hba1c in the diagnosis of type 2 diabetes: a systematic ... · 1 hba1c in the diagnosis of type 2...

1

HbA1c in the diagnosis of type 2 diabetes: a systematic review

Introduction

The use of HbA1c for diagnosis of type 2 diabetes is not currently recommended by

the World Health Organization (WHO) (WHO 2006). The reasons cited in the 2006

report included that HbA1c measurement was not widely available in many countries

throughout the world, global consistency in its measurement was problematic, and

that the HbA1c result is influenced by several factors including anaemia and

abnormalities of haemoglobin.

There is now renewed interest in HbA1c as a diagnostic criterion for diabetes.

Consequently this systematic review was undertaken to address this question.

Research Question

How does HbA1c perform in the diagnosis of type 2 diabetes based on the detection

and prediction of microvascular complications?

Methods

OBJECTIVE

To review best available evidence on the performance of HbA1c for the diagnosis of

diabetes, based on the detection and prediction of microvascular complications.

CRITERIA FOR CONSIDERING STUDIES FOR THIS REVIEW

Type of study

Cohort studies evaluating the association between HbA1c levels and prevalent or

incident microvascular complications.

Case-report, case-control and case-series studies and letters or commentaries were

excluded.

Type of participants Adults aged 18 years and older with or without diabetes.

Types of outcome measures

The following outcomes were included:

Main outcome

• HbA1c cut-point associated with prevalent or incident microvascular

complications associated with diabetes (e.g. retinopathy, microalbuminuria)

• Acceptable forms of analyzing data on this association including sensitivity and

specificity, ROC curve analysis, change point analysis, inspection of

decile/vigintile distribution, and inspection of continuous plots.

• Preference was given to studies using the most recent WHO diagnostic criteria,

however studies using older WHO or ADA diagnostic criteria were also included.

Other outcomes

2

• FPG and 2-h PG cut-points associated with prevalent or incident microvascular

complications (e.g. retinopathy, microalbuminuria)

• Sensitivity and specificity, ROC curve analysis, change point analysis, inspection

of decile/vigintile distribution, and inspection of continuous plots describing the

association between HbA1c, FPG or 2-h PG values and prevalent or incident

microvascular complications

Search methods for identification of studies

Electronic searches Databases were searched for relevant articles published between January 1990 and

September 2010. The January 1990 start date was selected because HbA1c

measurement was first developed in the late 70’s, did not become routinely used in

clinical practice until the late 80’s and the first reports relevant to this review were

published in the mid-90’s.

The following databases were searched:

• Medline

• Embase

• Pubmed

• Cinahl

• Psycinfo

• The Cochrane Library

A separate search strategy, specific for each electronic database was used for each

search. These searches can be found in Appendix 1.

Searching was carried out using a combination of keywords that cover all relevant

terminology for type 2 diabetes and the MESH terms HbA1c, type 2 diabetes,

diagnosis and complications. These searches were supplemented by reviewing

reference lists of relevant articles.

The relevance of articles was determined according to the inclusion and exclusion

criteria.

• Inclusion criteria require that the articles were conducted in humans (aged ≥ 18

years), contained cohorts with prevalent or incident cases of undiagnosed or

newly diagnosed type 2 diabetes, with diagnosis of diabetes based on the oral

glucose tolerance test (OGTT) or fasting plasma glucose (FPG) (using WHO

2006 or other established criteria); published in any language.

• Exclusion criteria were letters, commentaries, time series, case reviews or case-

control studies; all included participants had known diabetes.

METHODS OF THE REVIEW

Data Collection and analysis The inclusion of studies was assessed independently by two assessors. Articles were

only rejected on the initial screen if:

• the reviewer could determine from the title and abstract that the article was a time

series or case review or case-control study or letter or commentary;

3

• the study did not include measured HbA1c values

• the study did not report prevalent or incident microvascular complications

When a title/abstract could not be rejected with certainty, the full text of the article

was obtained for further evaluation.

Data abstraction was performed independently. Differences between reviewer’s

results were resolved by discussion and reanalysis of studies and by returning to the

relevant literature. A third reviewer was available to resolve any disagreement.

Assessing Study Quality and Level of Evidence Methodological quality of each study was assessed according to the Australian

National Health and Medical Research Council (NHMRC) criteria for assessing study

quality and grading the level of evidence (Appendix 2).

Quality assessment was not used as an exclusion criterion.

The GRADE (Grading of Recommendations Assessment, Development and

Evaluation) program was also used to generate summary of findings tables

(Schunemann et al. 2008).

Results The search strategy identified 9680 studies. The majority of these were found to be

irrelevant upon reading the title, requiring only 134 abstracts to be read. Of these, 11

met the inclusion criteria and were included in the review. A summary of reviewed

studies is detailed below and is summarised in the attached Tables.

HbA1c and the detection of prevalent microvascular complications McCance and colleagues (1994) performed a cross-sectional analysis of FPG, 2h

plasma glucose (PG) and HbA1c and the presence of microvascular complications

(retinopathy and nephropathy) associated with type 2 diabetes in Pima Indians aged

≥25 years (n=960) who were not receiving insulin or oral hypoglycaemic treatment at

baseline. The cut-points which achieved maximum sensitivity and specificity for

detecting retinopathy were ≥ 7.2 mmol/L for FPG (sensitivity 81%, specificity 80%),

≥ 13.0 mmol/L for 2h PG (sensitivity 88%, specificity 81%), and ≥ 7.0% for HbA1c

(sensitivity 78%, specificity 85%). Cut-points that were equivalent to the WHO 2h

PG criterion of ≥ 11.1 mmol/L (sensitivity 88%, specificity 76%) for detecting

retinopathy were ≥ 6.8 mmol/L for FPG (sensitivity 81%, specificity 77%) and ≥

6.1% for HbA1c (sensitivity 81%, specificity 77%). The prevalence of type 2 diabetes

detected using the optimal cut-points for FPG was 22%, 2 h PG 21% and for HbA1c

17%. The areas under the curve for nephropathy were not as good as those for

retinopathy.

Engelgau et al. (1997) performed a cross-sectional study of 1,018 Egyptians aged ≥

20 years to compare FPG, 2h PG and HbA1c for diagnosing type 2 diabetes and to

evaluate the performance of the WHO 1980 criteria. Of this population, 27% had

known diabetes (91% of whom were receiving antihyperglycaemic medication) and

8% had undiagnosed diabetes. Cut-points for each glycaemic measure were calculated

for OGTT defined diabetes as 1) the upper component of the fitted bimodal

distribution for each glycaemic measure, and 2) the presence of diabetic retinopathy.

4

The point of intersection of the lower and upper components that minimised

misclassification were ≥ 7.2 mmol/L for FPG, ≥ 11.5 mmol/L for 2h PG, and ≥ 6.7%

for HbA1c. The sensitivities for FPG, 2h PG and HbA1c were 84%, 90%, and 68%,

respectively; the specificities were all 100%. The prevalence of retinopathy increased

above the sixth decile for FPG values (median glucose 6.6 mmol/L in seventh decile)

and above the seventh decile for 2h PG (median glucose 14.4 mmol/L in eight decile)

and HbA1c (median HbA1c 7.6% in eight decile) values. When diabetic retinopathy

was used to define type 2 diabetes in the entire population, area under the receiver

operator characteristic curve (AROC) analysis revealed that both FPG (0.85) and 2h

PG (0.86) were superior to HbA1c (0.82; p < 0.01). In the total population, the

sensitivity and specificity for detecting diabetic retinopathy were approximately equal

for FPG, 2h PG and HbA1c cut-points of ≥ 7.8 mmol/l, ≥ 12.8 mmol/L, and ≥ 6.9%,

respectively.

In an analysis of NHANES III data on 2,821 individuals aged 40-74 years in whom

FPG, 2h PG and HbA1c were measured, all three measurements were strongly

associated with retinopathy (The Expert Committee on the Diagnosis and

Classification of Diabetes Mellitus 1997). The prevalence of type 2 diabetes increased

in the highest decile of each variable, corresponding to FPG ≥ 6.7 mmol/L, 2h PG ≥

10.8 mmol/L, and HbA1c ≥ 6.2%.

Miyazaki and colleagues (2004) compared FPG, 2h PG and HbA1c to diagnose type 2

diabetes based on the prevalence of retinopathy in a Japanese population of 1,637

subjects aged 40-79 years from the Hisayama study. Of these subjects, 2.3% had

diabetic retinopathy. All three measures were strongly associated with retinopathy.

The prevalence of retinopathy dramatically increased in the tenth decile of each

variable, corresponding to an FPG of ≥ 6.5 mmol/L, a 2h PG ≥ 11.0 mmol/L, and an

HbA1c of ≥ 5.8%. The prevalence of retinopathy in the tenth decile of FPG, 2h PG

and HbA1c was 16%, 20% and 20%, respectively. According to AROC analysis, the

optimal cut-points for the diagnosis of diabetes were 6.4 mmol/L for FPG, 11.1

mmol/L for 2h PG, and 5.7% for HbA1c. At these cut-points the three measurements

has identical sensitivity (87%) and similar specificity (87%-90%) for detecting type 2

diabetes. The AROC curve for detecting type 2 diabetes was not significantly

different between any of the three measurements (FPG 0.96, 2h PG 0.90, and HbA1c

0.95).

The association of FPG, 2h PG and HbA1c with retinopathy and microalbuminuria

was assessed by Tapp et al. (2006). Data were obtained from 2,182 participants with

retinal photographs and 2,389 with urinary albumin/creatinine results from the

AusDiab study (subjects aged ≥ 25 years). The prevalence of retinopathy in the first

eight deciles of FPG and HbA1c and the first nine deciles of 2 h PG was 7.2, 6.6, and

6.3%, respectively, showing no variation with increasing glucose or HbA1c (subjects

with known diabetes were excluded from these analyses). Above these levels, the

prevalence of retinopathy rose sharply to 18.6, 21.3, and 10.9%, respectively. The

thresholds for increased prevalence of retinopathy were ≥ 7.1 mmol/L for FPG, ≥ 6.1%

for HbA1c, and ≥ 13.1 mmol/L for 2h PG. The prevalence of microalbuminuria rose

more gradually across the deciles for each glycaemic measure. The thresholds were

less clear than for retinopathy, but were found at ≥ 7.2 mmol/L for FPG and ≥ 6.1%

for HbA1c, with no evidence of a threshold for 2h PG. For FPG the adjusted threshold

for retinopathy using a change point model was 8.5 mmol/L (95%CI 6.4-10.6%, p =

5

0.008) and for HbA1c ≥ 6.0% (95% CI 3.9-7.0%, p = 0.064). The association of 2h

PG and retinopathy was not assessed due to limited numbers, and there was no

significant thresholds observed for any measure of glycaemia with microalbuminuria

using change point models.

Ito and colleagues (2000a) evaluated FPG, 2h PG and HbA1c for the diagnosis of

diabetes based on the prevalence of retinopathy. The subjects were 12,208 Japanese

atomic-bomb survivors who underwent an OGTT between 1965 and 1997 (mean age

at initial test 59 years). The prevalence of retinopathy increased sharply and

significantly above the eighth decile with FPG (≥ 7.0 mmol/L), above the seventh

decile for 2h PG (≥ 11.0 mmol/L) and above the ninth decile of HbA1c (≥ 7.3%).

Wong and colleagues (2008) assessed data from three cross-sectional studies to

examine the relationship between FPG and retinopathy for the diagnosis of diabetes.

The three cohorts included 3,162 Australian subjects aged 45-97 years from the Blue

Mountains Eye Study (BMES), 2,182 Australian subjects aged 25-90 years from the

Australian Diabetes, Obesity and Lifestyle Study (AusDiab) and 6,079 US subjects

aged 45-84 years from the Multi-Ethnic Study of Atherosclerosis (MESA). The

prevalence of retinopathy was 11.5% in BMES, 9.6% in AusDiab and 15.8% in

MESA. Results indicate inconsistent evidence for a uniform glycaemic threshold for

retinopathy, with the suggestion of a continuous relationship. Across the three

cohorts, a FPG cut-point of ≥ 7.0 mmol/L had a low sensitivity ranging from 15-39%

for detecting retinopathy, with specificity between 81-96%. The AROC for FPG in

detecting retinopathy was low and ranged from 0.56 to 0.61. In a separate analysis,

the relationship between 2h PG and prevalent retinopathy was assessed in the

AusDiab cohort. A 2 hour plasma glucose cut-point of ≥ 11.1 mmol/L performed

worse than FPG in identifying prevalent retinopathy in this population, with a

sensitivity of 25%, specificity of 81% and AROC of 0.54. The authors also reported a

continuous relationship between prevalent retinopathy and glycated haemoglobin in

the MESA cohort, with change point models showing no evidence of a glycaemic

threshold.

The DETECT-2 collaboration conducted an analysis to determine whether there is a

glycaemic threshold for diabetic retinopathy (Colagiuri et al. Diabetes Care in press).

Three glycaemic measures, FPG, 2h PG and HbA1c, were examined. The analysis

included 12 studies from nine countries with a total of 47,364 participants aged 20-79

years with gradable retinal photographs. The prevalence of any retinopathy in people

with known diabetes was 23.1%, newly diagnosed diabetes 5.4%, impaired glucose

tolerance (IGT) 2.8%, impaired fasting glucose (IFG) 4.3% and normal glucose

tolerance (NGT) 4.0%. Based on visual inspection of vigintile distribution, there was

a glycaemic threshold for diabetes-specific retinopathy (moderate or more severe

retinopathy), at 6.4-6.8 mmol/L for FPG, 9.8-10.6 mmol/L for 2h PG and 6.4-6.8%

for HbA1c. When change point analyses with glycaemic measures plotted as the

continuous variable were used, no threshold was found for any measure of glycaemia

for diabetes-specific retinopathy. Based on ROC analyses, the optimal cut-points for

detecting diabetes-specific retinopathy in all subjects were 6.5 mmol/L for FPG, 12.4

mmol/L for 2 h PG and 6.3% for HbA1c. At these cut-points the AROCs, sensitivities

and specificities were 0.87, 82% and 81% for FPG; 0.89, 83% and 83% for 2h PG;

and 0.90, 86% and 86% for HbA1c.

6

HbA1c and incident microvascular complications A recent study by Massin and colleagues (in press, Archives of Ophthalmology)

compared the predictive values of baseline HbA1c and FPG for the development of

retinopathy over 10 years in 700 French subjects (aged 30-65 years at entry) from the

DESIR study. Of the study population, 235 had diabetes (treatment of FPG ≥ 7.0

mmol/L at least once over the preceding nine years), 238 always had NGT, and 227

had IFG at least once. The 44 subjects with retinopathy at 10 years had higher

baseline mean HbA1c (6.4 ± 1.6% vs. 5.7 ± 0.7%) and FPG (7.2 ± 2.7 mmol/L vs. 5.9

± 1.2 mmol/L) than those without retinopathy (both p < 0.0001). The 10-year

prevalence of retinopathy was 3.6% in the entire population and 16% for those with

HbA1c ≥ 6.5% and FPG ≥ 6.5 mmol/L. The 10-year prevalence of retinopathy was

3.3% for HbA1c < 6.0% and 6.8% for those with a higher HbA1c. An HbA1c of 6.0%

had a positive predictive value (PPV) of 6.8%, a negative predictive value (NPV) of

97%, a sensitivity of 16%, a specificity of 97%, and a positive likelihood ratio (PLR)

of 2.0 for 10-year retinopathy. For an HbA1c of 6.5%, these values were 15.9%, 97%,

7.9%, 97% and 2.4. For an FPG of 6.0 mmol/L these values were 8.6%, 97%, 27%,

90% and 2.6, while for a FPG of 6.5 mmol/L they were 17.4%, 97%, 21%, 96% and

5.7. A threshold above which retinopathy increased could not be determined from

these results due the small sample size and low frequency of 10-year retinopathy.

Van Leiden and colleagues (2003) evaluated the effect of HbA1c, among other risk

factors, on the incidence of retinopathy in 233 people aged 50-74 years with normal

and abnormal glucose metabolism from the Hoorn Study. Average follow-up was 9.4

years (range 7.9-11.0 years). The cumulative incidences of retinopathy among those

with normal, impaired, and diabetic glucose metabolism were 7.3%, 13.6%, and

17.5%, respectively. The cumulative incidence increased from 6.0% for those in the

lowest tertile of HbA1c to 20.7% for those in the highest tertile (p = 0.005 for trend).

The crude odds ratio for retinopathy were 2.01 and 2.71 for individuals with impaired

glucose metabolism and those with type 2 diabetes, respectively, compared with

individuals with normal glucose metabolism. The adjusted odds ratio for retinopathy

was 3.29 (95%CI 1.11-9.72) for the highest tertile of HbA1c at baseline. Limiting this

analysis to those without type 2 diabetes, the adjusted odds ratio for retinopathy in the

highest tertile of baseline HbA1c was 3.54 (0.94-13.37). Baseline HbA1c was

significantly higher in those who developed retinopathy at follow-up (6.1 ± 1.0%)

compared with those who did not (5.6 ± 1.0%, p = 0.03).

Prospective data were also reported by the McCance et al. (1994) on the development

to microvascular complications. However, as the data involved a combination of

measurement of HbA1 and measurement of HbA1c, it was considered inappropriate

for inclusion in this review.

7

Summary

1. The major objective of diagnosing diabetes is to prevent premature mortality and

complication-related morbidity. Therefore it seems logical to consider diagnosis in

terms of risk of complications.

2. Diagnostic criteria would ideally be derived from a study of outcomes and

complications in an untreated prospective cohort measuring different potential

diagnostic criteria at baseline. Alternatively outcomes could be compared with

different diagnostic criteria in intervention studies. A sub-group analysis of the

ADDITION study might have the power to examine this.

3. In the absence of the above information, the relationship of complications

(diabetes-specific) with direct or indirect measures of glucose can be examined,

either prospectively or in cross-sectional analysis.

4. Most of the data of the relationship of measures of glycaemia and retinopathy are

derived from cross-sectional studies. HbA1c levels associated with retinopathy

ranged from 5.8-7.3%. The DETECT-2 analysis pooled data from 47,364 people

and reported an HbA1c of approximately 6.5% as the threshold for diabetes-

specific retinopathy.

5. The DESIR study examined FPG and HbA1c and 10-year incident retinopathy. A

threshold above which retinopathy increased could not be determined due to small

sample and low frequency of 10-year retinopathy. An HbA1c of 6.5% had a PPV

of 15.9%, NPV of 97%, sensitivity of 7.9%, and specificity of 97%.

Acknowledgements

Funding for the systematic review was provided by the World Health Organization.

8

References

Colagiuri, S., C. M. Y. Lee, T. Y. Wong, B. Balkau, J. Shaw and K. Borch-Johnsen

(In press, Diabetes Care). "Is there a glycemic threshold for diabetic

retinopathy?".

Engelgau, M. M., T. J. Thompson, W. H. Herman, J. P. Boyle, R. E. Aubert, S. J.

Kenny, A. Badran, E. S. Sous and M. A. Ali (1997). "Comparison of fasting

and 2-hour glucose and HbA1c levels for diagnosing diabetes. Diagnostic

criteria and performance revisited." Diabetes Care 20(5): 785-791.

Ito, C., R. Maeda, S. Ishida, H. Harada, N. Inoue and H. Sasaki (2000a). "Importance

of OGTT for diagnosing diabetes mellitus based on prevalence and incidence

of retinopathy." Diabetes Res Clin Pract 49(2-3): 181-186.

Massin, P., C. Lange, J. Tichet, S. Vol, A. Erginay, M. Cailleau, E. Eschwege and B.

Balkau (In press, Archives of Ophthalmology). "HbA1c and fasting plasma

glucose as predictors of retinopathy at ten years: the French D.E.S.I.R. Study."

McCance, D. R., R. L. Hanson, M. A. Charles, L. T. Jacobsson, D. J. Pettitt, P. H.

Bennett and W. C. Knowler (1994). "Comparison of tests for glycated

haemoglobin and fasting and two hour plasma glucose concentrations as

diagnostic methods for diabetes." BMJ 308(6940): 1323-1328.

Miyazaki, M., M. Kubo, Y. Kiyohara, K. Okubo, H. Nakamura, K. Fujisawa, Y. Hata,

S. Tokunaga, M. Iida, Y. Nose and T. Ishibashi (2004). "Comparison of

diagnostic methods for diabetes mellitus based on prevalence of retinopathy in

a Japanese population: the Hisayama Study." Diabetologia 47(8): 1411-1415.

Schunemann, H. J., A. D. Oxman, J. Brozek, P. Glasziou, R. Jaeschke, G. E. Vist, J.

W. Williams, Jr., R. Kunz, J. Craig, V. M. Montori, P. Bossuyt and G. H.

Guyatt (2008). "Grading quality of evidence and strength of recommendations

for diagnostic tests and strategies." BMJ 336(7653): 1106-1110.

Tapp, R. J., P. Z. Zimmet, C. A. Harper, M. P. de Courten, D. J. McCarty, B. Balkau,

H. R. Taylor, T. A. Welborn and J. E. Shaw (2006). "Diagnostic thresholds for

diabetes: the association of retinopathy and albuminuria with glycaemia."

Diabetes Res Clin Pract 73(3): 315-321.

The Expert Committee on the Diagnosis and Classification of Diabetes Mellitus

(1997). "Report of the Expert Committee on the Diagnosis and Classification

of Diabetes Mellitus." Diabetes Care 20(7): 1183-1197.

van Leiden, H. A., J. M. Dekker, A. C. Moll, G. Nijpels, R. J. Heine, L. M. Bouter, C.

D. Stehouwer and B. C. Polak (2003). "Risk factors for incident retinopathy in

a diabetic and nondiabetic population: the Hoorn study." Arch Ophthalmol

121(2): 245-251.

WHO (2006). Definition and diagnosis of diabetes mellitus and intermediate

hyperglycemia. Geneva, World Health Organization.

Wong, T. Y., G. Liew, R. J. Tapp, M. I. Schmidt, J. J. Wang, P. Mitchell, R. Klein, B.

E. Klein, P. Zimmet and J. Shaw (2008). "Relation between fasting glucose

and retinopathy for diagnosis of diabetes: three population-based cross-

sectional studies." Lancet 371(9614): 736-743.

9

Table 1 HbA1c and prevalent microvascular complications – study characteristics

Author, year

and country

Subject no

and gender

(M/F)

Age

(years)

Prevalence of

diabetes (%)

Inclusion/ exclusion

criteria

HbA1c test method Glucose

method

Diabetes

diagnostic

criteria

Blood

sample

Colagiuri et

al. (in press,

Diabetes Care), International

47,364

22,127/

25,237

20-79 14.3 Age 20-79 years with

gradable retinal

photographs and data for at

least one measure of

glycaemia (FPG, 2h PG or

HbA1c)

Varies by study Varies by

study

WHO 1999 Varies by

study

Engelgau et

al. (1997),

Egypt

1,018

417/601

Mean:

45

35.6 ≥ 20 years old, Egyptian

(note: includes people with

known diabetes, many of

whom were receiving anti-

hyperglycaemic treatment)

Affinity chromatography

(Pierce Scientific)

CV: 6.0%

Glucose

oxidase

WHO 1980 Capillary

blood and

Serum

glucose

Expert

Committee

(1997), US

2,821

NR

40-74 NR NR NR NR NR NR

Ito et al.

(2000a),

Japan

12,208

6,440/5,768

58.6 ±

11.6

NR Japanese atomic bomb

survivors

HPLC Glucose

oxidase

WHO 1999 Venous

plasma

McCance et

al. (1994), US

– Pima Indian

960

384/576

≥ 25 14-26 depending

on measurement

and cut-point

(26.3 for 2-h PG ≥

11.1 mmol/L)

Pima Indian subjects ≥ 25

years of age not receiving

insulin or oral

hypoglycaemic treatment at

baseline

HPLC Potassium

ferricyanide

WHO 1985 Venous

plasma

Miyazaki et

al. (2004),

Japan

1,637 40-79 21-23 depending

on measurement

(21 for 2-h PG ≥

11.1 mmol/L)

Age 40-79 years, not

receiving insulin treatment

(note: includes people

receiving oral anti-

hyperglycaemic treatment)

HPLC Glucose

oxidase

WHO 1999 Venous

plasma

Tapp et al.

(2006),

Australia

2,476

1,114/1,362

Mean:

59

34.5 Age ≥ 25 years Boronate affinity HPLC

(Bio-Rad Variant

Haemoglobin Testing

System)

CV: < 2%

Olympus

AU600

analyser

WHO 1999 Venous

plasma

2-h PG = 2 hour plasma glucose; ADA = American Diabetes Association; BMI = body mass index; CV = coefficient of variation; HPLC = high-performance liquid

chromatography; NR = not reported; WHO = World Health Organization.

10

Table 2 HbA1c, FPG and 2-h PG cut-points associated with prevalent microvascular complications HbA1c FPG 2-h PG Study Complication

Optimum

cut-point

(%)

AROC Sensitivity

(%)

Specificity

(%)

Optimum

cut-point

(mmol/L)

AROC Sensitivity

(%)

Specificity

(%)

Optimum

cut-point

(mmol/L)

AROC Sensitivity

(%)

Specificity

(%)

Retinopathy

(ROC curve

analysis)

≥6.3 0.90 86 86 ≥6.5 0.87 82 81 ≥12.4 0.89 83 83

Colagiuri et

al. (in press, Diabetes Care)

Retinopathy

(visual inspection

of decile

distribution)

6.4-6.8 NR NR NR 6.4-6.8 NR NR NR 9.8-10.6 NR NR NR

Bi-modal:

- Entire

population

≥6.7

NR

68

100

≥7.2

NR

84

100

≥11.5

NR

90

100

Engelgau et

al. (1997)

Retinopathy#:

- Entire

population

≥7.6

0.82

NR

NR

≥6.6

0.85*

NR

NR

≥14.4

0.86*

NR

NR

Expert

Committee,

(1997)

Retinopathy

≥6.2 NR NR NR ≥6.7 NR NR NR ≥10.8 NR NR NR

Ito et al.

(2000a)

Retinopathy ≥7.3 NR NR NR ≥7.0 NR NR NR ≥11.0 NR NR NR

Retinopathy ≥7.0 NR 78 85 ≥7.2 NR 81 80 ≥13.0 NR 88 81

WHO equivalent ≥6.1 NR 81 77 ≥6.8 NR 81 77 ≥11.1 NR 88 76

McCance et

al. (1994)

ROC curve

analysis ≥5.7 0.95 87 90 ≥6.4 0.96 87 87 ≥11.1 0.90 87 90

Miyazaki et

al. (2004)



Microalbuminuria ≥6.1 NR NR NR ≥7.2 NR NR NR NR NR NR NR

Retinopathy§ ≥6.0 NR NR NR ≥8.5 NR NR NR NR NR NR NR

Tapp et al.

(2006)

Microalbuminuria NIL - - - NIL - - - NR NR NR NR

* Significantly different from HbA1c (p < 0.01); # Median decile value; § By change point analysis. 2-h PG = 2 hour plasma glucose; AROC = Area under the receiver

operator characteristic curve; FPG = fasting plasma glucose; NR = Not reported; ROC = receiver operator characteristic; WHO = World Health Organization.

11

Table 3 HbA1c and incident microvascular complications – study characteristics

Author,

year and

country

Subject no

and gender

(M/F)

Age

(years)

Follow-

up

(years)

Incidence of

diabetes (%)

Inclusion/ exclusion criteria HbA1c test method Glucose

method

Diabetes

diagnostic

criteria

Blood

sample

Massin et al.

(in press,

Archives of

Ophthalmol),

France

700

504/196

30-65 10 NR

Retinopathy:

6.3

Aged 30-65 years. Excluded if

uninterpretable retinal photographs

HPLC (Hitachi/Merck-

VWR) or

DCA 2000 automated

immunoassay system

(Bayer Diagnostics)

Glucose

oxidase

NR Venous

plasma

Van Leiden

et al. (2003),

Netherlands

233

124/109

50-74 9.4 NR

Retinopathy:

11.6

Aged 50-74 years from Hoorn,

Netherlands.

HPLC (Modular

Diabetes Monitoring

system; Bio-Rad)

Normal range: 4.3-6.1%

Glucose

dehydrogenase

WHO 1999 Venous

plasma

HPLC = high-performance liquid chromatography; NR = not reported; WHO = World Health Organization.

12

Table 4 HbA1c and FPG cut-points associated with incident diabetes complications

HbA1c FPG Study Complication

Optimum

cut-point

(%)

AROC Sensitivity

(%)

Specificity

(%)

Optimum

cut-point

(mmol/L)

AROC Sensitivity

(%)

Specificity

(%)

Massin et al.

(in press,

Archives of

Ophthalmol)

Retinopathy ≥ 6.0 NR 16 97 ≥ 6.5 NR 21 96

AROC = Area under the receiver operator characteristic curve; FPG = fasting plasma glucose; NR = Not reported.

13

Table 5. Evidence table for HbA1c and prevalent microvascular complications

Evidence Level of Evidence

Author (year),

population Level Study Type

Quality Rating Magnitude of

effect rating

Relevance

Rating

Colagiuri et al. (in press,

Diabetes Care),

International

N/A Pooled

Analysis High High High

Engelgau et al. (1997),

Egypt III-2 Cohort Medium High High

Expert Committee

(1997), US III-2 Cohort Medium Medium High

Ito et al. (2000a), Japan II Cohort High High High

McCance et al. (1994),

US – Pima Indian II Cohort High High High

Miyazaki et al. (2004),

Japan III-2 Cohort High High High

Tapp et al. (2006),

Australia III-2 Cohort High Medium High

Table 6. Evidence table for HbA1c and incident microvascular complications

Evidence Level of Evidence

Author (year),

population Level Study Type

Quality Rating Magnitude of

effect rating

Relevance

Rating

Massin et al. (in press,

Archives of Ophthalmology),

France

II Prospective

Cohort High Medium High

Van Leiden et al. (2003),

Netherlands II

Prospective

Cohort High Medium High

14

Table 7. GRADE table for HbA1c and detection of prevalent microvascular complications

Factors that may decrease quality of evidence

Outcome No. of

studies

Study

design Limitations Indirectness Inconsistency Imprecision Reporting bias

Final

quality Effect per 10001 Importance

True positives

(patients with prevalent complications)

3 studies2

(31,797 patients)

Observational None3 None None None Unlikely ⊕⊕⊕O

moderate

Prev 80%: 672

Prev 40%: 336 Prev 10%: 84

IMPORTANT

True negatives (patients

without prevalent

complications)

3

(31,797

patients)


moderate

Prev 80%: 172

Prev 40%: 516

Prev 10%: 774

IMPORTANT

False positives (patients

incorrectly classified as

having prevalent

complications)

3

(31,797

patients)


moderate

Prev 80%: 28

Prev 40%: 84

Prev 10%: 126

IMPORTANT

False negatives (patients

incorrectly classified as not having prevalent

complications)

3

(31,797

patients)


moderate

Prev 80%: 128

Prev 40%: 64

Prev 10%: 16

IMPORTANT

Inconclusive 4 4 studies (19,142

patients)

Observational – – – – – – – IMPORTANT

Cost Not reported – – – – – – – – NOT

RELEVANT

1 Based on combined sensitivity of 84% and specificity of 86%

2 One study contained pooled data from 8 studies with 29,819 participants

3 Although not a serious limitation, one study oversampled people with known diabetes

4 These 4 studies did not report information on sensitivity and specificity of HbA1c for predicting prevalent microvascular complications

15

Table 8. GRADE table for HbA1c and incident microvascular complications

Factors that may decrease quality of evidence

Outcome No. of

studies

Study

design Limitations Indirectness Inconsistency Imprecision Reporting bias

Final

quality Effect per 10002 Importance

True positives

(patients with incident complications)

1 study

(700 patients) Observational None None N/A2

Not

assessable3 Unlikely

⊕⊕OO

low

Prev 80%: 128

Prev 40%: 64 Prev 10%: 16

IMPORTANT

True negatives (patients

without incident

complications)

1


Not


⊕⊕OO

low

Prev 80%: 194

Prev 40%: 582

Prev 10%: 873

IMPORTANT

False positives (patients

incorrectly classified as

having incident

complications)

1


Not


⊕⊕OO

low

Prev 80%: 6

Prev 40%: 18

Prev 10%: 27

IMPORTANT

False negatives (patients

incorrectly classified as not having incident

complications)

1


Not


⊕⊕OO

low

Prev 80%: 672

Prev 40%: 336

Prev 10%: 84

IMPORTANT

Inconclusive 4

1 study (233 patients)

Observational – – – – – – – IMPORTANT

Cost Not reported – – – – – – – – NOT

RELEVANT

2 Based on combined sensitivity of 16% and specificity of 97%

2 Imprecision could not be assessed as confidence intervals were not reported

3 Inconsistency is not applicable with data from only one study 4 This study did not report information on sensitivity and specificity of HbA1c for predicting incident microvascular complications

16

Appendix 1

Search for HbA1c in the diagnosis of diabetes (search covers both sections: incident and

prevalent complications associated with HbA1c)

Search 1: Database: Ovid MEDLINE

Search Strategy:

--------------------------------------------------------------------------------

1 Diabetes Mellitus, Type 2/ (62685)

2 (type 2 diabetes or type II diabetes).tw. (42266)

3 (non?insulin dependent diabetes or NIDDM).tw. (7555)

4 1 or 2 or 3 (75337)

5 Hemoglobin A, Glycosylated/ (16909)

6 hba1c.tw. (8615)

7 h?emoglobin A1c.tw. (3166)

8 Glyco?h?emoglobin.tw. (653)

9 Glycated h?emoglobin.tw. (2802)

10 Glycosylated h?emoglobin.tw. (5302)

11 5 or 6 or 7 or 8 or 9 or 10 (24975)

12 Diagnosis/ (15662)

13 Diagnostic Tests, Routine/ (5441)

14 diagnos$.tw. (1271525)

15 exp Diabetes Complications/ (87841)

16 complication$.tw. (456183)

17 retinopath$.tw. (23219)

18 12 or 13 or 14 or 15 or 16 or 17 (1733665)

19 4 and 11 and 18 (4191)

20 limit 19 to (humans and yr="1990 - 2010") (3973)

***************************

17

Search 2 – Embase

No. Query Results

#22 #4 AND #12 AND #21 AND [humans]/lim AND [1990-

2010]/py 4132

#21 #13 OR #14 OR #15 OR #16 OR #17 OR #18 OR #19 OR

#20 2677720

#20 'retinopathy':ab,ti OR 'retinopathies':ab,ti 28677

#19 'complication':ab,ti OR 'complications':ab,ti 569736

#18 'diabetic retinopathy'/de 22102

#17 'diagnosis':ab,ti OR 'diagnostic':ab,ti OR 'diagnosed':ab,ti OR

'diagnoses':ab,ti 1545712

#16 'laboratory diagnosis'/de 35794

#15 'diagnostic procedure'/de 60017

#14 'diagnostic test'/de 46436

#13 'diagnosis'/de 805045

#12 #5 OR #6 OR #7 OR #8 OR #9 OR #10 OR #11 36114

#11 'glycosylated hemoglobin'/de 10917

#10 'glycosylated haemoglobin':ab,ti OR 'glycosylated

hemoglobin':ab,ti 6271

#9 'glycated haemoglobin':ab,ti OR 'glycated hemoglobin':ab,ti 3457

#8 'glycohaemoglobin':ab,ti OR 'glycohemoglobin':ab,ti 801

#7 'haemoglobin a1c':ab,ti OR 'hemoglobin a1c':ab,ti 2600

#6 hba1c:ab,ti 8781

#5 'hemoglobin a1c'/de 21712

#4 #1 OR #2 OR #3 103193

#3 'non insulin dependent diabetes':ab,ti OR 'noninsulin

dependent diabetes':ab,ti OR 'niddm':ab,ti 13423

#2 'type 2 diabetes':ab,ti OR 'type ii diabetes':ab,ti 55623

#1 'non insulin dependent diabetes mellitus'/de 87826

18

Search 3 – Pubmed

Search History

Search Queries Result

#18 Search #4 AND #10 AND #17 Limits: Humans, Publication Date from 1990 to 2010 7191

#17 Search #11 or #12 or #13 or #14 or #15 or #16 6064084

#16 Search retinopath*[Title/Abstract] 23768

#15 Search complication*[Title/Abstract] 474432

#14 Search diabetes complications[MeSH Terms] 86714

#13 Search diagnos*[Title/Abstract] 1320078

#12 Search diagnostic tests, routine[MeSH Terms] 5308

#11 Search diagnosis[MeSH Terms] 5190551

#10 Search #5 or #6 or #7 or #8 or #9 24308

#9 Search glycosylated haemoglobin or glycosylated hemoglobin[Title/Abstract] 18889

#8 Search glycated haemoglobin or glycated hemoglobin[Title/Abstract] 17668

#7 Search glycohaemoglobin or glycohemoglobin[Title/Abstract] 670

#6 Search hba1c[Title/Abstract] 9034

#5 Search hba1c[MeSH Terms] 16240

#4 Search #1 or #2 or #3 87454

#3 Search non?insulin dependent diabetes or niddm[Title/Abstract] 76711

#2 Search type 2 diabetes or type II diabetes[Title/Abstract] 75342

#1 Search type 2 diabetes[MeSH Terms] 60587

19

Search 4 – Cinahl

# Query Limiters/Expanders Results

S20 S4 and S11 and S18

Limiters - Published Date

from: 19900101-20101231;

Human

Search modes -

Boolean/Phrase

512

S19 S4 and S11 and S18 Search modes -

Boolean/Phrase 703

S18 S12 or S13 or S14 or S15 or S16 or S17 Search modes -

Boolean/Phrase 148780

S17 TI retinopath* or AB retinopath* Search modes -

Boolean/Phrase 1585

S16 TI complication* or AB complication* Search modes -


S15 TI diagnos* or AB diagnos* Search modes -


S14 (MH "Diagnosis, Laboratory") Search modes -

Boolean/Phrase 6119

S13 (MH "Diagnostic Tests, Routine") Search modes -

Boolean/Phrase 783

S12 (MH "Diagnosis") Search modes -

Boolean/Phrase 2056

S11 S5 or S6 or S7 or S8 or S9 or S10 Search modes -

Boolean/Phrase 5717

S10 TI ( glycosylated haemoglobin or glycosylated hemoglobin ) or

AB ( glycosylated haemoglobin or glycosylated hemoglobin )

Search modes -

Boolean/Phrase 837

S9 TI ( glycated haemoglobin or glycated hemoglobin ) or AB (

glycated haemoglobin or glycated hemoglobin )

Search modes -

Boolean/Phrase 390

20

S8 TI ( glycohaemoglobin or glycohemoglobin ) or AB (

glycohaemoglobin or glycohemoglobin )

Search modes -

Boolean/Phrase 66

S7 TI ( (haemoglobin a1c or hemoglobin a1c) ) or AB (

(haemoglobin a1c or hemoglobin a1c) )

Search modes -

Boolean/Phrase 816

S6 TI hba1c or AB hba1c Search modes -

Boolean/Phrase 2208

S5 (MH "Hemoglobin A, Glycosylated") Search modes -

Boolean/Phrase 3908

S4 S1 or S2 or S3 Search modes -


S3

TI ( (non insulin dependent diabetes or noninsulin dependent

diabetes or non-insulin dependent diabetes or niddm) ) or AB (

(non insulin dependent diabetes or noninsulin dependent

diabetes or non-insulin dependent diabetes or niddm) )

Search modes -

Boolean/Phrase 839

S2 TI ( (type 2 diabetes or type II diabetes) ) or AB ( (type 2

diabetes or type II diabetes) )

Search modes -


S1 (MH "Diabetes Mellitus, Non-Insulin-Dependent") Search modes -


21

Search 5 – Psycinfo

Database: PsycINFO

Search Strategy:

--------------------------------------------------------------------------------

1 (type 2 diabetes or type II diabetes).tw. (2117)

2 (non?insulin dependent diabetes or NIDDM).tw. (148)

3 1 or 2 (2248)

4 hba1c.tw. (493)

5 h?emoglobin A1c.tw. (199)

6 Glyco?h?emoglobin.tw. (19)

7 Glycated h?emoglobin.tw. (102)

8 Glycosylated h?emoglobin.tw. (311)

9 4 or 5 or 6 or 7 or 8 (883)

10 diagnosis/ (24273)

11 diagnos$.tw. (177767)

12 "Complications (Disorders)"/ (756)

13 complication$.tw. (11553)

14 retinopath$.tw. (379)

15 10 or 11 or 12 or 13 or 14 (189699)

16 3 and 9 and 15 (106)

17 limit 16 to (human and yr="1990 - 2010") (104)

***************************

22

Search 6 – Cochrane Library Current Search

ID Search Hits

#1 MeSH descriptor Diabetes Mellitus, Type 2 explode all trees 6415

#2 (type 2 diabetes or type II diabetes):ti,ab,kw 8982

#3 (non insulin dependent diabetes or non insulin dependent diabetes or non-insulin dependent diabetes or niddm):ti,ab,kw

2020

#4 (#1 OR #2 OR #3) 9548

#5 MeSH descriptor Hemoglobin A, Glycosylated, this term only 2656

#6 (hba1c):ti,ab,kw 1642

#7 (haemoglobin a1c or hemoglobin a1c):ti,ab,kw 788

#8 (glycohaemoglobin or glycohemoglobin):ti,ab,kw 69

#9 (glycated haemoglobin or glycated hemoglobin):ti,ab,kw 476

#10 (glycosylated haemoglobin or glycosylated hemoglobin):ti,ab,kw 3168

#11 (#5 OR #6 OR #7 OR #8 OR #9 OR #10) 4352

#12 MeSH descriptor Diagnosis, this term only 65

#13 MeSH descriptor Diagnostic Tests, Routine, this term only 293

#14 (diagnos*):ti,ab,kw 66662

#15 MeSH descriptor Diabetes Complications explode all trees 3896

#16 (complication*):ti,ab,kw 71382

#17 (retinopath*):ti,ab,kw 1861

#18 (#12 OR #13 OR #14 OR #15 OR #16 OR #17) 129125

#19 (#4 AND #11 AND #18), from 1990 to 2010 1141

23

Appendix 2

NHMRC Evidence Hierarchy: designations of ‘levels of evidence’ according to type of research question (NHMRC 2007)

Level Intervention Diagnostic accuracy Prognosis Aetiology Screening Intervention

I

A systematic review of level II

Studies

A systematic review of level

II studies


II studies


II studies

A systematic review of level II

studies

II A randomised controlled trial A study of test accuracy with: an independent, blinded comparison with a valid reference standard,

among consecutive persons with a defined clinical presentation

A prospective cohort study

A prospective cohort study A randomised controlled trial

III-1 A pseudorandomised controlled trial

(i.e. alternate allocation or some

other method)

A study of test accuracy with: an independent, blinded comparison with a valid reference standard,

among non-consecutive persons with a defined clinical presentation

All or none All or none A pseudorandomised

controlled trial

(i.e. alternate allocation or

some other method)

II-2 A comparative study with

concurrent controls:

▪ Non-randomised,

experimental trial

▪ Cohort study

▪ Case-control study

▪ Interrupted time series with a

control group

A comparison with reference

standard that does not meet

the criteria required for

Level II and III-1 evidence

Analysis of prognostic

factors amongst persons in

a single arm of a

randomised controlled trial

A retrospective cohort study A comparative study with


▪ Non-randomised,

experimental trial

▪ Cohort study

▪ Case-control study

III-3 A comparative study without


▪ Historical control study

▪ Two or more single arm

study ▪ Interrupted time series without a

parallel control group

Diagnostic case-control

study

A retrospective cohort study A case-control study A comparative study without


▪ Historical control study

▪ Two or more single arm

study

IV Case series with either post-test

or pre-test/post-test outcomes Study of diagnostic

yield (no reference

standard)

Case series, or cohort study of

persons at different stages of

disease

A cross-sectional study or

case series

Case series

(Source: NHMRC 2007)

24

Study Assessment Criteria

I. Study quality criteria

Systematic reviews 1. Were the questions and methods clearly stated?

2. Is the search procedure sufficiently rigorous to identify all relevant studies?

3. Does the review include all the potential benefits and harms of the

intervention?

4. Does the review only include randomised controlled trials?

5. Was the methodological quality of primary studies assessed?

6. Are the data summarised to give a point estimate of effect and confidence

intervals?

7. Were differences in individual study results adequately explained?

8. Is there an examination of which study population characteristics (disease

subtypes, age/sex groups) determine the magnitude of effect of the

intervention?

9. Were the reviewers' conclusions supported by data cited?

10. Were sources of heterogeneity explored?

Randomised controlled trials 1. Were the setting and study subjects clearly described?

2. Is the method of allocation to intervention and control groups/sites

independent of the decision to enter the individual or group in the study ?

3. Was allocation to study groups adequately concealed from subjects,

investigators and recruiters including blind assessment of outcome?

4. Are outcomes measured in a standard, valid and reliable way?

5. Are outcomes measured in the same way for both intervention and control

groups?

6. Were all clinically relevant outcomes reported?

7. Are factors other than the intervention e.g. confounding factors, comparable

between intervention and control groups and if not comparable, are they

adjusted for in the analysis?

8. Were >80% of subjects who entered the study accounted for at its

conclusion?%

9. Is the analysis by intention to intervene (treat)?

10. Were both statistical and clinical significance considered?

11. Are results homogeneous between sites? (Multi-centre/multi-site studies only).

Cohort studies 1. Are study participants well-defined in terms of time, place and person?

2. What percentage (%) of individuals or clusters refused to participate?

3. Are outcomes measured in a standard, valid and reliable way?

4. Are outcomes measured in the same way for both intervention and control

groups?

5. Was outcome assessment blind to exposure status?

6. Are confounding factors, comparable between the groups and if not

comparable, are they adjusted for in the analysis?

7. Were >80% of subjects entered accounted for in results and clinical status

described?

25

8. Was follow-up long enough for the outcome to occur

9. Was follow-up complete and were there exclusions from the analysis?

10. Are results homogeneous between sites? (Multicentre/multisite studies only).

Case-control studies 1. Was the definition of cases adequate?

2. Were the controls randomly selected from the source of population of the

cases?

3. Were the non-response rates and reasons for non-response the same in both

groups?

4. Is possible that over-matching has occurred in that cases and controls were

matched on factors related to exposure?

5. Was ascertainment of exposure to the factor of interest blinded to case/control

status?

6. Is exposure to the factor of interest measured in the same way for both case

and control groups in a standard, valid and reliable way (avoidance of recall

bias)?

7. Are outcomes measured in a standard, valid and reliable way for both case and

control groups?

8. Are the two groups comparable on demographic characteristics and important

potential confounders? and if not comparable, are they adjusted for in the

analysis?

9. Were all selected subjects included in the analysis?

10. Was the appropriate statistical analysis used (matched or unmatched)?

11. Are results homogeneous between sites? (Multicentre/multisite studies only).

Diagnostic accuracy studies 1. Has selection bias been minimised

2. Were patients selected consecutively?

3. Was follow-up for final outcomes adequate?

4. Is the decision to perform the reference standard independent of the test results

(ie avoidance of verification bias)?

5. If not, what per cent were not verified?

6. Has measurement bias been minimised?

7. Was there a valid reference standard?

8. Are the test and reference standards measured independently (ie blind to each

other)

9. Are tests measured independently of other clinical and test information?

10. If tests are being compared, have they been assessed independently (blind to

each other) in the same patients or done in randomly allocated patients?

11. Has confounding been avoided?

12. If the reference standard is a later event that the test aims to predict, is any

intervention decision blind to the test result?

(Sources: adapted from NHMRC1999, NHMRC 2000a, NHMRC 2000b, Liddle et al

96; Khan et al 2001)

26

Study quality – Rating The following was used to rate the quality of each study against the study type criteria

listed above.

High: all or all but one of the criteria were met

Medium: 2 or 3 of the criteria were not met

Low: 4 or more of the criteria were not met

27

II. Classifying magnitude of the effect

Ranking Statistical significance Clinical importance of

benefit

High Difference is statistically

significant

AND There is a clinically

important benefit for the

full range of estimates

defined by the confidence

interval.

Medium Difference is statistically

significant

AND The point estimate of effect

is clinically important

BUT the confidence

interval includes some

clinically unimportant

effects

Low Difference is statistically

significant|

OR

Difference is not statistically

significant (no effect) or

shows a harmful effect

AND

AND

The confidence interval

does not include any

clinically important effects

The range of estimates

defined by the confidence

interval includes clinically

important effects. (Source: adapted from the NHMRC classification (NHMRC 2000b)

III. Classifying the relevance of the evidence

Ranking Relevance of the evidence

High Evidence of an effect on patient-relevant clinical outcomes,

including benefits and harms, and quality of life and survival

Or

Evidence of an effect on a surrogate outcome that has been shown

to be predictive of patient-relevant outcomes for the same

intervention

Medium

Evidence of an effect on proven surrogate outcomes but for a

different intervention

Or

Evidence of an effect on proven surrogate outcomes but for a

different intervention and population

Low

Evidence confined to unproven surrogate outcomes.

(Source: adapted from the NHMRC classification (NHMRC 2000b)

hba1c in the diagnosis of type 2 diabetes: a systematic ... · 1 hba1c in the diagnosis of type 2...

Documents