Associations Between Continuous Glucose Monitoring-derived Metrics and Arterial Stiffness in Patients with Type 2 Diabetes


 Background Previous studies have suggested that high mean glucose levels and various glycemic abnormalities such as glucose fluctuation and hypoglycemia accelerate the progression of atherosclerosis in patients with type 2 diabetes. Although continuous glucose monitoring (CGM) that could evaluate such glycemic abnormalities has been rapidly adopted, the associations between CGM-derived metrics and arterial stiffness are not entirely clear. Methods This is an exploratory analysis of an ongoing prospective, multicenter, 5-year follow-up observational study. Study participants included 445 outpatients with type 2 diabetes who underwent CGM and brachial-ankle pulse wave velocity (baPWV) measurement at baseline. Associations between CGM-derived metrics and baPWV were investigated using multivariate regression models.Results In a linear regression model, all CGM–derived metrics were significantly associated with baPWV, but HbA1c was not. Some CGM–derived metrics related to intra-day glucose variability, hyperglycemia, and hypoglycemia remained significantly associated with baPWV after adjusting for possible atherosclerotic risk factors, including HbA1c. Based on baPWV ≥1,800 cm/s as indicative of high arterial stiffness, multivariate logistic regression found that some CGM-derived metrics related to intra-day glucose variability and hyperglycemia are significantly associated with high arterial stiffness even after adjusting for possible atherosclerotic risk factors, including HbA1c.Conclusions Multiple CGM-derived metrics are significantly associated with baPWV and high arterial stiffness in patients with type 2 diabetes who have no history of apparent cardiovascular disease. These metrics might be useful for identifying patients at high risk of developing cardiovascular disease.This study has been registered in the University Hospital Medical Information Network Clinical Trials Registry (UMIN000032325).

Importantly, recent clinical studies have suggested that glucose variability is more signi cantly associated with a higher subsequent incidence of myocardial infarction, acute heart failure, and cardiac death than the degree of hyperglycemic exposure indicated by HbA1c levels in patients with both type 2 diabetes and acute myocardial infarction [6]. Thus, evaluating various aspects of glycemic status may help identify patients with a high probability of developing CVD. In this regard, continuous glucose monitoring (CGM) has emerged as an optimal method to obtain a more complete pro le of blood glucose status that includes intra-day and inter-day glucose variations and patterns of hyperglycemia and hypoglycemia.
Recent clinical studies have demonstrated that standard deviation (SD), mean amplitude of glycemic excursion (MAGE), and time in range (TIR), which re ect intra-day glucose variation, are signi cantly associated with carotid artery intima-medial thickening [7,8]. Other studies have demonstrated that MAGE is associated with the presence and severity of coronary artery disease and vascular endothelial function, respectively [9,10].
Increased arterial wall stiffness re ects the state of atherosclerosis. The degree of arterial stiffness does not necessarily re ect arterial intima-media thickness or endothelial function. To assess arterial wall stiffness, brachial-ankle pulse wave velocity (baPWV) is a noninvasive parameter often used clinically. It is useful for evaluating the state of atherosclerosis and predicting CVD in patients with type 2 diabetes [11]. However, the relationship between CGM-derived metrics and arterial stiffness has not yet been fully clari ed.
In this explanatory study, we investigated the relationship between CGM-derived metrics and arterial stiffness in 455 patients with type 2 diabetes mellitus without a history of apparent CVD.

Study Design
This study is an exploratory sub-analysis of an ongoing, observational, prospective cohort study that aims to investigate the relationship between glucose variability evaluated with CGM and the incidence of composite cardiovascular events over a 5-year follow-up period, as described previously [12]. This study used baseline data from the cohort study. This study has been registered in the University Hospital Medical Information Network Clinical Trials Registry (UMIN-CTR), which is a non-pro t organization in Japan that meets the requirements of the International Committee of Medical Journal Editors (UMIN000032325).

Study Population
The study population consists of Japanese patients with type 2 diabetes who regularly attend the outpatient diabetes clinics of 34 institutions shown in Supplementary Table 1. The study design, inclusion criteria, and exclusion criteria were published previously [12]. Brie y, consecutive subjects were screened. Patients who meet the eligibility criteria were asked to participate in the present study. A total of 1,000 outpatients with type 2 diabetes under stable control and with no history of apparent CVD was recruited between May 2018 and March 2019. One patient withdrew consent. Among 999 subjects, 445 who underwent baPWV assessment were included in this exploratory sub-analysis.
The protocol was approved by the institutional review board of each participating institution in compliance with the Declaration of Helsinki and current legal regulations in Japan. Written informed consent was obtained from all participants after a full explanation of the study.

Biochemical Tests
Blood samples were obtained at visits after overnight fasting. Renal function, lipid levels, and HbA1c (National Glycohemoglobin Standardization Program) were measured with standard techniques. Urinary albumin excretion (UAE) was measured using a latex agglutination assay on a spot urine sample. Estimated glomerular ltration rate (eGFR) was calculated using a formula [13].

CGM With The Freestyle Libre Pro Device
The FreeStyle Libre Pro CGM (FLP-CGM) device (Abbott Japan, Tokyo, Japan), which measures glucose levels every 15 minutes for up to 14 days, was used in this study as previously reported [12]. Other than FLP-CGM use, there were no restrictions on participants' daily lives. Downloaded data sets were further analyzed. Glucose variability was assessed based on MAGE [14], SD, and coe cient of variation (CV).
MAGE was calculated as the arithmetic mean of the differences between consecutive peaks and nadirs, provided that the differences are greater than one SD of the mean glucose value. CV (%) was calculated by dividing SD by the mean of the corresponding glucose readings. The original statistical analysis plan (SAP) for this study was reported in the initial study protocol [12]. After the publication of the SAP, the Advanced Technologies & Treatments for Diabetes Congress proposed some CGM-derived metrics as useful clinical targets that complement HbA1c [15]. Thus, we updated the SAP by adding some CGMderived metrics to this study prior to database lock. Mean glucose was calculated from data collected during FLP-CGM use. TIR was de ned as the percentage of time spent in the target range between 3.9 and 10.0 mmol/L (time in range, TIR 3.9-10 mmol/L ), time above target glucose range (TAR > 10 mmol/L , TAR > 13.9 mmol/L ), and time below target glucose range (TBR < 3.9 mmol/L , TBR < 3.0 mmol/L ). Low blood glucose index (LBGI) and high blood glucose index (HBGI) formulae were implemented by converting glucose values into risk scores [16]. In addition, mean of daily differences (MODD) [17] in glucose levels and interquartile range (IQR) were calculated to assess inter-day glucose variability. MODD was calculated as the mean of the absolute difference between glucose levels measured at the same time on 2 consecutive days. IQR was calculated using values from the same time of day during the monitoring period. Since a previous study demonstrated that FLP-CGM was less accurate during the rst 24 hours (from the rst day to the second day) after insertion and during the last four days of its 14-day lifetime [18], we analyzed FLP-CGM data over the middle 8-day period.

Measurement Of BAPWV
At baseline, baPWV was measured using an automatic waveform analyzer (BP-203RPE form; Colin Medical Technology, Komaki, Japan), as described previously [19]. Brie y, measurement was performed in the supine position after 5 min of bed rest. Cuffs for occlusion and monitoring were placed snugly around both arms and both ankles. The pressure waveforms were then recorded simultaneously from the brachial arteries using the oscillometric method. All scans were conducted by well-trained observers at each institution. A previous study con rmed that baPWV measurements have high reproducibility [20]. Subjects with an ankle-brachial index ≤ 0.90 were considered to have peripheral artery disease; baPWV data of these individuals were excluded from this study.

Statistical Analysis
Results are presented as means ± SD for continuous variables or number (proportion) of patients for categorical variables. Comparisons between two groups were analyzed with Student's t-test and Wilcoxon's rank sum test for continuous data or the chi-square test or Fisher's exact test for categorical data as appropriate.
Subjects were categorized into two groups based on the baPWV value of 1,800 cm/s. We used this cutoff value recommended by the Japanese Circulation Society to identify subjects who are at high risk for developing CVD [21]. High arterial stiffness was de ned as baPWV ≥ 1,800 cm/s and low arterial stiffness was de ned as baPWV < 1,800 cm/s. Multivariate linear regression analysis was performed to investigate whether FLP-CGM-derived metrics are associated with baPWV when it was treated as continuous variable. In addition, multivariate logistic regression analysis was performed to investigate whether FLP-CGM-derived metrics are associated with high baPWV when it was treated as a dichotomous variable. Potential conventional risk factors evaluated with clinical, biochemical, or metabolic testing were included in the models based on clinical judgment. All statistical tests were two-sided with a 5% signi cance level. All analyses were performed using SAS software version 9.4 (SAS Institute, Cary, NC).

Baseline Clinical Characteristics
The baseline clinical characteristics of the 445 patients with type 2 diabetes are summarized in Table 1. Mean age was 65.9 ± 9.0 years, 67.6% were male, HbA1c was 7.08 ± 0.78% (53.9 ± 8.6 mmol/mol), and estimated duration of type 2 diabetes was 13.5 ± 8.3 years. Data are mean ± SD or number of patients (%).

Relationship Between FLP-CGM-derived Metrics and baPWV
We investigated the relationship between FLP-CGM-derived metrics and baPaWV in patients with type 2 diabetes when baPWV was treated as continuous variable. In a linear regression model, all calculated FLP-CGM-derived metrics were signi cantly associated with baPWV, although no signi cant association was observed between HbA1c and baPWV (Model 1 in Table 2). In Model 2, which adjusted for age and gender, all the aforementioned associations remained signi cant. After adjusting for variables in Model 2 plus body mass index (BMI) and duration of diabetes (Model 3), the associations between FLP-CGMderived metrics and baPWV remained signi cant, except for LBGI. Even after adjusting for variables in Model 3 plus HbA1c, systolic blood pressure (BP), lipid parameters, uric acid, eGFR, UAE, smoking, alcohol consumption, use of insulin therapy, use of angiotensin-converting enzyme (ACE) inhibitors and/or angiotensin II receptor blockers (ARBs), use of statins, use of anti-platelet agents, and presence of diabetic retinopathy (Model 4), the associations between the FLP-CGM-derived metrics of mean glucose, SD, CV, MAGE, TIR 3.9-10 mmol/L , TAR > 10 mmol/L , TAR > 13.9 mmol/L , TBR < 3.9 mmol/L , TBR < 3.0 mmol/L , and HBGI and baPWV remained signi cant (Table 2). Next, 445 subjects were divided into a high arterial stiffness group (n = 149) and a low arterial stiffness group (n = 296) based on the cut-off baPWV value of 1,800 cm/s. The clinical characteristics of the high and low arterial stiffness groups are summarized in Table 3. Subjects with high arterial stiffness were older, had longer duration of diabetes, lower BMI, higher systolic BP, higher UAE, higher prevalence of diabetic retinopathy, and lower eGFR. In addition, they were more frequently treated with sulfonylureas, dipeptidyl peptidase-4 inhibitors, insulin, and calcium channel blockers. There were signi cant differences in all FLP-CGM-derived metrics except for TBR < 3.9 mmol/L , TBR < 3.0 mmol/L , and LBGI between the two groups. Using this classi cation of baPWV, we investigated the relationship between FLP-CGM-derived metrics and high baPWV. In a multivariate logistic regression model, all FLP-CGM-derived metrics except for LBGI were signi cantly associated with high arterial stiffness, although no signi cant association was observed between HbA1c and arterial stiffness (Model 1 in Table 4). In Models 2 and 3, all FLP-CGMderived metrics except for LBGI remained signi cantly associated with high baPWV. In Model 4, SD, CV, MAGE, TAR > 13.9 mmol/L , and HBGI remained signi cantly associated with high baPWV.

Discussion
In this study, we demonstrated that most FLP-CGM-derived metrics are signi cantly associated with arterial stiffness, even after adjusting for various risk factors including HbA1c levels in 445 outpatients with type 2 diabetes and no history of apparent CVD. Notably, some FLP-CGM-derived metrics related to inter-day glucose variability and hyperglycemia identi ed subjects with high arterial stiffness who were at high risk for developing CVD.
In patients with type 2 diabetes, the main cause of arterial stiffness is damage to vascular walls caused by prolonged hyperglycemia. In fact, previous cross-sectional studies have indicated that higher HbA1c levels are associated with increased arterial stiffness in patients with type 2 diabetes [22,23]. In contrast, HbA1c was not associated with arterial stiffness in this study, a nding that was consistent with a different previous study [24]. On the other hand, this study found signi cant associations between FLP-CGM derived metrics related to hyperglycemia such as TAR > 13.9 mmol/L and HBGI. Unlike HbA1c, these parameters re ect remarkable hyperglycemia and were not affected by hypoglycemia. Thus, our data do not contradict the fact that the main cause of arterial stiffness is damage to vascular walls caused by prolonged hyperglycemia in patients with type 2 diabetes.
Previous cross-sectional studies demonstrated that CGM-derived metrics related to intra-day glucose variability are associated with carotid atherosclerosis, coronary atherosclerosis, or endothelial dysfunction in patients with type 2 diabetes [7][8][9][10]. This study also found that SD, CV, and MAGE are signi cantly associated with arterial stiffness; these variables re ect another aspect of atherosclerosis in patients with type 2 diabetes. Various factors are involved in the progression of arterial stiffness in patients with type 2 diabetes. Indeed, previous cross-sectional studies have demonstrated that conventional atherosclerotic risk factors such as age, BMI, duration of type 2 diabetes, glycemic control, dyslipidemia, systolic BP, eGFR, elevated uric acid levels, and albuminuria [25][26][27] are associated with arterial stiffness in patients with type 2 diabetes. Intriguingly, in this study FLP-CGM-derived metrics related to intra-day glucose variability were signi cantly associated with degree of arterial stiffness, even after adjusting for those possible conventional atherosclerotic risk factors. Taken together, this study highlighted the importance of intra-day glucose variability in terms of assessing the risk of arterial stiffness.
A recent study demonstrated that the incremental glucose peak, an indicator of glucose variability, during an oral glucose tolerance test is associated with arterial stiffness in patients with type 2 diabetes [28].
However, whether oral glucose tolerance test-derived incremental glucose peak re ects the pattern of glucose variability during usual living conditions has not yet been clari ed. In this regard, this is the rst study to provide evidence that FLP-CGM-derived metrics related to intra-day glucose variability evaluated during usual living conditions are signi cantly associated with arterial stiffness. Although the exact mechanism of how glucose variability contributes to arterial stiffness remains unclear, we propose the following possible scenarios. Previous studies have shown that glucose variability induces in ammation and overproduction of oxidative stress to a greater extent than chronic persistent hyperglycemia [29,30], leading to advanced AGE formation. The formation of AGEs is considered to be involved in arterial stiffness through cross-linking of collagen molecules and a subsequent loss of collagen elasticity [31].
Accordingly, vascular walls may be damaged by glucose variability more than by chronic persistent hyperglycemia.
In this study, FLP-CGM-derived metrics related to hypoglycemia were associated with arterial stiffness. Similarly, some recent studies have demonstrated that the acute effects of hypoglycemia include in ammation and endothelial injury in patients with type 1 diabetes [32,33]. In addition, a cross-sectional study demonstrated that repeated episodes of hypoglycemia are associated with preclinical atherosclerosis as evaluated by carotid and femoral ultrasonography and measurement of ow-mediated brachial dilatation in patients with type 1 diabetes [34]. Furthermore, we previously reported that a higher frequency of hypoglycemic episodes is associated with progression of carotid atherosclerosis in patients with type 2 diabetes [35]. Accordingly, physicians need to help prevent hypoglycemic episodes through assessing the risks of hypoglycemia with CGM in order to minimize arterial stiffness, especially in patients who have di culty detecting or who are completely unaware of hypoglycemia.
As discussed above, this study demonstrated that FLP-CGM-derived metrics are signi cantly associated with arterial stiffness. On the other hand, it is more important to screen for patients with a high potential of developing CVD in order to reduce the incidence of CVD. To achieve this goal, the Japanese Circulation Society proposed baPWV of 1,800 cm/sec as the cutoff value to identify subjects who are at high risk for developing CVD based on the results of several studies [36]. Based on this cutoff value, approximately 33% of study participants were de ned as being at high risk for CVD. Subjects with high arterial stiffness were older, had longer duration of diabetes, lower BMI, higher systolic BP, higher UAE, higher prevalence of diabetic retinopathy, and lower eGFR. They were more frequently treated with insulin and calcium channel blockers. Even after adjusting for these confounding factors, FLP-CGM-derived metrics related to intraday glucose variability such as SD, CV, MAGE, and metrics related to hyperglycemia such as TAR > 13.9 mmol/L and HBGI were signi cantly associated with high arterial stiffness. Given that FLP-CGM-derived metrics related to hypoglycemia, such as TIR 3.9-10 mmol/L and TAR > 10 mmol/L were not associated with high arterial stiffness, high postprandial glucose excursion amplitude may be a major contributor to increased arterial stiffness. Thus, based on FLP-CGM-derived metrics, focusing on reducing the amplitude of postprandial glucose excursion may be important to reducing the risk of both arterial stiffness and CVD development.
One strength of this study is its multicenter study design. Our study had certain limitations. First, the study was an exploratory study with a relatively small sample size. Second, the cross-sectional study design made it impossible to evaluate whether there was a causal relationship between FLP-CGMderived metrics and arterial stiffness. We only used arterial stiffness as a marker for the risk of CVD development. In this regard, we are currently conducting a long-term follow-up study in the same cohort that focuses on FLP-CGM-derived metrics and onset of primary CVD or changes in arterial stiffness.
Third, FLP-CGM-derived metrics were evaluated based on FLP-CGM measurements during a limited time. Due to this limitation, FLP-CGM-derived metrics related to inter-day glucose variability may be not associated with arterial stiffness after adjusting for various atherosclerotic risk factors. Repeated FLP-CGM measurements may be required to clarify the relationship between inter-day glucose variability and arterial stiffness. Fourth, we only recruited Japanese patients with type 2 diabetes. These constraints may limit the generalizability of our results.

Conclusion
In this study, we demonstrated that FLP-CGM-derived metrics related to intra-day glucose variability, hyperglycemia, and hypoglycemia are signi cantly associated with arterial stiffness, even after adjusting for various risk factors in patients with type 2 diabetes with no history of apparent CVD. In addition, based on FLP-CGM-derived metrics, high postprandial glucose excursion amplitude can identify subjects who are at high risk for developing CVD. Thus, these metrics could provide medical professionals with useful information for assessing the risk of CVD. Declarations