This study evaluated the effectiveness of a mobile application-assisted Lifestyle intervention integrated with CGM in improving glycemic control and body composition among the employees of manufacturing companies in South Korea. Moreover, it aimed to identify factors predicting responsiveness to the intervention. Following the 12-week intervention period, HbA1c, BMI, body weight, body fat mass, and tMFR of the participants significantly improved. Notably, these improvements were particularly pronounced in participants with elevated baseline HbA1c and BMI values. Furthermore, glycemic variability measured through CGM FSV was a significant predictor of intervention effectiveness, emphasizing the consistent association between stable glucose variability and favorable outcomes [11]. Moreover, our CGM data analysis revealed significant improvements in short-term glycemic variability indices, highlighting the benefits of the lifestyle intervention, which can be effective even in the early stages of intervention.
This study had several notable strengths. First, from a conceptual and methodological perspective, this study employed an integrated intervention model that combined mobile health technology, CGM feedback, and tailored health coaching in a structured sequential approach. This design aligned with the sequential behavior-change framework, starting with real-time glycemic monitoring and targeted feedback, advancing to active personalized behavioral modifications, and culminating in autonomous habit maintenance. Second, this study’s analytical rigor further enhanced its validity. Along with basic preintervention–postintervention comparisons, subgroup analyses were performed using robust criteria for defining responders, including sensitivity analyses with stricter thresholds. This study comprehensively explored individual variability in responsiveness to digital intervention by incorporating CGM-derived glucose variability data into multivariable logistic regression models, providing valuable insights into the key factors predicting the effectiveness of digital therapeutics. Finally, our findings were validated within a cohort of manufacturing workers, providing important academic evidence on how the management of chronic metabolic diseases, including diabetes care, should be structured and evaluated within this occupational group. Our results provided valuable insights into the tailored implementation and evaluation of digital health interventions specifically designed for manufacturing employees, considering their distinct occupational environment, characterized by irregular work schedules, sedentary behaviors, occupational stressors, and disrupted circadian rhythms. Thus, this study highlighted the potential of mobile technology and CGM-integrated interventions. It focused not only on enhancing metabolic outcomes but also on informing future occupational health strategies and policies aimed at effectively managing chronic conditions in workplace settings.
Previous studies have increasingly demonstrated the effectiveness of mobile health and digital therapeutics in managing chronic metabolic diseases [12, 13]. Previously, integrated mobile and CGM-based interventions have consistently shown HbA1c level reductions of 0.3–0.7%, alongside BMI improvements [14,15,16]. Our findings were in line with previously reported results in terms of the magnitude and direction of improvement. Furthermore, previous studies utilizing real-time CGM feedback have emphasized the advantages of individualized interventions in improving glycemic control [17, 18], which was consistent with our identification of CGM-derived glucose variability as a significant predictive marker. However, some earlier studies have indicated the limited effectiveness of CGM-based interventions on BMI reduction alone [19, 20]. In contrast, our results demonstrated significant BMI improvements, highlighting the crucial role of targeted interventions and coaching. Moreover, we identified significant short-term improvements in glycemic variability indices within the initial 2-week phase of CGM monitoring, characterized by significantly reduced spike counts, ARV, and FSV. These early changes indicated immediate behavioral adaptations in dietary habits and physical activity triggered by real-time glucose feedback. Improved glycemic stability can decrease oxidative stress, systemic inflammation, and overall metabolic burden even in a short-term period, thereby setting the stage for subsequent sustained metabolic improvements observed throughout the intervention [21, 22].
Clinically, this study offers compelling evidence supporting the utility of integrated mobile applications and CGM-based tailored interventions within workplace health promotion programs. Our findings emphasized the particular efficacy of this integrated approach among individuals with elevated baseline HbA1c and BMI, indicating that targeted strategies can enhance the efficiency of relevant interventions and subsequent clinical outcomes. Moreover, validating CGM-based glucose variability as a strong predictive indicator for responsiveness further equips clinicians and healthcare providers with practical tools for improving the precision and effectiveness of future interventions.
Glycemic variability indicators, such as FSV, are being increasingly recognized in clinical practice owing to their relevance beyond mean glucose and HbA1c. Glycemic variability is now recognized as a significant independent risk factor for a range of adverse clinical outcomes, including microvascular and macrovascular complications and even mortality in patients with diabetes. Several studies have reported that higher glycemic variability is significantly associated with increased risk of diabetic retinopathy, nephropathy, cardiovascular events, and all-cause mortality [23,24,25]. FSV has the potential to serve as a practical indicator for anticipating glycemic responses to interventions. In our study, lower baseline FSV was a significant predictor of being a responder following the intervention, highlighting the practical utility of this parameter. These findings were in line with prior literature, which indicated that blood glucose variability might reflect underlying oxidative stress [26]. Given the increasing availability of real-time CGM systems, markers such as FSV can play a pivotal role in personalizing diabetes management. This approach enables early identification of individuals at higher risk of poor glycemic outcomes and allows for tailoring interventions accordingly. Recent studies also supported the use of glycemic variability metrics, including FSV, for risk stratification and treatment response monitoring in people with type 1 and type 2 diabetes [27], which was consistent with our results.
In the current study, the total duration of the mobile application-assisted Lifestyle intervention was 12 weeks because it is widely recognized as the minimum duration necessary for the emergence of observable changes in laboratory markers and body composition [28, 29]. Notably, approximately 40% of participants had pre-existing diabetes, and prior studies have demonstrated that at least 12 weeks of lifestyle intervention is required to achieve statistically significant reductions in HbA1c levels [30]. Within this framework, we implemented CGM monitoring during the first 2 weeks to capture individual glycemic variability related to daily lifestyle patterns. This early monitoring phase involved providing personalized feedback to enhance participant motivation and engagement throughout the intervention period. In addition, glycemic variability indices measured during this initial CGM phase might help predict the long-term effectiveness of the interventions.
Despite its strengths, this study had several limitations. First, despite a multifaceted subgroup analysis, the absence of a control group precluded definitive conclusions about causality or the influence of confounding variables and natural disease progression. A control group could not be established because of workplace constraints. As a result, the causal interpretation of the findings was limited, although subgroup comparisons were conducted to provide some internal reference. Second, the relatively short intervention duration (12 weeks) limited our analysis of the long-term maintenance of behavioral changes and subsequent health outcomes. Third, the study population predominantly consisted of male employees (98%) working in physically labor-intensive manufacturing environments, which limited the generalizability of the findings. This male predominance likely reflected the gender composition typical of such workplaces and the potentially higher health-related motivation for lifestyle interventions observed among male workers in this sector. Fourth, practical barriers, including the costs associated with CGM devices and individualized coaching, as well as the required digital literacy, may have hindered the broader dissemination and application of such interventions in various real-world settings. We did not repeat CGM monitoring after the service because this study was based on a real-world service and was therefore designed with cost-effectiveness in mind. Further follow-up studies based on further CGM monitoring are needed to thoroughly verify the glycemic regulatory effects of this service. Finally, several key behavioral variables exhibited high rates of missingness. Although a complete case analysis was conducted, our approach may have introduced selection bias if individuals with complete data systematically differed from those with missing data. These behavioral variables were not included in the regression analysis because of the substantial proportion of missing values, which might have limited our ability to account for potential lifestyle-related confounders. This potential bias should be considered when interpreting the results. In addition, the lack of some key variables that could objectively measure application usability and engagement, such as the duration and frequency of participants’ use of the application and the frequency of their uploads of information related to dietary habits and physical activity, limited our analysis. Future research should incorporate more complete and systematically collected behavioral data to enhance the robustness and generalizability of the findings.