The relationship between homocysteine and new-onset atrial fibrillation in patients with acute myocardial infarction

Abstract

Objective

Elevated homocysteine (Hcy) are linked to an augmented risk of cardiovascular disorders. However, relationship between Hcy and in-hospital new-onset atrial fibrillation (NOAF) among patients experiencing acute myocardial infarction (AMI) remains uncertain. This study investigated the association between Hcy and incidence of NOAF in patients with AMI.

Materials and Methods

We conducted a retrospective cohort study, involving 3,347 patients with AMI. Participants were categorized into tertiles on the basis of Hcy levels during admission, and the outcome was the incidence of NOAF. Logistic regression models were employed to explored the associations between Hcy and NOAF when adjusting for potential confounders. Non-linear relationships were explored using generalized additive models (GAMs) and threshold effect analyses. Stratified analyses were performed to explore the robustness and potential interaction effects of other clinical factors on Hcy-NOAF relationship.

Results

Higher Hcy were consistently linked with a greater likelihood of in-hospital NOAF across all models (OR: 1.03, p < 0.0001). GAM analyses revealed a nonlinear association with a steeper increase in NOAF risk at higher Hcy levels. Threshold effect analysis identified a turning point at 34.27 μmol/L and below this point the association between Hcy and NOAF was stronger. Stratified analyses indicated a stronger in patients with smaller left atrial anterior-posterior diameter (LAAD).

Conclusion

A nonlinear association has been identified between elevated Hcy and the likelihood of in-hospital NOAF among individuals diagnosed with AMI, particularly among those with smaller LAAD measurements.

Introduction

It is common for patients with acute myocardial infarction (AMI) to develop atrial fibrillation (AF) [1,2]. New-onset AF (NOAF) occurring in approximately 4% to 12% of AMI cases [3–5]. AF onset during or after AMI exacerbates the overall cardiac burden, complicating patient management and worsening prognosis. Clinical outcomes are significantly worse in patients with NOAF after AMI, including increased heart failure incidence, greater thromboembolic risk, and higher mortality [5,6]. The clinical significance of NOAF in this patient population has been highlighted in several large-scale studies [7–9].

NOAF is known to be affected by clinical factors like age, left ventricular dysfunction, and heart failure, but the role of biological metabolites has not been fully elucidated [10]. The amino acid homocysteine (Hcy) has gained attention because of its associations with cardiovascular diseases [11]. Inflammation, oxidative stress and endothelial dysfunction all play a role in hyperhomocyteinemia (HHcy) -related atherosclerosis, stroke, and heart failure [12,13]. Our previous investigations also revealed that Hcy heightens the vulnerability of atherosclerotic plaques by promoting lipid accumulation and inflammatory response in macrophages [14,15].

Involvement of Hcy in arrhythmogenesis, particularly its relationship with NOAF in individuals experiencing AMI, remains unclear. Some observational studies suggest that heightened levels of Hcy may be linked to a greater likelihood of developing AF [16–18]. However, evidence from genetic and mechanistic studies is inconclusive [19,20]. The current study investigated the correlation between plasma Hcy and occurrence of in-hospital NOAF in patients admitted for AMI.

Materials and methods

Study population

A total of 4,212 patients were consecutively enrolled in this retrospective cohort study who had been diagnosed with AMI and underwent coronary angiography from January 2016 to December 2019. Data were retrieved from General Hospital of Ningxia Medical University’s big data research platform and information system.

AMI is diagnosed [21] when there is evidence of acute myocardial injury in addition to evidence of ischemia. An increase in cardiac troponin (cTn) levels exceeding the 99th percentile upper reference limit is required. The diagnosis is confirmed by signs of ischemia, new ischemic electrocardiogram (ECG) changes, pathological Q waves, and imaging indications of new myocardial damage or regional wall motion irregularities, or detection of coronary thrombi. ST-segment elevation myocardial infarction (STEMI) is characterized by new ST-segment elevation on the ECG in a minimum of two adjacent leads. Non-ST-segment elevation myocardial infarction (NSTEMI) denotes myocardial ischemia without ST-segment elevation on ECG.

Exclusion criteria: a history of atrial fibrillation (55 patients were excluded), ablation for arrhythmia (20 patients were excluded), valvular heart disease (86 patients were excluded), a previous diagnosis of AMI (399 patients were excluded), myocarditis (28 patients were excluded), hyperthyroidism (59 patients were excluded), and autoimmune disease or tumors (48 patients were excluded). Additionally, patients who were taking folic acid or vitamin B supplements within six months (98 patients), or whose baseline Hcy data were missing (72 patients) were excluded. Overall, 3,347 patients with AMI were recruited.

This study was approved by the Ethics Committee of the General Hospital of Ningxia Medical University (Ethics Review Number: 2020771) and was in accordance with the Declaration of Helsinki. The data are anonymous, and the requirement for informed consent was therefore waived by Institutional Review Board of General Hospital of Ningxia Medical University because of the retrospective nature.

Data collection

Information for this study included a variety of clinical, laboratory, ECG, medication, and coronary angiography parameters.

Baseline characteristics: Information on age, sex, and body mass index (BMI) was gathered in all participants. Medical history, including hypertension, diabetes, dyslipidaemia, stroke, post-myocardial infarction, AF, and history of percutaneous coronary intervention (PCI), was obtained. Smoking and alcohol consumption were also recorded. Other baseline information included the time of first medical contact (FMC), Killip classification, initial heart rate, systolic blood pressure (SBP), and diastolic blood pressure (DBP) upon admission.

Laboratory Data: Initial laboratory results were collected on admission before coronary angiography. White blood cell (WBC,10⁹/L), red blood cell (RBC,10¹²/L), platelet (10⁹/L), monocyte (10⁹/L), neutrophil (10⁹/L), alanine aminotransferase (ALT, U/L), aspartate aminotransferase (AST, U/L), creatinine (μmol/L), glucose (mmol/L), and D-dimer (µg/ml) were biochemically measured using a Siemens ADVIA 2400. Lipid profiles, including triglycerides (TG, mmol/L), total cholesterol (TC, mmol/L), low-density lipoprotein cholesterol (LDL-C, mmol/L), and high-density lipoprotein cholesterol (HDL-C, mmol/L), were measured via enzyme colorimetric methods. Plasma Hcy levels were measured by a circulating enzyme assay. All samples were processed and analysed at the Department of Laboratory Medicine, General Hospital of Ningxia Medical University as a routine admission laboratory test.

Echocardiographic parameters: All echocardiographic parameters were collected via the first echocardiography during hospitalization performed using a GE Vivid E9 ultrasound system. LAAD was measured using M-mode echocardiography in the parasternal long-axis view, measuring the maximum distance from the posterior edge of the interatrial septum to the inner border of the posterior left atrial wall during cardiac systole. LVEF was primarily assessed using Simpson’s biplane method in the apical four-chamber and two-chamber views.

Medication Use: Data concerning whether patients received dual antiplatelet therapy (aspirin plus clopidogrel or ticagrelor), statins, beta-blockers, and angiotensin-converting enzyme inhibitors (ACEIs)/angiotensin II receptor blockers (ARBs), anticoagulants or thrombolysis within the first 24 h of admission were recorded.

Coronary angiography and PCI data: Information on whether PCI treatment or primary PCI (time from symptom onset to first balloon inflation < 12 h) were recorded.

Homocysteine categorization

Plasma Hcy levels were measured at admission using automated clinical analysers. To investigate the potential non-linear relationship between Hcy and NOAF, we employed two complementary classification approaches. First, patients were divided into tertiles based on their admission Hcy concentration. The tertile cutoff points were determined by dividing the study population into three equal-sized groups: T1 (6.61–16.12 μmol/L, n = 1,116), T2 (16.13–23.72 μmol/L, n = 1,115), and T3 (23.73–134.38 μmol/L, n = 1,116). This classification approach allows for the investigation of potential threshold effects and provides clinically meaningful categorization without assuming linearity. This method is widely used in cardiovascular research to detect threshold effects and non-linear associations without imposing arbitrary cutoff values. This method is consistent with previous cardiovascular studies examining biomarker relationships and facilitates both statistical analysis and clinical interpretation of the results. Second, patients were classified based on the clinically established definition of HHcy, defined as plasma Hcy ≥ 15 μmol/L, resulting in normal Hcy (<15 μmol/L) and HHcy (≥ 15 μmol/L) groups. This binary classification aligns with previous clinical studies and facilitates comparison with established literature on Hcy-related cardiovascular outcomes.

Outcome assessment

NOAF is defined as the initial recorded instance of AF occurring during hospitalization after AMI, provided there is no prior history of AF. AF is diagnosed on the basis of ECG evidence of irregular RR intervals, the absence of distinct P waves, irregular atrial activation and episodes lasting at least 30 seconds²² regardless of whether it resolves spontaneously or requires medical intervention²³. All AMI patients underwent standard 12-lead or 18-lead ECG examination at admission, immediately after coronary angiography, and following PCI procedures. Continuous ECG monitoring was implemented for at least 72 h in all post-PCI patients and those with high-risk features (such as heart failure or hemodynamic instability). For patients reporting intermittent symptoms or showing concerning findings on standard ECG, 24-hour Holter monitoring was performed. Additionally, prompt 12-lead ECG examinations were conducted whenever patients reported symptoms potentially indicative of arrhythmias (such as palpitations or dizziness). The process of patient selection and grouping based on Hcy tertiles (T1, T2, T3) is illustrated in Figure 1.

Figure 1.

Flowchart of the study population.

AMI: acute myocardial infarction; AF: atrial fibrillation; Hcy: homocysteine; T1: tertile 1; T2: tertile 2; T3: tertile 3.

Statistical analysis

All statistical analysis was conducted employing Empower Stats (X&Y solutions, Inc. Boston MA, USA) alongside R software (version 4.2.0). In categorical data, frequencies and percentages are reported, whereas continuous variables are presented as means and standard deviation (SD). The chi-square test was used to analyse categorical variables and one-way analysis of variance was used for continuous data.

The relationship between Hcy and the risk of NOAF was evaluated using both univariate and multivariate logistic regression analysis. Odds ratios (ORs) with 95% confidence intervals (CIs) were calculated to estimate the risk. Potential confounding factors were selected based on clinical relevance and statistical significance in univariate analysis. The variance inflation factor of variables greater than 10 was excluded from the analysis to examine a correlation between Hcy levels and NOAF risk in patients with AMI. Several multivariate models were constructed with progressive adjustment for potential confounders. Both continuous and categorical (tertiles and HHcy) representations of Hcy were analysed.

An analysis of the nonlinear correlation between Hcy and NOAF risk was conducted using the generalized additive model (GAM). Two piecewise linear regression models were employed to assess threshold effect of Hcy level on NOAF probability. Stratified analyses and interaction tests were performed based on covariates fully adjusted model to investigate factors influencing the link between the Hcy and NOAF risk. For sensitivity analysis, a smooth curve analysis was conducted using the GAM stratified by LAAD tertiles. Statistics significance is defined as p value < 0.05.

Results

Baseline characteristics

Based on Hcy tertiles, individuals were stratified into three groups: T1 (6.61–16.12 μmol/L), T2 (16.13–23.72 μmol/L), and T3 (23.73–134.38 μmol/L). As shown in Table 1, significant demographic and clinical differences emerge across the tertiles. A higher proportion of men and older patients were found in the T3 group (p < 0.001 for both). Occurrence of smoking increased with higher Hcy levels, from 55.02% at T1 to 68.13% at T3 (p < 0.001). Conversely, diabetes mellitus and dyslipidemia tended to decrease across the tertiles. A history of stroke was more prevalent in T3 subgroup (p < 0.001). No differences were found in the time of FMC (p = 0.749) or in the distribution of STEMI/NSTEMI cases (p = 0.680), but the Killip classification indicated more severe heart failure in T3 subgroup (p = 0.012). Laboratory findings revealed that RBC and platelet counts decreased with higher Hcy levels, whereas creatinine and D-dimer increased significantly in T3 subgroup. Glucose, TG, and TC levels tended to decrease with increasing Hcy levels (p < 0.001 for all). In terms of echocardiography, LAAD increased and LVEF decreased significantly with higher Hcy levels (p < 0.001 for both). In terms of medication use, the frequency of clopidogrel use increased and the use of ticagrelor decreased with higher Hcy levels. For coronary angiography data, PCI rates were observed decreased in T3 group (p = 0.003). Overall, 3,347 patients with AMI were recruited, with 288 patients developing NOAF during hospitalization. Notably, the incidence of NOAF increased significantly across the tertiles, from 3.32% at T1 to 16.65% at T3 (p < 0.001).

Table 1.

Baseline characteristics according to tertiles of the Hcy levels.

Variables	Hcy, μmol/L				p value
	Total (6.61–134.38)	T1 (6.61–16.12)	T2 (16.13–23.72)	T3 (23.73–134.38)
	n = 3347	n = 1116	n = 1114	n = 1117
Demographic and clinical factors
Male, n (%)	2599 (77.65%)	789 (70.70%)	869 (78.01%)	941 (84.24%)	<0.001
Age, years	59.85 ± 12.21	57.33 ± 11.35	60.85 ± 12.08	61.38 ± 12.77	<0.001
BMI, kg/m2	24.69 ± 3.41	24.84 ± 3.43	24.68 ± 3.40	24.55 ± 3.40	0.142
Smoking, n (%)	2053 (61.34%)	614 (55.02%)	678 (60.86%)	761 (68.13%)	<0.001
Alcohol consumption, n (%)	680 (20.32%)	231 (20.70%)	233 (20.92%)	216 (19.34%)	0.604
Hypertension, n (%)	1600 (47.80%)	514 (46.06%)	543 (48.74%)	543 (48.61%)	0.359
Diabetes mellitus, n (%)	670 (20.02%)	312 (27.96%)	212 (19.03%)	146 (13.07%)	<0.001
Dyslipidemia, n (%)	1482 (44.28%)	540 (48.39%)	485 (43.54%)	457 (40.91%)	0.001
Stroke, n (%)	348 (10.40%)	91 (8.15%)	110 (9.87%)	147 (13.16%)	<0.001
Time of FMC, hours	14.42 ± 38.72	14.80 ± 39.58	14.76 ± 38.03	13.70 ± 38.58	0.749
Diagnosis, n (%)					0.680
NSTEMI	705 (21.08%)	244 (21.90%)	233 (20.92%)	228 (20.41%)
STEMI	2640 (78.92%)	870 (78.10%)	881 (79.08%)	889 (79.59%)
Killip classification, n (%)					0.012
I	2703 (80.76%)	933 (83.60%)	899 (80.70%)	871 (77.98%)
II	481 (14.37%)	145 (12.99%)	162 (14.54%)	174 (15.58%)
III	84 (2.51%)	19 (1.70%)	30 (2.69%)	35 (3.13%)
IV	79 (2.36%)	19 (1.70%)	23 (2.06%)	37 (3.31%)
SBP, mmHg	122.03 ± 21.57	122.86 ± 21.07	123.08 ± 21.19	120.16 ± 22.31	0.002
DBP, mmHg	76.15 ± 13.92	76.75 ± 13.55	76.60 ± 13.50	75.10 ± 14.64	0.008
Heart rate, bpm	80.09 ± 15.52	80.24 ± 14.84	79.28 ± 15.12	80.74 ± 16.53	0.078
Laboratory data
WBC,109/L	10.30 ± 3.60	10.24 ± 3.47	10.24 ± 3.52	10.42 ± 3.79	0.384
RBC, 1012/L	4.65 ± 0.63	4.70 ± 0.60	4.66 ± 0.59	4.59 ± 0.68	<0.001
Platelet, 109/L	226.74 ± 70.76	231.25 ± 64.92	226.48 ± 72.89	222.50 ± 73.91	0.014
Monocytes, 109/L	0.55 ± 0.24	0.53 ± 0.23	0.55 ± 0.25	0.55 ± 0.24	0.084
Neutrophil, 109/L	7.77 ± 2.57	7.66 ± 2.49	7.73 ± 2.50	7.92 ± 2.72	0.058
ALT, U/L	59.80 ± 92.97	59.63 ± 85.12	55.93 ± 47.13	63.84 ± 128.23	0.134
AST, U/L	156.69 ± 216.19	148.45 ± 190.30	155.53 ± 177.71	166.10 ± 268.96	0.020
Creatinine, μmol/L	74.38 ± 40.25	67.56 ± 32.56	73.82 ± 39.16	81.76 ± 46.56	<0.001
Glucose, mmol/L	7.74 ± 3.68	8.39 ± 4.27	7.68 ± 3.54	7.16 ± 3.02	<0.001
TG, mmol/L	1.69 ± 1.28	1.88 ± 1.46	1.64 ± 1.27	1.56 ± 1.05	<0.001
TC, mmol/L	4.24 ± 1.04	4.32 ± 1.04	4.24 ± 1.01	4.17 ± 1.05	0.003
LDL-C, mmol/L	2.23 ± 0.74	2.23 ± 0.75	2.23 ± 0.72	2.24 ± 0.77	0.860
HDL-C, mmol/L	0.93 ± 0.23	0.93 ± 0.24	0.93 ± 0.23	0.92 ± 0.23	0.283
D-dimer, μg/ml	0.60 ± 1.06	0.50 ± 0.87	0.57 ± 0.98	0.73 ± 1.27	<0.001
LAAD, mm	36.69 ± 4.55	36.06 ± 4.47	37.01 ± 4.45	37.00 ± 4.65	<0.001
LVEF, %	52.81 ± 9.69	53.92 ± 9.56	52.63 ± 9.63	51.90 ± 9.77	<0.001
Medications, n (%)
Aspirin	3327 (99.40%)	1110 (99.46%)	1106 (99.28%)	1111 (99.46%)	0.815
Clopidogrel	1560 (46.61%)	492 (44.09%)	497 (44.61%)	571 (51.12%)	0.001
Ticagrelor	1784 (53.30%)	623 (55.82%)	613 (55.03%)	548 (49.06%)	0.002
Statins	3335 (99.64%)	1111 (99.55%)	1111 (99.73%)	1113 (99.64%)	0.779
Beta blockers	2559 (76.46%)	857 (76.79%)	841 (75.49%)	861 (77.08%)	0.642
ACEI/ARB	1453 (43.41%)	500 (44.80%)	477 (42.82%)	476 (42.61%)	0.515
Anticoagulants	2810 (83.96%)	936 (83.87%)	933 (83.75%)	941 (84.24%)	0.947
Thrombolysis, n (%)	230 (6.87%)	78 (7.00%)	76 (6.82%)	76 (6.80%)	0.981
Coronary angiography
PCI, n (%)	2898 (86.59%)	995 (89.16%)	962 (86.36%)	941 (84.24%)	0.003
Primary PCI, n (%)	2056 (61.43%)	689 (61.74%)	684 (61.40%)	683 (61.15%)	0.959
Outcome
NOAF, n (%)	288 (8.60%)	37(3.32%)	65(5.83%)	186(16.65%)	<0.001

Data are expressed as the means ± SDs, median (interquartile ranges), or numbers (percentages).

Abbreviations: Hcy, homocysteine; BMI, body mass index; FMC, first medical contact; NSTEMI, non-ST-segment elevation myocardial infarction; STEMI, ST-segment elevation myocardial infarction; SBP, systolic blood pressure; DBP, diastolic blood pressure; WBC, white blood cells; RBC, red blood cells; ALT, alanine aminotransferase; AST, aspartate aminotransferase; TG, triglycerides; TC, total cholesterol; LDL-C, low-density lipoprotein cholesterol; HDL-C, high-density lipoprotein cholesterol; LAAD, left atrial anterior-posterior diameter; LVEF, left ventricular ejection fraction; PCI, percutaneous Coronary intervention; NOAF, new-onset atrial fibrillation.

As shown in Table S1, NOAF patients had a significantly higher incidence of hypertension and stroke, a lower incidence of dyslipidemia, a higher Killip classification, a lower SBP, and a faster heart rate. Furthermore, patients with NOAF had higher levels of certain laboratory markers such as neutrophils, AST, creatinine, Hcy and D-dimer, as well as lower levels of RBC, TG, TC, LDL-C and calcium. Patients with NOAF exhibited increased LAAD and decreased LVEF. Regarding medication use, patients with NOAF had significantly higher rates of clopidogrel use (55.90% vs. 45.73%, p < 0.001), lower rates of ticagrelor use (43.75% vs. 54.20%, p < 0.001), and higher rates of anticoagulant therapy (88.19% vs. 83.56%, p = 0.040) compared to those without NOAF. The result of boxplot, as depicted in Figure 2, indicated that patients with NOAF exhibited higher levels of Hcy compared to those without NOAF.

Figure 2.

Boxplot for hcy levels stratified by NOAF presence or absence in patients with AMI.

Univariate logistic regression analysis for in-hospital NOAF

As shown in Table S2, age, hypertension, stroke and severe heart failure (Killip II-IV) increased NOAF risk. Higher Hcy levels, WBC counts, and D-dimer levels linked a greater risk of NOAF, while RBC and TG levels linker to a decreased risk. Echocardiographic findings showed that increased LAAD (OR: 1.08, p < 0.001) was significantly linked to NOAF, whereas higher LVEF (OR: 0.98, p < 0.001) protected against NOAF. Medications such as clopidogrel and anticoagulants were linked to an elevated risk of NOAF, whereas ticagrelor was correlated to a lower risk. PCI was protective against NOAF (OR: 0.49, p < 0.001).

Multivariate logistic regression analysis of various models

Table 2 shows that all models examined consistently associated Hcy levels with a higher probability of NOAF. We constructed four progressively adjusted models to evaluate the robustness of this association: Model 1 was unadjusted to show the crude association between Hcy and NOAF. Model 2 was adjusted for basic demographic factors including sex and age. Model 3 was further adjusted for clinically important comorbidities and vital signs at admission, including hypertension, diabetes mellitus, dyslipidemia, stroke, Killip classification, SBP, and heart rate. Model 4, our fully adjusted model, included all variables from Model 3 plus laboratory parameters (WBC, neutrophil, AST, D-dimer, TG, TC, LDL-C, HDL-C) and echocardiographic measurements (LAAD and LVEF).

Table 2.

Multivariate logistic regression analysis of the association between hcy and in-hospital NOAF.

Exposure	Model 1	Model 2	Model 3	Model 4
	OR (95%CI) p	OR (95%CI) p	OR (95%CI) p	OR (95%CI) p
Hcy	1.03 (1.03, 1.04) <0.001	1.03 (1.03, 1.04) <0.001	1.03 (1.03, 1.04) <0.001	1.03 (1.03, 1.04) <0.001
HHcy
No	Reference	Reference	Reference	Reference
Yes	4.28 (2.78, 6.61) <0.001	3.50 (2.25, 5.44) <0.001	3.35 (2.15, 5.23) <0.001	3.83 (2.32, 6.31) <0.001
Hcy tertile
Low	Reference	Reference	Reference	Reference
Middle	1.81 (1.20, 2.73) 0.005	1.55 (1.02, 2.35) 0.039	1.51 (0.99, 2.31) 0.054	1.56 (0.99, 2.46) 0.057
High	5.83 (4.05, 8.38) <0.001	4.91 (3.38, 7.12) <0.001	4.66 (3.19, 6.80) <0.001	5.01 (3.32, 7.57) <0.001
p for trend	2.59 (2.17, 3.09) <0.001	2.42 (2.02, 2.91) <0.001	2.35 (1.96, 2.84) <0.001	2.46 (2.01, 3.01) <0.001

Model 1: adjusted for none.

Model 2: adjusted for sex and age.

Model 3: adjusted for sex, age, hypertension, diabetes mellitus, dyslipidemia, stroke, Killip classification, SBP and heart rate.

Model 4: adjusted for sex, age, hypertension, diabetes mellitus, dyslipidemia, stroke, Killip classification, SBP, heart rate, WBC, neutrophil, AST, D-dimer, TG, TC, LDL-C, HDL-C, LAAD and LVEF.

As a continuous variable, Hcy consistently showed a significant association with in-hospital NOAF across all models. The OR for Hcy was 1.03 (p < 0.001) in all models. At the bivariate level (HHcy: Hcy ≥ 15 μmol/L), patients with HHcy exhibited a markedly elevated likelihood of NOAF than those without HHcy. After adjusting for various confounding factors in Model 4, the OR remained significant at 3.83 (95% CI: 2.32-6.31, p < 0.001). For the categorical variable (Hcy was divided into three groups by tertiles), patients in the highest tertile had a heightened of NOAF probability. Even after full adjustment in Model 4, association remained significant, with an OR of 5.01 (p < 0.001). The trend test for Hcy tertiles was significant across all models, which indicated a dose-response correlation between elevated Hcy concentrations and NOAF occurrence.

Relationship between hcy and NOAF risk in patients with AMI

To further investigate the potential non-linear relationship between Hcy levels and NOAF risk, we performed GAM analysis using the fully adjustment Model 4 as described in multivariate analysis, which adjusted for sex, age, hypertension, diabetes mellitus, dyslipidemia, stroke, Killip classification, SBP, heart rate, WBC, neutrophil, AST, D-dimer, TG, TC, LDL-C, HDL-C, LAAD and LVEF. Figure 3 illustrates a nonlinear relationship between Hcy and NOAF probability in individuals diagnosed with AMI. As Hcy levels increase, the likelihood of NOAF increases significantly. Adjustment of the model for various confounders highlights the robust correlation between elevated Hcy and incidence of AF in individuals with AMI.

Figure 3.

Associations between the Hcy levels and in-hospital NOAF.

Threshold effect analysis shown in Table 3 demonstrates a noteworthy nonlinear correlation between Hcy and occurrence of in-hospital NOAF. In Model I (linear analysis), every unit increase in Hcy correlated with 3% elevated risk of NOAF (OR: 1.03, p < 0.001), while a SD increase in Hcy was linked to a 71% heightened risk (OR: 1.71, p < 0.001). Model II (nonlinear analysis) identified a turning point (K) at Hcy concentration of 34.27 μmol/L. Below this threshold, the risk of NOAF rose significantly with Hcy increase (OR: 1.09, p < 0.001), and each SD increase raising risk over 4-fold (OR: 4.16, p < 0.001). However, beyond this threshold, the association weakened, with 1% increase per unit (OR: 1.01, p = 0.013), and a 23% increase per SD (OR: 1.23, p = 0.013). The logarithm likelihood ratio test (LRT) confirmed that the nonlinear model (Model II) provided a significantly better fit than the linear model (Model I) (p < 0.001).

Table 3.

Analysis of threshold effect between Hcy levels and in-hospital NOAF.

Models	Hcy per-unit increase		Hcy per-SD increase
	OR (95%CI)	p value	OR (95%CI)	p value
Model I
One line effect	1.03 (1.03, 1.04)	<0.001	1.71 (1.54, 1.90)	<0.001
Model II
Turning point (K)	34.27		0.52
Hcy < K	1.09 (1.06, 1.11)	<0.001	4.16 (2.90, 5.98)	<0.001
Hcy > K	1.01 (1.00, 1.02)	0.013	1.23 (1.05, 1.45)	0.013
p value for LRT test*		<0.001		<0.001

Model I, linear analysis; Model II, nonlinear analysis. Adjusted for: sex, age, hypertension, diabetes mellitus, dyslipidemia, stroke, Killip classification, SBP, heart rate, WBC, neutrophil, AST, D-dimer, TG, TC, LDL-C, HDL-C, LAAD and LVEF). Model II differs significantly from Model I by the logarithm likelihood ratio test (LRT) by p < 0.05.

Association between the Hcy levels and in-hospital NOAF was identified nonlinear, after adjusting for all covariates in Model 4 (sex, age, hypertension, diabetes mellitus, dyslipidemia, stroke, Killip classification, SBP, heart rate, WBC, neutrophil, AST, D-dimer, TG, TC, LDL-C, HDL-C, LAAD and LVEF). Estimated values are indicated by the red lines, while the blue lines depict the associated 95% CIs.

Subgroup analysis and interpretation of the interaction results

Hcy were associated with NOAF risk in the subgroup analysis in Figure 4, with ORs ranging from approximately 1.02 to 1.06 per unit increase in Hcy. Notably, significant interactions were found for LAAD (p = 0.041 for interaction). Specifically, this relationship was more pronounced in patients with smaller LAADs, defined as LAAD less than 38 mm, which represents the median value in our study population and aligns with the upper limit of normal LA size according to current echocardiographic guidelines [24]. No noteworthy interactions were detected in the other subgroups, including gender, age, smoking status, hypertension, diabetes mellitus, stroke, LVEF, or medication use (clopidogrel, ticagrelor, and anticoagulants), which suggests that the effect of Hcy on AF risk is relatively consistent across these clinical characteristics and treatment regimens.

Figure 4.

Subgroup analysis based on Hcy levels and in-hospital NOAF in patients with AMI.

Adjustment was made considering all variables outlined in Model 4 except for the stratification element (sex, age, hypertension, diabetes mellitus, dyslipidemia, stroke, Killip classification, SBP, heart rate, WBC, neutrophil, AST, D-dimer, TG, TC, LDL-C, HDL-C, LAAD and LVEF).

Boxplots in Figure 5 exhibits that LAAD levels was higher in group with NOAF. Meanwhile, LAAD levels were lager in middle and high group of Hcy tertiles when compared with the low group. Figure 6 shows the correlation between Hcy and incidence NOAF stratified by LAAD. Patients with the largest LAAD (dotted blue line) showed a weaker association between Hcy levels and NOAF risk, whereas those with the smallest LAAD (solid red line) experienced the steepest increase in NOAF probability with increasing Hcy levels. The middle LAAD group (dashed green line) presented an intermediate increase in risk.

Figure 5.

Boxplots of LAAD levels in different groups.

A: LAAD levels in patients with AMI stratified by the presence or absence of NOAF.

B: LAAD levels in patients with AMI grouped by Hcy tertiles.

Figure 6.

Associations between Hcy levels and in-hospital NOAF stratified by LAAD levels.

Every stratification adjustment was made considering all covariates except for LAAD (sex, age, hypertension, diabetes mellitus, dyslipidemia, stroke, Killip classification, SBP, heart rate, WBC, neutrophil, AST, D-dimer, TG, TC, LDL-C, HDL-C, LAAD and LVEF).

Discussion

The main findings of this study are (Figure 7): 1. A notable independent relationship exists between elevated plasma Hcy and incidence of in-hospital NOAF among individuals diagnosed with AMI; 2. A nonlinear relationship with a threshold effect was observed with a strong association at Hcy levels below 34.27 μmol/L; 3. This relationship was more pronounced in patients with smaller LA size (LAAD < 38 mm).

Figure 7.

Graphical abstract (created by biorender.com).

This retrospective cohort study analyzed 3347 cases of patients with AMI, of which 288 cases development NOAF during hospitalization (A). In patients with AMI, Hcy levels were found to be nonlinearly related to the incidence of NOAF (B). Across different models, HHcy was an independent risk factor for NOAF in patients with AMI (C).

In our cohort of 3,347 patients with AMI, 288 patients (8.6%) developed in-hospital NOAF, which is consistent with previous studies reporting NOAF incidence rates ranging from 4% to 12% in patients with AMI [3–5]. This variation is influenced by differences in study populations, diagnostic criteria for AF, the timing of AF detection and type of AMI, and in-hospital or long-term follow-up outcomes. The current results fall within this range and reflect the specific characteristics of this cohort and the strict diagnostic criteria that was used.

The emergence of NOAF in individuals with AMI is of considerable clinical importance, as it correlates with a poor prognosis, including heart failure, stroke, bleeding, thrombosis, and overall mortality. The CREDO-Kyoto AMI Registry Wave-2 study [25] examined 6,228 Japanese AMI patients undergoing PCI and found that 7.9% developed NOAF during hospitalization, with significant impacts on mortality and stroke over the 5.5-year median follow-up. In the present, NOAF was slightly more prevalent during hospitalization following AMI than previously reported, which may be due to the complicated condition of patients in the acute stage of AMI and the perioperative period of PCI. The relatively high incidence observed highlights the need for routine monitoring and early intervention to prevent serious complications associated with NOAF.

Notwithstanding the significance of Hcy as a cardiovascular risk factor, including AF, its involvement in NOAF following AMI has been rarely been reported [26]. The present study revealed that older age, severe heart failure (Killip II-IV), higher WBC, greater LAAD and lower LVEF were not only been observed in higher Hcy levels, as shown in Table 1, but also related to the risk of NOAF after AMI (Table 2). Age, heart failure, inflammation, and atrial remodeling are considered as traditional risk factors for AF [27–29]. Results from the present investigation highlight the significant role of elevated Hcy levels as an independent risk factor, which can be explained through several potential mechanisms. Hcy induces endothelial dysfunction, promotes oxidative stress, and generates reactive oxygen species (ROS), which cause atrial remodelling and fibrosis [30]. Elevated Hcy also increases inflammation, enhances pro-inflammatory cytokines and promotes a prothrombotic state that increases AF incidence, especially in individuals with AMI [31–33]. These combined effects contribute to NOAF in patients with AMI and highlight the importance of monitoring and potentially targeting Hcy to reduce AF risk in this population.

As a modifiable risk factor for AF, elevated Hcy is identified in this study as similar to other existing studies. The ARIC and MESA cohort study by Kubota et al. demonstrated a significant association between elevated Hcy and incident AF across 7,133 participants, with each log₂-unit increase in Hcy associated with a 27% higher AF risk (HR 1.27, 95% CI 1.01–1.61) [34]. Similarly, meta-analysis of 11 studies by Rong et al. confirmed that patients with elevated Hcy had a significantly higher risk of AF (OR 2.21, 95% CI 1.16–4.21) and AF recurrence (OR 3.81, 95% CI 3.11-4.68) compared to those with lower Hcy levels [35]. Patients with paroxysmal and persistent AF have notably elevated Hcy levels. An earlier review indicated that Hcy may serve as a predictor of paroxysmal AF [18]. Consistent with our results, these investigations indicate that individuals exhibiting increased Hcy levels have a heightened propensity for developing AF and reinforce the involvement of Hcy as a significant cardiovascular risk factor.

A noteworthy aspect of this study is its application of GAM and threshold effect analysis to verify the nonlinearity between Hcy and NOAF risk. This analytical approach allowed us to identify a critical threshold of Hcy at 34.27 μmol/L, where the association with NOAF risk becomes more significant. This threshold was determined through a systematic statistical approach using two-piecewise logistic regression models. The optimal threshold (34.27 μmol/L) was identified as the point that maximized the likelihood ratio statistic when comparing the two-piecewise model against a single-line logistic regression model. This statistically derived threshold has clinical relevance as it substantially exceeds the commonly recognized upper limit of normal Hcy (15 μmol/L), suggesting that a considerable elevation of Hcy is required before the risk of NOAF increases significantly. Many studies, including those by Kubota et al. [34] and Rong et al. [35], have treated Hcy as a continuous risk factor with a linear association with AF; whereas our study provides more robust results based on Hcy as a diverse variable type. The risk of AF increases sharply at moderate Hcy concentrations but plateaus at higher concentrations, which suggests a threshold effect that has not been extensively explored in earlier research.

Additionally, the current study provides another novel insight into how structural factors, specifically LAAD interact with Hcy levels to influence AF occurrence. The relationships among Hcy, LAAD and AF are complex and important. The risk of AF is heightened with elevated Hcy, potentially due to mechanisms, such as endothelial dysfunction, oxidative stress, and atrial remodelling [36], whereas these processes contribute to structural changes in the atria [37], including enlargement of the LAAD, which is a known predictor of AF. In addition, elevated Hcy has been shown to influence myocardial fibrosis, which contributes to electrical remodelling and further increases the risk of AF [38]. This relationship suggests that Hcy not only acts as a risk factor but also interacts with structural heart changes to amplify the occurrence of AF. Patients with AMI already suffering from structural and biochemical deficits, the combined effect of Hcy and LAAD enlargement could significantly increase the risk of developing AF [39]. Therefore, careful monitoring of both Hcy levels and LAAD could be critical in assessing and managing AF risk in this population. In the current investigation, it was observed that LAAD was markedly greater in the higher Hcy group. Both LAAD and Hcy were identified as risk factors for AF. Importantly, after adjustment for multiple confounders, including LAAD and LVEF, Hcy remained independently associated with NOAF in patients with AMI. Interestingly, association was much stronger in patients with smaller LAADs. This may be explained that in patients with smaller LAADs may have less compromised atrial structure, which makes them more susceptible to the proarrhythmic effects of elevated Hcy. Conversely, in patients with larger LAADs, structural changes may dominate the risk and reduce the relative influence of Hcy. This distinction is particularly relevant in the acute post-AMI setting, where the pathophysiology of NOAF likely differs from that of chronic AF development seen in the general population. While chronic AF often follows progressive atrial enlargement, NOAF after AMI may be more dependent on acute triggers such as inflammation and oxidative stress—mechanisms through which Hcy exerts its pro-arrhythmic effects [40]. This finding highlights the importance of considering both Hcy levels and LAAD in NOAF risk stratification in patients with AMI, which suggests that the underlying and complex pathophysiological mechanisms warrant further investigation.

The clinical implications of these findings are substantial. Elevated Hcy levels may indicate higher NOAF risk in patients with AMI, particularly in those with a smaller LAAD. Targeted screening and early interventions, such as rhythm monitoring and Hcy-lowering therapies, could benefit these high-risk populations. Additionally, the threshold effect indicates that therapies reducing moderately elevated Hcy, such as folate and B-vitamin supplementation, might reduce NOAF risk. Validating these findings and exploring potential therapeutic strategies to mitigate NOAF risk in patients with AMI require further studies.

Several limitations exist in this study. First, its retrospective nature precludes establishing causality. Second, data from a single centre may limit the generalizability. Third, the diagnosis of NOAF was based on hospital records, which may introduce misclassification bias. Despite our comprehensive monitoring approach (including routine ECGs, continuous monitoring for high-risk patients, and symptom-triggered ECGs), the paroxysmal and sometimes asymptomatic nature of AF episodes may have resulted in some cases remaining undetected, potentially yielding conservative estimates of the association between elevated Hcy levels and NOAF. Additionally, residual confounding remains possible despite adjustments for known confounders. Factors such as genetic predisposition, detailed medication adherence, and lifestyle variables like diet and physical activity were not fully accounted for and could influence both Hcy levels and NOAF risk. Although we found no significant interaction between Hcy and medication use in our subgroup analyses, we cannot completely rule out potential drug-drug interactions or the influence of timing, dosage, and changes in medication regimens during hospitalization that might affect both Hcy metabolism and arrhythmogenesis.

Conclusions

The current investigation revealed a robust and nonlinear association between heightened Hcy and incidence of in-hospital NOAF in individuals with AMI, particularly in those with smaller LAADs. A threshold effect at 34.27 μmol/L offers fresh perspectives on the involvement of Hcy in the occurrence of AF. This study found a strong, nonlinear relation between elevated Hcy levels and the incidence of in-hospital NOAF in individuals with AMI. This was especially evident in those with smaller left atriums. Hcy below the threshold level of 34.27 μmol/L is particularly related to the occurrence of NOAF in those patients.

Funding Statement

This study was supported by the Postdoctoral research project of Shaanxi Province (2024BSHYDZZ004), the Clinical Research Funds of First Affiliated Hospital of Xi’an Jiaotong University (XJTU1AF2022LSL-014), the Key Research and Development General Project in Shaanxi Province, China (2023-YBSF-605) and the Natural Science Foundation of Ningxia, China (2023AAC02069).

Authorship contributions

CRediT: Ping Jin: Formal analysis, Funding acquisition, Methodology, Writing – original draft; Peng Wu: Formal analysis, Investigation; Juan Ma: Investigation; Yitong Bian: Formal analysis, Funding acquisition, Visualization; Huijuan Kou: Formal analysis, Funding acquisition; Xueping Ma: Funding acquisition, Project administration; Shaobin Jia: Methodology, Project administration; Qiangsun Zheng: Supervision, Writing – review & editing.

Ethics statement

This study was approved by the Ethics Committee of the General Hospital of Ningxia Medical University (Ethics Review Number: 2020771) and was in accordance with the Declaration of Helsinki. The data are anonymous, and the requirement for informed consent was therefore waived by Institutional Review Board of General Hospital of Ningxia Medical University because of the retrospective nature.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The datasets used in this study are available from the corresponding author on reasonable request.