Article Text

Download PDFPDF

Patent foramen ovale
  1. Vafa Alakbarzade1,
  2. Tracey Keteepe-Arachi2,
  3. Nazia Karsan2,
  4. Robin Ray2,
  5. Anthony C Pereira2
  1. 1Neurology, Royal Cornwall Hospitals NHS Trust, Truro, UK
  2. 2Saint George’s University Hospitals NHS Foundation Trust, London, UK
  1. Correspondence to Dr Vafa Alakbarzade, Royal Cornwall Hospitals NHS Trust, Truro TR1 3LQ, UK; vafa.alakbarzade.10{at}


Patent foramen ovale (PFO) is the most common anatomical cause of an interatrial shunt. It is usually asymptomatic but may cause paradoxical embolism, manifesting as stroke, myocardial infarction or visceral/peripheral ischaemia. PFO is a risk factor for stroke and may be associated with migraine with aura. New evidence suggests PFO closure reduces the risk of recurrent ischaemic stroke in a highly selected population of stroke survivors: those aged 60 years or younger with a cryptogenic stroke syndrome, a large right-to-left shunt, an atrial septal aneurysm and no evidence of atrial fibrillation. They benefit from percutaneous PFO closure in addition to antiplatelet therapy, rather than antiplatelet therapy alone. Current evidence does not support PFO closure in the treatment of migraine.

  • stroke
  • migraine
  • interventional
  • cardiology
  • cerebrovascular disease

Statistics from

Request Permissions

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.

What is a patent foramen ovale?

Patent foramen ovale (PFO) is a remnant of the fetal circulation persisting into adulthood1 (figure 1). It is implicated in several complex conditions including cryptogenic stroke, migraine, platypnoea–orthodeoxia syndrome and decompression sickness in divers.2–4

Figure 1

Patent foramen ovale anatomy.

Prenatally, oxygenated blood bypasses the lungs. The foramen ovale, a flap-like opening between the atrial septum primum and secundum, facilitates oxygenated fetal blood flow from the inferior vena cava through the right atrium, across the atrial septum to the systemic circulation. After birth, a fall in pulmonary vascular resistance in response to breathing and oxygenation reduces right heart pressure, which, coupled with rising left atrial pressures from increased pulmonary venous return, pushes the septum primum against the septum secundum, closing the foramen ovale.

In 75% of people, foramen ovale closure is complete by the age of 2. However, it persists in 25%, making it the most common congenital heart defect.5 An autopsy study from the 1930s reported a 29% incidence of small (2–5mm) defects and a 6% incidence of large (6–10mm) defects.6 A 1984 study corroborated these findings reporting an incidence of 27%.7 The cause of PFOs is poorly understood, but there may be a genetic predisposition.8


Postmortem studies have found that the mean PFO slit width in adults in 4.9 mm (range 1–19 mm).7 The average PFO size increases with age, suggesting that large PFOs may remain patent whereas smaller defects may close even during adulthood.7 The risk of paradoxical embolism rises with larger PFO size.9 Its anatomy is important to determine risk and to optimise treatment, informing the choice of specific closure device.10


Atrial septal aneurysm is significantly associated with PFO; it may even predict the presence of a PFO, raising the odds of one being present by 4.6 times.11 A transoesophageal echocardiography study comparing patients suffering embolic stroke (n=133) with controls (n=277) identified an atrial septal aneurysm in 15% of the stroke group versus 4% of controls.12 Furthermore, saline contrast injection identified PFO in 70% of the stroke group and 75% of controls with atrial septal aneurysms. Transthoracic echocardiography detected only 37.5% of atrial septal aneurysms.

The ‘Chiari network’ is a poorly understood embryological remnant that forms a web-like structure attached to the wall of the right atrium near the opening of the inferior vena cava. Although not generally considered pathological, one study of 1436 adults with PFO found that 83% also had a Chiari network.13 This study highlighted the functional implications of the Chiari network, with right-to-left shunting more common compared with controls (55% vs 12%, p<0.001), as well as being more common in those suffering cryptogenic stroke.


PFO may be a family trait conferring a genetic predisposition to stroke. In a study of 62 patients with ischaemic stroke and 62 age-matched and sex-matched control siblings, siblings of patients with PFO had a significantly higher stroke prevalence than siblings of those without PFO (61.5% vs 30.6%, p=0.015).14 Interestingly, this finding was specific to females.


PFO detection often relies on demonstrating a right-to-left shunt rather than actually seeing the PFO itself. Commonly used imaging modalities include echocardiography (transthoracic or transoesophageal) and transcranial Doppler. PFO diagnosis by transcranial Doppler involves identifying a shunt by detecting air microemboli in the middle cerebral artery following peripheral injection of agitated saline. Transoesophageal echocardiography remains the gold standard for detecting and visualising a PFO, with one study reporting a prevalence of 39% using transoesophageal echocardiography compared with 18% on transthoracic echocardiography and 27% on transcranial Doppler.15 Transoesophageal echocardiography also detected six PFOs smaller than 2 mm in size that had not been detected by transcranial Doppler, emphasising its superior resolution.

PFO detection using transoesophageal echocardiography involves injecting agitated saline into the cephalic vein in the antecubital fossa. Microbubbles then appear in the right atrium. The presence of a PFO enables microbubbles to pass into the left atrium, usually within a few heart beats of their appearance in the right atrium. Shunting is predominantly right to left, although there is some reverse shunting as the relative atrial pressures vary during the cardiac cycle and respiration. During transthoracic echocardiography, Valsalva techniques may raise the right atrial pressure, increasing the sensitivity for PFO detection16 (figure 2).

Figure 2

Patent foramen ovale (PFO) detected on transthoracic echocardiography: (A) transthoracic echocardiography detecting large PFO without saline contrast use; (B) transthoracic echocardiography with Valsalva induced rise in the right atrial pressure and increase saline contrast passage through the PFO. Note that the Valsalva technique causes the cardiac chambers to become smaller (confirming that the Valsalva has been effective) and then concentrates the bubbles in the right atrium.

The amount of contrast visualised in the left atrium does not allow an accurate estimate of the PFO size.17 However, imaging does allow risk stratification by identifying anatomical findings such as the presence of atrial septal aneurysm (table 1). Transoesophageal echocardiography is recommended to assess atrial septal anatomy in detail before PFO closure.17

Table 1


A PFO causes no harm in the vast majority of people. Therefore, incidental PFOs are not routinely closed. Closure can be performed percutaneously or surgically. Post-procedure antiplatelet medication or anticoagulation is indicated for 6 months and possibly longer in some patients.18 Thrombus formation is highest within the first 4 weeks after device implantation but extremely rare after 8–12 months.19 20

Percutaneous closure

Percutaneous closure may be undertaken using local anaesthetic with sedation or under general anaesthetic. It is performed under radiographic and echocardiographic guidance normally within 2 hours. A catheter is inserted into the femoral vein and the closure device introduced and advanced across the PFO and then deployed. There are several licensed closure devices. The in-patient stay is usually 1 day and the patient can return to normal activities within a few weeks.

Surgical closure

Surgical closure of PFOs is rare in the era of percutaneous equipment and lacks compelling trial evidence. In exceptional circumstances, surgical closure with thoracotomy and direct suturing of the PFO defect is a feasible option. Previously regarded as the gold standard for definitive closure of PFOs, studies have refuted this, citing high rates of residual shunting and postoperative recurrent embolism.21

Complications of PFO closure

Serious complications include death, haemorrhage requiring blood transfusion, cardiac tamponade, massive pulmonary embolism and occasional surgical rescue. In-hospital mortality during PFO closure is 0.3%.22 In the stroke prevention trials, atrial fibrillation was a significant complication occurring up to 3.7% of cases. Historically, complication rates varied considerably depending on choice of closure device. In one study assessing percutaneous closure, adverse event rates were 1.5% for major and 8.9% for minor events.22–26 Complications are more common in older people (>60 years) and in those with preceding ischaemic stroke.22 Other risks include vascular complications, haemorrhage, cardiac tamponade/perforation and pneumothorax or haemothorax.22 Recurrent neurological and peripheral embolic events occur in 0%–3.8% per year, possibly due to incomplete defect closure or thrombus formation on the device.27 Device thrombus formation is infrequent and satisfactorily managed with anticoagulation.28 Very rare complications include device migration, embolisation or erosion.29 Furthermore, there are few longitudinal data indicating how well closure devices function in the long term, when possibly in situ for decades.

Cryptogenic ischaemic stroke and PFO

Cryptogenic ischaemic strokes may be defined as symptomatic cerebral infarcts for which there is no identifiable cause after adequate evaluation to exclude lacunar infarction, large vessel stenosis above 50% or a cardiac cause such as atrial fibrillation.30 It is particularly difficult to judge how long to continue prolonged cardiac monitoring. The detection rate for atrial fibrillation has been reported to be 3.7% at 1 month, 8.9% by 6 months, 12.4% by 12 months, 21.1% by 2 years and 30.0% by 3 years.31 32

Cryptogenic stroke is a common clinical challenge, comprising up to 30% of ischaemic stroke.30 PFO manifests in approximately half of patients with cryptogenic ischaemic stroke,33–35 and it is often uncertain whether it is directly causative or an epiphenomenon. The RoPE (risk of paradoxical embolism) score was proposed to help decision-making by determining the likelihood of PFO contributing to cryptogenic stroke (table 2).35 Deep vein thrombosis (DVT) is more prevalent in people with cryptogenic stroke (20%) than in those whose stroke has a verified cause (4%).36 It is rare to find a thrombus in transit through a PFO with impending paradoxical embolism; most cases have been reported at autopsy.37 In people with an identified DVT, the usual initial treatment is anticoagulation, often for at least 3 months, before considering PFO closure. If long-term anticoagulation is necessary for other reasons (eg, recurrent DVTs), there is little evidence to inform a definitive decision regarding PFO closure and a multidisciplinary team decision (with neurology and cardiology) is advisable. In people at high risk of recurrent DVT, PFO closure could be considered.38

Table 2

Risk of paradoxical embolism (RoPE) score calculator

Initial randomised controlled trials23 25 26 did not verify the superiority of PFO closure over conservative therapy. However, PFO closure with the Amplatzer PFO Occluder device (figure 3) was superior to medical therapy alone in pre-specified ‘per-protocol’ and ‘as-treated’ analyses in the PC25 and RESPECT23 trials with few associated complications. Moreover, longer follow-up in the RESPECT cohort for almost 6 years compared with the original trial yielded 10 more strokes in the medical arm compared with the closure arm with a HR of 0.55,39 and stroke recurrence was significantly reduced by half in the closure arm (p=0.046).39 Furthermore, subgroup analysis suggested that PFO closure helped people with a large right-to-left shunt, an atrial septal aneurysm or those whose planned medical regimen (if they had been assigned to the medical therapy group) included antiplatelets rather than anticoagulants.39

Figure 3

Amplatzer PFO occluder.

Building on these observations, two trials with more selective inclusion criteria producing an ‘enriched’ study population, showed a marked reduction in stroke risk following PFO closure compared with conservative therapy alone (figure 4). The double-disk Amplatzer Septal Occluder devices were predominantly used in the CLOSE40 and REDUCE41 trials.

Figure 4

CLOSE and REDUCE trials inclusion criteria: modified from Saver45 and Kuijpers et al.46 ASA, atrial septal aneurysm; RLS, right-to-left shunt; TIA, transient ischaemic attack; VRF, vascular risk factor.

Using the RESPECT39 subgroup analysis, the CLOSE40 and REDUCE41 trials both stratified patients who had a stroke using particular PFO characteristics, especially large right-to-left shunt or atrial septal aneurysm. This approach differed from other PFO closure trials that recruited everyone with a PFO.40 41 The REDUCE41 trial also recruited patients whose MR brain scan had an ‘embolic’ pattern of infarcts rather than (for example) deep lacunar infarcts. CLOSE40 and REDUCE41 reported statistically significant reductions in stroke risk using the enriched stroke cohort. In the CLOSE trial, no patient in the intention-to-treat PFO closure group had a stroke, while stroke occurred in 14 patients in the antiplatelet-only arm (HR 0.03; 95% CI 0 to 0.26; p<0.001). The composite secondary outcome of stroke/TIA or systemic embolism occurred in significantly fewer patients in the PFO closure group compared with the antiplatelet-only group (3.4% vs 8.9%; HR 0.39; 95% CI 0.16 to 0.82; p=0.01).40 Similarly, in the REDUCE41 trial, recurrent ischaemic stroke occurred in six patients (1.4%) in the PFO closure group compared with 12 (5.4%) in the antiplatelet-only group (HR 0.23; 95% CI 0.09 to 0.62; p=0.002). In the REDUCE trial,41 the patients with ‘TIA’ had MRI evidence of infarction. PFO closure has not been trialled in patients with MRI-negative TIA.

Following these results, eight European societies proposed an official international position paper on management of PFO in the context of cryptogenic stroke.42 Individual data meta-analysis from the trials showed a relative risk reduction of up to 80% for recurrent stroke following PFO closure, with statistical significance reached in predefined patient subgroups such as those with atrial septal aneurysm, a large shunt or all PFO closure patients compared with those on antiplatelet treatment only.42 Meta-analysis of trials with an average follow-up of 3.8 years reported the number needed to treat with PFO closure to prevent one stroke overall was 37 (95% CI 26 to 68) and 21 in patients with high-risk PFO features (95% CI 16 to 61).42

Note that both the CLOSE40 and REDUCE41 trials compared patients who had a stroke taking antiplatelets to a PFO-closure group, in contrast to earlier trials which mixed anticoagulation and antiplatelets in the medical therapy arm. While cohort homogeneity produced positive results, there are still unanswered questions regarding using anticoagulation, including the directly acting oral anticoagulants. CLOSE40 investigators included an anticoagulation subgroup in the medical arm. PFO closure was better (no strokes against two in the anticoagulation arm) but with insufficient statistical power to make a definitive recommendation.

The recent PFO closure trials have limitations regarding their generalisability into clinical practice. First, these trials were open-label, end-point ascertainment studies, which increased the risk of the bias in the results. Second, the rate of venous thromboembolism (DVT and pulmonary embolism) was higher in the PFO closure group than in the medical therapy group in RESPECT.23 The predisposition to venous thromboembolism was particularly apparent in the subgroup of patients with prior unprovoked DVT or undefined thrombophilia.23 This subgroup represented only 4% of patients in the PFO closure group, but they accounted for 25% of venous thromboembolism during the trial. This suggests that patients who had a stroke with suspected thrombophilia or unprovoked DVT should probably be anticoagulated. It was not clear if PFO closure on top of anticoagulation benefited those patients who had a stroke. PFO closure trials also showed an increased rate of atrial fibrillation following closure.25 39 The identified atrial fibrillation appeared to be self-limiting and not recurrent, thereby not requiring long-term anticoagulation or moderating the clinical significance of PFO closure.39

The available evidence is compelling enough to recommend PFO closure in patients who had a stroke who meet the inclusion criteria for the CLOSE40 and REDUCE41 trials (figure 5). Indeed, a recent trial, DEFENSE-PFO (Device Closure vs Medical therapy for Cryptogenic Stroke Patients With High-Risk PFO), targeted this similar cohort and reported a 2-year ischaemic stroke rate of 10.3% in the medical group (5 out of 60 subjects) compared with zero (out of 60) in the closure group (p=0.013).43 Results from the Commissioning through evaluation report from the National Health Service England registry44 showed less benefit associated with PFO closure, in terms of reduced incidence of ischaemic stroke, possibly explained by patient selection or definitions of outcome. In summary, the decision to perform PFO closure in cryptogenic stroke should be made by a multidisciplinary team, which should include a neurologist or stroke physician and a cardiologist.44 Individualised decision-making is crucial before undertaking PFO closure usually to identify young patients who had a stroke (up to age 60 years) with no alternative cause for stroke despite a comprehensive work-up.45 MR brain scan findings suggesting ‘embolic’ infarcts, large right-to-left shunts and/or atrial septal aneurysm (figure 5).46 Alternatively, there is little evidence that PFO closure prevents stroke recurrence in elderly people or those with small right-to-left shunts, absent atrial septal aneurysm or lacunes on imaging.

Figure 5

PFO closure in patients who had a stroke. *Individual treatment decision needed based on bleeding risk. DOAC, direct oral anticoagulant; PFO, patent foramen ovale; VKA, vitamin K antagonist; TOAST, trial of ORG 10172 in acute stroke treatment; TOAST 1–4, (1) large-artery atherosclerosis, (2) cardioembolism, (3) small-vessel occlusion, (4) stroke of other determined aetiology; TOAST 5, stroke of undetermined aetiology (a) two or more causes identified, (b) negative evaluation, (c) incomplete evaluation; OAC, oral anticoagulant; RoPE, risk of paradoxical embolism; TIA, transient ischaemic attack. Adapted from23 40 41.

Migraine and PFO

Prevalence and associations

The possible association of migraine and PFO has been debated in the headache literature since the 1990s47 with many observational and two population-based studies examining the prevalence and possible associations of each.48 49 Migraine without aura is not associated with PFO, but the reported incidence of PFO is about 50% in people with migraine with aura.50–54 However, two large population-based studies have not replicated this finding, finding no significant difference in PFO rates among migraineurs and non-migraineurs.48 49 The mechanisms behind any potential association are speculative.55 Genetic susceptibilities may have a role: for example, the NOTCH3 gene56 is implicated in CADASIL, a disorder in which migraine with aura is often a dominant part of the clinical phenotype and there may be a high prevalence of right-to-left shunting.57 58

Another theoretical mechanistic link between PFO and migraine suggests that substances in the venous circulation including 5-hydroxytryptamine (5HT) or serotonin bypass the filter function of the pulmonary capillary bed through the PFO, reaching the brain in higher arterial concentrations and thereby triggering migraine.59 This theory evolved when three patients with migraine with carcinoid syndrome and congestive heart failure clinically improved following PFO closure.60 However, while 5HT does seem to play a role in migraine and drugs that act on 5HT receptors are used to treat migraine,61 the pathophysiology of migraine is complex and likely involves several neurotransmitter systems mediating painful and non-painful symptoms.62

Migraine, stroke and PFO

Migraine with aura has long been noted to increase the risk of stroke,63 in particular among women taking the combined contraceptive pill and/or smokers. Migraine with aura also features in the clinical phenotype of several genetic disorders that manifest stroke, such as CADASIL and MELAS. There is some evidence that stroke in migraineurs with aura is associated with PFO. The prevalence of a right-to-left shunt in people with a history of migraine with aura and stroke was significantly greater compared with patients with a history of migraine with aura but without stroke (84% vs 38.1%).59 Similarly, patients with cryptogenic stroke with or without migraine had a significantly higher prevalence of PFO.64 However, the CLOSE study showed that PFO closure did not reduce the frequency or severity of any type of migraine attacks in migraineurs with or without aura.40 The theory that subclinical emboli in the context of PFO trigger migraine and thereby provide a unifying mechanism linking the three conditions59 65 66 has contradictory evidence in the literature.67 68

PFO closure in migraine

Improvement in migraine following PFO closure was first published in 200069 followed by observational studies reporting migraine resolution in 10%–83% and improvement in 14%–83%. Complications affected up to 12%, the most common being atrial fibrillation and inguinal haematoma.70 These promising observational studies69 70 prompted three randomised controlled trials of PFO closure to improve migraine. The first prospective, randomised, double-blind, sham-controlled clinical trial, MIST (Migraine Intervention With STARFlex Technology), was conducted in 2008.71 It studied PFO closure using the STARFlex septal repair technology in patients with severe intractable migraine with aura and a moderate-to-large PFO against sham treatment (groin incision). It failed to demonstrate any significant improvement in migraine headache cessation or reduction in headache days between PFO closure or control at 6 months of follow-up. Stroke, atrial fibrillation, cardiac tamponade, retroperitoneal haematoma and pericardial effusion were among some of the reported adverse events in the implant group.

The PRIMA (Percutaneous Closure of Patent Foramen Ovale in Migraine with Aura) trial72 evaluated PFO closure using the double-disk Amplatzer Septal Occluder in intractable migraine. The study groups were randomised to PFO closure or medical management but the study was unblinded; the medical management group did not undergo an invasive procedure. The study did not meet its endpoint at 1 year (a 50% reduction in migraine frequency) and was terminated early owing to slow recruitment. However, there was a significant reduction in migraine with aura days (−2.4 vs −0.6 days) and attacks (−2.0 vs −0.5) at 1 year following PFO closure compared with the control group. There were five adverse events in the 53 PFO closure patients but no long-term complications.

The third trial was called PREMIUM (Prospective, Randomised Investigation to Evaluate Incidence of Headache Reduction in subjects with Migraine and PFO Using the Amplatzer PFO occluder to Medical Management).73 Subjects were randomised to PFO closure or right heart catheterisation, both with adjuvant medical management. There was no significant difference between the two groups in the ‘50% reduction in migraine attack’ outcome measure at 1 year. However, the PFO closure group experienced significantly fewer migraine days per month than the control group (3.4 vs 2.0, p=0.025), and significantly higher complete migraine remission for a year rates (8.5% vs 1%, p=0.01). In the subgroup of subjects who had aura present in >50% of migraine episodes, there was a significant reduction of at least 50% in migraine attacks at 1 year, and 15.4% (6/39) achieved complete remission compared with 2.5% (1/40) of controls.73 Similarly, in the PRIMA trial,72 post hoc analysis focusing on migraine with aura showed reduction in migraine with aura days per month (2.4 vs 0.6) and migraine with aura attacks (2.0 vs 0.5) in the PFO closure group compared with controls. In addition, 40% (16/40) of the PFO closure group were completely free of migraine with aura compared with 10% (4/40) of controls.72 This potential benefit from PFO closure in a minority of patients with migraine72 73 may be analogous to the ‘enriched’ population in the stroke trials but currently should be interpreted with circumspection and be hypothesis generating for future trials.

In summary, while prevalence and observational studies suggest an association between migraine with aura (in particular) and PFO, population-based studies do not confirm this. Interpretation of the studies is complicated by variability in study design, patient cohorts and sample sizes. Indeed, one study even called into question whether there is any relationship of PFO size and migraine.29 Similarly, while treatment studies initially suggested a potential benefit of PFO closure, this was not confirmed in larger randomised, controlled trials. Differing patient populations and diagnoses, implantation devices, elevated placebo responses, PFO sizes and residual shunt prevalence, as well as concomitant use of antiplatelet (and anti-headache) medication following implantation were likely confounding factors in these studies. It will be a challenge to demonstrate causality conclusively between two common and heterogeneous conditions: migraine, where at least 50% of the population carry the biological susceptibility, and PFO, which is present in 25% of the population. Further trials with enrichment of both the migraine cohort to reduce heterogeneity (eg, focusing on those with frequent or disabling aura), and further PFO stratification paralleling the recent stroke PFO closure trials (eg, based on characteristics such as size, large right-to-left shunt or presence of atrial septal aneurysm) may target subpopulations who are more likely to benefit from PFO closure. However, at present there is currently insufficient evidence to support routine closure of PFOs in patients with migraine.

So, what does a neurologist need to know?

PFO is common and affects about 25% of the population. It is normally asymptomatic. PFO may contribute to migraine, but there is currently insufficient evidence to advise its closure in people with migraine outside a clinical trial. PFO closure reduces the risk of recurrent ischaemic stroke in a highly selected population of stroke survivors: those who are 60 years or younger with a cryptogenic stroke syndrome, a right-to-left shunt and no evidence of atrial fibrillation. They benefit from percutaneous PFO closure in addition to antiplatelet therapy, rather than antiplatelet therapy alone. The number needed to treat over 3.8 years to prevent one stroke overall was 37 or 21 in patients with high-risk PFO features. Patients with recurrent or high-risk venous thromboembolism may benefit from anticoagulation without the need for additional PFO closure. Treatment decisions should be made collectively after multidisciplinary team discussion.

Key points

  • Patent foramen ovale (PFO) is the most common type of interatrial shunt, with a prevalence as high as 25%–30%.

  • PFO is associated with cryptogenic stroke, migraine with aura, systemic embolism, decompression sickness and platypnoea–orthodeoxia syndrome.

  • One in twenty people have higher risk PFO appearances, including large right-to-left shunt or PFO associated with atrial septal aneurysm, Eustachian valve or Chiari network.

  • Transoesophageal echocardiogram with saline contrast injection is the current standard for detecting PFO, but even so they are sometimes missed.

  • Patients with cryptogenic embolism require detailed, individualised multidisciplinary decision-making before considering PFO closure for stroke prevention.


We are grateful to Dr Arani Nitkunan, Dr Ajay Bhalla and Dr Declan O’Kane for reading and commenting on the manuscript.

Further reading

  1. 1.
  2. 2.



  • Contributors VA, TK-A and NK wrote the manuscript. RR and ACP reviewed, edited and approved the final version of the manuscript.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Patient consent for publication Not required.

  • Provenance and peer review Commissioned; externally peer reviewed by Tom Hughes, Cardiff, UK, and William Whiteley, Edinburgh, UK.