This article is only available in the PDF format. Download the PDF to view the article, as well as its associated figures and tables.
—We agree that calculation of an ROC curve can be a useful method for evaluating the trade-off between sensitivity and specificity of a test. We initially determined the ROC curve for our data, but chose not to present it because we were only interested in evaluating the characteristics of the glycosylated hemoglobin level at a very high specificity (≥95%) to minimize the number of false-positive diagnoses for diabetes. We did not believe that the data were sufficient to warrant publishing the "fine structure" of the curve between 95% and 100%. The important point is that, regardless of the cutpoint used, glycosylated hemoglobin levels and 2-hour tests are not identical. However, since glycosylated hemoglobin levels are used to evaluate and guide therapy for diabetes, we believe that it may be one of the most useful diagnostic tests for identifying patients who have diabetes requiring treatment.
Peters AL, Schriger DL, Davidson MB, Hasselblad V. Use of Glycosylated Hemoglobin Levels for Diagnosing Diabetes Mellitus-Reply. JAMA. 1997;277(3):211. doi:10.1001/jama.1997.03540270037018