Early use of three diabetes drugs and their lasting impact on high blood sugar

Poor glycemic control is an important risk factor for increased cardiovascular events in patients with type 2 diabetes (T2D) (1–3). Originally, the legacy effect refers to a phenomenon where an early and good glycemic control can yield long-lasting benefits in reducing patients’ risk of cardiovascular disease (CVD) even after the intensive therapy is discontinued (2, 4, 5). For example, the results of the United Kingdom Prospective Diabetes Study showed that in newly diagnosed T2D patients, compared to patients receiving conventional dietary therapy, patients receiving intensive glucose therapy at early stage had a lower risk of myocardial infarction and microvascular diseases for 10 years even after the intensive treatment is discontinued (6). Conversely, in other studies, researchers also found that a phenomenon known as legacy effect was observed that an early and good glycemic control can yield long-lasting benefits in reducing patients’ risk of CVD even after the intensive therapy is discontinued, while poor glycemic control at early stage similarly exhibits a legacy effect in increasing the risk of subsequent CVD (7, 8). A study in the United States found that among newly diagnosed T2D patients, compared to patients with HbA1c levels < 6.5% within one year of T2D diagnosis, patients with HbA1c levels ≥ 6.5% were associated with increased microvascular and macrovascular events over the subsequent 10 years (8). Regarding this phenomenon, Ceriello et al. recently reported that sodium-glucose transport protein 2 inhibitors (SGLT-2i), a novel class of anti-hyperglycemic agents that significantly reduce cardiovascular risk (9–11), can attenuate the legacy effect of poor glycemic control when used early in patients with T2D (12). According to their results, when using SGLT-2i in the first two years after diagnosis, the association between a high level of HbA1c in early stage of T2D and an increased subsequent risk of CVDs was no longer evident. However, there are limited studies so far that have reported the influence of early SGLT-2i introduction on the legacy effect of poor early glycemic control. Additionally, these studies did not conduct further analyses in different subpopulations. Therefore, the evidence supporting this finding is currently insufficient. Furthermore, it remains unclear whether other novel class of anti-hyperglycemic agents except SGLT-2i, such as glucagon-like peptide-1 receptor agonists (GLP-1RA) and dipeptidyl peptidase-4 inhibitors (DPP-4i), can have similar effects.

In recent years, the prevalence of T2D as well as the incidence of CVD in diabetic populations have been rising rapidly in China (13–15). Meanwhile, the utilization of novel anti-hyperglycemic agents such as SGLT-2i, GLP-1RA, and DPP-4i has also rapidly increased among Chinese diabetic patients (16, 17). These novel anti-hyperglycemic agents possess multiple pleiotropic effects, particularly, for SGLT-2i and GLP-1RA. Previous studies have reported that in addition to controlling blood glucose levels, these agents also have significant beneficial impacts on cardiovascular protection, renal protection, and improvement of cognitive function (18–20). However, no studies in China have reported the influence of early treatment with these novel anti-hyperglycemic agents on the legacy effect of poor early glycemic control. In this background, we conducted a retrospective population-based cohort study to investigate the influence of early use of SGLT-2i, GLP-1RA and DPP-4i on the legacy effect of hyperglycemia.

The two final cohorts included 21,477 and 22,493 patients with newly diagnosed T2D from 2010–2023 in YRHCD (Figure 2).

Based on the mean HbA1c levels during the early exposure period, each cohort was divided into HbA1c ≤ 7.0% and HbA1c>7.0% groups. In both cohorts, patients with different mean HbA1c levels differed significantly in factors including sex, smoking status, drinking behavior, education level, comorbidities, and use of antihypertensives, lipid lowering agents, and anti-hyperglycemic agents. Compared to patients with HbA1c ≤ 7.0%, patients with HbA1c>7.0% had a higher proportion of males, more likely to have smoked, and with lower education levels. In addition, patients in HbA1c>7.0% groups were also more likely to use of all kinds of anti-hyperglycemic medications (Table 1).

After adjusting for baseline covariates in Cox regression, results showed that compared to patients with mean HbA1c ≤ 7.0% during 1-year early exposure period, those with HbA1c>7.0% had a higher risk of incident CVD, with a HR of 1.165 (1.056–1.285). A similar pattern was observed in the cohort with a 2-year early exposure period, where the HR of patients with HbA1c>7.0% was 1.143 (1.044–1.252) (Table 2).

Each cohort was further divided into SGLT-2i/GLP-1RA users, DPP-4i users, and non-users based on treatments during the early exposure period. The three treatment groups differed significantly in mean age, proportion of males, smoking status, drinking behavior, education level, comorbidities, and most concomitant medications. Compared with non-users, SGLT-2i/GLP-1RA users and DPP-4i users had more male patients and more patients with poor early glycemic control. Besides, SGLT-2i/GLP-1RA users and DPP-4i users also had a higher proportion of patients with CCI≥2, and more users of statins, metformin, and insulin. Detailed characteristics of the study population by treatment groups are shown in the appendix (). 1

During the follow-up, there were 1,802 and 2,090 CVD events occurred in 1-year and 2-year exposure period cohort, respectively. The results of multivariate Cox regression analyses showed that in non-users group high level of HbA1c in early stage of T2D was significantly associated with incident CVDs with an HR of 1.152 (95%CI, 1.039-1.277) and 1.132 (95%CI, 1.028–1.246) compared with HbA1c controlled under 7.0% in 1-year and 2-year exposure period cohort, respectively. However, in patients initiating SGLT-2i/GLP-1RA within 1 or 2 years after T2D diagnosis, higher HbA1c level at baseline was not associated with CVD and the HRs were 1.017 (95%CI, 0.520–1.989) and 1.051 (95%CI, 0.539–2.050) respectively. In contrast, among DPP-4i users, patients with higher HbA1c level at baseline had higher risk of CVD and the HRs (1.535 (95%CI, 1.026–2.299) for 1-year and 1.311 (95%CI, 0.934–1.841) for 2-year exposure period cohort) were higher than that in non-users. The incidence of outcomes in different treatment groups by the mean HbA1c levels during the early exposure period in two study cohort are shown in the appendix (). 1

In subgroup analyses, results were generally consistent with the main analysis. For example, the legacy effect of mean HbA1c>7.0% in early stage of T2D was still able to be observed in patients <55 years old with an HR of 1.280 (95%CI, 1.062–1.542) for 1-year and 1.199 (95%CI, 1.012–1.422) for 2-year exposure period cohort. Moreover, in this subgroup, compared with non-users, the association between high level of HbA1c in the early stage of T2D and increased risk of later incident CVD was still lower and non-significant in SGLT-2i/GLP-1RA users, the HRs for 1-year and 2-year exposure period cohort were 0.547 (95%CI, 0.137–2.179) and 0.704 (95%CI, 0.174–2.845), respectively (Table 2).

In addition, the results of all sensitivity analyses were consistent with the main analysis (Table 3). First, after excluding outcome events occurred within the 30-day lag time to avoid reverse causality, SGLT-2i/GLP-1RA users still showed lower and non-significant associations compared to non-users, with an HR of 1.017 (95%CI, 0.520–1.989) respectively in the 1-year exposure period cohort, and an HR of 1.031 (95%CI, 0.525–2.022) in the 2-year exposure period cohort. Second, extending the washout period to 365 days also did not change the results in the primary analysis, with HRs of 0.993 (95%CI, 0.508–1.940) and 1.011 (95%CI, 0.515–1.988) in SGLT-2i/GLP-1RA users in the two cohorts respectively. Third, when subjects were restricted to patients newly diagnosed with T2D between 2018–2023, an HR of 0.949 (95%CI, 0.478–1.882) and 0.941 (95%CI, 0.455–1.946) was observed in SGLT-2i/GLP-1RA users of two cohorts respectively. Fourth, although subjects with missing data differed significantly from those without missing data in terms of age, follow-up duration, comorbidities, and use of antihypertensives, lipid lowering agents, and anti-hyperglycemic agents (Supplementary Table S5), the results after multiple imputation remained consistent with the primary analysis, with the HRs of 1.064 (95%CI, 0.549–2.063) and 1.028 (95%CI, 0.529–1.996) respectively in patients initiating SGLT-2i/GLP-1RA within 1 and 2 years after T2D diagnosis. Fifth, after adjusting competing risk of death using Fine-Gray subdistribution hazard model, the HRs for SGLT-2i/GLP-1RA users in 1-year and 2-year exposure period cohort were 1.042 (0.551–1.970) and 1.048 (0.576–1.907), respectively, which were consistent with the primary analysis. Sixth, when only patients who had been prescribed SGLT-2i at least once during the early period were classified as SGLT-2i users, and GLP-1RA use was included as a covariate, the results were consistent with the primary analysis. The association between high level of HbA1c in the early stage of T2D and increased risk of later incident CVD was still lower and non-significant in SGLT-2i users, with HRs of 0.929 (0.463–1.864), and 1.013 (0.508–2.018) in the two cohorts respectively. Seventh, after including cardiovascular death as an outcome event of interest, the results remained unchanged compared to the primary analysis, with HRs of 1.017 (0.520–1.989) and 1.051 (0.539–2.050) for SGLT-2i/GLP-1RA users in 1-year and 2-year exposure period cohort, respectively. Finally, after applying IPW to adjust for the covariate differences, the results remained similar to the primary analysis. The HRs for SGLT-2i/GLP-1RA users were 1.141 (0.613–2.123) and 1.243 (0.649–2.378) in 1-year and 2-year exposure period cohort, respectively (Table 3).

To our knowledge, this study is the first to investigate the potential effects of early use of SGLT-2i/GLP-1RA and DPP-4i on remedying the legacy effect of high level HbA1c in a Chinese population with newly diagnosed diabetes. In our study, based on different definitions of the early exposure period, two retrospective cohorts were established. The 1-year exposure cohort was comprised of subjects who had HbA1c measurement data within 1 year of their T2D diagnosis, while for the 2-year exposure cohort, subjects with HbA1c measurement data within 2 year after their T2D diagnosis were included. We found that poor glycemic control within 1 or 2 years after diabetes diagnosis did increase the risk of subsequent CVD. Moreover, compared to that in non-users of SGLT-2i, GLP-1RA and DPP-4i, in patients initiating SGLT-2i/GLP-1RA, high level of HbA1c in early stage of diabetes might not increase CVD risk compared lower level of HbA1c, indicating that these drugs might have the potential to mitigate or even eliminate the harmful CVD effects of high poor glycemic control within 1 or 2 years after diabetes. In contrast, DPP-4i did not present this effect. These results were consistent in subgroup analyses and across all sensitivity analyses.

Our findings were supported by previous studies. The legacy effect of high level of HbA1c in early stage of T2D on later CVD risk that we observed in both cohorts has been reported in several prior observational studies using real-world data. An observational study from U.S. found that HbA1c levels ≥6.5% during the early 0–1 years after diabetes diagnosis were associated with increased subsequent microvascular and macrovascular events (8). Another study based on population from the United Kingdom showed that patients with HbA1c consistently above 7% within 1 year of diabetes diagnosis had significantly higher CVD rates compared to those with HbA1c <7% (7). Additionally, in the study by Ceriello et al., researchers observed a significant trend toward an increasing risk of subsequent CVDs with progressively higher levels of mean HbA1c in the early 0–1, 0–2, and 0–3 years after diabetes diagnosis (12). Ceriello et al. also found that compared to non-users, patients using SGLT-2i during the early exposure period had a lower HR for the association between high level of HbA1c in early stage of T2D and increased later CVD risk, which was also consistent with our findings (12).

Relevant mechanistic studies indicate the legacy effect may be mediated through pathways including epigenetic modifications, non-enzymatic glycosylation of proteins and chronic inflammation, and endothelial dysfunction induced by hyperglycemia-related oxidative stress (5, 31, 32). Recent researches shows that SGLT-2i and GLP-1RA, which can provide cardiovascular benefits, may have the potential to mitigate these harmful legacy effects through their mechanisms of action. Multiple studies indicate that SGLT-2i can reduce oxidative and endoplasmic reticulum stress (33–37), while GLP-1RA can repair cardiovascular damage from hyperglycemia by modulating inflammatory responses and reversing endothelial dysfunction (38–40). These findings from mechanistic studies also supported the phenomena observed in our study.

In our study, we also observed an unexpected finding that DPP-4i use during the early exposure period after diabetes diagnosis did not present the potential to mitigate the harmful legacy effects of high HbA1c level. We thought the following factor may have contributed to this phenomenon. According to the current evidence, compared to SGLT-2i and GLP-1RA, DPP-4i has smaller and less definitive effects on reducing cardiovascular risk. A network meta-analysis of 23 clinical trials found that DPP-4i did not reduce the risk of any CVDs versus placebo, and was associated with higher cardiovascular risk compared to SGLT-2i and GLP-1RA (28). Similarly, multiple observational studies have also found that patients using SGLT-2i and GLP-1RA had lower risk of CVD compared to those using DPP-4i (41–43). However, our study was only an observational study, the potential of DPP-4i inhibitors in mitigating the harmful legacy effects of high HbA1c levels still requires further analysis through future mechanistic-based pharmacological or molecular research studies.

Our study investigated and compared the legacy effect of high level of HbA1c in early stage of T2D on later CVDs between different treatment groups using real-world data from China. In the sensitivity analyses, we considered the potential impact of time-related biases like reverse causality, the potential misclassification from shorter washout periods, the potential changes in diagnostic criteria for outcomes and measurements of some covariates over time, the potential influence of missing values in some categorical covariates, as well as the potential impact of competing risk of death and differences in baseline covariates. This ultimately verified the robustness of the results. Our results validated previous findings and also provided novel evidence on DPP-4i users and across age groups. However, there were also some limitations in our study. First, due to the late market approval of SGLT-2i and GLP-1RA in China, sample sizes of users of these drugs were small, so we could not determine if the non-significant interaction between treatment and glycemic level was due to insufficient power. The limited sample size also precluded analyses in more subgroups to explore results in different subpopulations. This limitation may also impact the generalizability and robustness of our findings, and can be addressed by updating the study results after more medication users have accumulated in the database over time. Second, the follow-up duration of this study was relatively short, and the treatment group had an even shorter follow-up duration due to the late introduction of SGLT-2i and GLP-1RA in China, so we could not observe long-term effects of SGLT-2i/GLP-1RA on CVD risk. Third, residual confounding from unmeasured factors like disease severity cannot be excluded. Fourth, based on the study design, we did not include patients with prevalent CVD, nor did we analyze the impact of SGLT-2i/GLP-1RA on the relapses of CVD, which warrants further investigation through additional research. Finally, the population in this study was from only one municipal district in China, whereas a previous study had shown that the legacy effect of early glycemic control may be influenced by patient characteristics⁷, thus caution should be taken when extrapolating the findings of this study to populations with different characteristics.

In summary, among Chinese patients with newly-diagnosed T2D and without CVD at baseline, a significant association between mean HbA1c levels >7% during the early exposure period of 0–1, 0–2 years after diagnosis and a higher risk of subsequent CVD compared with HbA1c ≤ 7% was observed. Compared with non-users of SGLT-2i, GLP-1RA and DPP-4i, this association was smaller and non-significant when patients receiving SGLT-2i/GLP-1RA during the early exposure period. These findings may provide some evidence support for more extensive and earlier use of novel anti-hyperglycemic agents with cardiovascular protective effects such as SGLT-2i and GLP-1RA.

The datasets used and/or analyzed during the current study are available from the corresponding author SZ but restrictions apply to the availability of these data, which were used under license for the current study and therefore are not publicly available. Data are available from the authors upon reasonable request and with permission of the Yinzhou District Center for Disease Control and Prevention. Requests to access these datasets should be directed to SZ,. siyan-zhan@bjmu.edu.cn

The studies involving humans were approved by the ethical review committee of the Peking University Health Science Center (IRB. No: IRB00001052-18013-Exempt). The studies were conducted in accordance with the local legislation and institutional requirements. Written informed consent for participation was not required from the participants or the participants’ legal guardians/next of kin in accordance with the national legislation and institutional requirements.

SD: Conceptualization, Data curation, Formal Analysis, Methodology, Validation, Writing – original draft, Writing – review & editing. HZ: Conceptualization, Methodology, Validation, Writing – original draft, Writing – review & editing. SC: Conceptualization, Writing – review & editing. YS: Resources, Software, Writing – review & editing. PS: Resources, Supervision, Writing – review & editing. HL: Resources, Supervision, Writing – review & editing. SZ: Conceptualization, Funding acquisition, Project administration, Supervision, Writing – review & editing.

The author(s) declare financial support was received for the research, authorship, and/or publication of this article. This work was supported by the National Natural Science Foundation of China (Grant No. 82330107 and 82204157), Peking University Medicine Fund for world’s leading discipline or discipline cluster development (BMU2023XY015), and the China Postdoctoral Science Foundation (grant NO. 2023T160032 and 2022M710251).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo.2024.1369908/full#supplementary-material↗