Electrocardiographic Indicators of Left Atrial Remodeling in Ischemic Stroke Patients with Atrial Fibrillation

Introduction

Atrial fibrillation (AF) is a modern epidemic related to the aging of society, and is particularly dangerous due to the direct connection between arrhythmia and ischemic stroke (IS). AF is an arrhythmia that causes adverse remodeling of the left atrium (LAR), known as atrial cardiomyopathy (AC) [1]. LAR refers to the spectrum of pathophysiological changes in atrial structure and mechanical function and the electric, ionic, and molecular milieu of the left atrium (LA) that most often occurs in response to stresses imposed by conditions such as hypertension, heart failure, diabetes mellitus, and obesity [2]. Remodeling forms the basis of AC, defined by a multi-society, expert consensus statement as any complex of structural, architectural, contractile, or electrophysiological changes affecting the atria with the potential to produce clinically relevant manifestations [2]. LAR is the link between morphological changes and electrophysiological abnormalities that contribute to the triggering and maintenance of AF, creating a vicious circle between AC and cardiac arrhythmias. Therefore, the pathomechanism of arrhythmia resulting from structural, functional, and electrophysiological remodeling may be reflected in specific P wave changes in the electrocardiogram (ECG) [3,4]. The presence of atrial fibrosis and fatty infiltration is the anatomic substrate of both AF and advanced interatrial block [4]. In patients with AF and after IS, structural and functional remodeling of both LA and right (RA) atria (AR) causes large differences in P indices, which are a manifestation of electrophysiological AR [5,6]. Prolongation of the P wave duration ≥110 ms, the dispersion of P ≥40 ms, and the P wave terminal force (PWTF) in the V1 lead (PWTFV1) ≤4000 mcV*ms are risk factors for recurrent AF, which increases the risk of embolic complications [7,8]. PWTFV1, P wave duration, and maximum P wave area are useful electrocardiographic markers that can be used to stratify the risk of incident IS [9]. The ARCADIA trial (AtRial Cardiopathy and Antithrombotic Drugs in prevention After cryptogenic stroke) use one of these measures, PWTFV1, as a key eligibility criterion for identifying cryptogenic stroke patients with potential underlying AC [10]. Although several clinical studies have shown the value of some ECG P wave markers in assessing the risk of IS, patients with atrial abnormalities without evident AF are not placed on oral anticoagulation [11]. This is controversial because it is a major diagnostic and therapeutic inconsistency, which potentially poses a risk of IS. There is an unresolved gap – whether the indication for anticoagulation should be the occurrence of AF or whether anticoagulation should be started earlier, after the electrophysiological and structural AR has been detected. More important, however, are the potential clinical consequences of changing the paradigm of anticoagulation initiation.

The sex differences in AF are vast. There are conflicting data as to whether sex plays a role in the association of various risk factors and the development of AF [12]. There are also female-/male-specific factors that contribute to the development of AF [13]. There are many gaps in our knowledge of the sex differences in AF, and many opportunities for future research.

The main aim of this study was to assess differences in LA enlargement indicating LAR markers based on the occurrence of IS among patients with AF and to evaluate sex differences in these markers. The results of this study may suggest future research directions for screening tools for patients at increased risk of AF and cardioembolic complications.

Material and Methods

STUDY DESIGN AND DATA COLLECTION:

The study material consisted of the medical histories of patients hospitalized in the Department of Neurology for cardioembolic stroke with a history of AF. These patients were selected because AR is to be expected in this group. The study included a group of 731 patients, with documented AF and after IS, hospitalized between 2000 and 2020 at the Department of Neurology. The exclusion criteria were: undocumented AF in the ECG, no ECG with sinus rhythm (SR) in the peri-stroke period (±6 months), implanted pacemaker or cardioverter-defibrillator, hemodynamically significant heart defects, and unsatisfactory quality of the ECG recording. Finally, the data of 256 patients (183 females and 73 males) over 18 years of age were evaluated (Figure 1). The collected study material involved medical documentation, which included ECG and trans-thoracic echocardiography (TTE) performed at admission to the hospital, as well as data considered in the study: age and comorbidities. The control group consisted of 70 people (44 females and 26 males) aged 68.43(±7.38) years with AF but without IS. The selection of the control group was based on a study of people with AF and after IS hospitalized for AF in the Department of Cardiology. Women predominated in both the study group and the control group, which facilitates comparison between the 2 groups and corresponds to similar proportions of patients in the daily treatment of people with similar conditions. Patient data from the study and control groups were manually extracted from the medical records archives. Patient documentation also contained information about the time between first diagnosis of AF and IS (in the study group) or hospitalization (in the control group). The study assessed structural AR (eg, LA diameter, area) and electrical remodeling (P wave indices). Results were compared among both sexes due to the well-known differences between males and females. Structural remodeling was assessed by echocardiographic parameters and electrical remodeling was assessed based on ECG parameters. The ECG and echocardiographic parameters considered in the study are listed in detail below.

ELECTROCARDIOGRAPHIC ASSESSMENT:

An ECG was routinely performed upon admission of patients to the hospital using the Aspel AsCard Orange v.07.105 device. The ECG was set for a paper speed of 25 mm/s and sensitivity of 1 (1 cm=1 mV). The ECG measurements were always performed by the same cardiologist blinded to clinical data. A magnifying glass with a scale of 0.1 mm, a scale length of 10 mm, and a lens diameter of 15 mm were used to measure the P wave in the ECG recording. This enables the analysis of the morphology of the P wave in available ECGs with an accuracy of 0.1 mm, which corresponds to 0.004 s of the duration of the P wave and enables the measurement of the P amplitude with an accuracy of 0.01 mV. The P wave duration was measured from the beginning of the ECG curve until its return to the isoelectric line. P amplitude was measured from the isoelectric line to the peak of the P wave (positive P amplitude) and from the isoelectric line to the lowest point on the P wave curve (negative P amplitude). For biphasic P waves, positive and negative P amplitudes were reported. PWDI was measured manually by subtracting the minimal P wave (P_min) duration from the maximal P wave (P_max) duration (P_max - P_min), measured by 12-lead surface ECG, from a single beat, and mean values for 3 complexes were calculated. The duration of the negative phase of the P wave in the V₁ lead was measured from the onset of the negative P wave in V₁ to the return of the wave to the isoelectric line. The amplitude of the negative phase of P in V₁ was measured from the isoelectric line to the deepest point of the P wave. The PWTFV₁ is the product of negative phase duration of the P wave in V₁ (ms) and negative phase amplitude of the P wave in V₁ (mcV).

All 12 ECG leads were analyzed. Then, the most commonly used P wave indices were calculated from the available parameters (Figure 2) [3,4]:

ECHOCARDIOGRAPHIC PROTOCOL:

Echocardiographic examinations were performed upon admission to the hospital using a Philips Epiq 7c Ultrasound System with a Philips X5-1 Sector Array Probe. Echocardiographic measurements were always performed by the same cardiologist, a different one from the one analyzing the ECG recordings and blinded to the clinical data. Measurements and calculations were performed based on current recommendations for cardiac chamber quantification by echocardiography in adults [14]. The following parameters were analyzed (Figure 3):

ATRIAL FIBRILLATION BURDEN:

In the simplest sense, AF burden refers to the amount of AF that an individual has. Classical AF burden indicators include the percentage of AF time in a specific period of observation [15]. Due to the retrospective nature of the study, continuous recording using telemonitoring or retrospective analysis of records in implantable devices was impossible [15]. Patients with implanted intracardiac devices were excluded from the study because the stimulated P waves would have impaired the assessment of P wave indices and thus falsified the assessment of electrical AR. Under these conditions, an attempt was made to estimate the burden of AF based on available data. Estimated AF burden was defined as the time from first AF diagnosis to IS onset, defined as a percentage of lifetime (estimated AF% burden). The time from first AF diagnosis to stroke onset was defined as estimated AF burden. In other words, 2 alternative indicators were used, which could be precisely determined based on available documentation:

STATISTICAL ANALYSIS:

Categorical variables are presented as counts (percentages). Normality was tested using the Shapiro-Wilk test. All continuous variables had normally distributed data. Descriptive statistics were reported as means with standard deviations (SD) for continuous variables for normally distributed data. For comparative studies involving 2 groups, the Mann-Whitney U test (for independent samples) was used. Correlations between categories were examined using the chi-square test (χ²) with Yates’ correction. Univariate comparisons between patients with and without stroke were performed using the Wilcoxon rank-sum test due to non-normal data distribution, and were assessed via the Shapiro-Wilk test with p<0.05 indicating statistical significance.

A multivariable logistic regression model was fitted to assess predictors of IS occurrence, using a binomial distribution with a logic link function. Predictors were selected based on their established clinical relevance to IS risk in AF, as evidenced by significant associations in univariate analysis, and informed by prior literature and clinical guidelines. Model fit was evaluated using the Hosmer-Lemeshow goodness-of-fit test (10 groups), with a non-significant p value (p>0.05) indicating adequate fit. The Tjur’s R² was calculated to estimate explained variance, and the rank correlation coefficient (rho) assessed agreement between observed and predicted probabilities.

Multicollinearity was examined using variance inflation factors (VIFs) and tolerance values, with VIF <3 and tolerance >0.1 as thresholds for acceptable collinearity. The time between first AF diagnosis and hospitalization – to-age ratio (AF% burden) was excluded from the final model due to high multicollinearity (preliminary VIF >10.0) with age and time from AF to IS. Model discrimination was evaluated via the area under the receiver operating characteristic (ROC) curve (AUC), with 95% confidence intervals (CI) computed using the DeLong method.

STATISTICAL TOOL:

Analyses were conducted using the R Statistical language (version 4.3.3; R Core Team, 2024) on Windows 11 x64 (build 26100), using the packages sjPlot (version 2.8.15) [16], performance (version 0.12.3) [17], report (version 0.5.8;) [18], pROC (version 1.18.5) [19], gtsummary (version 1.7.2) [20], gofcat (version 0.1.2) [21], MASS (version 7.3.60.0.10 [22], and dplyr (version 1.1.4) [23].

ETHICAL CONSIDERATION:

The consent to conduct the research was obtained from the local bioethics committee (Bioethical Committee at the Medical University of Lublin) – KE-0254/27/2013. The research was conducted in accordance with the principles contained in the Declaration of Helsinki [24,25]. After being fully informed about the aims, research methods, possible adverse effects and potential scientific, medical, and social values of this project, informed consent was obtained from all individuals included in the study. Because this was a retrospective study conducted at a university clinical center, consent to use for research purposes the data routinely obtained during a standard hospitalization is obtained from each patient in the hospital documentation upon admission.

Results

Patients with IS were older and most (222, 86.72%) had hypertension, resulting in a significant difference to the control group. Nevertheless, it cannot be concluded that patients with IS were characterized by multimorbidity, as heart failure, diabetes and ischemic heart disease occurred in only 1/4 to 1/3 of patients and similarly frequently in both groups (Table 1). The results of the presented study confirm that LAE in the study group is reflected in higher PWTFV1 values. 52,34% of the subjects in study group had a rate above 4000 mcV*ms (Table 1). PWD in the lead II and PWD in the lead III had a statistically significantly lower value in the group with IS compared to the control group. In contrast, patients with IS were characterized by a significantly higher value of PWD in the lead V1. As a result, the calculated PWDI was statistically significantly higher in the study group. Pmax was significantly longer in the control group than in the IS group. The proportion of patients with all 4 indices in a normal range (Pmax/PWTFV1/PWDNV1/PWDI) was higher in the control group than in the study group. Taken together, these data imply that electrophysiological remodeling was more advanced in the study group than in the control group.

Comparative analysis between sexes showed that females were older than males. In contrast, there were no significant differences in the incidence of cardiovascular disease and diabetes (Table 2).

There were several significant differences in P wave indices among patients with IS based on sex. PWTFV1 and Pmin were higher in the male group than in the female group, whereas values of PWD in the lead V2 were longer among females (Table 2). The PWDNV1 was significantly shorter in females than in males. There were also interesting differences in the frequency of certain abnormalities between females and males. A statistically significantly larger percentage of women had a higher prevalence of PWD in the lead I ≥110 ms than in males (Table 1). A lower percentage of PWTFV1 <4000 mcV*ms was recorded in males. Males were less likely than females to have at least 1 of the studied parameters in the normal range (considering PWTFV1, Pmax, PWDNV1). Only 13.7% of males had all 3 parameters in the normal range, while among females the percentage was 34.1% (p=0.002) (Table 2). The differences between the sexes were statistically significant (p<0.05).

Most patients had enlarged atria on TTE. Females had a significantly smaller LA and higher mean LVEF than males (Table 2).

Table 3 presents a comparison of clinical data and echocardiographic and electrocardiographic parameters in the study group with IS and in the female control group. The analogous comparison of clinical data and echocardiographic and electrocardiographic parameters in the study group with IS and in the male control group is presented in Table 4.

Univariate analysis of patient characteristics in an AF cohort (Table 5) revealed significant differences between those with and without IS. Patients with IS were older (mean age 76.20 years, SD=9.64) compared to those without (mean 68.43 years, SD=7.38; W=6673.5, p<0.001).

The time between first AF diagnosis and hospitalization was significantly longer in the IS group (mean 2.33 years, SD=1.15) than in the no-IS group (mean 1.18 years, SD=0.83; W=6892.0, p<0.001), indicating that prolonged AF duration increases IS risk, potentially due to progressive AR (Table 5). Similarly, the AF% burden was higher in the IS group (mean 3.03%, SD 1.46) versus the no-IS group (mean 1.70%, SD=1.16; W=6688.0, p<0.001), indicating that the relative duration of estimated AF exposure, adjusted for age, may further amplify risk (Table 5).

To further evaluate the results, multivariable analysis was conducted. We assessed predictors of IS occurrence using a multivariable logistic regression model. The model included 153 observations and 8 predictors, explaining 42.7% of the variance in IS occurrence. Predictors were selected based on their established clinical relevance to IS risk in patients with AF, including demographic factors (sex, age), comorbidities (hypertension), echocardiographic measures (LVEF, LAD), electrocardiographic parameters (PWDV1, Pmax), and the AF burden.

Model fit was evaluated using the Hosmer-Lemeshow test (÷ (8)=5.17, p=0.739), indicating no evidence of lack of fit (H0 not rejected). The rank correlation coefficient (rho) between observed and predicted probabilities was 80%, demonstrating good predictive accuracy.

Multicollinearity was assessed using variance inflation factors (VIFs) and tolerance values (Table 6). All included predictors had VIFs below 2 (range: 1.26–1.99) and tolerance values from 0.50 to 0.79, indicating low multicollinearity. The 95% confidence intervals for VIF and tolerance further support model stability, with no values indicating significant collinearity (eg, tolerance <0.1). However, the AF% burden was excluded from the final model due to high multicollinearity with age and time from AF diagnosis to hospitalization, as evidenced by preliminary VIF values exceeding 10.0.

The results of multivariable logistic regression analysis in Table 7 identified several predictors significantly associated with IS occurrence in patients with AF. A higher AF burden in the group with IS than in the group without IS (OR: 2.66, 95% CI: 1.54–5.20, p=0.001) indicates increased IS risk, demonstrating that prolonged exposure to AF can amplify thromboembolism risk. Hypertension (OR: 5.30, 95% CI: 1.27–24.34, p=0.025) also markedly elevated IS odds, reinforcing its role as a critical modifiable risk factor. Additionally, higher P wave duration in the lead V1 (PWDV1) values (OR: 1.05 per ms, 95% CI: 1.02–1.08, p<0.001) and lower Pmax length (OR: 0.94 per ms, 95% CI: 0.90–0.97, p<0.001) were linked to increased IS risk. The area under the curve (AUC) was 0.908 (95% CI: 0.861–0.955), reflecting excellent model performance (Figure 4)

Conversely, higher LVEF (OR: 0.92 per%, 95% CI: 0.84–0.99, p=0.030) was associated with reduced IS odds, suggesting a protective effect of preserved cardiac function. Age was associated with IS risk (OR: 1.08 per year, 95% CI: 1.00–1.17, p=0.043), consistent with its established role as a risk factor, but the effect size was small. Sex (OR: 1.91, 95% CI: 0.58–7.08, p=0.303) and LAD size (OR: 0.88 per cm, 95% CI: 0.28–2.72, p=0.829) were not significantly associated with IS, but echocardiography showed the LAA was larger in the IS patients compared to the control group (Table 7).

Discussion

LIMITATIONS:

A limitation of this study is its retrospective nature. The study included 731 patients, and SR was documented in 314 cases, but because of the exclusion criteria, only 256 patients were analyzed (Figure 1). Large randomized controlled trials are needed to assess the possible negative predictive value of all indicators of atrial conduction disturbances simultaneously, as suggested in the present study. Women constituted the majority in both the study and control groups, and patients in the study group were also older. In the control group, patients with AF and without IS were significantly younger than the IS patients, which may have caused this cohort to be slightly under-represented in the study. This may have affected the results of the statistical analysis in relation to the examined parameters to sex and age. The echocardiographic parameter of the size of the LA was the LAD and the surface areas of both atria were calculated in the apical four-chamber view. Volumetric parameters (eg, LAVI) that are currently recommended were not measured, because this retrospective study covered a period of up to 20 years, and at that time these indicators were not widely used. Echocardiographic parameters were not adjusted to BSA, which was impossible due to the retrospective nature of the study. This may result in a slightly distorted estimation of the relationship between atrial parameters and predictors significantly associated with IS occurrence in patients with AF, indicating no association between LAD and the occurrence of IS. However, this does not apply to LVEF, which does not require adjustment to BSA. Electrocardiographic indices were also not adjusted, but this was not performed in most previous studies. Multivariate analysis considers variables selected by the authors, which may be subjective. The selection of variables was based on available data in a retrospective analysis of medical documentation, and this is also likely a limitation of our study.

Conclusions

The main conclusion of the study is that LA enlargement can be found in surface ECG by P wave indices. ECG markers can predict stroke risk. However, the effect is smaller compared to hypertension and estimated AF burden, which appear to be stronger predictors. In patients with IS, higher indices of structural and electrophysiological LAR were observed. P wave indices and LA dimensions on echocardiography reached significantly higher values in the male group than in the female group, which is associated with an earlier occurrence of cardiovascular disease in males. However, sex was not an independent factor in the occurrence of IS. We found that IS was independently associated with the ECG markers PWDV1 and Pmax, and with the echocardiographic index LVEF.

Current research justifies the search for simple and easy-to-use screening tools (eg, useful in the GP’s office). Perhaps simple ECG tracing parameters could become a screening tool to reduce the risk of IS in the future. This study provides reasonable grounds to believe that, after validation, ECG indicators of LA electrophysiological remodeling, similarly to the LAVI as well as LVEF, should also be included in the new, extended CHA2DS2-VASc (currently: CHA2DS2-VA) risk scale.