hba1c as a screening tool for type 2 diabetes

35
Josh Rubenschuh Diagnostic Trial Review: HbA1c as a Screening Tool for Type 2 Diabetes Clinical Epidemiology Professor Brent Faught

Upload: jay-ruby

Post on 08-Feb-2016

18 views

Category:

Documents


1 download

DESCRIPTION

testing for diabetes

TRANSCRIPT

Page 1: HbA1c as a Screening Tool for Type 2 Diabetes

Josh Rubenschuh

Diagnostic Trial Review:

HbA1c as a Screening Tool for Type 2 Diabetes

Clinical Epidemiology

Professor Brent Faught

Page 2: HbA1c as a Screening Tool for Type 2 Diabetes

Introduction

Over the past 20 years the western world has seen obesity rates rise exponentially.

This unprecedented rise in obesity prevalence has been mirrored by a parallel rise in the

prevalence of type 2 diabetes mellitus. Type 2 diabetes is a metabolic disorder that is

characterized by high blood glucose resulting from peripheral insulin resistance. As of 2010

there were approximately 285 million people with the disease world wide as compared to only

30 million in 1985 (Guthridge, 2007). If left untreated diabetes can lead to many clinical

pathologies: hyperglycemia can contribute to the development of cardiovascular diseases like

myocardial infarction or stroke and the irregular glucose metabolism caused by diabetes can

lead to kidney and liver malfunction caused by ketoacidosis in the bloodstream.

The aetiology for type 2 diabetes is currently unknown, but it is hypothesized that

obesity causes insulin resistance, because adipose tissue stimulates a peripheral inflammatory

response, which is said to impair the sensitivity of insulin receptors throughout the body. When

peripheral insulin receptors function inadequately, the body cannot respond to insulin, and thus

the body is unable to regulate cellular glucose uptake and metabolism (Guthridge, 2007).

Currently, universal screening is not performed for type 2 diabetes as there is no

evidence that such a program would improve health outcomes. Screening for type 2 diabetes is

traditionally performed when a patient presents with high blood pressure or symptoms of

glucosuria, polyuria or polydipsia (Bennet, 2006). The diagnostic algorithm for diabetes

traditionally begins with a Fasting Plasma Glucose Test (FPGT), which measures the bodies

ability to metabolize a set amount of glucose within a defined period of time. If the FPGT

suggests impaired glucose metabolism then a blood sample is taken to confirm the presence of

insulin. If insulin is present in the blood it can be assumed that the body has become resistant to

Page 3: HbA1c as a Screening Tool for Type 2 Diabetes

insulin’s action, but if insulin is not present it can be assumed that the body’s insulin

production from the pancreas has failed (Bennet, 2006). In situations where the results of the

FPGT are inconclusive an Oral Glucose Tolerance Test can be performed, this test is more

expensive then the FPGT, but it is more sensitive and specific with regard to its diagnostic

abilities. There are tertiary levels of testing for type 2 diabetes that can diagnose the condition

with relative certainty, such as the hyperinsulinemic-euglycemic clamp test, but these tests are

rarely used in clinical practice because they are incredibly expensive and time consuming and

the secondary testing methods are adequately able to screen for diabetes (Bennet, 2006).

Type 2 diabetes is highly manageable and in some cases the condition can even be

cured. Diabetes management focuses on lifestyle interventions, lowering other cardiovascular

risk factors, and maintaining blood glucose levels within the normal range. Exercise is the

primary recommendation made to individuals suffering from type 2 diabetes because exercise

can stimulate insulin-independent glucose uptake into the muscles, which can control blood

glucose levels even in the presence of insulin receptor dysfunction (Guthridge, 2007). A low

glycemic diet and proper exercise program are usually sufficient to manage diabetes

effectively, but in cases where exercise is not possible or exercise is ineffective

pharmacological interventions are available. Biguanides and Thiazolidinediones are both drug

classes of insulin sensitizers that can effectively increase the insulin receptors affinity for the

hormone (Bennet, 2006).

With proper intervention and management strategies the prognosis for an individual

with type 2 diabetes is excellent. The patient can essential function unhindered by the

condition, but they may have to be cautious surrounding their dietary intake, because they will

Page 4: HbA1c as a Screening Tool for Type 2 Diabetes

be susceptible to adverse events if they consume high sugar foods in large quantities, or if they

fail to adhere to an exercise regimen (Bennet, 2006).

Question 1: Is This Test Potentially Relevant to my Practice?

In order for the HbA1c test to be a valuable screening tool it must

be useful within a general practitioners scope of practice, because primary

care facilities are the context in which diabetes screening would take place.

When using a screening tool, the clinician needs to ask whether or not the

screening test is relevant to their practice? This question is exceedingly

important because a clinician needs to evaluate whether the screening tool

in question is the best tool available in terms of utility and cost-

effectiveness. None of the articles directly addressed this question, instead

they focused on comparing the HbA1c test to the FPGT to determine which

test had higher efficacy and validity.

Focusing on the test itself, a prevailing consensus is starting to

emerge concerning the utility of Haemoglobin A1c (HbA1c) testing as a

front-line screening tool for diabetes (Saudek, 2008). After investigating the

efficacy and validity of the HbA1c test Herdzik et al. and Colagiuri et al.

have suggested that HbA1c is a better screening tool for diabetes than the

fasting plasma glucose test that is currently used for screening patients. In

contrast, Mannucci et al. determined that the HbA1c test did not warrant

inclusion in the screening protocol for diabetes.

HbA1c testing is an alternative to direct glucose measurements,

when screening patients for type 2 diabetes. HbA1c levels do not fluctuate

Page 5: HbA1c as a Screening Tool for Type 2 Diabetes

with day to day variations in plasma glucose levels; HbA1c levels are

representative of long term plasma glucose concentration (Saudek, 2008).

Haemoglobin A1c is produced in response to prolonged hyperglycaemia;

prolonged hyperglycaemia up-regulates the gene expression of HbA1c,

because this haemoglobin isoform is better than traditional haemoglobin at

transporting oxygen in glycaemic conditions. Increased production of the

HbA1c protein only occurs after prolonged glycaemic stress is placed on the

oxygen transport system.

HbA1c testing would improve a physicians ability to ascribe the

correct diagnosis of diabetes. Current glucose testing methods can be

rendered inaccurate by a person's recent diet and exercise regimen,

however HbA1c levels are indicative of long term glucose levels, so short

term lifestyle changes will not undermine the correct diagnosis when using

HbA1c testing (Saudek, 2008).

The HbA1c test also has excellent utility because type 2 diabetes is

highly treatable. Insulin sensitivity is inversely proportional to whole-body

adipose tissue content and plasma glucose levels (Saudek, 2008). The

detection of diabetes results in a patient being proscribed an exercise

regimen to decrease their whole body fat content and decrease their blood

sugar through insulin-independent muscle glucose uptake. Usually exercise

alone is sufficient to control or even eliminate type 2 diabetes. If adherence

to an exercise plan is unfeasible or if exercise proves ineffective, there are

also pharmacological interventions that can be employed.

Page 6: HbA1c as a Screening Tool for Type 2 Diabetes

Thiazolidinediones and Biguanides are drugs that sensitize the insulin

receptor, so that insulin mediated glucose uptake can return to normal and

diabetes can be controlled (Krentz, 2006). Exercise or drug interventions

are critical because prolonged untreated diabetes can have severe clinical

manifestations.

A series of practical considerations favour the use of HbA1c over

the Fasting Plasma Glucose Test, when screening for type 2 diabetes:

Firstly, the current screening tests for diabetes: the FPG test

require that the patient fast for a minimum of 8 hours prior to testing, but

the HbA1c test does not require any type of fasting (Saudek, 2008). Many

patients necessitate afternoon doctors appointments, therefore under the

current protocol patients would need to fast for the majority of the day,

which is both inconvenient and unhealthy. The need for a fasting blood

sample decreases the opportunity for a diagnosis of diabetes, because

some patients will want to avoid prolonged fasting, so they will avoid being

tested.

Secondly, HbA1c levels are not affected by short-term lifestyle

changes. A few days of dieting or increased exercise in preparation for a

doctor visit can significantly affect FPG, but HbA1c accurately reflects long-

term glycaemia (Saudek, 2008). Human metabolism, specifically blood

glucose regulation is highly dynamic and highly variable, which

necessitates that the screening test for diabetes must maintain high

validity in the face of sporadic fluctuations in glucose levels. HbA1c

Page 7: HbA1c as a Screening Tool for Type 2 Diabetes

maintains high validity in the face of lifestyle changes that may alter

glucose metabolism, while FPG is easily manipulated by short-term lifestyle

changes.

Thirdly, the HbA1c test is not a new or novel test within the realm

of a general practitioners practice, the HbA1c test is currently utilized by 78

percent of Canadian general practitioners for monitoring glycaemia in

diabetes patients. In the past HbA1c testing was not used in the screening

process because standardized and reliable diagnostic cut-offs values were

not available, but these standardized diagnostic values have since been

established (Saudek, 2008). The HbA1c would not require additional

training for physicians, so most physicians would not resist expanding the

use of HbAc1 within their practice.

Lastly, the HbA1c test costs approximately 68 dollars which is

significantly less than the 134 dollar cost of the FPGT (APRTG, 2012). Any

cost reducing strategy within the current healthcare funding structure is

highly beneficial. The inherent savings that would result from utilizing

HbA1c testing over the current tests could expand the number of people

who receive screening without increasing costs for the system. An increase

in the volume of patients screened could potentially unmask many of the

latent cases of type 2 diabetes within Canada.

According to the majority of literature surrounding HbA1c testing,

the HbA1c test should be placed into the screening protocol for diabetes

and from an economic perspective it should replace the FPGT, because it

Page 8: HbA1c as a Screening Tool for Type 2 Diabetes

outcompetes this test in terms of clinical efficacy and utility (Herdzik, 2002

& Coliguri, 2004).

Question 2: Has the Test Been Compared to a True Gold Standard ?

The hyperinsulinemic-euglycemic clamp is the ‘True Gold Standard’

for diagnosing type 2 diabetes, this "clamp" technique requires a steady IV

infusion of insulin to be administered into the left arm. The serum glucose

level is "clamped" at a normal fasting concentration by administering a

variable IV glucose infusion into the right arm. Numerous blood samples are

then taken over the course of the procedure to monitor serum glucose so

that a steady "fasting" level can be maintained (Castracane, 2003). In

theory, the IV insulin infusion should completely suppress hepatic glucose

production, thus gluconeogenesis and lipolysis should not confound the

test's ability to quantify how sensitive target tissues are to insulin. The

degree of insulin resistance should be inversely proportional to the glucose

uptake by target tissues during the procedure (Castracane, 2003). In other

words, the less glucose that's taken up by tissues during the procedure, the

more insulin resistant a patient is.

The hyperinsulinemic-euglycemic clamp is rarely used in the

clinical setting because it is very time consuming and exceptionally

expensive for a screening tool, with the materials and labour costing in

excess of 700 dollars per test. The high cost and technical difficulty

associated with the hyperinsulinemic-euglycemic clamp makes it

exceedingly difficult to use in a diagnostic trial (Castracane, 2003). All of

Page 9: HbA1c as a Screening Tool for Type 2 Diabetes

the diagnostic trials under investigation in this review chose to use the Oral

Glucose Tolerance Test (OGTT) as their ‘reference standard’ because this

test is highly valid and it is commonly used in the clinical setting because it

only costs about 130 dollars per test (APRTG, 2012).

All 3 studies under investigation compared the HbA1c test to the

‘reference standard’ Oral Glucose Tolerance Test. The OGTT encompasses

the administration of a standard dose of glucose orally and then 2 hours

later the plasma levels of glucose are measured to quantify insulin

dependant glucose uptake (Castracane, 2003). The problem with

comparing the HbA1c test to the OGTT reference standard is the fact that

the OGTT does not have 100 percent sensitivity and 100 percent specificity.

The reference standard, like almost every test is going to have inherent

error when attempting to determine the disease status of screened

patients. It is important to note that, in situations where the “True Gold

Standard” can not be administered, it is an acceptable practice to utilize

the most highly valid alternative test as the reference standard (Greenhalg,

1997). However, without definitively knowing the disease state of the

screened patients it could be erroneous to compare the congruency of

results between the OGTT and the HbA1c test. The congruency between the

OGTT and the HbA1c test may provide some insight into the efficacy of the

HbA1c test, but these results must be approached with caution, because

the OGTT has a relatively large chance of misdiagnosing and missing

diagnoses.

Page 10: HbA1c as a Screening Tool for Type 2 Diabetes

Question 3: Did this validation study include an appropriate spectrum of subjects?

When researchers design their studies, they attempt to get the

most appropriate spectrum of subjects to their disease of which they are

interested in. The more appropriate the spectrum of subjects in regards to

the disease, the more valid your results will be. What is very important is

prevalence. It is important that the studies have the same or very close

prevalence's of the disease as in general population when dealing with the

primary health care field. In order to be relevant and useful findings to the

general practitioner the prevalence of the study must be close to the

prevalence in the general population. If not, and the prevalence rates are

different it will influence the positive predictive value (PPV) and the

negative predictive value (NPV) and skew the findings. In all 3 of the

articles the prevalence of diabetes was similar to their general population

prevalence. The Mannucci et al (2002) study had a prevalence of 6.5% and

in Italy the prevalence of diabetes is 6.6% (Shaw, 2010); so no prevalence

bias is declared. The Colagiuri et al (2004) study had a prevalence of 7.4%

and in Australia the prevalence of diabetes is 5.7% (AIHW, 2011), so there

is a slight difference in prevalence which would slightly change the PPV and

NPV but not enough to declare a large prevalence bias. Finally, the Herdzik

et al (2002) study had a prevalence of 6.1% and the prevalence of diabetes

in Poland is 6.54% (Badave, 2010) which is similar and thus prevalence bias

is not declared. Overall, all three studies had similar diabetes prevalence's

as seen in the general population.

Page 11: HbA1c as a Screening Tool for Type 2 Diabetes

Furthermore, do theses studies have subjects that can be similar to

the general population and not filled with individuals with risk factors for

type 2 diabetes. If the study does not have an spectrum that is like the

general population there will be no external validity and be essentially

useless for the general practitioner. It has been found that the risk factors

for diabetes are age (older than 40), obesity, aboriginal/african/asian

populations, low SES and heart disease patients (Public Health Agency of

Canada, 2011). From what was given in the study, each study will be

analyzed in terms of their subject spectrum. The Mannucci et al study

subjects were all undiagnosed subjects, no smokers, majority of

commuters, volunteers from newspaper and television ads,

weight/height/sex/family history were all taken but not given in the study

except for a large number of obese individuals. This study did not have an

appropriate spectrum and thus the results may not be externally valid. We

don't know the risk factor distributions that may have an effect on the

results such as weight, family history and socioeconomic status. The issue

with not having this data means that the results might work in favour of the

researchers and be inaccurate. The Colagiuri et al study had an appropriate

spectrum because their subjects were from various ethnicities, all over 25,

from urban and non urban areas equally and had a sample size of 10,447

which would approximate a normal distribution and resemble the general

population. Thus, the results could be applied to the general practitioners

office scenario where he deals with the general population. The Herdzik et

Page 12: HbA1c as a Screening Tool for Type 2 Diabetes

al article had a somewhat appropriate spectrum in that the majority of risk

factors were distributed although the study consisted of only Caucasians.

Due to there only being Caucasians in this study, this articles findings are

not totally externally valid.

Question 4: Has Workup Bias Been Avoided?

In diagnostic trials it is essential to the validity of the study that the

gold standard or reference standard be done on every subject. There is a

potential bias in studies where this does not occur and this is called work up

bias. When work up bias occurs the gold standard or reference standard is

only done to those who tested positive leading to an overestimation of the

sensitivity. In all three of the articles work up bias has been avoided. All

subjects had the OGTT the reference standard for all three articles. Thus

there are no ramifications for any of the articles in terms of work up bias

because work up bias was absent in all three articles. In this case it was

justified that all individuals should get the OGTT because it is not a majorly

severe or too costly procedure thus all subjects were able to have it done.

Question 5: Has Expectation Bias Been Avoided ?

It is possible that when analyzing the results of a screening test a

researcher could be influenced by peripheral knowledge that is unrelated

to the actually results of the screening tool. For instance if the researchers

investigating the efficacy of HbA1c testing knew the disease state of the

patients they were screening, it is entirely possible that their outcome

Page 13: HbA1c as a Screening Tool for Type 2 Diabetes

expectations might influence the way they interpret the results of the

HbA1c test. Knowing the disease state of screened individuals or even

knowing patient symptoms can lead researchers to make a diagnosis that

may not be congruent with the actual test results.

An excellent way to prevent expectation bias is to blind those

involved with interpreting test results. If the researchers are privy to no

peripheral information about the patients disease status or symptoms,

they are less likely to be influenced by confounding information when

interpreting the test results, thus eliminating expectation bias (Greenhalg,

1997).

The articles by Herdzik et al. and Colagiuri et al. both explicitly

addressed the fact that the biochemists running their HbA1c tests were

completely blinded and had no previous knowledge of patient disease

status or symptoms. The blinding of research analysts suggests that the

results of these studies were not influenced by expectation bias, and thus

they retain relatively high validity. Unfortunately, the article by Mannucci

et al. made no mention of any blinding protocol; it appears that the

researchers administered the OGTT first and established the disease state

of each patient, then the same set of researchers administered the HbA1c

test. The fact that the researchers had a previous knowledge of patient

disease status prior to administering the HbA1c test may have potentially

biased their results based on their personal expectations. The HbA1c test

produces a serum concentration value, if the serum concentration falls

Page 14: HbA1c as a Screening Tool for Type 2 Diabetes

above or below a range of inconclusive values, then a definitive diagnoses

can be made, however there are intermediary ‘inconclusive’ plasma

concentrations where the determination of disease status is left up to the

“expert” analyst (Saudek, 2008). The interpretive nature of this test leaves

it exceptionally vulnerable to expectation bias, because if the test

produces an ‘inconclusive’ concentration, an individual who knows the

expected outcome will most likely classify the patient in accordance with

their preconceived disease status.

Question 6: Was the Test Shown to be Reproducible?

Essentially the test is reliable and reproducible if the same or

different observers perform the same test on two occasions on a subject

whose characteristics have not changed will they get the same result. All

three studies failed to report the validity and reliability of their statistical

measures, but because the HbA1c test is done by scientific analysis of

Hemoglobin A levels through a blood sample as long as the equipment is

working correctly the test should be reproducible with the same results

every time. Therefore, the test should be reproducible based on those

facts, although because the Mannucci et al. paper did not have an

appropriate spectrum the results might not be valid, and if something is

not valid it may not be reliable. Therefore the results from that study

might not be reliable whereas the results from the other two studies that

promote HbA1c use are reliable because they are more valid. McCarter et

al (2004) has tested the variation of the Hba1c test and found that there is

Page 15: HbA1c as a Screening Tool for Type 2 Diabetes

low intra-individual variation which is useful in the early detection of

diabetes, and thus HbA1c is meaningful for clinical workers and has been

shown to have strong reproducibility. The ramifications of not being

reliable or being reproducible will cause error in your statistical measures

because they will not be as precise as what they should be. The actual

specificity might not be actually what they report due to their errors and

unreliability.

Question 7: What are the Features of the Test as Derived From This Validation Study?

In the analysis of diagnostic trials and whether the test is valid, all

the above standards that have been previously talked about may be met

but the test might still be worthless because the sensitivity, specificity,

and other crucial features such as the PPV and NPV are too low, in other

words the test is not valid. What counts as acceptable is up for debate, in

terms of screening we are looking for a high specificity. in the case of

diabetes, we have a relatively common disease and so the specificity is

very important. For screening purposes, a test that produces a large

number of false positives would pose major problems to health

departments and cost to many resources. Thus, a high specificity is

essential. Furthermore, the positive predictive value and negative

predictive value are essential to this situation at the primary care level.

These values are essential because approximately 90% of the time we are

not sure if the individual has the disease but what is known is the outcome

of our clinical test. As well, Positive likelihood ratios and negative

Page 16: HbA1c as a Screening Tool for Type 2 Diabetes

likelihood ratios have clinical importance for the general practitioner. Each

study has different features which will be displayed in a summary table.

Table 1: Statistical features as derived from the Mannucci et al, Colagiuri

et al, and Herdzik et al papers in terms of the HbA1c diagnostic tool with

OGTT as the reference standard.

Statistical Feature Mannuci et al,

2002

Colagiuri et al,

2004

Herdzik et al, 2002

Sensitivity 98% 78.7% 73.7%

Specificity 30% 82.8% 93.2%

False Negative Rate 2% 21.3% 26.3%

False Positive Rate 70% 17.2% 6.8%

Prevalence 6.5% 7.4% 6.54%

Positive Likelihood

Ratio (PLR) 1.3 4.58 10.84

Negative

Likelihood ratio

(NLR)

0.31 0.26 0.28

Accuracy 64% 80.8% 83.45%

Positive Predictive

Value 8.83% 26.7% 42.3%

Negative Predictive

Value 99.4% 97.98% 98.1%

Page 17: HbA1c as a Screening Tool for Type 2 Diabetes

When looking at the PLR the Herdzik et al article has a PLR of 10.84

which essentially means that given a positive test result we can rule in

that they have the disease making the test effective, although Mannuci

claims the test has a PLR of 1.3 meaning there is no change in the

likelihood of having the disease. Colagiuri has a 4.58 PLR which would

agree with Herdzik in that it is an effective screening tool. Furthermore, an

effective screening tool for a common disease like diabetes must have a

strong specificity because we want to correctly identify those people who

do not have diabetes at the expense of not identifying people who actually

have the disease. This is logical in that diabetes in its early stages is not a

very virulent disease that can be treated readily. Colagiuri et al and

Herdzik et al both claim it has a high specificity 82.8% and 93.2%

respectively, whereas Mannuci et al claims the specificity is only 30% and

thus not an effective tool for use at the general practitioners clinic. All

three articles have similar prevalence's which mean, the prevalence won't

have differing effects on the PPV between the different articles; it can be

seen with high prevalence's that the PPV will increase. Thus even though

the previous factors of validity could have been complete in the Mannucci

et al article (which they aren't) the test would still not be valid in their

sense because of the low specificity, PLR and other features.

Question 8: Were Confidence Intervals Given?

Page 18: HbA1c as a Screening Tool for Type 2 Diabetes

Another useful tool in demonstrating the validity in a diagnostic

test and the corresponding statistical features is the use of confidence

intervals to account for statistical ambiguity. Through a basic statistical

concept; such as the following, as a sample size gets larger, the

confidence interval gets smaller it can be seen that it is vitally critical for

smaller studies to include confidence intervals to account for this

uncertainty. Unfortunately none of our studies include confidence intervals

for any of their statistical features, therefore we are uncertain how

accurate they are. Take specificity for example, any of the three articles

could've put down a number like 50 but the actual range could've been

20-80… how accurate is that. Due to this we have to question whether the

features are actually correct. For example, did Mannucci have a hidden

agenda and want to put down the HbA1c tool so that the fasting plasma

glucose test would be used instead at the clinic because him and his

researchers had ties to that test. All in all, if confidence intervals aren't

given we have to question the validity of the results. Although, in larger

studies such as the Colagiuri et al study whose sample size is 10447 the

confidence intervals would be inherently small due to the large sample

size and thus those results would be more precise.

Question 9: Has a Sensible “Normal Range” Been Derived?

When diagnosing diabetes using the HbA1c test, there is no exact

“normal range” that exists. There is simply no clean cut-off between what

would be normal and abnormal… but through PLR's and a ROC we can get

Page 19: HbA1c as a Screening Tool for Type 2 Diabetes

as close as we can to optimize our tests. In fact, Jesudason et al (2003)

and Tavintharan et al (2000) recommend the use cut off value of 6.2% to

be the best predictor of diabetes for the HbA1c test. In the Colagiuri study

the best predictive value was 5.3% calculated from the receiver operating

characteristic (ROC) curve. This cut off value was associated with a

specificity of 82.8% and a sensitivity of 78.7%. Herdzik et al did not claim

how they came up with their cut off value of 6.4%, but they had a

specificity of 93.2% and a sensitivity of 73.7% so the cut off value of 6.4

must be quite high on the ROC. Furthermore, the ROC curve analysis was

used by Mannucci to find a cut off point of 6.6%. Not having a universal

cut-off point is a problem in that studies will use different values this leads

to some studies showing more detection than others (affecting results i.e.

sensitivity/specificity) but also with different cut off points it is hard to

accurately compare the different studies. Furthermore, likelihood ratios

are important here to calculate the right range. Herdzik et al used the cut

off value of 6.4% and found a PLR of 10.84 which is very effective in the

ability to say they have the disease with a positive test result. This cut off

value resembles what other researchers found as the optimum cut off

point for increasing the specificity and PLR which are very essential for the

general practitioner use in the clinic with a common disease such as

diabetes. The ramifications of not having a sensible range would be the

possibility of having too high and too low of a cut-off point and might lead

Page 20: HbA1c as a Screening Tool for Type 2 Diabetes

to too many false negatives or false positives, respectively and also

depending on the circumstances.

Question 10: Has This Test Been Placed in the Context of

Other Potential Tests in the Diagnostic Sequence

The ultimate utility of the HbA1c diagnostic test depends on its

ability to be used in a diagnostic algorithm. The study by Colagiuri et al.

has suggested that the HbA1c test should be placed into the screening

algorithm as the secondary step towards diagnosing type two diabetes.

Dr. Colagiuri and his team of researchers believe the screening protocol

should begin by identifying individuals who have 3 or more risk factors for

diabetes, for example obesity, a family history of diabetes or high blood

pressure. Once susceptible patients have been identified they should be

screened using the HbA1c test and those who test positive according to

the HbA1c test would be screened using the OGTT. It is important to note

that Colagiuri et al. suggested that the HbA1c test replace the FPGT as the

secondary screening tool in the diagnostic algorithm based on their

findings that the HbA1c test is more sensitive and specific than the FPGT

and the HbA1c is less susceptible to confounding by short term lifestyle

changes. The findings of Colagiuri et al. definitely support replacing the

FPGT with the HbA1c test because if the FPGT was retained as the primary

screening tool for diabetes with such a low sensitivity it is entirely possible

that a large number of individuals with diabetes would go undiagnosed.

The primary screening tool needs to be highly sensitive in order to limit

Page 21: HbA1c as a Screening Tool for Type 2 Diabetes

the number of false negatives. Tests with higher specificity can be

employed later in the algorithm in order to screen out any false positives.

The study by Herdzik et al. has also suggested that the HbA1c test

be included within the screening protocol for type 2 diabetes, but they

believe the HbA1c test should follow the FPGT in the diagnostic algorithm.

Dr. Herdzik and his team found that the FPGT had a slightly higher

sensitivity then the HbA1c test, but the Haemoglobin A1c test was much

more specific, therefore they concluded that the HbA1c test should be

used as a secondary confirmatory test following primary screening by the

FPGT. The results obtained by Herdzik et al. support the conclusion that

the FPGT should be retained in the diagnostic algorithm and subsequently

followed by the HbA1c test, because the diagnostic sequence will utilize

the high sensitivity of the FPGT followed by the high specificity of the

HbA1c, their paired validity is higher then each test independently. Also

the researchers concluded that the HbA1c test has the ability to hold its

validity in the face of confounding factors such as short term lifestyle

changes that could influence the FPGT.

Finally, the study by Mannucci et al. argued that the HbA1c test

had no place within the diagnostic algorithm because the current

diagnostic thresholds for the test were inadequate for accurate

diagnostics. Dr. Mannucci and his team found that the FPGT had a higher

sensitivity and specificity then the HbA1c test, so they suggested that the

diagnostic protocol for diabetes remain the same with FPG as the primary

Page 22: HbA1c as a Screening Tool for Type 2 Diabetes

screening tool, followed by the OGTT. Based on the study findings Dr.

Mannucci and his team made the correct assessment because including a

test in the diagnostic algorithm that is superior to none of the current tests

is simply just a waste of resources.

Do the Conclusion and Recommendations fit the Content?

Based on the comprehensive analysis of all 3 diagnostic trials

analyzed within this review, the panel agrees with the conclusions of

Herdzik et al. and Colagiuri et al. These two studies have concluded that

the HbA1c test for screening diabetes should definitely be included within

the diagnostic protocol for the disease. This panel believes that the results

of the Mannucci et al. article conflicted with the other 2 studies due to

multiple errors within the studies methodology. The Mannucci et al. article

did not provide confidence intervals for their outcome measures and they

used a sample size that was unable to illicit sufficient power. Without

confidence intervals or sufficient power it is impossible to determine the

accuracy and precision of the statistical measures. In addition, Dr.

Mannucci and his team did not include an appropriate spectrum of

subjects within their trial and their study prevalence was much lower then

that of the external population. All of these flaws within their study

methodology call their results into question, so their negative measures

assessing the HbA1c test may not be valid or reliable.

Based on the congruent results of the Herdzik et al. and Colagiuri

et al. articles as well as peripheral research, it appears that the HbA1c test

Page 23: HbA1c as a Screening Tool for Type 2 Diabetes

has extremely high utility within the diagnostic protocol for type 2

diabetes. With the HbA1c tests newly established diagnostic cut-offs it

appears to be highly sensitive and specific. The HbA1c test is also

extremely useful in the diagnostic sequence because it retains high

validity in the face of environmental confounders that would influence the

glucose measurement tests. Additionally, the HbA1c test is a cost effective

alternative to the primary testing tools currently utilized within the

diagnostic spectrum for diabetes. Therefore, this panel concludes that the

HbA1c test is an extremely useful screening tool for diabetes mellitus and

it should undoubtedly be placed within the diagnostic algorithm as the

primary screening protocol.

Works Cited

AIHW. (2011). Diabetes prevalence in Australia: detailed estimates for 2011. Diabetes series no. 17. Cat. no. CVD 56. Canberra: AIHW. Viewed 24 March 2012.

APRTG. (2012) Biochemical Suppliers Ltd. Online store. Retrieved from: http://www.healthtestingcenters.com/hemoglobin-blood-test.aspx

Page 24: HbA1c as a Screening Tool for Type 2 Diabetes

Badave, Pol-Diab. Sieradzki, J. (2010). Diabetes: The Hidden Pandemic and its Impact on Poland. Diabetes. Novonordisk. 3: 8-15.

Bennet, C.M. Guo, M. Dharmage, S. C. (2006). HbA1c as a screening tool for the detection of type 2 diabetes: a systemic review. Journal of Diabetic Medicine. DOI 10.111. 1464-1484.

Castracane, V.D. Kauffman, B. (2003) Assessing Insulin Sensitivity: Diagnosing type 2 Diabetes Mellitus. The Journal of Clinical Endocrinology & Metabolism.

Colagiuri S, Cameron A, Hussain Z, Shaw J, Zimmet P. (2004). Screening for type 2 diabetes and impaired glucose metabolism: the Australian experience. Diabetes Care. 27: 367– 371.

Greenhalg, T. (1997). How to read a paper: Papers that report diagnostic or screening tests. British Medical Journal, 315:540-543

Guthridge, L. Eberhard, S. (2007). Guidelines for Diabetes: Executive Summary. European Heart Journal. DOI 10.1093.

Herdzik E, Safranow K, Ciechanowski K. (2002). Diagnostic value of fastingcapillary glucose, fructosamine and glycosylated haemoglobin in detecting diabetes and other glucose tolerance abnormalities compared to oral glucose tolerance test. Acta Diabetol. 39: 15– 22.

Jesudason DR, Dunstan K, Leong D, Wittert GA. (2003). Macrovascular risk and diagnostic criteria for type 2 diabetes. implications for the use of FPG and HbA1c for cost-effective screening. Diabetes Care. 26: 485– 490

Krentz, Andrew J. Bailey, Clifford R. (2006). Oral anti-diabetic agents: current role in type 2 diabetes mellitus. Journal of Applied Pharmacology. vol. 16 no. 4 1298-1313.

Mannucci E, Ognibene A, Sposato I, Brogi M, Gallori G, Bardini G et al. Fasting plasma glucose and glycated haemoglobin in the screening of diabetes and impaired glucose tolerance. Acta Diabetol 2003; 40: 181–186.

McCarter RJ, Hampe JM, Gomez R, Chalew SA. (2004). Biological variation in HbA1c predicts risk of retinopathy and nephropathy in type 1 diabetes. Diabetes Care. 27: 1259–1264

Public Health Agency of Canada (2011) Chronic Disease Risk Factors:

Page 25: HbA1c as a Screening Tool for Type 2 Diabetes

http://www.phac-aspc.gc.ca/cd-mc/risk_factors-facteurs_risque-eng.php

Saudek, Christopher D. Herman, William H. Sacks, David B. Bergenstal, Richard M. Edelman, David. Davidson, Mayer B. (2008). A New Look at Screening and Diagnosing Diabetes Mellitus. The Journal of Clinical Endocrinology & Metabolism. vol. 93 no. 7 2447-2453

Shaw, P. Sicree, T. & Zimmet. S. (2010). Global estimates for the prevalence of diabetes in 2010 and 2030. Diabetes Research and Clinical Practice, 87(1), 4-14.

Tavintharan S, Chew LSW, Heng DMK. (2000) A rational alternative for the diagnosis of diabetes mellitus in high risk individuals. Ann Acad Med Singapore. 29: 213– 218