Article Text

Importance of the underlying substrate in determining thrombus location in atrial fibrillation: implications for left atrial appendage closure
  1. Rajiv Mahajan1,
  2. Anthony G Brooks1,
  3. Thomas Sullivan1,
  4. Han S Lim1,
  5. Muayad Alasady1,
  6. Hany S Abed1,
  7. Anand N Ganesan1,
  8. Sachin Nayyar1,
  9. Dennis H Lau1,
  10. Kurt C Roberts-Thomson1,
  11. Jonathan M Kalman2,
  12. Prashanthan Sanders1
  1. 1Centre for Heart Rhythm Disorders, University of Adelaide and Royal Adelaide Hospital, Adelaide, Australia
  2. 2Department of Cardiology, Royal Melbourne Hospital and the Department of Medicine, University of Melbourne, Melbourne, Australia
  1. Correspondence to Professor Prashanthan Sanders, Department of Cardiology, Centre for Heart Rhythm Disorders, Royal Adelaide Hospital, Adelaide, SA 5000, Australia; prash.sanders{at}


Context The left atrial appendage (LAA) has been suggested to be the dominant location of thrombus in atrial fibrillation (AF) and has led to the development of LAA occlusion as a therapeutic modality to reduce stroke risk. However, the patient populations that would benefit most from this therapy are not well defined.

Objective A systematic review was performed to better define subgroups amenable to appendage closure.

Data sources The English scientific literature was searched using Pubmed through to March 1, 2011. Reference lists of relevant and review articles were screened to retrieve additional articles.

Study selection Studies were only included if they described the location of thrombus in left atrium. Case reports and case series describing less than 10 thrombi were excluded.

Data extraction Two reviewers independently extracted data and assessed quality of each study.

Results A total of 34 studies reporting on the location of atrial thrombus in patients with AF were included: 17 in valvular AF, 10 non-valvular AF and 8 in mixed valvular and non-valvular AF. Atrial thrombi were located outside the LAA in 56% (95% CI 53, 60) of valvular AF, 22% (95% CI 19, 25) in mixed cohorts and 11% (95% CI 6, 15) non-valvular AF. In non valvular AF, the studies with higher proportion of thrombi in the left atrial cavity had non-anticoagulated patients and a greater proportion of ventricular dysfunction and history of stroke.

Conclusion The location of atrial thrombus in patients with AF is dependent on the underlying substrate. In valvular AF, more than half the thrombi are located in the left atrial cavity. In the non-valvular AF group, a smaller proportion of thrombi were located outside the appendage. However, in certain subgroups (ie. non anti-coagulated, left ventricular dysfunction or prior stroke) the chances of left atrial cavity thrombus are higher.

  • Left atrium
  • thrombus
  • systematic review
  • atrial fibrillation
  • defibrillation
  • implantable cardioverter defibrillator
  • sudden cardiac death
  • holter ECG
  • radiofrequency ablation
  • arrhythmias
  • cardiac remodelling
  • hypertension
  • electrophysiology
  • cardiac arrest
  • remodelling
  • atrial arrhythmias
  • tachycardias
  • atrial flutter

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Cardioembolic stroke remains the most devastating complication of atrial fibrillation (AF). The risk of AF-related stroke increases from 1.5% per year between the ages of 50 and 59 years to 24% per year in patients >age of 80 years.1 In patients with AF, in the absence of rheumatic heart disease, the risk of stroke is fivefold greater after adjusting for other stroke risk factors, while in patients with rheumatic heart disease there is a 17.5-fold increased risk.1 ,2 There is also an increased incidence of silent cerebral infarction.3 Furthermore, evidence suggests that strokes in the setting of AF results in increased severity of the neurological deficit and are twice as likely to be fatal.4

Reports have suggested that the left atrial appendage (LAA) is the dominant location of thrombus in non-valvular AF and has led to the development of LAA occlusion as a therapeutic modality to reduce stroke risk.5 However, most of these reports have included only small and selected patient cohorts, such that the patient subsets that would benefit the most from occlusion of the LAA remain to be defined.6–8 In order to better understand the location of atrial thrombus and the determinants of thrombus location in AF, we undertook a systematic literature review.


Search strategy and selection criteria

The English scientific literature was searched using PubMed with keywords ‘atrial’ AND ‘fibrillation’ AND ‘left’ AND ‘thrombus’ contained in titles and abstracts of any record. The search was performed on 1 March 2011. The resultant citations were reviewed by two independent investigators (RM, AB). The articles of interest were those describing the location of thrombus in the left atrium (LA). In order to avoid publication bias, case reports and articles describing less than 10 thrombi, or those not clearly defining location of thrombus in the LA were excluded. All review articles were excluded but their reference lists were hand searched for all potential studies. Two additional references were included from reference list screening. Both these references had the selected keywords within them.9 ,10 Studies that were follow-ups of other included studies or a sub-study of the same cohort were also excluded. The studies were assessed against eight ‘quality’ variables to provide a guide to the extent to which the findings of each study could be relied upon, and to highlight any methodological flaws (see online supplementary appendix 1). None of the studies were excluded on the basis of quality.

Statistical analysis

Meta-analysis on the studies was performed using Stata V.11. Proportions were pooled across studies with sample size used to weight studies.11 The degree of heterogeneity in the proportion estimates across studies was assessed using the I-squared statistic,12 and a χ2 test of goodness-of-fit.


Search and synthesis of the literature

One thousand two hundred and four citations were identified after the initial literature search. They were combined with the supplementary hand searches to make a total of 1206. Of these, 1070 citations were excluded as they did not describe thrombi in LA in patients with AF, and 136 were selected for secondary review (figure 1). From this group, a total of 34 publications were identified after excluding small case series describing <10 thrombi and review articles.

Figure 1

Search criteria and flow diagram of the literature selection process.

Definition of study groups

Included studies were divided into three primary categories according to the substrate for AF: valvular AF (n=17), non-valvular AF (n=10) and the mixed non-valvular and valvular (n=8). Aberg et al reported statistics for both valvular and non-valvular cohorts and, as such, this study was included in both categories.3 We classified thrombi as LAA, if they were confined exclusively to the LAA. All other atrial thrombi were classified as left atrial cavity (LAC) category.

Identification of thrombus location

Studies reporting thrombus location in valvular AF mostly predated the non-valvular AF studies and used a variety of methodologies for thrombus location assessment. Thrombus location was examined via autopsy (3/17), surgical (9/17) or trans-oesophageal echocardiography (TOE) (5/17) in valvular AF. By contrast, the mixed and non-valvular AF groups used TOE with one exception (a non-valvular study which was autopsy based). When compared with direct examination of the LA during surgery, TOE has a sensitivity and specificity of 100% and 99%, respectively.14

Frequency of thrombi in the LAC

The proportion of thrombi located in the LAC was 56% (95% CI 53% to 60%; table 1) , 22% (95% CI 19% to 25%; table 2) and 11% (95% CI 6% to 15%; table 3) in the valvular, mixed and non-valvular AF cohorts, respectively.9 ,10 ,13–44 These data are further evaluated as shown in the Forest plot (figure 2), which presents the point estimate of thrombus location outside the LAA in all the studies. For thrombi located in the LA cavity, a majority of the studies did not further specify the location to different parts of the cavity.

Table 1

Location of thrombus in the left atrium in valvular atrial fibrillation

Table 2

Location of thrombus in the left atrium in mixed group

Table 3

Location of thrombus in the left atrium in non-valvular atrial fibrillation

Figure 2

Left atrial thrombus in the valvular, mixed and non-valvular atrial fibrillation cohorts.

Although the point estimate of the proportion of thrombi located in the LAC was only 11% in the non-valvular AF group (95% CI 6% to 15%), the heterogeneity statistic (I2>50, p=0.01) and forest plots demonstrated wide variation in this statistic (figure 2). Indeed, 22% (95% CI 13% to 31%) of thrombi for non-anticoagulated, non-valvular AF cohorts,13 ,35–37 were located in the LAC, whereas no thrombi (0%) were located in the LA cavity in appropriately anticoagulated patients prior to AF ablation (table 3).41–43 In the three non-valvular studies,38–40 where imaging was performed at a time when only part of the study patients were receiving anticoagulants, the proportion of thrombi located outside the appendage was intermediate (5.8%). In addition, the non-valvular AF cohorts with a greater proportion of LAC thrombi also had more patients with left ventricular dysfunction and history of stroke (table 4).

Table 4

Risk factors for stroke and anticoagulation status in non-valvular atrial fibrillation (AF) studies


This study analyses the location of atrial thrombus in patients with AF on the basis of the available literature, and identifies that 56% and 11% of atrial thrombi are located outside the LAA in valvular and non-valvular AF, respectively. However, 22% of atrial thrombi were located outside the LAA in non-anticoagulated non-valvular AF patients. In the non-valvular AF group, factors associated with thrombus outside the LAA included left ventricular dysfunction, history of stroke and anticoagulation status.

Valvular heart disease and AF

Data from surgical exclusion of LAA during valve replacement for valvular heart disease have demonstrated conflicting results regarding prevention of stroke in mitral valve disease with several cohort studies showing no benefit or even harm.45–53 While incomplete excision, which may be seen in up to one-third of patients, may also contribute to it,46–52 the current study provides further information on reasons for a lack of efficacy of this strategy. Indeed, in the current study, more than half the thrombi in the valvular AF are located in the LAC. These observations argue against a role of LAA occlusion in this cohort of patients.

Non-valvular AF

The PROTECT AF trial demonstrated the non-inferiority of the WATCHMAN device for LAA occlusion to warfarin therapy in non-valvular AF.7 Eighty seven per cent stopped warfarin after 45 days and 10% restarted warfarin for clinical reasons. It is worthwhile noting that there was a residual ischaemic stroke risk (3% vs 2% in the oral anticoagulant arm) despite LAA closure. Imaging of the brain for silent emboli, although not performed in this trial, would have shed more light on this issue. Moreover, as suggested by Maisel, this trial was too small to conclusively claim non-inferiority to warfarin, as reflected in the wide 95% CI (0.36 to 1.76).54 Similarly, the non-randomised PLAATO multicentre feasibility trial and its 5-year results have demonstrated the safety of the device with aspirin in patients of non-valvular AF who are not candidates for oral anticoagulants.8 ,55 In this study, the annual stroke rate over a period of 5 years was 3.8%. Although, there was no control arm, the estimated risk of stroke in a comparable population with an equivalent CHADS2 score (mean score 2.6) was estimated to be 6.6%. In addition, there have been equivocal results of LAA exclusion surgically during concomitant bypass in patients with non-valvular AF and risk factors for stroke.6

In the current study, although the risk of thrombus formation in LA cavity in non-valvular, AF was overall 11%, there was significant heterogeneity. The percentage of thrombus in the LAC was doubled to 22% in studies enrolling high risk, older and non-anticoagulant patients.13 ,35–37 These studies included patients with recent stroke, ventricular dysfunction or other risk factors which conferred high risk for stroke. Tsai et al subjected 219 chronic non-valvular AF patients (mean age 65 years) to TOE. One-third had a history of recent thromboembolism. Fifteen patients had LA thrombus, of which three were located in the LA cavity. Similarly, Brown et al evaluated 48 patients with a high proportion of left ventricular dysfunction or history of previous stroke. Left atrial thrombus was observed in 13 of 48 patients, two of these were located outside the LAA. The non-valvular AF group in the autopsy study by Aberg et al had coronary artery disease and hypertensive heart disease in three-quarters of the 506 autopsies studied. LA thrombi were located outside the LAA in 12 of 47 patients.

The studies performing screening for TOE prior to AF ablation enrolled low-risk patients for stroke. Moreover, the patients in these studies were optimally anticoagulated.41–43 Less than 10% of patients had LV dysfunction and stroke in each of these three studies. All the left atrial thrombi were localised to the LAA in this subgroup. Hence, the non-valvular AF group consists of studies with different risk of stroke and different anticoagulation status.

The reason for higher proportion of cavity thrombus in patients with ventricular dysfunction could possibly be related to left atrial enlargement, reduced cardiac output resulting in stasis, and possibly a greater activation of thrombogenic factors.

Clinical implications

This study demonstrates that more than half the thrombi in valvular AF are located out of the LAA, suggesting that LAA closure alone will not eliminate chances of thromboembolism in this group. Similarly, certain subgroups of non-valvular AF have higher propensity of LA cavity thrombi. Paradoxically, patients who are at higher risk for thromboembolism (ventricular dysfunction and previous stroke) are also at a higher risk of cavity thrombus.


The results of the current study were compiled using meta-analysis of primarily non-randomised observational data, rather than randomised controlled trial data. However, the technique of meta-analysis has come to be accepted in the literature to aggregate results from observational data, to facilitate synthesis of available evidence and generation of new hypotheses. Additionally, the subgroup statistical analysis could not be performed in the non-valvular group due to the small number of studies and a substantial proportion of them had a point estimate of 0%, which is difficult to model in the context of developing stable exact CIs and point estimates. Some of the non-valvular studies did not provide complete data regarding the prevalence of diabetes, hypertension, left ventricular dysfunction and stroke. Finally, the methods for detection of thrombus in this analysis were heterogeneous. However, it represents the summation of available literature and the basis on which LAA closure has been developed. Eventually, data from appendage closure studies in the coming future will test the hypothesis relating CHADS2 score to site of thrombus formation in non-valvular AF.


The location of thrombus in patients with AF is dependent on the underlying substrate. More than half the thrombi in valvular AF are located in the LA cavity, reflecting the conflicting results of LAA closure in valvular AF. In non-valvular AF, although the overall risk of thrombus location in LAC is low, there is significant heterogeneity. Certain subgroups of non-valvular AF (i.e., patients with ventricular dysfunction or history of stroke) may be at higher risk for location of thrombus in the LA cavity, especially if not receiving anticoagulants. These subgroups need to be better defined in a prospective manner to further refine the selection of patients suitable for LAA closure as a therapeutic modality for stroke prevention.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • This study was presented at the Annual Scientific Sessions of the Heart Rhythm Society, May 2011, San Francisco, USA and published in abstract form (Heart Rhythm 2011;8:S29).

  • Funding RM and HSA are supported by the Australian Postgraduate Award from the University of Adelaide. RM is supported by the Leo J Mahar Electrophysiology Scholarships from the University of Adelaide. AGB, KCR-T and PS are funded by the National Heart Foundation of Australia. HSL and MA are supported by a Postgraduate Medical Scholarship from the National Health and Medical Research Council of Australia. ANG is supported by the Michel Mirowski Fellowship from the Heart Rhythm Society. SN is supported by the Robert J Craig Electrophysiology Scholarship from the University of Adelaide. DHL is supported by a Postdoctoral Fellowship from the National Health and Medical Research Council of Australia.

  • Competing interests KCR-T reports having served on the advisory board of St Jude Medical. PS reports having served on the advisory board of St Jude Medical, Bard Electrophysiology, Biosense-Webster, Medtronic, Sanofi-Aventis, and Merck, lecture fees from St Jude Medical, Bard Electrophysiology, Biosense-Webster, Medtronic and Merck and research funding from St Jude Medical, Bard Electrophysiology, Biosense-Webster and Medtronic.

  • Provenance and peer review Not commissioned; internally peer reviewed.