Study Subjects and Design

The study aimed to enroll a total of 111 subjects based on sample size calculation. Recruitment of participants took place from July 2020 to November 2021. With the last patient out in February 2022. Subjects were selected from the investigator's established patient population based on specific inclusion and exclusion criteria. Eligible subjects had to be at least 21 years old with type 2 diabetes, planning to start basal insulin titration. They were required to have an HbA1c level of ≥ 7.5% (58.5 mmol/mol) within the past 3 months and a FBG target range set to 140 mg/dl (7.8 mmol/l) or lower as determined by the investigator. An HbA1c > 7.5% in the last 3 months was included to target those individuals in need of basal therapy titration as a high HbA1c is one of the main reasons to initiate basal insulin and a target group for the SMS service. Additionally, they needed to possess and use a mobile phone with SMS capabilities and access to the mobile phone network while at home (Fig. 1).

Fig. 1

Study overview. R on-site screening; V1 baseline visit with the handout of study supplies, start and training of insulin therapy and SMS service, HbA1c testing, and questionnaire administration. V2 Clarification of any questions and documenting any adverse events or device deficiencies, if any. V3 documentation of FBG stable in range, V4 HbA1c testing, data download, questionnaires, and documentation and documenting any adverse events or device deficiencies, V5 Documenting any adverse events or device deficiencies, if any. BOT basal oral-supported therapy. SMS Short Messaging Service; ICF informed consent form; FBG fasting blood glucose

Exclusion criteria included prior insulin therapy (except for gestational diabetes or < 1 week), current insulin therapy (e.g., prandial insulin, premixed insulin), impaired awareness of hypoglycemia with a history of regular hypoglycemia or recent hospitalization due to severe hypoglycemia within the previous 3 months, severe diabetes-related long-term complications (e.g., severe retinopathy, neuropathy, nephropathy requiring dialysis), pregnancy or plans to become pregnant, legal incompetence or limited legal competence, serious or unstable chronic physical or psychological conditions rendering subjects unable to understand the study's nature and scope or follow study procedures, and addiction to alcohol or other substances of abuse.

Subjects who met the inclusion criteria and did not meet any of the exclusion criteria were planned to be enrolled consecutively until the number of 111 subjects. Prior to study-related procedures, subjects signed the valid subject informed consent form approved by the Independent Ethics Committee.

Treatment

The study subject measured their FBG on 4 days after the baseline visit 1 without starting to inject insulin, thereby generating the mean baseline FBG (Fig. 1). The SMS service started on the fifth day after visit 1 requiring the subject to perform daily FBG measurements in the morning and injections of long-acting (basal) insulin in the evening. The subjects were asked by an SMS in the morning to enter their FBG value measured and by a SMS in the evening to enter their insulin dose injected into the responding SMS. This interaction between the subject and the SMS service was done until the FBG was stable in the target range according to the SMS service. The insulin titration scheme for this was preset based on the respective insulin manufacturers recommendation and could be changed at the investigator’s discretion. Once the optimal basal insulin dose had been found, the subject received a completion SMS. The subject continued to inject the final basal insulin dose every evening without performing regular FBG measurements in the morning. The insulin dose was not changed anymore. The SMS service stopped after the completion SMS. Regardless of whether the SMS service had been completed, visit 4 occurred 16 weeks ± 14 days after the baseline visit (visit 1). The subjects were reminded prior to visit 4 to measure their FBG on the 4 days before visit 4 in order to generate a mean FBG for this time point. These FBG values were not entered into the SMS service, but were automatically stored in the BG meter, which was downloaded by the HCP at visit 4. Over the entire study, subjects were allowed to perform BG measurements as recommended by their HCP using the BG meter supplied for the study. This specifically could include BG measurements to verify symptomatic hypoglycemia independent of the time of the day.

Outcomes

The primary objective of this study was to determine the percentage of subjects who achieved stable FBG within their individual target range after completing SMS-supported basal insulin titration by the latest at visit 4, which occurred at week 16 (± 14 days). The secondary objectives of the study included assessing various factors related to the treatment, such as the number of days until the FBG target range was initially reached, changes in HbA1c levels at visit 4 compared to baseline, FBG levels at visit 4 compared to baseline, total daily basal insulin dose at visit 4, the number of hypoglycemic events. Other secondary objectives were related to adherence to the SMS service requests, including response rate and response time to the service's SMS and changes in the following questionnaire scores: diabetes distress scale (DDS), Hospital Anxiety and Depression Scale (HADS), the Diabetes Medication System Rating Questionnaire (DMSRQ) and the 12-item Short-Form Health Survey (SF-12). Additionally, satisfaction regarding the use of the SMS service was assessed in a custom-made questionnaire. Finally, adverse events were recorded. The following adverse events were of special interest: symptomatic hypoglycemia episodes, hyperglycemia in combination with medical intervention, hypoglycemia in combination with medical intervention, ketosis, and ketoacidosis.

Statistical analysis

The analysis of the primary variable was based on a linear mixed probability model (LMM) for estimating the proportion of subjects in the target range with the study site as a random effect. The LMM is preferred over the generalized linear mixed model with logit link (GLMM) because of its greater numerical stability and direct interpretability. Therefore, the GLMM is reported as a sensitivity analysis only. The lower bound of the Wald-type one-sided 95% confidence interval (CI) for the model intercept, which measures the proportion of subjects in the target, was estimated.

All secondary endpoints were analyzed in an exploratory manner. The types of descriptive statistics used in this study are outlined in the following:

Type Categorical: Absolute (N) and relative frequencies (in %) per category together with 95% Wilson score confidence intervals. Where appropriate and present, the number of missing values as a “Missing” category was added.

Type Continuous: Number of observations (N), the number of missing values (missing), mean (mean), standard deviation (SD), median (median), quartiles (Q1/Q3), minimum and maximum (Min/Max) and a 95% confidence interval for the mean from t-distribution with N − 1 degrees of freedom (where reasonable).

Type Count: Total number of events (count), subject years under risk (subject years), incidence rate per subject year (rate), and a 95% confidence interval from Poisson distribution.

Riddle et al. [23] reported for their randomized controlled trial that 36.2% of the patients in the glargine group and 34.4% in the NPH insulin group (Neutral Protamine Hagedorn insulin) reached FBG levels ≤ 100 mg/dl (5.6 mmol/l) at the end of the study period (24 weeks). These proportions are in line with other trials, such as the randomized controlled trial of Hu et al. [24] who showed that 45.6% from their control group achieved FBG ≤ 7 mmol/l (≤ 126 mg/dl) after 12 weeks. However, this work was only reported in an abstract so that potential methodical issues cannot be analyzed. We therefore assume that 40% of patients might reach their individual FBG target range.

Sample size calculation is performed using the arcsine transformation, also termed angular transformation, which is A(p) = 2 arcsin arcsin √p, with A(p) being measured in radians. It is assumed that 15% of the patients drop out during follow-up. For the sake of comparison, sample size was also calculated using the standard approximation of the binomial to the normal distribution.

The following assumptions are used as basis for sample size calculation:

Usual care proportion of patients reaching the target level at the end of the study period: p0 = 0.4 and the SMS service p = 0.55.

Significance level: α = 0.05 one-sided.

Power: 1 − β = 0.9

Dropout: 0.15.

With these assumptions, a total of 111 patients are required for this study. The study was conducted in accordance with the Declaration of Helsinki, and approved by the Ethics Committee of Landesärztekammer Baden-Württemberg (protocol code DC000058, EUDAMED-Nr.: CIV-20-01-031566) on 20-Apr-2020. Informed consent was obtained from all subjects involved in the study.