Introduction

Heart failure (HF) is the terminal stage of several cardiovascular diseases. The incidence of HF is rising annually due to an aging population and the ongoing prevalence of metabolic risk factors like hypertension and diabetes [1]. In developed countries, the adult HF prevalence is approximately 1–2%, which can escalate up to 10% in those > 70 years [2, 3]. In China, the adult HF prevalence increased from 0.9 to 1.3% in 2003 and 2012–2015, with approximately 5 million additional cases during this period [4, 5].Despite significant advances in HF treatment, its prognosis remains poor, with an approximately 50% 5-year mortality rate [6]. This underscores the importance of accurately identifying the high-risk populations for providing early targeted interventions to effectively prevent HF onset.

The glomerular filtration rate estimated by serum creatinine (eGFRcr) is the most accurate indicator for assessing renal function. Moreover, a decline in eGFRcr is an independent risk factor for predicting the onset, progression, and poor prognosis of HF [7,8,9]. Compared to serum creatinine, cystatin C remains unaffected by muscle mass, age, and gender [10, 11]. Therefore, a stronger and more linear association has been observed between eGFRcys decline and the risks of HF, cardiovascular disease, and mortality [12, 13]. However, discrepancies between eGFRcr and eGFRcys are common. The difference between eGFRcys and eGFRcr (eGFRdiff), a novel variable, has significant predictive value for incident HF in the general population and chronic kidney disease patients (CKD) [14, 15].

However, existing research has mainly focused on the general population and CKD individuals. The predictive value of different eGFR metrics based on cystatin C and serum creatinine for HF incidence in diabetic patients at high risk for HF and CKD has not yet been assessed. Therefore, we investigated the predictive value of different eGFR metrics for HF incidence in the population with type 2 diabetes(T2D) by using the Kailuan study data.

Materials and methods

Study cohort

Being a prospective cohort study, health examinations for active and retired Kailuan Group personnel were conducted every two years from June 2006 to October 2007 in 11 hospitals, including Kailuan General Hospital and its affiliated hospitals. Follow-up assessments included incident HF and mortality. In the 6th health examination in 2016, cystatin C was investigated in T2D patients. We selected T2D patients who participated in this health examination and underwent cystatin C investigations as our study subjects. The inclusion criteria were: (1) Those who participated in the 2016 annual health examination and met the diagnostic criteria for T2D; (2) Patients with the availability of primary research data, including cystatin C and serum creatinine, and (3) Those willing to participate and provide informed consent. The exclusion criteria were: (1) Patients with a history of HF before the health examination and (2) Those having valvular and congenital heart diseases, respectively.

Collection of general clinical data and laboratory investigations

Patient data like age, gender, personal history, disease history, and medication usage were obtained through face-to-face interviews. We measured height, weight, blood pressure, heart rate, and relevant biochemical indicators by following previously published methods [16]. Smoking was defined as averaging at least one cigarette per day > 1 year or having quit smoking < 1 year ago. Additionally, body mass index (BMI) was calculated as weight / height2 (kg/m2).

Calculation and grouping of eGFRcr and eGFRcys

We used the 2012 CKD Epidemiology Collaboration (CKD-EPI) cystatin C equation and 2021 race-free CKD-EPI equations to calculate eGFRcys and eGFRcr, respectively [10, 17], and eGFRdiff = eGFRcys- eGFRcr.

The subjects were divided into 3 groups according to eGFRdiff level: negative eGFRdiff group: lower than − 15 mL/min/1.73 m2, with eGFRcys lower than eGFRcr; midrange eGFRdiff group: -15 to 15 mL/min/1.73 m2, with eGFRcys similar to eGFRcr; positive eGFRdiff group: 15 mL/min/1.73 m2 or greater, with eGFRcys higher than eGFRcr.

Diagnostic criteria.

T2D: The American Diabetes Association (ADA) Criteria for Diagnosis of Diabetes (2010) was referred [18].

1) History of T2D;

Or 2) Fasting blood glucose (FBG) ≥ 7.0 mmol/L;

Or 3) Two-hour blood glucose of ≥ 11.1 mmol/L in random plasma glucose test or oral glucose tolerance test;

Or 4) Hemoglobin A1c (HbA1c) ≥ 6.5%(47.5 mmol/mol).

HF: Chinese Guidelines for the Diagnosis and Treatment of Chronic Heart Failure (2018) was referred [19].

  1. 1)

    Symptoms and signs of HF, manifested as shortness of breath, fatigue, palpitations, fluid retention, as well as New York Heart Association (NYHA) heart function grade II and above;

  2. 2)

    Modified Simpson’s method: the left ventricular ejection fraction < 50% measured by echocardiography;

  3. 3)

    Plasma N-terminal pro-B-type natriuretic peptide ≥ 125 ng/L.

The diagnosis must meet conditions (1) as well as at least one of conditions (2) and (3).

Follow-up and endpoint events

After the completion of the 6th health examination, that is, the starting point of follow-up, trained medical staff reviewed the inpatient diagnosis and recorded the end-point events of the participants in the Affiliated Hospitals of Kailuan Group and the Designated Hospitals for Medical and Health Insurance of China every year. The end-point events ware defined as HF during the follow-up. The time of the first event was considered as the end-point for those with > 2 events, and the final follow-up date for those without HF was December 31, 2020. All diagnoses were confirmed by professional physicians according to the inpatient medical records.

Statistical analysis

Normally distributed measurement data were expressed as mean + sd. Multiple pairwise-comparison between different groups was conducted using a one-way analysis of variance. The least significant difference (LSD) test and Dunnett’s T3 test were used for evaluating the homogeneity of variance and heterogeneity of variance, respectively. Non-normally distributed data were presented as median and centiles (25th and 75th), while the comparison between the groups was performed using the Kruskal-Wallis rank sum test. Enumeration data were presented as frequency and percentage (n, %), and comparisons between groups were performed by the chi-square test. The Kaplan-Meier method was used to calculate the incidence of HF events in each group and the overall population, and a log-rank test was adopted to compare the difference in the incidence of HF.

eGFRcys and eGFRcr were categorized into the following four groups (ml/min/1.73m2 ): ≥90, 60–89, 45–59, ≤ 45.The eGFRdiff was assessed as a categorical and continuous variable. The effect of different eGFR groups and each 15 increases in eGFR on new-onset HF was studied using a multivariate Cox stepwise regression model. Model 1 unadjusted. Model 2 was adjusted for age and sex. Model 3 was further adjusted for SBP, BMI, TC, HbA1c, hemoglobin, smoking, anti-diabetic treatment, antihypertensive treatment, MI, and atrial fibrillation.

In addition, based on Model 1(age, sex), Model 2 (ARIC-sans-BNP model: age, sex, HR, BMI, SBP, HbA1c, hypertension and MI), C-Statistic, net reclassification index (NRI), and integrated discrimination improvement (IDI) were used to assess the ability of different eGFR to improve HF prediction models, respectively.

In order to avoid the influence of MI and hypertension on HF, sensitivity analysis was performed after excluding the above population.

SAS version 9.4 was used for the analysis (SAS Institute, Cary, NC, USA). All statistical analyses were double-tailed, with statistical significance set at P < 0.05.

Results

We included 7,967 participants who met the diagnostic criteria for T2D and underwent the 6th health examination as well as cystatin C and serum creatinine investigations.

Baseline characteristics

At baseline, the participants’ average age was 60.51 ± 10.03 years, with 6,300 (79.08%) males. The average systolic blood pressure (SBP) and HbA1c were 147.27 ± 20.58 mmHg and 7.57 ± 1.66%, respectively. The eGFRcys, eGFRcr, and eGFRdiff were 88.77 ± 22.73 mL/min*1.73 m2, 93.41 ± 15.43 mL/min*1.73 m2, and − 4.59 ± 18.72 mL/min*1.73 m2, respectively. The eGFR evaluated by serum creatinine and cystatin C showed inconsistent results, with eGFRcys values generally being lower than eGFRcr.

More than half of participants had a baseline eGFRdiff between − 15 and 15 mL/min/1.73 m2 (4679 participants [58.7%]; midrange eGFRdiff); 2280 participants (28.6%) had an eGFRdiff less than − 15 mL/min/1.73 m2 (negative eGFRdiff ), and 1008 participants (12.6%) had an eGFRdiff of 15 mL/min/1.73 m2 or greater (positive eGFRdiff). Compared with the other 2 eGFRdiff groups, participants in the negative eGFRdiff group were older, more often female, with higher baseline SBP, BMI, uACR and more anti-diabetic and antihypertensive treatment (Table 1; Fig. 1).

Table 1 Baseline characteristics overall and by eGFRdiff categories in participants
Full size table
Fig. 1
figure 1

A: Distribution of eGFRcr and eGFRcys in the total population; B: Distribution of eGFRdiff in the total population

Full size image

Cumulative incidence of HF events by eGFRdiff groups

Following a median follow-up time of 3.76 ± 0.72 years, 172 patients (2.16%) developed HF, and 517 patients (6.49%) died of all-cause mortality, respectively. The cumulative HF incidence in the midrange, negative, and positive eGFRdiff groups was 1.74%, 4.10%, and 0.61%, respectively. A log-rank test showed a significant difference in the cumulative incidence between the three groups (Fig. 2).

Fig. 2
figure 2

Incidence of heart failure by eGFRdiff category: lower than − 15 mL/min/1.73 m2, -15 to 15 mL/min/1.73 m2, 15 mL/min/1.73 m2 or greater. p < 0.01 for differences among curves using the log-rank test

Full size image

Multivariate cox regression analysis of the relationship between eGFR and new-onset HF

Using HF occurrence and grouping by different eGFR measures as the dependent and independent variables, the multivariate Cox regression analysis made adjustments for all traditional cardiovascular disease risk factors. The results showed that compared to the normal eGFRcr group, the HF risk increased progressively with reducing eGFRcr (HR values 1.84–3.61, all p < 0.05). Similarly, the HF risk increased progressively with decreasing eGFRcys, compared to the normal eGFRcys group (HR values 3.21–10.84, all p < 0.05). Moreover, the negative eGFRdiff group displayed a significantly increased risk of HF (HR 2.15; 95% CI: 1.57–2.94), while the positive eGFRdiff group had a significantly reduced HF risk when compared to the midrange eGFRdiff group (HR 0.40; 95% CI: 0.17–0.93). And each 15 mL/min/ 1.73 m2 higher eGFRdiff was associated with 34% (HR, 0.66; 95% CI, 0.58 − 0.47)lower risk of incident HF.

Table 2 Hazard ratios (HR) and 95% confidence intervals of different eGFR for HF
Full size table

Additional predictive value of eGFR for established HF risk models

In order to explore whether adding different eGFR measures to established HF risk models could improve predictive performance, we added continuous eGFRcr, eGFRcys, and eGFRdiff variables as well as grouped eGFRdiff to model 1 (age and sex) and model 2 (ARIC-sans-BNP model [20]), respectively. Addition of eGFRcr to the ARIC-sans-BNP model led to negligible improvement, eGFRcys and eGFRdiff outperformed eGFRcr in HF prediction beyond conventional risk factors, and the best model was the addition of eGFRcys to ARIC-sans-BNP model (Table 3).

Table 3 The additional predictive value of different eGFR for HF
Full size table

Subgroup analysis

No significant interactions were observed between eGFRdiff groups and age, sex, SBP, HbA1c, BMI, uACR, hypertension, or MI status on the impact of HF (p > 0.05). Subgroup analysis results showed that HF risk decreased as eGFRdiff increased in all subgroups, consistent with the overall population results.

figure a

Sensitivity analysis

In the primary study, the participants were divided into three groups based on eGFRdiff (in mL/min/1.73 m2 ). The result display that eGFRdiff was still significantly associated with incident HF when the participants were divided into quartiles of eGFRdiff in the further study (Additional file 1:Table S1). Even after excluding individuals with hypertension or MI at baseline, the relationship persisted(Additional file 1:Table S2).

Discussion

This study found that both eGFRcr and eGFRcys can independently predict HF incidence in individuals with T2D; however, common discrepancies have been observed between these two measures. Additionally, eGFRdiff, defined as eGFRcys − eGFRcr, is also an independent predictor of incident HF. Our results showed that the negative eGFRdiff group displayed a significantly increased HF risk, while the positive eGFRdiff group had a lower HF risk, compared to the midrange eGFRdiff group. However, eGFRcys exhibits superior additional predictive value for incident HF compared to eGFRdiff and eGFRcr in population with T2D, beyond traditional heart failure risk factors.

Our population with T2D revealed poorly managed metabolic indicators like blood pressure, blood glucose, and BMI. Many patients with T2D showed significant discrepancies between eGFRcr and eGFRcys, consistent with previous studies. Subsequently, our study also confirmed the independent predictive value of both these measures for incident HF, i.e., HF risk gradually increases as eGFRcr and eGFRcys decrease. For the same degree of decline in eGFRcys and eGFRcr, the risk of GFRcys-predicted incident HF was higher than that predicted by eGFRcr. This finding corroborates the results of Chen et al. [21] in the general population. Since cystatin C is less influenced by other factors compared to creatinine, we suggest that eGFRcys might more accurately reflect glomerular filtration rate and can identify high-risk individuals for incident HF better than eGFRcr.

Due to individual discrepancies between eGFRcys and eGFRcr, recent studies have confirmed the associations of eGFRdiff with cognitive decline [22], kidney failure [23], atrial fibrillation [24], and all-cause mortality [25]. In many studies on the general population [21], hypertensive population [26], and CKD patients [27], 23.8%, 29%, and 34% of subjects had an absolute eGFRdiff > 15, respectively. Moreover, in our type 2 diabetes population, the proportion of patients with an absolute eGFRdiff > 15 was higher (41.2%), suggesting more significant discrepancies between eGFRcys and eGFRcr in Chinese patients with T2D. However, further research is needed to confirm the distribution and potential clinical significance of individual eGFRdiff in this cohort due to the paucity of relevant data.

After adjusting for relevant influencing factors, regression analysis results showed that compared to the midrange eGFRdiff group, the negative eGFRdiff group displayed a 115% increased HF risk (HR = 2.15; 95% CI: 1.57–2.94), while the positive eGFRdiff group revealed a 60% decreased HF risk (HR = 2.15;95%CI:1.57–2.94). Thus, for every 15 mL/min/1.73 m2 eGFRdiff increase, the HF risk was reduced by 34%. In the CRIC baseline study [27], the increased and decreased risk of HF in the negative and positive eGFRdiff groups was not statistically significant, compared to the midrange eGFRdiff group. However, in time-updated analyses, the negative eGFRdiff group showed enhanced HF risk (HR = 1.99; 95% CI: 1.39–2.86), and the positive eGFRdiff group displayed reduced lower HF risk (HR = 0.67; 95% CI: 0.49–0.91) compared to midrange eGFRdiff. Due to such consistent results, the impact of the discrepancy between eGFRcys and eGFRcr on new-onset HF in the diabetic population was more pronounced than in the CKD population. This may be related to the baseline differences in eGFRcr and eGFRcys levels between the two populations; however, further research is needed to verify this finding.

Because 15 mL/min/1.73 m2 represents a clinically meaningful difference in eGFR that also distinguishes CKD stages, eGFRdiff was categorized based on a cutoff point of 15. In the sensitivity analysis, our results (Supplemental Table S1), which are in line with the conclusions of Carrero et al. [14], were identical even after grouping participants into eGFRdiff-based quartiles and repeating the Cox analysis with each additional standard deviation. Since hypertension and myocardial infarction are two major causes of HF, we performed a sensitivity analysis excluding these two groups to minimize their impact. Nevertheless, the results remained consistent (Supplementary Table S1). Due to gender imbalance and poor metabolic control in this population, we repeated Cox regression in different subgroups and obtained consistent results across all subgroups.

We not only confirmed that different eGFR measures can independently predict HF risk in type 2 diabetes patients, but also found that adding eGFRcr, eGFRcys, and eGFRdiff to established HF risk prediction models improves their predictive ability; however, eGFRcys provides superior additional predictive value compared to eGFRdiff. To our knowledge, this is the first study to evaluate how various eGFR measures in a type 2 diabetes population can improve traditional HF prediction models. To date, only Lees et al. [25] have confirmed that eGFRcys greatly improves cardiovascular disease prediction models than eGFRdiff in the general population.

No conclusive evidence is available regarding the mechanism underlying the association between eGFRdiff and incident HF. A possible explanation suggests the presence of “Pore Shrinkage Syndrome,” in which the glomerular basement membrane’s pore size decreases. Due to the larger molecular weight of cystatin C than creatinine, glomerular filtration of cystatin C decreases while creatinine filtration is unaffected. As a result, serum cystatin C levels increase and creatinine levels remain unchanged, thereby causing reduced eGFRcys compared to eGFRcr [28]. Subsequently, elevated pro-atherogenic protein levels in such patients could lead to the occurrence and development of cardiovascular diseases [29]. Another possible explanation is “sarcopenia,” as diabetic patients are more prone to sarcopenia compared to those with normal blood glucose levels [30]. Reduced muscle mass causes lower creatinine, higher eGFRcr, and significantly negative eGFRdiff values, respectively [31]. Thus, sarcopenia and HF share common pathophysiological pathways. A muscle metaboreceptor (ergoreceptor) contributes to the hemodynamic and autonomic responses to exercise by controlling the sympathetic, hypertensive, and hyperpnoic responses to exercise and may have a role in the vicious cycle of sympathetic activation, which is considered one of the central elements of HF pathogenesis [32, 33].

Our study had some limitations. Firstly, HF events were identified based on hospitalization diagnosis codes, which may vary across hospitals and exclude non-hospitalized HF patients. Secondly, our observational study design could not establish causality. Lastly, baseline serum creatinine and cystatin C measurements were taken only once and might have caused misclassification bias. However, the large sample size, a stable cohort, detailed influencing factors, and our robust results lend high credibility to our conclusions.

Conclusion

This study confirmed the independent predictive values of different eGFR metrics for HF risk in patients with T2D. Specifically, eGFRcys demonstrates greater sensitivity in predicting HF risk and can significantly enhance the predictive capability of traditional HF models. Our findings support a comprehensive use of cystatin C to estimate eGFR clinically, thereby emphasizing the need to assess eGFRcys rather than relying solely on eGFRcr or eGFRdiff values for HF risk stratification.

Additional file 1: Table S1. The effect of eGFRdiff on heart failure by quartile and each SD increase(Sensitivity analysis). Table S2. Hazard ratios (HR) and 95% Confidence intervals of different eGFRcr for heart failure (Sensitivity analysis).