Introduction
Type 2 diabetes mellitus (T2DM) patients with hemoglobin A1c (HbA1c) values in the normal range (< 6·0%) had the lowest risk of cardiovascular events [
1]. The goal of treatment of patients with T2DM is to reduce blood glucose levels, which can be attained through conventional and intensive glucose-lowering management, but this has also been strongly associated with adverse outcomes [
1‐
3]. The conclusions of recent randomized controlled trials that studied the effectiveness of intensive blood glucose management have not been consistent or conclusive [
4‐
9].
In addition to the level of glycemia, studies have recently identified glycemic variability as a potential risk factor for adverse outcomes in people with T2DM [
10]. Most studies evaluated HbA1c variability using the standard deviation (SD) or the coefficient of variation (CV) of HbA1c [
11‐
13]. However, both the SD and CV of HbA1c were difficult to interpret in clinical practice. In 2018, Forbes et al. proposed a new method to evaluate HbA1c variability, namely the HbA1c variability score (HVS). The HVS was calculated as the percentage of HbA1c level changes > 0·5% (5·5 mmol/mol) among all HbA1c measurements for an individual. This measure of HbA1c was much more clinically translatable.
Recent studies have also demonstrated that patients with T2DM with “low and stable” patterns of HbA1c over time have lower risks [
14,
15]. However, the relationship between HbA1c variability and blood glucose-lowering targets has not been well-studied. Thus, we used machine learning algorithms to cluster the HbA1c variability of participants into low, medium, and high levels based on the HVS and SD of HbA1c; this study aimed to investigate the optimal target of glucose-lowering treatment for patients with different HbA1c variability.
Discussion
In this post-hoc analysis, we found that HbA1c variability can significantly affect the efficacy of glucose-lowering treatments. Within different variability groups, intensive glucose-lowering treatment yielded varying outcomes. Specifically, in the low HbA1c variability group, intensive glucose lowering significantly decreased the risk of MACEs without increasing the risk of all-cause mortality. Conversely, in the high variability group, intensive glucose lowering significantly heightened the risk for both MACEs and all-cause mortality. Additionally, the optimal target for long-term mean HbA1c varied among different variability groups. Our findings potentially offer clinical implications in guiding patients with T2DM to determine their ideal glucose-lowering target.
Numerous studies have shown that in patients with T2DM, lower levels of HbA1c were associated with lower MACEs risks [
2,
21‐
23]. Our data also indicated that in the overall participants of the ACCORD study, the risk of MACEs increased with the rise of long-term mean HbA1c levels, and the risk of all-cause mortality was not related to long-term mean HbA1c levels. However, intensive blood glucose management, compared to standard blood glucose management, not only did not reduce the risk of MACEs but also increased the risk of all-cause mortality [
4]. This result was puzzling, suggesting that there could be some factors that have not received enough attention previously that affected the intensive management. Current research has already established that glycemic variability is an independent predictor of adverse outcomes in patients with T2DM [
24,
25]. In recent years, the variability of HbA1c has gradually gained attention, furthermore, many studies have demonstrated a significant association between the variability of HbA1c and adverse outcomes [
20,
26]. In 2018, a new index for measuring variability, HVS, has been proposed. Previous studies have reported a significant correlation between the HVS and adverse outcomes [
27,
28]. In these studies, using the HVS had many advantages over the SD and the CV, but still had some disadvantages. HVS is more translatable in clinical practice (as it can be interpreted as the percentage of total HbA1c measures that vary by 0·5% or 5·5 mmol/mol) [
29]. Although the HVS can show the frequency of HbA1c variability well, it still ignores the magnitude of HbA1c variability. Our research combines HVS with SD and systematically demonstrates the impact of HbA1c variability.
Previous studies aimed at evaluating the effect of intensive blood glucose treatment did not reach a consistent conclusion, which may be due to HbA1c variability. Of note, a secondary analysis using data from the Veterans Affairs Diabetes Trial (VADT) showed that blood glucose variability was only associated with adverse events in the intensive treatment group, suggesting that excessive blood glucose variability may neutralize the benefits of lower blood glucose levels [
30,
31]. However, our study found that higher variability in HbA1c brings higher risks regardless of treatment strategies. Our results found that in patients with low HbA1c variability, intensive treatment reduced the risk of cardiovascular events by 22% without increasing the risk of all-cause mortality; for these patients, whether considering the primary outcome or the risk of all-cause mortality, a lower mean HbA1c corresponded to a reduced risk. This is consistent with the conclusion of previous studies that patients with low and stable blood glucose levels have the lowest risk [
27]. Our research findings further emphasize the importance of HbA1c variability in the T2DM management, showing that long-term variability could serve as a factor guiding the glycemic control treatment in patients with T2DM. In patients with high HbA1c variability, intensive treatment not only failed to bring benefits but also significantly increased the risk of cardiovascular events and all-cause mortality, our results indicate that the altered risk of hypoglycemic events was insufficient to account for the increased adverse event risk observed in this subset of patients. This might be attributable to the significant rise in the incidence of non-hypoglycemic related SAE in the intensive treatment group. In these patients, maintaining a higher HbA1c level results in a lower risk. The level of mean HbA1c did not influence the risk of the MACEs. However, based on our results, patients with a mean HbA1c of around 7·88% had the lowest risk of all-cause mortality. A mean HbA1c below this value could lead to an increased risk, while values above this threshold did not impact the risk. These results suggested that in patients with T2DM, low levels of HbA1c should not be pursued excessively; instead, HbA1c should be controlled to maintain stability at the ideal level. However, due to the emergence of novel antidiabetic drugs such as Sodium-Glucose Co-Transporter 2 inhibitors (SGLT2i) and Glucagon-Like Peptide-1 Receptor Agonists (GLP-1RA), the current pharmacological treatment for T2DM has many differences from that used in the ACCORD study. Therefore, this result still requires further up-to-date research for confirmation.
Prior research has indicated that patients with high HbA1c variability presented with more cardiovascular risk factors at baseline [
32]. Therefore, the association between HbA1c variability and the risk of adverse events may not be a feature of HbA1c variability itself, but a sign of baseline differences in the patients’ characteristics. In this study, the association between HbA1c variability and the outcome remained robust even after adjustment for major cardiovascular risk factors, although some residual unadjusted risks may remain. The American Diabetes Association guidelines and American Heart Association scientific statement recommend individualization of HbA1c targets using a patient-centered approach: <7% (53 mmol/mol) for most nonpregnant adults; <6·5% for young patients with a long life expectancy and no significant cardiovascular disease; and less stringent targets (i.e., < 8%) for those with a history of severe hypoglycemia, limited life expectancy, and advanced microvascular or macrovascular complications [
33,
34]. Our findings suggested that HbA1c variability could be one of the risk factors guiding patients’ glucose lowering targets, and thus expanded the population that could benefit from intensive glucose lowering strategies. Our results showed that patients with high HbA1c variability had poor outcomes with intensive treatment for blood glucose control compared with other patients, and they also had greater HbA1c variability than the standard treatment group. According to our findings, a potential approach for patients with T2DM exhibiting long-term high HbA1c variability was to administer cautious glucose-lowering treatments. If the glucose-lowering effects were not significant, the strategy should be shifted towards stabilizing blood glucose levels. This could potentially prevent the increase in cardiovascular risks associated with aggressive glucose reduction. For patients with low HbA1c variability, the lower the mean HbA1c, the lower the risk. For those with moderate HbA1c variability, patients with a mean HbA1c level around 7·5% have the lowest risk. Meanwhile, for patients with high HbA1c variability, mean HbA1c does not influence the risk of adverse cardiovascular events, but a value around 7.8% is associated with the lowest all-cause mortality rate. Our results suggest that the long-term HbA1c variability of patients can guide the optimal blood glucose control target. How to establish a more effective risk grading criteria may be the future research direction for the treatment of T2DM.
Our study has many strengths. First, we systematically expounded the relationship between blood glucose-lowering treatment strategies and HbA1c variability for the first time. Second, our study benefitted from a large sample size and long follow-up period. Third, we used machine learning algorithms to classify the population into different HbA1c variability groups. However, our study has some limitations. First, the ACCORD study was not designed to evaluate the relationship between blood glucose-lowering treatment strategies and HbA1c variability; thus, the post-hoc analysis had its inherent limitations and could not lead to causal inferences. Additionally, the participants included in the ACCORD study did not represent the general population; thus, more general population studies are needed to verify our conclusions. Over the past decade, the therapeutic strategies for T2DM have undergone significant advancements. The treatment approach adopted in the ACCORD study is increasingly being supplanted by emerging medications, such as SGLT2i and GLP-1RA. Therefore, caution is warranted when interpreting and applying the findings of this study in clinical practice. Finally, as the high variability group had higher rates of smoking, more heart disease, higher BMI, higher cholesterol, and many other factors, there may be unobserved confounders such as exercise and diet to consider.
Conclusion
In conclusion, our findings suggest that long-term HbA1c variability can guide the long-term glycemic control targets for patients with T2DM. For those with low HbA1c variability, the lower the mean HbA1c, the lower the risk, and intensive glucose management can yield better outcomes. In patients with medium HbA1c variability, intensive glycemic control does not reduce the risk of cardiovascular events or all-cause mortality and a mean HbA1c level around 7·5% is associated with the lowest risk. Conversely, for those with high HbA1c variability, mean HbA1c does not influence the risk of MACEs. However, a value around 7·8% corresponds to the lowest all-cause mortality risk. For this group, having an excessively high value does not increase the all-cause mortality risk, but an excessively low value does elevate it. Further studies, especially those with samples that reflect the general population and research in the real-world setting of patients with T2DM undergoing the latest therapeutic approaches, are essential to validate the conclusions of this study.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.