The effect of education in obese type 2 diabetes adolescents and young adults based on continuous glucose monitoring and smartwatch-derived lifelog data
Article information
Abstract
Purpose
The increasing prevalence of type 2 diabetes (T2D) among adolescents and young adults (AYAs) is a major public health concern worldwide. This pilot study evaluated the effectiveness of lifestyle education in managing T2D in obese AYAs using continuous glucose monitoring (CGM) and smartwatch-derived lifelog data.
Methods
Seven obese AYAs and T2D aged 12–19 years were enrolled in this prospective interventional study. Patients continued to take their previously prescribed T2D medication. CGM data were collected for 10 days, followed by lifestyle education using CGM and smartwatch data. Outcomes, including anthropometrics, glycemic control, dietary intake, physical activity, and self-management skills, were reassessed after an additional 10 days.
Results
The median time in range increased from 58.1% (53.2%–75%) to 72% (64%–88%) (p=0.043) and time above range (>250 mg/dL) decreased from 10% (2.9%–18.6%) to 3.0% (1.0%–11.0%) (p=0.028). Median peak total caloric intake decreased from 2,854 (2,465–3,040) kcal/day to 2,091 (1,751–2,283) kcal/day and walking calorie expenditure increased from 163.9 (116.7–321.3) kcal/day to 180.2 (165.3–492.4) kcal/day (p=0.018 for both). The Summary of Diabetes Self-Care Activities score improved from 0.29 (0.05–0.43) to 0.33 (0.32–0.68) (p=0.043).
Conclusions
This integrated approach combining CGM and smartwatch-based education exhibited short-term effects on glycemic control, dietary habits, physical activity, and self-management skills in obese AYAs and T2D. Further studies are needed to confirm the long-term effectiveness of this strategy in this challenging population.
Hightlights
· Integrating continuous glucose monitoring and smartwatch-derived lifelog data into intensive lifestyle education proved effective in achieving short-term improvements in glycemic control, dietary habits, physical activity, and self-management among obese adolescents and young adults with type 2 diabetes. These results highlight the potential of this multimodal digital approach for managing type 2 diabetes in this challenging population.
Introduction
The increasing prevalence of type 2 diabetes (T2D) among adolescents and young adults (AYAs), which is closely associated with the increasing rates of childhood obesity [1-4], is a major health challenge globally. Early-onset T2D presents a more severe phenotype than adult-onset T2D, and is characterized by rapid disease progression, higher risk of complications, and poorer long-term outcomes [5]. Adolescence and young adulthood, which is characterized by physiological insulin resistance and lifestyle changes, is a critical period for the development of T2D, necessitating targeted interventions to address the unique needs of this population.
Lifestyle modification remains the cornerstone of T2D treatment in AYAs, along with obesity management encompassing dietary interventions, physical activity promotion, and behavioral therapy [6,7]. However, adherence to these strategies is often suboptimal in this age group owing to social and psychological barriers [8]. Continuous glucose monitoring (CGM) devices have emerged as transformative tools for diabetes management, providing real-time glucose data and encouraging self-management behavior. Although CGM has demonstrated efficacy in improving glycemic control in adults with T2D [9], its application in AYAs remains underexplored [10]. Evidence suggests that CGM can improve the quality of life and support behavioral changes in AYAs with T2D; however, barriers such as cost, stigma, and limited insurance coverage hinder its widespread adoption [11-13].
Digital health technologies, including wearable devices such as smartwatches and mobile applications (apps), have shown promise in promoting lifestyle changes. These tools facilitate real-time tracking of physical activity, dietary habits, and sleep patterns while providing actionable feedback to users [14-16]. Integrating CGM with smartwatch technology offers a novel multimodal approach to address the complex needs of AYAs with T2D. Such an approach not only leverages real-time data for personalized education, but also promotes sustained behavioral changes through interactive and user-friendly platforms.
This short-term pilot study aimed to evaluate the feasibility and effectiveness of combining CGM and smartwatch-derived lifelog data with intensive lifestyle education to manage T2D in obese AYAs. We hypothesized that this integrated intervention would improve glycemic control, dietary habits, physical activity level, and self-management skills. By addressing the unique challenges faced by this vulnerable population, our study aimed to provide insights into innovative strategies for sustainable diabetes care.
Materials and methods
1. Study design and patients
This prospective, interventional, single-group assignment trial was conducted at a single medical center in Seoul, Republic of Korea, from January 11, 2023 to May 31, 2023. The Institutional Review Board (IRB) of our institution reviewed and approved this study (IRB No. 2022-12-052-002), and written informed consent was obtained from the patients and their parents or legal guardians before enrollment. The eligible patients were AYAs aged 12–19 years with obesity (body mass index [BMI] >95th percentile) and T2D for >3 months. The inclusion criteria required agreement to wear the CGM device and smartwatch continuously and to maintain a food diary. The exclusion criteria included acute or chronic diseases other than T2D, obesity, or obesityrelated diseases (hypertension or dyslipidemia). Patients with mental illnesses or who were currently using medications that may affect glucose levels, such as steroids or traditional herbal medicines, prior CGM experience, history of hypoglycemia, or pregnancy, were excluded. Finally, 7 patients participated in the study.
2. Study protocol
Detailed information regarding the study protocol is provided in Supplementary Fig. 1. Patients visited our medical center 5 times, starting with a screening visit (visit 0). On visit 0, informed consent was obtained, followed by basic diabetes education and completion of a baseline questionnaire, the Summary of Diabetes Self-Care Activities (SDSCA-K) [17,18]. On visit 1, physical measurements and body composition analyses were performed, and blood was drawn if needed. Smartwatches and scales were provided to the patients. Training was provided on how to maintain a food diary, exercise, and sleep patterns in the ROTHY app. The ROTHY app automatically collects lifelog data through a smartwatch and manually enters the data. The first CGM was inserted into the patients. After 10 days of baseline data collection (days 1 through 10), 2 lifestyle education sessions based on the baseline data were conducted at visits 2 and 3. The first education session was conducted on days 11–17. On the second day of education on day 18 (visit 3), the second CGM was inserted, and the patients were trained on how to monitor the CGM. Follow-up data collection was performed on days 18–27. At visit 4, physical measurements, body composition analysis, CGM removal, and a questionnaire (SDSCA-K) were completed.
3. Variables
Primary outcomes included changes in glycemic control, dietary caloric intake, physical activity, and sleep patterns. The secondary outcomes were program adherence and self-efficacy. Glycemic control was assessed using the time in range (TIR), time above range (TAR), time below range, CGM average glucose, and coefficient of variation data from the CGM. Dietary intake was quantified based on a user-entered daily food diary for total energy, carbohydrate, protein, and fat consumption, using the ROTHY app and was calculated by a dietitian. Physical activity and sleep patterns were recorded as lifelog data, and sitting time was assessed using the lifelog data.
4. Data sources/measurement
Baseline anthropometric data, including height, weight, BMI z-score, waist circumference (WC), blood pressure (BP), and body composition, were measured. Height was measured using a Harpenden stadiometer (Holtain Ltd., United Kingdom) and weight was measured using a digital scale (150 A; Cas Co., Ltd., Korea). The height, weight, and BMI z-scores were based on the 2017 Korean National growth charts [19]. WC was measured at the umbilicus (cm). Body composition was analyzed using bioelectrical impedance analysis (InBody 770; Biospace Co., Korea). Fat percentage, fat mass, fatfree mass, and skeletal muscle mass were calculated from the body composition measurements. Baseline blood tests were performed to measure the glycated hemoglobin (HbA1c), aspartate aminotransferase, alanine aminotransferase (ALT), total cholesterol, triglyceride, high-density lipoprotein, and low-density lipoprotein levels. Family history of diabetes and related diseases (hypertension, dyslipidemia, and cardiovascular disease) was evaluated.
Data were collected using a CGM device (Dexcom G6; Dexcom, USA), daily food diaries, and smartwatch (Galaxy Watch 4). Physical activity data and sleep patterns collected from the smartwatch were transmitted to the ROTHY app. The patients used the ROTHY app to maintain their food diaries daily. This app displays the calories of each food item and allows users to select items such as meals, snacks, and drinks; upload a photo of their meal before eating; and save it on the ROTHY app. In addition, patients can log the type and amount of food, and time of day they planned to eat, and upload a post-meal photo, all stored on the app. Dietitians evaluated these daily food diaries and analyzed patient caloric intake and nutrient content of foods consumed. Because patient adherence to the food diary was unknown, it was assumed that at least 1 day of each 10-day period was completed. Therefore, the analysis was limited to peak caloric intake.
Two types of lifelogs were analyzed: (1) data automatically collected by the ROTHY app service and (2) data manually entered by the user through an event-logging feature in the ROTHY app. The automatically collected data included activities, steps taken, moving distance, walking calorie expenditure, sleep patterns, sleeping hours, saturation maximum and minimum, and low saturation (<90%) time during sleep. Events such as exercise or sleep could be entered manually. Exerciserelated data included duration, distance, calorie expenditure, speed, and heart rate. The smartwatch active wearing time (hr/day) was calculated using a combination of automatically collected walking time (hr/day) and manually entered exercise and sleep time. Sitting time (hr/day) was not included; it was included when patients wore a watch without movement upon awakening.
5. Statistical analyses
All data were nonparametric and are presented as medians and interquartile ranges. Statistical analyses were performed using IBM SPSS Statistics ver. 29.0 (IBM Co., USA). Baseline and follow-up data were compared using the Wilcoxon rank-sum test. Statistical significance was set at P<0.05.
Results
1. Baseline characteristics
Table 1 shows the baseline characteristics of 7 patients (1 female, 14.3%). The median age at T2D diagnosis was 15.2 years (interquartile range [IQR], 13.1–16.3), with a median BMI z-score of 2.6 (IQR, 2.1–2.7) and HbA1c of 11.3% (IQR, 10.8–13.3). A family history of diabetes was present in all but 1 patient. At enrollment, the median age was 17.4 (IQR, 14.9–18.0) years, with a median T2D duration of 1.3 (IQR, 0.8–2.6) years. The median weight was 91.2 (IQR, 87.4–116.6) kg with BMI z-score of 2.7 (IQR, 1.8–2.8) and WC of 98 (IQR, 92–115) cm. Body composition showed a fat percentage of 36% (IQR, 28.5%–38.2%), fat mass of 38.9 (IQR, 26.4–42.7) kg, fat-free mass of 64.3 (IQR, 56.5–73.2) kg, and skeletal muscle mass of 36.2 (IQR, 31.6–42.2 kg). Systolic BP was 146 (IQR, 134–164) mmHg, and diastolic BP was 88 (IQR, 76–95) mmHg. The median HbA1c was 6.7% (IQR, 6.2%–7.5%) with 2 patients (28.6%) having an HbA1c >7%. Liver function tests and lipid profiles are shown in Table 1. Four patients (57.1%) had an ALT level (IU/L) higher than the reference range of 5–45 (IU/L) [20]. One patient (14.3%) was taking ezetimibe, rosuvastatin, amlodipine, and olmesartan with metformin for high BP and dyslipidemia. Only 1 patient (14.3%) was receiving multiple daily insulin injections with metformin, whereas the other 6 patients were treated with metformin alone.
Baseline characteristics (n=7)
2. Outcome comparison
Fig. 1 shows the changes in BMI z-score and WC for each patient. Supplementary Table 1 shows the detailed median and interquartile values of outcome changes. Although no statistically significant differences were found, median weight, BMI, and BMI z-scores decreased. WC had the same median, from 98 (92–115) cm to 98 (92–113) cm, but the change was significant (P=0.046). Body composition rarely changed without significance.
(A) BMI z -score and (B) waist circumference changes from baseline to follow-up. Values are presented as median (interquartile range, IQR). BMI, body mass index. *P<0.05.
1) Primary outcomes: glycemic control, dietary intake, physical activity, and sleep patterns
Fig. 2 and Supplementary Table 1 present changes in glycemic control. Immediate follow-up of HbA1c was not performed; however, 4 patients had HbA1c levels of 6.8% (6.1%–8.5%) at a median of 67 d after study completion (Supplementary Table 1). The median CGM average glucose decreased from 188 (150–197) mg/dL to 175 (145– 175) mg/dL (P=0.043), TIR increased from 58.1% (53.2%– 75%) to 72% (64%–88%) (P=0.043), and TAR250 decreased from 10% (2.9%–18.6%) to 3.0% (1.0%–11.0%) (P=0.028) (Fig. 2A–C, respectively).
(A) CGM average glucose (mg/dL), (B) TIR (%), (C) TAR250 (%), changes from baseline to follow-up. Values are presented as median (interquartile range, IQR). Patient 2 was on both multiple daily insulin and metformin. CGM, continuous glucose monitoring; TIR, time in range; TAR, time above range; CV, coefficient of variation. *P<0.05.
All 7 patients had a peak total calorie intake over the age requirement [21]. Fig. 3 shows changes in peak total calorie intake, which decreased significantly from 2,854 (2,465–3,040) kcal/day to 2,091 kcal/d (1,751–2,283) (P=0.018). Supplementary Table 1 shows the detailed changes in dietary intake, where median peak carbohydrate decreased from 349 g/day (242–425) g/day to 250 (227–266) g/day (P=0.128), protein from 106 (100–119) g/day to 73 (54–95) g/day (P=0.018), and fat from 108 (89– 128) g/day to 86 (43–92) g/day (P=0.018).
Peak total calorie intake changes from baseline to follow-up. Values are presented as median (interquartile range, IQR). *P<0.05.
Fig. 4 shows changes in lifelog data. Fig. 4A shows the automatically collected step counts, walking distance, and walking calorie expenditure, and Fig. 4B shows manually entered exercise data. All median values in Fig. 4 (4A-1, 4A-2, 4A-3, 4B-1, and 4B-2) showed significant improvements except exercise calorie expenditure (Fig. 4B-3). Walking time increased from 0.74 (0.56–0.98) hr/day to 1.0 (0.91–1.89) hr/day (P=0.018), steps (×103/day) increased from 4.6 (3.1–6.9) to 4.9 (4.0–13.6) (P=0.028), walking distance from 3.3 (2.3–5.4) km/day to 3.6 (3.0– 11.0) km/day (P=0.028), and walking calorie expenditure from 163.9 (116.7–321.3) kcal/day to 180.2 (165.3–492.4) kcal/day (P=0.018). The number of exercise events per day increased from 0.7 (0.0–1.3) to 1.4 (1.1–2.0) (P=0.028), with corresponding increases in median exercise duration (P=0.018), but without significant change in median calorie expenditure (P=0.237). The median sitting time changed from 10.4 (7.4–12.4) hr/day to 11.3 (6.7–11.5) hr/day and sleeping time from 7.0 (6.5–8.6) hr/day to 6.7 (5.9–7.7) hr/day, but these differences were not significant. Supplementary Table 1 provides the detailed results.
Lifelog data changes from baseline to follow-up. (A) Automatically collected data. Values are presented as median (interquartile range, IQR). A1. Step counts (×103/day), A2. Walking distance (km/day), and A3. Walking calorie expenditure (kcal/day). (B) Manually entered exercise data. Values are presented as median (interquartile range). B1. Number of exercise events per day, B2. Total exercise time (min/day), and B3. Calorie expenditure per exercise event. *P<0.05.
2) Secondary outcomes: adherence and self-efficacy
The adherence of all 7 patients is shown as CGM and smartwatch active times in Supplementary Table 1 and Supplementary Fig. 2. Median CGM activity time was >99% at both baseline and follow-up. The median total smartwatch active wearing time changed from 8.1 (5.3– 8.5) hr/day to 9.8 (8.2–11.4) hr/day and this change was not significant. SDSCA-K score improved significantly from 0.29 (0.05–0.43) to 0.33 (0.32–0.68) (P=0.043).
Discussion
This pilot study demonstrated that integrating CGM with smartwatch-based lifestyle education led to significant short-term improvements in glycemic control, dietary intake, physical activity, and self-management skills among obese AYAs and T2D, in addition to ongoing T2D therapy (e.g., metformin). These findings suggest that a multimodal digital approach combining real-time glucose feedback and personalized lifestyle education using a wearable device can effectively promote behavioral changes and support the management of diabetes in this challenging population.
The intervention achieved a 23.9% increase in median TIR and a 70% reduction in median TAR250, along with 6 of 7 patients showing improved SDSCA-K scores. These results are consistent with previous studies of AYAs with T2D with CGM use showing behavioral and quality of life improvements. A total of 84% of patients wanted to use a CGM long-term, and the mean Pediatric Quality of Life inventory diabetes score significantly increased from 70 to 75 after using CGM [11,13]. However, these studies showed no improvement in CGM data (showing TIR change from 49.2% to 50.7%, 14% to 9%), even with feedback phone calls and insulin management. The main differences in our intervention were the 2 rounds of individualized intensive lifestyle education and the use of a smartwatch that focused on both lifestyle (diet, exercise, and sleep patterns) and glycemic control. This is consistent with a previous randomized control pilot study in which adult patients with T2D without insulin treatment improved their glycemic control with an interactive, personalized online treatment platform compared to a conventional management group (TIR increase; 32.2%±20.1% vs. 10.9%±21.0%) [22]. Visual feedback from CGM likely facilitates behavioral modifications by demonstrating postprandial glucose spikes, a strategy validated in qualitative studies [23].
Patients reduced their median peak daily total caloric intake by 26.7% and increased physical activity metrics (walking time, steps, exercise frequency, and walking calories burned), supported by smartwatch tracking. These findings align with the evidence that wearable devices enhance goal-setting through real-time feedback [24]. Notably, the absence of significant BMI changes despite behavioral improvements underscores the necessity for sustained interventions to impact adiposity. Previous studies have suggested that longterm lifestyle modifications can effectively manage T2D in AYAs, highlighting the critical role of lifestyle education alongside technological aids [25,26]. For instance, regular physical activity has been associated with significant reductions in HbA1c (by 0.8%) and BMI z-score; however, less than half of patients maintain consistent activity levels [27] and nearly 46.6% fail to adhere to lifestyle changes long-term, emphasizing the challenges of sustaining behavioral adjustments [28]. While this pilot study achieved full patient completion with variable adherence patterns, completion would most likely vary in longer studies. Prior research found no glycemic control difference between metformin-plus-lifestyle and metformin-alone groups, but this study uniquely integrated CGM and smartwatch technology to reinforce interventions [28].
There were some limitations to this pilot study. First, the nonrandomized design, small sample size (n=7), and absence of a control group limits the generalizability of the findings. The 10-day postintervention assessment period precludes the evaluation of long-term efficacy, which is a critical gap, given the chronicity of T2D. In addition, a potential selection bias exists because enrollment was managed by a single pediatrician, although this study attempted to minimize this bias by enrolling patients on a first-come, first-served basis within the study period. Relying on self-reported dietary data introduced potential recall bias, although photo-based logging via the ROTHY app partially mitigated this bias. The ROTHY app, which was previously available online, is no longer available for download after this study. This change hampered access to food diaries and limited adherence analysis. This highlights the complexities of conducting research that relies on commercial technologies and the need for strong partnerships with technology providers to ensure data continuity in future studies.
This study design allowed lifestyle therapy to be implemented while continuing existing T2D medications. It provides preliminary evidence that multimodal digital interventions can address critical gaps in AYA T2D management. However, achieving equitable implementation requires collaboration among clinicians, technology developers, and policymakers to ensure stable platform access and reimbursement structures. Although these results are encouraging, larger randomized controlled trials are needed to confirm long-term efficacy and to explore the feasibility of widespread clinical implementation. Future studies should evaluate the cost-effectiveness of this strategy and its effects on long-term diabetes-related outcomes in this challenging patient population.
In conclusion, integrating CGM and smartwatch technology with intensive lifestyle education for obese AYAs and T2D improved their glycemic control, dietary intake, physical activity, and self-management skills, demonstrating the promise of this multimodal approach in a short-term pilot study.
Supplementary materials
Supplementary Table 1 and Supplementary Figs. 1-2 are available at https://doi.org/10.6065/apem.2550026.013.
Comparison of anthropometric, glycemic control, dietary intake and lifelog data
Study protocol. SDSCA-K, Summary of Diabetes Self-Care Activities; CGM, continuous glucose monitoring. *Day 11–17 period may be extended by +3–7 days due to patient personal circumstances, holidays, etc. †The test was performed if results were not available within 3 months of enrollment. ‡Lifestyle education part 1 (on visit 2), data concerning physical measurements, body composition, daily food diaries, exercise routines, and sleep patterns were reviewed. Long-term weight control goals were set based on age, sex, body mass index, and lifestyle. Daily nutritional requirements were established for each patient according to weight goal. Part 2 (on visit 3), advanced training on the use of CGM devices was provided. Individualized plans including tailored meals and exercise and sleep routines, were developed.
Adherence changes. Continuous glucose monitoring (CGM) active time (%) (A) and smartwatch active wearing time (hr/day) (B) represent adherences to the study. Smartwatch active wearing time was defined as the combination of automatically collected walking time and the manually entered exercise and sleep time. IQR, interquartile range.
Notes
Conflicts of interest
No potential conflict of interest relevant to this article was reported.
Funding
This work was supported by Medical Science Research Institute grant from Kyung Hee University Hospital in 2022.
Data availability
The data that support the findings of this study can be provided by the corresponding author upon reasonable request.
Acknowledgments
We are especially grateful to Kil Yeon Lee (CEO of GI VITA and professor at the Department of Surgery at Kyung Hee University Medical Center, Seoul, Korea) for providing the CGM devices, smartwatches, and smartphone app (ROTHY app).
Author contribution
Conceptualization: MYH, HWJ; Data curation: HWJ; Formal analysis: HAW; Funding acquisition: MYH; Methodology: MYH, DHK, HWJ; Visualization: HAW; Writing - original draft: HAW; Writing - review & editing: HAW, MYH, JHC, KSS, DHK, HAW