The validity of basal blood glucose in the control of non-insulin-dependent diabetic patients

BACKGROUND: To validate basal glucemia as a control method for non-insulin dependent diabetes mellitus, and to determine the cut-off point that best characterizes good control. PATIENTS AND METHODS: A transversal, observational study of 256 patients who participated in a diabetes mellitus follow-up program during 1993. In the study, glucemia validity indicators were evaluated after making 2 X 2 tables and ROC (receiver operating characteristic) curves for the different values. Control values of glycated hemoglobin was used as to define a good (< 6.5%) and moderate (< 8%). RESULTS: The values of glucemia considered to be "good" as regards control (from 80 to 110 mg/dl, 4.4-6.05 mmol/dl) have good sensitivity (from 97.3% to 100%) and negative predictive values (from 85.7% to 100%) but extremely bad specificity (from 3.8% to 22.7%) and only moderate positive predictive values (from 59.5% to 64.1%) in reference to values of glycated hemoglobin of 6.5%. The same occurs for 8% as regards sensitivity (from 98.6% to 100%), negative predictive value (from 96.4% to 100%) and specificity from 2.1% to 14.5%). Positive predictive value worsens (from 27.8% to 30.3%). The most effective and most accurate values of glucemia in the ROC curves are 150 mg/dl (8.25 mmol/l) if the control of glycated hemoglobin is good, and 170 mg/dl (9.35 mmol/l) if it is moderate. CONCLUSIONS: The glucemia control figures recommended by consensus produce false positives when they are compared to glycated hemoglobin. In the analysis of effectiveness and ROC curves greater accuracy is obtained with glucemia values that are slightly higher than those recommended.