Impact of Chronic Kidney Disease Classification on New-Onset Atrial Fibrillation in the General Population　― The TAMA MED Project-AF and CKD ―

Abstract

Background: Atrial fibrillation (AF) and chronic kidney disease (CKD) are known risk factors for each other. In Tama City in Tokyo, 12-lead ECG and serum creatinine concentration have been included as essential examinations in specific health checkups to diagnose AF and CKD. In the present study, we investigated the impact of CKD classification on new-onset AF in the general population.

Methods and Results: Among 13,478 subjects aged 40–74 years at entry (age, 65.6±7.8 years; men, 42.0%), renal impairment with estimated glomerular filtration rate (eGFR) <60 mL/min/1.73 m² and proteinuria were found in 15.5% and 4.6%, respectively. CKD severity in individual subjects was classified according to a heatmap of the Japanese Society of Nephrology as 81.3% in the green, 15.1% in the yellow, 2.5% in the orange, and 0.9% in the red. Of those without AF in 2012, it had developed in 115 up to 2017; thus, the new-onset AF incidence rate was 2.6/1,000 person-years. Hazard ratios and 95% confidence intervals for new-onset AF in each CKD classification were 1.50 (0.93–2.41, P=0.097) in the yellow, 2.53 (1.03–6.23, P=0.044) in the orange, and 4.65 (1.47–14.70, P=0.009) in the red compared with the green as a reference.

Conclusions: CKD classification was significantly associated with new-onset AF in the general population. Thus, it would be useful for risk stratification of new-onset AF. Renal function evaluation is recommended in health checkups.

Atrial fibrillation (AF) is a common arrhythmia and an established risk factor for cardiogenic thromboembolism.^1,2 Accordingly, early detection of undiagnosed AF and prevention of new-onset AF prior to anticoagulation therapy are equally important to reduce AF-related stroke.³ On the other hand, renal impairment is a risk factor for various adverse events in the general population.^4,5 In general clinical practice, estimated glomerular filtration rate (eGFR) is widely used to evaluate renal function,⁶ and chronic kidney disease (CKD) is determined by eGFR <60 mL/min/1.73 m² and/or proteinuria.⁷ Because CKD is not only a precursor to endstage kidney disease (ESKD),⁸ but also a risk factor for new-onset AF,^9–11 CKD diagnosis is important for the prevention of new-onset AF as well as of ESKD in the general population. Therefore, in the present study, we aimed to investigate the impact of CKD classification determined by eGFR and proteinuria categories¹² on new-onset AF using data from the TAMA MED Project-AF and CKD.

In Tama City in Tokyo, 12-lead electrocardiogram (ECG) and serum creatinine concentration (sCr) have been included as essential examinations in specific health checkups, so-called “Tokutei kenshin”, to diagnose AF since 2008 and CKD since 2012.

Methods

Study Design of the TAMA MED Project-AF and CKD

The TAMA MED Project-AF¹³ and Project-CKD¹⁴ were conducted as retrospective cohort studies to clarify the prevalence and incidence of AF and CKD in the general population in conjunction with Project-Frail¹⁵ to research frailty. The study protocols and primary results have been reported.^13,14 Briefly, the studies conformed to the Declaration of Helsinki and were approved by the institutional ethics committee. A consecutive series of subjects who had national health insurance and underwent annual specific health checkups at a clinic or hospital belonging to the TAMA CITY Medical Association were recruited. All participants were aged 40–74 years at the time of entry. Age at the end of a fiscal year was adopted for analyses. All participants completed a questionnaire on self-reported past history, present illness, smoking status, and medications for hypertension, diabetes mellitus, and dyslipidemia. However, the kinds of drugs were unknown. Body height and weight, body mass index (BMI), waist circumference, and systolic and diastolic blood pressures were measured. Moreover, fasting plasma glucose or glycated hemoglobin (HbA1c), serum high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol, triglycerides, aspartate transaminase, alanine transaminase, γ-glutamyl transpeptidase, and urine sugar and protein levels were examined as original essentials. In addition, standard 12-lead ECG and sCr were examined as additional essentials.

AF was diagnosed by physicians directly or from reports based on automatic analysis of 12-lead ECGs in each clinic or hospital, regardless of the electrocardiograph model and vendor. To detect new-onset AF, individuals who did not have AF in 2012 were followed up for 5 years or until AF occurrence. CKD was diagnosed as eGFR <60 mL/min/1.73 m² and/or presence of proteinuria with qualitative urinalysis of (+) or higher.⁷ Diagnostic criteria for other diseases have been described previously.^13,14

Renal Function Evaluation and CKD Classification

Using the data for age, sex, and sCr values at the specific health checkups in 2012, eGFR was calculated by the equations of the Japanese Society of Nephrology:⁶ eGFR (mL/min/1.73 m²) = 194 × sCr (mg/dL)^–1.094 × age^–0.287 × (0.739 if female); and categorized as G1 for ≥90, G2 for 60–89, G3a for 45–59, G3b for 30–44, G4 for 15–29, and G5 for <15 mL/min/1.73 m². Proteinuria categories were modified for the qualitative test strips as A1 for (−) or (±), A2 for (+), and A3 for (++) or higher. CKD severity in individual subjects was classified according to heatmap data of the categories of eGFR (G1–G5) and proteinuria (A1–A3) in the Clinical practice guidebook for diagnosis and treatment of CKD 2012 of the Japanese Society of Nephrology,¹² which was modified for Japanese from the KDIGO (Kidney Disease: Improving Global Outcomes) 2012 clinical practice guideline,¹⁶ as follows: low risk or no CKD (green) for G1 or G2 with A1; moderately increased risk (yellow) for G1 or G2 with A2, G2 with A2, and G3a with A1; high risk (orange) for G1 or G2 with A3, G3a with A2, and G3b with A1; and very high risk (red) for G3a with A3, G3b with A2 or A3, and G4 or G5 regardless of proteinuria category.

Statistical Analysis

Data are presented as mean±standard deviation. The statistical significance of the differences in mean values was analyzed using Student’s t-test. Frequencies of parameters were compared using the chi-square test. The Kaplan-Meier curves were drawn to show the cumulative new-onset AF incidence from 2012 to 2017, and the time to events of each CKD classification were compared using the log-rank test. A Cox proportional hazard model was used to investigate the impact of CKD classification on new-onset AF. Hazard ratios (HRs) and 95% confidence intervals (CIs) for each CKD classification were calculated in 4 different models: unadjusted (Model 1), adjusted for age and sex (Model 2), excluded subjects with cardiac disease history (Model 3), and adjusted Model 3 for BMI and hypertension (Model 4) because BMI and hypertension were independent risk factors for new-onset AF in the subjects without a history of cardiac disease in our previous report on the TAMA MED Project-AF.¹³ The predictive ability of CKD classification for new-onset AF determined by the area under the receiver-operating characteristic curve (AUC) was compared with that of eGFR or proteinuria category by the DeLong’s test.¹⁷ Two-tailed P-values <0.05 were considered statistically significant. All statistical analyses were performed using the SPSS software version 23.0 (IBM Corporation, Armonk, NY, USA) and R version 3.6.1 (The R Foundation for Statistical Computing, Vienna, Austria).

Results

AF and CKD Prevalence in 2012

Characteristics of the participants in 2012 are shown in Table 1. Of the 13,478 subjects (age, 65.6±7.8 years; men, 42.0%), AF was found in 157 (121 men, 36 women); thus, AF prevalence was 1.2% (2.1% in men, 0.5% in women) in 2012. Numbers in each eGFR and proteinuria category are summarized in Table 2. Accordingly, CKD severity in individual subjects was classified as 81.3% in the green, 15.1% in the yellow, 2.5% in the orange, and 0.9% in the red (Table 3). Renal impairment with eGFR <60 mL/min/1.73 m² and proteinuria was found in 2,087 (15.5%) and 614 (4.6%), respectively (Table 2). Sum of numbers in CKD classifications of the yellow, orange, and red was 2,490; hence, CKD prevalence was 18.5% in 2012 (Table 3). There was a significant difference in the distribution of CKD classification between subjects with and without AF (P<0.001, Table 3). Consequently, CKD prevalence in subjects with AF was significantly higher than in those without AF (39.5% vs. 18.2%, P<0.001) (Table 3).

Table 1. Characteristics of Study Subjects in 2012

No. of subjects	13,478
Age, years	65.6±7.8
Sex, male	5,665 (42.0)
BMI, kg/m2	22.6±3.3
Systolic BP, mmHg	126.9±16.5
Diastolic BP, mmHg	75.1±10.5
eGFR, mL/min/1.73 m2	72.7±25.9
Current smoker	2,886 (21.4)
Past history	9,199 (68.3)
Present illness	2,792 (20.7)
Comorbidities
Cardiac disease*	741 (5.5)
Hypertension	5,613 (41.6)
Diabetes mellitus	1,650 (12.2)
Stroke/TIA	540 (4.0)
Metabolic syndrome	1,903 (14.1)
AF	157 (1.2)
Blood examinations
HDL-C, mg/dL	62.7±16.9
LDL-C, mg/dL	124.3±30.5
Triglycerides, mg/dL	114.1±83.5
HbA1c (NGSP),** %	5.7±0.7
AST, IU/L	24.5±13.4
ALT, IU/L	21.6±17.5
γ-GTP, IU/L	36.3±54.0
Urinalysis
Urine sugar	292 (2.2)
Urine protein	614 (4.6)
Medications
For hypertension	4,449 (33.0)
For diabetes mellitus	955 (7.1)
For dyslipidemia	2,582 (19.2)

Data are number of patients (%) or mean±SD. *Cardiac disease was self-reported. **HbA1c (JDS) +0.4%. γ-GTP, γ-glutamyl transpeptidase; AF, atrial fibrillation; ALT, alanine transaminase; AST, aspartate transaminase; BMI, body mass index; BP, blood pressure; eGFR, estimated glomerular filtration rate; HDL-C, high-density lipoprotein cholesterol; JDS, Japan Diabetes Society; LDL-C, low-density lipoprotein cholesterol; NGSP, National Glycohemoglobin Standardization Program; TIA, transient ischemic attack.

Table 2. Numbers and Prevalence of eGFR and Proteinuria Categories in 2012

CKD classification	Proteinuria categories				Total
	A1	A2	A3	No urine data
	UP (−) or (±)	UP (+)	UP (++) or (+++)
eGFR categories (mL/min/1.73 m2)
G1 ≥90	1,621 (12.0)	41 (0.3)	17 (0.1)	3	1,682 (12.5)	11,389 (84.5)
G2 60–89	9,330 (69.2)	282 (2.1)	84 (0.6)	11	9,707 (72.0)
G3a 45–59	1,717 (12.7)	81 (0.6)	40 (0.3)	3	1,841 (13.7)	2,087 (15.5)
G3b 30–44	150 (1.1)	28 (0.2)	25 (0.2)	2	205 (1.5)
G4 15–29	6 (0.0)	3 (0.0)	8 (0.1)	0	17 (0.1)
G5 <15	3 (0.0)	1 (0.0)	4 (0.0)	16	24 (0.2)
Unknown	2	0	0	0	2 (0.0)
Total	12,829 (95.2)	436 (3.2)	178 (1.3)	35 (0.3)	13,478
		614 (4.6)

Data are number of patients (%). : low risk (if no other markers of kidney disease, no CKD), : moderately increased risk, : high risk, : very high risk. CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate; UP, urine protein.

Table 3. Numbers and Prevalence of CKD Classifications in 2012

CKD classification	Analyses for AF and CKD prevalence in 2012			Analyses for new-onset AF incidence
	Overall participants	Subjects with AF**	Subjects without AF	Subjects without AF (for Models 1 and 2)	Subjects without AF or cardiac disease history (for Models 3 and 4)
Green (low risk)*	10,951 (81.3)	94 (59.9)	10,857 (81.5)	9,383 (82.4)	8,992 (82.8)
Yellow (moderately increased risk)	2,040 (15.1)	46 (29.3)	1,994 (15.0)	1,670 (14.7)	1,561 (14.4)
Orange (high risk)	332 (2.5)	11 (7.0)	321 (2.4)	252 (2.2)	232 (2.1)
Red (very high risk)	118 (0.9)	5 (3.2)	113 (0.8)	83 (0.7)	69 (0.6)
Unknown	37 (0.3)	1 (0.6)	36 (0.3)	–	–
CKD (sum of yellow, orange, and red)	2,940 (15.8)	62 (39.5)	2,428 (18.2)	2,005 (17.6)	1,862 (17.2)
Total	13,478	157	13,321	11,388	10,854

Data are number of patients (%). *If no other markers of kidney disease, no CKD. **P<0.001 vs. subjects without AF for distribution. AF, atrial fibrillation; CKD, chronic kidney disease.

Baseline Characteristics of Subjects Without AF in 2012

Of the 13,478 participants in 2012, 157 with AF at the time of entry, 37 without information on CKD classification, and 1,896 who underwent specific health checkup only once in 2012 were excluded. Consequently, a total of 11,388 subjects who did not have AF in 2012 and underwent annual checkups at least once between 2013 and 2017 were analyzed for new-onset AF incidence (Table 3). Baseline characteristics and CKD classification of subjects without AF in 2012 (n=11,388) are shown in Table 3 and Table 4. The characteristics of subjects who subsequently developed AF (new-onset-AF, n=115) or did not until 2017 (no-AF, n=11,273) are also separately shown in Table 4. Age and BMI as well as the frequency of men, any past history, cardiac disease, hypertension, history of stroke or transient ischemic attack, metabolic syndrome,¹⁸ and urinary protein and sugar levels in the onset-AF group were significantly higher than in the no-AF group, whereas HDL-C in the new-onset-AF group was significantly lower than in the no-AF group (Table 4). After exclusion of subjects with a cardiac disease history, the number of subjects without AF in 2012 was 10,854, which were used for additional Cox hazard model analyses (Models 3 and 4). Their CKD classifications are shown in Table 3.

Table 4. Baseline Characteristics of Study Subjects Without AF in 2012

	Overall	No-AF	New-onset-AF	P value***
No. of subjects	11,388	11,273	115
Age, years	65.4±7.3	65.4±7.3	67.2±4.3	<0.001
Sex, male	4,664 (41.0)	4,577 (40.6)	87 (75.7)	<0.001
BMI, kg/m2	22.6±3.2	22.6±3.2	23.8±3.8	<0.001
Systolic BP, mmHg	126.7±16.3	126.7±16.3	128.7±15.7	0.200
Diastolic BP, mmHg	75.1±10.5	75.1±10.5	76.2±9.6	0.230
eGFR, mL/min/1.73 m2	74.7±47.8	74.8±48.0	71.8±14.1	0.506
Current smoker	2,379 (20.9)	2,553 (20.9)	26 (22.6)	0.652
Past history	7,774 (68.3)	7,682 (68.1)	92 (80.0)	0.007
Present illness	2,306 (20.2)	2,283 (20.3)	23 (20.0)	0.947
Comorbidities
Cardiac disease*	534 (4.7)	506 (4.5)	28 (24.3)	<0.001
Hypertension	4,706 (41.3)	4,642 (41.2)	64 (55.7)	0.002
Diabetes mellitus	1,355 (11.9)	1,342 (11.9)	13 (11.3)	0.843
Stroke/TIA	432 (3.8)	416 (3.7)	16 (13.9)	<0.001
Metabolic syndrome	1,569 (13.8)	1,543 (13.7)	26 (22.6)	0.006
Blood examinations
HDL-C, mg/dL	63.0±16.9	63.0±16.9	59.6±16.5	0.030
LDL-C, mg/dL	124.9±30.1	124.9±30.0	119.8±37.1	0.069
Triglycerides, mg/dL	113.6±84.2	113.6±84.3	117.4±73.9	0.626
HbA1c (NGSP),** %	5.7±0.6	5.7±0.6	5.8±0.9	0.228
AST, IU/L	24.4±13.5	24.4±13.5	24.2±6.8	0.884
ALT, IU/L	21.6±17.9	21.6±18.0	21.0±10.4	0.699
γ-GTP, IU/L	35.5±47.8	35.5±48.0	40.9±34.8	0.228
Urinalysis
Urine sugar	228 (2.0)	219 (1.9)	9 (7.8)	<0.001
Urine protein	486 (4.3)	476 (4.2)	10 (8.7)	0.018
Medications
For hypertension	3,682 (32.3)	3,626 (32.2)	56 (48.7)	<0.001
For diabetes mellitus	767 (6.7)	757 (6.7)	10 (8.7)	0.399
For dyslipidemia	2,201 (19.3)	2,189 (19.4)	12 (10.4)	0.015

Data are no. of patients (%) or mean±SD. *Cardiac disease was self-reported. **HbA1c (JDS) +0.4%. ***Comparison between no-AF and onset-AF groups. Abbreviations as in Table 1.

New-Onset AF Incidence

Of the 11,388 subjects (age, 65.4±7.3 years; men, 41.0%) without AF in 2012, AF developed in 115 (1.0%) between 2012 and 2017. The overall incidence rate of new-onset AF was 2.6/1,000 person-years during an observation period of 5.0 years (44,182 person-years) (Figure 1). In each CKD classification, AF developed in 8 (0.9%) in the green, 21 (1.3%) in the yellow, 5 (2.0%) in the orange, and 3 (3.6%) in the red; therefore, the incident rates were 2.4/1,000 person-years in the green, 3.5/1,000 person-years in the yellow, 22.1/1,000 person-years in the orange, and 10.8/1,000 person-years in the red. There was a significant difference in the rates of new-onset AF among the CKD classifications (P=0.004 by log-rank test) (Figure 2).

Figure 1.

Kaplan-Meier curve for new-onset AF in Tama City between 2012 and 2017. AF, atrial fibrillation.

Figure 2.

Kaplan-Meier curves for new-onset AF in Tama City between 2012 and 2017 in each CKD classification. Green: low risk (if no other markers of kidney disease, no CKD), Yellow: moderately increased risk, Orange: high risk, Red: very high risk. AF, atrial fibrillation.

Impact of CKD Classification on New-Onset AF

Unadjusted HRs and 95% CIs for new-onset AF in each CKD classification were 1.50 (0.93–2.41, P=0.097) in the yellow, 2.53 (1.03–6.23, P=0.044) in the orange, and 4.65 (1.47–14.70, P=0.009) in the red compared with the green as a reference (Table 5, Model 1). After adjusting for age and sex, the respective values were 1.16 (0.72–1.88, P=0.545) in the yellow, 1.62 (0.66–4.01, P=0.295) in the orange, and 3.01 (0.95–9.54, P=0.061) in the red (Table 5, Model 2). After exclusion of subjects with cardiac disease history (n=10,854), HRs and 95% CIs for new-onset AF in each CKD classification were 1.11 (0.60–2.05, P=0.735) in the yellow, 3.32 (1.34–8.25, P=0.010) in the orange, and 4.33 (1.06–17.68, P=0.041) in the red (Table 5, Model 3). In addition, after adjusting for BMI and hypertension, they were 1.00 (0.54–1.84, P=0.986) in the yellow, 2.55 (1.02–6.39, P=0.046) in the orange, and 3.17 (0.77–13.09, P=0.110) in the red, respectively (Table 5, Model 4).

Table 5. CKD Classification and New-Onset AF (Cox Proportional Hazard Model)

Variables	Subjects without AF in 2012 (n=11,388)				Subjects without AF or cardiac disease history in 2012 (n=10,854)
	Model 1		Model 2		Model 3		Model 4
	HR (95% CI)	P value	HR (95% CI)	P value	HR (95% CI)	P value	HR (95% CI)	P value
CKD classification
Green	Ref.		Ref.		Ref.		Ref.
Yellow	1.50 (0.93–2.41)	0.097	1.16 (0.72–1.88)	0.545	1.11 (0.60–2.05)	0.735	1.00 (0.54–1.84)	0.986
Orange	2.53 (1.03–6.23)	0.044	1.62 (0.66–4.01)	0.295	3.32 (1.34–8.25)	0.010	2.55 (1.02–6.39)	0.046
Red	4.65 (1.47–14.70)	0.009	3.01 (0.95–9.54)	0.061	4.33 (1.06–17.68)	0.041	3.17 (0.77–13.09)	0.110
Age (/1-yr increase)			1.08 (1.03–1.12)	<0.001	–		–
Sex (men)			4.54 (2.96–6.97)	<0.001	–		–
BMI (/1-kg/m2 increase)			–		–		1.07 (1.01–1.14)	0.020
Hypertension			–		–		1.66 (1.07–2.57)	0.025

Model 1: unadjusted. Model 2: adjusted for age and sex. Model 3: excluded subjects with cardiac disease history. Model 4: adjusted Model 3 for BMI and hypertension. BMI, body mass index; CI, confidence interval; CKD, chronic kidney disease; HR, hazard ratio.

Predictive Ability of CKD Classification and eGFR and Proteinuria Categories for New-Onset AF

When the new-onset AF incidence was evaluated based on eGFR category, AF developed in 92 (0.9%) in G1 or G2, 21 (1.4%) in G3a, and 2 (1.4%) in G3b. No AF cases developed in either G4 or G5. HRs and 95% CIs for new-onset AF in each eGFR category were 1.64 (1.02–3.63, P=0.042) in G3a and 1.77 (0.44–7.17, P=0.427) in G3b compared with G1+G2 as a reference (Supplementary Table). On the other hand, when it was evaluated based on proteinuria category, AF developed in 105 (1.0%) in A1, 4 (1.1%) in A2, and 6 (4.4%) in A3. HRs and 95% CIs for new-onset AF in each proteinuria category were 1.23 (0.46–3.42, P=0.651) in A2 and 5.15 (2.26–11.72, P<0.001) in A3 compared with A1 as a reference (Supplementary Table).

The AUC of CKD classification for new-onset AF was 0.541, whereas those of eGFR and proteinuria categories were 0.527 and 0.523, respectively. There were no significant differences in AUCs between CKD classification and eGFR category (P=0.229) or between CKD classification and proteinuria category (P=0.284).

Discussion

In the present study, we investigated for the first time the impact of CKD classification based on a heatmap of the Japanese Society of Nephrology¹² on new-onset AF. The major findings were as follows. First, the AF and CKD prevalence was 1.2% and 18.5%, respectively, in subjects aged 40–74 years at entry in the general population in 2012. Second, the new-onset AF incidence rate was 2.6/1,000 person-years, and it gradually increased as CKD classification progressed. Third, CKD classification was independently associated with new-onset AF in this general population. However, the superiority of CKD classification for new-onset AF prediction to eGFR or proteinuria category was not confirmed.

AF and CKD Prevalence

In the present study, the AF prevalence was 1.2% in 2012, positioned just between 0.8% in 2008 and 1.4% in 2015, as we previously reported,¹³ indicating that AF prevalence might gradually increase yearly in the same city. To confirm it, investigation of secular trends in AF prevalence including data from other years would be necessary in future studies. On the other hand, the CKD prevalence was 18.5% in 2012, which was a little less than the 19.8% in 2015 that we previously reported.¹⁴ Similar to AF, because the CKD prevalence might have increased in recent years, this should also be clarified in future studies.

As shown in Table 3, CKD prevalence in subjects with AF was more than 2-fold higher than in those without AF, suggesting that AF had already developed in the high-risk subjects with CKD before 2012.

New-Onset AF Incidence and Known Risk Factors

In the present study, the new-onset AF incidence rate was 2.6/1,000 person-years during 2012 to 2017, which was comparable with the 2.5/1,000 person-years during the period between 2008 and 2015, as previously reported.¹³ Therefore, it is unknown from the present results whether the new-onset AF incidence rate has recently changed.

To date, several clinical conditions and comorbidities have been identified as independent risk factors for new-onset AF.^13,19–21 In the Framingham Heart Study, age, hypertension, diabetes mellitus, congestive heart failure, valvular disease, and myocardial infarction (only in men) were found to be significant risk factors for new-onset AF.¹⁹ A report from the Suita study in Japan revealed that age, systolic hypertension, overweight, excessive drinking, coronary artery disease, current smoking, arrhythmia other than AF, cardiac murmur, and non-HDL-C were independent risk factors for new-onset AF.²⁰ We also reported that older age, male sex, larger BMI, and hypertension were independent risk factors for new-onset AF in the general population without overt cardiac disease history.¹³

New-Onset AF Incidence and Renal Impairment

Because renal impairment is also reportedly a risk factor for new-onset AF,^9–11 renal function evaluation is important during annual health checkups in the general population. In Tama City, the sCr measurement was initiated from 2012; thus, no renal function parameter was included as a covariate to identify risk factors for new-onset AF in our previous report based on data in 2008.¹³ Renal function is evaluated by eGFR using sCr⁶ in general clinical practice. In addition, CKD severity classification based on heatmap data of eGFR and proteinuria category is recommended.¹² Proteinuria is reportedly associated with the AF prevalence²² and new-onset AF¹¹ independent of eGFR and the degree of renal dysfunction. Although eGFR is reported to be better than creatinine clearance for evaluating the association with AF prevalence,²¹ as to which is better in predicting new-onset AF between CKD classification and eGFR category remains unknown.

Impact of CKD Classification on New-Onset AF

Accordingly, we hypothesized that (1) CKD classification can be used for new-onset AF prediction, and (2) CKD classification is better as a new-onset AF predictor than the eGFR or proteinuria category only, because the CKD classification contains information on both eGFR and proteinuria.¹² As it turned out in confirming these hypotheses in the present study, CKD classification was significantly associated with new-onset AF in the general population (Table 5, Model 1). Although statistical significance in the HRs of the orange and red categories for new-onset AF disappeared after adjusting for age and sex (Table 5, Model 2), this result was expected because the equations for calculating eGFR already contain age and sex data.⁶ However, the trend in CKD classification did not change even after adjusting for age and sex. In contrast, age and sex were consistently associated with new-onset AF, as previously reported.¹³ Even when focusing on subjects without cardiac disease history, the new-onset AF incidence in CKD classifications of the orange and red was significantly higher than that of the green (Table 5, Model 3) as shown in the Model 1. Although statistical significance in HRs of the red for new-onset AF disappeared after adjusting for BMI and hypertension, probably due to small numbers in the red, the graduality of HRs was maintained (Table 5, Model 4) as in the Model 2. Therefore, CKD classification can be used for risk stratification of new-onset AF in the general population.

However, it may be difficult to use the CKD classification for new-onset AF prediction because the new-onset AF incidence rate in subjects with CKD is potentially not so high in the general population aged 40–74 years, as observed in the present study. In addition, because there was no significant difference in the predictive ability for new-onset AF between CKD classification and eGFR category or between CKD classification and proteinuria category, we failed to prove the second hypothesis. Therefore, CKD classification might not necessarily be essential for AF prediction. However, renal function evaluation using at least either eGFR or proteinuria category would be recommended in the general population.

On the other hand, when eGFR category was used, HRs for new-onset AF were significantly high in G3a only; and if the proteinuria categories were used, a significant high-HR was found in A3 only (Supplementary Table). These findings indicated that the impact of only eGFR or proteinuria category was limited for risk stratification of new-onset AF. In contrast, when CKD classification was used, the risk for new-onset AF gradually increased as CKD classification progressed (Supplementary Table). Therefore, on balance, CKD classification might be better for risk stratification of new-onset AF than only eGFR or proteinuria category, although the AUCs were comparable.

Study Limitations

First, the participants of specific health checkups were limited by age to between 40 and 74 years at entry, because elderly people aged ≥75 years move to the public late-elderly health insurance system in Japan. In addition, paroxysmal AF was not always detected at the time of health checkup. Therefore, the overall AF prevalence and incidence must be underestimated. Second, although the overall number of participants was more than 13,000, the numbers of subjects with CKD and subsequent new-onset AF were relatively small. One possible reason was that 62 subjects with AF and CKD at baseline in 2012 had to be excluded from the investigation for new-onset AF (Table 3). Third, the diagnostic accuracy of the automatic analysis of ECGs was not validated. Because there are many models and vendors of electrocardiographs, diagnostic algorithms were not unified. Fourth, GFR was not directly measured using 24-h urine collection and instead was estimated by equations based on sCr values,⁶ so it does not necessarily represent the actual GFR.²³ In addition, a change in renal function during the follow-up period was not considered in the present analysis. Finally, eGFR is generally calculated by equations of the Japanese Society of Nephrology⁶ in Japan, whereas it is estimated by the equations of the Modification of Diet in Renal Disease (MDRD)²⁴ or the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI)²⁵ worldwide. Therefore, the present results cannot necessarily be extrapolated to the subjects of other countries.

Conclusions

CKD classification was significantly associated with new-onset AF in the general Japanese population. Thus, it may be useful for risk stratification of new-onset AF. However, the predictive ability of CKD classification for new-onset AF was limited. Renal function evaluation using at least either eGFR or proteinuria category is recommended in health checkups.

Acknowledgments

We thank Mr. Junichi Murata for performing the statistical analyses. The present study was presented in part at the 67th Annual Scientific Session of the Japanese College of Cardiology (in Nagoya, Japan, September 13, 2019).

Source of Funding

The TAMA MED Project-AF and CKD are supported by the TAMA CITY Medical Association.

Disclosures

E.K. received remuneration from Daiichi-Sankyo, Bristol-Myers Squibb, and Ono Pharmaceutical. Other authors have declared that no conflicts of interest exist. W.S. is an Associate Editor of the Circulation Journal.

The study was partially supported by research funding from Boehringer Ingelheim, Astellas Pharma, Sumitomo Dainippon Pharma, Pfizer Japan, and Ono Pharmaceutical.

Data Availability

The deidentified participant data will not be shared.

IRB Information

The study protocols were approved by the Ethics Committee of the Nippon Medical School Tama-Nagayama Hospital with reference numbers 511 and 529.

Supplementary Files

Please find supplementary file(s);

http://dx.doi.org/10.1253/circj.CJ-20-0329

References

• 1.   Wolf PA, Abbott RD, Kannel WB. Atrial fibrillation as an independent risk factor for stroke: The Framingham Study. Stroke 1991; 22: 983–988.
• 2.   Feinberg WM, Blackshear JL, Laupacis A, Kronmal R, Hart RG. Prevalence, age distribution, and gender of patients with atrial fibrillation: Analysis and implications. Arch Intern Med 1995; 155: 469–473.
• 3.   Kodani E. Early detection of atrial fibrillation is the first step to prevent cardiogenic stroke: Usefulness of B-type natriuretic peptide. J Cardiol 2014; 65: 451–452.
• 4.   Go AS, Chertow GM, Fan D, McCulloch CE, Hsu CY. Chronic kidney disease and the risks of death, cardiovascular events, and hospitalization. N Engl J Med 2004; 351: 1296–1305.
• 5.   Nakayama M, Metoki H, Terawaki H, Ohkubo T, Kikuya M, Sato T, et al. Kidney dysfunction as a risk factor for first symptomatic stroke events in a general Japanese population: The Ohasama study. Nephrol Dial Transplant 2007; 22: 1910–1915.
• 6.   Matsuo S, Imai E, Horio M, Yasuda Y, Tomita K, Nitta K, et al. Revised equations for estimated GFR from serum creatinine in Japan. Am J Kidney Dis 2009; 53: 982–992.
• 7.   Levey AS, Eckardt KU, Tsukamoto Y, Levin A, Coresh J, Rossert J, et al. Definition and classification of chronic kidney disease: A position statement from Kidney Disease: Improving Global Outcomes (KDIGO). Kidney Int 2005; 67: 2089–2100.
• 8.   K/DOQI clinical practice guidelines for chronic kidney disease: Evaluation, classification, and stratification. Am J Kidney Dis 2002; 39: S1–S266.
• 9.   Watanabe H, Watanabe T, Sasaki S, Nagai K, Roden DM, Aizawa Y. Close bidirectional relationship between chronic kidney disease and atrial fibrillation: The Niigata preventive medicine study. Am Heart J 2009; 158: 629–636.
• 10.   Shang W, Li L, Huang S, Zeng R, Huang L, Ge S, et al. Chronic kidney disease and the risk of new-onset atrial fibrillation: A meta-analysis of prospective cohort studies. PLoS One 2016; 11: e0155581.
• 11.   Molnar AO, Eddeen AB, Ducharme R, Garg AX, Harel Z, McCallum MK, et al. Association of proteinuria and incident atrial fibrillation in patients with intact and reduced kidney function. J Am Heart Assoc 2017; 6: e005685.
• 12.   Japanese Society of Nephrology. Clinical practice guidebook for diagnosis and treatment of chronic kidney disease 2012. Tokyo: Tokyo Igakusha, 2012 (in Japanese).
• 13.   Kodani E, Kaneko T, Fujii H, Nakamura H, Sasabe H, Tamura Y, et al. Prevalence and incidence of atrial fibrillation in the general population based on national health insurance special health checkups: TAMA MED Project-AF. Circ J 2019; 83: 524–531.
• 14.   Kaneko T, Kodani E, Fujii H, Nakamura H, Sasabe H, Tamura Y, et al. Medical and dental visits of chronic kidney disease-diagnosed participants analyzed from the specific health checkups results in Japan: TAMA MED Project-CKD. J Clin Med Res 2020; 12: 115–121.
• 15.   Fujii H, Kodani E, Kaneko T, Nakamura H, Sasabe H, Tamura Y. “Yubi-wakka” (Finger-Ring) test: A tool to detect prefrailty in elderly populations, a pilot study. J Clin Med Res 2019; 11: 623–628.
• 16.   Kidney Disease: Improving Global Outcomes (KDIGO) CKD Work group. KDIGO 2012 clinical practice guideline for the evaluation and management of chronic kidney disease. Kidney Int Suppl 2013; 3: 1–150.
• 17.   DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: A nonparametric approach. Biometrics 1988; 44: 837–845.
• 18.   Oda E, Watanabe K. Japanese criteria of metabolic syndrome. Circ J 2006; 70: 364.
• 19.   Benjamin EJ, Levy D, Vaziri SM, D’Agostino RB, Belanger AJ, Wolf PA. Independent risk factors for atrial fibrillation in a population-based cohort: The Framingham Heart Study. JAMA 1994; 271: 840–844.
• 20.   Kokubo Y, Watanabe M, Higashiyama A, Nakao YM, Kusano K, Miyamoto Y. Development of a basic risk score for incident atrial fibrillation in a Japanese general population: The Suita Study. Circ J 2017; 81: 1580–1588.
• 21.   Yonezawa Y, Horinaka S, Shirakawa C, Kogure Y. Estimated glomerular filtration ratio is a better index than creatinine clearance (Cockcroft-Gault) for predicting the prevalence of atrial fibrillation in the general Japanese population. Hypertens Res 2018; 41: 451–459.
• 22.   Ohyama Y, Imai M, Kurabayashi M. Estimated glomerular filtration rate and proteinuria are separately and independently associated with the prevalence of atrial fibrillation in general population. PLoS One 2013; 8: e79717.
• 23.   Chancharoenthana W, Wattanatorn S, Vadcharavivad S, Eiam-Ong S, Leelahavanichkul A. Agreement and precision analyses of various estimated glomerular filtration rate formulae in cancer patients. Sci Rep 2019; 9: 19356.
• 24.   Levey AS, Bosch JP, Lewis JB, Greene T, Rogers N, Roth D. A more accurate method to estimate glomerular filtration rate from serum creatinine: A new prediction equation. Modification of Diet in Renal Disease Study Group. Ann Intern Med 1999; 130: 461–470.
• 25.   Levey AS, Stevens LA, Schmid CH, Zhang YL, Castro AF 3rd, Feldman HI, et al. A new equation to estimate glomerular filtration rate. Ann Intern Med 2009; 150: 604–612.