Supplementary Methods Chronic kidney disease stage 4 or 5 Current diabetic foot with ulcer, soft tissue infection or necrosis Unstable angina, myocardial infarction, or cerebrovascular accident over the preceding 6 months Vascular intervention involving the coronary or peripheral arteries over the preceding 6 months Type 1 diabetes mellitus The use of an insulin pump The use of medication that could affect plasma glucose levels* Uncontrolled hypertension (SBP 160 mmhg or DBP 100 mmhg) Diabetic ketoacidosis or hyperglycemic hyperosmolar syndrome over the preceding 6 months Aspartate aminotransferase or alanine aminotransferase levels X2.5 of the upper normal limit Pregnancy or lactation Disagreement to continue to practice appropriate contraception during the study Alcohol abuse or drug addiction Moderate to severe hearing or visual impairment Inability to continue using the smartphone application throughout the study period Participation in other clinical trials over the previous 3 months
A. Exclusion criteria *The medication that could affect plasma glucose levels was defined as follows: systemic corticosteroids (more than 30 mg of prednisolone or its equivalent) for more than 2 weeks; immunosuppressants that could affect plasma glucose levels, such as cyclosporine, tacrolimus, and sirolimus; nicotinic acid 1500 mg/day; systemic isotretinoin; medications for the treatment of obesity, such as orlistat, phentermine, phendimetrazine, and mazindol. SBP, systolic blood pressure; DBP, diastolic blood pressure B. Statistical analysis The efficacy of the mdiabetes was analyzed in the full analysis set, defined as a set of randomized patients meeting the inclusion and exclusion criteria with at least one follow-up value of HbA1c after the start of the study. The efficacy was also analyzed in the per protocol set. The safety of the program was analyzed in the safety analysis set, defined as a set of participants whose safety data was collected at least once after randomization. The data with missing values was analyzed using the last-observation carried forward method, except for the 7-point self-monitoring of blood glucose, which was analyzed in the set of participants who completed the measurement of glucose at all time points. The results are presented as the mean standard deviation (SD) for continuous variables, following a normal distribution, or the median and range for continuous variables not following a normal distribution. Categorical variables are expressed as frequency and proportion. The values of changes are presented as the least squares (LS) mean standard error (SE). To compare values of continuous variables at the same time point between the mdiabetes group and the plogbook group, the two-sample t-test was used for continuous variables. The Wilcoxon rank sum test was used for the parameters that did not exhibit a normal distribution. Changes at 24 weeks were compared to the baseline values with the paired t-test. To compare the values of categorical variables at the same time point between the mdiabetes group and the plogbook group, Pearson s chi-square test was used. The endpoint analysis with continuous variables was performed using the analysis of covariance (ANCOVA). The mean changes from baseline in each group were adjusted for baseline value, type of treatment (i.e., groups A to D), and site of study. For the endpoints with categorical variables, we used the Cochran-Mantel-Haenszel (CMH) test. The changes from baseline were adjusted for the type of treatment and the site of study. Statistical significance was assumed when p < 0.05. Statistical analyses were performed using SPSS version 21.0 (IBM Co., Armonk, NY, USA) and Prism 6.0 (GraphPad, San Diego, CA, USA).
Supplementary Figure 1. The scheme of mdiabetes system. The blood glucose level measured by a Bluetooth glucometer (MyHealthPoint TM, Infopia Co., Ltd., Anyang, Korea) was automatically transmitted to the mdiabetes application. Alternatively, the patient could manually record the measured blood glucose level on the mdiabetes application. Step counts were measured by an activity tracker (LG LifeGram LA11M-BS, LG Electronics, Seoul, Korea). The participants were instructed to enter the data manually with regard to food intake and exercise except walking. Upon entering the data, appropriate feedback messages were automatically sent to the participants. The website for the study staff included all data recorded through the mdiabetes system as well as vital sign, anthropometric/ laboratory data, and prescribed medications. When unacceptable hypoglycaemia or hyperglycemia happened, alarm messages were automatically sent to the study staff to manage the urgent problem. SPP, serial port profile; SMS, short message service
Supplementary Table 1. Criteria for rescue therapy Study period Fasting glucose (mmol/l) Frequency Week 1 week 6 15.0 3 consecutive days Week 7 week 12 13.3 3 consecutive days Week 13 week 24 11.1 3 consecutive days Anytime 17.8 any occurrence Rescue therapy was considered when self-checked pre-breakfast fasting glucose levels over 3 consecutive days were 15.0 mmol/l during weeks 0 to 6, 13.3 mmol/l during weeks 7 to 12, 11.1 mmol/l during weeks 13 to 24, or 17.8 mmol/l over the entire study period.
Supplementary Table 2. The insulin dose adjustment algorithm A. For basal insulin Intensive Less Intensive Least Intensive Glucose* (mmol/l) Lower TDD Higher Lower TDD Higher Lower TDD Higher TDD TDD TDD <3.3 2 4 3 6 4 8 3.4 4.4 1 2 2 4 3 6 4.5 5.5 No change 1 2 2 4 5.6 6.6 1 2 No change 1 2 6.7 7.7 2 4 1 2 No change 7.8 10.0 3 6 2 4 1 2 10.0 4 8 3 6 2 4 *Fasting glucose level B. For premixed insulin a. Adjustment of pre-dinner insulin dose based on the fasting glucose level in the morning Intensive Less Intensive Least Intensive Glucose* Lower TDD Higher Lower TDD Higher Lower TDD Higher (mmol/l) TDD TDD TDD <3.3 2 4 3 6 4 8 3.4 4.4 1 2 2 4 3 6 4.5 5.5 No change 1 2 2 4 5.6 6.6 1 2 No change 1 2 6.7 7.7 2 4 1 2 No change 7.8 10.0 3 6 2 4 1 2 10.0 4 8 3 6 2 4 * Fasting glucose level
b. Adjustment of pre-breakfast insulin dose based on the pre-dinner glucose level Intensive Less Intensive Least Intensive Glucose* Lower Higher Lower Higher Lower Higher (mmol/l) TDD TDD TDD TDD TDD TDD <3.3 3 6 4 8 5 10 3.4 4.4 2 4 3 6 4 8 4.5 5.5 1 2 2 4 3 6 5.6 6.6 No change 1 2 2 4 6.7 7.7 1 2 No change 1 2 7.8 10.0 2 4 1 2 No change 10.0 3 6 2 4 1 2 * Pre-dinner glucose level Insulin users were divided into two groups according to the total daily dose (TDD): lower TDD (less than 20 units) and higher TDD (20 units or higher). There were 3 categories of the intensity of control: intensive, less intensive, and least intensive. Initial intensity of glycemic control was set as less intensive. The insulin dose was determined by the median glucose level over the previous 3 days and the intensity category. The maximum allowed dose change was set at 10 units from baseline. If it was necessary to increase more than 10 units, the insulin titration specialist set a new baseline insulin dose after reviewing the condition of the patient. If unacceptable hypoglycemia and/or hyperglycemia persisted despite adjustment of treatment intensity, the patient would discontinue the study participation.
Supplementary Table 3. Examples of adjustment of the insulin treatment regimen Indication Adjustments No. of adjustments/no. of indicated events The glucose level is above the Intensification target range even after the dose of insulin was increased automatically 3 consecutive times based on the algorithm. Hypoglycemia (< 3.9 mmol/l) occurred twice in previous 7 days The insulin dose needs to be increased more than 10 units from baseline (the maximum increment that was allowed) according to the algorithm. (i.e. from least intensive to less intensive control, or from less intensive to intensive control) Attenuation (i.e., from intensive to less intensive control, or from less intensive to least intensive control) Setting a new baseline insulin dose 0/0 3/6 3/16 When the patient was indicated for insulin dose adjustment by the insulin titration specialist, an alarm message was generated to the medical staff. The research nurse gathered detailed information of the patient status, and it was discussed with the pre-assigned insulin titration specialist (E.K.K.). The insulin titration specialist reviewed the glucose records and made decisions either to approve the change of the treatment intensity category or set a new baseline insulin dose. The changes were made remotely through the website that was exclusively provided to the insulin titration specialist. The actual adjustments were made about 25% of all indicated events (6 cases). This was because most cases (11 cases) were indicated due to temporary problems (for example, omitting exercise or overeating for several days), which was improved at the time of considering adjustments, or nonadherence to the recommendations on insulin dose adjustment (4 cases). The remaining 1 case entered an incorrect dose of injected insulin. Instead of making adjustments, we educated the patients to properly use the mdiabetes by phone call.
Supplementary Figure 2. Modules in the mdiabetes system. Blood glucose levels measured by the Bluetooth glucometer (MyHealthPoint TM, Infopia Co., Ltd., Anyang, Korea) were automatically recorded in the mdiabetes application. When the glucose level was entered, an immediate feedback message was generated according to the glycemic control algorithm (A). Participants were encouraged to record daily food intake by using a detailed food database. A daily feedback message was produced according to the total daily calorie intake and proportion of major nutrients. The goal for daily calorie intake was predefined based on sex, BMI, and physical activity. For participants with BMI < 23 kg/m 2, feedback messages were focused on maintaining a healthy diet. Weight reduction focused feedback was provided for patients with BMI 23 kg/m 2. Participants could also set specific goals in improving their dietary habit, such as eating regularly, limiting alcohol, limiting sodium, and limiting refined grains and added sugar. These goals were posted on the main screen of the diet module (B). Energy expenditure was estimated using step counts measured by the activity tracker (LG LifeGram LA11M-BS, LG Electronics, Seoul, Korea) and manually entered exercise data. The number of steps was transmitted to the smartphone every hour, and the daily cumulative number was displayed on the screen. Similar to the diet module, participants could also enter their own exercise goals on the main screen of the physical activity module. The physical activity module provided an individualized exercise program based on age, BMI, and exercise capacity, which was assessed by a questionnaire at baseline. Daily physical activity targets increased gradually. It started from increasing daily activity, and there was a gradual increase in recommended duration and intensity of the scheduled exercise. For overweight patients with BMI 23 kg/m 2, an exercise program for weight reduction was provided (C). The mdiabetes contained a bulletin board, allowing users to communicate with each other. Participants shared experiences from their daily life and useful information. When the study participants posted medical questions, the research staff provided relevant answers (D).
A. Glucose monitoring module B. Diet module
C. Physical activity module D. Social networking
Supplementary Figure 3. Study population
Supplementary Table 4. Numbers of blood glucose measurement mdiabetes plogbook p value No. of fasting glucose measurement during the study Overall 140.3 33.0 149.7 23.2 0.024 Group A C 140.1 34.7 148.0 25.4 0.130 Group D 141.1 25.8 156.1 10.0 0.032 No. of total glucose measurement during the study Overall 180.3 67.3 167.6 51.8 0.151 Group A C 166.9 54.8 154.4 34.0 0.109 Group D 234.1 85.8 219.0 74.3 0.587
G lu c o s e (m g /d L ) G lu c o s e (m g /d L ) Supplementary Figure 4. Changes in 7-point self-monitoring of blood glucose after 24 weeks. A: There was no difference in baseline values between the two groups. B: After 24 weeks of intervention, the mdiabetes group had lower glucose levels compared to the plogbook group at pre-breakfast, pre-lunch, and post-dinner time. The data were analyzed by ANCOVA. B, breakfast; L, lunch; D, dinner. * p < 0.05, ** p < 0.01. A 2 6 0 2 4 0 2 2 0 2 0 0 1 8 0 1 6 0 1 4 0 1 2 0 m D ia b e te s, W 0 p L o g b o o k, W 0 p re -B p o s t-b p re -L p o s t-l p re -D p o s t-d p re -b e d tim e B 2 6 0 2 4 0 * * 2 2 0 2 0 0 1 8 0 1 6 0 * * * 1 4 0 1 2 0 m D ia b e te s, W 2 4 p L o g b o o k, W 2 4 p re -B p o s t-b p re -L p o s t-l p re -D p o s t-d p re -b e d tim e
Supplementary Table 5. Changes in blood pressure, body composition, fasting plasma glucose, and lipid profile
Values at baseline and week 24 are presented as the mean SD, and a two-sample t-test was used for the comparison of values between the groups. Value of change (baseline W24) are presented as the LS mean SE. The mean changes from baseline in each group were adjusted for baseline value, type of treatment (i.e., groups A to D), and the site of study. ANCOVA was used for the comparison of values between the groups. Supplementary Figure 5. Changes in the Summary of Diabetes Self-Care Activities (SDSCA) and WHOQOL-BREF scores after 24 weeks. A and B: In both the mdiabetes and plogbook groups, scores of all SDSCA domains did not change during the study. C and D: Among WHOQOL-BREF components, the quality of overall health in the mdiabetes group improved after 24 weeks compared to the baseline. The data was analyzed by the paired t-test. *** p < 0.001.
Supplementary Table 6. Changes in insulin doses and HbA1c of the insulin users mdiabetes (n = 18) plogbook (n = 17) Mean SD Median (Range) Total daily dose of insulin (units) 30.1 13.9 36.9 21.9 W0 26.5 (15.0-69.0) 28.0 (14.0-90.0) 31.8 13.6 37.1 22.0 W12 27.0 (18.0-69.0) 28.0 (14.0-90.0) 31.9 14.5 36.6 21.5 W24 31.0 (11.0-69.0) 28.0 (14.0-90.0) 1.7 2.9 0.2 2.4 W12 0 0.5 ( 2.0-9.0) 0.0 ( 7.0-6.0) 1.8 4.5 0.3 4.2 W24 0 2.5 ( 7.0-12.0) 0.0 ( 10.0-10.0) HbA1c (%) W0 W12 W24 W24 0 8.2 0.7 8.1 0.6 8.1 (7.2-9.5) 8.0 (7.3-9.9) 7.7 0.8 7.9 0.8 7.5 (6.6-9.1) 7.8 (6.9-10.5) 7.5 0.6 8.0 1.1 7.4 (6.7-8.7) 7.6 (6.7-11.7) 0.7 0.7 0.1 0.6 0.6 ( 2.4-0.2) 0.4 ( 1.0-1.8) p value 0.618 0.916 0.800 0.114 0.073 0.458 0.631 0.128 0.017
C h a n g e in H b A 1 c C h a n g e in H b A 1 c C h a n g e in H b A 1 c Supplementary Figure 6. Relationships between numbers of data input and the change in HbA1c in the mdiabetes group. A: The number of glucose measurement was negatively correlated with the change in HbA1c level (r = 0.265, p = 0.011). B and C: The number of food data input and daily step counts measured by the activity tracker had no correlation with the change in HbA1c. A N o. o f g lu c o s e m e a s u re m e n t B N o. o f fo o d d a ta in p u t 0 1 0 0 2 0 0 3 0 0 4 0 0 5 0 0 4 0 2 0 0 4 0 0 6 0 0 8 0 0 4 3 3 2 2 1 1 0 0-1 -1-2 -2-3 -4 r = -0.2 6 5 p = 0.0 1 1-3 -4 C N o. o f d a ily s te p s 0 5 0 0 0 1 0 0 0 0 1 5 0 0 0 2 0 0 0 0 4 3 2 1 0-1 -2-3 -4