Article Text

Download PDFPDF

A fasting plasma glucose level ≥7.8 mmol/l was sensitive but not specific for identifying poor glycaemic control
  1. Arthur T Evans, MD, MPH
  1. Cook County Hospital Chicago, Illinois, USA

    Statistics from

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

 QUESTION: In patients with type 2 diabetes mellitus who are not using insulin, how do fasting plasma glucose (FPG) measurements compare with those of glycosylated haemoglobin (HbA1c) in determining glycaemic control?


    Comparison of 2 laboratory tests.


    A family practice in Amsterdam, the Netherlands.


    1020 patients who were >40 years of age (mean age 64 y, 67% women), had type 2 diabetes (median duration 2.4 y), were not pregnant, and were not taking insulin.

    Description of test and diagnostic standard

    FPG levels were measured in venous plasma by the glucose oxidase method (Boehringer-Mannheim, Mannheim, Germany) and by the glucose dehydrogenase method (Merck, Darmstadt, Germany). HbA1c was measured using high performance liquid chromatography. The sensitivity and specificity of an FPG level of 7.8 mmol/l was calculated with reference to the HbA1c diagnostic standard cut points of 6.5% and 7.0%. These cut points were chosen in accordance with guidelines of the European NIDDM Policy Group and the American Diabetes Association.

    Main outcome measures

    The predictive value of different levels of FPG as an indicator of HbA1c level was analysed using receiver operating characteristic curves. Change in FPG level at 3 months was also assessed as an indicator of change in HbA1c level.

    Main results

    For the 2 HbA1c cut points that separated good from poor levels (6.5% and 7.0%), the areas under the curve did not differ (0.87 and 0.88, respectively; p=0.35). An FPG level ≥7.8 mmol/l detected 90% of patients with an HbA1c level ≥7.0%. An FPG level <7.8 mmol/l detected 66% of patients with an HbA1c level <7.0%. A change in FPG levels at 3 months of ≤0 (ie, no change or worsening) had a sensitivity of 57% and a specificity of 88% for detecting a deterioration in HbA1c of >0.5%.


    In patients with type 2 diabetes mellitus who were not receiving insulin, fasting plasma glucose level was highly correlated with glycosylated haemoglobin level. Except at extreme values (<7 and >12 mmol/l), fasting plasma glucose level was too imprecise to substitute for glycosylated haemoglobin level as a measure of glycaemic control.


    Clinicians have always considered the blood glucose and HbA1c tests as different—each valuable for complementary purposes. Therefore, the finding that a single FPG level does not accurately predict the HbA1c level is not surprising and will not change clinical practice.

    Nevertheless, some details in the article by Bouma and colleagues offer clinical utility. Because clinicians naturally distinguish between glucose levels as disparate as 8 mmol/l and 12 mmol/l (144 mg/dl and 216 mg/dl), it is clinically inappropriate to consider all values >7.8 mmol/l as equivalent. Likelihood ratios (LRs) allow for more useful clinical distinctions. When the data are recast as likelihood ratios, it appears that an FPG level >12 mmol/l effectively rules in poor glycaemic control (HbA1c level >7%), with an +LR >40. Similarly, an FPG level <7.8 mmol/l lowers the odds of poor glycaemic control by a factor of 0.15.

    Although the correlation found by Bouma and colleagues (r=0.77) is remarkably high (meaning that an FPG level can “explain” >50% of the variability in HbA1c level), perhaps a more clinically relevant question could be posed: can one accurately predict HbA1c levels from the average value of several glucose levels over an extended period? Older studies have suggested that averaging at least 3 FPG values can improve the correlation with HbA1c level almost to r=0.9.1

    Other studies have noted that non-fasting glucose levels may more accurately predict poor glycaemic control than do fasting levels. For example, HbA1c levels were more highly correlated with post lunch glucose levels (r=0.81) than with fasting levels (r=0.62), and post lunch glucose values of >11.1 mmol/l (>200 mg/dl) ruled in poor glycaemic control (HbA1c level >7%).2

    Clinicians must also be mindful that HbA1c values are not a perfect measure of glycaemic control. There are substantial interindividual variations in the affinity of haemoglobin and glucose, and rapid erythrocyte turnover can give a false sense of good glycaemic control.


    View Abstract


    • Source of funding: not stated.

    • For correspondence: Dr M Bouma, Institute for Research in Extramural Medicine, Van der Boechorststraat 7, 1081 BT Amsterdam, the Netherlands. Fax +31 20 44 48262.