Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Understand the anatomy and embryology of the interatrial septum and patent foramen ovale (PFO).
Develop an overview of the many clinical associations of a PFO.
Appraise the clinical evidence for and against closure of PFO.
Patent foramen ovale (PFO) is a common finding, occurring in up to 25% of people.1 ,2 An association between PFO and stroke has consistently been seen in up to 50% of patients without an identifiable cause, that is, the so-called cryptogenic stroke (CS) and only in 20% with an identified cause.3 ,4 Many studies have been published testing the hypothesis that paradoxical emboli through a PFO may be implicated, however the available evidence is mixed and conflicting,5 ,6 perhaps in part due to the low recurrence rate and long-term nature of these events. PFOs are associated with numerous other conditions including migraine with aura, decompression sickness, other venoarterial embolic phenomena and platypnoea orthodeoxia.
In this review we will describe the embryological development of the interatrial septum, discuss the diagnosis and clinical associations of PFO, as well as evaluate the available data for and against closure.
Anatomy and embryology
The embryological development of the interatrial septum and foramen ovale is complex, starting at 4–5 weeks post conception with fusion of ventral and dorsal endocardial cushions. Closure of the atrioventricular canal creates two cavities that develop into atria and ventricles and divide into left and right sides. Initially the septum primum grows from the roof of the atria towards the fused endocardial cushion (figure 1A), while the gap, the ostium primum, allows interatrial flow. Before complete atrial separation, a new communication, the ostium secundum, develops by fenestration of the superior region (figure 1B), allowing continuous right-to-left shunting of oxygenated blood from the umbilical arteries bypassing the fetal pulmonary circulation (figure 1C). Infolding of the atrial wall on the right aspect of the septum primum produces the septum secundum to overlap the ostium secundum (figure 1D). The two septae partially fuse and leave an uncovered part of the septum primum forming the fossa ovalis and anterosuperiorly the foramen ovale (figure 1E). Reversal of the pressure gradient on birth approximates the remaining flap and in the majority of the population the linings of ostium primum and ostium secundum fuse permanently closing the foramen.
PFO is believed to be present in up to one in four adults, from postmortem studies. One study of 965 normal hearts documented an increasing size and decreasing prevalence of PFO with age (34.3% up to 30 years, 25.4% in the fourth decade and 20.2% in the ninth and tenth decades).1 In 500 subjects who died due to acquired cardiovascular pathology PFO occurred in 15% of cases.7 Similarly, a PFO prevalence of 24% was found with transoesophageal echocardiography (TOE) among 585 subjects, age 45 years in a stroke prevention study.8 A larger, 1000 patient, TOE study found a lower prevalence of PFO of 9.2%, although this confirmed increasing frequency with age (12.96% vs 6.15% in patients aged 40–49 years vs 70–79 years).9
As most individuals with a PFO are asymptomatic, it is usually an incidental finding at autopsy,10 while antemortem presentation is often with a clinically associated condition and subsequent identification by one of the following methods.
Transcranial Doppler (TCD) studies flow patterns in the middle cerebral artery and, following the peripheral intravenous injection of agitated saline, monitors for air bubbles during normal respiration and manoeuvres that promote right-to-left shunting, such as a sniff and Valsalva. Valsalva manoeuvre increases intrathoracic pressure reducing systemic venous return resulting, on release, in a temporary increase in right atrial pressure producing a gradient allowing right-to-left shunting with a PFO.11 TCD with contrast has a greater sensitivity than transthoracic echocardiography (TTE) with contrast at identifying PFO with reduced specificity12 as it is unable to differentiate atrial, ventricular and pulmonary arteriovenous malformations (pAVMs).
TTE may identify a PFO using colour flow mapping; however, if suspected, TTE should be performed with air/saline or air/saline/blood contrast preferably injected from the femoral vein, with a recent meta-analysis suggesting little difference between agents.13 A PFO is confirmed by contrast in the left atrium within three cardiac cycles following opacification of the right atrium (figure 2); any longer than five cardiac cycles is suggestive of an intrapulmonary shunt.14 ,15 Shunt size can be quantified based by the number of bubbles seen in a single still frame (small: <10; moderate: 10–20; large: >20 bubbles) at rest and after provocative manoeuvres.16 Data is emerging that 3D TTE with contrast may be more effective than 2D, and similar to TOE with contrast at identifying PFO.14 ,17
Closer proximity of the TOE probe to the interatrial septum improves anatomical resolution although the invasive nature impairs Valsalva performance and reproducibility. Given the higher yield in some studies TOE may be used if PFO is not found non-invasively where there is a high clinical suspicion or to confirm the anatomy, including the location of the pulmonary veins, once the PFO has been demonstrated on TTE (figures 3 and 4).
Associations of PFO
Although striking images of large thrombi straddling a PFO following CS lend causality support (figure 6) numerous other conditions have been associated with an increased prevalence of PFO.
The first documented stroke in the presence of PFO was a young woman reported by Cohnheim in 1877.20 Subsequently, numerous studies have shown a significantly higher prevalence of PFO (averaging 40%) among subjects with CS compared with the general population—a significant effect irrespective of age.4 ,21–26 Studies suggest the size of PFO is correlated with risk in CS—suspected paradoxical embolisation was more common with a larger PFO than control or in those with known stroke aetiology.27 A large meta-analysis compared CS with strokes of known causes and demonstrated an OR for PFO of 3.16 (95% CI 2.30 to 4.35), for atrial septal aneurysms (ASAs) 3.65 (95% CI 1.34 to 9.97) and for PFO plus ASA 23.26 (95% CI 5.24 to 103.20).28 Whether anatomical variants, such as a persistent Eustachian valve, aneurysmal septum and prolonged PFO tunnel, are correlated with increased risk simply as they reflect a larger PFO is not clear.
Non-cerebral arterial embolisation
Systemic arterial embolisation to the limbs, gut and kidneys has been reported in association with PFO.29–33 The presence of PFO has been associated with young patients, without cardiovascular risk factors and non-atheromatous arteries, presenting with thrombotic ST elevation myocardial infarction in some,34 ,35 although not all, studies.36 ,37 Published cases of myocardial infarction with PFO often report the presence of pulmonary emboli,38–40 with the raised right-sided pressures increasing right-to-left shunting. A cardiac MRI study of patients with CS found 1 in 10 had subclinical myocardial infarctions on late gadolinium enhancement41—suggesting that this may be a more common manifestation than previously thought.
Thrombotic emboli travelling to the systemic circulation via PFO bypass the mechanical filter of the lung vasculature. Paradoxical infections are well recognised complications among children with congenital intracardiac shunts or adult patients with pAVMs42 Similarly, cerebral abscesses have been reported in patients with sepsis with a previously silent PFO.43
Migraine with aura
Migraines affect approximately 13% of the population aged 20–64 years, with 36% preceded by aura,44 and PFO is associated with migraine with aura in 40–60% compared with 20–30% controls.1 ,45–47 A PFO allows bypassing of the filtration activity of the lungs perhaps leading to increased systemic concentration of nitric oxide, kinins, serotonins or other vasoactive substances triggering migraine attacks.48 ,49 Alternative propositions include deoxygenation of arterial blood causing attacks and long-term shunting, lowering the threshold for migraine attacks.50
Decompression sickness can occur following diving or at high altitudes in pilots and astronauts.51 Inert gas bubbles formed during depressurisation may cause decompression sickness. Usually these migrate through capillaries or lymphatics to the pulmonary circulation and are expired. However, when present in large numbers, they swamp this filter to enter the arterial circulation and, after amplification by further inert gases within peripheral tissues, cause the acute vascular effects of decompression sickness. A PFO or pAVM facilitates the entry of inert bubbles in the arterial circulation. Decompression sickness following a normal non-provocative dive profile, where ascent is at an appropriate pace, is more often associated with a right-to-left shunt than following a provocative dive.52 The risk of decompression sickness also depends on the size of the PFO.53
Other conditions associated with PFO
Obstructive sleep apnoea
It has been suggested that the presence of elevated right heart pressure induces right-to-left shunting across the PFO contributing to hypoxia,54 and an increased severity of nocturnal episodes of hypoxia among patients with Obstructive sleep apnoea (OSA) with PFO may worsen the adverse vascular effects of this condition.55 Similarly desaturation can arise in a patient with transiently raised pulmonary artery pressures, such as following a pulmonary emboli or severe pneumonia.
An interesting entity strongly associated with PFO,56 ,57 platypnoea orthodeoxia is defined as an association of dyspnoea and arterial oxygen desaturation induced by upright posture and relieved by recumbency. The pathology is complex involving an intracardiac shunt, such as a PFO, combined with a mechanism to redirect flow through it—for instance, a persistent Eustachian valve and dilated aortic root causing stretching of the foramen increasing right-to-left shunting. In patients who exhibit such symptoms, closure of the PFO has been shown to be an effective treatment.10 Patients with severe liver dysfunction may develop platypnoea and orthodeoxia due to the hepatopulmonary syndrome, differentiated from a PFO with a bubble study (potentially while standing), which alters the urgency of treatment.58
In addition to associations in prevalence of PFO and events, studies have looked at stroke recurrence rates in the PFO population.
Mas et al59 prospectively examined 581 patients, aged 18–55 years, who suffered an ischaemic stroke without a definite cause. All received secondary prevention with aspirin and underwent TTE and TOE assessment for PFO and ASA at rest; following provocative manoeuvres, 37% were found to have a PFO, 2% an ASA and 9% had both. After a mean follow-up of 3.1 years, recurrent stroke occurred in 4.2% of patients with no interatrial septal abnormality, compared with 2.3% with a PFO alone. However, the rate was increased significantly to 15.2% with both a PFO and ASA. This low recurrence rate and parity without a combined PFO and ASA is consistent with smaller retrospective studies.60–62
Although retrospective studies have shown a strong association between the presence of PFO and CS, prospective studies, looking at the risk of first-time ischaemic stroke, have not shown PFO as an independent risk factor.8 ,63
Trials on PFO closure versus medical treatment in cryptogenic stroke
Medical approaches include the use of antiplatelets like aspirin, clopidogrel, dipyridamole or anticoagulants such as warfarin or novel oral anticoagulants. Some evidence favours anticoagulants in the presence of ASA and PFO,60 however with a higher risk of bleeding.64 ,65 A recent systematic review and meta-analysis of 15 clinical studies of medical treatment following CS or TIA demonstrated an absolute rate of recurrent ischaemic stroke of 1.6 events per 100 person-years (95% CI 1.1 to 2.1).66 In contrast, a retrospective analysis of 10 non-randomised studies of 1355 patients with PFO closure indicated a risk of recurrent stroke of 0–4.9% compared with a significantly increased rate of 3.8–12% in 895 medically managed patients from six trials.67 A large single-centre non-randomised trial by Windecker et al found the risk of stroke over 4 years following PFO closure had a non-significant trend towards fewer events compared with medical treatment (7.8% vs 22.2%, p=0.08). Analysis of patients with more than one cerebrovascular event at baseline found PFO closure was associated with a significantly lower risk of TIA or recurrent stroke (7.3% vs 33.2%, p=0.01).68 A propensity-matched comparison after median follow-up of 9 years showed a reduction in the composite end point (11% vs 21%, p=0.033) driven by a reduction in transient ischaemic attacks in the closure group.69
To date three prospective randomised trials have looked at the comparison of percutaneous closure of PFO versus medical treatment for secondary stroke prevention. A summary of these studies is presented in table 1. First, the CLOSURE 1 trial included 909 patients (age 18–60 years) who suffered CS or TIA within 6-months prior to recruitment and were randomised to PFO closure with a STARFlex device (NMT Medical) (n=447) or best medical treatment (n=462) and followed up for 2 years. The primary end point, a composite of stroke/TIA during 2 years of follow-up, death from any cause during the first 30 days, or death from neurological causes between 31 days and 2 years was 5.5% in the closure group compared with 6.8% in the medical-therapy group (HR=0.78, p=0.37).5 The respective rates were 2.9% vs 3.1% for stroke (p=0.79) and 3.1% vs 4.1% for TIA (p=0.44). The authors concluded that device closure did not offer a greater benefit than medical therapy alone for stroke/TIA prevention. Subsequently, the PC trial studied 414 patients under the age of 60 years who had suffered a previous stroke, TIA or peripheral embolic event. A total of 204 patients were randomised to Amplatzer PFO Occluder (St Jude Medical, Minnesota, USA) and compared with medical treatment in 210 subjects. The primary end point was a composite of death, non-fatal stroke, TIA or peripheral embolism with a mean 4-year follow-up. This end point occurred in 3.4% (7 device patients) vs 5.2% (11 medically treated subjects) (HR 0.63; 95% CI 0.24 to 1.62, p=0.34). Outcome for non-fatal stroke (1 vs 5: HR 0.2; 95% CI 0.02 to 1.72, p=0.14) and TIA (5 vs 7: HR 0.71; 95% CI 0.23 to 2.24, p=0.56) was also non-significant.70
Finally, the RESPECT trial enrolled 980 subjects (age 18–60 years) with PFO and CS within 270 days, randomised to medical therapy (n=481) with one or more antiplatelet (74.8%) or warfarin (25.2%) or PFO closure using the Amplatzer device (n=499). The primary outcome was recurrence of non-fatal stroke, fatal ischaemic stroke or early postrandomisation death defined as all-cause mortality. The mean follow-up was 2.6 years during which a target of 25 of events had occurred. Analysis was complicated by increased dropout in the medical arm resulting in a significantly lower patient/years follow-up. No deaths occurred and in the intention-to-treat cohort, 9 device patients and 16 medically treated patients had a recurrent stroke (HR 0.49; 95% CI 0.22 to 1.11, p=0.08). However, a significant difference was observed in the ‘as-treated’ analysis (5 vs 16; HR 0.27; 95% CI 0.10 to 0.75, p=0.007) as three strokes in the device arm occurred between randomisation and closure.6
These large multicentre randomised trials provide invaluable information, however all three had some important issues and limitations.71 First, all three encountered problems in recruitment due to off-label PFO closure highlighted by the frequency of non-research device closure; an estimated 1 000 000 implants occurred over the same time but only 2203 were recruited to studies. All studies suffered long recruitment times (CLOSURE 1 9 years, PC 13 years and RESPECT 10 years). CLOSURE 1 had to reduce its sample size potentially leading to underpowering, while with RESPECT and PC trials more dropouts occurred in the medical arm, some due to off-label PFO closure (table 1). The power of the PC trial was affected by a smaller observed (5.2%) than expected (12%) event rate in the medically treated patients.
Observational data suggests that patients with large PFO and coexisting ASA are at greater risk, however the CLOSURE 1 and PC trials recruited significant numbers with small shunts and relatively few with ASAs. In the CLOSURE 1 trial, only 52.9% of participants had a moderate or large shunt and only 36.6% had a coexistent ASA. The PC trial included 61% of patients with moderate or large shunts, and only less than a quarter with an ASA (23% in the device group and 24.3% in the medical group). The greater effect seen in the RESPECT trial may reflect the inclusion of 78% large or moderate shunts.
Early studies showed that complete PFO closure reduces risk of recurrent stroke,68 however in the CLOSURE 1 trial almost 14% had a significant degree of shunting at 6 months and at 2 years. Other concerns regarding this device arose as nearly a quarter of the strokes were in the first 30 days, 5.7% suffered postprocedural AF and major bleeding was noted in 2.6%; this could mask any potential benefits. The RESPECT study had significantly lower risk of residual shunt (6%) at 6 months while rates of atrial fibrillation and bleeding were 0.6% and 1.6%, respectively. In the CLOSURE 1 study an alternate cause of stroke or TIA was apparent in 87% of the closure group and 76% of the medically treated group, questioning the original patient workup.
Two recent meta-analyses have been published. The first, by Agarwal et al72 of 10 studies including CLOSURE 1 with 1886 patients compared device closure and medical therapy for recurrent neurological events. Overall recurrence rates were estimated at 0.79/100 patient years (CI 0.48 to 1.05) for device closure and 4.39 (CI 3.20 to 5.59) for medical management (relative risk 0.25 (CI 0.11 to 0.58)).72 A meta-analysis of the randomised control trials found a significant ‘intention-to-treat’ risk reduction for stroke and/or TIA in the device group (pooled HR=0.59; 95% CI 0.36 to 0.97, p=0.04).73
In summary there appears to be a real, albeit small, signal for reduced recurrent neurological events with device closure in carefully selected patients, which is reflected in the recommendations of the UK National Institute for Health and Care Excellence.74
Finally, although PFO closure is associated with higher initial expenditure there is data suggesting it is cost-effective in the long term, although this is based on a clear efficacy model.75 The hope is that the ongoing REDUCE and CLOSE trials, with more targeted inclusion criteria proposed by the RoPE investigators, will help better differentiate those for optimal medical therapy or device closure.76
Patients undergoing PFO closure for non-migraine indications have reported symptomatic improvement.77 A non-randomised study in refractory migraines demonstrated a significant improvement in symptoms and a total elimination of aura in patients undergoing PFO closure.78 The only published randomised study is the Migraine Intervention with STARFlex Technology (MIST) trial.79 This prospective sham-controlled study using the STARFlex device (NMT Medical) included patients with migraine with aura and a moderate or larger right-to-left shunt. The study failed to reach the primary end point of diary recorded headache cessation 6 months after randomisation. Furthermore, no significant difference was found for the secondary end points, assessing the frequency and severity of migraines over 3 months. A post hoc analysis revealed, when two extreme outliers were removed, a significant reduction in the median total headache days following PFO closure.
This negative result could reflect an overly ambitious primary end point or the likely multifactorial nature of migraine triggers, obscuring the effect of reduction of one mechanism. As before, the efficacy of the device has been questioned. Studies in this area are further complicated the fluctuant nature of presentation and severity of by this condition, particularly when a short follow-up is used.80 A follow-up study, the MIST II trial, started recruitment in 2006 but was halted due to poor uptake in 2008.81 Provisional presentation of the PRIMA study, using the Amplatzer occluder device, also failed to show a significant reduction in headache days at 1 year, although very few subjects were randomised to device closure (Transcatheter Therapeutics, 2014, unpublished data). Given the current evidence base, closure of PFO primarily for migraine reduction is not recommended outside of a trial, and should only be considered after thorough workup by an independent neurologist.
There are no prospective randomised studies of PFO closure in divers. A longitudinal study performed by Billinger et al82 followed up 104 recreational divers with a history of major decompression illness (DCI); 39 without PFO, 26 with a PFO who chose to undergo closure and 39 with a PFO who decided against closure. Over 5.3 years’ follow-up of 81 654 dives there were 5 major neurological DCI events—none in the non-PFO group, 0.5 ±2.5/10 000 dives in the PFO closure group and 35.8±102.5/10 000 dives in the PFO non-closure group. An experimental study evaluating 34 divers (19 with PFOs and 15 who had undergone device closure) in a hyperbaric chamber found the same number of venous bubbles was detected by TTE and TCD with complete elimination of arterial bubbles after simulated dives in the closure group.83 Although the data supports closure, it is advised that a cardiologist with a specialist interest in diving medicine reviews patients prior to attribution of causality to a potentially incidental PFO, because differentiation of alternative aetiologies of DCI is complex.
This is a rare condition and there are no published series comparing shunt closure versus conservative management. However, as reported reversal of the documented haemodynamic compromise on closure of the shunt is so clear, device closure is highly recommended in symptomatic individuals.56
Ideally, to improve the quality of the available data, we need the cooperation of clinicians, device companies and regulatory bodies to design and carry out an adequately powered study. This should aim to better risk-stratify patients, to identify the truly CS population, as well as predict risk of future events based on anatomy and shunt dynamics. Device closure with the current technology is associated with serious but infrequent complications and some residual shunt work should continue on better device design to reduce malpositioning, embolisation, clot formation and atrial tachyarrhythmias. Intracardiac echocardiography may allow for shorter procedure times and hospital stays by avoiding sedation and anaesthesia.84 Bioabsorbable technologies or the development of effective radiofrequency ablation techniques may reduce complications such as erosion. There remains an inextricable link between migraine with aura and the presence of PFO,85 and future research in this area would be valuable.
Given the need for careful consideration of numerous factors spanning cardiology, neurology and haematology for individual patients, there is a critical role for the multidisciplinary process in the optimisation of care. Recent developments of the heart team for complex valvular and coronary disorders have laid the essential groundwork for this new multidisciplinary group working.
A PFO is present in up to one in four people and is associated with a host of conditions including CS, DCI, OSA, migraine, paradoxical myocardial infarction and other distal embolisations. A large body of observational data point towards the benefit of device closure, however three randomised controlled trials have failed to send a conclusive answer. There were limitations in study design, recruitment process, inclusion criteria and device characteristics. An adequately powered study including a high proportion of patients with high-risk features using an efficacious device and careful inclusion criteria is still required. As all the current evidence favours intervention, the balance points towards improved outcomes with device closure versus medical therapy in carefully selected patients with CS.
Patent foramen ovale (PFO) is present in up to 25% of individuals.
There is a strong association between the presence of a PFO and cryptogenic stroke (CS), defined as a stroke without an identifiable cause.
Transoesophageal echocardiography with bubble contrast remains the first line of investigation in suspected cases.
The recurrence rate following a CS in those on optimal medical therapy is low.
Three randomised studies of device closure versus medical therapy in secondary prevention following CS have been completed: CLOSURE I, RESPECT and PC trials. All missed their primary end point of reducing the rate of recurrent stroke.
All three studies struggled with slow recruitment, high rate of off-label use of devices, and a heterogeneous study population in terms of PFO size and rate of atrial septal aneurysms.
In the largest study published to date, RESPECT, there was a signal towards improved outcomes with device closure. However further studies are required, and are ongoing, to clarify matters.
You can get CPD/CME credits for Education in Heart
Education in Heart articles are accredited by both the UK Royal College of Physicians (London) and the European Board for Accreditation in Cardiology—you need to answer the accompanying multiple choice questions (MCQs). To access the questions, click on BMJ Learning: Take this module on BMJ Learning from the content box at the top right and bottom left of the online article. For more information please go to: http://heart.bmj.com/misc/education.dtl
RCP credits: Log your activity in your CPD diary online (http://www.rcplondon.ac.uk/members/CPDdiary/index.asp)—pass mark is 80%.
EBAC credits: Print out and retain the BMJ Learning certificate once you have completed the MCQs—pass mark is 60%. EBAC/ EACCME Credits can now be converted to AMA PRA Category 1 CME Credits and are recognised by all National Accreditation Authorities in Europe (http://www.ebac-cme.org/newsite/?hit=men02).
Please note: The MCQs are hosted on BMJ Learning—the best available learning website for medical professionals from the BMJ Group. If prompted, subscribers must sign into Heart with their journal's username and password. All users must also complete a one-time registration on BMJ Learning and subsequently log in (with a BMJ Learning username and password) on every visit.
Curriculum topic: Congenital heart disease in adult patients
KNA and MM contributed equally to this study.
Contributors KNA and MM are equal first authors having provided the original draft and required revisions. AM provided editorial support and wrote the required questions. RR provided editorial support and specific assistance with the images. BC was the overall senior author providing the idea, guidance and editorial support.
Competing interests BC has received speaking fees from St Jude Medical and has worked on advisory boards for Boston Scientific and Medtronic.
Provenance and peer review Commissioned; externally peer reviewed.