T2DM is a chronic condition associated with a range of serious complications, including kidney damage, and peripheral neuropathy (1–4). Effective management of blood glucose levels is crucial for mitigating these complications (5, 6). Although intensive glycemic control appears to confer a near-lifelong benefit of cardiovascular health (7), the growing use of novel pharmacological agents and combination therapies has raised concerns regarding potential adverse effects on renal function (8, 9).

The ACCORD trial focusing on the assessment of conventional kidney-related outcomes have found that intensive glucose-lowering therapy are effective in reducing the risk of early-stage renal dysfunction (microalbuminuria) (10). However, these interventions have not been shown to significantly impact the progression to advanced stage of kidney diseases (kidney failure and all-cause mortality) (11, 12). In addition, previous studies analyzing the impact of intensive glucose-lowering interventions on composite renal endpoints have predominantly focused on the first occurrence of any endpoint, neglecting the varying severity levels of different outcomes. Consequently, there remains a lack of robust evidence regarding the effects on renal system function.

To address the limitations of using conventional endpoints or composite endpoints without considering severity gradations, this study employs the WOs measure within the win ratio method, incorporating multiple indicators that reflect the progression of renal dysfunction from early stages to mortality, providing a more comprehensive assessment of the impact of intensive glycemic control throughout the progression of kidney disease.

After excluding participants without baseline eGFR data (n = 48) or those with fewer than two eGFR measurements (n = 355), the analysis included a total of 9,848 participants, comprising 4,931 in the standard treatment group and 4,917 in the intensive treatment group (Supplementary Figure 1).

The median age was 62.7 years, and 38.3% participants were female. In terms of disease history, 54.6% used aspirin, and 34.9% had a history of cardiovascular disease. The prevalence of dyslipidemia and hypertension was 94.2 and 96.4%, respectively. Baseline laboratory measurements showed a mean eGFR of 91.1 mL/min/1.73 m², with 92.0% having an eGFR more than 60 mL/min/1.73 m². Blood pressure, LDL, HDL, BMI, and glucose values were comparable across groups (Table 1).

A decrease in eGFR slope is the main components in the HCE. Among participants in the intensive treatment group, the most frequently observed outcome was a decline in eGFR slope (56.1%), followed by a 40 and 50% reduction in eGFR. A similar trend was observed in the standard treatment group, where 55.3% of participants experienced a decline in eGFR slope. Severe eGFR declines to below 15 mL/min/1.73 m² were rare, occurring in only 0.3% of participants in both groups (Figure 1).

In this study, all patients who experienced persistent eGFR <15 mL/min/1.73 m² ultimately progressed to ESRD. Therefore, this event was incorporated into the higher-priority ESRD tier and did not serve as an independent level of comparison in the Win Odds analysis. As shown in Figure 2, tie rates exceeded 54.28% for all components except eGFR slope, which had a lower tie percentage of 39.49%, indicating greater discriminatory power. Using a tie-adjusted formula, the overall Win Odds was 1.03 (95% CI, 0.99 to 1.07), suggesting a slightly favorable trend for the placebo group compared to the intensive treatment group in terms of the HCE.

The Maraca plot demonstrates more dichotomous outcomes in the active group compared to placebo group, indicating that while the median rate of eGFR decline (shift to the right in the maraca plot) in active compared with the placebo group, the overall difference between groups was not statistically significant (Supplementary Figure 2).

Among 9,848 patients with type 2 diabetes at high cardiovascular risk receiving glucose-lowering therapy in the ACCORD trial, there were 1,987 cases of a 40% decline in eGFR, 840 cases of a 50% decline, 557 cases of a 57% decline, 33 cases of persistent eGFR < 15 mL/min/1.73 m², 286 cases requiring dialysis, and 622 died. For the composite kidney outcome, which included all-cause mortality, ESRD, and a 40% eGFR decline, the HR was 1.05 (95%CI, 0.97 to 1.13). Similarly, for the composite endpoint including all-cause mortality, ESRD, and a 57% eGFR decline, the HR was 1.07 (95%CI, 0.97 to 1.18). When analyzing the composite outcome of all-cause mortality, ESRD, and a 50% decline in eGFR, the HR was 1.09 (95%CI, 0.98 to 1.22). The WO for the composite kidney outcome was 1.03 (95%CI, 0.99 to 1.07), again revealing no significant difference between the treatment groups (Table 2). Across multiple HCE definitions, almost no significant differences were observed between treatment groups. For the relatively comprehensive HCE (Tiers 1–8), the Win Odds was 1.00 (95% CI, 0.95 to 1.04), and the HR was 1.02 (95% CI, 0.95 to 1.08). When albuminuria-related components (Tiers 6 and 7) were included, treatment effects remained neutral. A slight benefit was observed for the combination of Tiers 7 and 8, with a Win Odds of 1.07 (95% CI, 1.02 to 1.12) (Supplementary Figure 4). Kaplan–Meier survival curves (Supplementary Figure 3) showed minimal divergence between the treatment groups for kidney-related events, supporting the conclusion that intensive glycemic control did not significantly affect kidney outcomes.

This study used the win ratio statistical method to compare the effects of intensive and standard glucose-lowering treatments on HCE, finding no significant difference. Further analyses using Cox modeling across various kidney outcome combinations of differing severity also showed no significant effects.

Some studies investigating intensive glycemic control in patients with T2DM have showed its effectiveness in reducing the risk of early kidney damage. Specifically, the ACCORD trial, ADVANCE trial and the EDIC study reported a lower incidence of microalbuminuria in the intensive treatment group compared to the conventional group (24–26). Likewise, R. Bilous’ analysis of the UKPDS found that tighter glycemic control reduced the relative risk of proteinuria and significantly lowered the proportion of patients with a twofold increase in plasma creatinine levels (27). Furthermore, in the VADT-F trial, a significantly higher proportion of participants in the intensive treatment group retained normal kidney function at study completion (28). Considering the above, the use of isolated and non-continuous renal endpoint endpoints without a clear severity ranking may cause heterogeneities (29). By employing HCE analysis in our study, we conducted a systematic assessment by integrating and hierarchically ranking multiple renal outcomes, including all-cause mortality, ESRD, eGFR decline, and eGFR slope, based on their clinical severity, revealing that intensive glycemic control does not exert significant adverse effects on renal function.

Moreover, while several studies have utilized composite renal outcomes, they predominantly relied on first-event analyses, which may overlook clinically more severe but later-occurring events, such as ESRD or mortality. For instance, a retrospective cohort study found that intensive glucose lowering did not reduce the risk of persistent eGFR below 15 mL/min/1.73 m², doubling of serum creatinine levels, or ESRD (30). Similarly, a post hoc analysis of the ACCORD trial showed no significant reduction in the need for dialysis, or death from any cause in patients who received aggressive treatment (10). To address this limitation, our study utilized the WOs metric within the win ratio methodology, which prioritizes events based on clinical significance rather than chronological occurrence. This approach mitigates biases associated with traditional first-event analyses and is consistent with KDIGO and ERBP guidelines for standardized renal endpoint monitoring (18, 31, 32).

Our findings demonstrate the complementary strengths of Cox regression and the Win Odds approach in evaluating renal composite outcomes. While Cox models consider the time to first event irrespective of clinical severity, Win Odds emphasizes early and clinically significant events within a fixed 3-year window. This difference explains the limited contribution of rare but severe events, such as sustained eGFR <15 mL/min/1.73 m², in the Win Odds analysis. Notably, when albuminuria components were incorporated into the hierarchy, Win Odds revealed modestly favorable trends, highlighting the sensitivity of uACR as an early indicator of kidney injury. This is consistent with previous findings, such as those by An et al. (33), which reported that intensive HbA1c reduction may coincide with short-term eGFR decline in patients with elevated uACR. Given the limited number of events for certain uACR-related tiers, this trend should be interpreted cautiously due to the potential for false-positive findings. Collectively, these results underscore the value of hierarchical composite frameworks for capturing nuanced treatment effects across heterogeneous renal outcomes and the importance of including sensitive markers like uACR in future endpoint definitions.

Although the use of a composite outcome strengthens the validity of our findings, the relatively homogeneous patient population may limit the generalizability of the results. Additionally, the relatively short follow-up period restricts the ability to comprehensively evaluate the long-term effects of intensive glycemic control on renal function. Finally, the inclusion of renal status may be incomplete in our study, underscoring the need for more comprehensive assessments in future research.

Using hierarchical definitions for kidney endpoints, this study found no significant difference between the intensive and standard glucose-lowering groups. Sensitivity analyses consistently supported this conclusion, suggesting that intensive glycemic control may not provide a clear advantage over standard treatment in improving renal outcomes.