Relationship between glycemic intraday variations evaluated in continuous glucose monitoring and HbA1c variability in type 2 diabetes: pilot study

Background HbA1c variability is independent of mean HbA1c, and associated with mortality due to vascular complications. However, the significance of HbA1c variability is unknown at present. In this study, we used flash glucose monitoring (FGM) and evaluated glycemic intraday variations, and then examined the association with HbA1c variability. Methods We conducted a retrospective pilot study of 26 patients treated at the Outpatient department for type 2 diabetes mellitus (T2DM), and evaluated the following items associated with blood glucose levels and their changes/variations using FGM. The primary endpoint was factor(s) associated with standard deviation (SD) HbA1c over a 6-month period. To adjust for the effect of varying numbers of HbA1c measurements, we used the adjusted SD of HbA1c. Results There were significant correlations between mean HbA1c and each of glucose management indicator, maximum, percent time at glucose > 180 mg/day, mean of daily difference of blood glucose, and high blood glucose index. Adjusted SD HbA1c correlated significantly with percent time at glucose < 70 mg/dL and low blood glucose index. We estimated the regression coefficient of adjusted SD HbA1c using multivariate linear regression analysis, and noted that the presence of hypoglycemia affected Adjusted SD HbA1c (β = 0.130, SE = 0.044, P = 0.008). Hypoglycemia was noted in 17 patients, and adjusted SD HbA1c was significantly higher (p = 0.001) in the hypoglycemic group (0.22 ± 0.12%), compared with the non-hypoglycemic group (0.08 ± 0.05%). The cut-off value of adjusted SD HbA1c was 0.109% in the hypoglycemic group. Conclusions The results showed that HbA1c variability is associated with hypoglycemia. For patients with high HbA1c variability, we recommend evaluation for the presence of hypoglycemia and reconsideration of their treatment regimen including their glucose-lowering medications. Trial registration The study protocol and opt-out method of informed consent were approved by the ethics committees of the University of Occupational and Environmental Health (Trial registration: H27-186, Registered 25 Dec 2015) Supplementary Information The online version contains supplementary material available at 10.1186/s13098-021-00663-2.

Background One of the objectives of treatment of diabetes is prevention of future diabetes-related complications through glycemic control. HbA1c is the gold standard index used for monitoring the effectiveness of treatment of diabetes.
HbA1c reflects blood glucose level over the previous 3 months, and it is not affected by short periods of blood glucose variations, such as changes following meals and exercise [1,2]. In this regard, Continuous Glucose Monitoring (CGM) and Flash Glucose Monitoring (FGM) have been recommended for managing more specific glucose levels, and the time in range (TIR) and blood glucose levels of 180 or above and under 70 have become international standards for monitoring the response to diabetes treatment [3]. While CGM is a useful supplementary device, it is difficult to use in all patients during routine clinical practice. Clinical evidence indicates that keeping HbA1c < 7% is beneficial as it prevents the development of microangiopathy. The Diabetes Control and Complications Trial (DCCT) group has reported that the use of HbA1c as an index of glycemic control slows down the progression of diabetic retinopathy, nephropathy and neuropathy [4]. Furthermore, the effects of intensive glucose-lowering therapy on blood pressure and albumin excretion were still evident at 7 to 8 years after the end of the DCCT study; in the intensive therapy group, the odds ratio was 59% for the prevention of nephropathy and 84% for the prevention of progression to overt albuminuria. In this regard, a previous study of DCCT with 22-year follow-up reported 50% risk reduction in GFR (< 60 mL/min/1.73 m 2 ) in the intensive therapy group, coupled with significant reduction in the GFR decline rate [5,6]. Furthermore, the United Kingdom Prospective Diabetes Study (UKPDS) showed that for each 1% drop in HbA1c, the risk of microvascular complications reduces by 37% [7]. These trials verified the importance of elevated HbA1c as an indicator of chronic hyperglycemia. Another detailed blood glucose profile study using CGM concluded that HbA1c reflects the mean blood glucose level [8].
The ADVANCE trial [9] examined the effect of glycemic control, using HbA1c as the index parameter, and reported that intensive glucose-lowering therapy was not associated with a significant reduction in macrovascular disorders. Furthermore, the ACCORD trial concluded that intensive therapy was associated with significantly higher all-cause mortality and cardiovascular mortality hazard ratios [10]. These studies suggest that serious hypoglycemia linked to intensive therapy may underlie the increased risk of mortality. In this regard, recent studies have reported that large swings in blood glucose levels are associated with worsening of arteriosclerosis [11] and increased risk of mortality due to cardiovascular events [12]. Apart from blood glucose levels, the correlation between HbA1c variability, in addition to mean HbA1c, and mortality due to diabetic vasculopathies has also been analyzed [13][14][15]. However, to date, the true significance of HbA1c variability remains unknown.
We used the FGM system to examine the relationship between HbA1c variability and detailed glycemic profiles. The primary endpoint was the identification of background parameters and CGM indices that are associated with adjusted SD HbA1c. The latter term represents the variability of HbA1c over a 6-month period. The secondary endpoint was the identification of background parameters and CGM indices that are associated with the mean HbA1c over the 6-month period.

Subjects
We conducted a retrospective study from September 2018 to January 2019 at the Outpatient Clinic of the University of Occupational Medicine Hospital and University of Occupational Medicine Wakamatsu Hospital. The subjects were 26 patients with type 2 diabetic mellitus (T2DM), who had been on the FGMS ® System (FreeStyle Libre Pro System, Abbott Diabetes Care, Inc.) for at least eight days for evaluation of hemodynamics. The following inclusion criteria were applied in this study: (1) age between 30 and 80 years at the time of obtaining consent; (2) T2DM treated and followed-up at the Outpatients Clinic; (3) no changes (addition, switching, or discontinuation) to the glucose-lowering agents or their doses within the 4-week period before the commencement of the FGMS; (4) no changes to the T2DM treatment up to 6 months after the start of FGM. The following exclusion criteria were also applied. (1) Type 1 or secondary diabetes mellitus; (2) severe infection, before-or after surgery, or serious trauma; (3) renal dialysis; (4) severe hepatic dysfunction (AST ≥ 100 IU/L or ALT ≥ 100); (5) moderate or severe heart failure (NYHA/New York Heart Association Classification III or higher stage); (6) pregnant, lactating, or potentially pregnant patients; (7) treated with steroids or other drugs that affect blood glucose levels; and 8) history of macrovascular diseases. The following definitions were used for diabetic microangiopathies. Diabetic neuropathy was diagnosed by the presence of two or more clinical symptoms (bilateral spontaneous pain, hypoesthesia or paresthesia of the legs), absence of ankle tendon reflexes and decreased vibration sensations using a C64 tuning fork. Diabetic retinopathy was defined as mild or severe retinopathy based on the results of funduscopic examination by ophthalmologists. Diabetic nephropathy was defined as albumin-to-creatinine ratio ≥ 30 mg/g creatinine.
The study protocol and opt-out method of informed consent were approved by the ethics committees of the University of Occupational and Environmental Health (Trial registration: H27-186, Registered 25 Dec 2015).

Flash glucose monitoring system
The following parameters were measured using the data recorded by the FGM: average glucose level (AG), standard deviation (SD), coefficient of variation (CV), percent time at glucose level of 70-180 mg/dL (TIR: time in range), percent time at > 180 mg/dL (TAR: time above range), percent time < 70 mg/dL (TBR: time below range), maximum, minimum, glucose management indicator (GMI) [16], mean daily difference of blood glucose (MODD), low blood glucose index (LBGI) and high blood glucose index (HBGI) [17]. Hypoglycemia was defined as a glucose value less than 70 mg/dL as recorded by FGM. FGM was applied once at the start of the study and used for up to 14 days. The FGM data were recorded to avoid bias due to the insertion and removal of the FGM, or lack of stability of the unit. Since the MARD, which represents the accuracy of the sensor in FGM, is higher on the first day [18], and since most patients followed weekly routines in their daily living, we used data obtained from the second to seven days, and excluded those of the first day. We recorded the daily average value, and listed the average value for seven days. The mean percentage of time during which the FGM was active for all 26 cases was 97.4%, and the analysis included no missing values over the 7-day period.

Laboratory tests
HbA1c (%) was measured by HPLC using Tosoh HLC-723 G8 (Tosoh Co., Kyoto, Japan) and recorded as a NGSP (National Glycohemoglobin Standardization Program) value. eGFR (estimated glomerular filtration rate) was calculated as 194 × serum creatinine concentration (mg/dL) − 1.094 × age − 0.287 for men, and 194 × serum creatinine concentration (mg/dL) − 1.094 × age − 0.287 × 0.739 for women [19]. The SD HbA1c represents the SD of 3-7 HbA1c readings over a 6-month period from the time of starting the use of FGM. The SD HbA1c represents the SD of 3-7 HbA1c readings over a 6-month period from the time of starting FGM use. The median number of HbA1c measurements was 4.5. To adjust for the effect of different numbers of HbA1c measurements, we used in this study the adjusted SD of HbA1c, representing the SD of HbA1c divided by [n/(n-1)] 0.5 , where n is the number of HbA1c measurements [20]. On the other hand, the mean HbA1c reported in this study is the mean of the same measurements.

Statistical analysis
Continuous values are shown as mean values and categorical variables are expressed as count and percentage values. The Shapiro-Wilk test was used to test for normality; for statistical significance of the mean values of two groups, the Student's t-test was used if equal variance was confirmed by the F test, whereas Welch's t-test was used if it followed normal distribution. The Mann-Whitney U test was used if it did not follow a normal distribution, and Spearman's correlation analysis was used for testing the relationship between two variables. Univariate and multivariate linear regression analyses were used to estimate the regression coefficients for adjusted SD HbA1c. Multivariate analysis was performed with adjusted SD HbA1c as the dependent variable, and age, BMI, and presence of hypoglycemia as independent variables. Dummy variables were created for gender and presence of hypoglycemia, and then used in multiple regression analysis. The cutoff value of adjusted SD HbA1c was examined based on the ROC curve. The calculated sample size in the ROC analysis was 30 patients in total, assuming an area under the curve (AUC) of 0.80, power 0.80, with significance level of 5%, and a nonhypoglycemic group versus the hypoglycemic group of 2:1. A p value < 0.05 was considered significant. All analyses were performed using SPSS Statistical Software 25.0 (SPSS Inc., Chicago, IL).

Hypoglycemia versus non-hypoglycemia
We examined the difference between adjusted SD HbA1c and mean HbA1c with and without hypoglycemia (Fig. 2a, b). Adjusted SD HbA1c was significantly higher in the hypoglycemic group (p = 0.001) at 0.22 ± 0.12%, compared to the non-hypoglycemic group at 0.08 ± 0.05%. On the other hand, there was no significant difference in mean HbA1c between with or without hypoglycemia.
In order to evaluate the cutoff of adjusted SD HbA1c in the hypoglycemic group, a ROC curve was constructed and the area under the curved surface was calculated. ROC curve analysis showed a cutoff value of SD HbA1c  Table 2 Correlation with adjusted SD HbA1c and mean HbA1c The   Fig. 2c).

Discussion
We conducted a pilot study on HbA1c variability and CGM indexes, and showed that adjusted SD HbA1c is associated with the hypoglycemic indexes of TBR and LBGI. Although the association between HbA1c and CGM indexes has been analyzed previously [21,22], this is the first study that dissected the association between HbA1c variability and CGM indexes.
Variations in blood glucose levels are known to be involved in the progression of diabetes vasculopathies, and several studies reported the association of fasting blood glucose variability and postprandial blood glucose with vascular complications.
HbA1c variability has also attracted attention in recent years, with SD and CV used as scales of HbA1c variability. Bouchi et al. [13] reported that SD HbA1c, which is independent of the common cardiovascular risk factors, is associated with the risk of onset of cardiovascular disease (CVD) in T2DM patients. Hirakawa et al. [14] reported that high SD HbA1c is associated with increased risk of onset of vascular events and increased mortality rate in patients of the ADVANCE trial intensive therapy group. Furthermore, Orsi et al. [15] reported that HbA1c variability in T2DM patients is a strong independent predictor of all-cause mortality in T2DM. Although these reports have shown that HbA1c variability is associated with mortality due to diabetes vasculopathies, the significance of HbA1c variability remains unknown. This study is the first to show that HbA1c variability is associated with CGM-based hypoglycemic indexes. Since hypoglycemia is known to be associated with diabetic vascular complications and related mortality risk, future studies are expected to report on the association of HbA1c variability with risk of cardiovascular mortality.
The objective of glycemic control is to normalize blood glucose level. The United Kingdom Prospective Diabetes Study (UKPDS) [23] obtained data that confirmed the so-called legacy effect associated with intensive glycemic control; therefore, if that state can be maintained, then a long-term prognosis can be expected. On the other hand, intensive glycemic control also increases the risk of hypoglycemia. Particularly severe hypoglycemia must be avoided, as it is associated with poor prognosis. HbA1c is a useful index for evaluating hyperglycemia; however, since HbA1c cannot be used to evaluate hypoglycemia, it is recommended to set a lower limit blood glucose for HbA1c according to the conditions and risk of hypoglycemia [24,25]. Our study showed that high HbA1c variability increases the risk of hypoglycemia, and provided the cutoff value for adjusted SD HbA1c, above which the risk of hypoglycemia is increased. Our results call for evaluation of hypoglycemia in patients with high SD HbA1c.
As mentioned previously [4,7], correction of HbA1c is required in order to prevent microangiopathies. In agreement with previous studies, our results showed that the mean HbA1c was associated with the CGM