The KoreaN Cohort Study for Outcome in Patients With Chronic Kidney Disease (KNOW-CKD) is a nationwide prospective cohort study from nine tertiary centers in Korea (NCT0-1630-486, http://www.clinicaltrials.gov). KNOW-CKD recruited patients aged 20 to 75 who had CKD G1–G5 without KRT from 2011 until 2016. The detailed design and methods of KNOW-CKD have been previously published [17]. Among the 2,238 participants, 744 had T2DM. We excluded 37 patients whose baseline levels of HbA1c were unavailable. Therefore, 707 patients were included in the final analysis. The study was conducted following the Helsinki Declaration, and the research protocol was approved by the Institutional Review Board at each participating center (IRB approval number of Yonsei University Severance Hospital: 4-2011-0163). Written informed consent was obtained from all participants.

Socio-demographic information and medical history, such as age, sex, smoking status, drug history, and detailed personal and family medical history, were evaluated at enrollment. Hypertension was defined as self-reported hypertension, systolic blood pressure ≥140 mm Hg, diastolic blood pressure ≥90 mm Hg, or current use of antihypertensive drugs. Diabetes mellitus was defined as a history of diabetes mellitus, fasting glucose ≥126 mg/dL, or the prescribed use of glucose-lowering drugs. Body mass index (BMI) was calculated by body weight divided by the square of height. The Charlson comorbidity index was used to assess comorbid conditions.

Serum and urine samples were collected after overnight fasting at baseline and 6 months, and annually thereafter, according to the study protocol. The following laboratory variables were measured, hemoglobin, creatinine, total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol (LDL-C), triglycerides, albumin, ferritin and high-sensitive C-reactive protein (hs-CRP). Serum HbA1c concentrations were routinely measured using high-performance liquid chromatography (HPLC) at each visit. Serum creatinine was measured using the isotope dilution mass spectrometry-traceable method, and eGFR was calculated using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) creatinine equation [18,19]. Urinary protein excretion was determined using the urinary protein-to-creatinine ratio (UPCR; g/g).

The patients were followed up from enrollment to March 2020. The primary exposure of interest was time-updated and baseline HbA1c levels. Patients were classified into the following three groups according to HbA1c levels of <7.0%, 7.0%–7.9%, and ≥8.0%. We employed these cut-off values, which have been tested in previous clinical trials with aim to find clinical benefits associated with intensive glycemic control [9,10,12]. Additionally, HbA1c per 1.0% increase was used for the analysis, in which HbA1c was treated as a continuous variable.

The primary endpoint was a composite of major adverse cardiovascular events (MACEs) or all-cause mortality. MACE was defined as nonfatal myocardial infarction, unstable angina, percutaneous coronary intervention, coronary artery bypass graft, nonfatal stroke, and cardiac death [20]. Secondary endpoints included individual outcomes of MACEs, all-cause mortality, and CKD progression. CKD progression was defined as a ≥50% decline in eGFR from baseline or the onset of end-stage kidney disease (ESKD) that required dialysis or transplantation.

The baseline characteristics of the study population were described according to three categories of HbA1c level. Continuous variables were expressed as mean±standard deviation or medians with interquartile ranges for skewed data and compared using one-way analysis of variance. Categorical variables were expressed as numbers and proportions and analyzed by chi-square test. To explore the association between HbA1c levels and the risk of adverse outcomes, a time-varying Cox proportional hazards model was used for the primary analysis. In this analysis, all repeated measures, such as HbA1c, systolic blood pressure, BMI, serum albumin, eGFR, and drugs, were considered as time-varying exposures. Furthermore, we performed a conventional Cox proportional hazards regression model using baseline HbA1c levels. The adjusted model includes variables that showed statistical significance in the unadjusted model and well-known risk factors for cardiovascular events. The proportional hazard assumptions were confirmed using Schoenfeld residuals. Model 1 was adjusted for age, sex, BMI, systolic blood pressure, socioeconomic status, smoking status and Charlson comorbidity index. Model 2 further included the use of medications (renin-angiotensin-aldosterone system inhibitors and statins) and laboratory parameters (serum albumin, LDL-C, eGFR, and UPCR). The results of the hazard models were presented as hazard ratios (HRs) and 95% confidence intervals (CIs). Kaplan-Meier curve analyses for the cumulative incidence of the study outcomes were used to derive the incidence rates, and differences between the groups were compared by the log-rank test. Survival time was defined as the time interval between enrollment and the first onset of clinical outcomes. Patients lost to follow-up were censored on the date of the last examination. Adjusted restricted cubic splines with 3 knots were used to reveal the relationship between HbA1c levels and the risk of mortality. For the secondary analysis of MACE and CKD progression, a cause-specific hazard function for competing risk model was used. In this analysis, non-cardiac deaths that occurred before MACE and death that occurred before kidney outcome were treated as a competing risk and censored. Cumulative incidence function was used for the cumulative renal outcome curve and were compared using Gray’s test. The rate of kidney function decline per year was determined by the slope of eGFR obtained from a generalized linear mixed model. We additionally examined the effect modification among prespecified subgroups by age (<60 or ≥60 years), sex (male or female), BMI (<25 or ≥25 kg/m²), previous cardiovascular disease (yes or no), eGFR (<45 or ≥45 mL/min/1.73 m²), hs-CRP (<1 or ≥1 mg/L), and serum albumin (<4.0 or ≥4.0 g/dL). Statistical significance was defined as P<0.05, and all analyses were performed with Stata version 14.2 (Stata Corporation, College Station, TX, USA).

Demographic, clinical and laboratory details of participants according to HbA1c categories are presented in Table 1. The mean age was 59 years and 478 (67.6%) were men. The mean baseline eGFR was 40.3 mL/min/1.73 m² and the median proteinuria was 1.04 g/g. A histogram showing the distribution of HbA1c levels is presented in Supplementary Fig. 1. The mean and median HbA1c levels of all patients were 7.2% and 6.9%, respectively. The mean HbA1c levels of each group were 6.2%, 7.4%, and 9.1%, and the median levels were 6.3%, 7.3%, and 8.7%, respectively (Table 1).

Overall, there were no significant differences in age, Charlson comorbidity index, eGFR, and serum levels of albumin, LDL-C, ferritin, and hs-CRP. Patients with higher HbA1c levels were more likely to be women and current smokers, and those with higher levels of urinary protein excretion, and total serum cholesterol and triglycerides. Most patients were treated with renin-angiotensin-aldosterone system inhibitors (87%), with no difference in the use of this drug among the three groups. However, diuretics were used more frequently in patients with higher HbA1c levels. There was no difference in the use of oral antidiabetic medication among the three HbA1c categories. Not surprisingly, patients with uncontrolled glycemic control undertook insulin therapy more frequently (Supplementary Fig. 2).

During 3,358 person-years of follow-up, a total of 129 composite outcome events occurred (incidence rate, 3.78 per 100 person-years). The incidence rates of the composite outcome were significantly higher in patients with HbA1c levels of 7.0%–7.9% (3.98 per 100 person-years) and ≥8.0% (5.04 per 100 person-years) than in those with HbA1c levels of <7.0% (3.12 per 100 person-years) (Table 2). The Kaplan-Meier curve for the primary outcome also showed that the events were significantly higher in patients with higher HbA1c categories (Fig. 1A). In the time-varying Cox model, the adjusted hazard ratios (aHRs) for HbA1c levels of 7.0%–7.9% and ≥8.0% were 1.59 (95% CI, 1.01 to 2.49) and 1.99 (95% CI, 1.24 to 3.19), respectively, compared with HbA1c level of <7.0% (Table 3). In an additional analysis treating HbA1c as a continuous variable, the aHR per a 1.0% increase in HbA1c level was 1.17 (95% CI, 1.03 to 1.32) (Table 3). This association was similar to the baseline HbA1c-based model. The corresponding aHRs (95% CIs) for each HbA1c categories were 1.29 (95% CI, 0.82 to 2.04) and 2.03 (95% CI, 1.31 to 3.13), respectively (Supplementary Table 1). The continuous model also showed that a 1.0% increase in HbA1c was associated with 1.17-fold (95% CI, 1.03 to 1.33) higher risk of the primary outcome (Supplementary Table 1).

We also studied secondary outcomes for separate associations of HbA1c levels with the risk of MACE, all-cause mortality, and CKD progression. Consistent with the composite outcome events, there were graded increases in the individual incidence rates of MACE and all-cause mortality across the three HbA1c categories (Table 2). We also observed similar findings in the Kaplan-Meier curve analyses (Fig. 1B and C). In the analysis with MACE, the time-varying Cox model showed that HbA1c levels of 7.0%–7.9% and ≥8.0% were associated with a 2.17-fold (95% CI, 1.20 to 3.95) and 2.26-fold (95% CI, 1.17 to 4.37), respectively, higher risk of this outcome than HbA1c level of <7.0% (Table 4). Furthermore, the aHRs for all-cause mortality were 1.36 (95% CI, 0.68 to 2.72) and 2.08 (95% CI, 1.06 to 4.05) for the corresponding HbA1c categories. When HbA1c was used as a continuous variable, the HRs per a 1.0% increase in HbA1c level were 1.17 (95% CI, 0.98 to 1.40) and 1.23 (95% CI, 1.04 to 1.46) for MACE and all-cause mortality, respectively (Table 4). Similar trends were also observed in the analysis of baseline HbA1c levels (Supplementary Table 1). As previous studies showed a U-shaped association between HbA1c levels and mortality [14,15], we further examined this association by restricted cubic spline curve analysis. However, the results showed a relatively linear relationship between HbA1c levels and risk of mortality (Supplementary Fig. 3).

In contrast to the results above, there were no differences in the risk of adverse kidney outcomes among the three HbA1c categories. This association was consistent in both the timevarying and baseline Cox models (Table 4, Supplementary Table 1). In an additional analysis considering HbA1c as a continuous variable, a 1.0% increase in HbA1c level was not associated with CKD progression (Table 4). The cumulative incidence curve analyses confirmed this finding (Fig. 1D). We further evaluated if this association might differ by CKD severity because clinical benefits of intensive glycemic control with respect to adverse kidney outcome was observed in previous trials, in which patients with CKD G3 or greater were not included. Notably, for eGFR of ≥45mL/min/1.73m², patients with higher HbA1c levels showed greater decline rate of eGFR in generalized linear mixed models (Supplementary Table 2).

To validate our findings, we performed a sensitivity analysis in the tertile group of HbA1c levels. In line with the primary analysis, there was a graded association between HbA1c levels and the risk of the primary outcome. In the time-varying model, the aHRs for the middle and highest tertiles of HbA1c levels. were 1.25 (95% CI, 0.77 to 2.00) and 1.90 (95% CI, 1.19 to 3.05), respectively, compared with the lowest tertile. The corresponding HRs in the baseline Cox model were 1.51 (95% CI, 0.94 to 2.43) and 2.07 (95% CI, 1.29 to 3.33) (Supplementary Table 3).

Finally, we tested the effect of interactions between baseline HbA1c levels and prespecified subgroups by age, sex, BMI, prior cardiovascular disease, baseline eGFR, hs-CRP, and serum albumin level on the risk of the composite outcome. None of the subgroup factors was statistically significant, suggesting that the association of higher HbA1c levels with a higher risk of primary outcome existed in all subgroups (Fig. 2).