Diabetes was originally identified by the presence of glucose in the urine. Almost
2,500 years ago it was noticed that ants were attracted to the urine of some individuals.
In the 18th and 19th centuries the sweet taste of urine was used for diagnosis before
chemical methods became available to detect sugars in the urine. Tests to measure
glucose in the blood were developed over 100 years ago, and hyperglycemia subsequently
became the sole criterion recommended for the diagnosis of diabetes. Initial diagnostic
criteria relied on the response to an oral glucose challenge, while later measurement
of blood glucose in an individual who was fasting also became acceptable. The most
widely accepted glucose-based criteria for diagnosis are fasting plasma glucose (FPG)
≥126 mg/dL or a 2-h plasma glucose ≥200 mg/dL during an oral glucose tolerance test
(OGTT) on more than one occasion (1,2). In a patient with classic symptoms of diabetes,
a single random plasma glucose ≥200 mg/dL is considered diagnostic (1). Before 2010
virtually all diabetes societies recommended blood glucose analysis as the exclusive
method to diagnose diabetes. Notwithstanding these guidelines, over the last few years
many physicians have been using hemoglobin A1C to screen for and diagnose diabetes
(3). Although considered the “gold standard” for diagnosis, measurement of glucose
in the blood is subject to several limitations, many of which are not widely appreciated.
Measurement of A1C for diagnosis is appealing but has some inherent limitations. These
issues have become the focus of considerable attention with the recent publication
of the Report of the International Expert Committee that recommended the use of A1C
for diagnosis of diabetes (4), a position that has been endorsed (at the time of writing)
by the American Diabetes Association (ADA) (1), the Endocrine Society, and in a more
limited fashion by American Association of Clinical Endocrinologists/American College
of Endocrinology (5). This review will provide an overview of the factors that influence
glucose and A1C testing.
FACTORS CONTRIBUTING TO VARIATION IN RESULTS
Before addressing glucose and A1C, it is important to consider the factors that impact
the results of any blood test. While laboratory medicine journals have devoted some
discussion to the sources of variability in results of blood tests, this topic has
received little attention in the clinical literature. Factors that contribute to variation
can conveniently be divided into three categories, namely biological, preanalytical,
and analytical. Biological variation comprises both differences within a single person
(termed intraindividual) and between two or more people (termed interindividual).
Preanalytical issues pertain to the specimen before it is measured. Analytical differences
result from the measurement procedure itself. The influence of these factors on both
glucose and A1C results will be addressed in more detail below.
GLUCOSE MEASUREMENT
FPG
Measurement of glucose in plasma of fasting subjects is widely accepted as a diagnostic
criterion for diabetes (1,2). Advantages include inexpensive assays on automated instruments
that are available in most laboratories worldwide (Table 1). Nevertheless, FPG is
subject to some limitations. One report that analyzed repeated measurements from 685
fasting participants without diagnosed diabetes from the Third National Health and
Nutrition Examination Survey (NHANES III) revealed that only 70.4% of people with
FPG ≥126 mg/dL on the first test had FPG ≥126 mg/dL when analysis was repeated ∼2
weeks later (6). Numerous factors may contribute to this lack of reproducibility.
These are elaborated below.
Table 1
FPG for the diagnosis of diabetes
Advantages
• Glucose assay easily automated
• Widely available
• Inexpensive
• Single sample
Disadvantages
• Patient must fast ≥8 h
• Large biological variability
• Diurnal variation
• Sample not stable
• Numerous factors alter glucose concentrations, e.g., stress, acute illness
• No harmonization of glucose testing
• Concentration varies with source of the sample (venous, capillary, or arterial
blood)
• Concentration in whole blood is different from that in plasma
• Guidelines recommend plasma, but many laboratories measure serum glucose
• FPG less tightly linked to diabetes complications (than A1C)
• Reflects glucose homeostasis at a single point in time
Biological variation
Fasting glucose concentrations vary considerably both in a single person from day
to day and also between different subjects. Intraindividual variation in a healthy
person is reported to be 5.7–8.3%, whereas interindividual variation of up to 12.5%
has been observed (6,7). Based on a CV (coefficient of variation) of 5.7%, FPG can
range from 112–140 mg/dL in an individual with an FPG of 126 mg/dL. (It is important
to realize that these values encompass the 95% confidence interval, and 5% of values
will be outside this range.)
Preanalytical variation
Numerous factors that occur before a sample is measured can influence results of blood
tests. Examples include medications, venous stasis, posture, and sample handling.
The concentration of glucose in the blood can be altered by food ingestion, prolonged
fasting, or exercise (8). It is also important that measurements are performed in
subjects in the absence of intercurrent illness, which frequently produces transient
hyperglycemia (9). Similarly, acute stress (e.g., not being able to find parking or
having to wait) can alter blood glucose concentrations.
Samples for fasting glucose analysis should be drawn after an overnight fast (no caloric
ingestion for at least 8 h), during which time the subject may consume water ad lib
(10). The requirement that the subject be fasting is a considerable practical problem
as patients are usually not fasting when they visit the doctor, and it is often inconvenient
to return for phlebotomy. For example, at an HMO affiliated with an academic medical
center, 69% (5,752 of 8,286) of eligible participants were screened for diabetes (11).
However, FPG was performed on only 3% (152) of these individuals. Ninety-five percent
(5,452) of participants were screened by random plasma glucose measurements, a technique
not consistent with ADA recommendations. In addition, blood drawn in the morning as
FPG has a diurnal variation. Analysis of 12,882 participants aged 20 years or older
in NHANES III who had no previously diagnosed diabetes revealed that mean FPG in the
morning was considerably higher than in the afternoon (12). Prevalence of diabetes
(FPG ≥126 mg/dL) in afternoon-examined patients was half that of participants examined
in the morning. Other patient-related factors that can influence the results include
food ingestion when supposed to be fasting and hypocaloric diet for a week or more
prior to testing.
Glucose concentrations decrease in the test tube by 5–7% per hour due to glycolysis
(13). Therefore, a sample with a true blood glucose value of 126 mg/dL would have
a glucose concentration of ∼110 mg/dL after 2 h at room temperature. Samples with
increased concentrations of erythrocytes, white blood cells, or platelets have even
greater rates of glycolysis. A common misconception is that sodium fluoride, an inhibitor
of glycolysis, prevents glucose consumption. While fluoride does attenuate in vitro
glycolysis, it has no effect on the rate of decline in glucose concentrations in the
first 1 to 2 h after blood is collected, and glycolysis continues for up to 4 h in
samples containing fluoride (14). The delay in the glucose stabilizing effect of fluoride
is most likely the result of glucose metabolism proximal to the fluoride target enolase
(15). After 4 h, fluoride maintains a stable glucose concentration for 72 h at room
temperature (14). A recent publication showed that acidification of the blood sample
inhibits glycolysis in the first 2 h after phlebotomy (16), but the collection tubes
used in that study are not commercially available. Placing tubes in ice water immediately
after collection may be the best method to stabilize glucose initially (2,16), but
this is not a practical solution in most clinical situations. Separating cells from
plasma within minutes is also effective, but impractical.
The nature of the specimen analyzed can have a large influence on the glucose concentration.
Glucose can be measured in whole blood, serum, or plasma, but plasma is recommended
by both the ADA and World Health Organization (WHO) for diagnosis (1,2). However,
many laboratories measure glucose in serum, and these values may differ from those
in plasma. There is a lack of consensus in the published literature, with glucose
concentrations in plasma reported to be lower than (17), higher than (16,18,19), or
the same as (20) those in serum. Importantly, glucose concentrations in whole blood
are 11% lower than those in plasma because erythrocytes have a lower water content
than plasma (13). The magnitude of the difference in glucose between whole blood and
plasma changes with hematocrit. Most devices (usually handheld meters) that measure
glucose in capillary blood use whole blood. While the majority of these report a plasma
equivalent glucose value (21), this result is not accurate in patients with anemia
(22) (unless the meter measures hematocrit).
The source of the blood is another variable. Although not a substantial problem in
the fasting state, capillary glucose concentrations can be 20–25% higher (mean of
30 mg/dL) than venous glucose during an OGTT (23). This finding has practical implications
for the OGTT, particularly because the WHO deems capillary blood samples acceptable
for the diagnosis of diabetes (2).
Analytical variation
Glucose is measured in central laboratories almost exclusively using enzymatic methods,
predominantly with glucose oxidase or hexokinase (24). The following terms are important
for understanding measurement: accuracy indicates how close a single measurement is
to the “true value” and precision (or repeatability) refers to the closeness of agreement
of repeated measurements under the same conditions. Precision is usually expressed
as CV; methods with low CV have high precision. Numerous improvements in glucose measurement
have produced low within-laboratory imprecision (CV <2.5%). Thus, the analytical variability
is considerably less than the biological variability, which is up to 8.3%. Nevertheless,
accuracy of measurement remains a problem. There is no program to standardize results
among different instruments and different laboratories. Bias (deviation of the result
from the true value) and variation among different lots of calibrators can reduce
the accuracy of glucose results. (A calibrator is a material of known concentration
that is used to adjust a measurement procedure.) A comparison of serum glucose measurements
(target value 98.5 mg/dL) was performed among ∼6,000 laboratories using 32 different
instruments (25). Analysis revealed statistically significant differences in bias
among clinical laboratory instruments, with biases ranging from −6 to +7 mg/dL (−6
to +7%) at a glucose concentration of 100 mg/dL. These considerable differences among
laboratories can result in the potential misclassification of >12% of patients (4).
Similarly, inspection of a College of American Pathologists (CAP) survey comprising
>5,000 laboratories revealed that one-third of the time the results among instruments
for an individual measurement could range between 141 and 162 mg/dL (26). This variation
of 6.9% above or below the mean reveals that one-third of the time the glucose results
on a single patient sample measured in two different laboratories could differ by
14%.
OGTT
The OGTT evaluates the efficiency of the body to metabolize glucose and for many years
has been used as the “gold standard” for diagnosis of diabetes. An increase in postprandial
glucose concentration usually occurs before fasting glucose increases. Therefore,
postprandial glucose is a sensitive indicator of the risk for developing diabetes
and an early marker of impaired glucose homeostasis (Table 2). Published evidence
suggests that an increased 2-h plasma glucose during an OGTT is a better predictor
of both all-cause mortality and cardiovascular mortality or morbidity than the FPG
(27,28). The OGTT is accepted as a diagnostic modality by the ADA, WHO/International
Diabetes Federation (IDF) (1,2), and other organizations. However, extensive patient
preparation is necessary to perform an OGTT. Important conditions include, among others,
ingestion of at least 150 g of dietary carbohydrate per day for 3 days prior to the
test, a 10- to 16-h fast, and commencement of the test between 7:00 a.m. and 9:00
a.m. (24). In addition, numerous conditions other than diabetes can influence the
OGTT (24). Consistent with this, published evidence reveals a high degree of intraindividual
variability in the OGTT, with a CV of 16.7%, which is considerably greater than the
variability for FPG (6). These factors result in poor reproducibility of the OGTT,
which has been documented in multiple studies (29,30). The lack of reproducibility,
inconvenience, and cost of the OGTT led the ADA to recommend that FPG should be the
preferred glucose-based diagnostic test (1). Note that glucose measurement in the
OGTT is also subject to all the limitations described for FPG (Table 1).
Table 2
OGTT for the diagnosis of diabetes
Advantages
• Sensitive indicator of risk of developing diabetes
• Early marker of impaired glucose homeostasis
Disadvantages
• Lacks reproducibility
• Extensive patient preparation
• Time-consuming and inconvenient for patients
• Unpalatable
• Expensive
• Influenced by numerous medications
• Subject to the same limitations as FPG, namely, sample not stable, needs to be
performed in the morning, etc.
A1C MEASUREMENT
A1C is formed by the nonenzymatic attachment of glucose to the N-terminal valine of
the β-chain of hemoglobin (24). The life span of erythrocytes is ∼120 days, and consequently
A1C reflects long-term glycemic exposure, representing the average glucose concentration
over the preceding 8–12 weeks (31,32). Both observational studies (33) and controlled
clinical trials (34,35) demonstrate strong correlation between A1C and retinopathy,
as well as other microvascular complications of diabetes. More importantly, the A1C
value predicts the risk of microvascular complications and lowering A1C concentrations
(by tight glycemic control) significantly reduces the rate of progression of microvascular
complications (34,35).
Biological variation
Intraindividual variation of A1C in nondiabetic people is minimal (36) (Table 3),
with CV <1% (37). Variability between individuals is greater. Data derived from several
investigators imply that A1C values may not be constant among all individuals despite
the presence of similar blood glucose or fructosamine concentrations (38). Some investigators
have termed this a “glycation gap” and proposed that there are differences in the
rate of glycation of hemoglobin (“low and high glycators”) (39). Studies of twins
with type 1 diabetes support a genetic contribution to A1C values (40), and heritability
of the glycation gap was observed in healthy female twins (41). However, the glycation
gap is essentially a measure of A1C adjusted for fructosamine. Importantly, measurement
of fructosamine, which is glycated albumin and protein, suffers from several limitations
(42). In addition, some authors have questioned the statistical analysis (which is
not standard) used in determining the glycation gap and noted the statistical tautology
that the outcome is correlated with the residual from a regression (43). Importantly,
the postulate of a glycation gap remains unsubstantiated by data because glycation
rates cannot be measured accurately in vivo. In addition, the hemoglobin glycation
index (difference between observed A1C and that predicted from blood glucose) is not
an independent predictor of the risk of microvascular complications (43), and the
possible clinical significance of the glycation gap is unclear.
Table 3
A1C for the diagnosis of diabetes
Advantages
• Subject need not be fasting
• Samples may be obtained any time of the day
• Very little biological variability
• Sample stable
• Not altered by acute factors, e.g., stress, exercise
• Reflects long-term blood glucose concentration
• Assay standardized across instruments
• Accuracy of the test is monitored
• Single sample, namely whole blood
• Concentration predicts the development of microvascular complications of diabetes
• Used to guide treatment
Disadvantages
• May be altered by factors other than glucose, e.g., change in erythrocyte life
span, ethnicity
• Some conditions interfere with measurement, e.g., selected hemoglobinopathies
• May not be available in some laboratories/areas of the world
• Cost
Accumulating evidence supports the hypothesis that race influences A1C. Initial studies
in patients with diabetes reported statistically significant differences in A1C concentrations
among races (44). While adjusted for factors that may influence glycemia, it remains
possible that these differences may be due to variations in glycemic control. More
compelling support was provided in NHANES III where Mexican Americans and blacks had
higher average A1C values than whites (45,46). Similar findings were observed in adults
with impaired glucose tolerance in the Diabetes Prevention Program (47) and validated
in a cross-sectional analysis of two studies (48). Collectively these data suggest
that there are differences in A1C concentrations among racial groups. However, it
is not clear that these changes have clinical significance. A1C was measured in the
Atherosclerosis Risk in Communities (ARIC) study in 11,092 adults who did not have
a history of diabetes or cardiovascular disease (49). Consistent with prior publications,
blacks had mean A1C values 0.4% higher than whites. Nevertheless, race did not modify
the association between the A1C value and adverse cardiovascular outcomes and death
(49). Because follow-up revealed that blacks with biochemically defined incident diabetes
were significantly less likely than whites to report having received a diagnosis of
diabetes by a physician, the authors speculate that delays in diagnosis may explain
the higher A1C values in blacks.
The molecular mechanism underlying the racial and ethnic differences remains to be
established. Possibilities include differences in rates of glucose uptake into erythrocytes,
rates of intraerythrocytic glucose metabolism, rates of glucose attachment to or release
from hemoglobin or erythrocyte life span (50,51). Regardless of the mechanism, the
variations in A1C concentrations are relatively small (≤0.4%), and no consensus has
been reached on whether different cutoffs should be used for different races.
Preanalytical variation
Most factors that alter FPG do not significantly affect A1C concentrations. Acute
illness, short-term lifestyle changes (e.g., exercise), recent food ingestion, and
sample handling do not significantly alter A1C values (Table 3). Importantly, whole
blood samples are stable for 1 week at 4°C and for at least 1 year at −70°C or colder
(13,52).
The interpretation of A1C depends on the erythrocytes having a normal life span. Patients
with hemolytic disease or other conditions with shortened erythrocyte survival have
a substantial reduction in A1C (53). Similarly, individuals with acute blood loss
have spuriously low A1C values because of an increased fraction of young erythrocytes.
False increases in A1C have been reported with some methods in patients with hypertriglyceridemia,
hyperbilirubinemia, uremia, chronic alcoholism, or chronic ingestion of salicylates
(13). Because most interferences are method specific, in many cases they can be overcome
by selecting an appropriate method that is not subject to the interference.
Individuals with iron deficiency anemia have increased A1C and fructosamine concentrations
(54), both of which are reduced by therapy with iron (54,55). A mechanism for the
higher A1C was recently identified by the demonstration that malondialdehyde, which
is increased in subjects with iron deficiency anemia (54), augments glycation of hemoglobin
(56). However, the magnitude of the increase in A1C is probably small. Examination
of 10,535 adults without self-reported diabetes in NHANES III revealed that while
13.7% of women had iron deficiency, only 4.74% and 0.48% had A1C ≥5.5% or ≥6.5%, respectively
(57). Iron deficiency in women was associated with a small (odds ratio 1.39) yet significant
greater odds of A1C ≥5.5% but not with greater odds of A1C ≥6.5%. Iron deficiency
was rare in men (<0.5%) (57). Nevertheless, it would seem prudent to correct the iron
deficiency before measuring A1C in individuals with severe iron deficiency anemia.
Analytical variation
There are ∼100 different methods used to measure A1C. The most widely used commercial
methods use either antibodies (immunoassays) or cation-exchange chromatography (most
commonly high-performance liquid chromatography) to separate the glycated (A1C) from
the nonglycated hemoglobin (24). The National Glycohemoglobin Standardization Program
(NGSP) has been instrumental in standardizing A1C testing among laboratories (58,59),
particularly (but not exclusively) in the U.S. The NGSP has markedly improved the
performance of A1C testing (58). At the time of writing, the vast majority (93%) of
clinical laboratories that participate in CAP surveys use methods with between-laboratory
CVs <5% (www.ngsp.org). Within laboratory CVs for some methods are as low as <0.5%.
In addition, the International Federation for Clinical Chemistry (IFCC) developed
a reference method using mass spectrometry (or capillary electrophoresis) for A1C
measurement, which should result in international harmonization as it facilitates
traceability to a metrologically sound accuracy base. It is important to emphasize
that the IFCC method is technically complex, time consuming, and expensive and is
not designed for routine analysis of patient samples.
Hemoglobin variants affect some A1C measurements. The most common variants are HbS,
HbE, HbC, and HbD. A1C measurement is not appropriate in subjects homozygous for HbS
or HbC, with HbSC or with any other variant that alters erythrocyte survival. However,
A1C can be measured accurately in individuals heterozygous for HbS, HbE, HbC, or HbD
and in those with increased HbF, provided an appropriate assay is used (53,60). Only
∼4% of the 3,378 clinical laboratories that participated in the 2010 GH2 College of
American Pathologists survey (which measures A1C) use methods in which HbAS or HbAC
has clinically significant interference. In addition, if the sample is analyzed by
high-performance liquid chromatography method, careful inspection of the chromatogram
usually reveals the aberrant peaks produced by the variant hemoglobin. The presence
of a hemoglobin variant should be considered if A1C is >15% or if a large change in
A1C coincides with a change in laboratory A1C method (53). In these situations, hemoglobin
electrophoresis should be performed. It is important to emphasize that, like any other
test, A1C results that are inconsistent with the clinical presentation should be investigated.
PERSPECTIVE
Notwithstanding the use of glucose (FPG and/or the OGTT) as the “gold standard” for
the diagnosis of diabetes for many years, glucose testing suffers from several deficiencies.
The requirement that the subject be fasting at the time the blood is drawn is a considerable
inconvenience. While our ability to measure glucose has improved, inherent biological
variability can produce very large differences within and among individuals. In conjunction
with lack of sample stability, which is difficult to overcome in clinical practice,
these factors results in lack of reproducibility of glucose testing.
A1C, which reflects chronic blood glucose values, is routinely used in monitoring
glycemic control and guiding therapy. The significant reduction in microvascular complications
with lower A1C and the absence of sample lability, combined with several other advantages
(Table 3), have led to the recommendation by some organizations that A1C be used for
screening and diagnosis of diabetes (1). Accumulating evidence suggests that racial
differences in A1C values may be present, and the possible clinical significance of
this needs to be determined. Importantly, A1C cannot be measured in certain conditions.
Despite these caveats, A1C can be measured accurately in the vast majority of people.
A comprehension of the factors that influence A1C values and the conditions where
it should not be used will produce accurate and clinically meaningful results. The
convenience of sampling at any time without regard to food ingestion makes it likely
that measurement of A1C will result in the detection of many of the millions of people
with diabetes who are currently undiagnosed.