figure b

Introduction

Adjustment of bolus insulin dosage based on carbohydrate intake and frequent blood glucose monitoring (BGM) or continuous glucose monitoring (CGM) are suggested approaches to obtain optimal glycaemic control in type 1 diabetes treated with multiple daily insulin injections (MDI) [1,2,3,4,5].

Advanced carbohydrate counting, defined as bolus calculation based on carbohydrate ratios and insulin sensitivity [6], is known to improve HbA1c and treatment satisfaction compared with experience-based insulin bolus dosing [7,8,9]. Training in automated bolus calculation (ABC), rather than manual bolus calculation, may be even more efficient [8]. Conflicting results are available regarding influence on hypoglycaemia rate [10]. Sustained HbA1c reduction after introduction of ABC has been demonstrated even in a routine care setting [11].

Intermittently scanned CGM (isCGM) provides CGM data on request but has, until recently, had no alarms. Several studies have evaluated isCGM in terms of glycaemic effectiveness and person-related outcomes [12,13,14,15,16]. IsCGM has been found to decrease hypoglycaemia and glycaemic variability and increase treatment satisfaction in people with near-optimal HbA1c [17]. Predictors of HbA1c reduction are high baseline HbA1c [14, 16, 18,19,20] and frequent scans [18, 21].

Despite promising effects of several interventions, only a minority of people with type 1 diabetes achieve glycaemic goals [22]. While ABC presupposes user engagement and skills, isCGM is easily usable and widely desired. In our local setting, ABC is broadly introduced, while isCGM is only reimbursed in certain subpopulations. On this background, we intended to investigate the abovementioned interventions in the vast group of people with type 1 diabetes who have not yet obtained target glycaemic control, practised ABC or fulfilled the strict Danish criteria for isCGM reimbursement.

The primary aim of this trial therefore was to examine the effect of isCGM on glycaemic control compared with usual care, i.e. experience-based insulin bolus decision and BGM, in people with suboptimal HbA1c and naive to diabetes technology and advanced carbohydrate counting. In addition, we aimed to assess patient-reported outcomes among isCGM users compared with those allocated usual care, whether there is an additive effect of introducing ABC on top of isCGM and whether isCGM is superior to ABC.

Methods

Study design and participants

The study protocol was previously published, including details on usual care, interventional devices, screening and randomisation procedures, etc. [23].

The study is a randomised, controlled, multicentre, open-label trial carried out in five diabetes clinics in the Capital Region of Denmark. Inclusion criteria were age ≥ 18 years, type 1 diabetes ≥1 year, HbA1c > 53 mmol/mol (7.0%) and MDI with basal insulin ≥30% of total insulin dose. Exclusion criteria were ongoing practice of daily advanced carbohydrate counting, CGM/isCGM, NPH insulin, pregnancy, breastfeeding, gastroparesis, severe diabetes complications (i.e. proliferative retinopathy, myocardial infarction within the last 6 months), other medical or psychological conditions judged unsuitable for participation, participation in other diabetes-related clinical research projects, other drugs than insulin affecting glucose metabolism or the inability to understand the information and to give informed consent. The inclusion period was 1 October 2018 to 30 April 2020.

The trial was carried out in accordance with the Helsinki Declaration after approval by the Capital Regional Scientific Ethics Committee (H-17040573). The data collection was performed in accordance with the General Data Protection Regulation (P-2019-107). The trial is registered at ClinicalTrials.gov (registration no. NCT03682237).

Screening and randomisation

Possible candidates with interest in participation attended a screening visit, and those eligible were randomised 1:1:1:1 to usual care, ABC, isCGM or ABC+isCGM, with centrally prepared envelopes left unopened until after baseline blinded CGM data had been obtained. In the case of dropout, the participant’s allocation was not replaced [23].

Intervention

All participants attended a group course (4.5 h duration, 4–6 participants, week 0), five individual study visits (weeks 2, 4, 12, 24 and 26) and two telephone consultations (weeks 8 and 17) according to treatment allocation (electronic supplementary material [ESM] Fig. 1) [23]. All group courses were held by a study nurse and a study dietitian, each contributing to their areas of expertise. The screening and the midway visits were held with a study physician. Participants allocated ABC had a joint nurse–dietitian appointment for the first follow-up visit. The remaining study visits were held by a nurse solely for all participants. Participants not allocated ABC received common dietary advice for people with diabetes but were not taught bolus calculation based on carbohydrate ratios and insulin sensitivity. The ABC device selected was the Food and Drug Administration (FDA)-approved and Conformité Européenne (CE)-marked mobile phone application mySugr (mySugr, Austria) consisting of a bolus calculator adjustable on carbohydrate ratio, insulin sensitivity, target glucose value and insulin duration time. Before meals, the current glucose value and estimated carbohydrate intake are entered, and the user receives the suggested insulin bolus dose. Similarly, to correct hyperglycaemia outside meals, the user enters the current glucose value and receives the suggested corrective insulin dose. The isCGM device was the FDA-approved and CE-marked FreeStyle Libre Flash CGM system (Abbott Diabetes Care, Alameda, CA, USA), consisting of an upper arm sensor with a 14 day lifespan and a reading device (either an independent reader or a mobile phone application: FreeStyle Libre Link, Abbott Diabetes Care). The user scans the sensor with the reading device for on-demand values. Automatically obtained glucose values are continuously gathered for later review. The isCGM has a built-in bolus calculator, which can be used only with BGM measurements and not with scanned isCGM sensor values, which is why another bolus calculator was chosen for the study.

Hence, the group course elements were: general diabetes education (usual care and isCGM arms); training in carbohydrate counting, bolus calculation and mySugr application (ABC and ABC+isCGM arms); training in FreeStyle Libre Flash (isCGM and ABC+isCGM arms); and training in how to incorporate glucose trend arrows to adjust experience-based bolus dosing (isCGM arm) or the mySugr application settings (ABC+isCGM arm). Participants were also encouraged to continuously refer to isCGM data, i.e. current and average sensor glucose, time in range (TIR), as well as hypoglycaemic events on a weekly basis, and to adjust insulin or bolus calculator settings if needed. These different elements were also in focus at all study visits, where the study personnel titrated insulin doses based on the different types of glucose values, information on hypoglycaemic events, planned physical activity, etc. The overall aim was fasting and preprandial glucose values 4–6 mmol/l, postprandial glucose values <10 mmol/l, to prevent glucose values <3 mmol/l and to minimise glucose values <4 mmol/l [24]. Participants were encouraged to perform CGM scans ≥5 times daily (isCGM and ABC+isCGM arms, to decrease the risk of missing values, unless scanned every 8 h) or BGM ≥4 times daily (usual care and ABC arms, to secure a glucose measurement prior to all main meal boluses and before bedtime) and to use the ABC before carbohydrate intake and to correct hyperglycaemia (ABC and ABC+isCGM arms). Basal and bolus insulin doses, and ABC settings for those allocated ABC, were repeatedly evaluated during participation and adjusted if needed [6, 8]. Study healthcare professionals had received adequate education in the different group course contents and treatment modalities before first screening [23].

Sample size

The intervention was designed to have 80% power to detect a difference in mean TIR between isCGM and usual care arms of 75 min/day [25] with an SD of 120 min [2], and a two-sided α level of 0.05. A sample size of 160 (40 per arm) was needed and was further increased to 180 to account for discontinuation.

Data collection

At baseline and study end the participants wore a 14 day blinded CGM device (the FreeStyle Libre Professional CGM system, Abbott Diabetes Care, Oxon, UK) (participants in the isCGM arms wore both types of sensors for the last 2 weeks of participation), had blood and urine samples taken, had HbA1c locally analysed with high-performance liquid chromatography on a Tosoh G7 (Tosoh Bioscience, Japan), had body weight measured and filled in work sheets for daily reporting of insulin doses and the following questionnaires: the Problem Areas in Diabetes Questionnaire (PAID), the Diabetes Treatment Satisfaction Questionnaire (‘status’ and ‘change’ versions; DTSQs and DTSQc) [26], the Diabetes Empowerment Scale (DES short form) and Diabetes quality of life (ADDQoL-19) [27, 28]. At baseline, hypoglycaemia awareness was also assessed [29, 30]. IsCGM (scans/day) and ABC (application use/day) data and the number of weekly mild symptomatic hypoglycaemic episodes were collected at 2, 12 and 26 weeks. Adverse events were consecutively recorded.

Outcome measures

The primary outcome was the difference in change from baseline to study end between isCGM and usual care arms in TIR (minutes/day, obtained by blinded CGM). Due to consensus guidelines on glucose values obtained with CGM published after trial initiation, the range for TIR was redefined from 4–10 to 3.9–10 mmol/l [25].

Predefined and analysed secondary outcomes are: difference between groups in change in HbA1c (mmol/mol, %), TIR, time in hypo- (<3, <3.9 mmol/l) and hyperglycaemia (>10 mmol/l) (minutes/day), glycaemic variability (SD), total insulin dose (U kg-1 day-1), basal insulin dose (U kg-1 day-1), body weight (kg), urinary albumin/excretion rate (normo-, micro- and macroalbuminuria), diabetes distress, diabetes treatment satisfaction, diabetes psychosocial self-efficacy, diabetes quality of life and last mild symptomatic hypoglycaemia (<3 mmol/l, events/week) throughout participation.

In addition to the predefined outcomes, blinded CGM mean glucose (mmol/l), time in hyperglycaemia (>13.9 mmol/l), glucose CV, and TIR separately for daytime (06:00 hours to midnight) and night-time (midnight to 06:00 hours) [31] have also been analysed.

Statistical analysis

Changes in outcomes within arms over the intervention period and in comparison with the control arm (effects of the treatments) were modelled by linear mixed-effects models with random intercepts for participants and centre. Treatment arms and visits were included as fixed effects. The baseline level of the outcome was retained as an outcome in the model. The exact times of measurements were used. Outcomes for which the model residuals were not normally distributed were log-transformed prior to analyses. In the analysis of body weight, we adjusted for sex. Analyses were performed as intention to treat including all available data. Per-protocol (defined as ≥5 daily isCGM scans and ≥3 daily ABC insulin administrations) analyses were further performed for TIR, time in hypoglycaemia (<3.9 mmol/l), time in hyperglycaemia (>10 mmol/l) and HbA1c. Statistical significance was inferred at a two-tailed p < 0.05. The original p values are shown. Statistical significance for the secondary outcomes after adjustment for multiple testing (Benjamini–Hochberg method) [32] is also reported. Statistical analyses were conducted in R version 3.6.1 (The R Foundation for Statistical Computing, www.R-project.org) and SAS version 9.4 (SAS Institute, Cary, NC, USA).

Results

Population characteristics

In total, 184 MDI-treated adults with type 1 diabetes attended the screening visit, whereof 14 (8%) did not meet inclusion criteria due to HbA1c ≤ 53 mmol/mol (7.0%) (n = 8), use of glucose-lowering drugs other than insulin (n = 4), severe diabetes complications (n = 1) and conditions judged unsuitable for participation (n = 1).

The 170 eligible people were randomly assigned to usual care (n = 42), ABC (n = 41), isCGM (n = 48) or ABC+isCGM (n = 39). The slightly unbalanced numbers were due to unopened randomisation envelopes. The participants were followed at the diabetes centres: Steno Diabetes Center Copenhagen (55%); Nordsjællands Hospital, Hillerød (19%); Bispebjerg and Frederiksberg Hospital (14%); Amager and Hvidovre Hospital (9%); and Rigshospitalet (4%).

Baseline characteristics were comparable across arms: median (IQR) diabetes duration 18 (10–28) years, HbA1c 65 (61–72) mmol/mol (8.1% [7.7–8.7%]), mean BMI 26.7 (SD 3.9) kg/m2, age 47 (13.7) years and sex (63% male participants), apart from higher prevalence of antihypertensive medication (p = 0.010) and lower prevalence of unaware participants according to the Hillerød method (p = 0.013) in the ABC+isCGM arm (Table 1). The proportion of participants with TIR >70% at baseline was 5%, 3%, 13% and 0% for usual care, ABC, isCGM and isCGM+ABC (p = 0.056 for group difference).

Table 1 Baseline characteristics of the 170 study participants according to allocation: usual care, ABC, isCGM or ABC+isCGM
Full size table

Discontinuation and compliance

The discontinuation rate was 17% in the whole cohort. The rate was higher in the ABC arm (39%) compared with usual care (17%), isCGM (4%) and ABC+isCGM (10%) (p ≤ 0.026). The most frequent reasons for withdrawal from the ABC arm were time demands, workload and technical challenges (ESM Fig. 1). The ABC was used for bolus calculation 4 (3–6) times daily in the ABC arm and 3 (1–5) times daily in the ABC+isCGM arm at study end without group difference (4 [2–5] times/day at study end among all participants allocated ABC). The numbers of daily isCGM scans were 9 (7–12), 8 (6–12) and 5 (1–7) at 2 weeks after initiation, midway and at study end (isCGM and ABC+isCGM arms), without differences between groups, but with reduction over time in both arms (p < 0.001). The 141 completed participants were followed for 26 (24–28) weeks. The proportions of per-protocol users among the completers in the different intervention arms were 84% (ABC), 93% (isCGM) and 66% (isCGM+ABC). In the isCGM+ABC arm, 66% used the ABC and 97% used the isCGM per-protocol. Recording durations for blinded CGM were 13 (12–14) and 14 (12–14) days at baseline and study end.

Time in ranges

There was no significant improvement in the primary endpoint change in TIR from baseline to study end in the isCGM arm vs the usual care arm (% difference of 3.9 [−12–23], p = 0.660) (Fig. 1). The same was true for ABC and ABC+isCGM compared with usual care (p ≥ 0.061). Changes in time in hypo- and hyperglycaemia and in mean glucose were also comparable to usual care in all three intervention arms (p ≥ 0.105). Daytime TIR increased among ABC only participants (% difference of 22 [95% CI 1, 47]) and night-time hyperglycaemia increased among isCGM only participants (% difference of 258 [40, 814]) compared with usual care. In per-protocol analyses, TIR increased (% difference of 25 [2, 54]), but time in hypoglycaemia (<3.9 mmol/l) did not change significantly (p = 0.057) among ABC only users compared with usual care.

Fig. 1
figure 1

Changes in (a) TIR (3.9–10 mmol/l), (b) time in hypoglycaemia (<3.9 mmol/l), (c) time in hyperglycaemia (>10 mmol/l), (d) mean glucose, (e) glucose CV, (f) HbA1c, (g) treatment satisfaction (DTSQc), (h) psychosocial self-efficacy (DES short form) and (i) present quality of life (ADDQoL-19) scores in the three intervention arms: ABC, isCGM and ABC+isCGM, compared with usual care

Full size image

Findings on changes in TIR and time in hypo- (<3.9 mmol/l) and hyperglycaemia (>10 mmol/l) in those allocated ABC+isCGM vs those allocated isCGM only were comparable (p ≥ 0.241). Similarly, the results were comparable in those allocated isCGM only vs ABC only (p ≥ 0.118). The results on TIR and time in hypo- and hyperglycaemia among all participants using isCGM (isCGM or ABC+isCGM) were also comparable to those in the usual care arm (p ≥ 0.294). The proportion of participants with TIR >70% at study end was comparable across arms (9%, 16%, 20% and 17%; p = 0.594).

Other blinded CGM-derived endpoints and HbA1c

Compared with usual care, glycaemic variability (% difference of 11 [1, 20]) and glucose CV (% difference of 11 [4, 17]) decreased with ABC+isCGM, but did not change in the two other intervention arms compared with usual care. Mean glucose did not change in the intervention arms compared with usual care (p ≥ 0.314).

Compared with usual care, HbA1c decreased in the ABC+isCGM arm (4 [1, 8] mmol/mol) (0.4 [0.1, 0.7] %-point), but not among ABC or isCGM only participants (p ≥ 0.184). Similar results were found in the per-protocol analysis. The HbA1c reduction among all participants using isCGM (isCGM or ABC+isCGM) was greater than among those in the usual care arm (3 [0, 6] mmol/mol; 0.3 [0.0, 0.6] %-point; p = 0.049).

Patient-reported outcomes

Compared with usual care, treatment satisfaction (DTSQs, DTSQc) improved in the isCGM (5, 6 units; p < 0.001) and ABC+isCGM (4, 4 units; p < 0.012) arms. Psychosocial self-efficacy (0.3, p = 0.014), ADDQoL present quality of life (0.5, p = 0.018) and DTSQc-perceived frequency of hyperglycaemia (−1.2, p = 0.003) improved among ABC+isCGM vs usual care participants. The total ADDQoL score worsened in the isCGM arm (−0.5, p = 0.017), and, when solely analysing the ‘dietary freedom’ domain, the figure for change in impact on quality of life in the isCGM arm was −1.1 (p = 0.066) vs usual care.

Within-group changes

There were several significant within-group improvements from baseline to study end in glycaemic and patient-reported outcomes (Table 2, Fig. 2, ESM Table 1).

Table 2 Within-group changes in glycaemic and clinical variables according to allocation: usual care, ABC, isCGM or ABC+isCGM
Full size table
Fig. 2
figure 2

Within-group changes from baseline to study end in TIR (3.9–10 mmol/l), time in hypoglycaemia (<3.9 mmol/l) and time in hyperglycaemia (>10 mmol/l and >13.9 mmol/l) among participants allocated usual care, ABC, isCGM and ABC+isCGM. Data are shown for the 138 participants with data both at baseline and at study end. *Statistically significant within-group change (p < 0.05). There were no significant improvements in the intervention arms compared with usual care

Full size image

Adjustment for multiple testing

For the secondary outcomes, statistical significance of the intervention was maintained after adjustment for multiple testing with regard to glucose CV, treatment satisfaction (DTSQs and DTSQc) and DTSQc-perceived frequency of hyperglycaemia among ABC+isCGM participants, and treatment satisfaction (DTSQc) was maintained among isCGM only participants, compared with usual care. For other predefined secondary outcomes and additional analyses, statistical significance was lost when adjusting for multiple testing.

Safety

The change in mild hypoglycaemia episodes from baseline to study end in the intervention arms was comparable to usual care. The numbers of reported mild symptomatic hypoglycaemic weekly events were 2 (1–4) (usual care), 2 (1–4) (ABC), 5 (2.5–8) (isCGM) and 3 (1–7) (ABC+isCGM) (p < 0.001) 2 weeks after intervention initiation. The corresponding values for the whole cohort midway and at study end were 2 (1–3.5) and 3 (1–4), without differences across arms. There were six reported severe hypoglycaemic events (one event occurred before initiation of intervention) in five participants allocated usual care (n = 1), ABC (n = 2) and isCGM (n = 2). All participants fully recovered. There were no cases of ketoacidosis.

Discussion

In this large, randomised, controlled, multicentre trial, we approached MDI-treated people with type 1 diabetes, with suboptimal glycaemic control and naive to insulin pumps, CGM and advanced carbohydrate counting. Three different treatment approaches were evaluated compared with usual care: ABC, isCGM and the two combined. None of the treatment approaches led to a significant increase in TIR compared with usual care as measured by blinded CGM for 2 weeks at baseline and study end. Combined intervention with ABC+isCGM improved glucose CV and treatment satisfaction compared with usual care and resulted in a clinically relevant HbA1c reduction over time.

The main strength of our study is the robust design with large numbers, as well as the multicentre approach. Multicomponent interventions are probably pivotal in striving for glycaemic control, especially among those far from target. To our knowledge, the effect of a possible additive effect of ABC on isCGM has not yet been examined by others. However, the four-arm set-up with single and combined interventions may lead to interpretation difficulties. The main limitations are the recruitment difficulties and the higher than anticipated discontinuation rate in the ABC arm, resulting in a slightly underpowered arm. Other important weaknesses are the lack of TIR or HbA1c stratification at inclusion, resulting in somewhat unbalanced groups. Further, blinded CGM sampling for 2 weeks two times probably does not truly reflect blood glucose levels during the entire participation period. Whether participants’ behaviour was affected by blinded CGM wear, as indicated by a somewhat lower mean glucose than reflected by HbA1c, remains speculative.

We could not demonstrate any improvement in TIR by using isCGM vs usual care. To our knowledge, there are no other RCTs that have obtained blinded CGM data in their evaluation of isCGM efficiency. In our study, HbA1c was reduced by 3 mmol/mol (0.3%-point) among isCGM users. Recently, nationwide register-based studies from Sweden [33, 34], the Netherlands [19], Belgium [35] and the UK [18] including 1050–14,450 people with type 1 diabetes and baseline HbA1c of 62–68 mmol/mol (7.8–8.4%) were published. HbA1c reductions were found 8–24 months after isCGM initiation in four studies; however, the actual figures were rather modest: 1 [34], 3 [33], 5 [18] and 3 [19] mmol/mol (0.1, 0.3, 0.5 and 0.3%-point), and in the fifth study HbA1c was maintained [35]. Since we had provided an educational programme with structured group courses and close follow-up by an experienced team, we had probably expected greater HbA1c reduction by using isCGM. Our study population is characterised by people not obtaining the recommended HbA1c target but not necessarily with pronounced dysregulation, with whom one expects greater HbA1c reductions. In similar populations, greater effects on glycaemic variables have been found with real-time CGM with alarms [2, 3]. In our population, the HbA1c reduction in all isCGM users was greater than in the usual care arm, but then the effect of ABC was not taken into consideration, and the result lost its significance after multiple testing adjustments. Also, the recommended number and the actual figures of daily scans in the isCGM arms were lower than recent studies have demonstrated are needed to obtain target HbA1c [18, 21, 36]. Our finding of a reduced scanning frequency over time highlights the importance of adequate support to persistent compliance.

Existing data on severe hypoglycaemia and hospitalisations after initiation of isCGM are somewhat inconsistent, since there are reports on reductions [34, 35] but also a fourfold increase in hypoglycaemia admissions [16]. A newly published RCT in youth with poor glycaemic control reports increased glucose testing and treatment satisfaction with isCGM compared with BGM, but no translation to glycaemic improvement [37]. Others report moderate increases in mild hypoglycaemic events [14, 15], but fewer diabetic [2, 3] and ketoacidosis admissions [14], improved treatment satisfaction [17, 35, 38] and hypoglycaemia awareness [18], and less work absenteeism [19, 35] and diabetes distress [18] after initiating isCGM.

In our trial, only very few withdrew after isCGM allocation, and treatment satisfaction was markedly improved in the isCGM arm compared with usual care. Therefore, the initial sign of increased burden of diabetes on quality of life in the isCGM arm was surprising, but also lost its significance after multiple testing adjustments. One could speculate on whether the result may reflect a lack of tools to act on glucose data provided by isCGM. It is encouraging, however, that treatment satisfaction, psychosocial self-efficacy and present quality of life were improved in the combined intervention arm compared with usual care, although only treatment satisfaction maintained its significance after multiple testing adjustments. These findings might reflect the convenience of easy access to glucose values simultaneously with carbohydrate counting skills and a device for insulin bolus advice.

Our finding of time in hypoglycaemia of 7–11% at both baseline and study end is higher than recommended by consensus [25]. This is likely to be explained by the relatively limited sensor accuracy in the lower range [39]. This may also explain the relatively high symptomatic hypoglycaemia figures shortly after initiation of isCGM, as the open isCGM consists of the same sensor; we cannot exclude, however, that isCGM reveals otherwise unidentified hypoglycaemia.

Our clinical impression is that far too few in the Danish MDI-treated type 1 diabetes population have been trained in or perform daily bolus calculation based on carbohydrate ratios and insulin sensitivity, even though nutritional therapy has an integral role in diabetes management [40]. We found discontinuation rates lower than expected except among those allocated to ABC only. The workload of carbohydrate counting and the time and technical skills to seek bolus advice before administering rapid-acting insulin several times a day probably influence compliance. On the other hand, those continuing participation in the ABC arm reported improved treatment satisfaction and HbA1c was reduced. Similarly, among those allocated to ABC+isCGM, every third participant discontinued or reduced the use of ABC. Whether the workload of ABC+isCGM could be minimised with integrated tools remains speculative. Furthermore, the chosen ABC device did not hold features allowing bolus adjustments in relation to physical exercise, stress or illness, which are likely to be useful.

The aim of including ABC in this study was to evaluate its effect on top of isCGM and to compare ABC with isCGM. The group course and follow-up content, and how to incorporate isCGM data when adjusting the ABC settings, were developed for this specific study, based on manufacturer guidelines and in line with contemporary publications [41,42,43]. Whereas isCGM enables glucose testing and information on glucose trends and otherwise unrevealed glucose levels, ABC primarily entails focus on postprandial glucose values and may therefore be viewed as a factual insulin intervention. Introducing isCGM typically discloses low glucose values or declining glucose trends during sleep, which may explain the reduction of basal insulin dose and the increase in night-time hyperglycaemia in the isCGM arm, although these findings lost significance after multiple testing adjustments. It also seems as if ABC, at least among those that have the resources to perform it properly over time, has an overall, but especially daytime, glucose stabilising effect compared with usual care. We have not specifically investigated postprandial glucose values across arms in this study.

The study population is likely to be recognised by diabetes healthcare professionals worldwide, as people with type 1 diabetes not practising advanced carbohydrate counting and without access to isCGM/CGM are still common, due to limited resources either among users or within the healthcare system.

The key message is that although no improvement in TIR by isCGM alone was found, treatment satisfaction was markedly improved compared with usual care, and HbA1c was reduced by 3 mmol/mol (0.3%-point) over time, in people with type 1 diabetes with suboptimal HbA1c and naive to CGM and advanced carbohydrate counting. Combined intervention with ABC+isCGM improved glucose CV and treatment satisfaction compared with usual care and HbA1c was reduced by 5 mmol/mol (0.5%-point) over time. Discontinuation was most common among ABC only users. The two different interventions are shown to act on different aspects of blood glucose levels. The combined intervention appears advantageous, even though a considerable number can be expected to show resistance towards ABC.

Individualised preferences are likely to further improve the detected effects, but we still ought to consider other actions to obtain target glucose levels. Some may benefit from motivational counselling whereas others may profit from using insulin pumps, CGMs with alarms, automated devices such as hybrid closed-loop systems or new insulins. Future studies on alternative initiatives for people with long-lasting and MDI-treated type 1 diabetes, without reaching target glucose levels, are warranted.