We appreciate the valuable input from Puzone and colleagues regarding our paper, “Agreement between intermittent glucose concentrations and continuous glucose monitoring (CGM) in at-risk newborns.” Our research aimed to evaluate the feasibility and accuracy of non-invasive screening for neonatal hypoglycemia in at-risk newborns. The Dexcom G6 continuous glucose monitor, a factory-calibrated device, was employed as the non-invasive instrument in this investigation. It was compared with the i-STAT glucometer, a bedside laboratory-quality glucose concentration analyzer and the standard-of-care to intermittently determine glucose levels in our Newborn Nursery and NICU, which necessitated frequent needle sticks. Because the i-STAT analyzers are superior to point-of-care (POC) glucometers in precision, we compared glucose concentrations from the Dexcom G6 to the i-STAT, rather than less accurate POC glucometers.

Our results revealed a substantial and proportional discrepancy when compared to the laboratory-quality i-STAT glucose levels. From an analytic standpoint, we agree with Puzone et al that Mean Absolute Relative Difference (MARD) is not an optimal measure to characterize the performance of the Dexcom G6 compared to the i-STAT. In our paper, we presented the Bland-Altman analysis as the primary result, which is a more suitable statistical method to ascertain the potential adoption of the new screening approach in lieu of the current standard. Consequently, utilizing Bland-Altman analysis rather than MARD, we concluded that the Dexcom G6 was not accurate, particularly at the low glucose concentrations common in newborns, and was unsuitable as a trending device due to proportional bias. Moreover, we observed that retrospective sensor calibration in real-time did not mitigate the measurement bias.