We thank Mastrototaro et al. (1) for their comments. For more than 30 years, development of a reliable ambulatory continuous glucose monitor has appeared to be an insurmountable technological challenge. The Medtronic MiniMed continuous glucose monitoring system (CGMS), the first commercially available system, provides retrospective analysis of glucose levels, making confirmation of unexpected findings difficult. As stated by Mastrototaro et al., when our sensor data were evaluated using correlation tests, a high degree of correlation with capillary blood measurements was found, confirming several previous reports. However, simple correlation tests do not identify potentially important clinical discrepancies. The tools to evaluate the clinical efficacy of new monitoring systems are limited. We did not use the Clarke error grid because the boundaries used in the published version (2) are not consistent with the requirements of tight glycemic control. For example, reference/sensor glucose value pairs such as 200 vs. 90 or 80 vs. 160 would fall into the acceptable “B” zone, although today such differences are unacceptable. Indeed, in the data reported by Gross et al. (3), many points falling in the B zone may not be considered clinically acceptable today. Therefore, pending the publication of a validated modern version of the Clarke error grid or another validated tool, we used an admittedly more subjective tool, the clinical judgment of a diabetologist blinded to the subject’s identity.
Our profiles were generated using Solutions version 2 software, which was the version available at the time of the study and the one used in most previously published reports. Recalculation of our data with the newer software showed significant improvement, particularly in correcting the “midnight shift.” However, other discrepancies apparent in our study, and our ultimate conclusions, were unchanged.
Recently, McGowan et al. (4) used a different technique to validate CGMS readings in seven patients and found that in four, falsely low sensor readings “might have resulted in inappropriate reduction of overnight insulin dose,” a finding that supports our results. They conclude that many hypoglycemic episodes identified by the sensor may be spurious, thus questioning recent reports that showed an unexpectedly high incidence of asymptomatic nocturnal hypoglycemia.
We recognize the importance of this new technology and its inherent technical difficulties. We applaud Medtronic MiniMed for producing the first commercial system and for their continuing efforts to improve its reliability. The need for continued development is obvious, and clearly this technology will greatly improve our ability to treat diabetes. However, the current model has limitations that must be recognized, and new tools are needed to critically evaluate the clinical reliability of future devices.
References
Address correspondence to Dr. Muriel Metzger, Diabetes Unit, Hadassah University Hospital, P.O. Box 12000, Jerusalem, Israel. E-mail: [email protected].
