Patent foramen ovale closure and migraine: science and sensibility

Abstract

Migraine has been associated with patent foramen ovale (PFO), and PFO closure has become the most high-profile nonpharmacologic invasive therapy recommended for the prevention of recurrent migraine attacks, as well as for preventing further attacks in cryptogenic stroke. The results of Migraine Intervention with STARFlex Technology (MIST), a controversial but important recent randomized clinical trial (RCT) of PFO closure for migraine, do not support PFO closure for preventing migraine attacks. All patients with migraine, however, do not have a PFO, and the characteristic periodicity and predictability of migraine cannot be explained on the basis of paradoxical embolism through the PFO. Closure of the PFO or atrial septal defect can aggravate migraine suddenly. PFO increases in size with age, but migraine generally subsides with the passage of years. Serendipity does play a role in some medical discoveries, but in the absence of a logically defensible theoretical basis, chance and statistics can both become misleading. With soft end points, RCTs in migraine patients can generate conflicting and irreconcilable data. RCTs cannot supplant or substitute clinical common sense or justify serendipity. Scientific progress mandates that any serendipitous research must ultimately conform to the principles of the basic sciences surrounding the chance discovery. PFO closure for preventing migraine attacks is an unfortunate, but sobering, chapter in the migraine research saga.

Keywords::

• migraine
• paradoxical embolism
• patent foramen ovale

The foramen ovale is a hole between the right atrium and left atrium that is important during development of the fetus. Normally, a flap of tissue closes up the foramen ovale after birth, but in some individuals the flap does not produce complete closure. These individuals are said to have patent foramen ovale (PFO).

Recently, PFO and its closure has generated tremendous interest in interventional cardiology, as well as in neurology. For two decades, the potential role of PFO and atrial septal aneurysm (ASA) has been investigated for elucidating the genesis of ischemic stroke/cryptogenic stroke in young adults [1,2]. Currently, PFO closure is being actively investigated in randomized clinical trials (RCTs) as a seemingly logical management for cryptogenic stroke. In addition, an exciting but uncertain therapeutic and pathophysiological interface has also evolved between PFO, paradoxical embolism (PE) and migraine, with PFO closure purportedly offering an effective long-term (and possibly permanent) migraine preventive strategy – through abolition of presumed PE – superior to available preventive pharmacologic therapies. This article will reanalyze the theoretical and logical basis for advocating PFO closure as a definitive therapeutic measure for migraine.

Paradoxical embolism: a common feature or a clinical rarity?

The fundamental tenet surrounding PFO closure involves elimination of PE, whether the invasive procedure is used to prevent recurrent migraine attacks or cryptogenic stroke. PE itself is difficult, if not impossible, to document clinically. Barring a few post-mortem or echocardiographic studies showing thrombus lying in the PFO, there is little to conclude with scientific conviction that PE does occur on a frequent or regular basis in those suffering frequent migraine (as frequent as daily) attacks or recurrent cryptogenic stroke. Intriguingly, the case for PE as a clinically significant pathophysiologic mechanism has never been proved beyond reasonable doubt [2]. Nevertheless, the passage of air bubbles across PFO has evolved into an acceptable investigative surrogate for trans-atrial thromboembolism that has widely been adopted as a marker for possible and common real-life occurrence of PE in an individual.

Key limitations of RCTs and meta-analyses are that they do not seem to offer hope for any meaningful progress in this discipline (see later). A series of assumptions (myths) and omissions have carefully and (seemingly) justifiably laid the groundwork for the launch of the Migraine Intervention with STARFlex Technology (MIST) trial [3], as well as other trials for PFO closure in patients with migraine (or in patients with cryptogenic stroke). The results of RCTs of PFO closure on migraine attacks have proven contrary and controversial. While the links in the migraine–PFO–PE chain are far from robust, RCTs of PFO closure have given clinical respectability to this novel, but entirely serendipitous, invasive therapeutic approach. The phenomenology of migraine has been well defined. The validity of the assumptions surrounding the PFO–PE–migraine nexus can best be examined against established clinical characteristics of migraine.

PFO–PE–migraine nexus: the illusion of absolute certitude

Central to all known (such as ASA, small additional atrial septal defect[s], eustachian valve or Chiari network) or unknown static or dynamic aberrations associated with PFO that might confer a long-term risk for cerebrovascular events is occurrence of PE [2]. The single most important limitation to our understanding of PFO and its closure is that PE, in the vast majority of individuals with PFO, is a presumed and unrecorded – not a factual – event. All experimental studies of PFO closure for stroke or migraine prevention, therefore, rest on a major mechanistic assumption.

How robust is the assumed link between PE and PFO [2,4]? Cases demonstrating a thrombus actually traversing the PFO are relatively very few [2]. With 60–70 million Americans (20–25% of the general population) estimated to have a PFO [2], and, given the intense interest of interventional cardiologists studying PFO and pursuing PFO closure in clinical trials, the striking rarity of right atrial thrombus-in-evolution or -in-passage through the PFO – with or without ASA – is inexplicable. Second, the risk of first ischemic stroke in individuals with PFO remains unknown [2,5]. Third, even the presumed higher risk of recurrent stroke in patients with both PFO and ASA is controversial, while the mechanistic basis of the flapping action of ASA directing small clots coming from the inferior vena cava into the defect is purely speculative [2,6,7]. Fourth, no consistent link between stroke and Valsalva maneuver (VM) – that might lead to a transient right-to-left shunt (RLS) or to PE – has been established; also, the PFO subgroup at high risk for PE has not been consistently identified [2,8]. Fifth, the relationship between PFO size and PE appears to be complex. Remarkably, using balloon sizing for measurement, a link between PFO diameter and recurrent cryptogenic stroke [9], or between maximal potential PFO diameter and pre-closure neurologic event [10], could not be found. Furthermore, while PFO increases in size with age [2], the importance of PFO in stroke, as well as in migraine pathophysiology, inexplicably declines with advancing age [4]. Sixth, clinical states associated with hypercoagulability and clinical thromboembolism have not demonstrated any particular association either with PFO or with migraine attacks. Seventh, since PFO is present at birth, why does PE-mediated (cryptogenic) ischemic stroke or migraine not manifest in infants/toddlers or the preschool pediatric population [4,9]? While age is indeed an important risk factor for venous thrombosis with the risk of deep venous thrombosis rising every decade, states of hypercoagulability, such as dehydration, immobilization, surgery and genetic predispositions, are common in the pediatric population. Strokes attributable to atherothrombosis are indeed uncommon in children and adolescents; thromboemboli originating in the arterial circuit and resulting in strokes clearly do not involve the PFO. Eighth, while coughing, straining at stools, lifting or any other form of VM are common activities at all age groups, why does such ‘straining’ predispose only young adults (<55 years) with PFO to cryptogenic strokes? Parturition – perhaps the most common, severe, prolonged and repetitive form of VM – has rarely been reported to cause either thrombotic stroke or migraine attacks. Similarly, severe straining at stools – another common clinical circumstance associated with VM – is uncommonly mentioned in migraine or stroke literature. Ninth, presumably PE requires elevation of right atrial pressure above that of the left atrium, which has not itself been directly studied in persons with PFO [2]. Tenth, passage of air bubbles across the PFO is not tantamount to paradoxical and anti-gravity passage of a platelet–fibrin complex/plug, with the latter probably weighing several hundred times more than a stream of air-bubbles [4,201,202]. An intriguing reconsideration of the physics involved in presumed PE appears inevitable. Gas bubbles cross the PFO during transesophageal echocardiography, despite the fact that the right atrium is typically higher than the left atrium. Two issues appear important here. First, the air bubbles crossing the PFO defy physical laws that otherwise mandate that they rise to the top and float indefinitely in the right atrium. Blood in the right atrium (unlike the appendage) is not a stagnant pool, but a swirling fluid system within which the bubbles are likely distributed evenly against gravitational force. Second, the PFO is not a straightforward ‘hole’ in the heart, but a tunnel-shaped slit valve between the interatrial septum secundum (on the right side) and the interatrial septum primum (on the left side) [11]. Gas or thromboemboli crossing over from the right to the left cardiac circulations across the interatrial septum must traverse the tunnel-shaped PFO from the caudal to the cranial part (close to the roof) of the left atrium. Since the PFO valve mechanism may function properly even with a catheter across it [11], traversing the PFO is a complicated process. Unlike gas bubbles, thromboemboli cannot modify their size or alter their shapes to negotiate the tunnel-shaped PFO. In addition, the dimensions of the PFO tunnel can vary dynamically during the cardiac cycle. In these circumstances, transfer of gas will be markedly facilitated relative to the much heavier and much less malleable thromboemboli.

The fundamental basis for invasive/surgical interventions is clear: invasive intervention should prove superior (not just equivalent) to current medical therapy [2,4]. Presently, no medical or surgical intervention has proved superior to aspirin for prevention of thrombotic stroke [2,8]. That the risk of PFO closure in cryptogenic stroke is no higher than the long-term risk of complications of medical management – as suggested by some interventional cardiologists – appears incorrect [4]. Closure of PFO is not risk free; surgical interventions to manage complications of transcatheter procedures were needed in 8% of patients [12]. Thrombus formation on the device, cardiac perforation, infective endocarditis and other events requiring surgery have been reported [2,12]. Thromboembolism related to the AMPLATZER^® septal-occluder (AGA Medical, MN, USA) has been reported as late as 5 years after implantation [13]. It is also incorrect to assume that no antithrombotic agents might be required after placement of the PFO closure device [12,14], for which reason, as well as for age-related progression of atherosclerosis and widespread use of antiplatelet agents, long-term results of a comparison between PFO closure and medical managements becomes impractical [4]. In striking contrast to stroke prevention, the preventive role of aspirin in migraine is quite uncertain [15]. Currently, when we compare therapies in migraine, the comparison is between two uncertain, or at best, two serendipitous therapies [16]. Nevertheless, the life-threatening complications of PFO closure are unique and significantly raise the risk of carrying out this procedure for prevention of migraine attacks.

Ideally, complete PFO closure might be achieved with material that conforms to both sides of the interatrial septum, and has no risk of erosion, infection, arrhthymia or thrombogenicity. Even non-device PFO closure techniques are being considered, as well as devices with less and/or bioresorbable material. The hope that the next generation of PFO-closure devices/methods will offer better results is justified. What is still contentious is that the link between migraine and PFO remains nebulous (discussed later). The role of serendipity in science – both of the phenomenon itself, as well of as the mind prepared to recognize and register the serendipity – can neither be denied nor grudged. Serendipity, if effective, nevertheless must be brought back into the folds of logic and reason of science to gain general acceptance. Finally, improving technological finesse or sophistication of any invasive procedure cannot – and must not – be regarded per se as an important reason for its use in patients (see below). An almost irresistible tendency to select subjects of our investigations for their technical facility and clarity, rather than for their importance, seems to propel researchers. In the absence of a thorough theoretical background, PFO closure, however technologically sophisticated and well-promoted through high-profile RCTs, will remain highly controversial and completely empirical.

PFO closure in migraine & randomized clinical trials

Migraine is believed to be a cerebral cortical event that occurs in otherwise normal individuals. A large variety of disparate triggers can precipitate migraine. Such triggers vary between patients, and can also vary in the same migraine sufferer during different attacks. Most migraine patients can recount more than one or several triggers responsible for their headaches.

Results from earlier trials of PFO closure suggested a pathogenetic link between migraine and PFO; even the origin of migraine shifted conceptually to the heart. Recently published RCTs, however, do not support the intuitive urge to close PFO as a definitive therapeutic procedure that is superior to existing medical therapies either for stroke [2,6] or for migraine [3]. The speculative link(s) between migraine and PFO and the role of PFO closure as a possible but definitive therapeutic measure, nevertheless, continues to fascinate researchers, both cardiologists and neurologists [17–20]. Given the mathematical basis of the RCT (discussed later), unsurprisingly, PFO closure for managing migraine was found to hold promise in a more recent, but less rigorous, trial [21]. Propagators of hypotheses are never wrong, and, once published, hypotheses are never allowed to die [22]. The PFO–PE–migraine nexus has created an indelible niche in the minds of investigators. PFO closure for migraine is a classical example of the degree of difficulty involved in the falsification of an idea; such difficulty increases several fold when the notion derives strength from mathematics, as with RCT-based evidence (discussed later).

Faced with such contradictory data and obvious public disagreement about priority, protocol and propriety followed by litigation in the MIST trial, the state-of-the-art for preventive PFO closure seems pretty uncertain [4, 23,203]. US FDA approval for PFO closure devices for migraine or for cryptogenic stroke seems far less likely than just a couple of years ago. When we therapists speculatively and uncritically perform procedures (or accept pharmacologic therapies) about which we have but limited basic comprehension, we create myths – practically insoluble dichotomies in the context of both current knowledge and futuristic extrapolation. PFO closure for the management of migraine is precisely such a challenge to the commonly-held and seemingly logical belief that closure of a mechanical intracardiac defect with a mechanical device should yield positive clinical dividend. In other words, if there is indeed an intracardiac defect, why not go ahead and close the aberration? Might not closure of PFO prevent a cardiovascular catastrophy in the future, such as stroke or myocardial infarction due to PE? How could such an intuitive therapeutic strategy, then, possibly go wrong?

The greatest appeal – as well as the greatest peril – of the RCT is that it allows clinicians to carry out scientifically credible research without having to discern crucial clinical phenomena [16,24–29,204]. The RCT has goaded researchers to suspend or even jettison clinical judgment [24,30]. Clinical judgment – or logic or common sense – has come to be regarded not only as inconsequential in medical research, but is both feared and despised as an impediment to the systematic and unhindered orchestration of clinical trials. Consequently, the weakest part of such RCTs – generally sponsored by the pharmaceutical industry – is the research premise detailed in the introduction/background, wherein an insipid attempt is made to somehow cobble up a somewhat defensible theoretical prelude to the trial. To acquire theoretical substance, assumptions are weakly linked to other assumptions, in series or in parallel. Acquisition of scientific data is not an end by itself; data need a well-defined a priori matrix to be incorporated into an intelligent synthesis that transforms scientific efforts into truly meaningful and useful knowledge, bordering on wisdom. Wisdom, in turn, chips away constructively at notions to reveal a glimpse of the truth. The application of RCTs superbly evaluates average therapeutic efficacy, but is not designed or intended to address the basic scientific challenges in clinical taxonomy and data; RCTs tend to draw clinicians away from clinical reality and basic sciences [24,30]. Randomization is not a scientific method, but an invaluable statistical strategy for the mathematical exploitation of uncertainty [24].

“…although a modern jumbo jet comprises manifold per se sophisticated components … they remain a ‘heap of junk’ until assembled and trimmed to the perfect level of coordination that characterizes the flying wonder…”

– Bjorn Folklow

RCTs with poor research premises – such as trials of PFO-closure – create sophisticated ‘heaps of junk’ that cannot be converted or assimilated into an intelligible synthesis (‘the flying wonder’). The sophistication of RCT data comes from an ingenious two-pronged strategy: ‘blinding’ and ‘mathematics’ (discussed later).

The second major appeal of the RCT to migraine (and other primary vascular headache) research is the use of double-blind ratings. Double-blinding is presumed to be bias-free, even if imprecise and individually unspecified [24]. Unlike ‘sharp’ end points, such as pump failure or death in cardiovascular medicine, end points in primary headache research are quite ‘soft’. The RCT format converts ‘soft’ data, such as headache frequency, intensity and duration, into ‘hard’ data through visual analog or nonstandardized ordinal rating scales that are further subjected to mathematical logic. In RCTs, the absence of scientific clinical precision is not regarded as an important problem [24]. RCTs are designed to eliminate various forms of bias, but in the process RCTs embrace ‘chance’. RCTs cloak and color chance in mathematical hues and give respectability to much of the speculation in current clinical medicine – giving an Orwellian solidity to what would otherwise be classified as ‘pure wind’. Moreover, the RCT strategy assumes that the variable/entity under investigation will clinically remain consistent over the duration of the investigation. Given the intrinsically protean nature of migraine itself, it is hardly surprising that RCTs of PFO closure have not yielded consistent results and have aggravated the extant confusion. Even in the same migraine patient, two distinct headache attacks will only very rarely be precisely similar or identical in all clinically important aspects. These ‘insignificant impediments’ to philosophically precommitted, but determined, research efforts (through the RCT) have been eliminated through statistics, a mathematization that has laid the ground for profound biological confusion in PFO closure research. Finally, using numbers to determine clinical significance in medicine is the fundamental flaw in all numerical analyses in RCTs, whether through Bayesian analysis or numbers needed to treat or confidence intervals [24]. Overemphasis on the Bayesian analysis [31] does not fundamentally change the numerical nature of the analysis. Complex mathematical assumptions underlie all such endeavors. However, a complete and rigorous logical basis for all mathematical reasoning simply does not exist. Bertrand Russell and Alfred North Whitehead published ‘Principia Mathematica’ from 1910 to 1913; they failed to prove that mathematical reasoning was flawless. The intrinsic limitation of mathematical logic is an absolute that very rarely surfaces in biological, including medical, research. The RCT is simply a research tool based on and heavily biased towards mathematics; it cannot supplant biological common sense/logic or be allowed to waylay the research process. Capable as it is of creating several versions of truth – each being equally defensible, at least, mathematically – the RCT has, however, itself become the major quantitative constraining paradigm in primary headache research. Due to immense peer pressure, any rarely published challenge to the invincibility of the RCT, as well as to its basis – mathematics and statistics – is snuffed out immediately and rendered ineffective. Viewed unemotionally, the RCT can become an ethical minefield (see below) and is certainly not the only way to gain knowledge [32–34]. Jenner, Semmelweiss and Pasteur wasted no time with RCTs, but came out with ground-breaking medical discoveries. Remote from medicine, in the heartland of mathematics, the discoveries of buoyancy, of gravity or of the earth–sun relationship were the outcomes of careful observations coupled to inspired flashes of brilliance [205]. Data are a terminal facet of human thinking; what is truly original yet inscrutable are the preceeding thought processes.

PFO closure worsens migraine: the value of aberrant observations

No idea, Popper argues, cannot be improved upon [35]. Also, while no amount of evidence ever proves a hypothesis, any hypothesis might be refuted by a single piece of contradictory evidence. For the PFO–PE–migraine nexus, unexpected but sudden, dramatic and sustained worsening of migraine headache attack frequency or severity, or both, is the single piece of evidence that cuts through the Gordian knot of prophylactic/preventive PFO closure. Such apparently inexplicable cases clearly point to a hitherto unknown and unconsidered, but critical, disturbance of cranial physiology that underlies the outliers, and that can help make some sense of the area under the curve.

At 30 days, a significant worsening of migraine was observed in the majority of 57 patients who underwent PFO closure for cryptogenic stroke; patients with severe preprocedure migraine attacks were more likely to experience such clinical deterioration [36]. Furthermore, almost immediately after atrial septal defect (ASD) or PFO closure, dramatic worsening of migraine has been noted; prolonged opening of the PFO during the closure procedure has been implicated [37,38]. PFO closures, however, do not always cause noticeable or dramatic accentuation of migraine headache attacks [39,40], indicating a fundamental, but poorly understood, heterogeneity for migraine patients at the level of the cardiac interatrial septum (discussed later).

Worsening of the migraine state after PFO closure clearly underscores that such migraine attacks absolutely cannot be attributed to repetitive or frequent PE of platelet–fibrin plugs/thrombi/packets of serotonin or development of brain cortical spreading depression by any of these pathophysiologic mechanisms [16]. Generally, PFO closure is accompanied with antiplatelet therapy for 6–12 months or longer; platelet embolism as a cause of migraine attacks becomes highly unlikely in this situation. Elevated or changed levels of atrial natriuretic peptide (ANP) immediately after ASD closure has been suggested as a possible pathogenetic mechanism [37]. ANP is a physiological antagonist of arginine vasopressin (AVP); a possibly important functional interaction between ANP and AVP occurs both at the level of the renal-collecting tubules, as well as at the brain [41]. ANP appears to play a role as a counter-regulatory homone in many disorders characterized by volume expansion, including hypertension [42]. ANP lowers central neuronal hyperexcitability, as seen in experimental modulation of alcohol-withdrawal symptomatology [43]. A primary role for ANP in migraine pathophysiology is as unlikely as that for AVP [44]. Following ASD/PFO closure, elevated ANP levels would tend to fall to normal physiological levels.

Atrial septal defect, unlike PFO, is associated with a significant left-to-right shunt (LRS). We must now face the facts that indicate that correction or elimination of LRS (in ASD closure), and possibly RLS in PFO might both be involved in precipitation/aggravation of migraine attacks, at least in some patients. Since most patients of ASD do not manifest migraine, it might be safe to presume that the presence of LRS at the atrial level (and a relatively lower cardiac output) is ‘protective’, rather than being pathogenetic. Post-ASD closure, this protective influence over the occurrence of migraine attacks appears to be eliminated. Remarkably, correction of ASD increases the (cardiac) stroke volume and cardiac output by eliminating the shunt from the systemic to the pulmonary circulation [44]. ASD closure cannot possibly lead to cerebral ischemia, as suggested by some primary headache researchers [44]. The clock-like precision with which daily migraine with aura attacks can develop after correction of ASD [37] clearly indicates that some physiological system with a prominent circadian variation was affected by the surgical cardiac intervention; also, internal homeostatic adjustments reduced the impact of that variation after approximately 6 months [45]. Finally, the increased stroke volume following closure of ASD is usually slight, indicating that a low-volume, but highly vascular, cranial system was involved in post-procedure aggravation of migraine. The ocular choroidal circulation qualifies for such a cranial vascular bed, and has been implicated in migraine pathogenesis [44]. Such exacerbations of migraine subside with the passage of time [37]. Arterial-side cardiovascular adaptation – from the aortic root to the cranial circulation – to the slightly increased stroke/cardiac output might account for the gradual abatement of aggravated migraine after closure of ASD. Tissue creep at the level of the arterial tree, as well as the sclera, may also possibly be involved [46].

Post-PFO closure, the majority of the patients do not experience worsening of their headache attacks. Unlike ASD, PFO is not associated with any significant cardiac hemodynamic aberration; closure of PFO, therefore, has no cardiovascular hemodynamic impact. Moreover, for an entity that runs its course into decades, follow-up periods for 12–24 months after PFO closure for prevention of migraine are rather short and insufficient to exclude the placebo effect. Surgical or interventional procedures are not free from placebo effect. The aura surrounding an intervention, apart from its specific effect – if any – contributes to the outcome; expectations, experiences (both personal and that of friends and other patients), length and severity of illness, pre-intervention exposure to the procedure (through television or in magazines), and hopes and pride of both the interventionist, as well as the patient, determine the impact and duration of the placebo effect [47]. The flip side of the involvement of patients in the decision to undergo the intervention is poorly understood; such patients have a strong incentive to report benefit, otherwise they would have to admit to themselves, as well as to relatives and to well-wishers, that they made the wrong choice. If one is paying a large sum of money for the intervention, such an incentive is even stronger [47]. The placebo effect is amplified in disorders with a variable course. As it is difficult to conceive of any medical disorder more protean than migraine, the placebo effect of PFO closure should be regarded as significant. Contrary to the common perception, the RCT can amplify the placebo effect (see below) and confound statistical significance.

A discussion of pathophysiologic mechanisms involving PFO or ASD and their closures would be incomplete without a note on the VM. The VM is associated with common daily activities (laughing, coughing, sneezing, vomiting, straining at stools, lifting heavy weights, parturition, crying or shouting) that can alter both the cardiac stroke volume, as well as the systemic blood pressure [48]. At the end of the strain phase of VM, with massive inflow of blood into the chest, the right atrial pressure is raised briefly above the left atrial pressure. In the presence of PFO, VM can induce a transient RLS shunt during this phase. Migraine patients commonly maintain a low systemic blood pressure; consequently, the left atrial pressure is also likely to be relatively lower, and is likely to permit a RLS during this phase. This rush of blood then passes to the left side of the heart to raise the stroke output during the ‘overshoot/hypertensive’ phase of VM. Both these cardiovascular alterations can increase left-ventricular stroke volume and cardiac output transiently, along with a rise in cranial/ocular choroidal blood flow. Sustained improvement in some cases of migraine after PFO closure might involve dampening of changes in VM-related left-ventricular dynamics linked to common daily activities. Clearly, much that is important about PFO, interatrial septum and left heart dynamics and migraine remains unknown.

PFO closure for migraine: the ground reality of a cart-before-the-horse paradigm

Closing PFO while hoping (and praying) for migraine to subside long term is like watching the clouds in London to guess the weather in Tokyo. One of the major (un)scientific functions of the RCT of PFO closure – especially RCTs sponsored by the giant pharmaceutical industry – is to make both patients and therapists forget that the beginnings of such therapy were due to pure chance and not through rigorous application of scientific principles. Even more importantly, at least fiscally, it is the US FDA that needs to be convinced. Trials of PFO closure never mention the pathophysiologic absolute that migraine is a lateralizing cranial disorder. In the context of migraine (or cluster headache), lateralization (side-locked unilateral/bilateral or side-shifting location of headache) indicates without any ambiguity that the primary source or origin of headache lies at the level of the cranium. Simply because we are at a level of understanding that does not allow us to conceive of such a basic head-and-neck source of migraine headache, it does not mean that the latter is nonexistent or unimportant [16,44,46,49]. Till this issue is settled, all therapies for migraine – whether or not substantiated by RCTs – will remain empirical and/or serendipitous [16]. As a slowly unfolding drama in science, the RCT has come to powerfully reflect herd behavior. Acceptability of herd behavior is obvious in all human endeavors. Whether or not the RCT is scientifically appropriate, it is always politically correct. In the garb of the RCT, political correctness has, unfortunately, surreptitiously overtaken scientific propriety in medicine. Procedure is primary; ‘what’ you do has become far less important than ‘how’ you do it.

However, evidence from prospective clinical studies of PFO closure and stroke recurrence in patients with PFO has not matched our expectations [2–4]. Statistical association between PFO and migraine cannot by itself be regarded to indicate a pathogenetic link. The relationship between stroke and PFO will also still be regarded as weakly potential and uncertain [2]. The MIST trial is essentially a negative trial [18–20]. That negativity continues to be projected in a bizarre fashion as a beacon for the future, thereby keeping the gates for more clinical trials (and research grants) fully open. However, in an era where anything is possible and nothing can be disproved to death, irrational scepticism is characterized by an inability to accept the category of the absurd with the exhortation to “keep our minds open – until our brains fall out” [50].

The basic sciences surrounding PFO closure have, nevertheless, fallen far behind the technical skill with actual closure of the ‘defect’. In fact, mechanical PFO closure is a classic example of a device determining the science, rather than the other way around – that is, “this device is trying to make the anatomical defect a part of the disease process (cryptogenic stroke/migraine) and in fact it should be just the reverse” (Dr Norman S Kato, Cardiac Care Medical Group, Encino, CA, USA, member of FDA’s Circulatory System Devices Panel) [203]. Despite the emotionally charged issue for individual patients and for individual physicians faced with cryptogenic stroke or refractory migraine, huge gaps persist in the knowledge base between PFO, PE, stroke and migraine [2,4,16,28,29,44,49,51]. Nevertheless, members of the FDA’s Circulatory System Devices Panel have repeatedly concluded with total conviction that RCTs can alone decide once and for all whether closing a PFO with an endovascular device actually reduces the incidence of migraine [203]. The Panel avoids the more difficult question: is RCT of PFO closure for migraine justified in the first instance? In addition, the Panel does not acknowledge the gaps in comprehension between PFO, PE and stroke or migraine. Worse still, the Panel seeks to fill these gaps through RCTs. The RCT has intrinsic limitations that are not generally understood [24–27,32–34,51–54,205] or, as might be more likely, are understood, but are conveniently or deliberately ignored. Also, the RCT cannot settle the striking dichotomy between the practice and theory of PFO closure. Off-label abuse of the PFO closure device and the recent withdrawal of humanitarian device exemption [203] are other consequences of the cart-before-the-horse paradigm in PFO management. The purpose of clinical trials is not to crown potential therapeutic novelties with success, but to evaluate such therapies critically and dispassionately – a conceptual clarity that, for different reasons, often becomes elusive. RCTs that generate confusion rather than clarity increase, rather than decrease, the number of scientific assumptions surrounding the issue. Keeping the trial sponsor’s agenda in sight reflects a subliminal willingness to please the sponsors that blunts critical thinking and empathetic handling of patients – a harsh indictment, which is, unfortunately, only true (see below). As conclusions of RCTs might only be disproved by more rigorous RCTs, it is rare for a carefully structured medical assumption toasted through the RCT portal to meet its demise and to be labeled and buried as an inconsequential, misleading myth. Magnesium supplementation in acute myocardial infarction was one trend that indeed swept reason or logic aside for around a decade; sharp end points (death or acute pulmonary edema) following magnesium supplementation helped to disprove the hypothesis [54,204]. Magnesium infusion during acute myocardial infarction induces a double inotropic jeopardy [26], a theoretical absolute completely ignored by the proponents of magnesium supplementation. Similarly, magnesium supplementation for migraine – both for preventing and aborting headache attacks – captured the imagination of neurologists for quite some time till the novelty wore off, the confusion piled up and the logic of the exercise simply failed [16,26,49]. The fact that exogenous magnesium does not freely or readily cross the BBB and is highly unlikely to critically influence brain neuronal function has never been considered in-depth [16,49]. Since medical myths never die, but indefinitely and faithfully follow subsequent scientific fashion [22], RCTs have, to a certain degree, distorted medicine. Some RCTs stand as intellectual testaments to our fallibility never to acknowledge error. RCTs involving PFO closure seem to qualify for this dubious distinction. Overall, it is always better not to know and/or not to do than to know or do wrong, or to believe that an untruth is the ultimate, immutable truth.

PFO closure for migraine: ethico–political dimensions

Some cardiac interventionists justify PFO closure in well-educated patients [10], a trend that does not bode well for those who seem to be able to handle medical information for their own benefit. In an ironic, negative and disadvantaging twist, being formally and highly educated might shift the moral and legal responsibility of medical intervention away from the therapist to the patient, placing such patients at a higher risk for experimental therapies. That such thinking can surface at all is a caveat by itself. Written, informed consent is the inviolable component of the contract-like agreement between patients and therapists involved in RCTs. The key issue surrounding informed consent is whether we are empowering patients or persuading them. Doctors commonly still do not provide the information that patients really want [55]. The very thin line between ‘information’ and ‘persuasion’ can get blurred if scientists’ commitment to a belief is absolute, unwavering or overriding. With scientific preconviction, the information offered to potential trial candidates cannot be unbiased; also, little or no room is left for scientists to perceive or maneuver for errors or to explain unexpected results rationally to both in-trial patients, as well as to the scientific community. Finally, it is no secret that “there are institutional review boards across the country (USA) who just looked the other way when a PFO closure was somehow justified” (Dr Richard E Ringel – Johns Hopkins University, Baltimore, MD, USA, FDA Panel member, Circulatory System Devices) [203].

While the RCT is not necessarily unethical, scientific equipoise is maintained only when there are no trade-offs to be made [56]. Cut to the wire, patient equipoise (informed consent) is what justifies a trial [56]. To give meaningful or constructive consent, any potential trial subject must be capable of understanding the information imparted. The information provided by therapists/trialists might be concise or diffuse, helpful or obfuscating, complete or selective by virtue of significant omissions, uncomplicated or ‘loaded’ with subtle and/or embedded suggestions – all the while remaining within the bounds of legality and ‘truth’– truth as we understand or choose/pretend to understand at a given moment. Given the inexhaustible and protean database of migraine, information about migraine to potential recruits to RCTs can neither be complete, nor can ever be assimilated properly by lay people. Besides providing information, trial doctors commonly answer questions and channel interpretation of the ‘selected’ information. Patients’ need to believe in medical science and its practices, coupled with the enormous faith they repose in their doctors, facilitates the process of consent-taking and completes the circle of recruitment into RCTs. Successful recruitment into trials such as PFO closure or scalp application of botulinum toxin for the prevention of migraine attacks, however, does not indicate that appropriate information was imparted or understood. Very likely, the information offered in such trials in migraine patients was highly selective and implies a cosmetic or rite-of-passage exercise. Generally, patients do not understand methodology, but believe that their doctor would/should work in their own (the patient’s) interests [57].

Despite the best efforts to inform patients, they will rarely, if ever, be as well informed about their treatment options as their physicians [58]. More importantly, no amount of education can properly prepare patients to decide on PFO closure when therapists themselves are confused about most of the important issues. To expect migraine patients to make ‘informed’ and ‘appropriate’ decisions about a cardiac intervention for preventing neurological aura/headache attacks while trial scientists themselves do not comprehend well the basis of either migraine (headache or aura or both) or their own recommendation for PFO closure is nothing but an unacceptable form of exploitation of a gullible and vulnerable section of society. In such circumstances, clinical trials become more important than patients and are designed to force therapeutic consensus without resolution of key pathophysiologic facets [4,16,49]. Is it not cowardly to shift the responsibility for the decision to enroll onto the shoulders of frightened, despondent or desperate patients with little or no formal understanding of medical science [55]? Does a formally recorded, signed, freely-exercised choice by lay persons fully absolve clinical investigators from the restraints of their primary obligation to their patients? Legally, yes. Morally, no. Scientifically, such a proposition is as bizarre as would be an inquiry by the pilot of a commercial jet about the requisite flying speed and height from lay in-flight passengers. Law and legality itself is based on a moral bedrock modulated substantially by contemporary perception. As practiced today, ‘written, informed consent’ is the outcome of an uncertain balancing act between the interests of the patients and their physicians, with the fervent hope that physicians will themselves place their patients first – a hope that has long since evaporated [52,55,57–65].

Currently, RCTs are conducted without the information imparted to potential recruits being rigorously examined by a competent but ‘disinterested’ third party. The MIST study was approved by a “multicenter research ethics committee in the United Kingdom” [3]. Does this statement tell us that the fundamental premise(s) of the MIST study was examined in depth by trial-neutral experts in primary vascular headaches research, as well as in PFO research? While trialists can always muster facts to suit their investigative hypotheses and will always be better focused and informed about their project(s) than scientists on ethics committee panels, science is a process rather than facts somehow strung together. Blau’s caustic but pithy note on the RCT reads with compelling clarity and inimitable wit: “In clinical research, the double-blind often lead the blind into blind alleys or to nowhere” [66].

At the cutting edge of research, matters are nebulous to all, including investigators, reviewers and ethicists. Arguments in ethics are seldom conclusive, and most issues can be discussed threadbare with authority in diametrically opposing ways, with the circle of confusion remaining stubbornly intact. The perennial guide in ethical debates is the cardinal principle of mutual and/or reciprocal expectation [26,204]. The single most important ethical consideration for trial scientists is whether in similar circumstances they would themselves be prepared to accept similar medication or device implantation [26,204]. Given the limitations of both the PFO–PE link (see below), as well as the RCT itself, I do not believe that any cardiologist or neurologist with a PFO suffering intractable migraine would currently consent to – or allow her/his loved ones suffering similarly to consent to – implantation of an atrial occluder device. What is particularly disquieting is that the MIST trial Clinical Perspective [3] declares that “a double-blind sham-controlled study was both feasible and ethically justified in this condition”. Also, all patients in the MIST trial, including those in the sham arm, underwent general anesthesia [3]. Again, would any medical doctor with refractory migraine presently consent to undergo general anesthesia to participate in PFO closure trials? Finally, to date, there is no consensus on what truly constitutes refractory migraine. Inclusion of migraine patients into RCTs is, therefore, on a purely arbitrary basis. If this almost precedent-setting interventional practice of the MIST trial for a non-life-threatening illness goes unchallenged, a chilling fear resurfaces: “…ultimately, the individual patient will be counseled, even compelled, to do what is good for ‘society’” [62,65]. Rather than being subjects of experiments approved by a committee (institutional but in-house and uncritical or selectively critical) of strangers and ethicists, patients would be in safer hands if protected by their own doctors, whose only concern should be the patient’s welfare [62]. The patient would be poorly served by any balancing act; the welfare of the patient should always take precedence over the interests of society and science. In the RCT, great care is taken to maintain the pretense of looking after the patients’ interests. However, in the final analyses of RCTs with poor theoretical background, the patients’ welfare is generally wafted away in complex theories and statistical mathematics. In such clinical trials, patients no longer remain the end, but simply become the means to an end.

For interventional trials such as PFO closure for migraine, the following text should currently be regarded as mandatory for obtaining informed consent: “We do not know what causes migraine. We do not know how PFO closure might help prevent migraine aura or headache attacks. We do not definitively know the role of PFO closure for the management of stroke. Once the PFO is closed with a mechanical device, you might have residual/persistent shunting and/or require long-term antiplatelet therapy. Serious cardiac and extracardiac complications can develop following PFO closure. There is no generally accepted definition of refractory migraine.” While undoubtedly decimating or even completely aborting enrollment to PFO closure (and most other primary headache-related) RCTs, in the quest for scientific truth, such searing, personally singeing honesty is indispensable.

PFO closure for migraine: the heart of the matter

The chain of logic between PFO–PE–migraine, as well as the rationale for therapeutic PFO closure to prevent migraine attacks, is far more tenuous than the presumed links between PFO and cryptogenic stroke (see above). Such clinico–pharmacological absolutes (Table 1) will now be discussed. First, lateralization of headache, whether unilateral, bilateral or side-shifting in pattern is a pathognomonic feature of migraine. According to the PFO–PE–migraine hypothesis, passage of platelet–fibrin thrombi or vasoactive neuromediators including 5-hydroxytryptamine across the interatrial defect leads to cerebral ischemia, which in turn precipitates migraine attacks. To maintain that such putative emboli consistently, periodically (regularly or irregularly) and selectively stream into and lodge at the same segment of the brain circulation or parenchyma to effect hundreds of similar attacks of neurological aura (including scintillating scotomata), as well as lateralized headache over several decades in the typical migraine patient, stretches imagination to incredulity and is very unlikely [4]. Generally, emboli are distributed randomly across the systemic, as well as the cerebrovascular circulations; there appears to be no reason why putative paradoxical emboli should behave differently. No systemic influence, including PFO-linked PE, can satisfactorily explain the characteristic lateralization of migraine attacks [16,49]. Second, most individuals with PFO never have symptoms, including migraine attacks [2,4,16]. Furthermore, as reviewed, not all migraine patients have a PFO [16]. Third, PFO is present at birth and, unless closed spontaneously [67], will persist for life. However, migraine attacks do not begin at birth, but typically start at adolescence or early adult life; in addition, the attacks subside spontaneously during pregnancy (commonly) and with advancing age (generally). Pregnancy, however, can be associated with severe worsening of migraine. There is no evidence to suggest that PFO size varies with pregnancy. Furthermore, the size of a PFO increases from a mean of 3.4 mm in the first decade to 5.8 mm in the tenth decade of life; this feature results from the stretching of the valve of fossa ovalis with age- or size-based selection or both [67]. If the PFO–PE–migraine theory is indeed true, an increase in the size of PFO with age should accord consistently with a rising incidence of migraine attacks with aging. Additionally, there are several other diverse but well-known precipitating factors for migraine attacks that clearly have nothing to do with PE or the size of PFO. To maintain the PFO–PE–migraine hypothesis, the propensity to develop platelet–fibrin plugs and subsequent PE should also logically parallel both clinical events and precipitants of migraine. Next, a consistent temporal sequence between PE and the VM – that commonly aggravates migraine headache [68,69] – has not been found [2]. Changes in right atrial pressures following VM that might effect PE have also not been studied appropriately in migraine patients [2]. Finally, extended spontaneous remissions of migraine attacks are a characteristic and common clinical observation [69] that seems quite inconsistent with a key pathogenetic role for a persistent, steadily enlarging physical intra-cardiac ‘defect’, such as PFO. Fourth, embolic events assumed to be associated with PFO are characteristically infrequent and random and never occur predictably. Migraine, particularly menstrual migraine, generally has a characteristic and quite predictable periodicity besides random precipitation at other occasions [69]. Periodic migraine attacks, in the context of PFO, imply regularly timed or predictable PE – an impossible clinico–pathological scenario. Theoretically, PE per se is believed to be a completely unpredictable phenomenon [67]. Fifth, experimental or clinical states of excess 5-HT in humans ameliorate migraine headache or induce sustained remission of migraine attacks [70,71]. Abolition of excess 5-HT following the removal of carcinoid tumor causes migraine attacks to reappear [71]. Anomalous transfer of 5-HT across the PFO, therefore, is quite unlikely to precipitate migraine attacks. Sixth, migraine is twice as common in women as in men [69], but the incidence of PFO or of larger PFOs among migraine patients does not show a similar gender pattern [72]. Seventh, stress is the most common precipitant of migraine attacks [69,73]. While stress-associated catecholamine release can aggregate platelets and induce platelet–fibrin plug formation, migraine headache commonly develops significantly after subsidence of the stress (‘protective’ effect of stress) or the effects of other migraine precipitants [49,206]. The complex concatenation of physiological processes – an adaptive ‘system’ – that characteristically delays onset of migraine attacks in a wide variety of clinical and experimental circumstances has been discussed at length [49]. Eighth, propranolol, the prototypic migraine preventive pharmacologic agent, in the popular dose of 120 mg/day promotes platelet activation and aggregation [74] – the reverse of what might be expected with the PFO–PE–migraine hypothesis. Ninth, antiplatelet agents reduce formation of platelet–fibrin complexes, thereby limiting the putative PE–migraine cycle. However, in a placebo-controlled trial, 1500 mg of aspirin failed to offer significant prophylaxis for migraine [75]. Tenth, estrogen withdrawal appears to play a critical role in precipitating migraine attacks in women, while the level of prevention offered by estrogen supplementation is uncertain [73,76]. Estrogen, like propranolol, promotes aggregation of platelets [74] and thromboembolism is a dreaded side effect of such therapy. Eleventh, a high incidence of RLS has been seen in cluster headache patients (42.5%, 17 out of 40) [77]. PE has, strikingly, never been implicated in the pathogenesis of cluster headache. In addition, PFO closure has never been advocated for cluster headache, even though the patient’s suffering in this variant can be much worse than that in migraine. Twelfth, migraine with aura (MA⁺) patients typically have fewer, milder and shorter headache attacks than migraine without aura (MA^-) patients [78,79]. In trials of PFO closure for migraine, MA⁺ patients have consistently been seen to fare dramatically better than MA^- patients [3,76]. Consequently, a particular pathophysiological association has been assumed to prevail between MA⁺ and PFO [3,80]. However, there is little clinical echocardiographic evidence to support a stronger association of PFO–PE with MA⁺ over MA^-[72]. If the statistical trend seen in some studies of larger PFOs in MA⁺ were indeed clinically relevant, headache attacks in MA⁺ patients would have to be more frequent, more severe and more prolonged. The cohort of MA⁺ patients has less predictability of attacks with apparently greater susceptibility to placebo effects in clinical trials [27]. Even more importantly, approximately 70% of migraine patients experience both MA⁺ and MA^- on different occasions, thereby making any basic, or therapy-related, distinction between these two entities unpractical and highly controversial [78,79]. Thirteenth, occurrence of stroke in migraine patients is distinctly uncommon and does not show rising incidence with age [81,82]. Also, migrainous stroke typically affects a younger cohort expectedly associated with relatively smaller PFO defects. Women are also mostly afflicted with migrainous stroke [81,82], a clinical facet not reflected in the incidence or size of PFO defects in females (see above). Fourteenth, atrial fibrillation, a disorder clearly associated with thromboembolism, is not known to have any comorbidity linked to migraine attacks. Finally, neither the aura nor the headache represents the true beginning of migraine attacks; the onset of migraine-related pathogenetic or primary aberrations lies in the extended (several hours to a few days) prodrome [49]. Embolic events that might be related to PFO, as a rule, manifest suddenly, without prodromes.

Expert commentary

Migraine research faces a tardy but sobering realization: what is known regarding the pathophysiology of the disorder is far less important than what is unknown. The long shadow of serendipity and empiricism looms large over migraine-related therapeutics. RCTs do not reduce or eliminate empiricism; conversely, RCTs based on a series of weakly-linked assumptions invariably cloud issues further. In recent years, cardiologists and neurologists have aggressively pursued PFO closure for prevention of migraine (and cryptogenic stroke) at several headache research centers. Unexpectedly, the results of PFO closure for migraine prevention have proved negative in the MIST trial, the largest RCT to date. The MIST trial has been mired with unprecedented controversies between the trialists and their sponsor; certain procedural issues still remain hazy. Preliminary results (June 2010) of the CLOSURE I trial are also disappointing; PFO closure using the STARFlex^® device (NMT Medical, Inc., MA, USA) is not superior to the best medical therapy for preventing recurrent stroke or transient ischemic attack. News of the disappointing results has come as a blow to cardiologists, many of whom, were convinced that closing a PFO can cure or vastly reduce rates of cryptogenic stroke, transient ischemic attacks and migraine. Indeed, closure devices, including the STARFlex, are approved for stroke prevention via PFO closure in Europe and Canada, but have never been approved for this indication by the FDA.

The function of peer review is to prevent an idea being ‘sold’ or accepted prematurely or with dubious scientific validity. That PFO closure is no longer an idea but an invasive reality already carried out in hundreds of migraine patients on an imaginative and experimental basis is a telling indictment of the fallibility of the peer review system. The presumed association between PFO, PE and migraine stimulated several companies to develop catheter-based devices for closing PFOs. Funds for clinical trials aimed at demonstrating that PFO closure indeed reduces or cures migraines lured investigators genuinely convinced about PE or simply armed with belief in the PFO–PE–migraine nexus, as well as their own unshakeable faith in the invincibility of the RCT.

The RCT is neither infallible nor ethically irreproachable [22–30,33,49–66,83–95,203–205]. Although there appears to be no substitute for it, prepublication evaluation (of the RCT) by peer review is a complicated process that is susceptible to a host of negative influences [95–99]. Does peer review exist to keep egg off author’s faces [98]? Invited editorial comments accompanying published RCTs only rarely present in-depth analyses of the shortcomings of RCTs; the very basis of the RCT itself is hardly ever (never?) discussed. Postpublication critical review – the most cogent form of manuscript review – is published by most medical journals in a grudging fashion under correspondence usually in the back page(s). A prominent neurological periodical uniquely requires membership of the linked Neurological Association to consider any comment for publication. More often than not, meritorious correspondence is casually displaced by editorial fiat under the unscientific guise of ‘space constraint’. For most editors and referees, self-criticism seems not only unpalatable, but abhorrent [95]. For some reasons, these limitations of the RCT and of the peer review/medical journalism system are not perceived as critical by a large swathe of medical scientists, including both the migraine research establishment and the FDA. In science, reason and logic reign supreme, and, putting these values on the back burner is the safest way to ensure that we remain (perpetually?) lost in the sea of knowledge. Scientists are susceptible to both fashion and herd behavior. The bonding – more correctly, surrender – of critical thinking in biological sciences, particularly in medicine, to mathematics extracts (and will continue to extract) a heavy price in terms of wasted resources, therapeutic misadventures and unmitigated suffering. Prolongation of bleeding time caused by aspirin was initially disbelieved by peer reviewers and the manuscript was turned down [100]. In this post-modern age where multiple versions of the (medical) truth abound (most having been created by RCTs), the key question is whose ‘evidence’ is this anyway, and whose interests does it promote [99]? Pluralistic, fragmented webs of power and knowledge created by an accelerating technoculture do not allow any overarching vision to emerge or to sustain [99].

The best function of the history of medicine might be to temper the over-enthusiasm of the therapist. PFO closure for the prevention of migraine is a rather unfortunate chapter in medical history, particularly because this disillusion is being faced at a time when most researchers are overawed by technology and the power of mathematics – a form of pseudo-sophistication of medicine. Whereas clinical data management is becoming increasingly sophisticated [101], the primary research questions are being displaced. Having reached the peak of confusion with prophylactic PFO closure, we really do not know where to turn. The seductive quick-fix to migraine that PFO closure promised has proven illusionary, as well as diversionary. On the positive side, the heavy reliance on the RCT as a battering ram in medicine has itself been badly bruised. When we do recover from the mists of confusion, we would certainly be more likely to pay attention to issues of basic sciences that, indeed, matter [16,49]. A strange arrogance has crept into medicine through the portal of the RCT. To make any scientific progress at all in migraine, the allure, as well as the lure, of the RCT must be understood.

The need to decipher the biological nature or purpose of the PFO remains undimmed. This quest must begin with the understanding that some clinical truths are unprovable through the statistical system of reasoning, and are independent of the technicalities of the design, size and conduct of RCTs [88]. In other words, the devil is not in the details (of the clinical trials). Aggravation of migraine immediately after closure of PFO indicates that the interatrial ‘defect’ might have an adaptive role. The other feature that suggests an adaptive role for PFO is a high incidence (~30%) of a probe PFO at necropsy [102]. Widespread distribution of certain traits appear to be adaptive in nature, for example, the geographic distributions of sickle cell disease, ovalocytosis, thalassemia and glucose-6-phosphate dehydrogenase deficiency closely resemble that of malaria before the introduction of control measures, and suggest that these genetic disorders confer a degree of protection against death from falciparum malaria. High estimated prevalence of PFO also seems to indicate a ‘protective’ role in migraine patients (see above).

Regardless of its precise biological nature, some factor(s) obviously prevent(s) spontaneous closure of PFO in a sizable fraction of the population, including primary headache (migraine or cluster headache) patients. Medical scientists commonly share a misplaced disbelief in the intelligence of the body itself. “Our organs always improvise means of meeting every new situation … The body perceives the remote as well as the near, the future as well as the present” [103]. As Carrel has detailed in inimitable prose, we have little or no knowledge about the (cellular or molecular basis of) preparatory changes seen in pregnancy or following resection of one half of the thyroid or one kidney; also, the complexities of the development of the various structures of the eye evoke wonder. The genetic transmission of migraine is clear, although the precise mechanisms are poorly understood. Programmed predisposition to recurrent migraine attacks would be accompanied with certain endogenous protective or safety measures. While flow or pressure from the lower part of the right atrium opens the PFO, flow or pressure from the left side closes the PFO [104]. VM is a very common feature in humans that raises the pressure in the right atrium. A PFO would provide an anticipatory element of decompression (safety valve-like action) and prevent inordinate rises of pressure in the right side of the heart. Additionally, migraine is associated with higher elastase levels that have been implicated in the occurrence of cervical dissection [105]. Elastase is a metalloendopeptidase degrading specific elastin-type amino acid sequences; such enzymes are involved in matrix degradation [16]. Persistence of PFO likely involves an ‘anomalous’ matrix degradation at the level of the interatrial septum.

Five-year view

In hindsight, PFO closure for the prevention of migraine attacks will be remembered neither for its scientific precision nor its originality, but, for a grand orchestration of RCTs. The rush to close the PFO marks the baseless transition of PE from an assumed pathogenetic thromboembolic phenomenon to an established one. Since Cohnheim’s hypothesis that a clot passing through the PFO must have caused stroke followed by death in a woman in 1877 [106], we are no closer to understanding the precise role of the PFO [107]. Intuitive PFO closure stands delicately at the edge of scientific logic. In addition, the procedure appears to be indefensible ethically. Furthermore, enrollment into ongoing US trials of PFO closure for managing migraine has decreased dramatically. Preventive PFO closure highlights both the inadequacies of informed consent, as well as the subliminal processes by which the gatekeeper (physician) serves the system, rather than the patient [108]. We are overdiagnosing PFO and feeling compelled to close a mechanical defect with a mechanical device. Currently, no definitive pathophysiologic significance can be attributed to the PFO. In effect, we are treating the defect rather than the patient. To avoid earning the label of a ‘gadgeteer’, the medical scientist must learn the limitations of the gadget, as well as of the RCTs. Therapeutic PFO closure is also a disquieting reminder of the overdependence of society, in large, and medical research, in particular, on technology – science and technology (has) become so intertwined that that an effort at distinction is purposeless [109].

In summary, migraine (or cryptogenic stroke)-generating transient RLS/PE across the PFO/ASD has dominated our thinking and has become accepted as a clinical truth. That this ‘phenomenon’ is merely a bold extrapolation from studies with gas/air rarely, if ever, finds mention. The rarity of observed or established real-time transatrial platelet–fibrin complex PE has been lost in theoretical postulates surrounding RCTs of therapeutic PFO closure. PFO closure has evolved into a scientific and ethical minefield. The negative impact of PFO closure on human experimentation will ultimately dampen the unseemly rush to undertake RCTs in the many areas of clinical medicine. Undeniably, we need RCTs to evaluate therapies. What we do not need is the spectre of the RCT becoming the Wild West of medicine, and the stunting of critical thinking by (mathematical) statistics intrinsic to the RCT. The RCT has become a flashpoint between finance and medical science – a tribute to the human capacity to misuse every possible tool.

Table 1. Characteristics of migraine versus limitations of the patent foramen ovale–paradoxical embolism–migraine theory.

Migraine	PFO–PE–migraine theory
Headache attacks are cyclical and predictable in most patients, particularly in menstrual migraine [69,73]	Completely unpredictable phenomenon
Headache attacks usually lateralize over several decades (unilateral or bilateral) [69,73]	No plausible explanation
Characteristic migraine-related prodrome with significantly delayed (several hours or few days) onset of attacks [49,69,73]	No plausible explanation
Incidence of migraine F:M ≥2:1 [69]	No gender trend for PFO [72]
Migraine attacks usually do not begin at birth; characteristic remissions and exacerbations occur during the life of a patient [69,73].Migraine generally subsides with advancing age	Congenital defect present from birth; enlarges with age [67,72].PFO–PE theory cannot explain the frequent onset of migraine attacks in second/third decade, or sustained spontaneous remissions/exacerbations or the ‘protective’ effect of aging
‘Protective’ effect of LRS in ASD [4,29]. Closure of ASD can immediately precipitate or aggravate migraine	No such protection can be envisaged. Effective ASD/PFO closure precludes PE
‘Protective’ effect of stress [49]	No such protection can be envisaged
‘Protective’ effect of sustained systemic excess of 5-HT [70,71]	No such protection can be envisaged
‘Protective’ effect of estrogens [73,76]	No rational explanation. Estrogens activate platelets and promote thromboembolism
Propranolol frequently prevents migraine but aggregates platelets [74]	No rational explanation
Propranolol, as well as atenolol, do not affect brain circulation significantly. Atenolol does not influence brain neuronal function, but is as effective as propranolol in preventing migraine [16]	No rational explanation. No migraine preventive pharmacological agent affects PFO size or closure
Aspirin inhibits platelet aggregation but does not prevent migraine in controlled studies [75]	No rational explanation
Migraine-related stroke is rare and mainly affects younger women [81,82]	No rational explanation for rarity or for age/gender trend for stroke development in migraine patients

5-HT: 5-hydroxytryptamine; ASD: Atrial septal defect; F: Female; LRS: Left-to-right shunt; M: Male; PE: Paradoxical embolism; PFO: Patent foramen ovale.

Key issues

• • Patent foramen ovale (PFO) is estimated to be present in up to a quarter of the American population.

• • PFO is not found uniformly in all migraine patients.

• • PFO size is not related to the incidence or severity of migraine attacks.

• • Uncomplicated PFO is not associated with any known physiologic significance.

• • Most patients with PFO are asymptomatic.

• • Right atrial thrombus in-passage through the PFO is a clinical and post-mortem rarity.

• • The rise of right atrial pressure or paradoxical embolism has not been established in patients with migraine or cryptogenic stroke.

• • PFO-related paradoxical embolism can never be a regular, cyclical or predictable event, while menstrual migraine and migraine without aura attacks are frequently predictable.

• • PFO or atrial septal defect closure can immediately worsen the migrainous diathesis.

• • PFO closure can lead to complications that require surgical intervention.

• • PFO closure randomized clinical trials have yielded conflicting results due to the protean nature of migraine itself and the ‘soft’ end points.

• • PFO-related paradoxical embolism has no known link to the host of clinical circumstances/physiological stimuli that precipitate or remit migraine attacks.

• • The term ‘refractory migraine’ has not yet been defined.

• • PFO closure for the prevention of migraine raises the issue of misuse of the randomized clinical trial.

Financial & competing interests disclosure

The author has no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

No writing assistance was utilized in the production of this manuscript.