Relationship between right‐to‐left shunt, hypoxia, and epilepsy

Abstract

Objective

A right‐to‐left shunt (RLS) can mediate the hypoxic state, and hypoxemia is relevant for the development of drug‐resistant epilepsy (DRE). The objective of this study was to identify the relationship between RLS and DRE and further investigate the contribution of RLS to the oxygenation state in patients with epilepsy (PWEs).

Methods

We performed a prospective observational clinical study of PWEs who underwent contrast medium transthoracic echocardiography (cTTE) between January 2018 and December 2021 at West China Hospital. The collected data included demographics, clinical features of epilepsy, antiseizure medications (ASMs), RLS identified by cTTE, electroencephalography (EEG), and magnetic resonance imaging (MRI). Arterial blood gas was also assessed in PWEs with or without RLS. The association between DRE and RLS was quantified using multiple logistic regression, and the parameters of oxygen levels were furtherly analyzed in PWEs with or without RLS.

Results

A total of 604 PWEs who completed cTTE were included in the analysis, of which 265 were diagnosed with RLS. The proportion of RLS was 47.2% in the group of DRE, and the proportion of RLS was 40.3% in the group of non‐DRE. Having RLS was associated with DRE in multivariate logistic regression analysis (adjusted OR = 1.53, P = 0.045). In the analysis of blood gas, the partial oxygen pressure in PWEs with RLS was lower than those without RLS (88.74 mmHg versus 91.84 mmHg, P = 0.044).

Significance

Right‐to‐left shunt could be an independent risk factor of DRE, and low oxygenation might be a possible reason.

Key Points.

• There are approximately 47% of patients with drug‐resistant epilepsy (DRE) having right‐to‐left shunt (RLS), with a nonsignificant trend toward higher prevalence than non‐DRE.

• RLS is an independent risk factor after accounting for the other risk factors for DRE.

• Patients with epilepsy (PWEs) with RLS have a lower pressure of arterial oxygen (PaO₂) than those without RLS, which might be a potential underlying mechanism of DRE.

1. INTRODUCTION

Right‐to‐left shunt (RLS), with a prevalence of approximately 25%, ¹ is known to mediate hypoxemic events when microthrombi and deoxygenated venous blood from the right atrium directly enter the left atrium with oxygenated arterial blood. ² RLS can impair cerebrovascular autoregulation and trigger cortical spreading depression (CSD) by inducing ischemia/hypoxia, which predisposes patients to cryptogenic stroke, migraine, obstructive sleep apnea (OSA), and demyelination of the cortex. ³ , ⁴ The closure operation is currently considered as the first‐line treatment for cryptogenic stroke, severe OSA, and improvement of migraine frequency and severity. ⁵ , ⁶ RLS closure is associated with an improvement in oxygen saturation (SaO₂) and may convey collateral benefits to the neurologic system, especially in those with lung disease complications. ⁷ Epilepsy is characterized by unprovoked seizures with manifestations of abnormally synchronized electrical activity due to neuronal hyperexcitability. ⁸ Previous studies have found that chronic hypoxia can cause higher neuronal excitability, which can lower the epileptic seizure threshold involved in the increase of seizure susceptibility and thus lead to the development of drug‐resistant epilepsy (DRE). ⁹ Additionally, ictal hypoxemia is considered as one of the mechanisms underlying sudden unexpected death in epilepsy (SUDEP). ¹⁰ However, the association between RLS and epilepsy has never been explored, and the blood oxygen level of patients with epilepsy (PWEs) with and without RLS has never been compared either.

Nowadays, it has been found that the high prevalence of RLS was correlated with the commodity of migraine in PWEs, which could be an underlying cause of migraine in PWEs. ¹¹ Epilepsy is often comorbid with migraine, and they share similar symptom profiles, treatment options, and common pathogenic mechanisms, such as hypoxia, CSD, and microthrombi. ¹² , ¹³ Factors that acutely increasing oxidative stress in the brain, such as high altitude and psychosocial stress, can threaten neuronal viability, which is an immediate precipitant of both seizures and migraine attacks. ¹⁴ PWEs concomitant with migraine also have higher incidence of drug resistance and psychiatric disorders. ¹⁵ It has also been suggested that treatment for RLS was effective in reducing frequency of seizures and headache attacks in PWE with migraine. ¹¹ Nevertheless, the characteristics of epilepsy have never been investigated in PWEs with and without RLS. Therefore, we conducted a cohort study to test our hypothesis that RLS would be associated with DRE and PWEs with RLS could experience oxygen desaturation.

2. METHODS

2.1. Study population and data collection

This prospective, single‐center cohort consecutively recruited patients diagnosed with epilepsy from the Neurology Clinic, West China Hospital, Sichuan University, China. All enrolled subjects completed a participation pack containing an informed consent and a detailed information sheet of medical conditions. The current study was based on data from patients included in this cohort from January 2018 to December 2021. The inclusion criteria were as follows: (1) diagnosis of epilepsy; (2) age between 9 and 55 years; and (3) consent to participate in the study. Epilepsy was defined based on the International League Against Epilepsy (ILAE) 2014 criteria, ¹⁶ and all patients were diagnosed by two specialists independently. The exclusion criteria were as follows: (1) causes of epilepsy including tumor or neurotoxicity; (2) history of heart disease other than RLS; (3) clinical history of epilepsy, familial history, previous history, electroencephalography (EEG), and magnetic resonance imaging (MRI) could not be obtained; (4) coexistence of distinct respiratory, gastrointestinal, hematological, urinary, endocrine/metabolic diseases, and syndromes; and (5) unfinished contrast transthoracic echocardiography (cTTE) or RLS confirmed by cTTE at another hospital.

In addition to RLS identified by cTTE, variables for analysis included: age, gender, ethnicity, age of seizure onset, seizure symptoms, age of diagnosis, seizure frequency before and after antiseizure medications (ASMs), ASMs, classification of epilepsy based on the ILAE 2017 Classification of Seizure Types Basic Version, ¹⁷ smoking, alcohol, history of status epilepticus based on the 2015 ILAE definition, ¹⁸ febrile convulsions, prebirth, perinatal hypoxic complication, head trauma, encephalitis or meningoencephalitis, familial history of epilepsy, routine EEG and MRI. Medical histories were reviewed, and subjects also underwent a physical examination, electrocardiogram, urine routine, blood routine, and blood biochemistry before the cTTE examination to rule out other disorders. Patients were followed up from February 2018 to November 2022, and the last assessment focused on ascertaining seizure frequency, status epilepticus, usage of ASMs, and comorbidities. DRE was identified by 2010 ILAE definition ¹⁹ through outpatient visits or telephone interviews with patients or their legal guardians. The diagnosis of migraine was based on the International Classification of Headache Disorders (ICHD‐3 beta). ²⁰ Epileptiform discharges included spike waves, sharp waves, polyspikes, and spike/sharp wave complexes. Focal paroxysmal slow waves without epileptiform discharges were independently calculated. The MRI protocol with 1.5 or 3T (Tesla) machines included at least T1‐weighted, T2‐weighted, and fluid‐attenuated inversion recovery (FLAIR) sequences. Arterial blood gas was tested during their interictal period, and participants had no seizures during the day before and day of the measurements. Fasting arterial blood in the morning from the radial arteries was immediately analyzed using an on‐site blood gas analyzer (Cobas b123, Switzerland) after the interview at the outpatient department. Data were collected using the same standard forms in a Microsoft Excel file (V.2016). This study was approved by the Ethics Committee of West China Hospital of Sichuan University (ChiCTR‐OOC‐17011935 and ChiCTR2000031591).

2.2. Assessment of RLS

RLS was judged by transthoracic echocardiography (TTE) and with injection of contrast material conducted in tandem. First, TTE was performed using 1–5 MHz or 3–8 MHz multiplane transducers on an ultrasound machine (Philips IE33) for each patient. Once other cardiac diseases were identified by two experienced sonographers, patients were excluded. Next, the sonographers analyzed the presence of RLS by cTTE. Contrast medium containing microbubbles, made by an agitated solution prepared from 1 mL of blood and 1 mL of air with 8 mL of saline, was injected into the antecubital veins for increased sensitivity. ²¹ PWEs were assessed for RLS at rest, during a Valsalva maneuver (expiratory pressure at 60 mm Hg measured by manometer), and coughing. Second‐harmonic imaging was used to improve the identification and detection of microbubbles in PWEs with poor image quality. RLS was diagnosed if three or more detected microbubbles per frame appeared in the left atrium within three cardiac cycles. ²²

2.3. Data analysis

Univariate analysis was performed between DRE and non‐DRE. The chi‐squared or Fisher's exact test was used to compare categorical variables as appropriate. Continuous variables were expressed as mean ± SD and were compared using an independent‐sample t test. Variables with a P < 0.10 in univariate analysis were mainly included in multivariate logistic regression analyses carried out to assess the odds ratios (ORs) of the risk factors of DRE. Multicollinearity was assessed by the variable inflation factor (VIF), and variables were accordingly removed from the final model (VIF > 5). Bestglm package was also used to identify the best subset of assays for logistic regression model analysis. The best model was determined based on the lowest Akaike Information Criterion (AIC). A P < 0.05 was considered significant in logistic analysis.

We furtherly compared pressure of arterial oxygen (PaO₂), pH value, SaO₂, partial pressure of carbon dioxide (PaCO₂), and oxyhemoglobin concentration between PWEs with and without RLS to investigate the blood oxygen level. A P > 0.1 was reported as nonsignificant, whereas those with P < 0.1 were reported as statistical trends due to small samples of invasive arterial blood gas during the interictal period. The regression analysis was used to examine the interaction effects of RLS and PaO₂ in predicting DRE. All statistical analyses were conducted in R 4.0.1.

3. RESULTS

We carefully interviewed 1136 PWEs, and a total of 604 PWEs with determined drug responsiveness (DRE or non‐DRE) were finally included in the analysis of relationship between RLS and DRE. The 532 individuals who did not meet eligibility criteria included several subgroups: 101 (18.98%) not meeting our inclusion criteria, 7 (1.32%) with unfinished cTTE, 204 (38.35%) with no follow‐up visit after 2018, and 220 (41.35%) with undetermined drug responsiveness until the last follow‐up. The detailed reasons for exclusion are shown in Figure 1. The baseline characters of 220 PWEs with undetermined drug responsiveness and 604 PWEs with determined drug responsiveness were compared (Table S1). There were less PWEs with aura, frequency more than 1 per month before ASM start, focal epilepsy and more PWEs with diagnostic delay more than 5 years and unknown classification in the group of PWEs with undetermined drug responsiveness, while there were no differences in other aspects between PWEs with undetermined and determined drug responsiveness. Among 604 eligible PWEs, 51.16% (309/604) had a DRE outcome. The mean age of our analyzed cohort was 27.96 years old (9 to 55 years; SD = 11.00 years), and 54.5% were female patients. The overall prevalence of RLS was 43.87% (265/604) in our analyzed cohort.

FIGURE 1.

Flowchart of patient recruitment (ASM, antiseizure medication; cTTE, contrast medium transthoracic echocardiography; PWE, patient with epilepsy; TTE, transthoracic echocardiography).

The clinical features between the DRE and non‐DRE groups are presented in Table 1. Age distribution, ethnicity, and sex ratio in the DRE group were the same as those in non‐DRE group. Patients with DRE were more likely had younger seizure onset age, longer diagnostic delay, taking more ASMs, higher rate of perinatal hypoxic complication and head trauma history, and higher incidence of abnormalities in routine EEG and MRI (Table 1). Higher rate of RLS was also shown to be associated with DRE (OR = 1.32). There were 13 variables (P < 0.10 in univariate analysis) finally included in the logistic regression analysis. The results of logistic regression analyses are shown in Table 2. Multicollinearity was ruled out based on VIF, which were all <5. Finally, six risk factors for DRE were found, including seizure frequency ≥1 per month before ASM start (adjusted OR = 5.18, 95% CI [3.39, 7.91], P < 0.001), taking numbers of ASMs ≥ 2 (adjusted OR = 5.68, 95% CI [3.72, 8.67], P < 0.001), with aura (adjusted OR = 1.84, 95% CI [1.18, 2.87], P = 0.007), duration of epilepsy ≤10 years (adjusted OR = 0.60, 95% CI [0.38, 0.95], P = 0.029), normal MRI (adjusted OR = 0.49, 95% CI [0.30, 0.80], P = 0.004), and with RLS (adjusted OR = 1.53, 95% CI [1.01, 2.32], P = 0.045). To further analyze the potential association between RLS and DRE, we used the bestglm package to identify the best subset for logistic regression of DRE outcome. In addition to the factors mentioned in model 1, a history of status epilepticus (adjusted OR = 1.63, 95% CI [1.12, 2.37], P = 0.011) was also shown to be associated with DRE in model 2 (Table 2). RLS was still a positive independent risk factor of DRE (adjusted OR = 1.28, P = 0.044) in model 2. Although patients with DRE showed a higher prevalence of head trauma history, we did not observe significant association for DRE based on the logistic regression analysis of both models (Tables 1 and 2).

TABLE 1.

Comparison of the clinical features between the DRE and non‐DRE groups.

	DRE (n = 309)	Non‐DRE (n = 295)	P value
Gender (male, %)	132 (42.7)	143 (48.5)	0.156 a
Ethnicity (Han, %)	296 (95.8)	283 (95.9)	0.932 a
Age at the last follow‐up (y, mean ± SD)	28.48 ± 10.34	27.31 ± 11.64	0.156 b
Age at seizure onset (≤10 years, %)	74 (23.9)	53 (18.0)	0.071 a
Duration of epilepsy (≤10 years, %)	149 (48.2)	209 (70.8)	<0.001 a
Diagnostic delay (>5 years, %)	57 (18.4)	35 (11.9)	0.024 a
Aura (%)	140 (45.3)	73 (24.7)	<0.001 a
Bilateral tonic–clonic seizures (%)	222 (71.8)	226 (76.6)	0.181 a
History of status epilepticus (%)	38 (12.3)	28 (9.5)	0.269 a
Seizure frequency before ASM start (≥1 per month, %)	204 (66.0)	70 (23.7)	<0.001 a
RLS (%)	146 (47.2)	119 (40.3)	0.087 a
Classification (%)
Focal	286 (92.6)	235 (79.7)	<0.001 a
Generalized	18 (5.8)	37 (12.5)	0.004 a
Unknown	5 (1.6)	23 (7.8)	<0.001 a
Numbers of ASMs (≥2, %)	240 (77.7)	94 (31.9)	<0.001 a
Migraine (%)	63 (20.4)	45 (15.3)	0.100 a
Preterm birth (%)	6 (1.9)	6 (2.0)	0.935 a
Perinatal hypoxic complication (%)	22 (7.1)	10 (3.4)	0.041 a
History of febrile convulsions (%)	38 (12.3)	33 (11.2)	0.672 a
History of head trauma (%)	34 (11.0)	20 (6.8)	0.069 a
History of encephalitis or meningoencephalitis (%)	29 (9.4)	21 (7.1)	0.312 a
Family history of epilepsy (%)	16 (5.2)	16 (5.4)	0.893 a
Smoke (%)	15 (4.9)	18 (6.1)	0.500 a
Alcohol (%)	16 (5.2)	21 (7.1)	0.320 a
EEG (normal, %)	103 (33.3)	132 (44.7)	0.004 a
MRI (normal, %)	147 (47.6)	213 (72.2)	<0.001 a

Note: Bold indicates significant differences at P < 0.10.

^a χ² test.

^b Student's t test.

TABLE 2.

Logistic regression analysis showed negative or positive predictors of DRE.

	Coefficient	Standard error	OR (95% CI)	Adjusted OR (95% CI)	P value	VIF
Model 1
Age at seizure onset ≤10 years	0.0813	0.2812	1.44 (0.97, 2.14)	1.08 (0.63, 1.88)	0.773	1.1659
Diagnostic delay >5 years	0.1988	0.3042	1.68 (1.07, 2.65)	1.22 (0.67, 2.21)	0.557	1.1418
Duration of epilepsy ≤10 years	−0.5119	0.2353	0.38 (0.27, 0.54)	0.60 (0.38, 0.95)	0.029	1.2833
Seizure frequency before ASM start ≥1 per month	1.6445	0.2160	6.24 (4.37, 8.92)	5.18 (3.39, 7.91)	<0.001	1.0697
Numbers of ASMs ≥ 2	1.7367	0.2161	7.21 (5.02, 10.35)	5.68 (3.72, 8.67)	<0.001	1.0837
RLS	0.4249	0.2130	1.32 (0.96, 1.83)	1.53 (1.01, 2.32)	0.045	1.0440
Aura	0.6085	0.2277	2.52 (1.78, 3.56)	1.84 (1.18, 2.87)	0.007	1.1189
Focal onset	0.2357	0.5715	3.17 (1.91, 5.29)	1.27 (0.41, 3.88)	0.676	3.3781
Generalized onset	−0.2750	0.6469	0.43 (0.24, 0.78)	0.76 (0.21, 2.70)	0.674	3.1858
Perinatal hypoxic complication	0.0786	0.4887	2.18 (1.02, 4.70)	1.08 (0.42, 2.82)	0.872	1.0854
History of head trauma	0.5966	0.3784	1.70 (0.95, 3.03)	1.82 (0.86, 3.81)	0.112	1.1050
Normal routine EEG	0.0814	0.2333	0.62 (0.44, 0.86)	1.08 (0.69, 1.71)	0.727	1.2586
Normal MRI	−0.7071	0.2478	0.35 (0.25, 0.49)	0.49 (0.30, 0.80)	0.004	1.3891
Model 2
Duration of epilepsy ≤10 years	−0.3397	0.1246	0.38 (0.27, 0.54)	0.71 (0.56, 0.91)	0.007	1.0602
Seizure frequency before ASM start ≥1 per month	1.0029	0.1229	6.24 (4.37, 8.92)	2.73 (2.14, 3.47)	<0.001	1.0492
Numbers of ASMs ≥ 2	1.0252	0.1244	7.21 (5.02, 10.35)	2.79 (2.18, 3.56)	<0.001	1.0698
History of status epilepticus	0.4870	0.1915	1.34 (0.80, 2.24)	1.63 (1.12, 2.37)	0.011	1.0667
RLS	0.2441	0.1216	1.32 (0.96, 1.83)	1.28 (1.01, 1.62)	0.044	1.0197
Aura	0.4450	0.1265	2.52 (1.78, 3.56)	1.56 (1.22, 2.00)	<0.001	1.0193
History of head trauma	0.3268	0.2199	1.70 (0.95, 3.03)	1.39 (0.90, 2.13)	0.136	1.1004
Normal MRI	−0.3397	0.1246	0.35 (0.25, 0.49)	0.67 (0.52, 0.86)	0.001	1.0974

Note: Bold indicates significant differences at P < 0.05.

Abbreviations: CI, confidence interval; OR, odds ratio; VIF, variance inflation factor.

Furtherly, 64 PWEs consented to receive invasive arterial blood gas tests from the following cohort of 824 PWEs who completed cTTE. Among them, we excluded one patient whose ethnicity was Zang (who lived at altitudes above 2000 m). Finally, there were 24 patients with DRE, 20 patients with non‐DRE and 19 patients with undetermined drug responsiveness included in the analysis of blood gas. Although PWEs included in blood gas analysis were 4 years younger than those without blood gas tests, the other baseline characteristics of them were similar (Table S2). The details of patients tested for arterial blood gas are shown in Table 3. There was no significant difference in hemoglobin between PWEs with and without RLS (141.25 g/L versus 145.72 g/L, P = 0.273). PWEs with RLS had a lower level of PaO₂ (88.74 mmHg versus 91.63 mmHg, P = 0.044) than PWEs without RLS. Additionally, the level of pH value in PWEs with RLS showed a higher tendency without statistical significance compared with PWEs without RLS (7.396 versus 7.386, P = 0.065). However, there were no significant trends in SaO₂ (P = 0.242), PaCO₂ (P = 0.924) or oxyhemoglobin concentration (P = 0.797) observed between PWEs with and without RLS. We also visualized the interaction effect between RLS and PaO₂ in predicting DRE in the regression model by the visreg package. With the decrease in oxygen partial pressure, the probability of DRE had a higher trend in PWEs with RLS than that of PWEs without RLS, though not significantly (Figure S1).

TABLE 3.

Comparison of arterial blood gas in PWEs with and without RLS.

	RLS (n = 34)	No shunt (n = 29)	P value
Age at examination (years, mean ± SD)	25.30 ± 11.36	22.52 ± 10.67	0.526 a
Gender (male, %)	16 (47.1)	16 (55.2)	0.521 b
ASMs treatment outcome (%)
DRE	17 (50.0)	7 (24.1)	0.035 b
Non‐DRE	8 (22.9)	12 (41.4)	0.129 b
Undetermined	9 (25.7)	10 (34.5)	0.490 b
Age at seizure onset ≤10 years (%)	7 (20.6)	4 (13.8)	0.526 c
History of status epilepticus (%)	5 (14.7)	2 (6.9)	0.437 c
Migraine (%)	6 (17.6)	2 (6.9)	0.270 c
EEG (normal, %)	17 (50.0)	11 (37.9)	0.337 b
MRI (normal, %)	22 (64.7)	21 (72.4)	0.512 b
PaO2 (mmHg, mean ± SD)	88.74 ± 6.12	91.84 ± 5.81	0.044 a
PaCO2 (mmHg, mean ± SD)	38.64 ± 3.64	38.56 ± 3.46	0.924 a
SaO2 (%, mean ± SD)	97.74 ± 0.52	97.91 ± 0.62	0.242 a
Hb (g/L, mean ± SD)	141.25 ± 14.21	145.72 ± 17.83	0.273 a
oxyhemoglobin concentration (%, mean ± SD)	95.74 ± 0.85	95.67 ± 1.17	0.797 a
pH (mean ± SD)	7.396 ± 0.024	7.386 ± 0.021	0.065 a
C[HCO3‐] (mmol/L, mean ± SD)	23.24 ± 2.65	22.60 ± 2.04	0.293 a

Note: P < 0.1 were reported as trends.

^a Student's t test.

^b χ² test.

^c Fisher's exact test.

4. DISCUSSION

This is the first study to evaluate the association between RLS and DRE and provide a description of oxygenation in PWEs with RLS. The main findings of our study were that: (1) A tendency toward a higher rate of RLS in DRE than non‐DRE was observed although nonstatistically significant; (2) RLS was an independent risk factor after accounting for the other risk factors for DRE; (3) PWE with RLS desaturated more than those without RLS, which might be a potential underlying mechanism of DRE.

DRE affects 30%–40% of all PWEs, and half of DRE occurs with unknown reason. ²³ Several lines of evidence have shown that early age at onset, a long duration of epilepsy, high frequency of seizures before treatment, and structural lesions increase the risk for developing DRE; however, these factors only account for partial DRE occurrence. ²⁴ , ²⁵ In the present study, our findings showed that RLS remained as an independent risk factor of DRE after accounting for other risk factors.

RLS is an abnormal hemodynamic condition mainly caused by congenital structural malformation in the atrial septum, which allows anaerobic blood and small molecules to avoid pulmonary circulation by directly entering the arterial system. ²⁶ Xu et al. reported that the RLS positive rate was approximately 40% in the migraine group and epilepsy group, which was much higher than that in normal population. ²⁷ It has been reported that hypoxemia is more common in individuals with RLS than those without shunt. ²⁸ This hypoxic state caused by microemboli or unoxygenated peripheral blood may be responsible for those focal neurological deficits. ²⁹ Previous studies have identified that OSA with RLS experience more desaturation for a given level of respiratory disturbance than those without RLS. ³⁰ A marked improvement in hypoxemia immediately after RLS closure has also been widely reported. ³¹

Oxygenation is a prognostic factor of operation in patients with DRE, ¹⁰ and transient hypoxemia was also associated with average maximal ASMs withdrawal. ³² Animal studies also revealed that repetitive seizures appear when pH imbalance is induced by hypoxia. ³³ There are several possible mechanisms regarding the pathogenesis of hypoxia‐induced DRE. First of all, it has been found that hypoxia‐inducible factor‐1 alpha (HIF‐α) could upregulate P‐glycoprotein in refractory epilepsy, and inhibiting HIF‐α could suppress seizures. ³⁴ , ³⁵ HIF‐α is an important transcription factor that regulates gene expression in response to decreases in oxygen availability. ³⁶ Additionally, intermittent hypoxemia could mediate CSD and oxidative stress in the brain, which are also associated with DRE. ³⁷ CSD is a phenomenon that extends from the initial site of excessive neuronal excitability to the surrounding tissues followed by continuous electrical inhibition, which is a pattern of seizure initiation and termination. ³⁸ Recordings of CSD could serve as diagnostic and prognostic monitoring tools for DRE. ³⁹ Concerning the link between hypoxia and epilepsy, angiogenic edema caused by hypoxia is also linked to the formation of CSD, holding an important role in repetitive seizures. ⁴⁰ , ⁴¹ , ⁴² Moreover, reactive oxygen species (ROS) induced by hypoxia are also associated with the development of DRE, probably by upregulating P‐glycoprotein. ³⁴ Low blood oxygen levels could also cause a reduction in thalamic and hypothalamic volume, which contributes to the development of epilepsy, refractory epilepsy, and even sudden unexpected death in epilepsy. ⁴³ , ⁴⁴ PaO₂ is one of the most sensitive parameters of hypoxia. ⁴⁵ PWEs with RLS showed lower PaO₂ in our study, which might explain the result that RLS is an independent risk factor for DRE.

Furthermore, many clinical manifestations of stroke, migraine, and epilepsy, such as visual aura and visual field sensitivity reduction, could be evoked by insufficient oxygen supply during CSD, ⁴⁶ , ⁴⁷ and cessation of RLS can result in improvement of hypoxemia which can reduce these symptoms. ⁴⁸ , ⁴⁹ Consistently, several clinical reports have described improved symptoms of OSA and migraine, artery stiffness and vasodilation, nocturnal oxygenation, and overall quality of life after closure of RLS. ⁵⁰ , ⁵¹ Interestingly, it has also been reported that the frequency of seizures in three PWEs with migraine decreased significantly after PFO closure. ¹¹ Based on our analyses, RLS could be a prognostic or predictive biomarker of DRE and the detection of RLS could influence the decision‐making for PWEs with DRE.

Our study had several limitations. First, the study design was inherently subject to bias as well as the potential effect of unmeasured confounders. It is likely that our results are more generalizable to PWEs with early diagnosis, frequent seizures before ASMs start and focal epilepsy for a high‐rate exclusion of undetermined DRE and further monitoring of drug resistance is needed in our future study. Second, the small sample size underwent invasive arterial blood gas analysis limited the ability to detect a significant change and to be further grouped into mild, moderate and severe RLS, which would be investigated in our future studies. Third, cTTE had a lower specificity than TEE, whereas it had higher sensitivity in large RLS and could greatly reduce the pain and economic costs of patients. ⁵² Fourth, we did not use a negative control group of healthy individuals to compare the incidence of RLS in PWEs in analysis. Fifth, we did not directly recruit and compare PWEs before and after the cessation of RLS, which should be explored in future studies.

5. CONCLUSION

Our study suggests that RLS could be an independent risk factor of DRE and hypoxia might underlie DRE in PWEs with RLS. Based on these results, whether the screening and cessation of RLS can benefit PWEs should be investigated in the future.

AUTHOR CONTRIBUTIONS

Bosi Dong and Lei Chen contributed to conception. Bosi Dong, Yajiao Li, Shixu He, Qi Lai, Ximeng Yang, Hui Wang, and Yusha Tang contributed to acquisition. Bosi Dong and Shuming Ji contributed to analysis. Bosi Dong and Anjiao Peng contributed to drafting. Bosi Dong, Min Wu, Yunwu Zhang, and Lei Chen contributed to revising. Bosi Dong, Yajiao Li, Shuming Ji, Shixu He, Qi Lai, Ximeng Yang, Hui Wang, Yusha Tang, Anjiao Peng, Min Wu, Yunwu Zhang, and Lei Chen made final approval.

FUNDING INFORMATION

This research was supported by the National Natural Science Foundation of China (82271500) and Outstanding Youth Scientific and Technological Foundation project, Science and Technology Department of Sichuan Province (2020JDJQ0018).

CONFLICT OF INTEREST STATEMENT

None.

EPILEPSIA ETHICAL PUBLICATION STATEMENT

We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

ETHICS APPROVAL

This study involving human participants complied with the declaration of Helsinki and was approved by the Ethics Committee of Sichuan University.

PATIENT CONSENT STATEMENT

All patients provided written informed consent.

CLINICAL TRIAL REGISTRATION

This study involving patients with epilepsy or migraine was registered at the Chinese Clinical Trial Register (ChiCTR‐OOC‐17011935 and ChiCTR2000031591). The research on the relationship of patent foramen ovale and epilepsy, ChiCTR‐OOC‐17011935, registered July 10, 2017, http://www.chictr.org.cn/showproj.aspx?proj=20071; Patent foramen ovale closure for epilepsy: a study based on genomics, proteomics and fMRI, ChiCTR2000031396, registered March 30, 2020, http://www.chictr.org.cn/showproj.aspx?proj=50816.

Supporting information

Figure S1

Table S1

Table S2

Dong B, Li Y, Ji S, He S, Lai Q, Yang X, et al. Relationship between right‐to‐left shunt, hypoxia, and epilepsy. Epilepsia Open. 2023;8:456–465. 10.1002/epi4.12710

DATA AVAILABILITY STATEMENT

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.