Method for estimating confidence levels for measurements by blood glucose monitoring systems

A joint conference on self-monitoring of blood glucose (SMBG) has proposed that all glucose monitoring systems generate values that are within 10% of the actual blood glucose level 100% of the time. To estimate the confidence limits of blood glucose measurements made in a typical university ambulatory care setting and to ascertain whether they met the proposed standard, we performed duplicate determinations of blood glucose using a reflectance meter and applied to the measurements a method for calculating the closeness of measured values to a "true" mean. Based on paired measurements in 100 consecutive diabetic subjects, we were able to show that one measurement would be within 11.9% of a true mean value 95% of the time and within 8.4% of the mean 90% of the time. The 95 and 90% confidence limits for the average of two repeated measurements were calculated to be 8.4 and 5.9%, respectively. Our methodology can be applied to any set of SMBG values to calculate their confidence limits and to determine whether the measurements meet recommended standards.