Prevention of stroke in patients with patent foramen ovale


  • Conflict of interest: Bernhard Meier and Heinrich P. Mattle are the principal investigators of the PC Trial, which is supported by a grant from AGA Medical. Heinrich Mattle has received research grants, speaker's fees, and consulting honoraria from AGA Medical, and Bernhard Meier from AGA Medical, Johnson & Johnson, Occlutech, St. Jude Medical, and NMT Technologies. Krassen Nedeltchev is the principal investigator of the Swiss PFO Consortium (, which is supported by a grant from the Swiss National Research Foundation.

Heinrich Mattle*, Department of Neurology, Inselspital, University of Bern, Freiburgstrasse 10, CH-3010 Bern, Switzerland. E-mail:


Patent foramen ovale is found in 24% of healthy adults and 38% of patients with cryptogenic stroke. This ratio and case reports indicate that patent foramen ovale and stroke are associated, probably because of paradoxical embolism. In healthy people with patent foramen ovale, embolic events are not more frequent than in controls, and therefore no primary prevention is needed. However, once ischaemic events occur, the risk of recurrence is substantial and prevention becomes an issue. Acetylsalicylic acid and warfarin reduce this risk to the same level as in patients without patent foramen ovale. Patent foramen ovale with a coinciding atrial septal aneurysm, spontaneous or large right-to-left shunt, or multiple ischaemic events potentiates the risk of recurrence. Transcatheter device closure has therefore become an intriguing addition to medical treatment, but its therapeutic value still needs to be confirmed by randomised-controlled trials.


Paradoxical embolism is a commonly suspected cause of stroke, particularly in younger patients. The term paradoxical embolism was coined by J. Cohnheim in 1877 (1) to describe a condition in which emboli of venous origin enter the systemic circulation by being shunted from the right to the left atrium through a patent foramen ovale (PFO).

During foetal life, the PFO is an integral part of the circulation. The collapsed lungs substantially increase the resistance in the pulmonary circulation, requiring the blood to be shunted from the right atrium directly into the left atrium and systemic circulation. After birth, resistance in pulmonary circulation declines and facilitates flow through the lungs. The pressure in the right atrium falls below that of the left atrium, the septum primum is shoved against the septum secundum and the two generally fuse. However, in some persons, a slit-like opening persists, which is called PFO, and gives rise to (momentary) right-to-left shunts (RLS) (Fig. 1).

Figure 1.

 Aspect of patent foramen ovale (PFO) from the left atrial side in a cadaver heart. ASA, atrial septal aneurysm of the septum primum; SS, septum secundum.

A right-to-left shunt can be diagnosed easily using a bubble test and transcranial Doppler sonography (2). The gold standard to diagnose a PFO is transoesophageal echocardiography (3).

Patent foramen ovale is not the only but the most common RLS promoting paradoxical embolism, and with a prevalence of 24% in the general population (4). It accounts for up to 95% of RLS (5, 6), pulmonary shunts for only 5%, and atrial septal defects (ASD) for 1% or even less. In autopsy series, the prevalence of PFO ranges from 17 to 27% (7, 8) and from 16 to 73% in stroke patients (Tables 1 and 2) (9, 10).

Table 1.   Prevalence of PFO in patients with ischaemic stroke and nonstroke controls
with PFO
All nonstroke
% with
PFO in
% of PFOs
in controls
of PFO
=1/OR (in %)
  1. PFO, patent foramen ovale.

Lechat (11)19882460101006·002·6113·80401017<55
Webster (12)198820406405·671·9516·46501518<55
Chen (13)199115347403·721·2910·74441827<55
de Belder (14)199213643564·501·2116·7420522>55
de Belder (14)19925391395·590·6250·2513318<55
Cabanes (15)1993431009503·441·517·83431829<55
Job (16)1994387427631·410·722·77514371<55
Jones (17)19947262193·130·5717·18271132<55
Jones (17)199428194291830·900·511·571416112>55
Zahn (18)19955012011552·861·346·07422035<55
Zahn (18)199515684261·560·465·22221564>55
Del Sette (19)199826738502·901·197·11361634<55
Petty (20)2006392911085190·590·400·881321170All ages
Sum of all studies 323118322512401·691·402·06271859All ages
Table 2.   Prevalence of PFO in patients with cryptogenic strokes and in patients with known cause of stroke
First authorYear
patients with PFO
All cryptogenic
stroke patients
Patients with
known cause
of stroke
and PFO
All stroke
with known
cause of stroke
% with PFO in
patients with
% of PFOs
in controls
of PFO being
(in %)
Age categories
  1. PFO, patent foramen ovale.

Lechat (11)198820414193·571·0112·61492128<55
Webster (12)19881934166·330·6760·16561716<55
Jeanrenaud (21)1990811053·671·409·627300<55
Hausmann (22)1992929041·451·141·853100<55
Hausmann (22)199274510440·830·221·831623160>55
De Belder (14)199293510692·040·745·62261449All ages
Di Tullio (23)1992102112420·912·37184·524845<55
Di Tullio (23)19929246777·102·2022·9638814>55
Di Tullio (24)19939198251·910·566·55473252All ages
Cabanes (15)199336647365·332·0413·94561919<55
Ranoux (25)1993315411417·522·14143·705776<55
Job (16)1994274111333·861·4610·17663326<55
Homma (26)199416367383·541·2410·14441828All ages
Albers (27)1994325391200·280·081·001233353All ages
Jones (17)19944143121·200·216·88292583<55
Jones (17)19941057181371·410·613·27181371>55
Klötsch (28)1994314019719·433·8023·40782711All ages
Zahn (18)19955011818702·121·114·06422647All ages
Schminke (29)199533608404·891·9412·35552020All ages
Yeung (30)199616270152·461·563·875900<55
Yeung (30)1996278917791·590·793·20302263>55
Petty (31)1997225515612·040·924·52402549All ages
Roijer (32)199717673545·781·5920·9525617All ages
Serena (33)19983053381503·842·007·41572526All ages
Steiner (34)1998194212532·821·176·84452335All ages
Kanda (35)19981971304334·912·589·3427720All ages
Petty (20)200622133171581·640·833·25171161All ages
Handke (36)200777227342763·652·335·74341227>55
Force (37)200812622708·161·7538·1019312>55
Sum of all studies 602159433921933·322·853·87381530 
Sum of all studies of age categories <55 years 180336281685·773·659·13541717<55
Sum of all studies of age categories >55 years 142504876832·692·003·62281337>55

Several clinical conditions have been attributed to RLS or paradoxical embolism: ischaemic stroke (38), peripheral embolism (39, 40), brain abscess (41), decompression sickness in scuba divers (42), myocardial infarction (43), refractory hypoxaemia following right-sided myocardial infarction or severe lung disease (44), platypnoea–orthodeoxia syndrome (45), some forms of sleep apnoea syndrome (46), high-altitude pulmonary oedema (47) and migraine with aura (19, 48).

Although PFO is present in 24% of the general population, paradoxical embolism is relatively rare. After age adjustment to the European population, the annual incidence of first-ever ischaemic stroke is 139 per 100 000 inhabitants (49). However, in 60–70% of all ischaemic strokes, other causes are assumed so that the risk attributed to paradoxical embolism was estimated at 28 per 100 000 persons with PFO per year (50). In other words, during an average lifespan of 70 years, only 2% of all PFOs will be symptomatic. Given this low lifetime risk, no primary prevention is currently recommended and screening for PFO in asymptomatic persons can hardly be justified. Prevention of further events in persons with PFO and symptoms is the important issue. This review therefore focuses on secondary prevention.

Association of first-ever ischaemic stroke and PFO

Individual case reports have confirmed Cohnheim's observation of paradoxical embolism through a PFO (51, 52) (Fig. 2). Furthermore, several studies have assessed the frequency of atrial septal abnormalities in patients with stroke and transient ischaemic attacks (TIA) and kindled the interest in PFO and cryptogenic stroke, i.e. in stroke where none of the traditional stroke causes is present (11, 12, 23, 53–55). The authors postulated an association between PFO and ischaemic stroke. In 2000, a meta-analysis summarised the evidence that PFO, atrial septal aneurysm (ASA), or a combination of both abnormalities were more likely to be found in patients with stroke than in stroke-free individuals (38). Trials were categorised by age, clinical comparison, and abnormality.

Figure 2.

 Surgical view of a 1-cm-thick longish thrombus stuck in the patent foramen ovale (PFO) of a 45-year-old man admitted for pulmonary embolism.

Tables 1 and 2 show an updated meta-analysis derived from case–control studies. The updated meta-analysis confirms the association of PFO with stroke both when strokes are compared with non-stroke controls and when cryptogenic strokes are checked against strokes of known cause. The most important data published since Overell's meta-analysis in 2000 are a negative study with strokes and a population-based control group (20) and two positive studies that examined PFOs in elderly patients (36, 37); Petty and colleagues cast doubt on the association of PFO with stroke (56). The frequency of PFOs in their stroke group was only 13% compared with an average of 32% in other studies. This raises the question of whether the techniques to search for PFOs were adequate in Petty's stroke victims. Nevertheless, the updated meta-analysis remains positive for the association of PFO and stroke even after inclusion of this large negative study. On the other hand, the two studies addressing the comparison of PFO in cryptogenic stroke and stroke of known cause strengthened the data, indicating an association of PFO with stroke of unknown cause also in patients older than 55 years. In addition, older patients with cryptogenic stroke and PFO had significantly less severe atherosclerosis of the aorta than patients without PFO, strengthening the likelihood of a pathogenetic role of the PFO (36). Our updated meta-analysis demonstrates a higher prevalence of PFO in ischaemic stroke patients than in stroke-free individuals and a higher prevalence of PFO in cryptogenic stroke patients than in patients with stroke of known cause for all age groups. In addition, ASA concomitant to PFO is more frequent among patients with cryptogenic stroke than in patients with stroke of known cause, both in younger and in older age groups (36, 38).

The anatomical and functional size of the PFO and the mobility of the interatrial septum vary. This might be of clinical relevance and may influence treatment decisions. Patients with cryptogenic strokes have larger PFOs with more extensive RLS than patients with stroke of conventional cause (26, 34). Stroke patients with larger PFOs show more brain imaging features of embolic infarcts than those with small PFOs (34). The presence of a concomitant ASA is independently associated with multiple cerebral ischaemic lesions in PFO stroke, which may indicate an increased embolic risk (57). Another anatomical structure of relevance might be the Eustachian valve (EV). Before birth, the EV directs oxygenated blood from the inferior vena cava towards and across the PFO into the systemic circulation. In patients with a PFO, a persisting EV is frequently encountered. By directing the blood from the inferior vena cava to the interatrial septum, a persisting EV may prevent spontaneous closure of PFO after birth and may, therefore, doubly predispose to a paradoxical embolism (58).

The amount of shunting seems to be clinically significant as well. On the one hand, a case–control study linked increasing RLS at rest and high membrane mobility with a higher risk of recurrent brain embolism (59), and on the other, there is a high correlation between the size of the PFO on transoesophageal echocardiography and the amount of microembolic signals on transcranial Doppler sonography in stroke and TIA patients (60). In addition, atrial septal mobility predicts the degree of RLS across PFOs (61). Increasing deviation of the mobile interatrial septum is associated with larger PFOs, giving rise to more shunting, and thus a greater opportunity for a paradoxical embolism and an increased risk for stroke (61–63).

A hypercoagulable state might contribute to the risk of paradoxical emboli in PFO patients as well. At least two studies have shown an increased frequency of prothrombotic genotypes in such patients (64, 65), while other studies found no difference between PFO patients and controls (66, 67).

The current data indicate that there is a biological gradient of the risk of stroke associated with PFO, which grows with increasing amount of shunting and can be further enhanced by other factors such as a hypercoagulable state. ASA or EV in this context can be regarded as a marker of increased shunting. This biological gradient may also indicate that PFO is not only associated with cryptogenic stroke by chance but is also causally related via paradoxical emboli as a mediator. The presence of such a pathogenetic mechanism is supported by the PELVIS study with a 20% frequency of pelvic vein thrombosis in acute stroke patients with PFO as opposed to a rate of only 4% in patients with a known cause of stroke (68). It is rather amazing that the current diagnostic techniques were able to pick up pelvic vein thrombosis in 20% of cryptogenic stroke patients, considering the small diameter of thrombi causing occlusion of cerebral vessels and stroke in relation to the large size of veins. This might also explain why only a few calf thromboses are found as a source of cryptogenic strokes, which has been quoted often to argue against the mechanism of a paradoxical embolism through PFOs. Similar to the PELVIS study, a population-based cohort study showed a substantially increased long-term risk of subsequent arterial cardiovascular events including stroke in patients with venous thromboembolism also supporting the mechanism of paradoxical embolisation (69). Arrhythmias including atrial fibrillation related to the septal abnormality might represent an alternative mechanism by which PFO could induce strokes (70, 71). It is also conceivable that thrombi formed in the PFO tunnel or in the ASA are causing strokes. However, if this was a frequent mechanism, dislodgement of thrombi by the catheter crossing the PFO tunnel during PFO closure would be expected but such an event has yet to be described.

Risk of first-ever ischaemic stroke in persons with PFO

As outlined above, the association between PFO and ischaemic stroke has been demonstrated by individual case reports and case–control studies. Two population-based studies addressed the issue of PFO-related first-ever stroke risk prospectively: one in the Olmsted county, MN, and the other in the northern Manhattan population of New York (4, 9). Both studies failed to establish the role of PFO as an independent risk factor for stroke in the general population. There was only a non-significant trend towards a higher stroke incidence in persons with PFO in both studies.

The Olmsted county study (Stroke Prevention: Assessment of Risk in a Community, SPARC) enrolled 588 randomly sampled subjects, whose mean age was 67 years (4). A PFO was identified in 24% and an ASA in 2% of subjects. During a median follow-up of 5·1 years, cerebrovascular events (cerebrovascular-related death, ischaemic stroke, and TIA) occurred in 41 subjects (7%). After adjustment for age and comorbidity, PFO was not a significant independent predictor of stroke (hazard ratio 1·46, 95% CI 0·74–2·88, P=0·28). Even when looking at large-size PFOs, a significant risk for cerebrovascular events was not detected. The risk of stroke among subjects with ASA was nearly fourfold greater than that in those without ASA, but the proportional hazards regression analysis did not establish statistical significance (hazard ratio 3·72, 95% CI 0·88–15·71, P=0·074). The relatively small sample size and the advanced age of the study participants evoked some criticism within the scientific community (72, 73). The sample size of 1100 stroke-free subjects in the Northern Manhattan study (NOMAS) (9) was almost double that of the SPARC study. Patent foramen ovale was detected in 15% and ASA in 3% of the participants. During a follow-up of 6·6 years, ischaemic stroke occurred in 68 subjects (6%). After adjustment for demographic and risk factors, PFO was not found to be significantly associated with stroke (hazard ratio 1·64, 95% CI 0·87–3·09). Isolated ASA was associated with elevated stroke incidence (hazard ratio 3·66, 95% CI 0·88–15·30), but the coexistence of PFO and ASA was not (hazard ratio 1·25, 95% CI 0·17–9·24). Transthoracic echocardiography was used to detect PFO instead of the more sensitive transoesophageal echocardiography. Underdiagnosis of interatrial shunts may be responsible for the relatively low prevalence of PFO detected in the NOMAS study and may have biased the results.

Both SPARC and NOMAS suggest that the risk of stroke from a PFO in the general population is low. While the identification of PFOs that are just ‘innocent bystanders’ with a low risk does not appear to be meaningful, searching for PFOs that may expose a healthy individual to a relevant stroke risk does. Future studies might have to address the issue, which PFOs are dangerous enough to warrant primary preventive measures.

Risk of recurrent ischaemic stroke in patients with PFO

The natural course after cryptogenic stroke in patients with PFO needs to be further elucidated. Currently, the following treatment options for the secondary prevention of paradoxical embolism are in clinical use or evaluation: (1) antithrombotic treatment (ATT) with antiplatelet agents or anticoagulants, (2) transcatheter device closure (TDC) of PFO, (3) percutaneous suture closure of PFO, (4) percutaneous device-less closure and (5) surgical closure of PFO by open access or thoracoscopy.

Stroke recurrence under ATT

The Quality Standards Subcommittee of the American Academy of Neurology (AAN) recently published evidence-based guidelines for the management of patients with recurrent stroke, PFO, and ASA, after a critical review of the literature (74). The recommendations were largely based on one randomised-controlled trial (RCT) with blinded outcome assessments, the Patent Foramen Ovale in Cryptogenic Stroke Study (PICSS) (75), and three prospective-matched cohort studies (59, 76, 77). The first cohort study, the Lausanne study, reported 3·8% of recurrent, non-fatal, cerebral ischaemic events (stroke or TIA) per year during 3 years of prospective follow-up in 140 patients (76). Treatment type, administered in a non-random fashion (antiplatelets, anticoagulants, and PFO closure), was not related to stroke recurrence in the multivariate analysis. The largest cohort study, the French PFO/ASA study, was an unblinded trial of 581 patients with stroke. Of these, 37% of the patients had a PFO and 1·7% had a PFO combined with an ASA (77). All received acetylsalicylic acid (300 mg per day). After 4 years of follow-up, the risk of recurrent stroke was 2·3% in patients with PFO compared with 4·2% in patients without cardiac abnormalities. In the ‘La Sapienza’ study, recurrent cerebral ischaemic events (strokes or TIAs) occurred in 7·2% of 74 patients after a mean follow-up of 3 years, which is equal to 2·4%/year (59). Patients with TEE evidence of RLS at rest (i.e. without Valsalva provocation) or interatrial septal hypermobility were at an increased risk of stroke recurrence. If a composite of stroke and death was considered, the annual risk under ATT in patients with isolated PFO ranged from 1·5 to 3·7% in the cohort studies. The annual risk reported by the only RCT (PICSS) was twofold higher (7·2%), possibly due to age-related effects (mean age, 59 years compared with 44–53 years in the cohort studies).

PICSS was a sub-study of the randomised-controlled Warfarin-Aspirin Recurrent Stroke Study (WARSS). It compared acetylsalicylic acid and anticoagulation for the secondary prevention of stroke (75). A PFO was found in 39% of 265 patients with cryptogenic stroke and in 30% of 365 patients with a stroke of determined aetiology (P<0·02). For the entire group, the 2-year recurrent stroke or death rates did not differ between patients with and without a PFO (14·3% vs. 12·7%). In addition, when the groups with and without a PFO were analysed in relation to the efficacy of warfarin (mean international normalised ratio 2·04) or acetylsalicylic acid (325 mg/day), no significant differences were found. Neither the size of the PFO nor the coincidence of an ASA was associated with an increased risk of recurrent cerebrovascular events. In PICSS, acetylsalicylic acid and warfarin were equal in preventing recurrent stroke or death. In the subgroup of patients after cryptogenic stroke, the 2-year rate of recurrent stroke or death was lower, but not significantly, in patients with anticoagulation compared with antiplatelet therapy (8·8% vs. 16·9%, OR 0·47, 95% CI 0·22–1·04).

The main conclusion of the Quality Standards Subcommittee of the AAN was that PFO alone does not portend an increased risk of subsequent stroke or death in patients who have had a cryptogenic stroke and are treated medically (74). There were insufficient data to draw conclusions about isolated ASA. The results regarding patients with the combination of PFO and ASA were somewhat inconsistent: the French PFO/ASA study indicated a significantly increased risk, 15·2% (95% CI 1·8–28·6) after 4 years, while PICSS did not establish any association between the presence of PFO and ASA with stroke or death. The available data failed to demonstrate a difference between warfarin and acetylsalicylic acid effects on the risk of subsequent stroke or death among patients with a cryptogenic stroke and atrial septal abnormalities. However, it is important to note that there is a group of patients who should always be treated with anticoagulants for at least 3 months: those with concomitant deep vein thrombosis or pulmonary embolism (78).

Anatomical and functional studies and the French PFO/ASA study indicate that a coexisting ASA is a substantial and probably the most important potentiator of stroke risk in patients with PFO. ASA not only prevents spontaneous closure of the foramen after birth but it also produces frequent openings of the PFO cleft, perhaps even with every heart beat. Patients with PFO and ASA could be defined as ‘high-risk’ patients for recurrent events. There is also evidence that PFO patients with more than one previous event are at an increased risk of recurrent cerebral ischaemia (79). Furthermore, PFO size, degree of functional shunting, EV, and a coexisting hypercoagulable state likely are additional risk factors that could place a patient at high risk and need to be studied further.

Stroke recurrence after PFO closure

In addition to ATT, various procedures are used to close the PFO with the intention of preventing subsequent paradoxical embolism. Open-heart surgery is effective to close the PFO, but entails a broad range of perioperative complications, although mostly minor (80–82). Transcatheter PFO closure using radiofrequency thermal energy is less invasive but performed too poorly so far to be used in clinical routine (83, 84), and suture-based technology will require further refinements before safety and efficacy trials can be initiated (85). At present, percutaneous TDC is the most widely and probably the safest technique used to close PFOs.


Transcatheter device closure of ASD was first introduced in 1974 (86). The method consists of positioning a double-umbrella device on both sides of the interatrial septum through an approach over a femoral vein (Fig. 3). Several patient series treated with such devices have been reported, with recurrent cerebrovascular ischaemic events ranging from 0% to 3·4% (87–92). In the largest series reported to date, the annual recurrence rate of ischaemic events among 525 patients was 1·1% (six strokes, nine TIA, and two peripheral emboli in 1534 patient years) (93). Transcatheter device implantation is less invasive than open-heart surgery; nevertheless, complications can also occur. They include atrial wall perforation with pericardial effusion, device dislodgement or embolisation, device fractures, early and late free wall erosions, aortic regurgitation, thrombosis on the device and thrombus embolisation, septal fibrosis, arrhythmias, venous access complications, air embolism through the transseptal sheath, and death (94). Growing experience with more careful manipulation of catheters, wires, and devices, however, has increased the success rate of implantation close to 100% and decreased the rate of significant complications to <1%. In the large series of Wahl et al. (93), there were (2·5%) complications without any long-term sequelae. Significant complications reported to a manufacturer are even fewer with free wall erosions of only two among more than 11 000 implanted devices (95). Some authors advocate the use of transoesophageal or intracardiac echocardiography to guide device placement (96), but there is a growing group of cardiologists placing devices with the help of right atrial opacification with contrast medium under fluoroscopy (93). This seems to be more efficient, patient-friendly, and at least as safe compared with using transoesophageal or intracardiac ultrasound guidance.

Figure 3.

 Transoesophageal echocardiography showing the passage through a patent foramen ovale (PFO) (dashed arrow). The bottom panel shows a double-umbrella PFO occluder before (left) and after (right) release from the pusher cable. The PFO passage (dashed arrows) is delineated by a contrast medium injection.

The PFO occluding devices have different amounts of metal and fabric in their structure. With time, local inflammatory responses with lymphocytic infiltrates, the presence of foreign body giant cells, the cellular organisation of initial fibrin deposits, and a persisting immune response encapsulate the device by fibrous tissue and occlude the PFO permanently (97). Transoesophageal echocardiography at 6 months showed complete closure in 86% of patients, and a minimal, moderate, or large residual shunt in 9%, 3%, and 2%, respectively, with the most commonly used device (93). A residual shunt was a risk factor for recurrent ischaemic events (hazard ratio=3·4; 95% CI 1·3–9·2). Over the past few years, the market has seen numerous PFO occluding devices. Some have disappeared for a variety of reasons (98). A randomised trial has compared procedural complications and 30-day clinical outcome of three devices in 660 patients with a history of paradoxical embolism (99). The devices were the Amplatzer (Amplatzer PFO occluder, Plymouth, MN, USA), Helex (Gore HELEX Septal Occluder, Flagstaff, AZ, USA), and CardioSEAL-STAR-Flex (CardioSEAL Septal Occlusion System, Boston, MA, USA), and there were 220 patients per group. All PFO closures were technically successful. Exchange of devices or device embolisations that were retrieved was most frequently required for the Helex occluder, and thrombus formation on the device and paroxysmal atrial fibrillation were more common with the CardioSEAL-STAR-Flex occluder. Complete closure at 30 days was more frequent in the Amplatzer (65%, P=0·0005) and CardioSEAL-STAR-Flex (62%, P=0·0003) groups than in the Helex group (53%).

Even if device technology is advanced today and the delivery of devices is proven safe in skilled hands, the efficacy of the TDC still needs to be documented in randomised trials.

Comparisons of ATT and PFO occluding devices

Comparisons between ATT and TDC in patients with cryptogenic stroke are scarce and currently at the level of case–control studies. Windecker et al. (92) compared the risk of recurrent stroke in 308 patients with PFO who underwent TDC or received ATT alone. At 4 years of follow-up, TDC resulted in a non-significant trend towards risk reduction of recurrent stroke/TIA (8% vs. 22%; P=0·08) compared with ATT. Patients with more than one cerebrovascular event at baseline and those with complete PFO occlusion at follow-up after device implantation were at a lower risk for recurrent stroke or TIA compared with medically treated patients (7% vs. 33%, P=0·01; 7% vs. 22%, P=0·04, respectively). Schuchlenz and colleagues studied 280 patients treated with antiplatelets (n=66), anticoagulation (n=47), or TDC (n=167) and followed them for a mean of 2·6 years. The annual rate of recurrent cerebrovascular events was 0·6% after TDC compared with oral anticoagulation (5·6%, hazard ratio 0·06, 95% CI 0·12–0·29, P<0·001) and antiplatelets (13%) (100).

Patients with PFO and ASA constitute a population with a higher risk of cerebral ischaemia than patients with PFO alone (38, 77). Therefore, the question arises as to whether TDC is as effective to treat PFO and ASA compared with PFO alone. Wahl et al. (101) compared TDC in 141 patients with ASA and PFO and one or more paradoxical embolism and 220 patients with PFO alone. The procedure effectively abolished RLS and decreased septal hypermobility at the same time. Freedom from recurrent TIA, stroke, and peripheral emboli was 95% (ASA+PFO) and 94% (PFO) at 4 years, indicating that long-term prevention of recurrent events was as favourable in PFO+ASA patients as in patients with PFO alone.

Wöhrle and colleagues compiled recurrence rates of several series of PFO patients with medical treatment or after TDC. The annual rate of stroke or TIA after TDC was 1·3% (95% CI 1·0–1·8), which was lower than the annual rate of 5·2% (95% CI 4·4–6·2%) with medical treatment (94). This result is independent of the initial presence of an ASA and contrasts with the recurrence rate in patients with PFO and ASA during antiplatelet therapy (77).

To date, no results of RCT to compare recurrent ischaemic events after TDC closure or with medical treatment are available. There are five ongoing RCTs (PC-Trial in Europe, Canada, Brazil, and Australia, CLOSE in France, CLOSURE1 and RESPECT in USA, and GORE Reduce in USA and Scandinavia). Whenever possible, patients with cryptogenic stroke and PFO and their treating physicians should be encouraged to participate in one of these trials in order to clarify the best secondary preventive measure.

Unresolved issues in the management of patients with stroke and PFO

The annual risk of stroke recurrence is unlikely to exceed 1–2% in those aged <55 years. With advancing age, other risk factors and causes of stroke become more prevalent. As a result, the relative risk (RR) posed by a PFO decreases from 3·1 in patients aged <55 years to 1·1 in those older than 55 years (50). Assuming a 10% recurrence rate at 5 years in PFO patients receiving ATT and a 5% recurrence rate after TDC, an RCT has to include 620 patients per treatment arm in order to demonstrate a significant difference (α=0·95; power 90%). Assuming only a 5% recurrence rate in the ATT and a 2·5% rate in the TDC arm, the number of patients needed for the trial increases to 5098. Given the low recurrence rates under ATT and the risk of complications after TDC, it is crucial to identify ‘high-risk’ patients with a higher recurrence rate who may derive more or the most benefit from TDC. However, the effect of TDC in ‘high-risk’ patients has never been systematically tested. There is little doubt that RCT that compare ATT with TDC would provide powerful evidence for the optimal treatment of patients with cryptogenic stroke and PFO. Nevertheless, results from the ongoing randomised trials are not to be expected for the near future. Even if these trials are completed successfully, statistical differences may remain undetected with the planned sample sizes of 410 patients in the PC-trial, 900 in CLOSE, 800 in CLOSURE, 1500 patients in RESPECT, and 664 in GORE Reduce. Alternative data-gathering strategies, e.g. with multicentre prospective registries, are urgently needed.


The evidence linking PFO to cryptogenic stroke is strong. The meta-analysis of all case–control studies shows a consistent association, particularly when PFO is combined with ASA. The probability that PFO is an incidental finding is equal to one divided by the odds ratio. In all patients with cryptogenic strokes, this probability is 30% (95% CI 26–35%), 17% in patients younger than 55 years (95% CI 11–27%), and 37% in patients older than 55 years (95% CI 28–50%). An association makes pathophysiological sense and can be explained by paradoxical embolisation, which has been observed in several patients in vivo or autopsy and can be inferred and temporally related from venous thromboembolic studies as well. Moreover, there exists a biological gradient between the amount of shunting and the risk of recurrent stroke. One could even question the attribute ‘cryptogenic’ to a stroke that occurs without another identifiable aetiology but a septal shunt. According to current evidence, PFO could be regarded as an independent risk factor for stroke similar to hypertension, smoking, diabetes, hypercholesterolaemia, and other established stroke risk factors, although with less weight than the traditional risk factors. However, such a statement requires proof of a causal relationship between septal abnormality and stroke. To date, a causal relationship is likely but not proved. The association of PFO and stroke could be a chance finding. We all know that the selection of cases can bias case–control studies and give false answers that are not reproducible in RCT. A recent example is hormone replacement therapy, which was recommended after case–control studies, but is not advisable after the randomised-controlled women's health initiative (102). The results of case–control studies need to be confirmed by RCTs, and until results of such trials are available, we do not know whether there is actually a causal association and whether closure of the PFO has a beneficial effect.

Table 3 summarises the recommendations for the management of patients with septal abnormalities after ischaemic events according to current stroke guidelines. The AAN guidelines question the association of PFO with an increased risk of stroke (74). According to the AAN guidelines, antiplatelet agents are the prevention of first choice. The AHA/ASA and ESO guidelines also recommend antiplatelet agents for secondary prevention, while patients with hypercoagulable states or vein thrombosis should be anticoagulated (103–105). When strokes recur, the AHA/ASA and ESO guidelines recommend considering PFO closure. The ESO guidelines also recommend PFO closure for other high-risk patients, but leave the definition of ‘high risk’ open. ASA is probably the most important potentiator of the risk of PFO, but further research is needed to define such patients.

Table 3.   Recommendations of guidelines for management of patients with septal abnormalities after ischaemic events
GuidelineStatements and recommendations
Quality Standards Subcommittee of the AAN (74)PFO is not associated with an increased risk of subsequent stroke or death among medically treated patients with cryptogenic stroke.
The co-incidence of PFO and ASA possibly increases the risk of subsequent stroke (but not death) in medically treated patients aged <55 years.
In patients with a cryptogenic stroke and an atrial septal abnormality (PFO, ASD, or ASA), the evidence is insufficient to determine whether warfarin or acetylsalicylic acid is superior in preventing recurrent stroke or death, but minor bleeding is more frequent with warfarin.
There is insufficient evidence to evaluate the efficacy of surgical or endovascular closure.
American Stroke Association/American Heart Association (AHA/ASA) (103)For patients with an ischaemic stroke or TIA and a PFO, antiplatelet therapy is reasonable to prevent a recurrent event (Class IIa, Level B).
Warfarin is reasonable for high-risk patients who have other indications for oral anticoagulation such as those with an underlying hypercoagulable state or evidence of venous thrombosis (Class IIa, Level C).
Insufficient data exist to make a recommendation about PFO closure in patients with a first stroke and a PFO. PFO closure may be considered for patients with recurrent cryptogenic stroke despite medical therapy (Class IIb, Level C).
European Stroke Organisation (ESO) (104)Antiplatelet therapy after TIA or stroke.
In the presence of proven deep vein thrombosis (DVT) or atrial septal aneurysm, anticoagulation is recommended [Class IV, good clinical practice (GCP) (105)].
Consider endovascular closure of PFO in patients with cryptogenic stroke and high-risk PFO (Class IV, GCP) (104) guidelines.

Search strategy and selection criteria. References for this review were selected from the personal data bases of the authors collected for the purpose of clinical research on PFO. In addition, articles on PFO, PubMed, and Google were systematically searched for articles and studies related to PFO until the end of November 2008. Search terms were ‘patent foramen ovale’ and ‘paradoxical embolism’. PFO, patent foramen ovale.


We thank Pietro Ballinari for statistical advice.