This study was an exploratory, 12-month, randomized, open-label, parallel-group, single-center randomized controlled trial (RCT) in a tertiary care hospital in Barcelona, Spain.

The study protocol (Clinical Trials identifier NCT03819335), consent forms, and patient information sheets were approved by the Institutional Review Board of Hospital de la Santa Creu i Sant Pau (IIBSP-SUG-2018-01). All participants provided written consent before the study procedures were started.

The inclusion criteria were as follows: adults (≥ 18 years), T1DM with duration > 1 year, basal-bolus insulin regimen with multiple daily injections (MDI), last (< 3 months) HbA1c ≥ 53 mmol/mol (7%) and < 75 mmol/mol (9%), knowledgeable about carbohydrate counting and functional insulin treatment, and with regular use of smartphone or tablet (android/iOS).

The exclusion criteria were as follows: use of an app for diabetes management at study entry, utilization of real-time or intermittent continuous glucose monitoring (CGM), being pregnant or planning pregnancy, and having any disease or clinical condition that might interfere with the study protocol (e.g., active cancer, severe mental disorder, or planning for surgery).

Participants were recruited at routine follow-up appointments in the Endocrinology Department. If they were eligible, after reading an information sheet and signing the informed consent form, the participants were randomized to the intervention group (IG) or the control group (CG), ratio 1:1, by one of the investigators (GC) using a sequence generated with https://www.sealedenvelope.com/ by a second investigator (RC). Allocation concealment was ensured.

The CG regularly used their glucometer Accu-Chek® Aviva Expert bolus calculator (which provides bolus advice according to glycemia and carb intake) and were scheduled for five face-to-face visits with an endocrinologist (at baseline, 3, 6, 9, and 12 months). The IG were helped to download and install the mySugr® app, received a compatible meter (which automatically uploads blood glucose values to the app via Bluetooth), and were instructed in their use. They were scheduled for five visits, three face-to-face (at baseline, 6, and 12 months) and two teleconsultations (at 3 and 9 months) using the mySugr® Care web platform.

mySugr® is an app for diabetes management from Roche Diabetes Care® and has Free and Pro versions. Their features include a diabetes diary with automatic collection and analysis of data on glycemia (including estimation of HbA1c) and the possibility to manually enter information on food intake, physical activity, and insulin dose (Fig. 1). It is compatible with CGM systems and includes integration with Google Fit type motion sensors (to automatically collect physical activity data), a bolus calculator (activated for IG), reminders about BGM, and the possibility to save pictures of food consumed in the Pro version. The Pro version in the Spanish language was provided to the participants in the study.

We followed the Consolidated Standards of Reporting Trials (CONSORT) reporting guidelines (Supplementary Table 1).

mySugr screenshots of diabetes diary and bolus calculator © 2021 mySugr GmbH

At baseline and at the 12-month visit, patients were given the following questionnaires which they filled in: the Diabetes Empowerment Scale-Short Form Spanish version (DES-SF-S), diabetes-related tasks, the Diabetes Quality of Life Spanish version (EsDQOL), and the Diabetes Distress Scale Spanish version (DSS-S).

The DES-SF-S was used to measure psychosocial self-efficacy (Supplementary Table 2). The original DES has 28 items divided into three subscales, as follows: managing the psychosocial aspects of diabetes, assessing dissatisfaction and readiness to change, and setting and achieving goals [24]. The reduced version has free access and includes eight items, assessed through an ordinal scale of five answers [25]. The result is obtained after the sum of all items in which higher values are related to higher perceptions of psychosocial self-efficacy.

The diabetes-related tasks questionnaire was developed ad hoc to assess tasks related to and time requirement for diabetes self-management (Supplementary Table 3). It has five items rated in an ordinal scale from 1 to 4. A higher score means that the patient has the appropriate tools and needs less time and fewer tasks for diabetes self-management.

The EsDQOL questionnaire has 46 items which measure four domains that are highly relevant to treatment perceptions, as follows: satisfaction with treatment, impact of treatment, worry about social/vocational issues, and worry about the future effects of diabetes (Supplementary Table 4) [26]. All items are scored on a 5-point scale where a lower value means a better QOL. The Spanish version has been validated and has free access [27].

The DSS-S was used to assess diabetes-related distress (Supplementary Table 5). It is a free-access 17-item instrument with four subscales (emotional burden, physician distress, regimen distress, and interpersonal distress) where each item is rated on a 6-point scale. An average score of < 2.0 reflects little or no distress, between 2.0 and 2.9 moderate distress, and ≥ 3.0 high distress. A total or subscale score ≥ 2.0 is considered clinically significant [28].

Data regarding age, sex, highest level of education, employment status, body mass index, systolic and diastolic blood pressure, heart rate, diabetes duration, insulin therapy, and presence of diabetic complications were registered at baseline with information from electronic health records and via physical examination. HbA1c was measured with a point-of-care method (DCA, Siemens DCA 2000+) at baseline and after 6 and 12 months.

At baseline and at all follow-up visits, the frequency of BGM, mean blood glucose (BG) and standard deviation (SD), coefficient of variation (CV), and high and low blood glucose index (HBGI and LBGI) were obtained from the glucometer download.

Moreover, patients in the IG filled in an ad hoc satisfaction questionnaire (Supplementary Table 6) and were asked if they still used the app after the end of the study.

Recruitment began in March 2019 and finished in February 2020, and the study ended in May 2021.

The primary outcome was to compare the change from baseline to 12 months in empowerment, assessed after completion of the DES-SF-S questionnaire, between the IG and the CG.

Secondary endpoints were glucose-related outcomes, namely, change from baseline of mean BG and SD, CV, HBGI, and LBGI at each follow-up visit and change from baseline in HbA1c at 6 and 12 months.

Other secondary endpoints included change from baseline in self-management (measured via the diabetes-related tasks questionnaire), change in QOL (evaluated with EsDQOL), and distress related to diabetes (evaluated with DSS-S), from baseline to endpoint, and change in adherence. Adherence was assessed with the rate of attendance at face-to-face visits and teleconsultations and, at each follow-up visit, the frequency of BGM and uptake of recommendations proposed (percentage of recommendations prescribed by the physician and accepted by the patient at the next visit). IG participants were also asked if they still used the app 3 months after the end of the study.

Finally, for IG participants, satisfaction with the mySugr® app was assessed with a specific questionnaire (ad hoc) at the end of the study.

The study was designed as an exploratory study so that a formal sample size calculation to test the hypothesis that mySugr® app use with teleconsultation led to an improvement of empowerment at 12 months would not be performed. We did not use previous studies assessing empowerment [19] for calculation of sample size because they were performed in a different population and health system. Given that it was an exploratory study, the sample size was set at 15 patients per group. Considering that there could be a dropout rate of 10%, a minimum of 33 patients had to be enrolled.

Categorical data were expressed as frequencies and percentages. The Kolmogorov-Smirnov test was used to examine whether the quantitative data had normal distribution, and the data were expressed as mean and SD or median (P25, P75) accordingly. Differences were tested using the Mann-Whitney U test for non-normal distribution of continuous variables and Student’s T test for normally distributed continuous variables. Statistics were performed using IBM-SPSS version 26.0 (Chicago, IL). The data were analyzed according to the intention-to-treat principle.