Of a total of 231,102 Glic™ platform users, 12,727 individuals were eligible based on the study eligibility criteria. Their data were extracted on 19th October 2023 and these individuals were included in this study. Of those included, 11,007 (86.5%) reported their glucose monitoring method to be BGM, while 1720 (13.5%) reported using fCGM. Overall, 70.5% of individuals had T1DM, 20.9% T2DM, 4.5% GDM and 4.1% LADA. There was a significant higher proportion of individuals with T1DM among fCGM users (T1DM 87.6% vs. other types of diabetes 12.4%) compared to BGM (T1DM 67.8% vs. other types of diabetes 32.2%) (p < 0.001). The mean overall age was 35.61 years (SD 16.14), and fCGM users were significantly younger (31.62 years, SD 17.32) than those who perform BGM (36.23 years, SD 15.86) (p < 0.001). Overall, 39.1% were male, with a significant higher proportion of males among fCGM users (41.9%) compared to BGM (38.6%) (p = 0.011). Most individuals (57.0%) were from the Southeast region of Brazil, followed by 17.6% from the South region, 13.7% Northeast region, 9.1% Central-west region and 2.5% North region. There was a significant higher percentage of fCGM users (59.0%) in the Southeast region compared to BGM (56.7%) (p < 0.001). The mean number of years of diabetes diagnosis was 9.10 years (SD 10.41), with fCGM users having a significant longer diagnosis time (10.65 years, SD 11.51) compared to BGM (8.87 years, SD 10.21) (p < 0.001). The mean body mass index (BMI) was 25.66 kg/m2 (SD 5.34), with fCGM users having a significant lower BMI (23.73 kg/m2, SD 4.60) compared to BGM (25.94 kg/m2, SD 5.38) (p < 0.001). Meanwhile, the mean Z-score in children and adolescents (age < 18 years) was 0.13 (SD 1.05), with no significant difference between fCGM and BGM (p = 0.408). In terms of use of the Glic™ platform, on average Glic™ users used the platform to input their glucose levels 35.82 times (SD 125.43), with fCGM users using Glic™ more frequently (50.84 times, SD 241.41) than BGM (33.48 times, SD 95.13) (p = 0.009). Table 1 presents the characteristics of the study population.

Comparing the different types of diabetes, individuals with T1DM were younger (30.85 years, SD 14.48) than those with GDM (33.55 years, SD 5.95), LADA (41.31 years, SD 12.58) and T2DM (51.00 years, SD 13.45). Interestingly, only for T1DM, fCGM users were younger (29.39 years, SD 16.34) than those who perform BGM (31.15 years, SD 14.05) (p = < 0.001). On the contrary, for T2DM, fCGM users were older than those who perform BGM, with mean ages of 55.37 years (SD 15.70) and 50.86 years (SD 13.35) (p = 0.002) in T2DM. Meanwhile, for GDM and LADA, there were no significant differences in age between fCGM and BGM. Majority of individuals were female for T1DM (62.4%) and LADA (67.9%); however, most individuals were male for T2DM (53.7%). Similarly to the overall study population, there were more males using fCGM compared to BGM in T1DM (41.5% vs. 36.9%, p = 0.001), while there were no statistically significant differences in sex percentages in T2DM and LADA between the two glucose monitoring types. Regarding the country region, for T1DM, GDM and LADA, there were similar regional distribution to the overall study population, with more individuals living in the Southeast region (56.8%, 59.3%, and 60.6%, respectively), followed by South region (18.8%, 17.5%, and 18.5%, respectively), Northeast region (12.7%, 13.4%, and 12.0%, respectively), Central-west region (9.7%, 7.7%, and 7.0%, respectively), and North region (2.0%, 2.1%, and 1.9%, respectively). However, only for T1DM there was statistically significant difference in region distribution comparing those using fCGM compared to BGM, with a higher proportion of fCGM users living in the Southeast region compared to BGM (59.4% vs. 56.3%, p < 0.001). For T2DM, there was a different pattern, with more individuals living in the Southeast region (56.7%), followed by Northeast region (17.2%), South region (13.6%), Central-west region (7.9%), and North region (4.6%), but also with no statistically significant difference in region distribution between individuals using fCGM compared to BGM. In terms of years of diabetes diagnosis, T1DM individuals had the longer diagnosis time (11.29 years, SD 10.81), followed by LADA (6.99 years, SD 7.58), T2DM (4.58 years, SD 8.03) and GDM (0.83 years, SD 1.39). For T1DM, fCGM users had a shorter diagnosis time compared to BGM, with mean number of years of diabetes diagnosis of 11.02 (SD 11.72) and 11.35 (SD 10.62) (p = < 0.001), respectively. Conversely, for T2DM and GDM, fCGM users had longer diagnosis time compared to BGM, with mean number of years of diabetes diagnosis of 9.78 (SD 11.88) and 4.41 (SD 7.81) (p = < 0.001) in T2DM, and of 3.00 (SD 5.64) and 0.79 (SD 1.16) (p = 0.003) in GDM. Meanwhile, there were no significant differences in years of diabetes diagnosis between LADA individuals who used fCGM or BGM (p = 0.444). Individuals with T2DM (29.33 kg/m2, SD 5.45) and GDM (29.08 kg/m2, SD 5.85) had a higher BMI than LADA (25.12 kg/m2, SD 4.26) and T1DM (24.33 kg/m2, SD 4.65). Similarly to the overall study population, the BMI was lower in fCGM users compared to BGM in T1DM and T2DM, while there were no statistically significant differences in GDM and LADA. Regarding the frequency of Glic™ use, individuals with LADA (47.11 times, SD 128.17) were the ones who used Glic™ more frequently, followed by T1DM (42.24 times, SD 132.21), GDM (24.35 times, SD 48.11) and T2DM (14.42 times, SD 109.37). Contrary to the overall study population, there were no statistically significant differences in frequency of Glic™ use between fCGM users and BGM in all four types of diabetes. Supplementary Tables 1, 2, 3 and 4 presents the characteristics of the study population by type of diabetes.

In terms of basal insulin prescription, overall, 64.8% had a prescription, with more fCGM users being prescribed a basal insulin (83.8%) than those who perform BGM (61.8%). The type of basal insulin most used were long-acting analogs (62.2%), followed by ultralong-acting analogs (19.3%) and NPH (17.8%). Among the long-acting analogs, Glargine was the most preferred insulin (98.1%), meanwhile among ultralong-acting analogs it was Degludec (93.9%). Comparing basal insulin prescription patterns, fCGM users were prescribed ultralong-acting analogs more frequently (43.5%) than those who perform BGM (14.2%) (p < 0.001). Regarding basal insulin doses, the mean total basal daily dose was 26.73 units/day (SD 15.35) and the mean total basal daily dose per kilogram was 0.41 units/kg/day (SD 0.21), with fCGM users having significantly lower doses than BGM, with mean total basal daily doses of 22.14 units/day (SD 13.74) and 27.68 units/day (SD 15.49) (p < 0.001), respectively, and the mean total basal daily doses per kilogram of 0.36 units/kg/day (SD 0.19) and 0.42 units/kg/day (SD 0.22) (p < 0.001), respectively. The mean number of daily injections of basal insulin was 1.48 (SD 1.42), with fCGM users having significantly lower mean number of daily injections (1.20, SD 0.80) than BGM (1.53, SD 1.51) (p < 0.001). The mean percentage of TBD/TDD was 56% (SD 18%), with fCGM having significantly lower ratio (55%, SD 17%) than BGM (57%, SD 18%) (p = 0.015).

In terms of bolus insulin prescription, overall, 55.5% had a prescription, with more fCGM users being prescribed bolus insulin (79.2%) than those who perform BGM (51.8%). The type of bolus insulin most used were short-acting analogs (74.7%), followed by ultrashort-acting analogs (16.1%) and human regular (9.2%). Among the short-acting analogs, aspart the most preferred insulin (53%), followed by glulisin (23.9%) and lispro (23.2%). Comparing bolus insulin prescription patterns, fCGM users were prescribed ultrashort-acting analogs more frequently (37.2%) than those who perform BGM (11.0%) (p < 0.001). The mean total bolus daily dose was 22.77 (SD 22.61) units/day, with fCGM users having significantly lower doses (19.93, SD 18.61) than BGM (23.45, SD 23.43) (p < 0.001). The mean total bolus daily dose per kilogram was 0.36 (0.35) units/kg/day, with no significant differences between fCGM and BGM.

In terms of oral anti-diabetics, only 12.4% of the study population had a prescription of an oral anti-diabetic, with fCGM having a significantly lower percentage of use (6.4%) than BGM (13.3%) (p < 0.001). Of the different classes of oral anti-diabetics, metformin was the most used (51.2%), followed by combinations of metformin + sulfonylurea (9.5%), combinations of metformin + SGLT2 inhibitors (9.3%), and SGLT2 inhibitors (6.4%), among other classes and combinations. SGLT2 inhibitors were used in a significantly higher percentage of fCGM users (18.2%) than those who perform BGM (5.5%) (p < 0.001). Table 2 presents detailed information on treatment patterns.

Comparing the different types of diabetes, individuals with LADA had the highest percentage of basal insulin prescription (88.5%), followed by T1DM (79.5%), T2DM (22.8%) and GDM (7.3%), with more fCGM users being prescribed basal insulin than those who perform BGM, except for LADA for which there were not significant difference. The type of basal insulin most used were long-acting analogs for T1DM (65.8%) and LADA (55.7%), while human NPH was the one most used for GDM (83.3%) and T2DM (58.7%). Comparing basal insulin prescription patterns, fCGM users were prescribed ultralong-acting analogs more frequently than those who perform BGM for all types of diabetes. Regarding bolus insulin treatment, individuals with LADA had the highest percentage of bolus insulin prescription (72.1%), followed by T1DM (71.3%), T2DM (10.1%) and GDM (2.6%), with more fCGM users being prescribed bolus insulin than those who perform BGM in all types of diabetes, except LADA. The type of bolus insulin most used were short-acting analogs for T1DM (77.0%) and LADA (66.3%), while human regular was the one most used for GDM (60.0%) and T2DM (45.4%). Comparing bolus insulin prescription patterns, fCGM users were prescribed ultrashort-acting analogs more frequently than those who perform BGM for all types of diabetes, except GDM. More detailed information on the prescription patterns per type of diabetes are presented in Supplementary Tables 5, 6, 7 and 8.

In terms of glucose data manually inserted in the app by the user, there were only 841 individuals in the study population of Glic™ app users who attended the criteria for the glycemic control analyses. Of those, 657 had T1DM, 96 had T2DM, 47 had GDM and 41 had LADA. Considering that T1DM and LADA represented more than 80% (698/841) of this glucose data cohort, have similarities in terms of pathophysiology of endogenous insulin deficiency and treatment, and that the population profile of these two types of diabetes was similar between fCGM users and BGM (See Supplementary Table 9), the glucose data analyses were grouped for these two types of diabetes. Comparisons of population profile between all eligible patients and this glucose data cohort are also presented in Supplementary Table 10. The mean 24 h glucose level was 180.24 mg/dL (SD 46.76), with fCGM users having a non-significant lower glucose level (174.58 mg/dL, SD 39.40) than those who perform BGM (181.41 mg/dL, SD 48.10) (p = 0.212). When this population of T1DM and LADA were split into adults (> 18 years) and children and adolescents (< 18 years), the mean 24 h glucose level in adults was lower (179.04 mg/dL, SD 47.34) than in children and adolescents (185.69 mg/dL, SD 43.81). In both age groups, fCGM users had a non-significant lower glucose level than those who perform BGM, with glucose levels of 174.27 mg/dL (SD 38.90) and 179.80 mg/dL (SD 48.55) (p = 0.467) in adults and of 175.18 mg/dL (SD 40.83) and 190.75 mg/dL (SD 44.52) (p = 0.072) in children and adolescents, respectively. Table 3 presents the detailed glucose levels analyses for the T1DM and LADA individuals, including nocturnal, pre-prandial and post-prandial glucose levels. Supplementary Tables 11 and 12 presents the detailed glucose levels analyses for the adults and children and adolescents with T1DM and LADA. For T2DM, the mean 24 h, nocturnal, pre-prandial and post-prandial glucose levels were 144.28 mg/dL (SD 48.33), 156.36 mg/dL (SD 70.19), 139.77 mg/dL (SD 39.95), 161.98 mg/dL (SD 68.53), respectively. For GDM, the mean 24 h, nocturnal, pre-prandial and post-prandial glucose levels were 102.47 mg/dL (SD 9.02), 110.24 mg/dL (SD 17.95), 88.43 mg/dL (SD 7.76), 112.12 mg/dL (SD 9.65), respectively. Comparisons between fCGM users and BGM were not possible for these two types of diabetes as there were only two and one fCGM users among these T2DM and GDM individuals, respectively.

In terms of factors associated with a better glycemic control, in a population of 376 individuals with all four types of diabetes for which there were complete data, it was found in this present study that a higher frequency of use of Glic™ and a higher percentage of TBD/TDD were associated with a higher odds of being euglycemic. On the contrary, obesity was associated with dysglycemia. The use of fCGM tended to be associated with euglycemia compared to BGM; however, this association was not statistically significant. There were no associations with a better glycemic control for all other factors, including age, country region, years of diabetes diagnosis, class of basal and bolus insulin, number of basal insulin injections and carbohydrate/insulin ratio (Fig. 1).

Forest plot of factors associated with a better glycemic control

BMI, body mass index, CI, confidence interval; GLP1, glucagon-like peptide 1; OR, odds ratio; SD, standard deviation; TBD/TDD, total basal dose/total daily dose

Regarding the switch of glucose monitoring method from BGM to fCGM, of the 211 individuals included in this analysis, 35% were dysglycemic when using BGM. This number dropped to 30% after switching to fCGM, but without statistical significance (p = 0.24). There was also no difference in the therapeutic regimen (basal and bolus insulin) after changing from BGM to fCGM.