Screening for Diabetes and Pre-Diabetes With Proposed A1C-Based Diagnostic Criteria

There is substantial interindividual variation in A1C at the same levels of glycemia.

“Moreover, there is substantial interindividual variation in A1C at the same levels of glycemia (20,21), reflecting differences in red cell penetration, glycation, hemoglobin species, red cell half-life, vitamin and medication status, and others factors (22).”

• cites [20] Nathan, 2008, Translating the A1C assay into estimated average glucose values, Diabetes Care sourceDisputed

  The Conclusions describe a tight A1C–average-glucose relationship and say that, for many if not most patients with diabetes, no important factors affect it. They report that substantially greater interindividual variation was found in a separate study of children, not in this study. The paper acknowledges some variation and clinical exceptions, but its characterization runs contrary to the claim of substantial variation at the same glycemia. (LLM, conf=0.82)

  Checked by a reasoning model — may be wrong.

  Experimental — accuracy not yet measured
• cites [21] Rohlfing, 2002, Defining the relationship between plasma glucose and HbA1c: analysis of glucose profiles and HbA1c in the Diabetes Control and Complications Trial, Diabetes Care sourceNot checked — no full text of the source

Problems with the IEC criteria have been reported with diverse populations.

“Further studies would be required before our findings with the ADA criteria could be generalized to other racial/ethnic groups, but problems with the IEC criteria have been reported with diverse populations (8,9,12).”

• cites [8] Kramer, 2010, A1C and diabetes diagnosis: The Rancho Bernardo Study, Diabetes Care sourcePartly supported

  The Results report low agreement between the IEC-recommended A1C cutoff of 6.5% and diabetes diagnosis by ADA criteria in this cohort of older adults. The Conclusions also compare the finding with NHANES data, but this paper does not establish that the problem occurs across diverse racial or ethnic populations. (LLM, conf=0.83)

  Experimental — accuracy not yet measured
• cites [9] Christensen, 2010, Moving to an A1C-based diagnosis of diabetes has a different impact on prevalence in different ethnic groups, Diabetes Care sourceSupported
• cites [12] Lorenzo, 2010, Performance characteristics of the new definition of diabetes, Diabetes Care sourceSupported

Better test characteristics appear to be provided by measuring random plasma glucose or glucose 1 h after a 50-g oral glucose challenge.

“Although A1C screening may be of limited value, better test characteristics appear to be provided by measuring random plasma glucose (25) or glucose 1 h after a 50-g oral glucose challenge (16); either could be obtained opportunistically during office visits, at any time of day, and without a prior fast.”

• cites [25] Ziemer, 2008, Random plasma glucose in serendipitous screening for glucose intolerance: screening for impaired glucose tolerance study 2, J Gen Intern Med sourcePartly supported

  The Results and Discussion report that random plasma glucose screened for undiagnosed diabetes with an AROC of 0.80 and performed well regardless of meal status or time of day; the Protocol says it required no prior fast. The copy does not establish that its test characteristics are better than A1C screening. The separate 50-g challenge claim is attached to citation 16, not this source. (LLM, conf=0.82)

  Experimental — accuracy not yet measured
• cites [16] Phillips, 2009, Glucose challenge test screening for prediabetes and undiagnosed diabetes, Diabetologia sourceSupported

The IEC A1C criteria have recently been compared with testing with fasting glucose or OGTTs in various populations to diagnose diabetes and high-risk/pre-diabetes.

“The IEC A1C criteria have recently been compared with testing with fasting glucose or oral glucose tolerance tests (OGTTs) in various populations to diagnose diabetes (7–13) and high-risk/pre-diabetes (10,11,13), but the ADA A1C criteria have not been studied.”

• cites [7] Carson, 2010, Comparison of A1C and fasting glucose criteria to diagnose diabetes among U.S. adults, Diabetes Care sourceSupported
• cites [8] Kramer, 2010, A1C and diabetes diagnosis: The Rancho Bernardo Study, Diabetes Care sourcePartly supported

  The Results and Conclusions compare the A1C ≥6.5% diabetes threshold with fasting-glucose and OGTT criteria in older adults. The paper discusses missed pre-diabetes, but our copy does not show a comparison of IEC A1C high-risk/pre-diabetes criteria with those glucose tests. (LLM, conf=0.82)

  Experimental — accuracy not yet measured
• cites [9] Christensen, 2010, Moving to an A1C-based diagnosis of diabetes has a different impact on prevalence in different ethnic groups, Diabetes Care sourceSupported
• cites [10] van't Riet, 2010, Relationship between A1C and glucose levels in the general Dutch population: the new Hoorn study, Diabetes Care sourceSupported
• cites [11] Mohan, 2010, A1C cut points to define various glucose intolerance groups in Asian Indians, Diabetes Care sourceSupported
• cites [12] Lorenzo, 2010, Performance characteristics of the new definition of diabetes, Diabetes Care sourcePartly supported

  The Methods and Results compare the IEC A1C diabetes threshold with fasting-glucose and WHO criteria incorporating two-hour glucose testing in IRAS participants. The paper also examines the A1C 6.0–6.4% high-risk category, but does not directly compare A1C with glucose-based criteria for diagnosing pre-diabetes. Its findings come from one study population; the sentence cites several papers for the broader claim about various populations. (LLM, conf=0.83)

  Experimental — accuracy not yet measured
• cites [13] Zhou, 2010, Performance of an A1C and fasting capillary blood glucose test for screening newly diagnosed diabetes and pre-diabetes defined by an oral glucose tolerance test in Qingdao, China, Diabetes Care sourceSupported

At similar glucose levels, A1C levels are higher in blacks than whites.

“Our findings that the proposed A1C criteria result in more false-positive and fewer false-negative results in blacks compared with whites are consistent with previous reports that at similar glucose levels, A1C levels are higher in blacks than whites (23).”

• cites [23] Herman, 2007, Differences in A1C by race and ethnicity among patients with impaired glucose tolerance in the Diabetes Prevention Program, Diabetes Care sourceSupported

The ADA criteria resulted in a distribution of diagnoses more similar to OGTT by identifying more subjects as high risk and fewer as normal.

“The ADA criteria resulted in a distribution of diagnoses more similar to those with the OGTT criteria by identifying more subjects as high risk and fewer subjects as normal.”

The proposed A1C criteria missed 82-94% of individuals with pre-diabetes.

“The proposed criteria missed 70% of individuals with diabetes, 71–84% with dysglycemia, and 82–94% with pre-diabetes.”

Using OGTT results, 5.8% of combined study subjects had new diabetes.

“Using OGTT results, 5.8% of the combined study subjects had new diabetes, and 36% had pre-diabetes (Table 1).”

False-positive results were more common in black subjects than white subjects.

“There were also racial differences, with false-positive results being more common in black subjects and false-negative results being more common in white subjects.”

OGTT revealed diabetes in 5.2% of combined study subjects.

“OGTTs revealed pre-diabetes in 35.8% and diabetes in 5.2% of combined study subjects.”

The proposed A1C criteria missed 70% of individuals with diabetes.

“The proposed criteria missed 70% of individuals with diabetes, 71–84% with dysglycemia, and 82–94% with pre-diabetes.”

A1C provided ROC curve areas ≤0.70 for dysglycemia or pre-diabetes.

“A1C provided receiver operating characteristic (ROC) curve areas for diabetes of 0.79–0.83, but ROC curve areas were ≤0.70 for dysglycemia or pre-diabetes.”

A1C provided ROC curve areas for diabetes of 0.79-0.83.

“A1C provided receiver operating characteristic (ROC) curve areas for diabetes of 0.79–0.83, but ROC curve areas were ≤0.70 for dysglycemia or pre-diabetes.”

OGTT revealed pre-diabetes in approximately 36% of combined study subjects.

“OGTTs revealed pre-diabetes in 35.8% and diabetes in 5.2% of combined study subjects.”

• Also stated in: § RESULTS · plain

  “Using OGTT results, 5.8% of the combined study subjects had new diabetes, and 36% had pre-diabetes (Table 1).”

The proposed A1C criteria missed 71-84% of individuals with dysglycemia.

“The proposed criteria missed 70% of individuals with diabetes, 71–84% with dysglycemia, and 82–94% with pre-diabetes.”

A1C-based categorizations were mostly incorrect for OGTT diagnoses of both pre-diabetes and diabetes.

“A1C-based categorizations were mostly incorrect for OGTT diagnoses of both pre-diabetes and diabetes.”

A1C testing resulted in more normal diagnoses and fewer high risk and diabetes diagnoses with both IEC and ADA criteria compared with OGTT.

“Thus, compared with an OGTT, A1C testing resulted in more normal diagnoses (categorizations) and fewer high risk and diabetes diagnoses with both the IEC and ADA criteria.”

False-negative results were more common in white subjects than black subjects.

“There were also racial differences, with false-positive results being more common in black subjects and false-negative results being more common in white subjects.”

The percentage of incorrectly identified high-risk subjects with A1C screening for pre-diabetes was lower with ADA criteria: 96% of whites and 86% of blacks with IEC vs 87% of whites and 73% of blacks with ADA.

“However, the percentage of incorrectly identified high-risk subjects (false-negative plus false-positive results) with A1C screening for pre-diabetes was lower with the ADA criteria: 96% of whites and 86% of blacks with the IEC criteria vs. 87% of whites and 73% of blacks with the ADA criteria.”

Repeating the analyses with a higher fasting plasma glucose cutoff for normal glucose tolerance (FPG <110 mg/dl) did not change the results.

“Repeating the above analyses with a higher fasting plasma glucose cutoff for normal glucose tolerance (FPG <110 mg/dl) did not change the results (see supplementary Tables 2, 3, and 4, and supplementary Fig. 1, available in an online appendix).”

With IEC criteria for high risk, 94% of whites vs 77% of blacks with pre-diabetes by OGTT would have false-negative results.

“With the IEC criteria for high risk, 94% of whites vs. 77% of blacks with pre-diabetes by OGTT would have false-negative results, whereas 1.7% of whites vs. 9.0% of blacks without pre-diabetes would have false-positive results (P < 0.02 that whites have more false-negative and less false-positive results).”

At lower ADA cutoff for high risk, 78% of whites vs 56% of blacks with pre-diabetes by OGTT would have false-negative results.

“At the lower ADA cutoff for high risk, there would be fewer false-negative but more false-positive results: 78% of whites vs. 56% of blacks with pre-diabetes by OGTT would have false-negative results, whereas 8.5% of whites vs. 18% of blacks without pre-diabetes would have false-positive results (P < 0.05 that whites have more false-negative than false-positive results).”

Approximately equal sensitivity and specificity would be provided by cutoffs of 5.3-5.5% for pre-diabetes or dysglycemia and 5.5-5.8% for diabetes.

“Approximately equal sensitivity and specificity would be provided by cutoffs of 5.3–5.5% for pre-diabetes or dysglycemia and of 5.5–5.8% for diabetes.”

At lower ADA cutoff for high risk, 8.5% of whites vs 18% of blacks without pre-diabetes would have false-positive results.

“At the lower ADA cutoff for high risk, there would be fewer false-negative but more false-positive results: 78% of whites vs. 56% of blacks with pre-diabetes by OGTT would have false-negative results, whereas 8.5% of whites vs. 18% of blacks without pre-diabetes would have false-positive results (P < 0.05 that whites have more false-negative than false-positive results).”

Incorrect identifications across the three datasets averaged 70% for diabetes.

“Incorrect identifications across the three datasets averaged 90 and 71% for pre-diabetes (by IEC and ADA criteria, respectively, combining false-positive and false-negative results), 84 and 65% for dysglycemia (all false-negative results), and 70% for diabetes (all false-negative results).”

Of those with diabetes by OGTT, 78% of whites vs 51% of blacks would have false-negative results (P < 0.01).

“Of those with diabetes by OGTT, 78% of whites vs. 51% of blacks would have false-negative results (P < 0.01), whereas 0.3% of whites vs. 1.8% of blacks without diabetes would have false-positive results (P = 0.15).”

Of those without diabetes by OGTT, 0.3% of whites vs 1.8% of blacks would have false-positive results (P = 0.15).

“Of those with diabetes by OGTT, 78% of whites vs. 51% of blacks would have false-negative results (P < 0.01), whereas 0.3% of whites vs. 1.8% of blacks without diabetes would have false-positive results (P = 0.15).”

Both IEC and ADA criteria are insensitive for diabetes (<34% at A1C cutoff ≥6.5%).

“Both the IEC and the ADA criteria are specific for diagnosing diabetes and pre-diabetes (>90% specificity above A1C cutoffs of 6.0%) but insensitive for both diabetes (<34% at A1C cutoff ≥6.5%) and pre-diabetes (lowest with IEC criteria: <31% above the cutoff of 5.7% and <13% above the A1C cutoff of 6.0%).”

Only for diabetes were areas under ROC curve ≥0.80 (0.82 in SIGT, 0.80 in NHANES III, and 0.83 in NHANES 2005-2006).

“Only for diabetes were areas under the ROC curve areas ≥0.80 (0.82 in the SIGT study, 0.80 in NHANES III, and 0.83 in NHANES 2005–2006), whereas for pre-diabetes they were lowest (0.65 in SIGT study, 0.62 in NHANES III, and 0.68 in NHANES 2005–2006), and intermediate for dysglycemia (0.67 in the SIGT study, 0.66 in NHANES III, and 0.70 in NHANES 2005–2006).”

The impact of A1C-based criteria differs by race.

“In addition, the impact of the A1C-based criteria differs by race.”

With IEC criteria for high risk, 1.7% of whites vs 9.0% of blacks without pre-diabetes would have false-positive results.

“With the IEC criteria for high risk, 94% of whites vs. 77% of blacks with pre-diabetes by OGTT would have false-negative results, whereas 1.7% of whites vs. 9.0% of blacks without pre-diabetes would have false-positive results (P < 0.02 that whites have more false-negative and less false-positive results).”

The proposed A1C diagnostic criteria are racially discrepant for screening.

“The proposed A1C diagnostic criteria are insensitive and racially discrepant for screening, missing most Americans with undiagnosed diabetes and pre-diabetes.”

The proposed A1C diagnostic criteria miss most Americans with undiagnosed diabetes and pre-diabetes.

“The proposed A1C diagnostic criteria are insensitive and racially discrepant for screening, missing most Americans with undiagnosed diabetes and pre-diabetes.”

With NHANES 2005-2006 data, 43-52 million non-Hispanic U.S. adults with pre-diabetes would be missed by screening with A1C.

“With use of NHANES 2005–2006 data, ∼5.9 million non-Hispanic U.S. adults with unrecognized diabetes and 43–52 million with pre-diabetes would be missed by screening with A1C.”

With NHANES 2005-2006 data, approximately 5.9 million non-Hispanic U.S. adults with unrecognized diabetes would be missed by screening with A1C.

“With use of NHANES 2005–2006 data, ∼5.9 million non-Hispanic U.S. adults with unrecognized diabetes and 43–52 million with pre-diabetes would be missed by screening with A1C.”

The proposed A1C diagnostic criteria are insensitive for screening.

“The proposed A1C diagnostic criteria are insensitive and racially discrepant for screening, missing most Americans with undiagnosed diabetes and pre-diabetes.”

• AssumptionRests on an assumption: Identification of diabetes and pre-diabetes can permit management to prevent complications or delay progression from pre-diabetes to diabetes.

Such variation would be expected to make it more difficult for A1C measurements to distinguish small differences in glycemia, e.g., pre-diabetes versus normal.

“Such variation would be expected to make it more difficult for A1C measurements to distinguish small differences in glycemia, e.g., pre-diabetes versus normal.”

• AssumptionRests on an assumption: A1C within an individual fluctuates little from day to day, whereas an OGTT can reflect day-to-day differences in insulin secretion and insulin action.

With extrapolation from NHANES 2005-2006, A1C screening with IEC and ADA criteria would incorrectly identify diabetes in 470,000 black and 6 million white non-Hispanic Americans.

“With extrapolation from the NHANES 2005–2006 dataset, A1C screening with the IEC and ADA criteria would incorrectly identify diabetes (false-positive or false-negative results) in 470,000 black and 6 million white non-Hispanic Americans.”

Abnormal results on a single OGTT was highly predictive of subsequent diabetes in Pima Indians.

“Moreover, abnormal results on a single OGTT was highly predictive of subsequent diabetes in Pima Indians (24), and differences found on repeat studies should be included in the variability of the ROC curve analyses.”

• cites [24] McCance, 1994, Comparison of tests for glycated haemoglobin and fasting and two hour plasma glucose concentrations as diagnostic methods for diabetes, BMJ sourceNot checked — no full text of the source

The IEC criteria would incorrectly identify high risk in 6 million black and 42 million white non-Hispanic Americans.

“The IEC criteria would incorrectly identify high risk in 6 million black and 42 million white non-Hispanic Americans.”

A1C testing for pre-diabetes would be incorrect for 43 million Americans using ADA criteria and 52 million using IEC criteria.

“A1C testing for pre-diabetes would also be incorrect for 43 million Americans using the ADA criteria and 52 million using the IEC criteria, with the majority of those being pre-diabetic by an OGTT but normal by A1C (Table 2).”

The study did not search for factors that can be problematic for A1C measurements, such as hemoglobinopathies and anemia.

“We also did not search for factors that can be problematic for A1C measurements, such as hemoglobinopathies and anemia.”

• AssumptionRests on an assumption: Factors such as hemoglobinopathies and anemia can be problematic for A1C measurements.

A1C and OGTT differ intrinsically.

“It is beyond our analysis to address the accuracy of OGTT versus A1C, but they differ intrinsically.”

• AssumptionRests on an assumption: A1C within an individual fluctuates little from day to day, whereas an OGTT can reflect day-to-day differences in insulin secretion and insulin action.

OGTT is a gold standard used for identification of diabetes and pre-diabetes around the world.

“We evaluated the proposed criteria as screening tests in three populations, compared with the OGTT as a “gold standard” used for identification of diabetes and pre-diabetes around the world (14).”

• cites [14] 2006Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Use of the proposed A1C criteria would not be a good screening strategy to identify the many Americans who currently have unrecognized diabetes or pre-diabetes as found with OGTTs.

“Although our findings support diagnosing diabetes by the proposed A1C criteria as rarely false positive, use of the proposed criteria would not be a good screening strategy to identify the many Americans who currently have unrecognized diabetes or pre-diabetes as found with OGTTs, a diagnostic standard currently used around the world.”

• AssumptionRests on an assumption: Identification of diabetes and pre-diabetes can permit management to prevent complications or delay progression from pre-diabetes to diabetes.

One limitation is performance of only a single OGTT.

“One limitation is performance of only a single OGTT, but such an approach reflects clinical practice.”

• AssumptionRests on an assumption: A single OGTT reflects clinical practice.

Extrapolating NHANES 2005-2006 findings to the non-Hispanic American adult population, A1C testing would incorrectly identify diabetes in 6.5 million.

“Extrapolating the NHANES 2005–2006 findings to the non-Hispanic American adult population, A1C testing would incorrectly identify diabetes in 6.5 million (false-positive and false-negative results).”