Variation of glucose time in range in type 1 diabetes

Abstract Introduction The aim of the study was to assess the variation of glucose time in range (TIR) for persons with type 1 diabetes who perform intermittently scanned continuous glucose monitoring (isCGM). Methods Glucose data for 8 weeks were analysed for 166 persons. TIR was calculated over four consecutive 2 weeks periods. Sixty‐one of the persons had two downloads with an interval of >3 months. Results A total of 140 individuals (84%) used multiple daily injection, and 26 (16%) used continuous insulin infusion. The within‐individual standard deviation (SD) for TIR was 6.3% corresponding to 95% limits of agreement for the difference between two TIR values of ±17.6%. Mean TIR calculated from the first and last 2 weeks was 52.2 ± 17.1% and 53.7 ± 16.4%, respectively (difference 1.5%, SD of the difference 10.4%, p = .07). For persons with two downloads separated by months, the SD of the difference in TIR was 12.6%. Conclusions The 95% limit of agreement for TIR is vast for persons using isCGM. It is difficult to draw firm conclusions regarding systematic differences when individual TIR from 2 weeks are compared. This may not be valid for users of insulin pumps with closed‐loop insulin delivery.


| INTRODUC TI ON
Continuous glycaemic monitoring (CGM) has given rise to several new glycaemic metrics as valuable alternatives to haemoglobin A1c. [1][2][3][4] The time in range (TIR) of glucose values is strongly correlated with mean glucose and HbA1c, [5][6][7][8] but TIR also reflects diurnal glucose variation and is independent of individual physiological factors influencing the rate of glycation. 9,10 Another advantage of TIR is that the effect of intervention can be evaluated after a few weeks, while only minor changes in HbA1c can be expected. 8,11 It is common clinical practice to present TIR and other glycaemic indices in a one-page condensed ambulatory glucose profile based on 14 days of CGM. 2,4 Information about spontaneous changes in TIR is sparse. This information is necessary to determine whether an observed clinically relevant change in TIR can be considered statistically significant or within the range of normal variation. TIR is an important outcome measure in clinical trials, and study dimensioning depends on information about the reproducibility of TIR.
Intermittently scanned continuous glucose monitoring (isCGM) is the most widespread form of CGM. The aim of the present study was to describe variation in TIR and other glycaemic metrics in persons with type 1 diabetes using isCGM.

| MATERIAL AND ME THODS
We had unrestricted access to isCGM (Freestyle Libre, Abbott) in the diabetes outpatient clinic for adults with type 1 diabetes in Regional Hospital Silkeborg, Denmark. 12 Glucose data were evaluated from all available downloads to the Diasend platform in the period February to November 2019. The study population comprises 169 non-pregnant individuals with type 1 diabetes of whom 61 had two downloads with an interval of more than 3 months. Glycaemic metrics from this cohort have previously been described in detail. 8 Glycaemic metrics were evaluated for periods of 2 weeks for the last 8 weeks before download. Period 1 was weeks one and two before download and period 4 was weeks seven and eight before download. For persons with two downloads, glycaemic metrics for the last 14 days from the first and second downloads were compared.
Mean glucose was calculated as the mean of glucose values Collection of clinical data was approved by the local institution.
No ethical approval was needed for this observational study.

| Statistical analysis
Data that were normally distributed by visual inspection of Q-Q plots are presented as mean ± SD, and paired data compared with Student's paired t-test. Data for periods 1-4 were analysed by repeated measured one way ANOVA with calculation of SD between and within subjects. The 95% limits of agreement, that is, the normal distribution prediction interval for the difference between two measurements for the same individual, was calculated as within  the difference in TIR (last-first measurement) was 1.7% (95% CI: −1.6-4.9) p = .31, SD of the difference 12.6% ( Table 2).
The risk that the difference between two TIR values by chance is larger than 10 percentage point is 26.4%, and the risk for a difference larger than 5% is 57.7%. If TIR is calculated from weeks 1-4 and weeks 5-8, the 95% prediction limits are ±13.1% (data not shown). The 95% prediction limits for mean glucose from 2 weeks are ±36 mg/dl (2.0 mmoL/L) and ± 0.9% (9.4 mmol/mol) for GMI (Table 1).
With the precautions needed due to winsorizing at the 10th percentile level of TBR, our results indicate that the limit of agreement for the ratio between two TBR values is between 3.86 and 0.26 (=1/3.86). Note: Data are mean ± SD or median (IQR). TBR is log transformed data after winsorizing TBR <0.6% (corresponding to the 10th percentile level for all four periods) and presented both as mean ± SD and geometric mean +/÷ tolerance factor. period, the 95% prediction limit is high (±13.1%). The international consensus statement for interpretation of CGM data recognize that even a small (5%) increase in TIR is associated with a glycaemic benefit. 1 A change in TIR of 10 percent point is considered clinically relevant for changes in retinopathy or albuminuria. 14,15 However, in the present study, a random change of more than 10 percentage point between two measurements of TIR is expected in more than 26.6% of the cases. In pregnancy, even a 5% change in TIR is clinically important. 16 This study has some limitations. First, the result of isCGM was not blinded and TIR results cannot be considered truly spontaneous.

| DISCUSS ION
The patients were expected to correct excursions in glucose and F I G U R E 1 A Bland-Altman plot of the difference TIR period 1 minus TIR period 4 for 166 persons with type 1 diabetes plotted against their average. The dotted line denotes the mean difference 1.5% (SD 10.4%) and 95% prediction interval (Limits Of Agreement) from −18.9 to 21.8%. TIR is calculated from 2 weeks glucose data TA B L E 2 Glycaemic metrics for 61 persons calculated from 2 weeks with an interval of more than 3 months Note: Data are mean ± SD or median (IQR). TBR is log-transformed data after TBR values <0.6% are winsorized and presented both as mean ± SD and geometric mean +/÷ tolerance factor. † Hodges-Lehman median difference and (95%CI).
*The active CGM time is statistical significant longer on the second measurement than in the first measurement.
increase TIR, but changes in TIR were small and statistically insignif-

ACK N OWLED G EM ENTS
We thank statistician Aparna Udupi, Biostatistical Advisory Service, Faculty of Health, Aarhus University, Denmark, for data management.

FU N D I N G I N FO R M ATI O N
The study was financially supported by the Rosa and Asta Jensen Foundation.

KWH has received honorarium as an advisory board member for
Abbott Laboratories A/S, Denmark. BMB has no conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.