Atrial fibrillation (AF) is the most common cardiac arrhythmia and is associated with a 5-fold risk for stroke (1). Recent evidence stated left atrial thrombus (LAT) and left atrial spontaneous echo contrast (LASEC) as risk factors of cardiogenic embolism in atrial fibrillation patients (2). Although transesophageal echocardiography (TEE) is still considered the gold standard to exclude LA/LAA thrombus, TEE requires special skills for proper performance and interpretation. Additionally, it is a relatively invasive test, usually performed with the patient under conscious sedation. Therefore, a potentially non-invasive and efficacious method allowing identification of LAT/SEC with reliability and accuracy comparable to TEE would be of significant clinical value.

The current guidelines for anticoagulant therapy for stroke prevention in NVAF depend on CHADS₂ and CHA₂DS₂-VASc scores (3). However, the predictive power of the CHADS₂ and CHA₂DS₂-VASc scores for the presence of LAT in NVAF is not satisfactory (c-statistics 0.55~0.70) (4, 5). Thrombus can be found even in some patients with low CHA2DS2-VASc scores (6). As such, it is of scientific interest to establish a stronger predictive model that incorporates factors associated with LAT/SEC based on clinical and ultrasonic data.

A powerful model that estimates LAT/SEC presence can assist cardiologists to identify high-risk patients and lead to a rational therapeutic choice. As a result, many efforts on the peri-procedural estimation of LAT/SEC have been made previously (5, 7). However, there is still no accurate model to predict LAT/SEC. Owing to this lack of a specific and practical predictive method, the development of a predictive model that incorporates factors associated with LAT/SEC based on peri-procedural clinicopathologic data becomes desirable.

In this study, we applied nomogram analysis, which can provide individualized, evidence-based, and highly accurate risk estimation. To our knowledge, we have established the first nomogram for peri-procedural LAT/SEC risk estimation in NVAF. The objectives of this study include to (1) investigate the predictive power of the CHADS₂ and CHA₂DS₂-VASc scores for the presence of LAT/SEC, (2) identify the clinical predictors of LAT/SEC, and (3) establish nomogram for LAT/SEC risk estimation in NVAF.

During the study period, 1,813 consecutive NVAF patients were collected. The patients were divided into the training (1,239, 68.3%) and validation cohorts (574, 31.7%). The baseline data were similar between the training and validation cohorts. The mean age of the participants was 62.6 ± 9.4 years. Of the total participants, 64.4% were males. LAT/SEC was found in 260 (21.0%) and 124 (21.6%) patients in the training and validation cohorts, respectively. The clinical and demographic characteristics of the patients are presented in Table 1.

The results of the univariate and multivariate logistic analysis are presented in Table 2. In the LAT/SEC group, patients had a higher prevalence of hypertension, prior stroke/TIA, non-paroxysmal AF, CAD, and larger left atrial diameter. Likewise, the LAT/SEC group had higher values of CHADS₂ and CHA₂DS₂-VASc scores, but lower left ventricular ejection fraction and estimated glomerular filtration rate. However, the prevalence of vascular disease and diabetes mellitus was similar in the two groups. The multivariate analysis showed that risk factors such as age (OR = 1.04, 95% CI: 1.02–1.06, and P = 0.001), LAD (OR = 1.21, 95% CI: 1.16–1.26, and P = 0.001), LVEF (OR = 0.92, 95% CI: 0.89–0.94, and P = 0.001), previous stroke/transient ischemic attack (OR = 1.79, 95% CI: 1.23–2.60, and P = 0.002), hypertension (OR = 1.44, 95% CI: 1.02–2.06, and P = 0.041) and non-paroxysmal AF (OR = 2.76, 95% CI: 1.99–3.85, and P = 0.001) remained independently associated with LAT/SEC.

ROC curve analysis was used to investigate the predictive power of the CHADS₂ and CHA₂DS₂-VASc scores concerning LAT/SEC. The results showed that the c-statistic of the CHADS₂ and CHA₂DS₂-VASc scores were 0.608 and 0.606, respectively. Furthermore, we developed a LAT/SEC risk estimation nomogram based on the results of the multivariate logistic analysis (Figure 1). The bootstrap validation method was used to internally validate the resulting model.

The nomogram demonstrated a very good predictive power in estimating the risk of LAT/SEC, with an unadjusted C index of 0.836. Besides, calibration plots graphically showed good agreement on the presence of LAT/SEC between the risk estimation by the nomogram and TEE confirmation. In the validation cohort, the nomogram displayed a C index of 0.794 for the estimation of LAT/SEC risk. Also, our result indicates that the observed frequencies and the estimated probability of LAT/SEC presence showed a good calibration curve for the risk estimation (Figures 2A–D). The decision curve shows the clinical usefulness of the nomogram (Figures 2E,F). In this analysis, the final decision curve showed that for a threshold probability between 10 and 80%, the model had positive net benefit.

The optimal cutoff value of the total nomogram scores was determined to be 97.0. The sensitivity, specificity, positive predictive value, and negative predictive value were 83.8, 71.9, 44.2, and 94.4%, respectively in the training cohort, and 70.2, 74.4,43.1, and 90.1%, respectively in the validation cohort (Table 3).

The present study that established the first nomogram for LAT/SEC risk estimation in patients with NVAF found that a new model composed of age, LAD, LVEF, previous stroke or transient ischemic attack, HTN and non-paroxysmal AF had a better performance for the prediction of LAT/SEC compared to CHADS₂ and CHA₂DS₂-VASc scores.

The prevalence of LAT/SEC for patients with AF varied in previous studies (2, 4, 13). In our study, the prevalence of LAT/SEC was 21.2% in NVAF population. In the present study, two variables that are not included in the CHA₂DS₂-VASc score were found to be the top predictors of LAT/SEC. Two predictors include LAD and non-paroxysmal AF. Therefore, it may be reasonable to consider NVAF patients with the enlarged left atrium and non-paroxysmal AF as candidates for more intensive medical follow-up. The association between these factors and LAT/SEC has also been reported in previous studies (7, 14–16). For instance, left atrial enlargement has been shown to associate with LAT/SEC, which is a surrogate marker of stroke risk (14, 17). Earlier evidence also reported that the possibility of thrombus formation increases with an enlarged LA cavity (14, 18). Although many mechanisms can explain the association between left atrial enlargement and thrombus formation, the mechanism that involves changes in left atrial hemodynamics, such as the existence of turbulences, reduced flow velocity, increased blood stasis and endothelial injury could be speculated as plausible mechanisms (19).

As earlier mentioned, non-paroxysmal AF remained a significant predictor for LAT/SEC in our study. Although the current ESC guidelines do not list AF type or AF burden among factors affecting the probability of LAA thrombus formation, few studies have shown that persistent or permanent AF carries a higher risk of stroke than paroxysmal AF (20). Relative to paroxysmal AF, non-paroxysmal AF shows greater structural remodeling and endocardial fibroelastosis of the atria and appendage, both of which are likely to contribute to thrombus formation (21, 22).

In our study, there were still 103 (14.9%) patients with LAT/SEC among the low-risk group (classified based on CHA₂DS₂-VASc score). In addition, the c-statistics of the CHADS₂ and CHA₂DS₂-VASc scores were 0.608 and 0.606, respectively, suggesting CHADS₂ and CHA₂DS₂-VASc scores had relatively weaker predictive performance in discriminating LAT/SEC compared to the new model. There could be two reasons that contribute to the observed phenomenon. Firstly, the models share the same risk factor with atherosclerosis. Consequently, CHADS₂ and CHA₂DS₂-VASc scores may predict stroke through the mechanism of atherosclerosis but not via mechanisms that involve cardiogenic embolism. Secondly, these risk scores do not incorporate other risk factors that are highly linked to thrombo-embolic risks, such as echocardiographic components, biochemical concentrations, and coagulation parameters that are known for predisposing stasis of blood within the left atrium and appendage.

In this study, our multivariate analysis revealed several predictors of LAT/SEC. By combining these predictors of LAT/SEC, we constructed a nomogram model. Interestingly, the newly constructed model demonstrated a strong discriminatory performance to identify patients with increased risk of LAT/SEC. The prognostic relevance of such a model of clinical risk factors has not been prospectively studied in the past. According to our results, the discriminatory performance of the new composition score was even stronger than CHADS₂ and CHA₂DS₂-VASc scores. For clinical use of the model, we recommend 97.0 as the cutoff value, and patients with a score of 97.0 or more should be considered as a high-risk group for LAT/SEC. Based on these predictions from the nomogram, the new model might serve as a substitute of TEE for NVAF patients who cannot tolerate TEE and provide references about whether to stop anticoagulants after procedural in the follow-up.

Our study has some limitations. First, this analysis was based on data from a single institution, thus it is necessary to validate the results from other centers. Second, this analysis is a retrospective study, some specific markers which might be associated with LAT/SEC such as left atrial appendage morphology, markers of endothelial dysfunction, and inflammation were not included in the nomogram. Moreover, the current study lacks data on left atrial volume, a more accurate marker to assess left atrial size. Third, the nomogram achieved a good predictive accuracy, with a cutoff point of 97.0, however, it demonstrated a significant proportion of false-positive and false-negative rates in the training (28.1 and 16.2%, respectively) and validation cohort (25.6 and 29.8%, respectively), which indicates replication of such study is of crucial importance in the future study to confirm the power of the utilized model in clinical decision making. Finally, the present study included only NVAF patients who underwent TEE before ablation or cardioversion intervention, therefore our results may be limited to NVAF patients who are candidates for ablation or cardioversion interventions.

Our study showed that Age, LAD, LVEF, HTN, previous stroke or transient ischemic attack, and Non-paroxysmal AF were the risk factors of LAT/SEC from AF patients. By combining these risk factors of LAT/SEC, a nomogram was constructed. The model provides an optimal peri-procedural estimation of LAT/SEC risk in patients with non-valvular atrial fibrillation.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by the First Affiliated Hospital of the Dalian Medical University Institutional Review Board. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.

XY and YX designed this study. ZL, QL, and YT were in charge of data analysis and data collection. ZL drafted the article. FL and TH did the critical revision of article. TC and LG conducted the data collection. All authors have read and approved the final manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.