Predictors of Left Atrial Thrombus andSpontaneous Echo Contrast in RheumaticValve Disease Before and After Mitral

Valve ReplacementMehmet Ozkan, MD, Cihangir Kaymaz, MD, Cevat Kirma, MD, Ali Civelek, MD,

Ali Riza Cenal, MD, Cevat Yakut, MD, and Ubeydullah Deligonul, MD

In this study we aimed to analyze, with reference tomitral regurgitation (MR), the incidence and predictors ofleft atrial (LA) thrombus and spontaneous echo contrastin patients with rheumatic valve disease before and aftermitral valve replacement. The incidence of LA thrombusis known to be less in patients with MR. The impact ofmitral valve replacement on this beneficial effect has notbeen studied in detail. The study included 169 consecu-tive patients (59 men and 110 women, average age40 6 13 years) with rheumatic mitral valve disease whounderwent transesophageal echocardiographic exami-nation 1 to 3 days before and within 7 days (mean4.0 6 1.3) after mitral valve replacement using mechan-ical prostheses in a single institution. The preoperativeincidence of echocardiographic LA spontaneous echocontrast (SEC) was 1.1%, 30%, and 54%, and the inci-

dence of thrombus was 1.1%, 13%, and 17% in thegroups with MR, combined mitral stenosis 1 MR, andisolated mitral stenosis, respectively. In the MR group,SEC and thrombus incidence increased significantly aftersurgery. The independent predictors for postoperativethrombus development were atrial fibrillation, postoper-ative SEC, and preoperative thrombus. Thrombus re-curred after surgery in 64% of 14 patients who hadsurgical thrombectomy. The presence of postoperativeMR was associated with decreased risk of postoperativeSEC and thrombus development. The interaction be-tween MR and SEC and thrombus both before and aftersurgery provides further support for the protective effectof MR against LA thrombus formation. Q1998 by Ex-cerpta Medica, Inc.

(Am J Cardiol 1998;82:1066–1070)

In clinicopathologic series, and recently in echocar-diographic studies, the incidence of left atrial (LA)

thrombus formation has been found to be lower inpatients with mitral regurgitation (MR) than in thosewith nonregurgitant valves despite comparable LAsize and atrial fibrillation.1–7 Several studies indicatedthe potential for early LA thrombus formation aftermitral valve replacement.8–20The presence of MR dueto paravalvular leaks after valve surgery also showeda favorable effect on LA thrombus development.16

However, the impact of mitral valve replacement onthe interaction between MR and LA thrombus forma-tion has not been studied prospectively. In this study,we aimed to analyze, with reference to MR, the inci-dence and predictors of LA thrombus and spontaneousecho contrast (SEC) formation in a large series ofconsecutive patients with rheumatic valve disease be-fore and after mitral valve replacement.

METHODSThe study population included 169 consecutive

patients (59 men and 110 women; average age 40613 years) with rheumatic mitral valve disease whounderwent transesophageal echocardiographic exami-nations 1 to 3 days before and within 7 days (mean

4.0 6 1.3) after mitral valve replacement with use ofmechanical prostheses in a single institution betweenNovember 1995 and June 1997. The diagnosis ofrheumatic valve disease was based on a history ofrheumatic fever or documentation by echocardiogra-phy or pathology of the characteristics of rheumaticvalve involvement including (1) leaflet thickening,deformation, and retraction, (2) fusion, shortening,and fibrosis of the subvalvular apparatus, and (3)accompanying aortic or tricuspid valve involvement.Written informed consent was obtained from eachpatient before entry into the study, and the protocolwas approved by the institutional review board.

Echocardiographic technique and descriptions:Transesophageal echocardiography (TEE) was per-formed after.4 hours of fasting, under topical anes-thesia with lidocaine and conscious sedation with in-travenous midazolam. For TEE, a 5-MHz multiplaneprobe, and for transthoracic echocardiography a 3.25-MHz transducer connected to a Vingmed CFM 800system (Horten, Norway) were used. The mitral valvearea was measured by pressure half time21 and directplanimetric methods. MR was diagnosed by colorDoppler documentation of mitral regurgitant jet bothon transthoracic echocardiography and TEE. The se-verity of MR was determined from the videotaperecordings of multiplane TEE according to previouslypublished guidelines.22 A mild, moderate, and severeregurgitation was diagnosed when the regurgitant jetarea was,4 cm2, 4 to 8 cm2, and.8 cm2, respec-tively. The presence of systolic pulmonary vein flowreversal on the TEE indicated severe MR.

From the Kosuyolu Heart and Research Hospital, Kosuyolu, Istanbul,Turkey; and the Cardiology Section, University of Nebraska MedicalCenter, Omaha, Nebraska. Manuscript received February 9, 1998;revised manuscript received and accepted June 15, 1998.

Address for reprints: Ubeydullah Deligonul, MD, The University ofTexas Health Center at Tyler, 11937 U.S. Highway 271, Tyler, Texas75708-3154.

1066 ©1998 by Excerpta Medica, Inc. 0002-9149/98/$19.00All rights reserved. PII S0002-9149(98)00556-6

Patients were divided into 3 groups: Isolated mitralstenosis (MS) was diagnosed as mitral valve area of,1.5 cm2 associated with no or mild regurgitation(MS group). The severe MR group included all pa-tients with echocardiographically severe MR. A diag-nosis of mixed MS and MR (MS1 MR) was madewhen moderate MR was associated with a valve area,1.5 cm2 (MS 1 MR group). Spontaneous echocontrast was diagnosed when swirling, smoke-likeecho densities in the left atrium, distinct from near-field and high gain artifacts, were documented on theTEE.23 Thrombus was diagnosed when fixed or mo-bile, soft, homogenous echo densities were detected inthe left atrium, in the LA appendage, or on the mitralvalve prosthesis. Mitral prosthetic valve thrombus wasclassified as obstructive when mitral valve area by thepressure half-time method was,1.5 cm2. String-like,mobile structures.1 mm in diameter and 5 to 15 mmin length were diagnosed as strands.

Surgical technique: Mitral valve replacement wasperformed using standard open-heart surgery with thepatient on cardiopulmonary bypass. Intraoperatively,anticoagulation with heparin was maintained to keepactivated clotting time.450 seconds. Anticoagula-tion was started using subcutaneous heparin (5,000 U,twice daily) and warfarin (5 mg/day) 12 to 24 hoursafter surgery when no further bleeding from the chestdrains was documented. The oral anticoagulation dosewas adjusted to reach an international normalized ratioof .2.5 by day 7.

Statistical methods: The continuous variables arepresented as mean6 SD. A chi-square test was usedto compare the incidence of SEC and thrombus in the3 groups before and after surgery. The rates of SECand thrombus in the preoperative versus postoperativeperiod were compared using the Wilcoxon rank test.Continuous variables were compared using analysis ofvariance in the MS, MS1 MR, and MR groups, andthe unpaired Student’st test in the groups with orwithout SEC or thrombus. A stepwise multiple regres-sion analysis (SPSS for Windows) was used for de-termination of the predictors of postoperative SEC andthrombus in the overall series and in each of the 3groups. The following variables were entered into theinitial model: age, gender, atrial fibrillation, LA diam-eter, preoperative MR (severe), postoperative MR (se-

vere), preoperative SEC, preoperative thrombus, post-operative SEC, postoperative thrombus, and prostheticmitral valve thrombus. A p value,0.05 was selectedfor entering or removing the variables into the model.A p value ,0.05 was accepted as significant. All pvalues,0.001 were recorded as p,0.001.

RESULTSBefore surgery, 87 patients had MR, 30 patients

had MR1 MS, and 52 had MS. All patients receivedmechanical valve prostheses (bileaflet in 88 patients,and single disc in 81). Atrial fibrillation was present in101 patients before surgery and persisted in 99 aftersurgery. The remaining 68 patients were in sinusrhythm before surgery, with postoperative atrial fibril-lation development in only 1. The LA size was 5.661.2 and 5.26 1.1 mm before and after surgery,respectively (p5 NS). Postoperative MR was trans-valvular in 5 and paravalvular in 16 patients.

Incidence of spontaneous echo contrast and throm-bus (Table I): Before surgery, SEC was documented in38 patients (22%) and thrombus in 14 (8%). Throm-bus was located in the left atrium only in 6 patients, inthe LA appendage in 2 patients, and in both locationsin 6 patients. The preoperative incidence of SEC was1.1%, 30%, and 54% in the groups with MR, MS1MR, and MS, respectively. Similarly, preoperativeintraatrial thrombus incidence increased from 1.1% inthe MR group to 13% in the MS1 MR group, and upto 17% in the MS group (see Table I for p values).

Surgical thrombectomy was performed in all 8patients with LA thrombus, and appendage ligationwas performed in the 8 patients with appendagethrombus. After surgery, LA SEC was documented in63 patients (37%, p,0.025 compared with the pre-operative rate of 22%), and thrombus was documentedin 28 (17%, p,0.017 compared with the preoperativerate of 8%). Most of this increase in thrombus inci-dence was due to new thrombus development in theMR group. In the MR group, in contrast to the other 2groups, the SEC and thrombus incidence increasedsharply after surgery (from 1.1% to 38%, p,0.001,and from 1.1% to 18%, p,0.001, respectively).

Location and type of intraatrial thrombus after sur-gery: Postoperative thrombus was documented in 28patients and was located at the atrial septum in 7, at

TABLE I Incidence of Left Atrial SEC and Thrombus Before and After Valve Replacement

Preoperative Postoperative

SEC (%) Thrombus (%) SEC (%) p Value* Thrombus (%) p Value*

MR group (n 5 87) 1 (1.1) 1 (1.1) 33 (37.9) ,0.001 16 (18.4) ,0.001MS 1 MR group (n 5 30) 9 (30) 4 (13) 11 (36.6) NS 5 (16.6) NSMS group (n 5 52) 28 (53.8) 9 (17.3) 19 (36.5) NS 7 (13.4) NS

p† for MR vs MS 1 MR ,0.001 ,0.015 NS NSp† for MR vs MS ,0.001 ,0.001 NS NSp† for MS vs MS 1 MR p ,0.06 NS NS NS

Total (n 5 169) 38 (22.4) 14 (8.3) 63 (37.2) ,0.025 28 (16.5) ,0.017

*p value for preoperative versus postoperative incidence.†p value for between-group comparisons. All p values ,0.001 are shown as ,0.001.

VALVULAR HEART DISEASE/PREDICTORS OF THROMBUS AFTER MITRAL REPLACEMENT 1067

the atrial septum and posterior LA wall in 9, and at theposterior wall only in 3. The thrombus showed diffusemural attachment in 2. The thrombus was found lim-ited to the LA appendage in 7 patients. The thicknessof mural thrombus averaged 6.56 1.3 mm. Isolatedatrial strands were noted in 2 patients. Thrombusinvolving the mitral valve prosthesis was seen in atotal of 38 patients (9 with concomitant LA throm-bus). This was associated with hemodynamic obstruc-tion in 14 of 38 patients (6 with concomitant LAthrombus). In the MR, MS1 MR, and MS groups,postoperative isolated LA thrombus developed in 10,1, and 4 patients, isolated LA appendage thrombus in2, 3, and 2 patients, and combined LA1 appendagethrombus in 4, 1, and 1 patient, respectively.

Table II demonstrates the univariate pre- and post-operative predictors of postoperative spontaneous echocontrast and thrombus in all 169 patients: The presenceof postoperative SEC increased the likelihood ofthrombus and vice versa. Preoperative increased LAdiameter and atrial fibrillation was associated withhigher risk of postoperative SEC, but only atrial fi-brillation was related to the presence of postoperativethrombus. Although the incidence of SEC and throm-bus increased significantly after surgery in the MRgroup, preoperative severe MR was not a predictor ofpostoperative SEC or thrombus. Among the postoper-ative variables analyzed, LA thrombus as well asmitral prosthetic thrombus, especially the obstructivetype, were more frequent in patients with postopera-tive SEC. The presence of postoperative MR wasassociated with decreased risk of postoperative SECand thrombus development.

Table III demonstrates that the postoperativethrombus was associated with larger LA diameter onlyin the MR group (p,0.001).

Recurrent thrombus development after thrombec-tomy or left atrial appendage ligation: In 9 of 14 pa-tients who had thrombectomy during surgery, throm-bus recurred after surgery (64%). None of these pa-tients were in the MR group. These 9 patientsconstituted 32% of the total postoperative thrombus

cases. Among the 8 patients who had LA appendageligation, 3 with external ligation had recurrent throm-bus development on the LA side of the suture line. Inthe postoperative period, patients with and withoutthrombus had similar international normalized ratiovalues (1.76 0.4 vs 1.86 0.5, respectively). Nosystemic embolization was noted within the first 30days after surgery. During 6 months of follow-up,total or near total resolution of thrombus was docu-mented in 12 patients (43%).

Independent predictors of postoperative left atrialthrombus and spontaneous echo contrast development(Table IV): When all patients were combined, the inde-pendent predictors for postoperative thrombus devel-opment by stepwise multiple regression analysis werethe presence of preoperative thrombus, atrial fibrilla-tion, and postoperative SEC. Atrial fibrillation andpostoperative SEC were the independent predictorsfor postoperative thrombus in the MR group. In boththe MS1 MR and MS groups, however, the presenceof thrombus before surgery was the only predictor forpostoperative thrombus formation. Postoperative SECwas predicted by atrial fibrillation, postoperative LAthrombus, and postoperative no or mild MR. Thepredictive power of atrial fibrillation for postoperativeSEC was also evident in each of the 3 patient groups.

DISCUSSIONThere are 2 important results of this study: First, a

significant proportion of patients developed significant

TABLE II Parameters Measured Before and After Valve Replacement and Postoperative SEC and Thrombus

ParametersSEC (1)(n 5 63)

SEC (2)(n 5 106)

Thrombus (1)(n 5 28)

Thrombus (2)(n 5 141)

PreoperativeAge (yr) 41.7 6 12.7 39.4 6 13.2 41.0 6 14.5 40.1 6 12.7Women (%) 45 (71.4) 65 (61.3) 9 (32.1) 91 (64.5)LA diameter (cm) 5.7 6 1.1 4.9 6 1.0* 5.8 6 1.3 5.1 6 1.06Atrial fibrillation (%) 55 (87.3) 61 (57.5)† 27 (96.4) 73 (51.8)†Preoperative mitral regurgitation, severe (%) 33 (52.3) 54 (50.9) 16 (57.1) 71 (50.3)

PostoperativeSEC (%) — — 22 (78.5) 41 (29)†LA or appendage thrombus (%) 22 (34.9) 6 (5.6)* — —Mitral prosthetic thrombus (%) 19 (30) 19 (17.9)‡ 9 (32.1) 29 (27.1)Obstructive mitral prosthetic thrombus (%) 10 (15.8) 4 (3.8)‡ 6 (21.4) 8 (5.6)‡Postoperative mitral regurgitation, mild (%) 8 (12.7) 6 (5.6)‡ 1 (3.6) 13 (9.2)‡Postoperative mitral regurgitation, moderate (%) 1 (1.5) 4 (3.8) 0 5 (3.5)‡Postoperative mitral regurgitation, severe (%) 1 (1.5) 15 (14.1)‡ 0 16 (11.3)‡

* p ,0.001; †p ,0.005; ‡p ,0.05.

TABLE III Left Atrial Diameters in Groups With or WithoutPostoperative Left Atrial Thrombus

Thrombus (1)(cm)

Thrombus (2)(cm)

MR group 6.3 6 1.2* 4.7 6 1.8†

MS 6 MR group 5.7 6 1.1 4.8 6 1.5MS group 4.6 6 1.7 4.7 6 1.2

*p ,0.05; †p 5 0.001 (MR group compared to other 2 groups).

1068 THE AMERICAN JOURNAL OF CARDIOLOGYT VOL. 82 NOVEMBER 1, 1998

valvular and/or LA thrombus, and LA SEC (a likelyprecursor for thrombus) in the very early postopera-tive period after mitral valve surgery. Second, therewas a sharp postoperative increase in SEC and throm-bus incidence in patients who underwent valve re-placement for severe MR. An additional finding wasthe high recurrent thrombus rate after surgical throm-bectomy.

Postoperative LA thrombus may develop as earlyas during operation.15 In an autopsy series, LA throm-bus involving the interatrial septum incision site, pos-terior wall, or the sewing ring was found in 61% ofpatients who died within 60 days of mitral valvereplacement.10 Thrombus localization by TEE in ourstudy was found to be similar. Early echocardio-graphic detection of thrombus development after mi-tral valve replacement was reported initially by Mikellet al.9 Scott et al,13, using TEE, reported the incidenceof postoperative thrombus development as 16%, in-creasing to 55% in patients who had had systemicembolism. Our results are also in concordance withthe study of Bonnefoy et al16 who reported a thrombusincidence of 12% and an SEC incidence of 40%postoperatively. They also found a significantly de-creased incidence of SEC in patients with paravalvularMR.

In our series, obstructive mitral prosthetic throm-bus was a univariate predictor of SEC and intraatrialthrombus. However, this variable was not significantin multivariate analysis, suggesting the more powerfuleffects of other factors such as atrial fibrillation. Pre-operative thrombus appeared to be a significant riskfactor in the MS 1 MR and MS groups, despitesurgical thrombectomy. Endocardial damage duringthrombus removal or chronic changes in LA endocar-dium may be responsible for recurrent thrombo-sis.10,13,17,24

LA SEC and thrombus are caused by stasis of theblood in the left atrium precipitated by atrial fibrilla-tion, larger LA size, and low cardiac output.19,20,25–27

Our study clearly demonstrated a marked increase inthrombus development risk after mitral valve replace-ment in patients with MR. This was more prominent

in patients who also had atrial fibril-lation and a large LA diameter inaddition to MR. More than half ofthe patients with postoperativethrombus had severe MR at baseline.Interestingly, patients who devel-oped moderate to severe MR aftersurgery (paravalvular leak) had a sig-nificantly lower incidence of postop-erative SEC and thrombus. These re-sults bring direct support to the pre-viously suggested protective effectof MR against LA thrombus devel-opment.1–7 According to the resultsof our study, correction of MR byvalve replacement counteracts thisbeneficial effect and unmasks the

thrombus-facilitating effects of atrial fibrillation andof the large LA diameter.

1. Coulshed N, Epstein EJ, McKendrick CS, Galloway RW, Walker E. Systemicembolism in mitral valve disease.Br Heart J 1970; 32:26–34.2. Wood P. Diseases of the Heart and Circulation. 2nd ed. London: Eyre andSpottiswoode, 1956:502–604.3. Wanishsawad C, Weather DL, Buell JC. Mitral regurgitation and left atrialthrombus in rheumatic mitral valve disease: a clinicopathological study.Chest1995;108:677–681.4. Mosvowitz C, Mosvowitz HD, Jacobs LE, Meyerowitz CB, Podolsky LA,Kotler MN. Significant mitral regurgitation is protective against left atrial spon-taneous echo contrast and thrombus as assessed by transesophageal echocardi-ography.J Am Soc Echocardiogr1993;6:107–114.5. Hwang JJ, Shyu KG, Hsu KL, Chen JJ, Kuan P, Lien WP. Significant mitralregurgitation is protective against left atrial spontaneous echo contrast formation,but not systemic embolism.Chest1994;106:8–12.6. Karatasakis GT, Gotsis AC, Cokkinos DV. Influence of mitral regurgitationon left atrial thrombus and spontaneous echocardiographic contrast in pa-tients with rheumatic mitral valve disease.Am J Cardiol 1995;76:279 –281.7. Blackshear JL, Pearce LA, Asinger RW, Dittrich HC, Goldman ME, Zabalo-goita M, Rothbard RM, Halperin JL. Mitral regurgitation associated with reducedthromboembolic events in high risk patients with nonrheumatic atrial fibrillation.Am J Cardiol1993;72:840–843.8. Roberts WC, Morrow AG. Mechanisms of acute left atrial thrombosis aftermitral valve replacement. Pathologic findings indicating obstruction to the leftatrial emptying.Am J Cardiol1966;18:497–503.9. Mikell FL, Asinger RW, Rorke T, Hodges M, Sharma B, Francis GS. Twodimensional echocardiographic demonstration of left atrial thrombi in patientswith prosthetic mitral valves.Circulation 1979;60:1183–1190.10. Ben-Shachar G, Vlodaver Z, Joyce LL, Edwards JE. Mural thrombosis of theleft atrium following replacement of the mitral valve.J Thorac Cardiovasc Surg1981;82:595–600.11. Akalin H, Ozyurda U, Corapcioglu T, Uysalel A, Sonel A. Successfulnonsurgical therapy of mural thrombosis of the left atrium after mitral valvereplacement.J Thorac Cardiovasc Surg1988; 95:733–734.12. Chow WH, Lee WT, Tai YT, Cheung KL. Free-floating ball thrombus in leftatrium after mitral valve replacement: surgical removal following embolization tothe aorta.Am Heart J1990;120:1463–1465.13. Scott PJ, Essop R, Wharton GA, Williams GJ. Left atrial clot in patients withmitral prostheses: increased rate of detection after recent systemic embolism.IntJ Cardiol 1991;33:141–148.14. Gonzales-Alujas T, Evangelista-Masip A, Garcia Del Castillo H, Tornos-MasP, Soler-Soler J. Recurring free floating thrombus in the left atrium in a patientwith mitral prosthesis.Chest1994;106:303–304.15. Cheung AT, Levin SK, Weiss SJ, Acker MA, Stenach N. Intracardiacthrombus: a risk of incomplete anticoagulation for cardiac operations.AnnThorac Surg1994;58:541–542.16. Bonnefoy E, Perinetti M, Girard C, Robin J, Ninet J, Barthelet N, Lehot JJ,Touboul P. Systemic transesophageal echocardiography during the postoperativefirst 24 hours after mitral valve replecement.Arch Mal Coeur Vaiss1995;88:315–319.17. Peterson LM, Fisher RD, Reis RL, Morrow AG. Cardiac operations inpatients with left atrial thrombi: incidence and prevention of postoperativeemboli.Ann Thorac Surg1969;8:402–406.18. Tunick PA, Schulman IC, Kronzon I. Fistulous tract within a left atrial

TABLE IV Independent Variables of Postoperative SEC and Thrombus Formationby Stepwise Logistic Regression Analysis

Spontaneous Echo Contrast Thrombus

MR group Atrial fibrillation* Atrial fibrillation*Postoperative thrombus† Postoperative spontaneous

echo contrast†MS 1 MR group Atrial fibrillation‡ Preoperative thrombus†

MS group Atrial fibrillation† Preoperative thrombus†

Absence of postoperative mitralregurgitation*

Overall group Atrial fibrillation§ Atrial fibrillation*Absence of postoperative mitral

regurgitation*Preoperative thrombus§

Postoperative thrombus‡ Postoperative spontaneousecho contrast§

*p ,0.05; †p ,0.001; ‡p ,0.01; §p ,0.0001.

VALVULAR HEART DISEASE/PREDICTORS OF THROMBUS AFTER MITRAL REPLACEMENT 1069

thrombus: an unusual route for prosthetic mitral regurgitation demonstrated bytransesophageal echocardiography.Am Heart J1994;127:716–719.19. Michalis LK, Thomas MR, Smyth DW, Why HJ, Monaghan MJ, Jewitt DE.Left atrial spontaneous echo contrast assessed by TEE in patients with eithernative mitral valve disease or mitral valve replacement.J Am Soc Echocardiogr1993;6:299–307.20. Skudicky D, Skoularigis J, Essop M, Rothlisberger C, Sareli P. Prevalenceand clinical significance of mild paraprosthetic ring leaks and left atrial sponta-neous echo contrast detected on transesophageal echocardiography three monthsafter isolated mitral valve replacement with a mechanical prosthesis.Am JCardiol 1993;72:848–850.21. Hatle L, Angelson B, Tromsdal A. Noninvasive assessment of atrioven-tricular pressure half-time by Doppler ultrasound.Circulation 1979;60:1096 –1104.22. Helmcke F, Nanda N, Hsiung M, Soto B, Adey CK, Goyal RG, GatewoodRP. Color Doppler assessment of mitral regurgitation with ortogonal planes.Circulation 1987;75:175–183

23. Beppu S, Nimura Y, Sakakibara H, Nagata S, Park YD, Izumi S, Ueoka M,Masuda Y, Nakasone I. Smoke-like echo in the left atrial cavity in mitral valvedisease: its features and significance.J Am Coll Cardiol1985;6:744 –749.24. Ullal SR, Kluge TH, Hill JD, Keith WJ, Gerbode F. Left atrial thrombi inmitral valve disease.J Thorac Cardiovasc Surg1971;6:932–937.25. Burchfiel CM, Hammermeister KE, Krause-Steinrauf H, Sethi GK, Hender-son WG, Crawford MH, Wong M. Left atrial dimension and risk of systemicembolism in patients with a prosthetic heart valve. Department of VeteransAffairs Cooperative Study on Valvular Heart Disease.J Am Coll Cardiol1990;15:32–41.26. Hwang JJ, Kuan P, Lin SC, Chen WJ, Lei MH, Ko YL, Cheng JJ, Lin JL,Lien WP. Reappraisal by transesophageal echocardiography of the significance ofleft atrial thrombi in the prediction of systemic arterial embolization in rheumaticmitral valve disease.Am J Cardiol1992;70:769–773.27. Stroke Prevention in Atrial Fibrillation Investigators. Predictors of thrombo-embolism in atrial fibrillation: II. Echocardiographic features of patients at risk.Ann Intern Med1992;116:6–12.

1070 THE AMERICAN JOURNAL OF CARDIOLOGYT VOL. 82 NOVEMBER 1, 1998