Potential source of cerebral embolism in migraine with aura

Methods.

A total of 113 patients, consecutively referred by the Headache Outpatient Clinic (Dr. Dalla Volta) for migraine with aura (MA+, mean age 34 ± 12 years, M/F ratio 24/89), were compared with 53 patients with migraine without aura (MA−, mean age 36 ± 13 years, M/F ratio 9/44) and with 25 age-matched nonmigraine subjects (mean age 31 ± 10 years, M/F ratio 7/18). MA− patients were consecutively referred in the 2 months after the completion of the study on MA+ patients. Controls were selected among nonmigraine subjects of the hospital staff standardized for age and gender (table 1). The diagnosis of migraine with or without aura was based on the criteria adopted by the International Headache Society.8

Both patients and controls were subjected to a semi-structured interview including detailed information about current cigarette smoking (CS), use of oral contraceptives (OCs), and personal or family history of deep vein thrombosis, ankle swelling on standing, and varicose veins (henceforth referred to as venous troubles [VT]). In migraine patients the average monthly frequency of headache attacks in the last year was recorded as well as the aura-related symptoms (visual, sensory, or motor). During history-taking, particular care was paid to recording symptoms of more than 24 hours’ duration that might have been attributable to a cerebrovascular insult. Both patients and controls underwent a full neurologic examination aimed at detecting possible residual focal cerebral signs.

PFO was assessed with transcranial Doppler sonography with IV injection of agitated saline. The technique is described in detail elsewhere9; briefly, it consists of the injection of a contrast-enhancing agent into a peripheral vein (20 mL of previously shaken saline so as to generate air microbubbles) while recording the flow velocity of the right middle cerebral artery with a hand-held transcranial Doppler probe placed on the temporal bone. In the case of interatrial right-to-left shunt, either spontaneous or produced by a Valsalva strain, the air microbubbles contained in the injected bolus escape the pulmonary filter and can be easily detected in the brain vessels as transient spikes on the velocity spectral curve within 7 seconds of the injection. Later signals may be due to either intrapulmonary shunts or surviving microbubbles after the pulmonary passage. In the current study only patients fulfilling the temporal criteria for cardiac right-to-left shunt were considered positive. The method has been validated and provides 90% sensitivity and 100% specificity, with an overall diagnostic accuracy of 95%, in comparison with transesophageal echocardiography.9

The assessment of PFO was performed both during normal breathing and after a Valsalva strain. In positive cases a semi-quantitative grading was adopted: permanent PFO if bubbles could be detected during normal breathing, functional grade 1 PFO if single spikes appeared only after Valsalva strain, and functional grade 2 PFO if confluent spikes were recorded after Valsalva strain. This grading is roughly related to PFO size, reflecting a progressively larger amount of shunted blood from functional grade 1 to permanent PFO (P. Berlit, personal communication, 1996). To avoid any possible bias toward a positive finding, the Doppler operator was blinded to the presenting diagnosis.

No systematic brain imaging was performed. In 24 MA+ patients and 8 MA− patients with PFO a standard MRI with T1, T2, and proton density sequences was carried out.

Univariate statistical analyses were performed with chi-square and odds ratio (OR) calculations. Multivariate analyses were further conducted to evaluate the effect of possible confounding variables (SPSS computer program, release 5.0; SPSS, Chicago, IL).

Results.

MA+ patients had reported an average monthly frequency of four attacks in the last 12 months (range 0 to 30). Most patients had a mixture of attacks with and without aura, and the ratio of migraine with aura to total migraine was extremely variable both between and within patients. Overall, migraine with aura attacks accounted for 70% of the total. A total of 109 patients (96%) had typical visual auras, 51 (45%) had sensory symptoms, and 23 (20%) motor troubles. Fifty patients experienced a mixture of visual, sensory, and motor troubles in the prodromal phase of migraine attacks. No patient reported prolonged auras exceeding 1 hour. None had a history or actual physical signs of focal cerebrovascular disease. Twenty-two (19%) reported current CS, 18 (16%) use of OCs, and 24 (21%) some form of VT. A right-to-left shunt was found in 54 patients (48%)—in 30 (56% of PFO+ patients) PFO was functional grade 1 and in 24 (44% of PFO+ patients) functional grade 2. In three patients (15% of PFO+ patients) PFO was permanent. Twenty-four of 54 patients with right-to-left shunt underwent a standard MRI, which was normal in all patients.

In MA− patients the average frequency of attacks was eight per month in the last year (range 1 to 25). The prevalence of CS was 23% (12/53), OC use 23% (12/53), and VTs 19% (10/53). Overall, 12/53 patients (23%) had PFO. The type distribution was as follows: functional grade 1 in eight (67% of PFO+ patients) and functional grade 2 in four (34% of PFO+ patients). One patient had permanent PFO. MRI was performed in eight patients, in all cases with negative results.

In control subjects the prevalence of CS was 40% (10/25), OC use 24% (6/25), and VTs 24% (6/25). Overall, PFO was detected in 5 of 25 individuals (20%)—functional grade 1 was present in 2 (40% of PFO+ patients) and functional grade 2 in 3 (60% of PFO+ patients). None had permanent PFO.

There was a statistically significant difference in the prevalence of PFO between MA+ patients (48%) and MA− patients (23%) (OR = 3.13, 95% confidence interval [CI] = 1.41 to 7.04, χ² = 9.52, p = 0.002) and between MA+ patients (48%) and control subjects (20%) (OR = 3.66, 95% CI = 1.21 to 13.25, χ² = 6.46, p = 0.01), whereas between MA− patients (23%) and control subjects (20%) there was no difference (OR = 1.17, 95% CI = 0.32 to 4.45, χ² = 0.07). Because of the unbalanced number of patients, further analyses were conducted comparing the first 50 MA+ patients with the last 63 MA+ patients and each of the former cohorts with MA− patients and controls. The results, shown in table 2, are in agreement with the general picture.

There was no clear trend for a different proportion of OC use or VTs between the three groups. The 40% prevalence of CS in control subjects was not statistically different from the 23% prevalence in MA− patients or the 19% prevalence in MA+ patients.

On linear correlation analysis, sex, VTs, and CS were not significant. Therefore, we further evaluated the association between migraine with aura and PFO in a logistic regression model in which migraine with aura was entered as a dependent variable and PFO was included as covariate. The results confirmed the highly significant statistical association between migraine with aura and PFO (relative risk 3.32; 95% CI 1.74 to 6.34, p = 0.0033).

The proportion of conventional stroke risk factors (hypertension, diabetes, cardiac disease) was absolutely negligible. No patient reported premature strokes among first-degree relatives.

Discussion.

Migraine with aura has long been suspected as a risk factor for stroke in young adults based on circumstantial evidence: focal alterations in cerebral blood flow sometimes approaching the ischemic threshold have been repeatedly detected during the aura phase,10 stroke may occur during the course of a typical migraine attack,11 headache with features similar to migraine may occur in 25% of strokes,12 and platelet hyper-aggregability has been documented in migraine patients.13

In epidemiologic surveys, migraine has been involved as an independent risk factor for stroke in the Physicians Health Study,5 and migraine with aura represented the only risk factor in women below the age of 35 in a recent report from the Italian National Research Council Study Group on Stroke in the Young.6 So far, however, the mechanism of the association has remained elusive. Platelet hyper-aggregability, endothelial abnormalities, and reduced cerebral blood flow in the aura phase have all been in turn speculatively advocated as possible determinants.5,6

Our findings suggest a further mechanism by which the stroke risk might be enhanced in migraine with aura but not migraine without aura. The 48% prevalence of PFO found in MA+ patients is, in fact, of the same order of magnitude as that reported in cryptogenic stroke patients,1-3,14 in whom PFO is incriminated as the causative agent. By contrast, the prevalence of PFO in MA− patients is 23% in our series, not different from controls and close to the figure reported in autopsy series and in echocardiographic studies in normal individuals.15,16

Our findings likely do not result from an overestimation bias. Patients were consecutively referred from the Headache Outpatient Clinic with an established diagnosis of primary headache. A careful history-taking and clinical examination were performed to rule out cerebrovascular accidents, and the Doppler examiner was unaware of the clinical diagnosis.

A referral bias is also unlikely because migraine patients were attending the Headache Outpatient Clinic on a regular basis, the proportion of visual to other types of aura was the same as that reported in the literature,11 the clinical characteristics of the aura were in all respects typical, and, finally, in 32 PFO+ patients (24 MA+ and 8 MA−) MRI disclosed no abnormality. In addition, conventional stroke risk factors such as CS and use of OCs were equally represented in MA+ and MA− patients and controls. Therefore, our results can hardly be attributed to misclassification of cerebrovascular patients in the MA+ group.

Rather, migraine with aura, unlike migraine without aura, doubles the chance of harboring a PFO with the same frequency encountered in patients with cryptogenic stroke. The association between PFO and migraine with aura becomes even more impressive if spontaneous right-to-left shunt is taken into account, which was found in 8/113 MA+ patients (7%), 1/53 MA− patients (1.8%), and none of the controls. Migraine with aura may thus predispose to stroke by increasing the risk of paradoxical cerebral embolism.

PFO is frequently associated with atrial septal aneurysm (ASA), another potential source of cerebral emboli,15 and the association of both abnormalities further increases the risk of recurrent strokes.17 The transcranial Doppler technique prevents the assessment of ASA prevalence, which may represent a drawback of the present study. However, this is a minor pitfall, in our opinion because there is no reason to believe that the proportion of PFO+ patients with concomitant ASA is different in MA+ patients compared with MA− patients and normal subjects.

It is tempting to speculate whether paradoxical cerebral embolism may also be involved in the aura of migraine attacks. A thorough discussion on the pathogenesis of aura is beyond the scope of this report. However, small amounts of blood may be shunted to the left atrium even during normal breathing,18 platelet-sustained activation has been demonstrated in MA+ patients,13 and focal areas of hypoperfusion close to the ischemic threshold may be found in occipital regions during the aura phase.10,19 Also, paradoxical emboli, although for hitherto unexplained reasons, may have a particular propensity for the posterior circulation.20 Therefore, in some patients with activated coagulation, such as women taking OCs, and in some circumstances, such as after physical exertion, paradoxical microembolization of platelet clots to the terminal branches of posterior circulation may occur. This might trigger a focal transient ischemia, which in turn might lead to the generation of the aura-associated spreading oligoemia.10,19 A somewhat similar hypothesis has been proposed for TGA attacks,7 and recent anecdotal evidence suggests that migraine attacks may be reduced by warfarin.21

Whether anticoagulants have any place in the prophylaxis of migraine attacks or migraine-related strokes will require further studies. However, the possibility of individual assessment of PFO may have immediate practical implications. A systematic search for PFO in MA+ patients may be warranted to identify individuals with permanent or functional grade 2 PFO who are at increased risk of paradoxical cerebral embolism.22 In these patients a detailed investigation of coagulation abnormalities known to increase the risk of deep vein thrombosis, such as protein S and C deficiency or activated protein C resistance,23 should be performed and OCs precautionally discouraged in female patients.