A position statement of the Australian Diabetes Society

Reimbursement by Medicare of the costs of measuring glycated haemoglobin (HbA1c) for the diagnosis of diabetes mellitus was recently approved. An HbA1c value of 48 mmol/mol (6.5%) or more constitutes a positive result, suggesting the diagnosis of diabetes mellitus. This test provides an alternative to traditional glucose-based methods of diagnosis; it does not replace them. The correct use of the test may facilitate earlier diagnosis of people with elevated mean blood glucose levels who are at increased risk of long-term diabetes-specific microvascular complications. HbA1c assessment will be used predominantly for the diagnosis of type 2 diabetes mellitus.

It is important that medical practitioners who elect to use the test for diagnostic purposes understand its nature, its limitations and its benefits. The latter were outlined in the position paper of the HbA1c Committee of the Australian Diabetes Society published in this Journal in 2012.1 We recommend that medical practitioners read the earlier paper in conjunction with this new implementation document. Assessment of HbA1c levels during pregnancy is not discussed in this article.

This position statement of the Australian Diabetes Society is endorsed by the Royal College of Pathologists of Australasia and the Australasian Association of Clinical Biochemists.

Medicare reimbursement

The Medicare Benefits Schedule (MBS) entry for item 66841 describes the test as the “Quantitation of HbA1c (glycated haemoglobin) performed for the diagnosis of diabetes in asymptomatic patients at high risk”. When used for this purpose, the cost of the test can be reimbursed only once during a 12-month period.2 The brevity of the MBS description raises certain questions.

1. High-risk patients

Only patients at high risk of undiagnosed diabetes should be tested. These are patients with either (i) a medical condition or ethnic background associated with high rates of type 2 diabetes, or (ii) an Australian type 2 diabetes risk (AUSDRISK) score of 12 or greater, placing them at increased risk of diabetes.1,3,4

The HbA1c test should not be used to randomly or systematically screen undifferentiated groups of patients. Without prior knowledge of the medical status of an individual, the test may not correctly diagnose patients as having diabetes (see section 5, below).

2. Asymptomatic patients

Medicare restricts diagnostic HbA1c assessment to asymptomatic patients. However, many symptoms of diabetes are, in isolation, non-specific; eg, tiredness and blurred vision. Patients presenting with such symptoms should be considered asymptomatic and appropriate for HbA1c testing if at high risk of developing diabetes. If one or more symptoms that suggest diabetes are present in a low-risk patient, blood glucose tests should be used.

Patients who have multiple classical symptoms of diabetes (weight loss, polyuria, polydipsia, blurred vision etc.), however, are not asymptomatic, and should have the diabetes diagnosis confirmed by blood glucose assessment; high blood glucose levels would be expected in these cases. Further, patients with rapidly evolving diabetes can theoretically have normal HbA1c levels because blood glucose levels have not been elevated for a significant period of time.

3. Repeat assessment of HbA1c levels

An HbA1c test result of less than 48 mmol/mol (6.5%) indicates that diabetes is unlikely. As the test will have been performed in a high-risk patient, it should be repeated 12 months later, according to the National Health and Medical Research Council (NHMRC) guidelines.4 These patients should also be given appropriate lifestyle advice.5,6

Labelling people with an HbA1c value slightly under 48 mmol/mol (6.5%) with prediabetes is not recommended, as there is uncertainty about using HbA1c levels to define prediabetes. This is consistent with the position of the World Health Organization.7 However, an HbA1c level of 42–47 mmol/mol (6.0–6.4%) suggests a higher risk of developing diabetes than that based on the AUSDRISK score alone; these individuals will also be at increased risk of the cardiovascular complications.8,9 They should be counselled about lifestyle measures (weight loss, dietary change, exercise) and assessed for other modifiable cardiovascular risk factors (hypertension, dyslipidaemia, smoking). Unless they develop symptoms of diabetes, additional blood glucose measurements should not be performed to diagnose diabetes. They may have blood glucose levels consistent with impaired fasting glucose, impaired glucose tolerance, or diabetes, but, with an HbA1c level below 48 mmol/mol (6.5%), they are at minimal risk of developing microvascular complications.10 Even if diagnosed with diabetes or prediabetes, lifestyle advice should be the major intervention. Their HbA1c levels should be re-assessed 12 months later.

4. Confirmation

The NHMRC guidelines indicate that abnormal blood glucose levels in an asymptomatic patient should be confirmed to establish a diagnosis of diabetes. A single elevated HbA1c result is accepted by Medicare as evidence for established diabetes, although other organisations (WHO, American Diabetes Association) recommend that diagnoses made by HbA1c testing be confirmed by follow-up testing.7,11

The diagnosis of diabetes has employment, insurance, financial and lifestyle implications, so it is important that it is correct. Although a more reliable laboratory measure than blood glucose levels,1 HbA1c levels do vary within a narrow range over time, both in individuals and during the measurement process. Errors during sample labelling and handling can also occur. It is therefore recommended that a confirmatory test be performed on a different day, ideally as soon as possible and before any lifestyle or pharmacological interventions commence; if delayed, a normal follow-up result may reflect the effects of treatment. A result below 48 mmol/mol (6.5%) does not confirm diabetes, and these patients should generally be advised that they do not have diabetes. They are, however, at high risk of its developing, and they should be managed accordingly (section 3).

If both HbA1c and glucose levels are elevated in an individual, the diagnosis of diabetes is confirmed. If only one of the values is elevated, the relevant test should be repeated to confirm the diagnosis.

There is an apparent conflict between these practice guidelines and the Medicare regulations (one diagnostic HbA1c test in a 12-month period). Medicare recognises a single elevated HbA1c measurement as establishing a diabetes diagnosis; this entitles the patient to four monitoring HbA1c tests in each subsequent 12-month period. We therefore recommend that the first monitoring test be performed before any interventions are initiated. A positive result in this test confirms the diagnosis and sets the baseline for clinical management. A result below 48 mmol/mol, if the test is performed appropriately, means that the patient should be classified as not having clinical diabetes, but they should have a further diagnostic HbA1c test 12 months later.

5. Abnormal measurements

In a small but important minority of people, HbA1c levels are not a reliable indicator of plasma glucose levels. An inappropriately low HbA1c value is the major concern, as the diagnosis of diabetes will be missed in such patients. The possibility of medical conditions that invalidate the HbA1c result should be considered in all patients with an unexpectedly low HbA1c result, as discussed in our earlier paper.1 In summary, HbA1c assessment may not be appropriate in patients with significant chronic medical disease, anaemia or abnormalities of red blood cell structure (Box 1). A full blood count may reveal red blood cell abnormalities suggestive of a haemoglobinopathy or haemolytic anaemia, but a normal full blood count does not exclude the possibility of such conditions. Certain ethnic communities more frequently have underlying haemoglobin abnormalities, and this should be discussed with the testing laboratory when appropriate. Emerging methodologies are minimising this problem.

Early identification of diabetes and associated complications provides an opportunity to start effective preventive treatment that reduces the subsequent development or progression of macrovascular and microvascular disease.1–3 However, diabetes remains undiagnosed in up to 50% of people with the disorder.2–4

Delayed diagnosis is due in part to the use of an algorithm that relies on the assessment of glucose levels and, if the results are equivocal, a follow-up oral glucose tolerance test (OGTT).5 This complicated algorithm can significantly delay informing and educating the patient.6 In contrast to glucose testing, assessment of glycated haemoglobin A (HbA1c) requires no fasting.7 This makes it more suitable for opportunistic testing, and results in fewer missed diagnoses.8

The American Diabetes Association (ADA) guidelines have included laboratory HbA1c testing for diagnosing diabetes since 2010.7 HbA1c testing is endorsed by the World Health Organization,9 and is recommended for diagnosing diabetes in the United Kingdom and New Zealand.10,11 In 2012, the Australian Diabetes Society (ADS) expert committee advised that HbA1c assessment can be used to diagnose diabetes, and applied for a corresponding Medicare Benefits Schedule (MBS) rebate.12,13 This was added on 1 November 2014 after the Medical Services Advisory Committee (MSAC) provided advice to the Minister for Health in mid 2014.13,14

MSAC and the ADS agreed that point-of-care (POC) HbA1c testing was not within the scope of this application. Laboratory HbA1c tests from the Kimberley region of northern Western Australia are analysed in Perth, over 1600 km away, which can lead to significant delays in receiving test results. In a number of studies, POC HbA1c results have been shown to be closely correlated with laboratory-based results15–17 and the POC HbA1c process has been accepted by Australian Aboriginal Medical Services for the management of diabetes.18

The availability of immediate results is likely to further improve diagnosis of diabetes in remote areas and the timeliness of starting treatment. We aimed to determine whether a combination of POC and laboratory HbA1c testing, in real-world settings using existing processes for HbA1c assessment, is an effective method of testing for diabetes in remote Kimberley Aboriginal people when compared with the standard glucose algorithm.19

Methods

Data were collected by local health care providers from 1 September 2011 to 30 November 2013 at six Kimberley sites. All Aboriginal and Torres Strait Islander people in the Kimberley region aged 15 years and older are regarded as being at high risk of developing diabetes, and local protocols recommend that they be tested annually.19 Aboriginal and/or Torres Strait Islander people who did not have confirmed diabetes and who were due for diabetes testing and attending participating clinics were invited to take part. A cross-sectional design was used to compare the effectiveness of a new model of detecting diabetes (the HbA1c algorithm) with the standard model (the glucose algorithm) in diagnosing prediabetes and diabetes. The study was conducted on an opportunistic basis according to available clinic resources.

HbA1c and glucose tests

Clinic staff recorded patient consent, determined fasting status, conducted initial POC HbA1c tests and collected blood for laboratory HbA1c and glucose tests. Blood for the laboratory tests was not always collected on the same day as the POC test, as some remote clinics are only able to send laboratory samples for analysis once a week and often collect samples only on the day of transport. For the OGTT, we used a standard 2-hour 75 g/300 mL glucose load (Carbotest, Lomb Scientific).5

Capillary blood HbA1c concentration was measured in a finger-prick blood sample collected by primary health care providers and analysed on a DCA 2000+ Analyzer (Siemens/Bayer). This was the only POC machine available in Kimberley clinics during the period of the study, and is routinely used for assessing diabetic control. The aim of our study was to look at real-world practice, so we did not attempt to change the way staff undertook testing or to maintain or calibrate the machines.17

Venous whole blood samples were collected in fluoride/oxalate tubes (glucose test) or EDTA tubes (HbA1c test). Normal procedures at each clinic were used to store (whole blood was stored at 4°C) and transport samples to one of the three Kimberley PathWest laboratories. Whole blood samples for HbA1c testing were transported to the PathWest laboratory in Perth. The distances from the study sites to the Kimberley laboratories ranged from 2 km by road to 650 km by air, and from the Kimberley laboratories to Perth between 1677 km and 2224 km by air. Samples were received in Perth in 1–8 days.

Venous plasma HbA1c levels were measured as part of routine PathWest work. PathWest uses an automated immunoassay, with anticoagulated whole blood specimens automatically haemolysed by HbA1c haemolysis reagent in the predilution cuvette on a Cobas Integra 800 analyser (Roche Diagnostics). Total haemoglobin levels were measured colorimetrically, while HbA1c levels were determined by immunoturbidimetric assay. The ratio of the HbA1c concentration to that of total haemoglobin in the specimen yielded the final proportionate HbA1c measurement.

Venous plasma glucose (PG) levels were measured by enzymatic assay (glucose oxidase spectrophotometric dry chemistry) on a Vitros 250 Analyser (Ortho Clinical Diagnostics) at the Kimberley PathWest laboratories.

Comparison of the HbA1c and glucose algorithms

Participants were classified independently by each model for detecting diabetes. We then determined, using the same participants for each model, whether the HbA1c algorithm was more likely than the glucose algorithm to:

• provide a rapid definitive result;
• detect cases of diabetes.

Initial classification by the HbA1c algorithm used the POC HbA1c measurement, or the first laboratory HbA1c measurement if the POC test was not completed (five cases). Subsequent diagnosis of prediabetes and diabetes was based on POC and/or laboratory HbA1c measurements as follows.

• Normal: POC or laboratory HbA1c < 39 mmol/mol (< 5.7%).
• Prediabetes: laboratory HbA1c = 39–46 mmol/mol (5.7%–6.4%).
• Diabetes: two diagnostic HbA1c measurements ≥ 48 mmol/mol (≥ 6.5%).

While ADA and WHO guidelines require two diagnostic laboratory results for diagnosing diabetes in asymptomatic patients,9,20 we used diagnostic POC results, with laboratory results confirming the diagnosis.

We based the initial classification using the glucose algorithm on the first laboratory PG measurement. Subsequent diagnosis of impaired glucose tolerance and diabetes was based on follow-up glucose tests as follows.

• Normal: PG level < 5.5 mmol/L.
• Indeterminate (equivocal result): fasting PG (FPG) level 5.5–6.9 mmol/L, or random PG (RPG) level 5.5–11.0 mmol/L with no completed follow-up tests.
• Impaired glucose tolerance: OGTT 2-hour PG level 7.8–11.0 mmol/L.
• Diabetes: one diagnostic OGTT result (FPG level ≥ 7.0 mmol/L or 2-hour PG level ≥ 11.1 mmol/L), or two diagnostic glucose measurements (FPG level ≥ 7.0 mmol/L and/or RPG level ≥ 11.1 mmol/L).

It was expected that participants with an initial laboratory HbA1c value greater than 39 mmol/mol (5.7%) or a glucose level greater than 5.5 mmol/L were to be followed up with further tests (FPG, OGTT, HbA1c) to confirm the abnormal result (Appendix). The researchers provided regular reminders to clinics to support follow-up. Participants were diagnosed with diabetes if they met any of the HbA1c-, OGTT- or glucose-based diagnostic criteria for diabetes, based on all measurements collected during the course of the study.

Barriers to and enablers of screening by each algorithm for detecting diabetes were documented during the study.

Statistical analysis

All analyses were performed with Stata, version 13 (StataCorp). We used the McNemar test for paired nominal data to compare differences between the two algorithms regarding the rates of screening for prediabetes and diabetes; between the rates of their diagnosis; and between the rates of obtaining a definitive result. Point estimates were presented with 95% CIs; the exact P value was used. P < 0.05 was defined as statistically significant.

Ethics approval

Ethics approval was obtained from the Human Research Ethics Committee of the University of Western Australia, the Western Australian Aboriginal Health Ethics Committee and the Western Australian Country Health Service (WACHS) Human Research Ethics Committee. This project was supported by the Kimberley Aboriginal Health Planning Forum Research Subcommittee.21

Results

Two hundred and fifty-five participants were enrolled and assessed by both models for detecting diabetes (Box 1). Their median age was 36 years (range, 17–79 years) and 152 participants (59.6%) were female.

Participants were significantly more likely to receive a definitive test result with the HbA1c algorithm (250 of 255 participants; POC test completed in all but five cases) than with the glucose algorithm (214 of 255 participants: 41 cases were not screened or returned indeterminate results; McNemar odds ratio [OR], 10.0; 95% CI, 3.6–38.5; P < 0.001; Box 1). HbA1c results were also received much more rapidly; only six of 255 participants (2.4%) had not received a result within 7 days. In contrast, 56 (22.0%) of 255 participants had not received a definitive result within 7 days with the glucose algorithm, and 52 (20.4%) had not received a definitive result within 28 days (P < 0.001).

Of the 255 participants, follow-up laboratory tests were planned for the 168 (65.9%) with abnormal initial laboratory measurements (Appendix). These participants were significantly more likely to be followed up with laboratory HbA1c tests than with fasting FPG tests or OGTTs (92 v 72 participants; P = 0.0051). Fifty-seven OGTTs were completed, including 49 of 117 (41.9%) requested tests.

Participants were significantly more likely to be diagnosed with diabetes by the HbA1c algorithm (15 diagnoses) than with the glucose algorithm (4 diagnoses; McNemar OR, 12.0; 95% CI, 1.8–513; P = 0.003). While 16 of 255 participants were identified as having diabetes by one or both algorithms, an additional four participants were identified by supplementary OGTTs performed because of the study design, giving a total of 20 participants diagnosed with diabetes (Box 2). The five cases not identified with the HbA1c algorithm (cases 16–20) had initially been classified as prediabetes; at the time of the diagnostic OGTT, one person had a HbA1c measurement that was also diagnostic for diabetes (Box 2).

Of the 187 participants whose initial glucose measurements were normal, eight (4.3%) were diagnosed with diabetes — either as the result of two diagnostic HbA1c results (cases 4–7) or of a diagnostic OGTT result (cases 17–20) (Box 2). For seven participants, diabetes was confirmed within 2 months, and for the eighth participant, after 6 months. None of the 137 participants with a POC HbA1c measurement of less than 39 mmol/mol were subsequently diagnosed with diabetes.

Discussion

Our study showed that testing for diabetes by HbA1c assessment in Aboriginal populations in remote towns and communities was significantly more likely to detect diabetes than glucose testing, and that a definitive result could be obtained more rapidly.

It is of concern that only four of the 20 participants diagnosed with diabetes were identified with the glucose algorithm. In our study, six participants were diagnosed by OGTT, but four of these had normal FPG values. While our results are based on small numbers, this suggests the FPG is not sufficiently sensitive as a screening test in this population.

In contrast, all participants with a confirmed diagnosis of diabetes were identified by the HbA1c algorithm as having either diabetes (15 cases) or prediabetes (five cases). Those classified as having prediabetes are expected to be followed up more frequently, reducing the chance of diabetes in these patients being missed for any length of time. HbA1c testing is clearly more likely to detect diabetes than glucose testing.

Our study had a relatively complicated protocol because it compared two diagnostic algorithms. Unsurprisingly, adherence to these protocols was not always complete, but adherence to OGTT (42%) was better than in another recent Australian study (27%),22 and better than would be expected from the usual experience of remote Aboriginal health practice. This improvement was probably due to regular reminders provided by the researchers. Initial adherence to HbA1c testing was excellent, but only 55% of those who required follow-up HbA1c assessment were actually tested, with staff reporting the following barriers to follow-up testing:

• Waiting for participants to return while fasting, so that blood for HbA1c and glucose tests could be collected at the same time.
• High workforce turnover. Some clinicians who were reviewing results did not realise that some “patients” were study participants, assuming they had already been diagnosed with diabetes; they considered HbA1c values of 48–52 mmol/mol (6.5%–6.9%) as indicating good glycaemic control and decided that a further HbA1c test was unwarranted.
• Field officers tasked with bringing participants to the clinics often had extensive lists of patients; reviewing blood tests was often a lower priority.
• Participants refusing the OGTT or leaving before the 2-hour sample had been collected.

Our results highlight the need for simplified testing regimens. For asymptomatic individuals, ADA and WHO recommend a confirmatory test as soon as practical.9,20 However, as HbA1c testing is more robust and reproducible than glucose testing, MSAC does not consider a confirmatory test necessary;13 further, there is only one MBS rebate for HbA1c screening in any 12-month period.14

Reducing the number of laboratory tests will further improve screening and diagnosis, and the Kimberley HbA1c algorithm has been updated to reflect this (Box 3). If the POC HbA1c result is abnormal (≥ 39 mmol/mol [5.7%]), we recommend that venous blood be collected on the same day for a confirmatory laboratory HbA1c test. If the POC result is high (the cut-point could be determined locally) then baseline assessments (eg, cardiovascular risk factors, kidney function) for newly diagnosed patients can be requested at the same time as the confirmatory laboratory HbA1c test, again potentially reducing the number of clinical visits and venepunctures needed.

An additional advantage to the patient of an immediate result is that they know straight away whether the result is normal (55% in our study),17 which would avoid the need to return for a clinical review (charged to MBS). Other potential cost savings include reduced specimen collection (patient and staff time, consumable costs) and transportation of blood samples from remote communities to Perth. Consumer costs, such as patient time, can affect compliance23 and are thus an important element of diabetes testing in this population. If patients are not having other laboratory blood tests, POC testing will avoid the need for venepuncture, which may make screening more acceptable to some patients, potentially increasing uptake. This can be expected to further improve screening and diagnosis in remote areas and the timeliness of initiating management, including lifestyle advice and pharmacotherapy.

Further research is required to determine the cost-effectiveness of the Kimberley HbA1c algorithm compared with screening in remote locations by laboratory HbA1c testing alone, and for using simplified POC accreditation processes.17

Potential barriers to implementing HbA1c testing include a lack of familiarity with the test as a diagnostic tool. Further, as the MBS rebate has only recently been announced, it is not currently a routine test for people without diabetes. Together with updated protocols, changes to the way laboratories report HbA1c results of diagnostic tests will be needed to highlight new diagnoses of prediabetes and diabetes. Modifications of how medical software processes these tests may also be needed, as well as extensive education for health service providers.

A change from glucose to HbA1c testing would substantially increase the number of people needing active follow-up (93 instances of prediabetes, compared with 7 of impaired glucose tolerance; Box 1), potentially leading to overdiagnosis of prediabetes. While this would add to the workload of health service providers, identifying more people at increased risk could also improve targeting of attempts to reduce risk early in the disease process. Further research is required to consider the appropriateness of the prediabetes cut-point used in our study and whether differing levels of intervention are appropriate within the HbA1c 39–64 mmol/mol (5.7%–6.4%) range.

The strength of our study is that it reflects the practicalities of diabetes detection in remote locations; staff were trained in-house to use the POC HbA1c analyser and existing processes for laboratory HbA1c measurement were used. The limitations of our study included the use of a convenience sample that may not be representative of the entire Kimberley adult Aboriginal population. Further, not all participants with abnormal laboratory measurements could be located for follow-up laboratory tests, resulting in incomplete data, and there were variations in adherence to the study protocol for follow-up tests.

Our study shows that adopting the Kimberley HbA1c algorithm may simplify the testing process in previously undiagnosed individuals and provide a timelier and more accurate diagnosis of diabetes for Aboriginal people and other high-risk remote populations in Australia and elsewhere in the world.

1 Comparison of the glycated haemoglobin (HbA1c) and standard glucose algorithms for classifying participants

---

POC = point-of-care.

---

* Initial classification was based on the POC HbA1c measurement, or on the first laboratory HbA1c measurement if the POC test was not completed (five cases). Subsequent diagnosis of prediabetes was based on the first laboratory HbA1c measurement, or on the POC measurement if the laboratory test was not completed (four cases). Diagnosis of diabetes was based on the POC and/or laboratory HbA1c measurements. † Initial classification was based on the first laboratory glucose measurement. Subsequent diagnosis of impaired glucose tolerance and diabetes was based on follow-up laboratory glucose measurements or an oral glucose tolerance test. ‡ Two participants each had one diagnostic laboratory HbA1c result that was not followed up; in contrast to American Diabetes Association guidelines, which require a confirmatory test in asymptomatic patients,20 the Australian Medical Services Advisory Committee13 does not consider a confirmatory test to be necessary for diagnosing diabetes. § Seven participants had impaired glucose tolerance; one participant was diagnosed with gestational diabetes 378 days after enrolling.

3 Kimberley algorithm for screening for and diagnosing diabetes using point-of-care (POC) and laboratory (lab) glycated haemoglobin (HbA1c) testing19,24

A different approach may provide more effective early detection of diabetes

Diabetes poses a considerable health threat in Australia and is predicted to soon become the largest contributor to the burden of disease in this country.1 There are an estimated 1 million Australians with diabetes, and another 2 million at high risk of developing the disease.2 Many people with diabetes remain undiagnosed and an important strategy for reducing the disease burden is to detect and treat it earlier in order to minimise the risk of its devastating complications.

The current glucose-based protocol endorsed by the National Health and Medical Research Council (NHMRC) for screening for undiagnosed diabetes is cumbersome, time-consuming and inconvenient, and it impedes the widespread implementation of diabetes screening programs.3 This protocol requires large numbers of individuals to have oral glucose tolerance tests (OGTTs), but fewer than 1 in 3 of those who should complete an OGTT do so.4 In 2011, the World Health Organization endorsed the assessment of glycated haemoglobin (HbA1c) levels as a diagnostic test for diabetes,5 a recommendation adopted by the Australian Diabetes Society (ADS) in 2012.6 While the HbA1c test is more user-friendly and does not require fasting, there are clinical situations in which it may not provide an accurate assessment of diabetes, such as people with certain haemoglobinopathies or conditions that alter red blood cell turnover.6

HbA1c testing in remote communities

An important question is how well HbA1c testing detects undiagnosed diabetes in real-life contexts. The study by Marley and colleagues in this issue of the Journal7 compared the glucose-based algorithm recommended by the NHMRC with an HbA1c-based algorithm as applied in a remote Australian Aboriginal community. The HbA1c-based algorithm used an initial point-of-care (POC) HbA1c assessment followed by laboratory HbA1c assay if needed. Participants were significantly more likely to receive a definitive result within 7 days and to be diagnosed with diabetes using the HbA1c algorithm than with the glucose-based protocol. The study also highlighted the increased likelihood of follow-up with HbA1c testing; only 42% of participants with an equivocal glucose result underwent an OGTT as recommended by the NHMRC guideline. Since not all participants had undergone both HbA1c and OGTT assessments, a comparison of the accuracy of the two procedures for diagnosing diabetes was not possible. Nevertheless, the study clearly showed that the HbA1c-based algorithm detected more cases of diabetes, is more likely to be completed as recommended, and delivered more rapid results.

Can the findings of this study be applied more broadly in Australia? Potentially, but not, unfortunately, while the current Medicare Benefits Schedule (MBS) restrictions on diagnostic HbA1c tests apply.8 The costs of POC HbA1c testing for diabetes diagnosis are not reimbursed by the MBS. This is a major problem, not only for remote communities with restricted access to laboratory services, but also in any situation where POC testing could be used to exclude the likelihood of undiagnosed diabetes. The accuracy of POC testing is often raised as a concern, but an established quality control program operates in the Aboriginal Medical Services, and a similar program could be extended to other settings.9 The Marley et al study could have provided more information on the comparative accuracy of POC and laboratory HbA1c assays had all participants undergone both POC and laboratory HbA1c assessments. In this regard, it should be remembered that there are many inherent inaccuracies in glucose testing related to methodological and procedural techniques, as well as substantial intra-individual biological variability.

A second restriction is that an MBS reimbursement is available for only one diagnostic HbA1c test in a 12-month period. This has implications for performing a second, confirmatory HbA1c test before an asymptomatic person is diagnosed with diabetes, as recommended by Australian and international guidelines.3,5,6 Confirming an initial result is essential in light of the significant lifelong implications of being diagnosed with diabetes, especially when the laboratory result is close to the diagnostic cut-point. This MBS restriction is a missed opportunity, as not all individuals who receive an initial abnormal blood glucose result have follow-up tests, meaning that some individuals are incorrectly diagnosed with diabetes.

Practical questions that need to be resolved

According to the MBS regulations, a single elevated laboratory HbA1c test result establishes the diagnosis of diabetes. Once diagnosed with diabetes, the individual can have up to four MBS-reimbursed HbA1c tests in a 12-month period to monitor their diabetes. This provides a possible solution to the dilemma of making a diabetes diagnosis based on only one abnormal HbA1c result, especially if the result is borderline positive and the individual is asymptomatic. The second HbA1c test (the first post-diagnosis monitoring HbA1c test) could be performed within a short time, before any changes in the management of the patient, and could be used to confirm the diabetes diagnosis.

Another point worth noting is that neither the WHO nor the ADS endorse a particular HbA1c range for diagnosing prediabetes. While the American Diabetes Association suggests that an HbA1c value of 39–47 mmol/mol (5.7%–6.4%) is equivalent to prediabetes as defined by glucose testing,10 this remains an area of ongoing debate, particularly concerning the appropriate lower HbA1c cut-point.

For many years, we have struggled to implement glucose-based diabetes screening and case detection algorithms. As shown by Marley and colleagues,7 HbA1c testing provides an opportunity to overcome many of the barriers to implementing effective screening programs. Although there are certain clinical limitations to HbA1c testing that doctors must bear in mind, the pragmatic approach adopted by Marley and colleagues would facilitate the earlier detection of diabetes in many people, and provide the opportunity to intervene earlier to reduce the personal, family and societal burden of diabetes.