INTRODUCTION. To find an optimal critical line in the fasting plasma glucose (FPG)-HbA1c plane for identifying diabetes in participants with impaired fasting glucose (IFG) and thereby improve the efficacy of using FPG alone in diabetes screening among American Indians.
RESEARCH DESIGN AND METHODS. We used FPG, 2-h postload glucose (2hPG), and HbA1c measured in the 2,389 American Indians (aged 45–74 years, without diabetes treatment or prior history of diabetes) in the Strong Heart Study (SHS) baseline (second) examination. Participants were classified as having diabetes if they had either FPG £ 126 mg/dl or 2hPG ≥ 200 mg/dl, as having IFG if they had 110 £ FPG < 126 mg/dl, and as having normal fasting glucose (NFG) if they had FPG < 110, according to the American Diabetes Association (ADA) definition. Logistic regression models were used for identifying diabetes (2hPG ≥ 200 mg/dl) in IFG participants. The areas under the receiver operating characteristic (ROC) curves generated by different logistic regression models were evaluated and compared to select the best model. A utility function based on the best model and the cost-to-benefit ratio was used to find the optimal critical line. The data from the second examination were used to study the effect of the time interval between the successive diabetes screenings on both the FPG criterion and the optimal critical line.
RESULTS. A total of 37% of all subjects with new diabetes at baseline and 55.2% of those in the second exam had 2hPG ≥ 200 but FPG < 126. There was a very large portion of IFG participants with diabetes (19.3 and 22.9% in the baseline and second exam, respectively). Among the areas under the ROC curves, the area generated by the logistic regression model on FPG plus HbA1c is the largest and is significantly larger than that based on FPG (P = = 0.0008). For a cost-to-benefit ratio of 0.23888, the optimal critical line that has the highest utility is: 0.89 × HbA1c + 0.11 × FPG = 17.92. Those IFG participants whose FPG and HbA1c were above or on the line were referred to take an oral glucose tolerance test (OGTT) to diagnose diabetes. The optimal critical line is lower if a successive diabetes screening will be conducted 4 years after the previous screening.
CONCLUSIONS. FPG ≥ 126 and 2hPG ≥ 200, as suggested by the ADA, are used in-dependently to define diabetes. The FPG level is easy to obtain, and using FPG alone is suggested for diabetes screening. It is difficult to get physicians and patients to perform an OGTT to get a 2hPG level because of the many drawbacks of the OGTT, especially in those patients who already have FPG < 126. It is also impractical to conduct an OGTT for everyone in a diabetes screening. Our data show that 37% of all subjects with new diabetes in the SHS baseline exam and 55.2% of those in the second exam have 2hPG ≥ 200 but FPG < 126. These cases of diabetes cannot be detected if FPG is used alone in a diabetes screening. Therefore, although the small portion of diabetes in the NFG group (4.7% in the base-line and 6.9% in the second exam) may be ignored, those cases of diabetes among IFG participants (~20% in our data) need further consideration in a diabetes screening. It may be worthwhile for those IFG participants identified by the optimal critical line to take an OGTT. The optimal critical line and time interval between successive diabetes screenings need further study.