Author + information
- Received July 28, 2005
- Revision received September 26, 2005
- Accepted October 3, 2005
- Published online January 17, 2006.
- Irene Meissner, MD⁎∥,⁎ (, )
- Bijoy K. Khandheria, MD†,
- John A. Heit, MD⁎,
- George W. Petty, MD⁎,
- Sheldon G. Sheps, MD‡,
- Gary L. Schwartz, MD‡,
- Jack P. Whisnant, MD§,
- David O. Wiebers, MD⁎∥,
- Jody L. Covalt¶,
- Tanya M. Petterson∥,
- Teresa J.H. Christianson∥ and
- Yoram Agmon, MD⁎
- ↵⁎Reprint requests and correspondence:
Dr. Irene Meissner, Department of Neurology, Mayo Clinic, 200 First Street SW, Rochester, Minnesota 55905
Objectives We sought to determine the association between patent foramen ovale (PFO), atrial septal aneurysm (ASA), and stroke prospectively in a unselected population sample.
Background The disputed relationship between PFO and stroke reflects methodologic weaknesses in studies using invalid controls, unblinded transesophageal echocardiography examinations, and data that are unadjusted for age or comorbidity.
Methods The use of transesophageal echocardiography to identify PFO was performed by a single echocardiographer using standardized definitions in 585 randomly sampled, Olmsted County (Minnesota) subjects age 45 years or older participating in the Stroke Prevention: Assessment of Risk in a Community (SPARC) study.
Results A PFO was identified in 140 (24.3%) subjects and ASA in 11 (1.9%) subjects. Of the 140 subjects with PFO, 6 (4.3%) had an ASA; of the 437 subjects without PFO, 5 had an ASA (1.1%, two-sided Fisher exact test, p = 0.028). During a median follow-up of 5.1 years, cerebrovascular events (cerebrovascular disease-related death, ischemic stroke, transient ischemic attack) occurred in 41 subjects. After adjustment for age and comorbidity, PFO was not a significant independent predictor of stroke (hazard ratio 1.46, 95% confidence interval 0.74 to 2.88, p = 0.28). The risk of a cerebrovascular event among subjects with ASA was nearly four times higher than that in those without ASA (hazard ratio 3.72, 95% confidence interval 0.88 to 15.71, p = 0.074).
Conclusions These prospective population-based data suggest that, after correction for age and comorbidity, PFO is not an independent risk factor for future cerebrovascular events in the general population. A larger study is required to test the putative stroke risk associated with ASA.
The relationship between paradoxical (right-to-left) embolism from a patent foramen ovale (PFO) and stroke remains controversial despite numerous studies implicating this condition with the risk of stroke. Methodologic weaknesses driving the lack of conclusive results in published data include the retrospective, circumstantial, and anecdotal nature of analyses; poor selection of controls; and, importantly, data that are unadjusted for age and comorbidity (1). Cardiac embolism causes an estimated one-fifth of ischemic strokes (2), 60,000 to 110,000 of which result from PFO according to some accounts (1,3,4). The variability in reported stroke risk from PFO ranges from <1% (5,6) to as high as 17.5% (7). These figures underscore the magnitude of this controversy.
The Stroke Prevention: Assessment of Risk in a Community (SPARC) study was designed to estimate the prevalence of potential risk factors for stroke in the population. The prevalence of PFO was found to be 25.6% (8), which is consistent with autopsy reports of the frequency of PFO in patients with normal hearts (3,9,10). This high prevalence of PFO in combination with considerable enthusiasm for PFO closure, as a result of the recent advent of catheter-closure devices, creates a pressing need for prospective, methodologically sound data regarding the long-term risk of PFO. The SPARC study we describe is the largest transesophageal echocardiography (TEE)-based study of stroke outcome in a random sample of the population.
The resources of the Rochester Epidemiology Project in Rochester, Minnesota, were used to enumerate and randomly sample the entire 45-years-or-older Olmsted County population. These study subjects were not volunteers. The study design and initial results of the first phase of the SPARC study have been published previously (8,11). An age- and gender-stratified random sample of the Olmsted County population consisting of 1,475 residents were chosen, of whom 230 were ineligible by predetermined exclusion criteria (terminal illness, dementia precluding informed consent, severe disability, or esophageal disease precluding TEE), and 657 declined to participate. The final SPARC study sample consisted of 588 randomly sampled subjects (approximately one-half of those eligible) who consented to multimodality testing, including medical record review, home medical interview, TEE, carotid ultrasonography, and repeated blood pressure measurements. A comparison of comorbid conditions, age, and gender among participants and a random sample of eligible non-participants (decliners) showed no significant differences, thus confirming that participants were a representative sample of the population (11). This prospective follow-up study was approved by the Mayo Foundation Institutional Review Board.
End point definitions were specified before initiation of the study, and updated information was obtained from the medical records of the subjects five years after the original SPARC study assessment. Three of the 588 subjects declined authorization for medical research under Minnesota Statute 144.335. Baseline data collected for these three subjects indicated that none had PFO; one had atrial septal aneurysm (ASA). Of the 585 subjects available for follow-up, 8 were missing PFO assessment (7 had unsuccessful TEE and 1 was uncooperative for the cough and Valsalva maneuvers). TEE was successful in 577 subjects.
Baseline transthoracic echocardiography and multiplane TEE (12) were conducted according to standard practice guidelines with commercially available ultrasonographic instruments (8,13). The heart and thoracic aorta were scanned for the presence of potential embolism sources (14). The interatrial septum (IAS) was viewed primarily in the transverse midesophageal four-chamber view and the longitudinal biatrial-bicaval view. All TEEs were reviewed systematically by a cardiologist who was blinded to the subject’s comorbid history.
Patent foramen ovale
A PFO was defined as a right-to-left interatrial shunt diagnosed with intravenous injections of agitated saline while the patient was at rest and with provocative maneuvers (cough or release of Valsalva or both). The PFO grade was dichotomized as maximal shunting ≤10 versus >10 bubbles. Visualization of a “hole” only was not sufficient for this analysis.
We defined ASA according to criteria previously published by Agmon et al. (13) and Hanley et al. (15): 1) diameter of the base of the aneurismal portion of the IAS 15 mm or more and either 2) protrusion of the IAS, or part of it, 15 mm or more beyond the plane of the IAS or 3) phasic excursion of the IAS during the cardiorespiratory cycle 15 mm or more in total amplitude.
The data resources of the Rochester Epidemiology Project (16) were used to identify the inception cohort of residents of Olmsted County, Minnesota, with a first lifetime deep venous thrombosis, pulmonary embolism, or chronic thromboembolic pulmonary hypertension during the 30-year period 1966 through 1995. Consistent criteria were used to identify venous thromboembolism (VTE) (17). Only those incident VTEs occurring before entry into the SPARC study were considered as contributing to the baseline risk of cerebrovascular events.
Cerebrovascular events (CVDs) were transient ischemic attack, cerebral infarct, or death (either primary or contributing cause) as the result of the aforementioned conditions.
Survival free of CVDs was estimated with the Kaplan-Meier product-limit method, starting at the initial SPARC visit. Subjects were censored at the last medical visit if they were lost to follow-up (two subjects), at the date of the medical record abstraction, or at the date of death if not due to CVD. Univariate Cox proportional hazards regression analyses unadjusted, adjusted for age and male gender and adjusted for previously derived risk factors (previous myocardial infarction, atrial fibrillation, age, and gender), were used to determine the effect of PFO, ASA, and VTE on CVD. An exploratory stepwise Cox proportional hazards regression model included age and gender. All other baseline variables were allowed to compete, including previous myocardial infarction, atrial fibrillation, PFO, ASA, VTE, and the remaining variables listed in Table 1.A p value of 0.05 was used to determine whether a variable would enter or leave the model. A follow-up analysis was done in which a p value of 0.10 was used to determine whether a variable would enter or leave the model. All two-way interactions between all variables in the final stepwise model were explored whenever sample size was sufficient. The proportional hazards assumption was examined for the final stepwise multivariate risk factor model. The proportional hazards assumption is invalid if the ratio of the hazards depends on time.
Forty-one subjects had a qualifying stroke. Fifty-one subjects died before having CVD and were censored because of death. Of the 585 subjects, 493 (84.3%) were alive at the censor date. Patent foramen ovale was found in 140 of 577 subjects (24.3%); 8 subjects were missing the PFO measurement (TEE unsuccessful). Eight additional subjects had visualization only. Eleven subjects had an ASA. Of the 140 subjects with PFO, 6 (4.3%) had an ASA; of the 437 subjects without PFO, 5 (1.1%) had an ASA (two-sided Fisher exact test, p = 0.028). A total of 12 subjects had a previous VTE: 6 had a PFO, 5 did not have a PFO, and 1 was missing the PFO measurement (uncooperative for cough and Valsalva maneuvers) (Table 2).
Of the 140 subjects with PFO, 12 had a cerebrovascular ischemic event at follow-up; none of the 12 subjects had associated ASA. The Kaplan-Meier estimate of survival free of CVD at five years was 91% for subjects with PFO and 93% for those without PFO (Fig. 1).Patent foramen ovale was not associated with an increase in the hazard of CVD in any of the models (all p values ≥ 0.25) (Table 3).No interactions between PFO and any of the adjusting variables were statistically significant. Although the hazard ratios were slightly more than 1.0 (range, 1.27 to 1.46), the 95% confidence intervals were wide and contained 1.0. Furthermore, PFO grade (Table 1) was not related to CVD in the univariate and multivariate analyses adjusted for age, gender, and other risk factors (all p values > 0.75).
The Kaplan-Meier estimate of survival free of CVD at five years was 81% for those with ASA and 93% for those without ASA (Fig. 2).Atrial septal aneurysm was strongly, but not significantly, associated with an increase in the hazard of CVD. The hazard of CVD in subjects with ASA was nearly three times higher than the hazard in those without ASA in the unadjusted model and the model adjusting for age and gender (p ≤ 0.16), although only 2 of the 41 subjects with subsequent CVD had ASA; neither had associated PFO. After adjustment for age, gender, and clinical risk factors, the hazard was nearly four times higher (hazard ratio 3.72, 95% confidence interval 0.88 to 15.71, p = 0.07) (Table 3).
The Kaplan-Meier estimate of survival free of CVD at four years was 83% for subjects with VTE and 93% for those without VTE (Fig. 3).Venous thromboembolism was not associated with an increase in the hazard of CVD (all p ≥ 0.20) (Table 3). Estimates of the hazard ratio varied widely (0.77 to 2.52 for the completely adjusted and unadjusted models, respectively). No interactions between previous VTE and the adjusting variables were statistically significant (all p > 0.30).
Results from both stepwise models are shown in Table 4.As in previously reported analyses (18), when the p value to enter and leave was set to 0.05, previous myocardial infarction and previous atrial fibrillation were the only variables to enter the model after age and gender were included. When the p value to enter and leave was set at 0.10, ASA and a history of diabetes entered the model, in addition to previous myocardial infarction and atrial fibrillation, and were marginally significant.
In this population-based study, a detailed analysis of multiple variables found that PFO is a highly prevalent condition (24.3%). After adjustment for age and risk factors, PFO was not an independent predictor of CVD. More subjects with a PFO had an ASA (4%) than did subjects without a PFO (1%, p = 0.03), but having a PFO was not significantly associated with an increase in the risk of a cerebrovascular end point.
A previous incident of VTE also was not significantly associated with an increase in CVD. The association between ASA and CVD is less clear. Only 11 subjects had an ASA, 2 of whom had a cerebral ischemic event; hence, the estimated increased risk was nearly four-fold. However, despite the small numbers, this association is consistent univariately and after risk-factor adjustment, a suggestion that it is not spurious. Given the small number of strokes in this population and the even smaller number of subjects with ASA, clearly a larger study is needed to test for an association between ASA and subsequent stroke (19).
Why the difference between results from the SPARC study and the literature? Review of the PFO detection rates reported in the literature show wide variability, suggesting an underestimation of PFO in patients with stroke of known cause and an overdetection rate in tertiary referral centers where nonincident, recurrent strokes are evaluated (20–25). This presumption is supported by a recent Mayo Clinic study that compared patients undergoing TEE for noncryptogenic stroke with those referred for cryptogenic stroke. The frequency of PFO was lowest in patients with cardioembolic stroke at 10.5%, a finding suggesting that the unblinding of echocardiographers to mechanisms of stroke at the time of TEE (e.g., atrial fibrillation, valvular disease, ventricular thrombus) may have resulted in lower intensity efforts to search for PFO. A comparison of contrast injections between controls and patients also supported an underdiagnosis of PFO in patients with known causes of stroke. Cardiac disorders in which left atrial pressures are increased also may limit the sensitivity of contrast echocardiography for detecting right-to-left shunting (19).
Reported detection rates for patients referred for cryptogenic stroke approach 50%, whereas the rates are low for patients with stroke of known cause (Fig. 4).The combination of overdetection of PFO in referred cases of cryptogenic stroke and underdiagnosis of PFO in patients with known causes of stroke may result in an overestimation of the reported hazard ratios for cryptogenic stroke associated with PFO. The similar detection rates of PFO in a random population sample and in population-based cases of cryptogenic stroke challenge the popular opinion that PFO is necessary and causal for stroke occurrence. The prospective, population-based, random subject selection design of this study obviates the biases in a study evaluating cause and effect once the end point has already occurred. This study is unique in that it is the largest prospective randomly selected, population-based TEE study to date. However, there were relatively few cerebrovascular events, a factor potentially limiting the ability for this study to detect a statistically significant hazard. The 95% confidence interval for PFO indicates that the hazard ratio could be as low as 0.65 or as high as 2.5 and still be consistent with this study’s conclusions. However, the hazard ratio estimate for PFO was very modest (1.28), did not change after adjusting for age and gender, and increased only slightly after additionally adjusting for myocardial infarction and atrial fibrillation—findings suggesting that this result would hold true with more cerebrovascular events, even if the 95% confidence interval narrowed with a larger sample size. These are the risks for the population as a whole; in an individual patient, PFO may be associated with an embolic cerebrovascular event.
Patent foramen ovale is a common occurrence, occurring in one in four subjects in the population. Because of methodologic inconsistencies, detection rates of PFO in different populations vary widely, leading to potentially inaccurate conclusions regarding its association with stroke. Patent foramen ovale does not seem to be an independent risk factor for stroke in the general population. These data should provide guidance for management strategies for patients with PFO, particularly those with comorbid cardiovascular conditions. The stroke risk related to ASA is less clear and will require further follow-up of larger population samples.
Supported in part by research grant NS-06663 from the National Institute of Neurological Disorders and Stroke.
- Abbreviations and Acronyms
- atrial septal aneurysm
- cerebrovascular events
- interatrial septum
- patent foramen ovale
- Stroke Prevention: Assessment of Risk in a Community study
- transesophageal echocardiography
- venous thromboembolism
- Received July 28, 2005.
- Revision received September 26, 2005.
- Accepted October 3, 2005.
- American College of Cardiology Foundation
- Meier B.,
- Lock J.E.
- Albers G.W.,
- Comess K.A.,
- DeRook F.A.,
- et al.
- Ranoux D.,
- Cohen A.,
- Cabanes L.,
- et al.
- Holmes D.R. Jr.,
- Cabalka A.
- Agmon Y.,
- Khandheria B.K.,
- Meissner I.,
- et al.
- Hanley P.C.,
- Tajik A.J.,
- Hynes J.K.,
- et al.
- Meissner I.,
- Khandheria B.K.,
- Sheps S.G.,
- et al.
- Cabanes L.,
- Mas J.L.,
- Cohen A.,
- et al.