Glycemic Control by Treatment Modalities: National Registry-Based Population Data in Children and Adolescents with Type 1 Diabetes Open Access

New technologies introduced into clinical practice usually lead to an improvement of T1D management and/or the well-being of persons with diabetes [1]. Nonetheless, reaching satisfactory glycemic outcomes still represents a challenge for a majority of children and adolescents with T1D (CwD). Recent data from the global SWEET initiative comprising 25,654 CwD treated in large centers on 6 continents showed that only 37% and 21% of CwD achieved the target HbA1c of <58 mmol/mol (7.5%) and <53 mmol/mol (7%), respectively [2]. The search for optimal therapeutic strategy at the population level thus remains an ongoing process.

The latest breakthrough in the treatment options was the introduction of hybrid closed-loop (HCL) systems. Several randomized controlled trials (RCTs) and real-life studies showed that HCL systems represent a safe and effective tool for improving the glycemic control in CwD [3‒7]. However, no direct comparison of HCL to other treatment modalities has been available on a population level yet. Furthermore, the “transition” period with only part of CwD switching to the novel therapy and multiple treatment modalities running in parallel is a unique opportunity for assessing the real-life effects of HCL.

The assessment of glycemic control became more complex after the introduction of CGM. The routinely used CGM-derived characteristics include times in glycemic ranges which provide more information than the well-established HbA1c. Specifically, the assessment of time in range (TIR), consensually defined as a time spent between 3.9 and 10.0 mmol/L [8], has become a golden standard of glucose control evaluation. Nevertheless, TIR as a single parameter does not adequately illustrate the whole picture of patients’ glycemic profiles – this parameter alone does not indicate whether the out-of-range readings are generally too low or too high [9]. Therefore, a new parameter named glycemia risk index (GRI) has been recently proposed as a suitable marker of glycemic state which especially emphasizes time in hypoglycemia as an important glycemic parameter [9]. Although several clinical studies aiming to compare the effect of the use of different treatment modalities on HbA1c, severe hypoglycemia, and diabetic ketoacidosis (DKA) in both children and adults with T1D have been performed [10, 11], the CGM-derived parameters have not been analyzed on pediatric population level with respect to age and type of treatment to date. Interestingly, Shah et al. [12] recently proposed 14 days of CGM sampling duration as an optimal period for estimation of GRI in adults with T1D.

Our hypothesis was that CwD treated with HCL achieve better results compared to those treated with other modalities. The aim of this study was to compare parameters of T1D control (HbA1c, times in glycemic ranges, GRI) in different groups of CwD using different treatment modalities using the data from the nation-wide pediatric diabetes registry. Additionally, we aimed to test the selected predictors of TIR and GRI (sex, age, T1D duration, and center size) on a population level.

The study was performed using the data from the nation-wide Web-based pediatric diabetes registry ČENDA (acronym for the Czech words for Czech National Pediatric Diabetes Database) described in detail elsewhere [13]. In short, this system collects anonymized data on every patient aged 0–18.99 years with any type of diabetes who is attending one of the participating pediatric diabetes outpatient clinics. More than 90% of all CwD in the country are included. In 2021, 52 out of 56 centers across the country provided patient data to the registry. A total of 4,066 CwD data were available. ČENDA is a benchmarking system that allows continuous anonymous online comparisons of selected key attributes of diabetes control, diabetic comorbidities, and acute (severe hypoglycemia and DKA) and chronic complications among diabetes clinics. All laboratory values are analyzed locally and batch transferred to the registry at least once a year. External quality control is performed for HbA1c measurements once a year in all biochemical laboratories in the country. HbA1c testing was performed using the IFCC standards in mmol/mol, with all measured records of HbA1c used for statistical analysis. This study was made exclusively on the basis of the annual data from 2021. ČENDA registry was approved by the Ethical Committee of the Motol University Hospital and registered with the Bureau for Personal Data Protection.

Type of treatment (multiple daily injections [MDIs], continuous subcutaneous insulin infusion without hybrid closed-loop functionality [CSII], and continuous subcutaneous insulin infusion with HCL) was collected and the date of the treatment modality change (if applicable) was registered in all patients. The ČENDA registry collects data on CGM usage based on medical reports and prescription data provided by the reporting diabetologist. Data on the CGM use are categorized according to the time spent on CGM in the past year: no use, <19%, 20–39%, 40–69%, 70+%. Time spent in standard glycemic categories [8] over the last 2 weeks before the visit was collected at every visit in children using the CGM.

Two commercial HCL systems were available in the country in 2021: t:slim X2 with Control-IQ algorithm (Tandem Diabetes Care, San Diego, CA, USA) was introduced in December 2020 (switching from the Basal-IQ algorithm by remote update) and Medtronic MiniMed 780G (Medtronic Inc. Minneapolis, MN, USA) since January 2021 sequentially replacing the previous pump generation with a predictive low glucose suspend function (MiniMed 640G). Insulin pumps without any of the automatic functions (Dana Diabecare RS [Sooil, Seoul, South Korea] and Accu-Chek Insight [Roche, Basel, Switzerland]) were also used by some Czech children in 2021. In addition, some patients used the unofficial do-it-yourself HCL system AndroidAPS [14] – information on the use of this system was also collected. All types of the abovementioned insulin pumps including consumables and all types of CGMs (Guardian 3 and 4 [Medtronic Inc. Minneapolis, MN, USA], DexCom G6 [DexCom, San Diego, CA, USA], FreeStyle Libre [Abbott, Chicago, IL, USA]) are fully covered without any significant financial burdens for CwD in the Czech Republic [15]. All centers caring for CwD have equal access to prescription of modern technologies.

Continuous variables were summarized as means with standard deviation or as medians with interquartile range (IQR), as appropriate. For each patient, possible multiple measurements of HbA1c and/or CGM-derived data were averaged into one value per patient and variable. Linear regression models were used to assess the differences in HbA1c and CGM-derived variables between modality groups; in case of post hoc pairwise comparisons, Tukey method was used to check for multiple comparisons. Multiple linear regression models were also used to explore the effect of other factors such as sex, age, duration of diabetes, and center size. In case of HbA1c measurements as a response, the values were log transformed.

The study flowchart is shown in Figure 1. Overall, of 4,066 CwD reported to the ČENDA registry in 2021, 3,860 (94.9%) were treated for T1D. 3,251 (84.2%) of them fulfilled the inclusion criteria and their data were included in the statistical analysis. A majority of them (67.3%, 2,187) were treated with MDI; 1,064 patients (32.7%) used the insulin pump. Any type of CGM data over 70% of the time were reported in 2,340 CwD (72.0%); in 2,025 of them (86.6%), data on times in glycemic ranges were available. HbA1c was known in 99.0% of CwD, whereas in 34 (1.0%) only CGM-derived parameters but no HbA1c was available. The basic demographic characteristics of the study group are presented in Table 1.

Mean annual frequency of HbA1c testing was 3.3 ± 1.1/year and the median of means of the CwD was 56.5 mmol/mol (IQR 42.4–70.6 mmol/mol). The analysis of the mean annual HbA1c by treatment and monitoring modalities is shown in Table 2 and Figure 2 and illustrates the effectiveness of CGM for lowering HbA1c. Both noCGM groups achieved significantly higher HbA1c values in comparison with CGM users (Table 2, p < 0.001). No statistically significant difference was observed in HbA1c between MDI rtCGM, CSII, and HCL (Tukey post hoc p values 0.11, 0.36, and 0.93). These results have not changed after the adjustment for age and diabetes duration. Significant difference in the medians of HbA1c was observed between MDI, isCGM, and MDI rtCGM groups (5.2 mmol/mol [0.5%], p < 0.001).

Mean times in glycemic ranges for the whole study group of 2,025 CwD with available CGM data were as follows: TIR 66.3%, VLow 1.7%, Low 5.3%, High 20.1%, VHigh 6.6%. Significant differences in these parameters were observed by treatment modalities (Fig. 3, Table 3). The highest TIR was reported in CwD using HCL (76.5%, p < 0.001 vs. any other group) followed by MDI rtCGM and CSII with TIR 68.8% and 69.0% (ns between each other), respectively. The lowest TIR was observed in the MDI isCGM group (60.1%, p < 0.001 vs. any other group) which has also showed the highest proportion of time spent in hypoglycemia (9.1%, VLow and Low aggregated). The HCL group presented with the lowest times in Low and VLow.

The relationship between TIR and age by treatment modalities and T1D duration is illustrated in online supplementary Figure 1 (for all online suppl. material, see https://doi.org/10.1159/000530833). While the data derived from patients treated with HCL and CSII show a constant, age-independent TIR, significant decrease in TIR with age is noticeable in the MDI groups (online suppl. Fig. 1a). The course of TIR and T1D duration shows typical decrease in TIR in the first years following the T1D onset in all but HCL groups (online suppl. Fig. 1b). The heatmap combining T1D duration and age shows the homogeneous distribution of high TIR within the HCL group while all the other groups presented with strong age and T1D duration dependency (online suppl. Fig. 2).

The median GRI for the whole group of 2,025 patients was 41.6 ± 28.2 (IQR) and mean 44.2 ± 20.8 (SD). After the division to the study groups, the lowest GRI median (29.1, IQR 22.3–37.9, p < 0.001 vs. all other groups) was observed in the HCL group, followed by MDI rtCGM and CSII (38.8, IQR 28.2–53.2, and 40.1, IQR 30.2–49.2, respectively) (Table 3; online suppl. Fig. 3). There were no significant differences in median GRI between the latter groups. The highest GRI was present in CwD treated with MDI and monitored by isCGM (p < 0.001 vs. all other groups) (Table 3; online suppl. Fig. 3). Online supplementary Figure 4 shows the hypoglycemia/hyperglycemia components revealing higher glycemic stability and lower risk of both hypo- and hyperglycemias in CwD using HCL. CwD using the isCGM had higher hypo- and hyperglycemic component of GRI than other groups.

The changes of GRI with age and T1D duration by treatment modalities are illustrated in online supplementary Figure 5. GRI increases with age in the MDI isCGM group but remains constant in MDI rtCGM and CSII groups and decreases in HCL (online suppl. Fig. 5a). The relationship between GRI and T1D duration generally corresponds to TIR (online suppl. Fig. 5b). GRI increased marginally with T1D duration (LM model, p = 0.056).

Sex, age, T1D duration, and center size were included into univariate linear regression model as potential predictors of TIR and GRI for particular study groups. The analysis revealed T1D duration as a significant predictor of lower TIR and higher GRI. Age predicts lower TIR in MDI groups only, whereas higher GRI was predicted by age in the MDI isCGM group only. An increase of age by 1 year predicts a decrease in GRI by 0.39 (p = 0.006) (online suppl. Table 1).

The introduction of insulin pumps with HCL functionality and routine use of CGM systems in people with T1D raised legitimate questions about their effectiveness. Up to our best knowledge, we present the first registry-based pediatric population data comparing key parameters of glycemic control of the six frequently used treatment modalities. Our data support the concept that HCL is the most effective way to achieve optimal glycemic control regardless of age.

Although the results of clinical studies published since 2013 have demonstrated the effectiveness of HCL for T1D control [3‒7, 16], this modality entered the common clinical practice only in the last 2 years. Since 2020, CwD living in high-income countries were given the opportunity to use the HCL technology, but the proof of its efficacy on the population level was still missing. Our registry-based pediatric data from a country with full availability of modern technologies [15] describe significant differences in TIR and GRI between HCL and other modalities but not in HbA1c in the first year after the introduction of this technology. The discrepancy in results between HbA1c and CGM-derived parameters is probably caused by the fact that HbA1c value is not an ideal marker of glycemic control mainly because it does not take the fluctuations of glycemia into account. This concept is also supported by the lowest time spent in hypoglycemia observed in the HCL group as hypoglycemia is an established factor in decreasing the HbA1c value. The additional explanation why no HbA1c difference between HCL and other groups using rtCGM was demonstrated lies in the fact that the study was performed in a population with a median HbA1c of 53.3 mmol/mol (7.0%) in 2021 (ČENDA 2021 annual report) which is lower than most of the other registries of pediatric diabetes [2]. As the lowering of HbA1c is challenging in CwD with favorable baseline parameters, studies performed in CwD with higher baseline/background HbA1c might present different results [3, 5].

Currently available HCL systems have their limits in terms of improving the diabetes control. Interestingly, most studies performed in adults and adolescents – irrespective of whether they were RCTs or real-life – showed a TIR around 75% in the HCL group [17‒19] which corresponds precisely to our population-wide results. Recently, a German single-center study described significant differences in TIR between pre-school (73%) and school (69%) CwD treated with HCL [3]. Interestingly, we did not observe any such age dependence which is of particular note especially in adolescence, the period of life with the lowest percentage of CwD achieving the optimal glucose control [20‒22]. We can therefore hypothesize that the continuing implementation of currently available HCL technologies would be able to diminish or even erase the age-specific differences in diabetes control but will probably not be effective enough to achieve normoglycemia on a population level.

Our results further add to the basic paradigm of current diabetology, i.e., the necessity for frequent use of CGM. The apparent difference is noticeable in HbA1c (Fig. 3) with frequent CGM users (>70% of the time) achieving lower HbA1c by 12–14 mmol/mol (1.1–1.3%) than the nonusers or intermittent users, which is also clinically relevant for the increased risk of late complications development [22, 23]. This fact is in line with the RCT Comisair study showing that rtCGM is a key tool for lowering HbA1c in adults with T1D irrespective of the treatment modality (MDI vs. CSII) [24]. Our results (nonsignificant difference in HbA1c, TIR, and GRI between MDI rtCGM and CSII) are in line with this conclusion, but our observation of better performance of HCL in most of the CGM-derived parameters over the other groups monitored with rtCGM adds an important argument for HCL implementation in CwD as a treatment method of choice keeping in mind that HCL has been associated with an increase in BMI with potential implications on the cardiovascular risk factors [25].

The assessment of self-monitoring efficacy between isCGM and rtCGM in CwD treated with MDI is still a subject of debates among diabetologists and people with diabetes [26]. In our study, we present consistently favorable results of the MDI rtCGM compared to MDI isCGM group in all of the analyzed parameters of glycemic control in a large group of CwD. Similar results with higher TIR and reduced time in hypoglycemia on rtCGM were published by Hásková et al. [27] and Visser et al. [28] using the RCT design. Interestingly, a study of 666 Italian children observed similar HbA1c (7.6% vs. 7.5% DCCT) but lower TIR (49% vs. 56%) in MDI isCGM versus rtCGM groups [11]. Although this notion might be subject to the selection bias in our registry-based observational study (CwD using isCGM could be less focused on glycemic targets then rtCGM users), our data are in accordance with the concept that the presence of alarms likely provides an effective means of hypo- and hyperglycemia prevention in CwD on a population level.

The CGM data from the ČENDA registry are quite unique in the stratification of the time of CGM use as most of the registry-based studies present CGM usage only categorically (yes/no/not known) which tends to oversimplify as CwD are stratified according to the percentage of their CGM use in real life. Finer and in fact more realistic categories of CGM use applied by the ČENDA registry increase the validity of the presented results and belong to the strength of this study. The CGM metrics were collected over the last 2 weeks before the visit as these data correlate strongly with 3 months of mean glucose, TIR, and hyperglycemia metrics [29] as well as GRI [12].

GRI is one of the newest metrics of diabetes control – here we present its first use in population-based pediatric data. The patterns of GRI in four assessed treatment modalities generally correspond well to the time in glycemic ranges analysis and substantially less with HbA1c owing to the relation with the calculation method of GRI. The mild decrease of GRI with age is however still somehow surprising and needs other clarification or possibly even the adjustment of the index formula. Our GRI results are currently impossible to compare with other registry-based studies as the only study was based on a limited number of CGM tracing gained from a heterogenous group of adults [9].

The study has several limitations. First, as in other population-based registries, we cannot prove any causality. We thus cannot exclude that children using HCL are better educated/motivated or have a more stimulating family environment. Nevertheless, the number of participants was large in all of the analyzed groups which should decrease the potential bias. Second, this study was limited to 1 year and we thus cannot exclude the “new therapy effect” during the first months of the HCL use. Long-term follow-up study is therefore warranted. Third, we cannot present data on the frequency of acute complications by treatment modalities because the ČENDA registry does not collect the exact dates of severe hypoglycemia or DKA episodes and we thus cannot link the event with the currently used treatment modality in a population with rapidly changing treatment preferences. However, published data uniformly conclude that HCL does not increase the risk of severe hypoglycemia or DKA [5‒7]. Lastly, presented data cannot be linked with pubertal status as data on pubertal stage are not collected in the registry.

In conclusion, this population-based study shows that HCL technology is superior to other treatment modalities in CGM-derived parameters in CwD regardless of age. Therefore, HCL should be considered a treatment modality of choice to all CwD fulfilling the indication criteria.

The authors thank the remaining members of the ČENDA Study Group (in alphabetic order of cities) – Drs: J. Vyžrálková, M. Pejchlová (Brno), M. Schubert (Bruntál), L. Kocinová (Česká Lípa), I. Röschlová (Frýdek-Místek), A. Kudličková (Hodonín), A. Benešová (Chomutov), J. Češek (Chrudim), M. Adam (Jablonec), P. Kracíková (Jičín), P. Vlachý, M. Svojsík (Jihlava), M. Jiřičková (Jilemnice), K. Poločková (Karviná), J. Kytnarová (Prague-VFN), K. Dimová (Kladno), S. Fialová (Kroměříž), M. Vracovská (Klatovy), J. Sivíčková, P. Pelcová (Karlovy Vary), E. Hlaďáková (Kyjov), J. Bartošová (Liberec), M. Kulinová (Mladá Boleslav), N. Filáková (Ostrava – City Hospital), M. Romanová (Prague-FNKV), M. Šnajderová, S. Koloušková (Prague – Motol), L. Týce (Náchod), E. Farkaš (Nový Jičín), Z. Ježová (Nové Město na Moravě), M. Honková (Opava), B. Červíčková (Pardubice, Trutnov), T. Farová (Písek), K. Fiklík (Plzeň), J. Malý (Sokolov), M. Gregora (Strakonice), J. Malý (Svitavy), A. Lysáková (Šumperk), J. Chocholová (Tábor), P. Eichl (Teplice), O. Michálková (Třebíč), M. Struminský (Třinec), M. Procházka (Ústí nad Orlicí), H. Vávrová (Vsetín), P. Gogelová (Zlín), and P. Mikyška (Znojmo). Marie Kajprová and Jakub Šumník are gratefully thanked for help with data entry.

Written informed consent was obtained from parents/guardian/next of kin for all patients aged under 18 years. Adult participants have given their written informed consent to participate in the study. The ČENDA registry was approved by the Ethical Committee of the Motol University Hospital (approval Nol. 154/2012) and registered with the National Bureau for Personal Data Protection. Ethical approval is not required for this particular study in accordance with local or national guidelines.

Z.S. and L.P. reported speakers’ honoraria from Medtronic, Abbott, and A-Import.

This study was supported by grants from the Czech Diabetes Society and the Czech Ministry of Health (conceptual support project to research organization 00064203 – FN Motol and AZV grant NU21-01-00085).

Z.S., V.N., Le.P., P.K, P.V., Ja.S., R.P., D.N., J.V., Ji.S., K.K., B.O., Lu.P., A.S., S.P., and O.C. contributed to study concept and design and acquisition of data. Z.S., O.C., and M.P. drafted the manuscript. M.P. provided statistical analysis. All authors approved the final version of the manuscript. Z.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Data are not publicly available due to ethical reasons. Further inquiries can be directed to the corresponding author.