Mitral Valve Strands and the Risk of Ischemic Stroke in Elderly Patients
Background and Purpose Strands are thin and filamentous attachments on the cardiac valves shown by transesophageal echocardiography. Their nature and their potential for embolization are largely unknown. The objective was to estimate the risk of brain infarction in patients with mitral valve strands.
Methods Using transesophageal echocardiography, we compared the frequency of strands on native mitral valves in 284 consecutive patients admitted with brain infarction and 276 control patients, all older than 60 years. In a second part, case subjects were followed up over a 2- to 4-year period, and the risk of recurrence of brain infarction was estimated in patients with and without strands.
Results In the case-control study, mitral valve strands were found in 22.5% of the case patients and in 12.1% of the control subjects. In case subjects, mitral valve strands were more frequent in those with mitral valve dystrophy (52.4% versus 37.4%; P=.03). Strands were not associated with mitral valve prolapse, annular calcifications, or left atrial spontaneous echocardiographic contrast. After adjustment for age, sex, and mitral valve dystrophy, the odds ratio for ischemic stroke among patients with mitral strands was 2.2 (95% confidence interval, 1.4 to 3.6; P=.005). The frequency of strands was not different in patients with a known cause of brain infarction (24.4%) from that in patients with no other apparent cause (20.9%). During 646 per 100 person-years of follow-up, the incidence of recurrent brain infarction was 6.0 person-years in patients with strands and 4.2 in those without. In the Cox analysis, including potential confounders and poststroke treatment, mitral valve strands did not appear as independent predictors of recurrent brain infarction (relative risk, 1.3; 95% confidence interval, 0.5 to 3.0; P=.54).
Conclusions The present study shows an independent association between mitral valve strands and the risk of brain infarction. However, the lack of an increased relative risk of recurrence raises doubts about the potential causal relation with brain infarction in patients aged 60 years or older.
Strands are highly mobile, threadlike filaments, seen on cardiac valves.1 2 They have recently been recognized on native3 4 5 and prosthetic6 7 8 valves with the use of transesophageal echocardiography, a new window to the optimal imaging of cardiac structures.9 The exact nature of strands and their potential for embolization are still debated. In his pathological report, Magarey2 described filiform excrescences (Lambl’s excrescences) on the auricular surface of mitral valves that were more common with increasing age and thickened valves. He hypothesized that Lambl’s excrescences might be the result of successive depositions of fibrin. Giant Lambl’s excrescences resulting in papilliform masses have been reported to be associated with emboli.10 11 12 It has been suggested that valve strands, as seen by transesophageal echocardiography, may represent Lambl’s excrescences.3
The aims of the present study were to determine the risk factors associated with mitral valve strands and the frequency of strands in consecutively admitted patients with brain infarction compared with control subjects. In addition, we sought to evaluate the relative risk of recurrent brain infarction in case subjects with and without strands.
Subjects and Methods
As part of a study evaluating the frequency and clinical significance of aortic arch plaques in patients older than 60 years,13 we undertook an independent case-control study to assess the association between brain infarction and strands on native mitral valves. We excluded patients and control subjects who had valvular prostheses. A follow-up study was also performed in case subjects to determine the risk of recurrent brain infarction in patients with and without strands.
Three hundred thirty-eight patients older than 60 years admitted to neurology departments for brain infarction were included between September 1991 and October 1993. The following risk factors were recorded: hypertension, hypercholesterolemia, cigarette smoking, diabetes mellitus, past myocardial infarction, peripheral artery disease, and atrial fibrillation. Patients underwent a diagnostic workup, including cranial CT imaging or MRI of the brain, ultrasound examination of cervical arteries, 12-lead electrocardiography, and transesophageal echocardiography. Patients were then divided according to the presumed cause of the stroke: likely cause, lacunar infarct, another possible cause, and no apparent cause.13 Three patients with valvular prostheses were excluded, leaving a total of 335 patients in the final sample.
The control group consisted of 276 patients with no history of brain embolization or peripheral embolization who underwent transesophageal echocardiography for cardiac conditions such as mitral or aortic valvulopathy, ischemic or idiopathic left ventricular systolic dysfunction, atrial fibrillation, and miscellaneous causes. The latter included endocarditis, intracavitary masses, suspicion of aortic dissection, pulmonary embolism, and inadequate transthoracic echocardiogram. Risk factors were recorded at the time of examination.
Transesophageal echocardiography was performed by trained cardiologists who had no information on the etiology of brain infarction. The examinations were recorded on videotapes for further off-line analysis. Transesophageal echocardiography was performed within 2 weeks after the onset of stroke, according to standard techniques.9 We used commercially available imaging systems (VingMed CFM 700, CFM 800, Hewlett-Packard Sonos 1000, and Acuson 128XP) with a 5-MHz single-plane probe in 428 patients, a biplane probe in 110 patients, or a multiplane probe in the last 73 patients.
A complete examination of cardiac structures for cardiac and aortic sources of embolism was performed. Left atrial and appendage thrombus, spontaneous echocardiographic contrast, atrial septal aneurysm, patent foramen ovale, mitral valve prolapse, and atherosclerotic plaques in the ascending aorta and arch were defined according to previously described criteria.13 14 Mitral valve strands were recognized as thin and filamentous attachments on the atrial surface of mitral leaflets, ≤1 mm in width, 1 to 20 mm in length, moving independently from the valves, in and out of the imaging plane (Fig 1⇓). They were differentiated from ruptured chordae tendineae, which appear as a fluctuating linear echo with high-frequency fluttering chordae in the left atrial cavity during systole. Sessile or pedunculated valvular lesions >1 mm in width were not described as strands. Mitral valve dystrophy was defined as a valve thickness ≥3 mm by transesophageal two-dimensional echocardiography.15 Mitral annular calcifications were defined as the presence of a cluster of dense high-intensity echoes between the posterior left ventricular wall and the posterior mitral leaflet.16
Case patients were followed up for 2 to 4 years to assess new vascular events, as described elsewhere.17 The incidence of recurrent brain infarction was compared in patients with and without strands.
In the case-control analysis, risk factors for strands were studied in case patients and control subjects, including demographic factors and cardiac morphological abnormalities. The frequencies of strands in both groups were then compared and expressed as a crude odds ratio. Adjusted odds ratios were estimated through unconditional logistic regression.
To estimate the impact of the referral diagnosis in control subjects on the risk estimation, we used the data of one center that included the largest number of case and control subjects (Saint-Antoine Hospital, 57% of the sample). Control subjects were classified in four broad categories of indication for transesophageal echocardiography. The frequency of strands and odds ratios for brain infarction in patients with strands were then estimated. Homogeneity of the odds ratios was tested by the Breslow and Day test.18
In the follow-up study, we computed rates of incidence by dividing the number of new cases of brain infarcts by the number of person-years in the groups of patients with and without strands. Cox models were used to estimate the relative risk of recurrent brain infarction in patients with mitral valve strands after controlling for potentially confounding factors. Two models were created: the first one included variables such as sex, age, and treatment; the second model included the same variables and potential causes of brain infarction: carotid stenosis >70%, plaques of the aortic arch, and atrial fibrillation. All reported probability values are two-tailed. Data were analyzed with the use of SAS19 and BMDP software.20
Videotapes of the transesophageal examinations were reviewed by one of us (A.C.), randomly and without knowledge of the status of the subject. Because we used strict criteria to recognize the presence of strands, we considered the information to be missing when the examination of the mitral valve was considered suboptimal for this specific question, mainly for technical reasons (≤10 cardiac cycles in the tape). Thus, information was lacking in 12% of the patients. In addition, the videotaped examinations of 53 randomly selected patients and control subjects were also reviewed by a senior echocardiographer (C.C.), according to the same protocol. The level of agreement between the two observers was 86.8%, and the Cohen’s κ index was 0.65.
The general characteristics and main risk factors for stroke of patients and control subjects are shown in Table 1⇓. Overall, case patients were older and more frequently women. With regard to risk factors for brain infarction, hypertension and cigarette smoking were more frequent in patients, as well as high-grade atherosclerotic plaques of the ascending aorta and proximal arch. Atrial fibrillation was equally frequent in both groups.
Information on strands was missing in 12% of the patients (51 case subjects and 20 control subjects). In the remaining patients, mitral valve strands were found in 22.5% of the patients (64/284) and 12.1% of the control subjects (31/256), yielding a crude odds ratio of 2.1 (95% confidence interval, 1.3 to 3.4; P<.005).
Risk factors for strands were studied separately in patients and control subjects (Table 2⇓). Patients with strands were more frequently men, although not significantly, and they more frequently had mitral valve dystrophy. The presence of strands was independent of age. Control subjects with strands were significantly older, but there was no significant difference in sex or frequency of mitral valve dystrophy. There was no difference in patients with or without strands in both groups with regard to other potential risk factors for brain infarct.
A multivariate unconditional logistic regression model was used to control for potentially confounding variables: age, sex, and mitral valve dystrophy. The adjusted odds ratio was 2.2 (95% confidence interval, 1.4 to 3.6; P=.001). A model that also included hypertension, cigarette smoking, atrial fibrillation, aortic arch plaques, left atrial spontaneous echocardiographic contrast, and atrial septal aneurysm gave a comparable estimate, with an adjusted odds ratio of 2.1 (95% confidence interval, 1.3 to 3.6; P=.005).
The frequency of strands in patients according to the cause of stroke was as follows: 24.4% (22/90) in patients with a likely cause, 21.6% (11/51) in patients with a lacunar infarct, 23.1% (12/52) in patients with a possible cause, and 20.9% (19/91) in patients with no other apparent cause.
Frequency of Strands in Control Subjects According to Referral Diagnosis
Among 149 control patients referred to the Saint-Antoine center for transesophageal echocardiography, the four following categories of cardiac conditions were considered: mitral valve disease (n=61), ischemic and idiopathic cardiomyopathy (n=39), aortic valve disease (n=11), and miscellaneous causes (n=38). Table 3⇓ shows the risk estimates when case patients of Saint-Antoine Hospital (n=175) were compared with each control category. The corresponding crude odds ratios varied from 1.4 in the miscellaneous group to 3.7 in the cardiomyopathy group. However, homogeneity between the odds ratios was not rejected by the Breslow and Day test (Breslow and Day statistic=2.45, P=.484). The global crude odds ratio for the Saint-Antoine center was 2.2, close to that found in the overall population.
Follow-up Study in Patients
Among our study population of 335 patients, 51 had missing information on strands. Thus, 284 patients were followed for a mean of 2.3 years. During the 646 person-years of follow-up, a recurrent brain infarction was observed in 29 cases. The incidence was 6.0 per 100 person-years in patients with strands (7 events) and 4.2 in patients without strands (22 events). Information on post–qualifying event treatment was available in 264 case patients: 66.3% (175/264) were treated with antiplatelet drugs, 20.8% (55/264) were treated with oral anticoagulants, and 12.9% (34/264) received no antithrombotic treatment. In patients with strands, the incidence rate of stroke was 7.1 per 100 person-years in patients receiving antiplatelet drugs, 6.9 in those receiving oral anticoagulants, and 0 in those with no antithrombotic treatment.
In the multivariate Cox model adjusted for age and sex, the relative risk of stroke in patients with strands was 1.3 (95% confidence interval, 0.6 to 3.1). In a model including age, sex, treatment, and likely sources for stroke (atrial fibrillation, carotid stenosis, aortic arch plaques), the relative risk was 1.3 (95% confidence interval, 0.5 to 3.0; P=.54). The two Kaplan-Meier curves in patients with and without mitral valve strands were not significantly different from one another (Fig 2⇓).
The results of the present case-control study indicate an independent association between mitral valve strands detected with the use of transesophageal echocardiography and the risk of brain infarction. This could occur if the frequency of mitral valve strands was overestimated in case subjects. We believe this hypothesis unlikely for several reasons.
First, the risk might be upwardly biased in our case-control study. Consecutiveness of patient recruitment, systematization of the transesophageal echocardiographic examination, and review of the videotapes by an echocardiographer unaware of patient status were used to avoid a systematic bias in the estimation of the frequency of strands. In fact, the knowledge of patient status may influence the rigor for the search of strands. In the only study with a design comparable to our study, Lee et al3 described mitral valve strands in 11 of 50 patients (22%) referred for ischemic stroke or transient ischemic attacks, an estimation close to that we found. In a recent report by Freedberg et al,4 the estimated risk was found to be much higher than that found in our study. However, that study included only patients referred for transesophageal echocardiography, and the referral diagnosis for control subjects was not indicated, leaving the possibility of an overestimation of the risk.
Second, overestimation of the risk might be related to an underestimation of the frequency of strands in control subjects.21 This point is critical in studies in which exposure is determined through a noninvasive procedure. Control subjects referred for transesophageal echocardiography represent a selected group of patients.22 To test this hypothesis, we studied the frequency of strands by referral diagnosis. We found that the frequency of strands varied in the four categories of indication for transesophageal echocardiography in control subjects from 7.7% of those referred for left ventricular dysfunction to 18% of those referred for aortic valve disease. Remarkably, the miscellaneous group, which is often less exposed to recruitment bias, had the highest frequency of patients with strands (18.4%). When case subjects were compared with this group, the estimated odds ratio was not significantly increased and was close to the relative risk obtained in the follow-up study. Strands were detected in 2.3% of control subjects in the study of Freedberg et al4 and in 0.3% of those evaluated in the study of Tice et al.5 Thus, in both studies the incidence of mitral valve strands in control subjects was under the lowest estimate found in our study. These results highlight the caution needed when estimating the risk associated with strands in case-control studies when control subjects are referred for transesophageal echocardiography.
In the follow-up study, the risk estimate for recurrent brain infarction was 1.3, which is not significantly different from 1.0. The risk estimate is not directly comparable to the one estimated in the case-control study because only case subjects were followed up and recurrent strokes were studied. In addition, the relatively small number of recurrent brain infarctions might restrict our ability to detect a significant increase of the risk in patients with strands. Another explanation for the lower than expected risk estimate in the follow-up study might be that since their qualifying vascular event, patients were treated with antithrombotic drugs, thus reducing the risk of recurrent brain infarction. However, our results did not confirm this hypothesis because we found that brain infarction recurrence in patients with strands was the lowest in those who received no antithrombotic therapy.
Our study contains other evidence against strands as a major source of embolism. If strands were a powerful source of brain embolism, one could expect a higher frequency of strands in patients with no other detectable cause. In fact, we found a similar prevalence of strands in patients with a likely source of brain infarction and in those without any detectable cause.
In our study as well as that of Lee et al,3 mitral valve thickening as shown by transesophageal echocardiography was associated with a higher incidence of mitral valve strands compared with valves without thickening. This fact supports the assumption that strands may represent a degenerative process of leaflet tissue. However, the exact nature of valvular strands remains undetermined. Lee et al3 suggested that mitral valve strands were consistent with Lambl’s excrescences. Whether strands on native mitral valves identified by transesophageal echocardiography and Lambl’s excrescences described in pathological reports10 12 23 represent the same morphological entity is currently unknown.
We found that strands were significantly associated with the risk of stroke. However, our data suggest that this association might be related in part to a referral bias in control subjects. Nevertheless, our follow-up study indicates that, in patients with strands, no specific treatment, medical or surgical, has to be considered.
Participating Institutions and Investigators
The following institutions and investigators participated in the French study of Aortic Plaques in Stroke (FAPS).
Paris, Saint-Antoine Hospital, Pierre and Marie Curie University: Department of Neurology—Pierre Amarenco, MD (principal investigator); Olivier Heinzlef, MD; Christian Lucas, MD; Pierre-Jean Touboul, MD (cranial ultrasound study design); Jean-Luc Gérard, MD; Valérie Adraï, MD; Didier Rougemont, MD; Marie-Germaine Bousser, MD; Department of Cardiology—Ariel Cohen, MD, PhD (coprincipal investigator); Christophe Chauvel, MD; Bouziane Benhalima, MD; Catherine Albo, MD; Éric Abergel, MD
Grenoble, Centre Hospitalier et Universitaire de Grenoble: Stroke Unit—Marc Hommel, MD (local principal investigator); Gérard Besson, MD; L. Vercueil; Department of Cardiology—Bernard Bertrand, MD (local coprincipal investigator)
Besançon, Centre Hospitalier et Universitaire de Besançon: Department of Neurology, Jean Minjoz Hospital—Thierry Moulin, MD (local principal investigator); Didier Chavot, MD; Laurent Tatu, MD; Department of Cardiology, Saint-Jacques Hospital: Yvette Bernard, MD (local coprincipal investigator)
Lille, Centre Hospitalier et Universitaire de Lille: Department of Neurology, Roger Salengro Hospital—Didier Leys, MD (local principal investigator); Philippe Rondepierre, MD; Christian Lucas, MD; Department of Cardiology, Cardiology Hospital—Luc Goulard, MD (local coprincipal investigator); Ghislaine Deklunder, MD; Elie Chamas, MD
Dijon, Centre Hospitalier et Universitaire de Dijon: Department of Cardiology, Bocage’s Hospital—Sylvie Falcon, MD (local principal investigator); J.-E. Wolf, MD; Department of Neurology, General Hospital—Maurice Giroud, MD
The following served as committee members of the FAPS study:
Echocardiography—Ariel Cohen, MD, PhD (echocardiography study design and reviewer of all echocardiography examinations); Bernard Bertrand, MD; Christophe Chauvel, MD; Yvette Bernard, MD
Data Monitoring and Coordinating Center—Pierre Amarenco, MD
Data Analysis—Christophe Tzourio, MD, PhD, Unité 360, Recherches Epidémiologiques en Neurologie et Psychopathologie, Institut National de la Santé et de la Recherche Médicale, Paris
This study was supported by grants from the Institut National de la Santé et de la Recherche Médicale (CNEP 92CN23) and from the Direction de la Recherche Clinique de l’Assistance Publique-Hôpitaux de Paris (922601).
Reviews of this article were directed by Mark L. Dyken, MD.
A complete list of the participants in this research study appears at the end of this article.
- Received April 10, 1997.
- Revision received May 28, 1997.
- Accepted May 28, 1997.
- Copyright © 1997 by American Heart Association
Lambl VA. Papillare exkrescenzen an der semilunarklappe des aorta. Wien Med Wochenschr. 1856;6:244-247.
Magarey FR. On the mode of formation of Lambl’s excrescences and their relation to chronic thickening of the mitral valve. J Pathol Bacteriol. 1961;61:203-208.
Lee RJ, Bartzokis T, Yeoh TK, Grogin HR, Choi D, Schnittger I. Enhanced detection of intracardiac sources of cerebral emboli by transesophageal echocardiography. Stroke. 1991;22:734-739.
Tice FD, Slivka AP, Waltz ET, Orsinelli DA, Pearson AC. Mitral valve strands in patients with focal cerebral ischemia. Stroke. 1996;27:1183-1186.
Iung B, Cormier B, Dadez E, Drissi MF, Tsezana R, Viguier E, Caviezel B, Michel PL, Samama M, Vahanian A, Acar J. Small abnormal echos after mitral valve replacement with bileaflet mechanical prostheses: predisposing factors and effect on thromboembolism. J Heart Valve Dis. 1993;2:259-266.
Albers GW, Comess KA, Derook FA, Bracci P, Atwood JE, Bolger A, Hotson J. Transesophageal echocardiographic findings in stroke subtypes. Stroke. 1994;25:23-28.
Benhalima B, Cohen A, Chauvel C, Abergel E, Albo C, Elhadad S, Hobeiche M, Khireddine M, Rozensztajn L, Valty J. Etude morphologique par échocardiographie transoesophagienne et aspects cliniques de la rupture de cordages mitral chez le sujet âgé. Arch Mal Coeur. 1995;88:345-352.
Breslow NE, Day NE. The Analysis of Case-Control Studies. Lyon, France: International Agency for Research in Cancer; 1980:1. Statistical Methods in Cancer Research.
SAS User’s Guide: Volume 1, Version 6. 4th ed. Cary, NC: SAS Institute Inc; 1994.
BMDP Statistical Software Manual. New York, NY: John Wiley & Sons, Ltd; 1992.
Schlesselman JJ. Case-Control Studies: Design, Conduct, Analysis. London, UK: Oxford University Press; 1982.