Current issues of ACP Journal Club are published in Annals of Internal Medicine


Hemoglobin A1c was a better alternative to fasting plasma glucose testing than fructosamine for diabetes screening

ACP J Club. 1992 May-June;116:84. doi:10.7326/ACPJC-1992-116-3-084

Source Citation

Tsuji I, Nakamoto K, Hasegawa T, et al. Receiver operating characteristic analysis on fasting plasma glucose, HbA1c, and fructosamine on diabetes screening. Diabetes Care. 1991 Nov;14:1075-7.



To assess the diagnostic properties of fasting plasma glucose (FPG), HbA1c, and fructosamine levels as screening tests for diabetes.


Cross-sectional study of screening tests.


Health clinic in Osaka, Japan.


All patients ≥ 40 years visiting the clinic during 2 months in 1989 and diabetic patients seen during the subsequent 5 months were included, with the exception of diabetic patients taking insulin or oral hypoglycemic agents. Mean age (SD) of the 619 patients was 51.1 (7.3) years; 65% were men.

Description of tests and diagnostic standard

FPG, HbA1c, and fructosamine levels were measured in blood drawn from an antecubital vein after an overnight fast. A 75-g oral glucose tolerance test (OGTT) was then done and plasma glucose measured at 60 and 120 min. Patients were classified as either diabetic (FPG ≥ 8mM or plasma glucose [PG] at 120 min ≥ 11 mM, or both), impaired glucose tolerant (FPG < 8 mM and PG at 120 min, 8 to 11 mM), or nondiabetic (FPG < 8 mM and PG at 120 min < 8 mM).

Main results

469 patients were nondiabetic; 88, glucose intolerant; and 62, diabetic. Mean (SD) FPG was 5.06 mM (0.46) for nondiabetic patients,5.40 mM (0.54) for patients with impaired glucose tolerance, and 7.42 mM (2.09) for diabetic patients; mean HbA1c levels were 5.67% (0.41%), 5.85% (0.46%), and 7.50% (1.49%); mean fructosamine levels were 2.24 mM (0.26), 2.33 mM (0.23), and 2.80 mM (0.47); and plasma glucose levels after 120 min during the OGTT were 5.89 (1.10), 8.84 (0.83), and 15.06 (3.67) (P < 0.05 for analysis of variance for each test). When patients with impaired glucose tolerance were grouped with nondiabetic patients in the disease-absent group, the area (SE) under the receiver operating characteristic (ROC) curves for FPG (0.944 [0.020]) and HbA1c (0.935 [0.022]) was similar, and in both cases was larger than that for fructosamine (0.856 [0.031], P < 0.05). When patients with impaired glucose tolerance were grouped with diabetic patients in the disease-present group, the area under the ROC curve for FPG (0.785 [0.016]) was similar to that for HbA1c (0.753 [0.026]) and larger than the area under the curve for fructosamine (0.706 [0.026] P < 0.05).


HbA1c was diagnostically accurate and useful in screening patients for diabetes; fructosamine was less accurate.

Sources of funding: Ministry of Education (Japan) and Japanese Association of Social Insurance.

Address for article reprint: Dr. I. Tsuji, Room 6030D, Department of Epidemiology, School of Public Health, The Johns Hopkins University, 615 North Wolfe Street, Baltimore, MD 21205, USA.


Should we be screening asymptomatic, nonpregnant patients for diabetes mellitus? The American Diabetes Association recommends that persons with one or more risk factors or in high-risk groups be referred to a physician for evaluation and testing (1). Critical review of screening for diabetes mellitus, however, indicates that evidence is still insufficient that identifying patients with asymptomatic diabetes and initiating treatment during the asymptomatic phase yields better outcomes than if treatment is initiated once symptoms develop. Screening, therefore, is not recommended by some experts (2).

Tsuji and colleagues have shown that HbA1c performs nearly as well as fasting plasma glucose in separating diabetic from nondiabetic and impaired glucose-tolerant individuals in a case-finding setting. This finding confirms earlier reports of the usefulness of HbA1c as a screening test for diabetes mellitus. This study population, primarily white-collar workers, most of whom were reported to be asymptomatic, presenting for an employer-sponsored health check-up, is representative of a subset of patients that the internist might consider screening for diabetes. The performance of HbA1c and fasting plasma glucose were assessed using the area under the ROC curve and are, therefore, independent of disease prevalence. As such, the results should be applicable to groups not represented in the study sample.

Measurement of HbA1c is not influenced by food intake, physical activity, or metabolic stress and can be done on a random blood sample. Clinicians may find that the practical advantages of the HbA1c make it preferable to FPG as a screening test for diabetes mellitus, given its similar diagnostic accuracy as shown by this study. However, the circumstances in which such screening is appropriate remain to be defined.

Thomas Mattimore, MD
University of CaliforniaLos Angeles, California, USA


1. American Diabetes Association. Position statement: screening for diabetes. Diabetes Care. 1989;12:588-90.

2. Singer DE, Samet JH, Coley CM, Nathan DM. Screening for diabetes mellitus. Ann Intern Med. 1988;109:639-49.

1997 Editorial update: The use of a new diagnostic threshold for diabetes (FPG ≥ 7.0) may however modify these results. Expert Committee on the Diagnosis and Classification of Diabetes Mellitus. Report of the Expert Committee on the Diagnosis and Classification of Diabetes Mellitus. Diabetes Care. 1997; 20:1183-97.