The DirecNet study protocol involving the short-term T1D participants is available on the DirecNet website, http://direcnet.jaeb.org/viewpage.aspx?pagename=home
, and is briefly summarized here. The study was approved by the DirecNet Data and Safety Monitoring Board and the institutional review boards at each of the five DirecNet centers. A parent or guardian gave written consent, and each subject gave written assent. Major eligibility criteria included the following: 1
) clinical diagnosis of T1D between 6 and 52 weeks, 2
) positive antibody (GAD65 and islet cell autoantigen 512) determination, and 3
) age 8 to <19 years. Participants were asked to wear a Guardian Clinical (Medtronic MiniMed, Northridge, CA) for 3–7 days prior to a hypoglycemic clamp test. The Guardian Clinical is the same as a Guardian Real-Time CGM, but it is a “blinded” device, meaning the glucose values are recorded in the receiver but the values are not visible to the patient. The C-peptide concentration during MMTT was measured at Northwest Lipid Metabolism and Diabetes Research Laboratories (Seattle, WA) using Tosoh AIA 1800 (Tosoh, Montgomeryville, PA). Among the 20 participants who completed both tests, 1 had a C-peptide response <0.2 nmol/L during MMTT and 3 did not wear the CGM device. This analysis included the other 16 participants who had a mean ± SD peak C-peptide level of 0.96 ± 0.42 nmol/L, ranging from 0.46 to 1.96 nmol/L, and a median duration of glucose data of 77 h (25th–75th percentile 62–117), ranging from 34 to 154 h. A1C was measured using the DCA Vantage analyzer (Siemens Healthcare Diagnostics).
These 16 short-term T1D participants were matched for age and A1C to 34 participants with ≥5 years T1D duration (“longer-term”) from the JDRF CGM Randomized Clinical Trial cohort (3
). CGM data were taken from an approximately 7-day period of blinded sensor use at baseline prior to randomization. The median duration of glucose data was 150 h (25th–75th percentile 131–181), ranging from 97 to 241 h in the 26 (76%) participants using the Guardian Clinical device and 8 (24%) using the FreeStyle Navigator (Abbott Diabetes Care, Alameda, CA). A1C was measured using the DCA 2000 analyzer (Bayer, Tarrytown, NY) in 32 (94%) participants and Tosoh G7 (Tosoh) in 2 (6%) participants. The correlation between the DCA 2000 assay and the Tosoh G7 assay is r
= 0.94 (5
In addition, the 16 short-term T1D participants were matched for age to 26 nondiabetic participants who were diabetes-antibody negative from a separate JDRF-funded study (2
), including 12 (46%) participants who used the Guardian Clinical device for 3 days and 14 (54%) who used a blinded Navigator for 5 days. The median duration of the glucose data was 71 h (25th–75th percentile 67–103), ranging from 51 to 141 h. A previous study showed that results were similar comparing the Navigator and Guardian Clinical CGM devices (mean glucose 98 ± 11 and 98 ± 9 mg/dL, respectively (2
). A1C was measured using the DCA 2000 for 23 (88%) participants and Tosoh G7 for 3 (12%) participants.
CGM glucose indices were calculated giving equal weight to each of the 24 h of a calendar day. As the long-duration group had a greater amount of data than the short-term T1D and nondiabetic groups, group comparisons were made using both 3 and 6 days of data. Since the results were similar regardless of the length of data used (Supplementary Table 1
), the data from ~6 days of sensor wear are presented here. Coefficient of variation (CV) as a measure of glucose variability was defined as SD divided by mean glucose.
Means ± SD or median (25th–75th percentiles) was summarized as appropriate to the distribution for demographic and clinical characteristics and various measures of glucose separately for each of three groups. The comparisons between short-term T1D participants and nondiabetic participants were performed using regression models based on van der Waerden rank normal scores adjusted for age, sex, and race/ethnicity (white vs. nonwhite). Comparisons of the two T1D groups were also adjusted for A1C level. In addition, comparisons of SD between groups were adjusted for mean glucose because of the high correlation between SD and the mean glucose. The models were not adjusted for CGM device type because the short-term T1D group only used the Guardian Clinical. Previous studies have shown that the accuracy of the Real-Time system and FreeStyle Navigator are similar (6
). Moreover, the results were similar when the analysis was restricted to the Guardian Clinical data (data not shown). Analyses were performed with SAS version 9.2 (SAS Institute, Cary, NC). Only P
values <0.01 were considered statistically significant because multiple statistical comparisons.