Abstract
Objective: To examine the association between patent foramen ovale (PFO) and atrial septal aneurysm (ASA) and stroke. Method:— Data from case-control studies that examined the relative frequency of PFO, ASA, or both, in all patients with ischemic stroke, cryptogenic stroke, and known stroke cause as well as control subjects were included. Trials were categorized by age, clinical comparison, and abnormality. Combined OR were calculated using fixed effect (FE) and random effect (RE) methods.
Results: Comparing patients with ischemic stroke with control subjects using RE, OR for all ages was 1.83 (95% CI, 1.25 to 2.66) for PFO (15 studies), 2.35 (95% CI, 1.46 to 3.77) for ASA (nine studies), and 4.96 (95% CI, 2.37 to 10.39) for PFO plus ASA (four studies). Homogeneous results were found within the group younger than age 55: using FE, OR was 3.10 (95% CI, 2.29 to 4.21) for PFO, 6.14 (95% CI, 2.47 to 15.22) for ASA, and 15.59 (95% CI, 2.83 to 85.87) for PFO plus ASA. For patients older than age 55, using FE, OR was 1.27 (95% CI, 0.80 to 2.01) for PFO, 3.43 (95% CI, 1.89 to 6.22) for ASA, and 5.09 (95% CI, 1.25 to 20.74) for PFO plus ASA. Comparing cryptogenic stroke with known stroke cause, heterogeneous results were derived from total group examination using RE: OR was 3.16 (95% CI, 2.30 to 4.35) for PFO (22 studies), 3.65 (95% CI, 1.34 to 9.97) for ASA (five studies), and 23.26 (95% CI, 5.24 to 103.20) for PFO plus ASA (two studies). In patients younger than age 55, using FE the OR was 6.00 (95% CI, 3.72 to 9.68) for PFO; only one study examined ASA or PFO plus ASA. In patients aged 55 years or older, three studies produced heterogeneous results for PFO: using RE, OR was 2.26 (95% CI, 0.96 to 5.31); no data were available on ASA prevalence.
Conclusions: PFO and ASA are significantly associated with ischemic stroke in patients younger than 55 years. Further studies are needed to establish whether an association exists between PFO and ischemic stroke in those older than 55.
Individual case studies and series in the 1980s postulated a causal role for interatrial septal abnormalities—patent foramen ovale (PFO) and atrial septal aneurysm (ASA)—in the etiology of embolic stroke.1,2⇓ Numerous case-control studies have been published since examining the frequency of interatrial septal abnormalities in patients with stroke or TIA in comparison with control subjects. Similarly designed studies aiming to establish whether such abnormalities were more likely to be found in individuals with “cryptogenic” stroke (stroke without other clear cause) followed. Ambiguous findings have left clinicians uncertain as to whether they should investigate for such abnormalities, and whether positive results should be regarded as causal or incidental. Published reviews have provided an interpretation of the literature without systematically evaluating available evidence.3 Studies differ in the age groups examined and in the control groups chosen, but similarities can be found in design and methodology such that they may be grouped together using simple, clinically intuitive criteria. Although the results of such an exercise cannot address the appropriate management of such patients, a systematic review and meta-analysis has implications for current investigational and therapeutic strategies, and for further research.
Methods.
A systematic search was made using both MEDLINE and BIDS (Bath Information and Data Services) bibliographic databases for the key words patent foramen ovale, atrial septal aneurysm, and right-to-left shunt. This resulted in 2738 references, which were individually assessed. All studies that examined the prevalence of PFO, ASA, or a combination of both in a stroke population were examined in full. All those that met the criteria defined below were included in the meta-analysis. We also searched the bibliographies of all included studies, many excluded studies, and any review articles for additional suitable studies. Both English and foreign language journals were examined. Unpublished data were not sought. Included were 1) case-control studies that compared prevalence of PFO or ASA in patients with ischemic stroke or TIA with nonstroke control patients using a validated diagnostic technique; 2) case-control studies that compared prevalence of PFO or ASA in patients with cryptogenic stroke with patients with known stroke cause using a validated diagnostic technique; and 3) case-control studies that compared prevalence of PFO or ASA in patients with cryptogenic stroke with nonstroke control patients using a validated diagnostic technique. Clear definitions of the italicized terms are given in Definitions (see below).
The three broad comparisons above were each divided into total (patients of all ages), young (patients ≤55 years), and old (≥55 years) groups. Differing inclusion criteria dictated that patients aged 55 may be included in either category. If planned analysis of a different age group above or below age 55 was stated in the objectives or methods section of the paper, results are included within the relevant comparison. If such an age subgroup was studied as a post hoc analysis within a larger study (i.e., it may have been defined and analyzed after knowledge of results) the data were excluded. Two studies divided patients into groups older or younger than age 50,4,5⇓ and one used an age cut-off of 60 years.6 These were included in the separate age analyses as if 55 years was the division used.
Both fixed effects (Mantel-Haenstzel) and random effects (Der-Simonian Laird) methods of meta-analysis were used,7 and the combined OR from each method was tabulated. Using a significance level of 10%, homogeneity of trials was assessed for each comparison. If significant heterogeneity was not detected, results were presented graphically (forest plot) using the fixed effects method. If significant heterogeneity was present, reasons for this were investigated, and the comparison was presented graphically using the random effects method. Sensitivity analysis was performed for each group of trials, and possible sources of bias were examined. Funnel plots were constructed to investigate publication and related bias.8 Explanations of the italicized terms are given in the Appendix.
Definitions.
Atrial septal abnormalities.
Patent foramen ovale. PFO was defined as presence of right-to-left shunt at interatrial level. One study included patients with atrial septal defects in the published data,4 defined as a defect in the septum primum or secundum, or a defect >0.5 cm wide in the region of the fossa ovalis, with left-to-right flow. The investigators reported that the relative frequencies in the case and control groups were not substantially altered if analysis was restricted to patients with PFO (i.e., just those with right-to-left shunt). All other trials excluded patients with atrial septal defects.
Atrial septal aneurysm.
ASA was defined by base width of 1.5 cm or greater, with at least 1.1 cm excursion into either the left or the right atrium, or a sum of the total excursion into the left or right atrium of 1.1 cm or greater.9 Some studies used stricter criteria.10 All studies used transesophageal echocardiography (TEE). Base width was not mentioned in one paper.11
Diagnostic technique.
When more than one technique was used in a study, the most sensitive was chosen for analysis.
Transcranial Doppler ultrasound.
Transcranial Doppler (TCD) ultrasound assessment of PFO involved gelatin or saline contrast. Timing of microbubble spike(s) ranged between 4 and 20 seconds after injection. Examinations were performed at rest, with coughing, and with Valsalva maneuver in most studies.
Transthoracic echocardiography.
Transthoracic echocardiography (TTE) of PFO used gelatin or saline contrast. Diagnosis required appearance of at least one microbubble of contrast in the left atrium within four cardiac cycles of opacification of the right atrium.
Transesophageal echocardiography.
TEE for diagnosis of PFO used gelatin or saline contrast and required contrast opacification (at least 1 microbubble) of the left atrium within 3 seconds or four cardiac cycles of opacification of the right atrium. All studies examining ASA used TEE, and the echocardiographic definition is given above.
Stroke subclassification.
Nonstroke control patients.
This group included normal volunteers, hospitalized patients, or individuals receiving an echocardiographic examination for another reason who were compared with patients with stroke or TIA.
Ischemic stroke or TIA.
This group included patients with sudden clinical focal neurologic deficit consistent with the diagnosis of stroke (confirmed as ischemic by CT or MRI) or TIA (lasting less than 24 hours). One study did not mention cranial imaging.12 One study included patients with peripheral (arterial) embolus13; three others examined such patients, but they were removed from the analysis where possible.6,14,15⇓⇓
Known stroke or TIA cause (noncryptogenic stroke).
Levels of clarification of cause differed considerably between studies. Minimum requirements were assessment of cardiac rhythm (with electrocardiography), assessment of presence of carotid stenosis by Doppler ultrasound (except one study that used a clinical bruit12), and assessment of alternative cardioembolic source (by TTE). Many earlier papers used angiography in their assessment.16-18⇓⇓ Assessment of procoagulant markers differed, and if conducted these were used for classification.11,18-21⇓⇓⇓⇓ Many studies used accepted stroke data bank criteria in their definition of cryptogenic stroke, namely those of the National Institute of Neurological Disorders and Stroke (NINDS)21-24⇓⇓⇓ and the BADISEN (Banco de datos de Ictus de la Sociedad Espanola de Neurologia).25 These systems classify strokes as of determined cause (cardioembolic, lacunar [small vessel], large artery atherosclerotic, or unusual but determined cause) or cryptogenic. All other studies used a similar system, and for this analysis noncryptogenic stroke includes the following causes: arterial dissection, carotid stenosis >50% (one study used stenosis ≥31%5), intracranial atherosclerosis with stenosis >50% of the corresponding vessel, angiitis, migrainous infarction, coagulopathies, systemic disorders (e.g., lupus or Hughes syndrome), atrial fibrillation (chronic or paroxysmal on electrocardiography), recent (within 6 weeks) myocardial infarction, dilated cardiomyopathy, rheumatic mitral stenosis, mitral or aortic vegetations or prostheses, left atrial or left ventricular tumor or thrombus, spontaneous left atrial echo contrast, and complex atheroma between the aortic valve and the left subclavian artery origin. Mitral valve prolapse was regarded as a cause of stroke in some trials,13,21-27⇓⇓⇓⇓⇓⇓⇓ but generally as a risk factor (see below). Two studies28,29⇓ assessed as “undetermined” those individuals with more than one cause of stroke. These patients were defined as having noncryptogenic strokes for this analysis.
Cryptogenic stroke or TIA.
Patients without predetermined cause for stroke, as described above, were defined as having cryptogenic stroke or TIA. One study included patients with atrial fibrillation and spontaneous echo contrast in its cryptogenic stroke group.30 Another classified those with atrial fibrillation on TEE or Holter monitoring as cryptogenic.31 One study excluded (i.e., defined as noncryptogenic) patients with hypertension (blood pressure ≥170/95).12 In all other studies hypertension was regarded as a risk factor.
Risk factors.
Those elements that increased the risk, but were not necessarily the cause of, stroke (e.g., hypertension, diabetes, hypercholesterolemia, smoking, alcohol use, oral contraceptive pill use, migraine, and, generally, mitral valve prolapse [see above]) were termed risk factors. ASA was regarded as a risk factor in studies examining PFO and PFO was regarded as a risk factor in those studies examining ASA.
Results.
Details of the comparisons performed in the total category (all ages) (table 1), in the young (≤55 years) (table 2), and in the old (≥55 years) (table 3) will be considered in turn.
Associations between interatrial septal abnormalities and stroke in patients of any age
Associations between interatrial septal abnormalities and stroke in young patients (≤55 years)
Associations between interatrial septal abnormalities and stroke in older patients (≥55 years)
Patent foramen ovale.
Stroke patients versus nonstroke control subjects.
The “total” comparison of ischemic stroke patients to nonstroke control subjects yielded 15 studies for PFO.4,6,11,13-17,25,26,32-36⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓ Significant heterogeneity was detected, which appeared largely to result from the different ages of participants. If the trials are divided into two separate groups—positive studies6,11,13,15-17,32,33⇓⇓⇓⇓⇓⇓⇓ and neutral or negative studies4,14,25,26,34-36⇓⇓⇓⇓⇓⇓—and the ages of included patients are compared, a difference emerges between the mean age of patients in the positive studies (44.8 years) and in the neutral or negative trials (61.1 years) (p = 0.022). No significant trends were observed in diagnostic technique employed, attempts to blind observers to the clinical diagnosis, choice of control subjects, or retrospective versus prospective design. A funnel plot did not suggest publication bias.
Homogeneous results were obtained in the younger age group (figure 1A). This comparison included nine trials,4,6,11,13,16,17,26,32,33⇓⇓⇓⇓⇓⇓⇓⇓ which produced a common effect measure of 3.10 (95% CI, 2.29 to 4.21). A funnel plot was asymmetric, suggesting possible bias toward the publication of positive results. In the older age group (figure 1B), more heterogeneous results were evident.4,6,13⇓⇓ The OR from both fixed and random effects analysis crossed the line of no effect. The largest study showed the least association, but examination of individual trials provided no explanation for the difference in effect.
Figure 1. Comparison of prevalence of patent foramen ovale (PFO) in patients with ischemic stroke and nonstroke control subjects, classified according to age: less than 55 years (A) and more than 55 years (B). Individual studies are listed on the left; P denotes prospective studies and R retrospective studies. Total patient numbers (N) and those with PFO (n) are shown for the experimental and control groups in each study, and total numbers are provided at the foot of the figure. OR for individual studies are represented by black boxes (▪), the size of which corresponds to the weight attached to each, and are presented with 95% CI (thin black line). Results to the right of the line of no effect (OR = 1) denote a positive association of PFO with ischemic stroke. The combined OR is presented with 95% CI at the bottom right (♦). A chi-square test (for heterogeneity) is shown at the bottom left.
Cryptogenic stroke versus known stroke cause.
This comparison was characterized by numerous small studies estimating a reasonably consistent positive effect. Twenty-two studies were included.4-6,11,13,14,16-28,36-38⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓ Chi-square test for heterogeneity was significant, but the studies that were the source of the heterogeneity detected within the group14,21,28⇓⇓ were well conducted, with clear definitions of cryptogenic stroke. The funnel plot was symmetric.
When considering the younger age group, more consistent results were obtained (figure 2A). Effects were homogeneous (p = 0.29), and the nine eligible studies4,5,11,16-18,21,26,27⇓⇓⇓⇓⇓⇓⇓⇓ gave a symmetric funnel plot. Data were scarce in the older age group, with each study employing a different diagnostic technique (TTE,21 TEE,4 and TCD5) and the two most recent studies estimating a consistent nonsignificant effect (figure 2B).
Figure 2. Prevalence of patent foramen ovale (PFO) in patients with cryptogenic stroke and with known stroke cause, classified according to age: less than 55 years (A) and more than 55 years (B). Individual studies are listed on the left: P denotes prospective studies and R retrospective studies. Total patient numbers (N) and those with PFO (n) are shown for the experimental and control groups in each study, and total numbers are provided at the foot of the figure. OR for individual studies are represented by black boxes (▪), the size of which corresponds to the weight attached to each, and are presented with their 95% CI (thin black line). Results to the right of the line of no effect (OR = 1) denote a positive association of PFO with known stroke cause. The combined OR is presented with 95% CI at the bottom right (♦). A chi-square test (for heterogeneity) is shown at the bottom left in each figure.
Cryptogenic stroke patients versus nonstroke control subjects.
The heterogeneity evident in this group of comparisons enabled trials to be divided into those with a point OR of 1 to 2,4,14,36⇓⇓ those with a point OR of 2 to 46,12,25,26,30,31⇓⇓⇓⇓⇓ and those with a point OR >4.11,13,16,17⇓⇓⇓ Analysis of these three groups showed a trend toward greater patient age in trials with a lower OR (p = 0.08 by analysis of variance [ANOVA]). The groups did not differ in diagnostic technique, presence of blinding, choice of control subjects, or design. The group with an OR of 2 to 4 tended to have less detailed criteria for the diagnosis of cryptogenic stroke, and one trial in each group included lacunar stroke within their “cryptogenic” classification. A funnel plot did not suggest bias. Again, subdivision into age bands revealed much more homogeneous results, both in the young,4,11,16,17,26⇓⇓⇓⇓ and in the old.4,12⇓
Atrial septal aneurysm.
Stroke patients versus nonstroke control subjects.
In this comparison, we found that the negative studies34-36⇓⇓ recruited older patients than did the positive studies6,9,11,15,39,40⇓⇓⇓⇓⇓ (p = 0.014). Once more, funnel plot appearance did not suggest publication bias. Comparisons in the young11,15,39,40⇓⇓⇓ and old15,39⇓ groups were homogeneous, and estimated a clear increased prevalence of ASA in patients with stroke compared with control subjects.
Cryptogenic stroke versus known stroke cause.
The heterogeneity evident in this comparison (total) was due to the presence of one negative–neutral study.28 This study was well conducted (although blinding of echocardiographers was not mentioned) with clear definitions of cryptogenic stroke. It found an increased prevalence of ASA in lacunar stroke. The four remaining studies11,29,36,37⇓⇓⇓ estimated a combined OR of 5.30 (95% CI, 2.70 to 10.42). Only one study defined and studied a younger age group.11 No studies were available for analysis in patients over age 55 years.
Combination of patent foramen ovale and atrial septal aneurysm.
For comparison of patients with stroke and nonstroke control subjects, homogeneous results were obtained for the total group9,11,15,34⇓⇓⇓ and the younger group,11,15⇓ albeit with extremely wide confidence limits. Only one study conducted in old patients was found.15 For cryptogenic stroke versus known stroke cause, assessment yielded a significant OR with a very wide CI in the total group;11,37⇓ there was only one study in the younger group.11 Data on the final comparison for the combination of both lesions (cryptogenic stroke versus nonstroke control subjects) came from a single study.11
Discussion.
The literature on the prevalence of interatrial septal abnormalities in stroke is both extensive and confusing. Erroneous conclusions can be drawn from observational data on a small number of subjects viewed in isolation. A number of studies have failed to show a significant association of interatrial septal abnormalities with stroke in the young.4,13,26,39,40⇓⇓⇓⇓ Consideration of the greater body of literature supporting an association is biased, and fails to provide proper assessment of the level of association. The data presented in this analysis allow a number of firm conclusions to be reached, in addition to suggesting areas requiring further research.
Meta-analysis of observational studies has been criticized because it can generate “spurious precision” from disparate data sets. It may, however, clarify an extensive and confusing literature by organization and collation of the available information. Predefined inclusion criteria, systematic examination of trials to detect sources of heterogeneity and bias, and the consistent results found in separate age bands support the methods presented here. We did not seek unpublished data when conducting this analysis. Publication and related bias was examined by visual inspection of funnel plots, and asymmetry was not often evident. Funnel plots, however, remain an insensitive measure, open to different interpretations.7 Referral bias presents a major source of difficulty, particularly in those studies employing retrospective design or TEE. Some studies included only those referred for TEE, introducing bias.9,15,22-24,34,38⇓⇓⇓⇓⇓⇓ Such populations are likely to be similar to those with “embolic” stroke.39 Numerous studies only investigated for, or reported, either ASA or PFO, so the associations demonstrated may partly be with both lesions in conjunction rather than with either alone. Blinding of observers was often ignored or poorly defined, which is of concern given the degree of interobserver disagreement in diagnosis.41 Such disagreement may result from different diagnostic criteria or from methodological inconsistencies,42 and may partly explain the widely different detection rates in the studies included. For PFO, these ranged from 10% to 44% for stroke, 31% to 77% for cryptogenic stroke, 4% to 25% for known stroke cause, and 3% to 22% for control subjects. For ASA, detection rates ranged from 2% to 17% for stroke, 4% to 25% for cryptogenic stroke, 0.2% to 22% for known stroke cause, and 0% to 15% for control subjects. Prevalence of PFO declines with advancing age, while average lesion size increases.43 Variation in prevalence also stems from the different diagnostic techniques employed. Although the specificity and sensitivity of TCD in comparison with TEE are well defined,44 TTE underestimates the presence of PFO and ASA.45 The different extents to which cryptogenic stroke was investigated for and defined, and the details of that definition, did not exert any clear effect on study conclusion. Recruitment rates were lower in TEE studies (65% to 76%), with investigators experiencing difficulties with patient consent, tolerance of the procedure, and in obtaining satisfactory images.
When classified on the basis of age and abnormality detected, published studies demonstrate a fairly homogeneous effect, despite differences in design. The unequivocal finding that both PFO and ASA are associated with ischemic stroke in the young is important. The implication for planned investigation is that PFO should be sought in young (≤ age 55) patients, and if found should not be regarded as incidental. Although less frequently detected, ASA is more strongly associated with ischemic and cryptogenic stroke than PFO. Because reliance on TCD and TTE will not provide the clinician with accurate information on this and other potentially important cardiac embolic sources,3 this meta-analysis supports wider use of TEE in young patients with stroke.
Less work has been conducted in the older age group (≥ age 55), and the results presented highlight the areas in which data are lacking. The finding that heterogeneity within total comparisons is eliminated by grouping into age bands, and that negative trials are more likely to stem from the inclusion of older patients, suggests that age exerts an effect on the relationship between interatrial septal abnormalities and stroke that is clinically relevant. Other stroke causes and risk factors are more likely to play their part in the old, and the association of PFO with both total and cryptogenic ischemic stroke in those older than age 55 years remains unconfirmed. The OR for cryptogenic stroke compared with known stroke cause in those over age 55 nearly reached significance, whereas results in the stroke versus control comparison were much more equivocal. These findings are consistent with the hypothesis that a true effect exists that is easier to detect once patients with more common causative factors have been excluded.
Does the association demonstrated in the young imply causation? Both abnormalities in conjunction are consistently more strongly associated with ischemic stroke than either alone, and are associated with higher rates of recurrent stroke.46 This, coupled with the reported finding that larger PFO23,26,38⇓⇓ and ASA11 are more strongly associated with cryptogenic stroke than smaller abnormalities and are more likely to lead to recurrence,47 suggests a “dose-response” relationship supporting causality. In addition, the relationship is biologically plausible, cerebral ischemia or infarction resulting either by paradoxical embolism from a venous source or by in situ thrombosis at an atrial level. Despite such arguments, documentation of venous thrombosis in patients whose stroke mechanism is felt to be paradoxical embolism has proved unrewarding.18
Secondary prevention for stroke patients with PFO is a subject of considerable debate. Neither open heart surgery48 nor transcatheter closure49 guarantees freedom from recurrent events. A recent surgical series,50 with flawless clinical outcome, outlined proposed criteria for surgical correction that were reached after many years of both clinical practice and research in this area. The ongoing Patent Foramen Ovale in Cryptogenic Stroke (PICSS) substudy of the Warfarin Aspirin Recurrent Stroke Study (WARSS) may lack sufficient power to determine best medical therapy as stroke recurrence rate is likely to be low,46 and will provide no information regarding the role of corrective techniques. Active investigation for atrial septal abnormalities, and large-scale prospective data collection, as proposed by Kasner et al.,51 may improve estimation of the risks and benefits of the different therapeutic strategies available, and provide useful evidence on which to base therapeutic decisions.
Appendix7,8⇓
The magnitude of statistical diversity that exists between the results of different sets of data is termed heterogeneity. The chi-square test for heterogeneity assesses whether the differences between the effect estimated by each data set can be assumed to be a consequence of random sampling variation. If this test is not significant, outcomes in the individual studies are said to be homogeneous, in that they estimate a single underlying effect.
Fixed effects meta-analysis considers that the difference between the effect found in each study is exclusively due to random variation. Thus, if all the studies were infinitely large they would give identical results. Random effects meta-analysis assumes a different underlying effect for each study and takes this into consideration as an additional source of variation.
A funnel plot is a plot of effect estimate against sample size used to assess validity and to detect bias in meta-analyses. These are skewed and asymmetrical in the presence of publication and other bias. A forest plot is a figure comprising the estimate and 95% CI for each trial, together with an overall estimate and CI.
Acknowledgments
Acknowledgment
The authors thank Iain Sim for technical advice.
Footnotes
-
Additional material related to this article can be found on the Neurology Web site. Go to www.neurology.org and scroll down the Table of Contents for the October 24 issue to find the title link for this article.
- Received February 9, 2000.
- Accepted June 14, 2000.
References
Letters: Rapid online correspondence
REQUIREMENTS
If you are uploading a letter concerning an article:
You must have updated your disclosures within six months: http://submit.neurology.org
Your co-authors must send a completed Publishing Agreement Form to Neurology Staff (not necessary for the lead/corresponding author as the form below will suffice) before you upload your comment.
If you are responding to a comment that was written about an article you originally authored:
You (and co-authors) do not need to fill out forms or check disclosures as author forms are still valid
and apply to letter.
Submission specifications:
- Submissions must be < 200 words with < 5 references. Reference 1 must be the article on which you are commenting.
- Submissions should not have more than 5 authors. (Exception: original author replies can include all original authors of the article)
- Submit only on articles published within 6 months of issue date.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- Submitted comments are subject to editing and editor review prior to posting.
You May Also be Interested in
Hemiplegic Migraine Associated With PRRT2 Variations A Clinical and Genetic Study
Dr. Robert Shapiro and Dr. Amynah Pradhan