877-P: Limits of Correlation Coefficient Analysis in Determining the Minimal Duration of CGM Data Needed to Estimate Time Below Range

Abstract

To design a clinical trial it is important to know for how long CGM data should be collected to accurately assess time in different glucose ranges (time in ranges). Several studies approached this problem through the computation of the correlation coefficient (R2) between a metric computed in a month-long trial and in several shorter windows of increasing duration. The minimal duration (MD) granting R2>threshold (e.g., 0.9) is then used to estimate the long-term metric. Here, we focus on time below range (TBR), defined as time <70 mg/dl [%]. We first implemented the R2-based analysis on trials of different duration: A1, A2, A3, lasting 100, 200, 300 days (d), respectively, obtained selecting portions of the same Study A (n=45, duration=360 d, sensor=Abbott Freestyle Libre). Table 1 shows that the longer the trial duration, the larger the resulting MD. Notably, all the obtained trial duration fractions are similar. Then, the analysis was repeated for other two trials of equal duration: Study B (n=85, duration=240 d, sensor=Dexcom G5) and A4, obtained selecting 240 d from Study A. Although B and A4 refer to different T1D populations and different sensors, the resulting trial duration fraction is the same. In conclusion, the R2-based analysis yields different results based on the duration of the considered dataset, and seems to identify only the fraction of data needed to match the reference TBR. Disclosure N. Camerlingo: None. M. Vettoretti: None. M. Cigler: None. A. Facchinetti: None. G. Sparacino: None. J.K. Mader: Advisory Panel; Self; Becton, Dickinson and Company, Eli Lilly and Company, Medtronic, Prediktor Medical, Sanofi-Aventis. Speaker’s Bureau; Self; Abbott, Eli Lilly and Company, Medtronic, Novo Nordisk A/S, Roche Diabetes Care, Sanofi-Aventis. P. Choudhary: Advisory Panel; Self; Abbott, Eli Lilly and Company, Insulet Corporation, Medtronic. Research Support; Self; European Union, JDRF. Speaker’s Bureau; Self; Dexcom, Inc., Novartis AG, Novo Nordisk A/S, Sanofi-Aventis. S. Del Favero: Research Support; Self; Dexcom, Inc. Funding Innovative Medicines Initiative 2 Joint Undertaking (777460); European Union; European Federation of Pharmaceutical Industries and Associations; T1D Exchange; JDRF; International Diabetes Federation; The Leona M. and Harry B. Helmsley Charitable Trust

Design of clinical trials to assess diabetes treatment: Minimum duration of continuous glucose monitoring data to estimate time‐in‐ranges with the desired precision

“…The consensus panel of Battelino et al 2 recommended a minimum CGM duration of 14 days to accurately estimate time‐in‐ranges. 10 , 19 However, it was recently shown in 20 , 21 that this approach 10 , 19 may produce inconsistent results in different datasets. The approach proposed in this work offers an alternative way to overcome this limitation.…”

“…The consensus panel of Battelino et al 2 recommends assessing 14 days of CGM, as the most recent 14 days of CGM data provide a good approximation of time‐in‐ranges collected over a 3‐month period. 10 , 19 However, this indication is empirical, 20 , 21 and the literature lacks a description of how the precision in the estimation of time‐in‐ranges improves as the trial duration increases.…”

Comparison Between Closed-Loop Insulin Delivery System (the Artificial Pancreas) and Sensor-Augmented Pump Therapy: A Randomized-Controlled Crossover Trial

“…Correlation analysis indicates that a duration of 4 weeks might be needed to reliably obtain representative numerical estimates of time-in-target range and mean glucose outcomes, 27 although using a correlation analysis to obtain this duration has been recently questioned. 28 Despite the significant improvement in time-in-target range with automated insulin delivery, the numerical reduction in mean glucose level of 0.4 mmol/L did not reach Table 3. Comparisons Between the Automated Insulin Delivery and Sensor-Augmented Pump Therapy During the Day (06:00-00:00) and Night (00:00-06:00)…”

An analytical approach to determine the optimal duration of continuous glucose monitoring data required to reliably estimate time in hypoglycemia

“…More important, it is easy to verify that the empirical method proposed in Ref. 17 yields different results based on the duration of the reference dataset 20 .…”