

Glycated albumin is of growing interest as a biomarker of diabetes control and was cleared for clinical use by the FDA in 2020. We measured glycated albumin in stored serum samples from the nationally representative NHANES. Most samples were pristine (had never undergone a freeze–thaw cycle), but some were non-pristine. We demonstrated that the performance of glycated albumin was excellent in pristine samples but problematic in non-pristine samples. Our results have major implications for the use of glycated albumin data in NHANES, a publicly available data set. Analyses of glycated albumin in NHANES 1999–2004 should be limited to pristine samples only.