Individualized coefficient of variability cut-off values for reducing the risk of hypoglycemia in Chinese type 1 diabetes mellitus (T1DM) patients

To the Editor: Type 1 diabetes mellitus (T1DM) is a progressive disease caused by severe islet β-cell function impairment, resulting in high glucose variability (GV) in patients.[1] While intensive insulin treatment enabled T1DM patients to achieve near-normal glucose, reflected by reduced hemoglobin A1c (HbA1c) levels, stricter HbA1c targets were associated with an increased risk of hypoglycemia, which would compromise patients' quality of life and survival.[2] Recent studies suggest that GV reflects hypoglycemia better than HbA1c alone, and continuous glucose monitoring (CGM) provides a more comprehensive glycemic profile for better glucose indices assessment.[3] Therefore, we aimed to explore the relationship between the coefficient of variability (CV) of blood glucose and hypoglycemia in T1DM patients with different HbA1c levels and to identify the optimal cut-off values of CV for reducing the hypoglycemia risk in Chinese T1DM patients. Inclusion criteria was shown in Supplementary Material, https://links.lww.com/CM9/B600. The study was approved by the Ethics Review Committee of the Second Xiangya Hospital of Central South University (No. 2019-198). All participants provided signed informed consent. Ultimately, 517 participants were enrolled. The details of their clinical characteristics, biochemical indices, and CGM data were collected and had been described previously.[4] The study flow chart was shown in Supplementary Figure 1, https://links.lww.com/CM9/B600. The targets were set according to the international consensus guidelines on CGM: time in range (TIR, 3.9–10.0 mmol/L) ≥70%, time above range (TAR, >10.0 mmol/L) <25%, time below range (TBR, <3.9 mmol/L) <4%, and CV <36%; target HbA1c levels <7.0% for adults and <7.5% for minors are recommended.[5] Statistical methods were provided in Supplementary Material, https://links.lww.com/CM9/B600. According to the Chinese guidelines for HbA1c targets, participants were categorized into two groups: low-HbA1c (L-HbA1c) group (n = 169, median [Q1, Q3]: 6.5% [6.1%, 6.9%]) and high-HbA1c (H-HbA1c) group (n = 348, 8.8% [7.8%, 10.3%]). Analysis of the CGM-derived metrics showed that the L-HbA1c group had lower estimated HbA1c (eHbA1c) levels, time above range (TAR, >10.0 mmol/L), mean blood glucose (MBG), standard deviation (SD), mean amplitude of glycemic excursions (MAGE), largest amplitude of a glycemic excursion (LAGE), CV, and high blood glucose indices (HBGI) (all P <0.001), but higher proportions of TAR <25%, TIR ≥70%, and TBR <3.9 mmol/L (all P <0.001) than the H-HbA1c group. However, there was no difference in the proportion of TBR <4% (P = 0.139) or LBGI (P = 0.246), suggesting that even after adjusting for HbA1c levels, intensively treated individuals still had an increased risk of hypoglycemia, this finding implied that factors other than HbA1c played a significant role [Supplementary Table 1, https://links.lww.com/CM9/B600]. We further explored the role of CV in hypoglycemia risk by dividing all participants into four groups based on the achievement of the glucose CV threshold (<36%) and the HbA1c goals (adults <7% and minors <7.5%). The groups were L-CV/L-HbA1c, L-CV/H-HbA1c, H-CV/L-HbA1c, and H-CV/H-HbA1c. Then, we stratified the analysis according to CV within the two HbA1c groups [Supplementary Table 2, https://links.lww.com/CM9/B600]. Results showed that CV was correlated closely with measures of hypoglycemia in both the L-HbA1c and H-HbA1c groups. We then investigated the correlation between multiple independent variables and TBR using logistic regression [Supplementary Table 3 and Figure 1, https://links.lww.com/CM9/B600]. Results showed that both CV and HbA1c were positively correlated with TBR (all P <0.001), while other variables, including age, diabetes duration, and sex, showed no significant correlation with TBR (all P >0.05) and were therefore not included in further analyses. Recent research has demonstrated that the threshold for CV varies in different populations with diabetes, and these differences may be related to the population, type of diabetes, and treatment. Therefore, we aimed to select a more suitable CV as the achievement goal for further exploration of CGM-associated target achievement, especially for TBR <4% in T1DM patients. We used receiver operating characteristic (ROC) analysis to calculate CGM-related indices in the L-HbA1c and H-HbA1c groups [Figure 1A]. Our results indicated that CV was the best discriminator of hypoglycemia as detected by CGM in T1DM patients (area under the curve [AUC] 0.845, 95% confidence interval [CI] = 0.785–0.905; AUC 0.815, 95% CI = 0.766–0.864, respectively, for L-HbA1c and H-HbA1c), followed by MBG, TIR, and HbA1c. Although there was a slight increase in AUC for CV combined with HbA1c in both groups, no significant difference in diagnostic efficacy was observed in either group.Figure 1: (A) Cut-off values for the prediction of TBR <4% based on ROC analysis. (B) Comparisons of selected CGM parameters in patients with lower HbA1c stratified by CV 35%. *P values <0.001 for the CV ≤35% group vs. the CV >35% group. AUC: Area under the curve; CGM: Continuous glucose monitoring; CV: Coefficient of variation; FPR: False positive rate; LBGI: Low blood glucose index; MBG: Mean blood glucose index; ROC: Receiver operating characteristic; TAR: Time above range; TBR: Time below range; TIR: Time in range; TPR: True positive rate.Furthermore, we explored the optimal CV cut-off values for assessing TBR <4% among participants whose HbA1c level was on target or not. A cut-off point of 35% was the optimal CV value for participants who achieved their target HbA1c, with a sensitivity of 70% and a specificity of 82%. The optimal CV cut-off in the HbA1c non-target group was 36%, with a sensitivity of 88% and a specificity of 64%, which was in line with international recommendations [Supplementary Table 4, https://links.lww.com/CM9/B600]. Finally, we stratified patients whose HbA1c levels were on target (n = 169) using a CV cut-off of 35% and compared CGM parameters (especially those related to hypoglycemia risk) between the two groups, CV ≤35% (n = 100) and >35% (n = 69). Significant differences were observed in TIR, TAR, and hypoglycemic metrics. The group with CV ≤35% demonstrated significantly higher TIR (85.29% vs. 64.30%, Z = –7.286, P <0.001) and lower hypoglycemic metrics—including TBR (2.32% vs. 5.90%, Z = –5.797, P <0.001) and LBGI (2.3% vs. 4.3%,Z = –6.392, P <0.001)—than the group with CV >35%, as well as lower TAR (8.41% vs. 25.66%, Z = –5.858, P <0.001). Additionally, more participants achieved the clinical targets of TIR ≥70% (82.0% vs. 49.3%, χ2 =24.096, P <0.001), TBR <4% (85.0% vs. 40.6%, χ2 =41.527, P <0.001), and TAR <25% (79.0% vs. 27.5%, χ2 =53.071, P <0.001) in the CV ≤35% group than in the CV >35% group [Figure 1B]. This study reveled that T1DM patients with HbA1c levels within clinical goals still had unsatisfactory TBR and a high rate of TBR<4%. Lower HbA1c did not reliably indicate a lower risk of hypoglycemia, as HbA1c alone explained only a small portion of the hypoglycemia risk. Given that patients who did not meet the HbA1c goal mostly had higher CV values (>36%), we hypothesized that achieving a target CV of 36% may modify the relationship between HbA1c and hypoglycemia in T1DM patients. This study showed that the patients who achieve the target of CV <36% had a significantly reduced risk of hypoglycemia, CV seemd to be a more appropriate indicator for assessing hypoglycemia risk in T1DM patients, particularly those with low HbA1c levels, independent of whether the HbA1c goal was achieved. The study also demonstrated that CV positively correlated with TBR and TBR <4% in T1DM patients. Furthermore, this study revealed that 35% was the optimal CV cut-off for participants who achieved target HbA1c levels, achieving a target of CV <36% may be easier than achieving the HbA1c target for T1DM patients included in the study (53.6% vs. 32.6%) and that reduced TBR and increased rates of TBR <4% were more pronounced in individuals meeting the more stringent target of CV <35%. Overall, our findings suggest that a stricter CV target may be more beneficial for individualized blood glucose control in T1DM patients with hypoglycemia risk than a more stringent HbA1c target alone. However, our conclusions should be confirmed through prospective studies with longer CGM duration, and it is essential to include T1DM patients with diabetic complications to provide evidence that a stricter CV target would slow the development of diabetic complications in those who achieved the HbA1c target. Funding This work was supported by a grant from the National Key Research and Development Program of China (No. 2018YFC2001005). Conflicts of interest None.