We performed a retrospective analysis of patients with AF who underwent their initial RFCA at the First Affiliated Hospital of Zhengzhou University and the China-Japan Friendship Hospital between July 2019 and October 2022. The inclusion criteria were hospitalization for initial RFCA and available pre-ablation measurements of serum creatinine and cystatin C. We excluded patients with severe valvular heart disease, uncontrolled thyroid dysfunction, congenital heart disease, or cardiomyopathy, as well as those who were lost to follow-up or died during the follow-up period. This research followed the principles outlined in the Declaration of Helsinki and obtained approval from the ethics committees of the two hospitals (Approval Nos. 2023-KY-0327 and 2022-KY-043). The study design and participant flow are illustrated in Fig. 1.

Fig. 1.

The flowchart illustrates the selection of patients. AF, atrial fibrillation; RFCA, radiofrequency catheter ablation.

The data included various modules, such as clinical characteristics, comorbidities, medications, laboratory tests, and echocardiographic assessments. The duration of AF was determined by calculating the time from the initial diagnosis to the date of RFCA. Persistent AF was defined as AF that lasts longer than seven days. Complications, and preoperative medications, including amiodarone, angiotensin converting enzyme inhibitors (ACEI)/angiotensin receptor blocker (ARB), and statins, were extracted from the electronic medical records. All laboratory measurements were derived from fasting blood samples drawn from the patient’s peripheral veins prior to the ablation procedure. Echocardiographic measurements, including ejection fraction (EF), left ventricular end-diastolic dimension (LVED), and left atrial diameter (LAD), were recorded before the ablation. The ablation methods employed for each patient were documented in the ablation records.

The CKD-EPI equations, recommended by Kidney Disease: Improving Global Outcomes (KDIGO), are widely utilized for estimating eGFR [14]. In this study, we calculated eGFR using three methods based on the CKD-EPI formula: serum creatinine (eGFRcr), serum cystatin C (eGFRcys), and a combination of both (eGFRcrcys) [8, 9]. To classify kidney function, we used a threshold of 60 mL/min/1.73 m², as an eGFR below this value is a widely recognized indicator of reduced kidney function in population-based research. The specific calculation formula is as follows:

eGFRcr = 142 $\times$ min(Scr/k,1)^α $\times$ max(Scr/k,1)^-1.200 $\times$ 0.9938^age $\times$ 1.012 (if female) where Scr is serum creatinine, k is 0.7 for females and 0.9 for males, and $\alpha$ is –0.241 for females and –0.302 for males.

eGFRcys = 133 $\times$ min(Scys/0.8,1)^-0.499 $\times$ max(Scys/0.8,1)^-1.328 $\times$ 0.9962^age $\times$ 0.932 (if female) where Scys is serum cystatin C.

eGFRcrcys = 135 $\times$ min(Scr/k,1)^α $\times$ max(Scr/k,1)^-0.544 $\times$ min(Scys/0.8,1)^-0.323 $\times$ max(Scys/0.8,1)^-0.778 $\times$ 0.9961^age $\times$ 0.963 (if female) where Scr is serum creatinine, Scys is serum cystatin C, k is 0.7 for females and 0.9 for males, and $\alpha$ is –0.219 for females and –0.144 for males.

The RFCA technique has been previously detailed [15]. Briefly, all patients underwent RFCA guided by an electroanatomical mapping system (CARTO, Biosense Webster, Diamond Bar, CA, USA). After a three-dimensional reconstruction of the left atrium, circumferential pulmonary vein isolation was performed in all patients with AF. In persistent AF, additional linear ablations (roof line, mitral isthmus, and cavotricuspid isthmus) were routinely performed, and complex fractionated atrial electrogram ablation was considered if necessary. Superior vena cava (SVC) isolation was applied if induced tachycardia suggested an origin in the SVC or if active SVC potentials were detected. Bidirectional conduction block of all lines was confirmed immediately and reassessed after a 30-minute observation. If the pulmonary veins showed reconnection, re-isolation was done. When AF persisted despite these measures, sinus rhythm was restored using synchronized direct current cardioversion.

After the procedure, unless contraindicated, all patients received oral anticoagulant therapy for up to three months, along with a short-term regimen of antiarrhythmic drugs to reduce early recurrence.

The primary outcome of this study was the recurrence of AF within one year following ablation. Recurrence was identified as any AF, atrial flutter, or atrial tachycardia episode exceeding 30 seconds, documented via electrocardiogram or 24-hour Holter monitoring following a 3-month blanking period [16, 17]. Patients were scheduled for follow-up assessments at 3, 6, 9, and 12 months, conducted either through outpatient visits or telephone consultations. At each interval, electrocardiogram and 24-hour Holter monitoring were recommended. Additionally, patients experiencing symptoms indicative of AF recurrence were advised to undergo immediate electrocardiogram and 24-hour Holter monitoring for confirmation.

Continuous variables are expressed as the mean $\pm$ standard deviation or median (interquartile range), depending on distribution normality, and analyzed using the t-test or Mann-Whitney test. Categorical variables are summarized as frequencies and percentages, with comparisons made using the $\chi$² test where applicable. Event-free survival post-ablation was evaluated using Kaplan-Meier curves and the log-rank test.

Univariate and multivariate Cox regression analyses were conducted to investigate the relationship between different eGFR calculation methods and AF recurrence. Variables that showed statistical significance in the univariate analysis (p 0.05) as well as those deemed clinically relevant were incorporated into the multivariate analysis model. Three models were used to address potential confounders: Model 1 did not account for any adjustments; Model 2 adjusted for age, sex, and body mass index (BMI); and Model 3 additionally incorporated variables identified in univariate Cox regression with p 0.05 (AF type, AF duration $\ge$2 years, hemoglobin, LAD, and CHADS₂ score). A collinearity test was conducted for each model to avoid overfitting. Hazard ratios (HRs) and 95% confidence intervals (CIs) were reported.

Receiver operating characteristic (ROC) curves, including the area under the curve (AUC), were employed to assess the predictive value of different eGFR measurement methods, both alone and in combination with a baseline risk model (age, sex, BMI, AF type, AF duration $\ge$2 years, hemoglobin, LAD, and CHADS₂ score). The incremental predictive value of introducing eGFR calculation methods into the baseline model was assessed with the C-statistic, integrated discrimination improvement (IDI), and continuous net reclassification improvement (NRI). The “Survival” R package was used to calculate the C-statistic, and the “survIDINRI” R package was used for IDI and NRI. Statistical significance was defined as a two-sided p-value 0.05. All analyses were performed in SPSS (version 26.0, IBM Corporation, Armonk, NY, USA) and RStudio (version 4.4.1, R Foundation for Statistical Computing, Vienna, Austria).

After screening, 523 patients were included in the analysis. By the 1-year follow-up, 174 (33.3%) experienced AF recurrence following RFCA. The baseline characteristics and a comparison of the recurrence and non-recurrence groups are summarized in Table 1. The mean age was 60.35 $\pm$ 12.16 years, with 191 (36.5%) being female and 217 (41.5%) having persistent AF. The median eGFRcr was 95.09 (83.40, 103.74) mL/min/1.73 m², while mean eGFRcys and eGFRcrcys values were 75.59 $\pm$ 21.45 and 84.99 $\pm$ 19.71 mL/min/1.73 m², respectively, and their distributions are illustrated in Fig. 2. Notably, eGFRcr exhibited a pronounced right-skewed distribution, whereas eGFRcys approximated a normal distribution.

Fig. 2.

Distribution of estimated glomerular filtration rate (eGFR) by different equations. (A) Distribution of eGFR calculated using serum creatinine (eGFRcr). (B) Distribution of eGFR calculated using cystatin C (eGFRcys). (C) Distribution of eGFR calculated by combining creatinine and cystatin C (eGFRcrcys).

Compared with those who did not experience recurrence, patients in the recurrence group were more likely to have persistent AF, an AF duration $\ge$2 years, and higher CHADS₂ scores (all p 0.05). They also had lower triglyceride, hemoglobin, and hemoglobin A1c levels, as well as reduced eGFRcys and eGFRcrcys, and a larger left atrial diameter (all p 0.05).

The predictive value of different eGFR measures for AF recurrence was assessed using ROC analysis (Fig. 3A), revealing an AUC of 0.585 (p = 0.001) for eGFRcys, which slightly outperformed both eGFRcr (AUC = 0.537, p = 0.179) and eGFRcrcys (AUC = 0.576, p = 0.004). The optimal cutoff values were determined as follows: eGFRcr at 92.978 mL/min/1.73 m² (sensitivity 48.9%, specificity 59.6%); eGFRcys at 64.280 mL/min/1.73 m² (sensitivity 42.0%, specificity 73.6%); and eGFRcrcys at 76.093 mL/min/1.73 m² (sensitivity 42.0%, specificity 71.9%).

Fig. 3.

Receiver operating characteristic (ROC) curves for predicting atrial fibrillation (AF) recurrence. (A) ROC curves of individual estimated glomerular filtration rate (eGFR) measures: eGFR based on creatinine (eGFRcr), eGFR based on cystatin C (eGFRcys), and combined creatinine-cystatin C eGFR (eGFRcrcys). (B) ROC curves showing the predictive performance when each eGFR measure was added to the basic model.

The proportions of patients with eGFRcr, eGFRcys, and eGFRcrcys below 60 mL/min/1.73 m² were 31 (5.9%), 129 (24.7%), and 52 (9.9%), respectively. In the Kaplan-Meier analysis, patients were initially stratified by the 60 mL/min/1.73 m² cutoff, showing no significant difference in AF recurrence for eGFRcr (log-rank p = 0.197; Fig. 4A) and eGFRcrcys (log-rank p = 0.110; Fig. 4E). However, patients with eGFRcys 60 mL/min/1.73 m² exhibited a significantly higher AF recurrence rate (log-rank p = 0.010; Fig. 4C). When stratified by the optimal cutoff values, patients with eGFRcr $\le$92.978 mL/min/1.73 m² showed no significant difference in AF recurrence (log-rank p = 0.068; Fig. 4B), while those with eGFRcys $\le$64.280 mL/min/1.73 m² (log-rank p 0.001; Fig. 4D) and eGFRcrcys $\le$76.093 mL/min/1.73 m² (log-rank p = 0.002; Fig. 4F) had significantly higher recurrence rates.

Fig. 4.

Cumulative incidence of AF recurrence stratified by eGFR measures. (A) AF recurrence by eGFRcr with 60 mL/min/1.73 m² cutoff. (B) AF recurrence by eGFRcr with optimal cutoff. (C) AF recurrence by eGFRcys with 60 mL/min/1.73 m² cutoff. (D) AF recurrence by eGFRcys with optimal cutoff. (E) AF recurrence by eGFRcrcys with 60 mL/min/1.73 m² cutoff. (F) AF recurrence by eGFRcrcys with optimal cutoff.

In the univariable Cox regression analyses for the recurrence of AF (Supplementary Table 1), persistent AF (HR = 1.364, 95% CI: 1.013–1.837; p = 0.041), duration of AF $\ge$2 years (HR = 1.786, 95% CI: 1.321–2.414; p 0.001), lower hemoglobin levels (HR = 0.988, 95% CI: 0.980–0.996; p = 0.005), greater LAD (HR = 1.032, 95% CI: 1.010–1.055; p = 0.005), higher BMI (HR = 1.044, 95% CI: 1.002–1.088; p = 0.038), and higher CHADS₂ scores (HR = 1.133, 95% CI: 1.013–1.267; p = 0.029) were each significantly associated with AF recurrence. Other parameters examined did not reach statistical significance.

In the multivariable Cox regression analyses (Table 2), when eGFR was treated as a continuous variable, only eGFRcys consistently showed a significant inverse association with AF recurrence across all three models (Model 1: HR = 0.988, 95% CI: 0.981–0.995, p = 0.001; Model 2: HR = 0.986, 95% CI: 0.978–0.994, p = 0.001; Model 3: HR = 0.990, 95% CI: 0.982–0.998, p = 0.019). eGFRcrcys was also significantly associated with AF recurrence in Models 1 and 2 (p = 0.003 and p = 0.004, respectively) but only showed borderline significance in Model 3 (p = 0.067). By contrast, eGFRcr was not significantly associated with AF recurrence.

When categorizing eGFR by 60 mL/min/1.73 m², there was no significant association between any of the three types of eGFR 60 mL/min/1.73 m²—eGFRcys, eGFRcr, or eGFRcrcys—and AF recurrence after thorough adjustment. When stratified by the optimal cutoff values, patients with eGFRcys $\le$64.280 mL/min/1.73 m² exhibited a significantly higher risk of AF recurrence across all three models (Model 1: HR = 1.743, 95% CI: 1.290–2.356, p 0.001; Model 2: HR = 1.838, 95% CI: 1.313–2.572, p 0.001; Model 3: HR = 1.601, 95% CI: 1.133–2.263, p = 0.008). Similarly, patients with eGFRcrcys $\le$76.093 mL/min/1.73 m² had a significantly higher recurrence risk across all models (Model 1: HR = 1.622, 95% CI: 1.200–2.192, p = 0.002; Model 2: HR = 1.686, 95% CI: 1.200–2.367, p = 0.003; Model 3: HR = 1.457, 95% CI: 1.026–2.068, p = 0.036). In contrast, patients with eGFRcr $\le$92.978 mL/min/1.73 m² did not show a significant increase in AF recurrence risk in any model.

Adding eGFRcr to the baseline model increases the AUC from 0.657 to 0.659, reflecting a minimal gain in predictive ability. By contrast, incorporating eGFRcys into the baseline model enhances the AUC to 0.673, and eGFRcrcys yields a similar yet slightly lower AUC of 0.669 (Fig. 3B).

As shown in Table 3, the baseline risk model achieves a C-statistic of 0.633 (95% CI: 0.591–0.675). The inclusion of eGFRcr leads to little improvement, with the C-statistic only rising to 0.634 (95% CI: 0.592–0.676) and no significant changes in either IDI (p = 0.358) or continuous NRI (p = 0.498). By contrast, eGFRcys notably raises the C-statistic to 0.649 (95% CI: 0.608–0.691) and significantly enhances both IDI (p = 0.020) and continuous NRI (p = 0.030). Similarly, adding eGFRcrcys increases the C-statistic to 0.645 (95% CI: 0.603–0.686) and yields a significant IDI (p = 0.020), though its continuous NRI remains non-significant (p = 0.109).