Factors Affecting Continuous Glucose Monitoring Results: A Meta-Analysis ()
1. Introduction
Diabetes is a group of heterogeneous diseases, commonly characterized by elevated blood glucose levels, or hyperglycemia. It can be classified into Type 1 diabetes and Type 2 diabetes. Type 1 diabetes (T1DM) results from immune-mediated destruction of pancreatic β-cells, leading to impaired insulin secretion, and is mostly characterized by absolute insulin deficiency. Type 2 diabetes (T2DM) involves reduced insulin action (insulin resistance) along with progressive loss of β-cell function, typically starting as relative insulin deficiency. This often results in disrupted glucose-dependent insulin secretion by the receptors. Both types can develop into severe hyperglycemic symptoms such as polyuria, thirst, fatigue, unexplained weight loss, vision impairment, and increased risk of infections, ketoacidosis or non-ketotic hyperosmolar syndrome, with potential for coma. Chronic hyperglycemia can also lead to dysregulation of insulin secretion and/or action, and is associated with long-term damage and functional impairments of various tissues and organs (eyes, kidneys, nerves, heart, and blood vessels) as well as cancer.
With the increasingly advanced development of treatment and monitoring technologies for diabetes patients domestically and internationally, there are some differences in the prevalence of diabetes across countries [1]. According to data from the National Health and Nutrition Examination Survey (NHANES) in the United States from 2013 to 2023, the prevalence of diabetes among adults has not shown a significant increase. However, patients who have been diagnosed with diabetes have experienced worsening conditions after 10 years. In China, the incidence of diabetes is also rising annually, with Type 2 diabetes (T2DM) being predominant [2]. It is particularly important to manage diabetes patients both in hospitals and at home due to the large number of cases. Poor management that leads to consistently high or low blood glucose levels in diabetes patients can cause irreversible damage to target organs and even result in life-threatening complications [3]. Therefore, it is crucial for diabetes patients to continuously and stably monitor their blood glucose levels and maintain them within the prescribed normal range. However, traditional blood glucose monitoring techniques often involve invasive procedures, causing discomfort and pain to patients, which decreases their compliance and motivation for self-monitoring. Diabetes patients need a precise and comfortable way to monitor blood glucose in order to improve their quality of life. The development of continuous glucose monitoring technology can address this issue and provide significant relief for such patients [4].
Continuous dynamic blood glucose monitoring (CGM) refers to the technology that continuously monitors glucose concentrations in the interstitial fluid of subcutaneous tissue through glucose sensors [5]. This technology provides continuous, comprehensive, and reliable 24-hour blood glucose information, helping to understand the trends and characteristics of blood glucose fluctuations. It enables effective blood glucose monitoring and better management for diabetes patients, and is favored by clinical patients, doctors, and nurses [6]. However, the results from continuous glucose monitoring (CGM) can also be influenced by certain factors [4]. This study aims to consider and eliminate these influencing factors, allowing for more systematic and optimized management of patients’ blood glucose levels. It involves a comprehensive systematic evaluation and analysis of recent research on CGM, providing a reliable evidence-based basis for clinical blood glucose management.
2. Methods
2.1. Inclusion and Exclusion Criteria
1) Inclusion Criteria: a) Participants aged ≥ 18 years; b) Populations diagnosed with diabetes or not yet diagnosed with diabetes; c) Diabetes patients who are receiving or not receiving treatment; d) Observational studies.
2) Exclusion Criteria: a) Individuals who cannot understand or refuse to use continuous glucose monitoring; b) Less than 70% usage time of continuous glucose monitoring; c) Participants with impaired decision-making capacity or without decision-making capacity; d) Participants who have died or whose continuous glucose monitoring data is missing.
2.2. Literature Retrieval Strategy
Computer searches were conducted in PubMed, Embase, Web of Science, and the Cochrane Library, with a search period covering from the establishment of each database until March 2025. The search employed English subject terms combined with advanced search strategies. The search query was: Advanced search ((((((Continuous dynamic blood glucose monitoring, influence factor) OR (Interstitial fluid blood glucose, influence factor)) AND ((Continuous dynamic blood glucose monitoring, management) OR (Interstitial fluid blood glucose, management) OR (Freestyle Libre, influence factor) OR (Freestyle Libre, management)))))). Subject term search: “Continuous dynamic blood glucose monitoring” [Mesh] AND (influence factor, management). Search results yielded a total of 526 articles, which were included in the meta-analysis. The search process is illustrated in Figure 1.
2.3. Search Results
We retrieved a total of 526 articles from four electronic databases. Using EndNeto software, we excluded 425 ineligible articles. For the remaining 101 articles, we entered information such as the title, author, and publication year, and the system automatically identified and excluded 14 duplicate studies. Finally, we conducted an in-depth review of the remaining 87 articles. Among these: 17 were animal experiments, 31 were studies involving participants under 18 years old, 14 articles were missing, and 15 were reviews. Ultimately, 11 studies met our requirements and were included for analysis.
Figure 1. Flowchart of literature screening.
2.4. Data Extraction
The research team for this study consisted of four members. During the retrieval and analysis of articles, three researchers selected articles and extracted data based on inclusion and exclusion criteria. They then cross-verified the selections. If any contentious issues arose, the other three researchers were consulted to resolve them. Articles were preliminarily included by reviewing their titles and abstracts. After identifying and excluding duplicate articles, each article was then read in detail to finalize the selection of appropriate articles. Data were extracted from these chosen articles and included information such as: author, country, publication year, study type, sample size, statistical tests (M ± SD), and outcome indicators. This information was compiled into a summary table, such as Table 1. The primary outcome indicator was MARD (Mean Absolute Relative Difference); the smaller the value, the more accurately continuous glucose monitoring reflects glucose levels [7]. Secondary outcome indicators included Time in Range (TIR) %, Glycemic Variability (CG-MBY), Precision Absolute Relative Difference (PARD), etc. as detailed in Tables 1-4.
Table 1. Diagram of general characteristics.
Table 2. Article quality evaluation form.
2.5. Quality Assessment
The methodological quality of non-randomized controlled trial designs was evaluated based on the Risk of Bias in Non-randomized Studies—of Interventions (ROBINS-I) tool, for quality assessment of the articles. The ROBINS-I scale uses a “dot system” to rate the included studies in three aspects: before intervention grouping, during intervention grouping, and after intervention grouping. The ratings include: Before intervention grouping—confounding bias, participant selection bias; During intervention grouping—intervention classification bias; After intervention grouping—bias due to deviations from intended interventions, missing data bias, outcome measurement bias, and selective reporting bias [19]. There are a total of seven components. If all components are rated as low risk, the study is considered low risk. If any component is rated as moderate, high, or very high risk, the study is determined to be of the highest level of risk. If any component is rated as no information, the study is defined as having no information. Specific ratings are detailed in Table 5.
Table 3. Article quality evaluation form.
Table 4. Article quality evaluation form.
|
|
|
Non-exposure group (control group) |
Exposure group (Experimental group) |
Author |
Year |
Influencing factors |
Non-exposure (experimental group) Sample size (control group) |
Eating high-carbohydrate before exercise (MARD%) |
High-carbohydrate intake: Standard difference before exercise |
Exposure (experimental group) Sample size (control group) |
Eating high-carbohydrate exercise (MARD%) |
Eating high-carbohydrate exercise with standard difference during exercise |
Matzka M; |
2024 |
Sports |
199 |
15.7 |
14.1 |
101 |
17.1 |
13.6 |
Bauhaus H; |
2023 |
Sports |
519 |
17 |
10 |
519 |
17 |
9 |
Table 5. Article overall quality evaluation form.
Author |
Year |
Total offset risk assessment |
Kevin Hanson |
2024 |
|
Medium-risk bias |
Avari P |
2023 |
|
Medium-risk bias |
Toyota M |
2021 |
|
Medium-risk bias |
Narasaki Y |
2024 |
|
Low-risk bias |
Villard O |
2022 |
|
Low-risk bias |
Matzka M |
2024 |
|
No information |
Villa-Tamayo |
2024 |
|
Medium-risk bias |
Continued
Bauhaus H |
2023 |
|
High-risk bias |
Olafsdottir AF |
2022 |
|
High-risk bias |
Eichenlaub M |
2025 |
|
Medium-risk bias |
Pleus S |
2022 |
|
High-risk bias |
Green 
: Low-risk bias; Yellow 
: Medium-risk bias; Red 
: High-risk bias; Black 
: Extremely high risk bias; Circle 
: No information.
2.6. Data Collection and Statistical Analysis
All included studies used consistent measurement tools, enabling a meta-analysis of combined quantitative data. The mean and standard deviation of the scores on the humanistic care ability scale in each study were summarized using Stata SE.14, and presented using weighted mean difference (WMD) effect sizes and 95% confidence intervals (CI). The Cochrane Q test and I2 statistics were utilized with I2 values of 25%, 50%, and 75%, indicating low, moderate, and high heterogeneity, respectively [20]. When I2 > 50%, and p < 0.05, a random-effects model was used.
3. Results
3.1. Overall Results
Results, as shown in Figure 2, indicate that among the 11 studies that could affect continuous glucose monitoring (CGM) outcomes, the analysis produced χ2: 3438.01, I2 = 99.7%, P = 0.266 > 0.05. There were no significant differences in the effects of influencing factors between the groups.
Figure 2. Forest plot of factors affecting continuous glucose monitoring.
3.2. Subgroup Analysis of the Study
3.2.1. Product Performance Result
As shown in Figure 3, indicate that among the 11 studies that could affect continuous glucose monitoring (CGM) outcomes, the analysis yielded χ2: 2046.12, I2 = 99.8%, P = 0.269 > 0.05. There were no significant differences in the effects of influencing factors between the groups.
Figure 3. Subgroup analysis of influencing factors of product performance in continuous ambulatory blood glucose monitoring results forest plot.
Figure 4. Subgroup analysis of influencing factors of hemodialysis based on continuous ambulatory glucose monitoring results, forest plot.
3.2.2. Hemodialysis
As a result, for example: Figure 4, among the 11 studies that might affect the results of continuous dynamic glucose monitoring (CGM), the analysis results showed that χ2: 30.7, I2 = 93.8%, P = 0.413 > 0.05. There is no difference in the effect of each influencing factor between groups.
3.2.3. Exercise Result
As shown in Figure 4, indicate that among the 11 studies that could affect continuous glucose monitoring (CGM) outcomes, the analysis yielded χ2: 0.62, I2 = 0.0%, P = 0.432 > 0.05 (see in Figure 5). There were no significant differences in the effects of influencing factors between the groups.
Figure 5. Subgroup analysis of exercise influencing factors in continuous ambulatory glucose monitoring results Forest plot.
4. Discussion
4.1. Product Performance
With continuous technological advancements, the methods for continuous glucose monitoring (CGM) have been consistently updated and developed in numerous studies. These advancements have reduced interfering factors, improved monitoring signals and sensitivity, and even allowed for stable and accurate blood glucose readings in critically ill patients experiencing hypoxia. In recent years, many manufacturers have been committed to product improvements, calibration, algorithms, and technological updates, making CGM more precise and reliable [21]-[27]. This has facilitated easier self-management of blood glucose for patients, reducing pain and anxiety and increasing patient compliance with self-management. The stability of CGM provides clinical value for healthcare providers monitoring blood glucose in different patients, improving work efficiency and effectively preventing hypoglycemia [28]. It also offers healthcare professionals a reliable basis for clinically guiding patients in dietary and medication management [29].
4.2. Hemodialysis
Diabetes is a leading cause of end-stage renal failure. Concerns have been raised about whether continuous glucose monitoring (CGM) results might be inaccurate in hemodialysis patients. However, analysis indicates that CGM accuracy is not impacted in these patients. In hemodialysis patients, CGM can effectively prevent adverse events such as hypoglycemia [30]. It also aids in guiding insulin medication during dialysis, providing valuable insights. CGM can be a significant benefit for blood glucose management in hemodialysis patients [31] [32].
4.3. Exercise
Regarding exercise, there is some debate about whether it interferes with continuous glucose monitoring (CGM) results. Some studies suggest that exercise may impact CGM results, but this remains controversial [33] [34]. From the analysis, exercise does not appear to significantly affect the accuracy of the results. This could be related to product performance, as research indicates that the Dexcom G6 product shows better accuracy during exercise [35].
5. Conclusion
In summary, product performance, hemodialysis, and exercise do not excessively impact the accuracy of continuous glucose monitoring (CGM) results. Although some studies suggest that peritoneal dialysis, cardiac surgery, surgical operations, radiofrequency ablation, and high-altitude living environments can affect CGM results, the research is limited [7] [36]-[39]. There are no recent RCTs or observational studies to serve as the best evidence. In the future, we hope more similar studies will be published to provide a foundation for accurately assessing patient blood glucose levels in clinical settings.