In the current study, 137 (39.0%) of the total 351 outpatients with T2DM experienced AH episodes, in which Level 1 AH and Level 2 AH accounted for 61.3% and 38.7%, respectively. 85 (62.0%) of the AH patients experienced NAH and 25 (18.2%) exclusively NAH. Multivariate logistic regression analysis demonstrated that patients with younger age, lower HbA1c, and higher SBP levels were significantly associated with increased risk of AH. While after further grading of AH, male sex and DPP4i regime were shown to be associated with lower risk of Level 2 AH.
Previously, Chico et al.  detected unrecognized hypoglycemias in 46.6% (14 out of 30) of the type 2 patients with diabetes by using CGMS and noticed that 73.7% of all events occurred at night, tremendously provoking attention from academia. Further, Gehlaut et al.  applied CGMS on 108 non-hospitalized patients with T2DM to detect the episodes of hypoglycemia. During the five-day study, 53 (49.1%) participants were captured of hypoglycemia by CGMS, but only 24.5% (13 out of 53) had self-reported symptoms, while the majority (75%) of patients were unaware of any hypoglycemia episodes. The prevalence of NAH was 73.6% (37 out of 53), and hypoglycemia occurred only at night was 20.8% (14 out of 53). In our study, none of the participants reported symptomatic hypoglycemia, and 39.0% of the target patients experienced AH events, which was in parallel with previous studies. Notably, all patients experienced hypoglycemic episodes claimed they had no symptoms of hypoglycemia. We speculate that the heterogeneity of symptom response in diabetic individuals could took into consideration and selective bias might play a role. It is also possible that the patients were asymptomatic due to recurrent episodes of hypoglycemia.
Although plenty of studies [12, 17, 22, 23, 25] had indicated that older age, longer diabetes duration, and microvascular complications were independently correlated with increased prevalence of hypoglycemia, in our present study, multivariate logistic regression analysis found no significance of these variables in predicting AH events except for younger age in the Level 2 AH group. Various explanations could be proposed. Firstly, most prior studies were conducted during hospitalization, and they focused merely on the insulin-used population with poor glucose control to investigate the incidence of hypoglycemia. However, the target population of our study was T2DM outpatients whose status was relatively stable with "well-controlled" glucose levels. Differences in samples may contribute to differences of results. Secondly, we used the CGMS to detect the hypoglycemia episodes, which is expected to discover additional unrecognized hypoglycemia events than traditional SMBG.
Previous studies [24, 28] had demonstrated that intensive insulin treatment or sulfonylurea (SU) could predict the episodes of hypoglycemia in T2DM. Our study also showed a higher rate of insulin therapy (52.8% vs. 36.9%) in the Level 2 AH group than in the non-AH group. The univariate analysis (Additional file 1: Table S1) indicated that patients with the use of insulin were associated with higher risk of Level 2 AH (P = 0.018). However, this association did not exist when accounting for other influencing factors in the multiple logistic regression models. This may be due to the substantial effect of the adjusted confounders. Additionally, neither univariate nor multivariate analysis demonstrated the association between SU and the risk of AH. Overall, our study observed that outpatients treated with insulin or SU did not predispose to asymptomatic hypoglycemia, which was consistent with the conclusion of Monnier et al. [22, 29]. This interesting finding reminds us to screen all diabetic patients for a history of hypoglycemia, rather than focusing only on those using insulin or insulin secretagogues.
In our study, the univariate logistic analysis indicated that DPP4i regime was associated with lower risks of Level 1 and Level 2 AH, yet the association between DPP4i and Level 1 AH was substantially attenuated after further adjusting for other confounders. This may stem from the fact that, in our present study, elder age was a protective factor for AH, and the average age of the Level 1 AH group was substantially higher than that of the Level 2 AH group (55.3 ± 12.9 vs. 47.0 ± 15.8). Thus, after controlling for age and other affecting factors, the protective effect of DPP4i in the Level 1 group disappeared. Overall, DPP4i have been proved to be substantially associated with lower risks of hypoglycemia . Further studies are needed to confirm the causal relationship between DPP-4i regime and the risk of AH.
The link between the diagnosis of hypertension and a higher risk of hypoglycemia had been identified in previous studies [27, 28]. In the present study, 12.8% of the subjects had combined hypertension. Multifactorial analysis showed that higher SBP level was predisposed to AH episodes despite the well-controlled blood pressure levels (125.5 ± 75.9 mmHg) of participants. Thus, the role that strict control of SBP may play in patients who were vulnerable to AH needs further attention. Additionally, our preliminary analysis did not observe a significant association between gender and AH, but when further dividing AH episodes into Level 1 and Level 2 AH, we noticed that women are prone to Level 2 AH, which was in accordance with a prospective study by Zhang et al. . The exact reason was unclear. As such, a larger sample size is needed in the future to verify the relationship between gender and AH.
Prior research found that lower HbA1c was an independent predictor of AH [28, 30, 31]; this was also highly salient in the current cohort. Currently, HbA1c is known to be the most used parameter to assess glycemic control and important index in the treatment of hyperglycemia, which has been used as the primary endpoint for many CGM studies . However, new data support the need to devote attention to TIR and TAR, for a comprehensive evaluation of glycemic control among the diabetes population [12, 23, 32]. Indeed, some clinicians may choose to target the reduction of the TAR and minimize hypoglycemia, thereby arriving at more time in the target range. Whereas, recent recommendations from an international consensus on TIR emphasized that targets should be individualized and each 5% increase in TIR correlated with clinically meaningful benefits . For the present study, in terms of lower HbA1c levels and higher percentages of TIR, the AH group seemed to had better blood glucose control than the non-AH group, but when integrated with MAGE, SD, and TAR, the higher glucose variability (GV) of the AH group is noteworthy. Accordingly, given the substantially high risk of AH in T2DM outpatients, clinicians should be vigilant in preventing hypoglycemia in such populations and avoid aggressively attempting to achieve near-normal glucose or HbA1c levels in whom the risks of lower glycemic targets may far outweigh the potential benefits on diabetic complications. So far, SD, MAGE, and MBG measured by CGMS, as key metrics for GV, has been confirmed to be closely correlated with hypoglycemia events [22, 33,34,35]. Providing unprecedented access to a range of new indicators of glucose control, CGMS, which would help clinicians and patients with diabetes to overcome the limitations of HbA1c and SMBG, should be brought into the routine management of diabetes to facilitate prompt therapy adjustment.
Our study had several strengths. Firstly, compared with similar domestic studies, the sample size of our study is larger. Secondly, after installing the CGMS device, all the participants returned home without any changes of routine life, making the risks of AH highly reliable. Besides, several weaknesses should be noted. First, our CGM data only covered three days of glucose metrics, which might underestimate the incidence of AH. Second, since we did not collect data of other possible causes of hypoglycemia, such as alcohol consumption and physical activities, it is impossible to determine whether these factors play a role in the occurrence of AH in T2DM. Finally, our study is a retrospective study, which has recalling bias. Therefore, further large samples of prospective studies are needed to consolidate the conclusions of this study, including long-term follow-up of the hypoglycemic outpatients after treatment and lifestyle adjustments.