Is Patent Foramen Ovale a Modifiable Risk Factor for Stroke Recurrence?
Although the prevalence of a patent foramen ovale (PFO) in the general population is ≈25%, it is approximately doubled among cryptogenic stroke (CS) patients. This has generally been attributed to paradoxical embolism, and many physicians recommend PFO closure to prevent recurrence. However, the benefit of PFO closure in patients with stroke has not been demonstrated. Furthermore, the epidemiology of stroke recurrence in patients with CS with PFO versus without PFO and in those with large right-to-left shunts versus small right-to-left shunts has yielded results that appear difficult to reconcile with the hypothesis that paradoxical embolism is an important cause of stroke recurrence. The purpose of this review is to critically examine the epidemiologic evidence that PFO is a potentially modifiable risk factor for stroke recurrence in patients with CS. The evidence suggests that many patients with CS and PFO have strokes that are PFO attributable, but many have strokes that are unrelated to their PFO. We introduce the concept of “PFO propensity,” defined as the patient-specific probability of finding a PFO in a patient with CS on the basis of age and other risk factors. We show that this value is directly related to the probability that CS is PFO attributable. Because there is substantial heterogeneity in both PFO propensity and recurrence risk among patients with PFO and CS, stratification for PFO closure by these joint probabilities will likely prove crucial for appropriate patient selection.
Even though it was first proposed as a cause of stroke in 1877, paradoxical embolism through a patent foramen ovale (PFO) remains controversial.1 To some, it is a common cause of cryptogenic stroke (CS), whereas to others, it is a medical curiosity. Some authors express doubts regarding the importance of paradoxical embolism, even when a PFO is discovered in a patient with no other apparent stroke etiology.2 Others advocate the prophylactic closure of nearly any discovered PFO,3 even though incidental PFOs can be found in ≈25% of autopsies.4 The debate has been intensified in equal measure by the absence of definitive, randomized data regarding the effectiveness of PFO closure for stroke and the confusing and paradoxical nature of the epidemiologic evidence that PFO is a risk factor for stroke recurrence. The purpose of this review is to critically examine that evidence.
Is PFO a Risk Factor for Stroke Recurrence?
The PFO in Cryptogenic Stroke Study (PICSS) was one of the largest studies to examine the role of PFO in stroke.5 Figure 1 shows stroke recurrence rates in patients enrolled in PICCS with and without PFO. This figure, which demonstrates near-identical stroke recurrence in patients with and without PFO, is often invoked by those who caution against the routine closure of PFO in patients with stroke, the apparent logic being that if PFO is not associated with an increased risk of stroke recurrence, then PFO closure is unlikely to reduce this risk. Furthermore, when PFO patients in this trial were subgrouped into those with small versus large PFOs, there was a strong trend for higher recurrence rates in those with smaller PFOs; the 2-year event rates in patients with no, small, and large PFO were 15.4%, 18.5%, and 9.5%, respectively. This again appears to speak against the importance of paradoxical embolism as a risk factor for stroke and stroke recurrence.
However, this straightforward interpretation of the results of PICSS is likely to be misleading for several reasons. First, despite its name, PICSS enrolled all patients with stroke for whom transesophageal echocardiography (TEE) was performed, including both CS patients and patients with a stroke of determined cause. Indeed, only 250 of the 630 patients enrolled in PICSS had a CS. Thus, most patients in PICSS had PFO-unrelated strokes, whether a PFO was discovered at TEE or not. Indeed, when understood in this light, the trend to lower recurrence in patients with large versus small PFO becomes understandable: the group of patients with smaller PFOs has a stroke recurrence rate similar to those without a PFO because they have stroke mechanisms that are similar; the group of patients with larger PFOs is presumably “enriched” for patients with PFO-related CSs, which may have a lower recurrence risk than do other stroke subtypes. Thus, the fact that patients with large PFOs have the same or even a lower recurrence rate than do those without a PFO does not imply that large PFOs do not increase the risk of recurrence in the patients who actually have them.
Indeed, similar results were seen among patients with CS. A recent meta-analysis by Almelkhafi et al6 examined 4 studies that compared recurrence rates in patients with CS, with versus without PFO. They determined that the presence of a PFO does not increase the risk of recurrence of either stroke or transient ischemic attack (pooled relative risk with vs without PFO=1.1; 95% CI, 0.8 to 1.5) or of stroke alone (pooled relative risk=0.8; 95% CI, 0.5 to 1.3). However, just as in PICSS (which included all stroke patients), these similarities in recurrence risks among those with CS does not imply that PFO is not a risk factor for stroke recurrence in patients who have them. The result simply indicates that if a PFO is pathogenic, then the strokes that it causes are roughly as likely to recur as CSs caused by other occult mechanism (eg, occult paroxysmal atrial fibrillation or subthreshold aortic atheroembolic disease).
Indeed, one of the most consistent findings is that CS patients with PFO have a substantially lower prevalence of conventional stroke risk factors than do CS patients without PFO. Conventional stroke risk factors in the 3 largest studies examining stroke recurrence in CS are shown in Table 1 for patients with and without PFO.7–9 These large and consistent differences have several important implications: (1) PFO is pathogenically important in the index CS, because otherwise, patients with and without this anatomic variant would be expected to be similar; (2) PFO is an important risk factor for stroke recurrence because it single-handedly compensates for the shortfall in other risk factors such that those with and without PFO have similar recurrence rates; and (3) even before a TEE is obtained, the presence or absence of conventional stroke risk factors can be used to estimate the likelihood of finding a PFO. It can be seen that, among CS patients, younger patients without hypertension and/or diabetes and/or coronary artery disease are much more likely to have a PFO than do patients with these risk factors.
Another finding that has been somewhat confusing is that CS patients with small and large right-to-left shunts have roughly equivalent stroke recurrence rates. This has been found consistently in the 3 largest studies that have examined this issue,7,9,10 whether shunting is measured by TEE or transcranial Doppler. The absence of an effect of shunt size might be explained in part by the poor reliability of shunt measurement, which depends on multiple factors including volume status, procedural technique, and interreader reliability. However, another explanation also deserves attention: patients with smaller PFOs may be more likely to have occult mechanisms other than paradoxical embolism, with recurrence risks that are the same or even higher than for those with paradoxical embolism. The finding of similar recurrence risks should therefore not be interpreted as implying that CS patients with small and large shunts found on TEE or transcranial Doppler are equally likely to have a future paradoxical embolism.
PFO in CS: Pathogenic or Incidental?
Thus, the fact that patients with a CS and a PFO may have stroke mechanisms that are either PFO related (for example, paradoxical emboli) or PFO unrelated and that it is not possible with certainty to segregate these 2 groups, substantially complicates the clinical epidemiology as it does the clinical care of these patients. Closing a PFO that is totally incidental to a stroke is not likely to decrease stroke recurrence and will only add procedural and device-related risks. Although it is typically not possible in an individual patient to determine with certainty whether a PFO discovered in the setting of a CS is pathogenic (as opposed to incidental), it is possible to determine the fraction of CSs attributable to the presence of a PFO among patients in whom a PFO is discovered.
This is shown in Figure 2, with a PFO prevalence of 40% among those with CS and 25% among a control population. If one assumes that the prevalence of PFO among those with CS unrelated to a PFO is 25% (based on the control rate), the figure suggests that ≈50% of PFOs discovered in the setting of CS would be incidental. This is an application of Bayes’ theorem. Using Bayes’ theorem to solve for the probability that PFOs are incidental yields the following Equation11,12:
Probability that PFO is incidental in CS cases=prevalence of PFO in controls×(1−prevalence of PFO in CS cases)/prevalence of PFO in CS cases×(1−prevalence of PFO in controls).
Thus, as PFO prevalence in CS patients decreases, the probability that a discovered PFO will be incidental increases. If the prevalence of PFO in CS patients were equivalent to that in a control population (for example, 25%, then the probability that a discovered PFO is incidental increases to 100% (that is, this would be the expected rate if PFO were not a risk factor for cryptogenic stroke).
The equation has another interesting property in that the right-hand side is numerically equivalent to the inverse of the odds ratio in case-control studies.11,13 This permits the easy conversion of the odds ratio of case-control studies associating the prevalence of PFO in CS patients versus controls to the more clinically relevant estimation of the probability that a discovered PFO is incidental.
Figure 3 shows the results of a meta-analysis of case-control studies comparing PFO prevalence in patients with CS versus those with stroke of known cause. Only studies without age exclusions were included in this analysis. Panel A shows the results expressed conventionally as an odds ratio and shows that 16 of 17 studies demonstrated a higher prevalence of PFO in CS patients compared with patients with stroke of known cause, a remarkably strong and consistent result. According to the Bayesian transformation (panel B), these results yield an estimated summary probability of 33% (95% CI, 28% to 39%) that a PFO discovered in a patient with a CS is incidental.11
However, there is tremendous between-study heterogeneity in this estimate. Between-study differences that might account for this variability include the stringency of the operational definition of CS (that is, the work-up either required or typically performed to rule out other causes), with more stringent studies yielding lower probabilities of incidental PFO. Bias due to more rigorous TEE examination of CS patients compared with that in patients with known mechanism (that is, in unblinded studies) would tend to diminish the estimate of incidental strokes.
The characteristics of patients enrolled in a study can also affect the PFO-attributable fraction among patients with PFO and CS (Table 2). Among studies enrolling only younger patients, the estimated probability of an incidental PFO is considerably lower; among older patients, considerably higher. The data suggest that when a PFO is found together with an atrial septal aneurysm in the setting of a CS, this finding is rarely incidental.11
The equation can also be used to estimate, in an individual patient, the likelihood that a discovered PFO is, or is not, incidental, based on the prevalence of PFO in CS patients with similar characteristics. We call this individualized prevalence estimate “PFO propensity” (that is, the probability that a CS patient has a PFO or not based on other characteristics). As shown in the Table, among CS patients, PFO propensity will be increased by younger age and the absence of conventional stroke risk factors. Figure 4 illustrates the relation between the PFO-attributable fraction (that is, the fraction of CSs attributable to the PFO in those with both CS and PFO) based on PFO propensity, from the equation. Because conventional logistic-regression modeling can be used to estimate PFO propensity, these models can be used to stratify patients by probability that a PFO is pathogenic versus incidental.
We have shown that there is compelling yet indirect evidence that PFO is an important risk factor for stroke recurrence after an initial CS. However, the stroke recurrence rate in patients with PFO and CS is relatively low, estimated at ≈2% per year.4 It is also clear that many of the recurrent strokes, as with the index events, may not be due to paradoxical embolism. Indeed, even among the proportion of patients in whom a stroke is PFO attributable, other mechanisms aside from paradoxical embolism have been proposed, such as in situ thrombus formation or increased susceptibility to atrial fibrillation.14,15 Not all mechanisms would be addressed with mechanical closure.
Thus, the margin of potential benefit for PFO closure is narrow. Even a relatively low rate of procedure- and device-related complications could nullify most or all of the potential benefit. Although case series consistently show very low rates of stroke recurrence after device placement and suggest benefit for closure compared with medically treated patients,16 for interventions with a low margin for benefit, these observational studies are unreliable, and we must await the completion of randomized, clinical trials, which are ongoing.
Regardless of the findings of trials, this discussion suggests that some patients might benefit from PFO closure and others might be harmed, whereas most will remain stroke-free with or without the implanted device. Careful patient selection for closure and for the trials that study closure will therefore be crucial. Where risk is determined by multiple patient characteristics, use of risk models may be important for selecting patients who can benefit.17 Risk stratification strategies for PFO closure, however, need to include stratification not only for stroke recurrence risk but also, more importantly, for the probability that the index event was itself PFO attributable. Even if many CS patients can benefit from closure, testing the procedure in a population of CS patients with many incidental PFOs may falsely suggest that the procedure is of no benefit.
Source of Funding
This work was partially funded by National Institutes of Health/National Institute of Neurological Disorders and Stroke grant R01 NS062153.
David Thaler is a consultant to AGA Medical. David Kent reports no conflicts of interest.
- Received June 29, 2010.
- Accepted July 16, 2010.
Tong DC, Becker KJ. Patent foramen ovale and recurrent stroke: closure is the best option: no. Stroke. 2004; 35: 804–805.
Meier B. Closure of patent foramen ovale: technique, pitfalls, complications, and follow up. Heart. 2005; 91: 444–448.
Homma S, Sacco RL. Patent foramen ovale and stroke. Circulation. 2005; 112: 1063–1072.
Homma S, Sacco RL, Di Tullio MR, Sciacca RR, Mohr JP. Effect of medical treatment in stroke patients with patent foramen ovale: patent foramen ovale in Cryptogenic Stroke Study. Circulation. 2002; 105: 2625–2631.
Almekhlafi MA, Wilton SB, Rabi DM, Ghali WA, Lorenzetti DL, Hill MD. Recurrent cerebral ischemia in medically treated patent foramen ovale: a meta-analysis. Neurology. 2009; 73: 89–97.
Serena J, Marti-Fabregas J, Santamarina E, Rodriguez JJ, Perez-Ayuso MJ, Masjuan J, Segura T, Gallego J, Davalos A. Recurrent stroke and massive right-to-left shunt: results from the prospective Spanish multicenter (CODICIA) study. Stroke. 2008; 39: 3131–3136.
Lamy C, Giannesini C, Zuber M, Arquizan C, Meder JF, Trystram D, Coste J, Mas JL. Clinical and imaging findings in cryptogenic stroke patients with and without patent foramen ovale: the PFO-ASA Study. Atrial Septal Aneurysm. Stroke. 2002; 33: 706–711.
Alsheikh-Ali AA, Thaler DE, Kent DM. Patent foramen ovale in cryptogenic stroke. incidental or pathogenic? Stroke. 2009; 40: 2349–2355.
Berthet K, Lavergne T, Cohen A, Guize L, Bousser MG, Le Heuzey JY, Amarenco P. Significant association of atrial vulnerability with atrial septal abnormalities in young patients with ischemic stroke of unknown cause. Stroke. 2000; 31: 398–403.
Mugge A, Daniel WG, Angermann C, Spes C, Khandheria BK, Kronzon I, Freedberg RS, Keren A, Denning K, Engberding R. Atrial septal aneurysm in adult patients: a multicenter study using transthoracic and transesophageal echocardiography. Circulation. 1995; 91: 2785–2792.