Association between Deep White Matter Hyperintensities and Right-to-left Shunt in Migraine (CAMBRAIN): A Cross-sectional Multicenter Study

Deep white matter hyperintensities (DWMHs), often identied by hyperintense lesions on T2-weighted magnetic resonance imaging (MRI), were discovered to have a higher prevalence in migraine patients. A right-to-left shunt (RLS), which is also prevalent in migraineurs, could potentially contribute to the formation of DWMHs by induction of controversial embolism and endothelial dysfunction. In this cross-sectional study, we aim to evaluate the association between RLS and the prevalence of DWMHs in patients with migraine. and c-TCD examinations. The exclusion criteria of our study were 1) a neurological disorder involving inpatient history; 2) severe artery stenosis detected by TCD or carotid artery ultrasonography; and 3) conditions that result in an inability to undergo a standard c-TCD examination, including an insucient temporal window, inadequate cubital venous access, and inability to perform the Valsalva maneuver (VM) because of a severe heart or lung disease.


Introduction
Migraine is a common headache disorder and causes a great burden of disability (1), yet the pathology of the disease still remains unclear. Recent literature reveals that migraine is a neurological disorder, and multiple brain areas including cortex, thalamus, hypothalamus, and brain stem are involved in different phases of a migraine attack (2). Also, structural changes including deep white matter hyperintensities (DWMHs) and subclinical brain infarctions were found associated with migraine (3,4). A higher prevalence of subclinical focal DWMHs has been reported in migraine patients (5,6). Moreover, this heavier burden of DWMHs has been associated with female and elderly patients (3,7). Nevertheless, the higher prevalence of DWMHs in the migraine population cannot be fully explained by traditional vascular risk factors. It has been proposed that microvascular ischemia are caused by recurrent hemodynamic changes (8).
A right-to-left shunt (RLS) is a situation where an abnormal pathway between pulmonary and systemic circulations exists. Contrast-enhanced transcranial Doppler (c-TCD), with its high sensitivity and speci city for detecting both intracardiac and extracardiac shunts, is widely used in the detection of RLS. Patent foramen ovale, present in 20-25% of the general population and 40-60% of migraine patients (9)(10)(11)(12), is usually accountable for intracardiac RLS. A higher reported prevalence of RLS in patients with migraine has led to the hypotheses that RLS could trigger migraine attacks in some patients and also contribute to the formation of DWMHs, by the induction of controversial embolism and endothelial dysfunction (7,9,10). Observations on the association between RLS and WMHs have been made, with highly varying conclusions. In this study, we evaluated the association between the presence, grades, and subtypes of RLS and the prevalence of DWMHs in patients with migraine.

Study design
This is a cross-sectional, observational study, which was performed in 14 hospitals from July 2018 to March 2020 (see Additional le 1).

Participants
In this study, we consecutively enrolled patients aged between 18 and 55 years, and the recruitment procedure is shown in a owchart (Fig. 1). The inclusion criteria for our study were 1) diagnosis of migraine by a neurologist from a headache clinic according to the International Classi cation of Headache Disorders III (13); 2) ability to provide informed consent to participate in the study; and 3) completion of the required tests, including brain MRI, carotid artery ultrasonography, transcranial Doppler (TCD), and c-TCD examinations. The exclusion criteria of our study were 1) a neurological disorder involving inpatient history; 2) severe artery stenosis detected by TCD or carotid artery ultrasonography; and 3) conditions that result in an inability to undergo a standard c-TCD examination, including an insu cient temporal window, inadequate cubital venous access, and inability to perform the Valsalva maneuver (VM) because of a severe heart or lung disease.

Demographics, migraine features, and other clinical information measurement
A standard questionnaire on demographic information and migraine features was conducted under guidance of a neurologist. Clinical histories including hypertension, hyperlipidemia, heart disease, diabetes, and stroke or transient ischemic attack (TIA) were collected based on self-reported history or doctor diagnoses. The Hamilton Depression and Anxiety Rating scales and the Montreal Cognitive Assessment scale were used to assess mood and anxiety symptoms and cognitive status, respectively, during the interview.

TCD protocol
In order to achieve a standard procedure in TCD examination, experienced TCD investigators from participating centers underwent a training program for 3 months at the First Hospital of Jilin University before the patient enrollment started. The investigators were blind to the migraine status of patients. TCD examinations followed our standard procedure using a TCD detector (EMS-9A; Delica, China) at all participating centers, and the same machine was used to conduct c-TCD to detect RLS. The procedure of c-TCD has been detailed in our previous work (10). Brie y, a hand-held 2 MHz probe was used to monitor the left middle cerebral artery with the patient in a supine position. A mixture of 9 mL of normal saline solution, 1 mL of air, and a drop of the patient's own blood was used as our medium. The procedure was performed rst with normal breathing, and then two more times with a 10 s VM, with a 5-min interval between each procedure. RLS was diagnosed when at least one microbubble (MB) was detected in the TCD spectrum. Permanent RLS was characterized by an RLS detected at rest, whereas a provoked RLS was characterized by RLS only detected after a VM. Several category systems for c-TCD grading have been used (11,14). Based on the standards reported by Wessler et al., Spencer et al., and our previous research (10,15,16), we assessed RLS with a ve-level category system: grade 0 = negative; grade I = 1 ≤ MBs ≤ 10; grade II = 10 < MBs ≤ 25; grade III > 25 MBs and no curtain; and grade IV = curtain (when a single bubble cannot be identi ed). In this study, the c-TCD results are presented as RLS − (grade 0), smallto-moderate shunts (grade I and grade II), and large shunts (grade III to grade IV) (Fig. 2).

MRI protocol and DWMHs rating
Whole-brain MRI images (thickness = 5 mm) on 1.5T scanners were obtained; on the axial plane, sequences included were pulse sequences of diffusion weighted sequences (DWI and ADC), FLAIR T2weighted sequences, and spin echo T1-weighted sequences. On both axial and sagittal planes, a spin echo T2-weighted sequence was obtained. Two neuro-radiologists blinded to any clinical information evaluated the digital images for prevalence and grades of DWMHs. DWMHs were de ned as hyperintense lesions located in the supratentorial deep white matter, present on both standard and FLAIR T2-weighted sequences, with normal or slightly low signal intensities on T1-weighted sequences (Fig. 3). Periventricular white matter hyperintensities, de ned as WMH within 10 mm of the ventricular system, were excluded from this study. Virchow-Robin spaces were excluded as lesions with T2 hyperintensity that were not hyperintense on FLAIR T2 sequence. DWMHs were rated in each cerebral lobe following the Scheltens scale, and were scored as follows: 0) no lesions; 1) < 3 mm and n < 6; 2) < 3 mm and n ≥ 6; 3) 4-10 mm and n < 6; 4) 4-10 mm and n ≥ 6; 5) ≥ 11 mm and n ≥ 1; and 6) con uent lesions (17). Lesions found at basal ganglia and infratentorial spaces were not included in this study. Final DWMH load scoring values from two raters were averaged.

Statistical analysis
We used Spearman's χ 2 , 2-tailed Fisher's exact, unpaired t, and Mann-Whitney U tests to evaluate differences in the distributions and means of characteristics between study groups when appropriate. Using logistic regression analysis, we examined the odds ratio (OR) and 95% con dence intervals (CI) for DWMH presence based on RLS grading (RLS − /small-to-moderate shunt/large shunt), sex, age, BMI, disease duration, attack frequency, migraine subtypes (migraine with aura [MwA] and migraine without aura [MwoA]), family history, hypertension, hyperlipidemia, and cognitive impairment. All analyses were conducted with IBM SPSS 24.0 (Chicago, IL, USA), and signi cance level was set at p < 0.05.

Clinical characters of participants
In all, 237 participants were included in the statistical analysis, with a mean age of 39.3 ± 11.7 years, of which 78.1% were women. Among all 237 participants, 107 (45.1%) were diagnosed for the rst time with migraine, 32 (13.5%) were diagnosed with MwA, 22 (9.3%) with chronic migraine, and 7 (3.0%) with medication-overuse migraine.
Eight patients had missing data, 1 on the use of oral contraceptives, 3 on caffeine intake, and 4 on exercise habits, all of whom were diagnosed with MwoA. These cases were removed in the analyses of speci c variables.

Association between RLS and DWMH
Overall, 115 (48.5%) patients had RLS, 52 (45.2%) of whom had large shunts. RLS was detected in 48.3% of MwoA patients and 50% of MwA patients, with no statistically signi cant differences between migraine subtypes.
DWMHs were identi ed in 138 patients (58.2%). Between the RLS + and RLS − groups, the prevalence rate of DWMHs did not differ signi cantly (RLS + 57.4% vs RLS − 59.0%, p = 0.74). We also found no statistically signi cant association between DWMHs and RLS subtypes or grades (Table 1). Clinical characteristics in DWMH + and DWMH − patients First, we analyzed the differences in all demographic, migraine-related, and other clinical characteristics between the DWMH + and DWMH − groups. Signi cant differences in DWMH prevalence with regard to age, BMI, disease duration, and family history were found in single variable analyses ( Table 2). Diabetes and coronary heart disease could not be analyzed because of the small sample size (one patient in the DWMH − group had coronary heart disease; 7 patients had diabetes: 4 in the DWMH + group and 3 in the DWMH − group). None of the participants reported a history of stroke or TIA. After controlling for other variables, age remained as an independent risk factor for the presence of DWMHs. There was also a positive association between a diagnosis of MwA and the presence of DWMHs. No statistically signi cant association between RLS grading and DWMHs was found, and family history showed a negative association with DWMHs, statistically (Table 3).

Discussion
Through this observational study, we aimed to clarify if there is an association between DWMHs and RLS in migraine patients, especially in those with large or permanent shunts. The prevalence rate of DWMHs in our study was 58.2%, which is slightly higher than that reported in previous studies (7), despite the younger age (39.3 ± 11.7 years) of our participants. The overall prevalence of RLS in our study (48.5%) was similar to that of our previous work (46.1%) (10), but lower compared to other studies (9,18). Also, a higher prevalence rate of RLS in patients diagnosed with MwA, as reported in the literature (10,19), was not found in our study. This was probably because of the limited sample size of patients with MwA.
Based on our ndings, DWMHs in migraine patients have no association with RLS, regardless of the subtypes or grades. Neither patients with permanent RLS nor those with large shunts have an increased risk of presenting a higher prevalence of DWMHs. After controlling for other variables, age and aura seem to be independent risk factors for an increased prevalence of DWMHs.
WMHs have always been considered as a phenomenon of small vessel disease (20), and the most consistent risk factors for WMHs are age and hypertension (21). Current literature suggests that migraine is a risk factor for WMHs (5,22), raising questions regarding whether migraine patients are more vulnerable to the development of WMHs and how these may affect them.
The association between RLS and WMHs in migraine patients has long been debated. The majority of published papers have reported no association between RLS and WMHs (7,9,23); however, some research has shown a positive association (24,25). There is evidence that migraine patients have a higher prevalence of RLS (9,10), and it is believed that paradoxical embolism caused by RLS is related to cryptogenic stroke and silent brain infarcts in migraine (26). Therefore, some authors suggest that RLS is involved in WMHs through a similar mechanism (27).
However, as no widely accepted theory on the pathology of WMHs has been proposed, it is possible that periventricular WMHs (PVWMHs) and DWMHs have differences in pathology, according to neuroimaging studies (28). High burdens of PVWMHs are more often associated with clinically isolated syndromes and cognitive impairment diseases (28,29), whereas DWMHs have a higher prevalence among migraine patients. In the CAMERA series studies, only DWMHs in women were found to have a different prevalence between migraineurs and controls (3,5). Furthermore, DWMH lesions that seem to progress in migraine patients as pre-existing microstructural white matter changes have been observed years before the emergence of visible focal WMHs on conventional MRI (30). Furthermore, a study on migraine and tension-type headache patients suggested that small DWMHs are associated with RLS in young migraineurs (25). Therefore, we may assume that if paradoxical embolisms caused by RLS play a part in WMHs in migraine patients, it is most likely associated with DWMHs, rather than PVWMHs. In addition, similar to the case in cryptogenic stroke, patients with permanent RLS and large shunts should be at a higher risk (31,32).
In this study, c-TCD instead of heart sonography is used to detect RLS due to its higher sensitivity (33). Although c-TCD cannot distinguish the different levels of RLS, both intracardiac and extracardiac shunts can cause paradoxical embolism. Moreover, compared to heart sonography, c-TCD has a better detection rate on extracardiac shunts (34). A different system for rating RLS grades was used in this study instead of the one that most former studies on this topic have used, mainly dividing RLS into three grades: negative, less than 10 MBs, and more than 10 MBs. This decision was based on our former research, which has shown that the prevalence rate of large shunts (de ned as > 25 MBs or curtain on TCD spectrum) in migraineurs is signi cantly higher than that in healthy controls (10). Thus, when it comes to a possible paradoxical embolism caused by RLS, we believe that large shunts under our category system should more likely be deemed a probable cause. However, results from our study add further evidence to the lack of association between RLS and DWMHs, even in patients with large or permanent shunts. Therefore, it is not likely that RLS contributes to the high prevalence of DWMHs in the migraine population through paradoxical embolism.
Whether this higher burden of DWMHs is a consequence of normal aging or migraine as a risk factor remains unclear. There is consistent evidence of age-related accumulation of both PVWMH and DWMH burden (21,36), suggesting that WMHs might be part of the normal aging pattern in brain tissue. However, an imaging pattern study on brain aging in the general population has shown that WMHs contribute to beyond-normal brain aging imaging ndings (36). DWMH progressions in migraine patients have been reported in some longitudinal studies, suggesting that female migraineurs or patients with recurrent headache attacks may develop heavier DWMH burdens (3,4). Therefore, it is possible that this association between age and DWMHs in migraine patients is an outcome of both aging and the impacts of migraine-related factors over the years.
The difference between sexes, as reported in several studies (6,30), was not detected in our study. A possible reason is the different proportion of women on oral contraceptives in different study populations. In our study, only 8 out of the 185 women reported a history of oral contraceptive use, compared with 25% of women who had a history of over 15 years of oral contraceptive use in the CAMERA study (30). Considering that the use of oral contraceptives can be a risk factor for vascular incidents, the low rate of oral contraceptive usage in our study group could account for this difference.
We did not nd any association between DWMH prevalence and cognitive impairment, which may be explained by the mild load of DWMHs in migraine, in line with a previous study (3). Furthermore, a recent study on neuroimaging suggests that instead of DWMHs, PVWMHs are associated with cognitive impairment (28).

Migraine features
With regard to migraine features, we found a positive association between DWMHs and aura, which is consistent with a previous nding (37). Cortical hyper-excitability is thought to lower the threshold for cortical spreading depression, which could be responsible for aura in migraine. This cortical spreading depression might be caused by ischemic events, which possibly links the association of aura with DWMHs (38).
Disease duration and attack frequency, as reported in the literature (9,24), were found to have no association with DWMHs in our study. Contrary to what we expected, we found a negative association with family history. However, it must be mentioned that patients with a self-reported positive family history in our study were associated with younger age and longer disease duration. Therefore, it would be too early to draw a conclusion based on our current data.

Strengths and limitations
This is a cross-sectional multicenter study with 14 participating hospitals, which to some extent makes our data representative of Chinese migraine patients from headache clinics. To the best of our knowledge, few reports on this issue have focused on patients with large shunts. Therefore, our study adds new evidence clarifying that RLS was found to have no association with DWMHs in migraine even in patients with large shunts (> 25 MBs or curtain on spectrum).
Our study presents a series of limitations. First, this study is lacking in healthy controls due to the di culty in obtaining MRI images from healthy populations. Since we focused on the association between RLS and DWMH in migraine patients, our conclusion still has some reference value in clinical practice. Second, because of our limited sample size, only 13.5% of the patients were diagnosed with MwA. In addition, as 45% of our participants were diagnosed with migraine for the rst time, participants in our study have suffered from severe migraine attacks less often. Therefore, conclusions from our study may not be applicable to severe migraine patients. For future research, work on DWMH progression in different subtypes of migraine and in patients and controls who have undergone patent foramen ovale closure surgery should be conducted.

Consent for publication
This manuscript does not contain any data that could reveal a participating individual's personal information.
Availability of data and materials The datasets used and analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.

Funding
This work was supported by the Natural Science Foundation of China (NSFC, Grant No. 81971620).
Authors' contributions ZYS analyzed and interpreted the patient data, and was a major contributor in writing the manuscript. XY and WSB contributed to the original idea of this work, and provided guidance in data interpretation and manuscript writing. All other listed authors were contributors in patients' enrollment and data acquirement. All authors read and approved the nal manuscript. Figure 1 Flow chart of recruitment and participation in this study.  Distribution of deep white matter hyperintensity loads in different grades and subtypes of RLS patients.
No statistically signi cant differences in DWMH loads were found with regard to RLS grades (a) or subtypes (b). DWMH: deep white matter hyperintensity; RLS, right-to-left shunt