BACKGROUND: This study was designed to assess the accuracy, reliability, and contribution to clinical decision-making of two commercially available continuous glucose monitoring (CGM) devices using a novel analytical approach. STUDY DESIGN: Eleven individuals with type 1 diabetes and five with type 2 diabetes wore a Guardian RT (GRT) (Medtronic Minimed, Northridge, CA) or DexCom STS Continuous Monitoring System (DEX) (San Diego, CA) device for 200 h followed by an 8-h laboratory study. A subset of these subjects wore both devices simultaneously. RESULTS: Subjects produced 1,902 +/- 269 readings during the ambulatory phase. During the laboratory study we found: lag time of 21 +/- 5 min for GRT and 7 +/- 7 min for DEX (P < 0.005); mean absolute relative difference of 19.9% and 16.7%, respectively, for GRT and DEX; and glucose exposure (the ratio of study device/laboratory reference device [YSI Instruments, Inc., Yellow Springs, OH] area under the curve) of 95 +/- 6% for GRT and 101 +/- 13% for DEX. Reliability measured during laboratory study showed 82% for DEX and 99% for GRT. Clarke Error Grid analysis (YSI reference) showed for GRT 59% of values in zone A, 34% in zone B, and 7% in zone D and for DEX 70% in zone A, 28% in zone B, 1% in zone C, and 1% in zone D. Bland-Altman plots (YSI standard) yielded for DEX 3 mg/dL (95% confidence interval, -78 to 84 mg/dL) and for GRT -21 mg/dL (95% confidence interval, -124 to 82 mg/dL). Six of eight subjects completed both home and laboratory simultaneous use of DEX and GRT. Lag times were inconsistent between devices, ranging from 0 to 32 min; area under the curve revealed a tendency for DEX to report higher total glucose exposure than GRT for the same patient. CONCLUSIONS: CGM detects abnormalities in glycemic control in a manner heretofore impossible to obtain. However, our studies revealed sufficient incongruence between simultaneous laboratory blood glucose levels and interstitial fluid glucose (after calibrations) to question the fundamental assumption that interstitial fluid glucose and blood glucose could be made identical by resorting to algorithms based on concurrent blood glucose levels alone.