Glycated albumin: An old yet new biomarker in people with diabetes and the potential for novel applications

ABSTRACT

Glycated albumin (GA), the product of nonenzymatic glycation of serum albumin, reflects the mean glucose levels over the preceding approximately 2 weeks, bridging the gap between daily self‐monitoring of blood glucose and the roughly 3‐month window covered by glycated hemoglobin (HbA1c) measurement. First quantified in the early 1980s, GA was long overshadowed by HbA1c, but at present, interest has revived for two reasons. First, HbA1c assessment proves unreliable when the erythrocyte lifespan or hemoglobin glycation is perturbed, as in patients with renal failure, hemolytic anemia, hemoglobin variants, and iron deficiency. Second, continuous glucose monitoring has drawn attention to postprandial excursions and shortterm glucose variability that are not captured by HbA1c. This review revisits GA's biology, measurement, standardization, clinical utility, limitations, and evidence linking it to diabetic complications. Because it responds within 2 weeks to treatment and is markedly influenced by postprandial glucose excursions, GA supports early monitoring of therapy, and its ratio to HbA1c serves as an index of postprandial glycemia and insulin secretion. However, it must also be recognized that conditions that alter albumin turnover change GA levels independently of blood glucose levels. The prognostic value of GA is strongest in patients on maintenance hemodialysis, although recent studies in various cohorts and a meta‐analysis have expanded situations in which GA measurement is of prognostic value. New measurement approaches, including finger‐prick sampling, weekly home measurement, and noninvasive saliva and tear assays, point toward a renewed role for GA measurement in everyday glucose monitoring in individuals with diabetes.

INTRODUCTION: WHY GLYCATED ALBUMIN NOW?

For more than four decades, hemoglobin A1c (HbA1c) has been the cornerstone of glycemia monitoring in people with diabetes, and its close relationship with the risk of microvascular complications demonstrated in landmark prospective trials underpins current treatment targets ¹ . Nevertheless, the prognostic value of HbA1c rests on two assumptions that are violated in many clinically important situations: that the erythrocyte lifespan is normal at about 120 days, and that the rate of hemoglobin glycation is not perturbed by red cell or iron metabolism. Neither of these assumptions holds across all people with diabetes.

At the same time, continuous glucose monitoring (CGM) has illuminated features of glycemia, including postprandial excursions, day‐to‐day variability, and nocturnal hypoglycemia episodes that HbA1c is designed to average away. This has revived interest in biochemical markers that would capture shorter term or different dimensions of glycemia and invited a fresh look at where each marker ranks in practice, a renewed interest that is reflected in the multiple recent reviews on glycated albumin (GA) ² , ³ .

GA was first quantified as a clinical glycemia index in the early 1980s ⁴ , ⁵ , and early comparative studies confirmed its value as a marker of shorter term glycemia control ⁶ , ⁷ . The absence of a simple, reproducible assay limited its uptake until an enzymatic method that allows routine quantification without prior separation of albumin was introduced in 2002 ⁸ , ⁹ . Adoption of GA measurement was particularly rapid in Japan, where measurement of serum GA is now an established, reimbursed laboratory test.

Our research group has approached GA not only as a substitute parameter for HbA1c, but as a marker that can be measured in new ways and used for new purposes. We have developed assays that quantify GA in self‐sampled finger‐prick blood ¹⁰ and non‐invasively in saliva and tears ¹¹ , ¹² . We have demonstrated in randomized studies that frequent GA measurement with rapid feedback can drive behavioral changes in people with type 2 diabetes ¹³ , ¹⁴ . In this review, we revisit what GA is, where it came from, how it is measured, and what clinical evidence is available, before considering where GA might contribute best in the future.

BIOLOGICAL AND PHYSIOLOGICAL BASIS

The chemistry of albumin glycation

Like HbA1c, GA is a product of the nonenzymatic Maillard reaction. Circulating glucose reacts with free epsilon‐amino groups of lysine residues on albumin to form an unstable Schiff base, which undergoes Amadori rearrangement to a stable ketoamine (fructosyl‐lysine) adduct. Several lysine residues are glycated in vivo; Lys‐525 is the principal site, accounting for roughly one‐third of the total glycation, followed by glycation at Lys‐199, Lys‐281, and Lys‐439 ¹⁵ , which recent mass spectrometry of plasma and tear albumin has reconfirmed ¹⁶ . Because the adducts are reported as a percentage of the total albumin, the value is partly self‐correcting for the prevailing albumin concentration. Albumin has a plasma half‐life of about 17 days, against that of roughly 120 days for erythrocytes, so that the serum level of GA (hereinafter simply, GA) reflects the mean glucose levels over the preceding approximately 2 weeks (Table 1).

Table 1.

Characteristics of the main glycemic measures: the period each reflects, its sensitivity to postprandial changes, and conditions that produce spurious values

Characteristic	SMBG/CGM	Glycated albumin (GA)	Hemoglobin A1c (HbA1c)
What it reflects	Blood or interstitial glucose measured directly in real time	Fraction of serum albumin that is glycated (normalized for the serum albumin concentration) 17	Fraction of hemoglobin that is glycated
Time window reflected 7 , 18 , 19	Minutes to hours (real time)	Approximately 2 weeks	Approximately 2–3 months (weighted toward recent glucose levels)
Sensitivity to postprandial peaks	Captured directly	More sensitive to postprandial peaks than HbA1c 20 , 21 , 22	Low (excursions averaged out)
Assay standardization	Point‐of‐care; device‐dependent	Comparatively well standardized, but no international standardization yet 23 , 24 , 25	Internationally standardized (IFCC, NGSP)
Conditions in which the glycemia level is overestimated (spuriously high)	—	Hypothyroidism 26 , 27 , 28	Iron deficiency, incl. in late pregnancy 29 , 30 , 31 Glycation‐enhancing Hb variant, for example, Hb Himeji 32
Conditions in which the glycemia level is underestimated (spuriously low)	—	Nephrotic syndrome/heavy proteinuria 33 , 34 Chronic liver disease 35 , 36 , 37 , 38 Obesity or high visceral fat 39 , 40 , 41 , 42 , 43 Hyperthyroidism 26 Smoking, hyperuricemia, hepatic steatosis 44 , 45 , 46 , 47	Renal failure/hemodialysis 48 , 49 , 50 , 51 , 52 , 53 , 54 , 55 , 56 , 57 Erythropoiesis‐stimulating agents, recovering anemia 49 , 52 Hemolytic anemia/after transfusion 58 , 59 Glycation‐impairing Hb variant 60 , 61 Neonates (fetal hemoglobin) 62 , 63 , 64
Conditions for which it is especially useful/preferred	Real‐time control; detecting hypoglycemia and day‐to‐day variability	When HbA1c is unreliable (patients with renal failure or on dialysis, anemia, Hb variants, iron deficiency, or pregnancy 29 , 65 , neonates, etc.) Early evaluation of treatment response and postprandial or acute change Post‐gastrectomy 66	Long‐term control and complication risk assessment (a validated surrogate) 1

The time window, kinetics, and limits of conversion formulas

The kinetics of the GA response to changing glucose levels were first modeled in experimental diabetes ⁵ , ⁶⁷ and then analyzed in people with diabetes. In a classic kinetic study, GA fell with a half‐time of about 17 days, and its weighting function extended back roughly 40 days, against about 100 days for HbA1c ⁷ . Consistent with this, GA declined measurably within 2 weeks of starting treatment, well before any change in HbA1c can be detected ⁶⁸ . This kinetic advantage is the basis of using GA for early monitoring of the response to therapy (Section “Clinical utility: When does GA provide additional information?”).

A related point is the conversion formulas that link GA to HbA1c, described in a comprehensive review by Koga and Kasayama ¹⁹ . The most widely used relationship, HbA1c (%) = 1.73 + 0.245 × GA (%), was derived under stable glycemic conditions with a measurement error model ⁶⁹ . Its existence is the very reason GA has often been treated merely as a surrogate. The formula holds, however, only at steady state. Under conditions when blood glucose levels rise quickly, as at the onset of fulminant type 1 diabetes, or fall quickly, as at the start of intensive insulin therapy, GA and HbA1c diverge because they reflect average glucose levels over different periods. Applying the conversion in these dynamic states introduces systematic error, and the two markers must then be read independently. GA is interchangeable with HbA1c only when the blood glucose is stable, and it is precisely when glucose is changing that GA is most useful.

Sensitivity to postprandial excursions

GA is influenced to a greater degree by postprandial glucose excursions than HbA1c: studies pairing these markers with CGM show that GA is more closely correlated than HbA1c with indices of glycemic variability, including the mean amplitude of glycemic excursions, the standard deviation of blood glucose, and postprandial plasma glucose ²⁰ , ²¹ , ²² . This selective sensitivity is what makes the GA:HbA1c ratio informative about the postprandial contribution to overall glycemia (Section “The GA:HbA1c ratio as a derived index”).

MEASUREMENT AND STANDARDIZATION

From chromatography to enzymatic assay

GA measurement has advanced through several generations. Early approaches quantified fructosyl‐lysine as furosine after acid hydrolysis ⁴ , or separated glycated from non‐GA by ion exchange and boronate‐affinity high‐performance liquid chromatography (HPLC) ⁷⁰ , ⁷¹ ; however, these methods were labor‐intensive and poorly suited to routine throughput. The decisive advance came with the introduction of a fully enzymatic assay ⁸ , ⁹ : an albumin‐specific proteinase releases glycated amino acids, which a ketoamine oxidase converts to measurable hydrogen peroxide, while total albumin is measured in parallel, allowing GA to be calculated without separating albumin beforehand. This chemistry underlies commercial reagent systems such as the Lucica GA‐L kit ⁷² , and a dry‐chemistry point‐of‐care format based on the same principle has been described ⁷³ . Use of this enzymatic assay has become widespread because it is rapid, runs on conventional analyzers, and works on both serum and plasma ³ .

Reference ranges

Reference values for GA have been established in large populations. GA was measured in approximately 3.14 million Japanese blood donors during 2009–2010, yielding sex‐ and age‐specific distributions with an adult upper reference limit near 16% ⁷⁴ . A comparable range was also found in healthy American volunteers ⁷⁵ , suggesting applicability across populations. GA in healthy infants is substantially lower than that in healthy adults and rises with age, correlating more strongly with the logarithm of age and with serum albumin than with plasma glucose ⁶² ; age‐appropriate reference ranges are therefore essential in neonates and infants.

Standardization: Current status and challenges

In Japan, the enzymatic assay is embedded in routine laboratory work and covered by national health insurance. Internationally, interlaboratory and inter‐method variability remain issues that wait to be addressed ²³ , ²⁴ . By contrast, HbA1c has undergone decades of international standardization (IFCC, NGSP) that set a benchmark, which has not yet been established for GA. No equivalent global standardization program for GA exists at present, which limits direct comparison across assay systems and is an important caveat when GA‐based evidence is generalized across countries. Among the intermediate‐term markers, GA measurement is better standardized than fructosamine measurement and is less vulnerable to preanalytical variables ²⁵ , although certified reference materials and calibrators will still be needed for international harmonization.

CLINICAL UTILITY: WHEN DOES GA PROVIDE ADDITIONAL INFORMATION?

Early monitoring of the response to treatment

The defining practical advantage of GA is that it responds to any change in therapy within 2 weeks. In people with diabetes with marked hyperglycemia in whom blood glucose levels are brought under control with intensive insulin therapy, GA falls substantially within weeks while the HbA1c remains largely unchanged at first ⁶⁸ . Koga and colleagues showed that early change of GA reliably predicts the HbA1c response at 3 months ⁷⁶ , an approach also validated for sitagliptin ⁷⁷ , glimepiride ⁷⁸ , and treatment intensification in hospitalized people with diabetes ⁷⁹ . Clinicians can therefore adjust treatment sooner rather than have to wait for 3 months to confirm an HbA1c response. GA also detects relapse earlier: after an inpatient diabetes education program, GA rose at the point of deterioration while HbA1c was still drifting downward because of its longer lag ⁸⁰ . These uses of GA are summarized in a clinical review by Koga and Kasayama ¹⁹ .

Acute elevations of blood glucose

Because GA responds to changes in blood glucose within 2 weeks (Section “Early monitoring of the response to treatment”), it rises quickly whenever glucose climbs acutely, for example, after medication nonadherence or dose reduction, or at the new onset of hyperglycemia, while HbA1c still reflects the average glucose levels over the preceding 2–3months. Fulminant type 1 diabetes is an extreme case: blood glucose rises explosively over days, so that GA climbs rapidly while HbA1c stays near normal. The GA:HbA1c ratio is therefore markedly elevated at the time of diagnosis of fulminant type 1 diabetes ⁸¹ , ⁸² , with a milder pattern at the onset of autoimmune acute‐onset type 1 diabetes ⁸³ ; a high GA:HbA1c ratio at presentation signals very recent hyperglycemia and helps distinguish fulminant type 1 diabetes from type 2 diabetes ⁸¹ .

Postprandial hyperglycemia

GA is influenced to a greater degree by postprandial glucose excursions than HbA1c (Section Sensitivity to postprandial excursions), so that it is especially informative when postprandial hyperglycemia predominates. After gastrectomy, rapid gastric emptying produces pronounced postprandial spikes, which result in disproportionate elevation of GA as compared with HbA1c, making it the preferred index for monitoring in this population subset ⁶⁶ .

Pregnancy and iron deficiency

HbA1c is confounded in pregnancy by hemodilution, altered erythrocyte turnover, and iron deficiency. In late pregnancy, HbA1c, but not GA, is significantly elevated in iron‐deficient women even in the absence of diabetes ²⁹ , and the same artifact affects women with iron deficiency anemia and diabetes ³⁰ . Thus, iron deficiency itself, common in premenopausal women, causes elevation of HbA1c independently of glycemia, while leaving GA unaffected ³¹ . Reference intervals for GA across a normal pregnancy term in Japanese women have been defined ⁶⁵ , providing the normative basis for using GA for monitoring women with gestational diabetes.

Neonates, infants, and neonatal diabetes

Fetal hemoglobin persists in the neonatal period and interferes with standard HbA1c assays, so that HbA1c is generally uninterpretable in neonates regardless of the glycemia status, whereas GA measurement in cord blood is feasible and interpretable ⁶³ . In neonatal diabetes caused by KCNJ11 mutations, GA tracked glucose and responded appropriately to treatment while HbA1c did not ⁸⁴ . GA is lower relative to glycemia in neonates and infants with diabetes than in older age groups, partly because the serum albumin is lower, so that the interpretation must be age‐adjusted ⁶² , ⁶⁴ .

Diagnosis and population screening

GA has been evaluated as a diagnostic test, with a cutoff of about 15.5% yielding good sensitivity and specificity in large Chinese ⁸⁵ and Japanese ⁸⁶ population studies. Combining GA with fasting plasma glucose improved the detection rate of undiagnosed diabetes beyond that using fasting glucose alone ⁸⁷ , largely by capturing postprandial hyperglycemia excursions that fasting glucose misses. The enzymatic platform makes such screening practical at scale ⁷² .

PITFALLS: CONDITIONS THAT DISTORT INTERPRETATION

GA reflects the glycemia status faithfully only when albumin metabolism is normal. Conditions that accelerate albumin catabolism, increase albumin loss, or slow albumin synthesis change GA independently of blood glucose, while those that shorten the erythrocyte lifespan or alter hemoglobin glycation invalidate HbA1c instead. Knowing which marker is distorted, and in which direction, is the basis for rational marker selection; the contrasting patterns are summarized in Table 1.

Conditions that are associated with falsely reduced serum GA values

Nephrotic syndrome and heavy proteinuria lower the GA relative to the degree of glycemia by accelerating albumin turnover and then shortening the glycation exposure time of each albumin molecule. Below the nephrotic range of proteinuria, near 3.5 g per day, the effect is small ³³ , but over this range, GA underestimates glycemia and the GA:HbA1c ratio falls in the overt nephropathy stage ³⁴ . In chronic liver disease, the altered turnover of albumin and erythrocytes affects GA and HbA1c levels, respectively ³⁵ , ³⁶ , ³⁷ , ³⁸ . Obesity independently lowers GA but not the HbA1c in both people with ³⁹ , ⁴⁰ and without ⁴¹ , ⁴² diabetes, and GA is inversely related to the fat mass and visceral adipose tissue even in people with normal glucose tolerance ⁴³ . Hyperthyroidism accelerates albumin catabolism and lowers GA ²⁶ , while each of smoking ⁴⁴ , ⁴⁵ , hyperuricemia ⁴⁶ , and raised serum aminotransferase levels with hepatic steatosis ⁴⁷ is associated with relatively lower GA, plausibly through inflammation‐driven catabolism.

Conditions that are associated with falsely elevated serum GA values

Hypothyroidism slows protein catabolism, prolongs albumin half‐life, and allows GA to accumulate independently of glucose. GA is spuriously elevated in overt hypothyroidism and falls with thyroid hormone replacement ²⁶ , ²⁷ . In one report, a person with type 2 diabetes who became hypothyroid from excessive seaweed intake had a GA of 26.9% with a discordantly low HbA1c, with this discrepancy prompting the thyroid diagnosis ²⁸ .

Conditions in which HbA1c is unreliable and GA is preferred

In patients with renal failure or on maintenance hemodialysis, uremia shortens the erythrocyte lifespan, anemia reduces the pool of older glycated red cells, and erythropoiesis‐stimulating agents speed red cell turnover, all of which lower HbA1c as measurement artifacts. The magnitude of shortening of the erythrocyte lifespan by uremia can be substantial: directly measured erythrocyte lifespan falls from a mean of about 126 days in good health to about 29 days in cases with hemolytic anemia ⁵⁸ . HbA1c underestimates the average blood glucose in people with diabetes on hemodialysis, while GA reflects it more accurately ⁴⁸ , ⁴⁹ , ⁵⁰ , ⁵¹ , ⁵² , ⁵³ , ⁵⁴ , ⁵⁵ . The same holds in pre‐dialysis stage 4–5 chronic kidney disease ⁵⁶ , ⁵⁷ . In patients with hemolytic anemia and posttransfusion states, shortened red cell survival lowers the HbA1c while GA remains unaffected, so that the GA:HbA1c ratio is higher in people with diabetes who are not on dialysis and have anemia than in those without anemia ⁵⁹ . Hemoglobin variants can distort HbA1c in either direction, and the HbA1c:GA ratio helps classify them; glycation‐enhancing variants such as Hb Himeji are associated with falsely high HbA1c values with a low GA:HbA1c ratio ³² , while glycation‐impairing variants are associated with falsely reduced HbA1c levels with a high GA:HbA1c ratio ⁶⁰ , ⁶¹ . Finally, iron deficiency raises HbA1c, but not GA (Section Pregnancy and iron deficiency), so that GA is more reliable in iron‐deficient pregnant and premenopausal women ²⁹ , ³⁰ , ³¹ .

COMPARISON WITH OTHER MARKERS AND THE GA:HbA1c RATIO

GA vs fructosamine and 1,5‐anhydroglucitol

Among the intermediate‐term markers, fructosamine reflects total glycated serum protein over the preceding roughly 1–2 weeks and is correlated more closely with the GA than with the HbA1c ¹⁷ , ⁸⁸ ; on the other hand, GA is specific to albumin, and being expressed as a percentage of the total serum albumin, is normalized to the albumin concentration ¹⁷ ; a recent appraisal judged the overall diagnostic efficiency of GA exceeds that of fructosamine across many settings ²⁵ . 1,5‐anhydroglucitol falls once blood glucose exceeds the renal threshold of about 180 mg/dL and glycosuria develops ⁸⁸ , but it is confounded by alpha‐glucosidase inhibitors ⁸⁹ , and unlike GA, was found to be not independently associated with incident diabetes after adjustment for HbA1c ⁹⁰ .

GA vs CGM‐derived metrics

CGM‐derived metrics such as time in range, the glucose management indicator, mean glucose, and variability represent different, real‐time metrics complementary to GA. CGM has also prompted more demanding goals such as time in tight range, proportion of the day with blood glucose in the range of 70–140 mg/dL ⁹¹ ; whether GA, with its short‐term, postprandial‐weighted window, can serve as a biochemical correlate of such targets is an emerging question: early evidence in type 2 diabetes has shown that GA is correlated inversely with time in tight range (TITR), with a GA of about 17.4% identifying a TITR above 50% in one study ⁹² ; however, data remain limited. GA has also been shown to be correlated significantly with CGM‐derived mean glucose and the amplitude of glycemic excursions ²¹ , ²² , ⁸⁸ , although the agreement is imperfect, because CGM samples glucose changes in days while GA integrates the levels over weeks. As CGM expands in people with type 2 diabetes, including those on basal insulin ⁹³ , ⁹⁴ , GA has been shown to be correlated more closely with the time in range than HbA1c or fructosamine and to reach a predictive level earlier ⁹⁵ . Because cost, burden, and access still limit the use of CGM ⁹⁶ , GA is especially valuable as a biochemical stand‐in for time in range in people who do not or cannot use CGM.

The GA:HbA1c ratio as a derived index

The GA:HbA1c ratio is higher in type 1 than in type 2 diabetes (about 3.2 vs 2.9) ⁹⁷ , ⁹⁸ , and is chiefly useful for flagging confounders discussed in Section Pitfalls: conditions that distort interpretation: it increases under conditions that lower HbA1c relative to GA (hemolysis, erythropoiesis‐stimulating agents, impaired‐glycation hemoglobin variants) and decreases under conditions that lower GA relative to HbA1c (obesity, hepatic steatosis, nephrotic syndrome), so an unexpectedly discordant ratio should prompt a search for these conditions. Cross‐sectional data also suggest this ratio as an index of postprandial glycemia and insulin secretory function, since it is directly correlated with postprandial glucose and inversely correlated with insulin secretory capacity ⁹⁸ , ⁹⁹ , ¹⁰⁰ , ¹⁰¹ ; as the ratio is derived and influenced by albumin metabolism, it warrants the same caution as GA itself.

ASSOCIATION WITH COMPLICATIONS AND OUTCOMES

The shape of the evidence

The link between HbA1c and microvascular complications rests on interventional evidence from the DCCT and UKPDS, which showed that lowering of HbA1c was associated with a reduced risk of retinopathy, nephropathy, and neuropathy ¹ , ¹⁰² ; this is what makes HbA1c a validated surrogate endpoint. There are no equivalent interventional trials for GA, for which all available evidence at present is observational, mostly from cross‐sectional and cohort studies with the usual risk of residual confounding. This does not, however, make GA inferior to HbA1c for assessing complications: where HbA1c is itself biased, notably in patients on maintenance dialysis, GA can provide better prognostic information than the established HbA1c. What is plainly lacking for GA is trial evidence that acting on the serum GA levels changes outcomes. This section summarizes the observational evidence linking GA to diabetic complications and mortality.

Microvascular complications

GA has been demonstrated to be associated with microvascular outcomes in cross‐sectional analyses. In the Atherosclerosis Risk in Communities (ARIC) Study, GA, fructosamine, and HbA1c were each associated with albuminuria and retinopathy, and the associations for GA and fructosamine remained significant even after adjustment for HbA1c, suggesting that these parameters capture glycemia information that HbA1c does not fully capture ¹⁰³ . Community‐based data confirm these findings: in the Hisayama Study, GA and other glycemia measures were associated with the prevalence of diabetic retinopathy and helped define diagnostic glycemia thresholds ¹⁰⁴ . Interpretation across the stages of diabetic nephropathy requires care because the GA:HbA1c ratio is pushed down in the overt nephropathy stage by albuminuria and pushed up in end‐stage renal disease by the greater HbA1c bias from uremia ³⁴ .

Macrovascular disease and mortality

Many observational studies have linked serum GA to the presence/absence and severity of coronary artery disease in people with type 2 diabetes ¹⁰⁵ , ¹⁰⁶ , ¹⁰⁷ , ¹⁰⁸ , ¹⁰⁹ . Among 829 people who underwent coronary angiography, GA was more closely correlated with the number and severity of significant stenoses than HbA1c ¹¹⁰ , consistent with the contribution of short‐term and postprandial glycemia excursions to the plaque burden. GA has also been shown to predict progression of carotid atherosclerosis ¹¹¹ , ¹¹² , ¹¹³ and in‐stent restenosis ¹¹⁴ , and studies in community‐based cohorts have extended these findings beyond selected clinical populations: in the Hisayama Study, higher serum GA was associated with incident cardiovascular disease ¹¹⁵ and with carotid atherosclerosis ¹¹⁶ . These associations are also biologically plausible: in vitro, GA has been shown to activate pro‐atherogenic and pro‐fibrotic signaling in vascular ¹¹⁷ , ¹¹⁸ , endothelial ¹¹⁹ , ¹²⁰ , ¹²¹ , ¹²² , macrophage ¹²³ , and mesangial cells ¹²⁴ , ¹²⁵ , and to be cytotoxic to retinal pericytes ¹²⁶ , ¹²⁷ . Serum GA has been shown to be associated with mortality as well: in older adults in the ARIC study, HbA1c, fructosamine, GA, and fasting glucose were each associated with mortality, with HbA1c retaining the clearest prognostic value ¹²⁸ , and in the United States general population, serum GA was shown to be comparably associated with all‐cause and cardiovascular mortality to HbA1c ¹²⁹ ; furthermore, a higher GA:HbA1c ratio was also associated with a greater mortality risk ¹³⁰ . An important caveat is that the cardiovascular studies come disproportionately from Asian cohorts, with inconsistent adjustments in the studies for albumin metabolic confounders; therefore, the causal relationship remains uncertain.

Other complications

Beyond the vascular system, GA has been linked to other diabetes‐related outcomes: in the Hisayama Study, higher GA levels predicted sarcopenia and dynapenia in older adults ¹³¹ , extending its associations beyond the classical micro‐ and macrovascular complications.

Studies in people on hemodialysis

Among the settings reviewed above, hemodialysis provides the clearest example of a situation with prospective evidence of GA predicting outcomes where HbA1c is not reliable. Because HbA1c is biased downward here, its association with the outcomes is weak, whereas GA has been demonstrated to be of consistent prognostic value across independent cohorts: GA, but not HbA1c, predicted the all‐cause mortality and cardiovascular hospitalization in 444 people with diabetes on dialysis ¹³² , predicted mortality and cardiovascular events in 503 people initiated on hemodialysis ¹³³ , predicted cardiovascular hospitalization and length of stay in one cohort ¹³⁴ , and elevated GA was shown to be predictive of the long‐term mortality in patients with end‐stage renal disease ¹³⁵ . A point–counterpoint exchange captures the ongoing debate ¹³⁶ , ¹³⁷ ; it was demonstrated in a very large cohort that HbA1c still shows a J‐shaped association with cardiovascular mortality in patients on dialysis despite its bias ¹³⁸ . A 2023 meta‐analysis of 17 cohorts covering more than 80,000 participants found that higher GA levels were associated with cardiovascular and all‐cause mortality in a dose‐responsive manner, regardless of the dialysis status ¹³⁹ . On balance, there is prospective evidence favoring the use of GA, although interventional evidence is lacking; also, the survival benefit of better control appears to be attenuated in people with preexisting cardiovascular disease ¹⁴⁰ .

What evidence is still needed

No prospective trials have used GA as a treatment target or shown that GA‐guided care improves outcomes, even in patients on dialysis, in whom the observational case is strongest. Macrovascular studies have not consistently adjusted for albumin metabolic confounders, and CGM‐era evidence is still nascent. The main confounders are nutrition, inflammation, and renal function, as described in Section Pitfalls: conditions that distort interpretation, so that outcome studies should record serum albumin, and nutritional and inflammatory markers as covariates. Future studies should classify populations by the degree of renal function impairment, use objective complication endpoints, and ideally test whether acting on GA changes them. Until then, GA should inform decisions with clear awareness of both its strengths and the current limits of available evidence for its usefulness.

TOWARD CLINICAL IMPLEMENTATION AND THE VALUE OF GA FOR MONITORING

A pragmatic framework for when to add GA monitoring

The preceding sections suggest two broad circumstances in which GA provides more practical information than HbA1c. The first is when HbA1c is unreliable as a marker of the glycemia status, for example, in patients with renal failure and those on dialysis, and patients with anemia, hemoglobin variants, or iron deficiency, and pregnant women. The second is when monitoring of short‐term or postprandial changes is especially relevant, as at the start of treatment or in people with acute onset of hyperglycemia. Recent reviews converge on positioning GA as an attractive alternative where HbA1c measurement provides biased or unreliable results ² , ³ , ²⁵ .

Rather than providing a fixed algorithm for introducing GA whenever a given condition is present, we prefer to present this framework as a support for decision‐making. A rigid algorithm is probably premature at this time. Although current guidelines acknowledge GA as an alternative glycemia biomarker, routine glycemia monitoring frameworks still center on HbA1c and CGM ¹⁴¹ , ¹⁴² . The evidence for GA‐specific population thresholds and monitoring intervals also remains incomplete. Two observations favor a responsive use of markers: in Japan, while the delivery of recommended care has improved, the proportion of people reaching treatment targets remains suboptimal ¹⁴³ ; furthermore, HbA1c‐target achievement varies by season, being lowest in winter ¹⁴⁴ . Both underscore the need for a marker that reflects recent changes quickly enough to prompt timely action; however, GA is best framed as a complement to HbA1c rather than a replacement at present.

GA as a driver of lifestyle behavior changes

Tight glycemia control reduces the risk of complications in people with diabetes, but the benefit is preventive and often distant: the macrovascular payoff may take years to emerge ¹⁰² and mixed results are obtained even with intensified multifactorial intervention ¹⁴⁵ . However, the lifestyle changes required to maintain this control are often difficult to sustain ¹⁴⁶ , ¹⁴⁷ . This is where the time window covered by GA measurement becomes relevant in a new way. HbA1c, which reflects the average blood glucose levels of the previous 2–3 months, responds too slowly to serve as feedback on recent lifestyle behaviors. CGM reflects these behaviors in real time and improves glycemia control in people with type 2 diabetes ⁹³ , ⁹⁴ , ¹⁴⁸ , but its usefulness is limited by the cost, burden, and/or access ⁹⁶ ; self‐monitoring of blood glucose requires frequent finger pricks ¹⁴⁹ . Use of digital tools such as smartphone applications and telemedicine can improve the usefulness of HbA1c modestly ¹⁵⁰ , ¹⁵¹ , ¹⁵² . Measurement of GA, on the other hand, which reflects the average blood glucose levels of the previous 1–2 weeks, sits between the other two options: results are obtained fast enough to reflect the effects of recent lifestyle behaviors and require only periodic sampling.

We tested this idea directly. In a randomized controlled study, bi‐weekly GA measurement with same‐day feedback over 8 weeks induced behavioral changes and led to reductions in body weight and fat mass, and the rising GA trend seen in the control group was absent in the intervention group ¹³ . In another randomized controlled study, once‐weekly home GA measurement, with the results delivered through a smartphone application, combined with daily self‐review of lifestyle behaviors, significantly reduced the GA and HbA1c levels, as well as body weight and waist circumference ¹⁴ . GA can therefore function as a behavioral feedback marker, besides as a diagnostic/monitoring index. Because the stigma around having diabetes can undermine self‐care ¹⁵³ , ¹⁵⁴ , ¹⁵⁵ , ¹⁵⁶ , low‐burden care/monitoring that provides timely feedback may help sustain engagement, consistent with guidelines on positive health behaviors ¹⁵⁷ , although this remains to be tested.

Weekly GA by mail and the logistics of home testing

A practical route to frequent monitoring is home sampling combined with postal delivery. We have established an HPLC assay that measures GA in self‐sampled finger‐prick blood, with the results agreeing closely with venous enzymatic measurement (R‐squared 0.988, slope near 1.0); the samples remain stable for several days at ambient temperature, and the levels in mailed samples correlated almost perfectly with those stored in the laboratory (R‐squared above 0.99) ¹⁰ . This allows weekly GA measurements without visits to the clinic that can fit in with the rhythm of lifestyle adjustments and support telemedicine. It requires only an accurate enough assay method to detect week‐to‐week change and a logistics chain that preserves sample integrity.

Complementarity with CGM

GA and CGM complement rather than compete with each other. Where both are available, each covers the gaps of the other, and because GA does not depend on the hemoglobin level, it stays valid even when the erythrocyte lifespan is shortened ⁵⁸ . Just as importantly, CGM is still out of reach for many people because of cost, burden, and/or supply, and for such people, GA delivers much of the short‐term, postprandial‐weighted information that CGM provides from a single blood sample. Thus, GA can serve as a biochemical surrogate for estimating the time in range when CGM cannot be used or when it is too early in the treatment course to rely on HbA1c ⁹⁵ .

TOWARD LESS INVASIVE GA MEASUREMENT

Rationale

Even finger‐prick sampling is invasive and can deter frequent testing. Use of noninvasive samples, principally saliva and tears, collected without a needle prick for measurement of GA would lower the burden further on the individual and make truly frequent monitoring realistic.

Salivary and tear GA

GA can be detected in both saliva and tears. Using liquid chromatography with mass spectrometry, we found that the GA levels in tears and saliva were correlated significantly with the blood GA levels, with the correlations remaining strong even after adjustments for the body mass index, age, and nephropathy stage ¹¹ . Subsequently, we measured salivary GA by HPLC after antibody‐based purification and found that it was strongly correlated with the blood GA levels in both fasting and postprandial samples, suggesting that salivary GA could eventually become a substitute for blood GA and serve as a promising candidate for home testing ¹² . Independent work using a small‐volume nano‐liquid chromatography platform confirmed that the tear GA levels were also closely correlated with the plasma GA levels. Using the well‐established Lys‐525 as the primary quantitation site, the authors proposed Lys‐136 and Lys‐137 as alternative glycation sites, which could be useful for future assay design ¹⁶ .

Advantages and challenges

The advantages are clear: saliva or tear sample collection is painless and frequently repeatable, potentially reducing the burden and cost for the individual. The challenges are equally clear. Albumin concentrations in saliva and tears are far lower than those in the blood, and natural tear films provide only a limited sample volume (1–5 μL). Therefore, reliable measurement of GA in tears/saliva requires the use of highly sensitive platforms—such as HPLC or nano‐liquid chromatography coupled with mass spectrometry. Furthermore, a preanalytical purification and concentration step is often essential, not only to achieve sufficient sensitivity but also to ensure that the analyte falls within the quantitative dynamic range of the assay ¹² , ¹⁶ . The oral environment, including the recent food intake and oral health status, may influence measurements of GA in saliva samples, and available sample volumes are very small in tear sample measurements of GA ¹⁶ . None of these approaches are yet standardized, and the studies conducted so far have been exploratory and small. The decisive evidence is still missing; namely, confirmation by a prospective controlled study that regular noninvasive GA measurement is a cost‐effective method for improving glycemia control ¹¹ , ¹² .

CONCLUSION: PROSPECTS FOR GA 2.0

GA is an old marker, as it was first quantified in the 1980s, that is undergoing renewal. Its classical roles, as an alternative to HbA1c when HbA1c is unreliable and as an index of short‐term and postprandial glycemia excursions, are now expanding with the development of new ways for its measurement and application. Finger‐prick sampling with postal delivery, weekly home measurement results read through a smartphone application, and noninvasive sampling (saliva or tear samples) have changed GA from a marker measured a few times a year into one that can almost be followed up as a vital sign. GA is better correlated with the time in range than HbA1c ⁹⁵ ; it can predict cardiovascular events and mortality across dialysis as well as non‐dialysis populations ¹³⁹ ; and it can drive behavior changes when used to provide frequent feedback ¹³ , ¹⁴ .

Based on these developments, we envision a course of care in which people with diabetes can check their glycemia status each week at home (Figure 1), without a needle, and receive fast enough feedback to connect the results to the previous week's meals and activities, with GA measurement serving as the biochemical companion to CGM and digital self‐management tools ¹⁵⁰ . The challenges are also equally clear and surmounting them will require more than enthusiasm. International standardization of GA, still lagging behind that of HbA1c, is a prerequisite for comparable results, and prospective trials using GA as a target with hard outcomes are needed to move beyond observational evidence. The noninvasive methods must be standardized and validated, and GA measurement must move from an acknowledged alternative for evaluating the glycemia status to an integral part of monitoring frameworks that would still center on HbA1c and CGM ¹⁴¹ , ¹⁴² . With these steps, the same molecule that has been measured for nearly 40 years could support development of a genuinely new model for diabetes monitoring: the glimmering promise of GA 2.0, an old biomarker used in new ways.

Figure 1.

The glycated albumin (GA) 2.0 concept. Low‐burden sampling (finger‐prick blood, saliva, or tears) enables weekly home GA measurement by high‐performance liquid chromatography or enzymatic assay, with results returned by post or smartphone application. Rapid feedback over 1–2 weeks facilitates behavior change and complements continuous glucose monitoring and HbA1c, with the aim of better and more sustainable glycemia control.

FUNDING INFORMATION

No funding was received for this research.

DISCLOSURE

The authors declare no other conflicts of interest.

APPROVAL OF THE RESEARCH PROTOCOL, INFORMED CONSENT, REGISTRY AND THE REGISTRATION NO. OF THE STUDY/TRIAL, ANIMAL STUDIES

This article is a review of previously published literature and reports no new human or animal data; formal ethics approval was therefore not needed for the review itself. However, the primary studies cited were approved by the relevant institutional review boards and conducted in accordance with the principles of the Declaration of Helsinki (as revised in Fortaleza, 2013).

DATA AVAILABILITY STATEMENT

Data sharing not applicable to this article as no datasets were generated or analysed during the current study.