This data analysis investigates the impact of Vively's comprehensive metabolic health solution, which integrates continuous glucose monitoring (CGM), a supportive mobile application, and healthcare professional support. The objective of this research is to identify potential areas for developing machine learning glucose prediction models by examining the effects of these integrated solutions on users' health outcomes.‍

Diabetes and prediabetes are growing global health concerns, with millions of people affected and numbers continuing to rise. Managing blood glucose levels is critical to prevent complications like cardiovascular disease, kidney damage, and nerve damage. However, traditional finger-prick tests only offer brief snapshots of glucose levels. They often miss key fluctuations throughout the day, which are essential for understanding and improving metabolic health.

Continuous Glucose Monitoring (CGM) technology has emerged as a powerful tool to address this gap. By measuring glucose in the interstitial fluid every few minutes, CGMs provide a clearer picture of daily glucose trends. This real-time data allows users to:

• Track and respond to glucose changes throughout the day
• Identify personal triggers and patterns
• Make more informed decisions about diet, exercise, and other habits

The Vively app works alongside CGM devices to enhance this experience. It helps users visualise their glucose data, offers personalised insights, and delivers educational content to support lasting behaviour change. The app aims to make metabolic health more accessible and actionable for individuals with diabetes, prediabetes, or anyone interested in optimising their glucose control.

This report explores the impact of CGM technology paired with the Vively app on blood glucose metrics and overall health outcomes. Drawing on data from a large group of users over a period of up to 21 months, it evaluates how this combined approach influences key indicators like HbA1c, time in range, glucose variability, and BMI.

It also outlines the methodology used in the analysis, including the user population, data sources, and measurement techniques. The findings offer valuable insights into how CGM and digital tools like Vively can support lifestyle change and improve metabolic outcomes—while also highlighting areas for future research, such as longer-term use and population diversity.

Literature review

Continuous glucose monitoring (CGM) is revolutionising diabetes management by providing detailed glucose data that enables more informed decision-making by patients and clinicians. CGM systems measure interstitial glucose levels via a sensor inserted under the skin and generate an internationally standardised ambulatory glucose profile (AGP) report depicting glycaemic patterns.

The history, sensing mechanisms, and recent advancements in CGM technologies were reviewed by Lee et al. (2021). The two main types of CGM systems are real-time CGM, which provides continuous data passively, and intermittently scanned CGM (isCGM) or flash glucose monitoring (FGM), where the user must actively scan the sensor to obtain readings. Most real-time CGMs employ electrochemical sensing using glucose oxidase and either monitor generated hydrogen peroxide or use a redox mediator. The Eversense implantable CGM uniquely utilises a fluorescent polymer sensing mechanism. Ongoing CGM innovations aim to extend sensor lifetime, improve accuracy and usability, and incorporate artificial intelligence.

Lin et al. (2021) reviewed the evidence on clinical efficacy, user experience, and cost-effectiveness of CGM in diabetes. For type 1 diabetes, multiple randomised controlled trials (RCTs) have demonstrated that CGM improves glycated haemoglobin (HbA1c), increases time in range (TIR), and reduces hypoglycaemia compared to self-monitoring of blood glucose (SMBG). These benefits are especially pronounced in those with suboptimal glycaemic control and hypoglycaemia unawareness. In type 2 diabetes, RCTs have shown HbA1c reductions of 0.3–0.5% with CGM vs. SMBG, mainly in insulin-treated patients. Flash CGM appears particularly effective in younger type 2 patients with higher HbA1c using less frequent SMBG.

Beyond glycaemia, CGM improves diabetes treatment satisfaction and quality of life and reduces diabetes distress. Cost-effectiveness analyses suggest CGM is cost-effective in type 1 diabetes through reducing costly complications. Limitations include high device costs, potential sensor inaccuracies, alarm fatigue, and skin irritation. Ongoing uptake of CGM is expected as the technology advances and evidence accumulates.

Some researchers are exploring CGM-derived metrics to predict progression to type 2 diabetes in at-risk individuals. Colás et al. (2019) used Detrended Fluctuation Analysis (DFA) to quantify the complexity of glucose time series obtained by CGM in patients with hypertension. Their optimised DFA algorithm, not using the standard pre-processing step of time series integration, demonstrated significant predictive power for type 2 diabetes onset over a median 33-month follow-up. The DFA complexity metric provided distinct information from conventional CGM variability measures in a principal component analysis. This novel application of CGM and complexity analysis may help target preventive interventions.

Martens et al. (2021) conducted an RCT evaluating real-time CGM in 175 type 2 diabetes patients on basal insulin, managed in primary care. Over 8 months, HbA1c decreased by 1.1% with CGM vs. 0.6% with SMBG (p=0.02). CGM also significantly increased TIR by 15% and reduced hypoglycaemia compared to SMBG. This supports CGM as a powerful tool for optimising insulin therapy in real-world type 2 diabetes care.

Zahedani et al. (2023) investigated the effectiveness of a digital health application integrating wearable data and behavioral patterns to improve metabolic health. The study enrolled 2,217 participants with varying glucose levels, including those in the normal range, prediabetes, and type 2 diabetes. Participants used CGM for 28 days, logged food intake, physical activity, and body weight via a smartphone app, and received personalized recommendations based on their preferences, goals, and observed glycemic patterns. The study found significant improvements in metabolic health markers, particularly in non-diabetic participants, including reduced hyperglycemia, glucose variability, and hypoglycemia. Participants also experienced weight loss and improvements in healthy eating habits. The personalized approach and potential for scalability make this technology promising for type 2 diabetes prevention and treatment.

In summary, the evidence convincingly demonstrates that CGM improves glycemic control and patient-reported outcomes in both type 1 and type 2 diabetes, with a particularly strong benefit in type 1. Flash CGM appears effective in selected type 2 patient subgroups. The cost-effectiveness of CGM in type 1 diabetes is established, with further studies needed in type 2. Novel applications of CGM, such as complexity analysis and integration with digital health apps, may help predict and prevent type 2 diabetes onset. With ongoing technological advances and increasing evidence and accessibility, CGM has the potential to become the standard of care in diabetes management. Expanding CGM use in clinical practice can help alleviate the health and economic burden of the global diabetes epidemic.

Methodology

This study analysed data from Vively app users, including individuals with diabetes, prediabetes, and normal glucose levels. Participants were provided with a CGM device every three months, with the option to purchase and use additional devices. The Vively app offered support and guidance on CGM use, as well as access to healthcare professionals for optimising outcomes.

Data were collected from the first and last CGM readings for each participant, with the duration of app use ranging from 14 days to 21 months. Key metrics were calculated, including average blood glucose, time in range (TIR), estimated HbA1c, glucose variability, and BMI. TIR was defined as the percentage of time spent within the target glucose range of 70-180 mg/dL (3.9-10.0 mmol/L) for individuals with diabetes and 70-140 mg/dL (3.9-7.8 mmol/L) for those without diabetes. Glucose variability was assessed using the coefficient of variation (CV) of glucose readings.

Comparisons were made between the first and last CGM readings to evaluate changes in the aforementioned metrics over time. Subgroup analyses were conducted based on factors such as gender, age (19-30, 31-64, 65-100 years), and the involvement of practitioners or health coaches in the participants' care. Additionally, changes in metrics were specifically examined for women with polycystic ovary syndrome (PCOS).

Key metrics calculation:

• Changes in average blood glucose, time in range, HbA1c, variability, and BMI for each group (diabetes, prediabetes, normal)
• Comparison of metrics based on gender, age, and practitioner/health-coach consultation
• Comparison of users with different service timeframes
• Changes in metrics for women with PCOS

Data analysis

Data source and data processing

All data is stored in the “cgm_db” database, eight tables are used here.

Key preprocessing:

• The cgm readings are grouped by patient_id and sensor_id to calculate the estimated HbA1C to categorize diabetes, prediabetes and normal users. Observed those very low HbA1Cs are from CGMs with limited rows of data (far less than 14 days); those CGMs are deleted and assigned the groups again, resulting in patients with three groups as expected.

• Choose users with multiple CGMs, and there are 1812 users left.

• Calculating Time in Range, for diabetes TIR is 3.9-10 mmol/L, for others is 3.9-8 mmol/L
• Calculating variability, the coefficient of the variability is used here.

• Spikes are calculated based on the below thresholds. The events are detected from the highest values. If two different events happen with a gap of less than 6 rows, they will be updated to one event. This is because sometimes the spikes are around the thresholds; the fluctuation will make them two events at almost the same time. Then, the average spike counts are calculated from the total spikes divided by CGM days (end_date-start_date).

Results

All users

The users in this study utilised the Vively service for periods ranging from 14 days to 21 months (e.g., patient_id 13858). The analysis reveals improvements across all metrics for diabetes and pre-diabetes users. However, for users with normal glucose levels, the improvements in HbA1c, TIR, and GV were not as pronounced, although a decrease in daily spike counts was observed.

This finding suggests that CGM enables easier management of glucose spikes, empowering users to take effective actions to reduce these spikes. In contrast, other measures such as HbA1c, TIR, and GV are influenced by multiple factors, including sleep, stress levels, and overall lifestyle, which may be more challenging to improve solely with the help of CGM.

Users who use multiple CGMs are not the same users who upload smart scale readings or have BMI scores. Additionally, the total number of users with BMI scores is small. Therefore, we are examining all users with BMI scores, rather than just those who also use multiple CGMs. A decrease over all three groups can be observed.

Gender

We have 1511 out of 1812 users provide gender information. There isn't a significant difference between male and female users in terms of the effect after using multiple CGMs. It's noteworthy that similar numbers of men and women with diabetes (37 and 39) and prediabetes (59 and 52) are using multiple CGMs. However, among normal users, there are significantly more females than males, with 960 females compared to 364 males, almost three times as many.

Age

Among multiple CGM users with age information, 89.6% fall into the 31-64 age group, totalling 1353 users. The glucose level change status is consistent with the overall user distribution. In the 19-30 age group, there is only one user with diabetes and three with prediabetes. Similarly, in the 65-100 age group, there are six users with diabetes and seven with prediabetes.

‍

PCOS group

A similar trend is observed in the PCOS group, showing improvement across all measures, although not distinctly.

Overall, users have shown improvement, but using a CGM does not guarantee this outcome. For example, when comparing HbA1c levels, the average result from 942 users showed improvement, while 870 users showed worsened levels. The numbers of improved and worsened users are listed in the table below. Further study is needed to understand the factors contributing to these different outcomes. Subsequent analysis will group users based on timeframe, gender, age, and professional healthcare support to delve deeper into these findings.

Timeframe

Considering the substantial impact of the timeframe for using CGMs, we have classified users into three categories based on the service duration: 0-3 months, 3-6 months, and over 6 months.

Overall, the relationship between the outcome and duration is not linear. While normal and prediabetic users show similar outcomes with longer usage, diabetic users experience greater improvement when using CGMs for more than 6 months, with an average decrease of 1.1 in estimated HbA1C compared to improvements ranging from 0.2 to 0.7 in other groups.

However, when we examine the group of improved users, their progress is particularly noticeable, especially among those with diabetes. There is a clear trend of improvement with longer CGM usage. For instance, in the table below, diabetic users show an improvement in HbA1C of 1.1 when using the CGM for less than 3 months. This improvement increases to 1.32 after another 3 months, and further to 1.58 for usage exceeding 6 months.

Practitioner and health coach

In addition to basic CGM readings, we examined the impact of health coaches and practitioners on improvement. However, we did not find a significant effect from their support. This lack of impact may be due to the limited comparison groups, as fewer diabetes and prediabetes users using multiple CGMs seek help from health coaches and practitioners. Although the ratio of diabetes and prediabetes users engaging with health coaches and practitioners is similar to or higher than the normal group, there are fewer samples in these groups.

It's worth noting that long-term users, especially normal users who may be experiencing alarming situations, prefer to work with practitioners. Almost half of the users who have been using CGMs for longer than 6 months are working with practitioners.

Table of user ratio - health coach

Table of user ratio - practitioner

‍

If we examine the contingency table comparing the improvement status for users with and without practitioner support, we can see that there is a modest difference in the percentage of users who improved. Users who had the assistance of practitioners showed a 55.83% improvement rate, which is approximately 5 percentage points higher than the 50.57% improvement rate for users without practitioner support.

‍

Regarding the lack of a comparison group, consider the users in the 3-6 months group, for instance. While an improvement in estimated HbA1C is observed, there is a decrease in GV. However, because there are only three diabetic users in this group, the findings are not statistically significant.

‍

Findings

The data analysis reveals several findings:

• Improvements in average blood glucose, time in range, HbA1c, variability, and BMI were observed across diabetes and prediabetes users between the first and last CGM readings.
• CGM helps in monitoring and reducing spikes across all users. Particularly for users with normal glucose levels who showed a decrease in daily spike counts despite less pronounced improvements in other metrics.
• The trend of improvement with longer CGM usage is clearly demonstrated for users with diabetes. This is more obvious in the improved groups.
• The gender disparity in the normal group, with nearly three times as many female users as male users, is a noteworthy finding. It might be worth discussing potential reasons for this difference and its implications.
• Fewer elderly users utilise the service, with the majority of users (89.6%) falling into the 31-64 age group. This demographic distribution could have implications for targeting and tailoring the service, as well as for further division of the groups in future analyses.
• Long-term users are more likely to be working with practitioners.
• While overall improvements were observed, individual user outcomes varied, with some users showing worsened metrics. Further investigation into the factors contributing to these different outcomes could be valuable.
• The limited number of users with diabetes and prediabetes in certain comparison groups when considering the support from health coaches and practitioners (e.g., age groups, 3-6 months duration) affects the statistical significance of the findings for these subgroups.

Discussion and interpretation of findings

The findings suggest that the Vively app and continuous glucose monitors (CGMs) can effectively support users in improving glucose control—particularly in managing glucose spikes. Here’s a breakdown of the key takeaways:

Spike control across all users:
Reductions in daily glucose spikes were observed in all groups, including those with normal glucose levels. This highlights CGM's role in helping users recognise and act on real-time glucose fluctuations, even when traditional metrics remain unchanged.

Improvements in long-term users:
Users who engaged with CGM and the app over longer periods saw greater improvements, especially in the diabetes group. This indicates that sustained use helps individuals better understand their patterns and adjust behaviour over time.

Potential synergy with professional support:
Long-term users were more likely to be working with healthcare practitioners. This reinforces the benefit of combining real-time CGM insights with professional guidance to personalise care.

Gender disparity in normal glucose users:
A higher proportion of female users was noted in this group. This may be influenced by awareness levels, health-seeking behaviours, or referral patterns. Future outreach could focus on balancing adoption across genders.

Age-based opportunities for growth:
Most users fell within the 31–64 age range. Fewer older adults were represented, possibly due to:

• Limited familiarity with mobile apps and wearables
• Affordability concerns or fixed incomes
• A lack of perceived need, particularly in well-managed individuals

Real-time insights promote proactive action:
CGMs empower users to take immediate steps, like adjusting meals or activity levels based on personalised glucose responses, fostering better metabolic outcomes.

Future opportunities for Vively:
To enhance its impact, Vively could consider features like:

• Practitioner dashboards or secure data sharing
• Integration with electronic health record systems
• In-app messaging to connect users with healthcare providers

These findings support the value of combining CGM data with app-based insights, especially when paired with ongoing professional care.

Conclusion

While the overall results demonstrate the potential benefits of the Vively app and CGM technology, the variability in individual user outcomes underscores the complexity of glucose management. Factors such as medication adherence, lifestyle choices, and comorbidities may influence the effectiveness of the intervention. Further research exploring these factors could help identify strategies to optimize outcomes for all users.

The limitations of the current analysis, such as the small sample sizes in certain comparison groups, highlight the need for ongoing data collection and analysis. As more users adopt the technology and generate data over longer periods, more robust conclusions can be drawn about the impact of the Vively app and CGM technology on glucose control and health outcomes.

In conclusion, the findings of this analysis suggest that the Vively app and CGM technology can be valuable tools for improving glucose control, particularly in terms of reducing glucose spikes. The trends observed in relation to usage duration, age, and practitioner involvement provide insights into potential strategies for optimising the impact of the intervention. However, further research is needed to address the limitations of the current analysis and explore the factors influencing individual user outcomes.

Limitations and future scope

The current analysis is limited by

• The service duration: It would be beneficial to analyse data from a larger cohort of users who have been consistently using the service for at least 12 months.
• A more controlled scheme is helpful to distinguish the outcome of different groups: For example, comparing outcomes between users who received varying durations or frequencies of health coaching or between users with different levels of engagement with the app's features and challenges could provide valuable insights into the most effective strategies for supporting glucose control.
• Limited data on specific user groups: The current analysis included a relatively small number of users with specific conditions, such as PCOS, long-term users with health coaches, which limits the generalisability of the findings to these populations.
• Understanding factors contributing to worsened outcomes: It is important to investigate the factors contributing to worsened outcomes among some users. Future research could explore the characteristics, behaviours, and experiences of users who did not benefit from the intervention to identify potential barriers to success and inform strategies for improving outcomes.
• Assessing behavioral changes: The current analysis primarily focused on changes in glucose control metrics and related health outcomes. However, understanding the impact of the Vively app and CGM technology on users' behaviors, such as dietary choices, physical activity, and sleep patterns, could provide valuable insights into the mechanisms underlying the observed improvements. Future studies could incorporate data on these behavioral factors, such as food logs or sleep tracking data, to examine how the intervention influences lifestyle choices and the subsequent impact on glucose control.
• Potential self-selection bias: It is possible that the users who chose to use the Vively app and CGM technology for an extended period were more motivated or engaged in their health management compared to the general population. This self-selection bias could potentially influence the observed outcomes. Future research could employ randomized controlled trials or other study designs to minimize the impact of self-selection bias and provide a more rigorous evaluation of the intervention's effectiveness.

Reference

Lee, I., Probst, D., Klonoff, D., & Sode, K. (2021). Continuous glucose monitoring systems: Current status and future perspectives of the flagship technologies in biosensor research. Biosensors and Bioelectronics, 181, 113054. https://doi.org/10.1016/j.bios.2021.113054

Lin, R., Brown, F., James, S., Jones, J., & Ekinci, E. (2021). Continuous glucose monitoring: A review of the evidence in type 1 and 2 diabetes mellitus. Diabetic Medicine, 38(5), e14528. https://doi.org/10.1111/dme.14528

Colás, A., Vigil, L., Vargas, B., Cuesta-Frau, D., & Varela, M. (2019). Detrended fluctuation analysis in the prediction of type 2 diabetes mellitus in patients at risk: Model optimization and comparison with other metrics. PLoS One, 14(12), e0225817. https://doi.org/10.1371/journal.pone.0225817

Martens, T., Beck, R. W., Bailey, R., Ruedy, K. J., Calhoun, P., Peters, A. L., ... & MOBILE Study Group. (2021). Effect of continuous glucose monitoring on glycemic control in patients with type 2 diabetes treated with basal insulin: A randomized clinical trial. JAMA, 325(22), 2262-2272. https://doi.org/10.1001/jama.2021.7444""

Zahedani, A. D., McLaughlin, T., Veluvali, A., Aghaeepour, N., Hosseinian, A., Agarwal, S., ... & Snyder, M. (2023). Digital health application integrating wearable data and behavioral patterns improves metabolic health. NPJ Digital Medicine, 6(1), 216.