Evidence of shared genetic risk factors for migraine and rolandic epilepsy
Address correspondence to Dr. Deb K. Pal, Department of Clinical Neuroscience, Institute of Psychiatry, King’s College London, London SE5 8AF, U.K. E-mail: Deb.Pal@iop.kcl.ac.uk
Purpose: Evidence for a specific association between migraine and rolandic epilepsy (RE) has been conflicting. Children with migraine frequently have electroencephalographic (EEG) abnormalities, including rolandic discharges, and approximately 50% of siblings of patients with RE exhibit rolandic discharges. We assessed migraine risk in RE probands and their siblings.
Methods: We used cohort and reconstructed cohort designs to respectively assess the relative risk of migraine in 72 children with RE and their 88 siblings using International Classification of Headache Disorders (ICHD-2) criteria. Incidences were compared in 150 age and geographically matched nonepilepsy probands and their 188 siblings. We used a Cox proportional hazards model, using age as the time base, adjusting hazard ratios (HRs) for sex in the proband analysis, and for sex and proband migraine status in the sibling analysis.
Results: Prevalence of migraine in RE probands was 15% versus 7% in nonepilepsy probands, and in siblings of RE probands prevalence was 14% versus 4% in nonepilepsy siblings. The sex-adjusted HR of migraine for an RE proband was 2.46 [95% confidence interval (CI) 1.06–5.70]. The adjusted HR of having ≥1 sibling with migraine in an RE family was 3.35 (95% CI 1.20–9.33), whereas the HR of any one sibling of a RE proband was 2.86 (95% CI 1.10–7.43).
Discussion: Migraine is strongly comorbid in RE and independently clusters in their siblings. These results suggest shared susceptibility to migraine and RE that is not directly mediated by epileptic seizures. Susceptibility gene variants for RE may be tested as risk factors for migraine.
Reciprocal comorbidity between epilepsy and migraine, two common childhood conditions, has long been recognized (Froelich et al., 1960; Lennox & Lennox, 1960), and the association with migraine has been shown to apply across idiopathic, cryptogenic, and symptomatic classes of epilepsies (Ottman & Lipton, 1994). Among idiopathic epilepsies, benign occipital epilepsy has the strongest association with migraine (Andermann & Zifkin, 1998). Surprisingly, however, evidence of migraine comorbidity with rolandic epilepsy (RE), the more common idiopathic partial epilepsy, is conflicting (Santucci et al., 1985; Bladin, 1987; Giroud et al., 1989; Septien et al., 1991; Andermann & Andermann, 1992; Wirrell & Hamiwka, 2006). This is despite the fact that electroencephalography (EEG) findings of children with migraine show a higher than expected prevalence of centrotemporal sharp waves (CTS, or “rolandic discharges”) (Ziegler & Wong, 1967; Kinast et al., 1982), an association that points to a probable link between migraine and rolandic epilepsy. Because CTS waves are inherited in an autosomal dominant manner (Bali et al., 2007), one would expect siblings of children with RE, to be at increased risk for migraine.
We therefore aimed to determine whether (1) the risk of migraine is higher in children with RE than in children without epilepsy; and (2) whether siblings of RE children are at higher risk of migraine than siblings of children without epilepsy. Both RE and migraine are believed to be complex genetic disorders (Bali et al., 2005; Wessman et al., 2007; Strug et al., 2009). If migraine is both comorbid with RE and aggregates in RE siblings, then we would be able to conclude that RE and migraine share some underlying genetic risk factors. We would also have evidence to reject the hypothesis that migraines, in the context of epilepsy, are mediated by epileptic seizures. Answering both research questions will advance our understanding of the etiology of migraine in families with RE and allow for anticipatory diagnosis, counseling, and intervention for migraine in RE patients and families.
Patients and Methods
We assessed the comorbidity of migraine and RE using a cohort design, treating RE versus nonepilepsy as the exposure, and migraine as an age-dependent outcome. To assess the coaggregation of migraine in siblings of children with RE, we used a reconstructed cohort design. The resulting effect estimates compare migraine risk among siblings of RE children to that among siblings that are unexposed (siblings of children without epilepsy). We collected data from 72 RE patients, 150 children without epilepsy, the 88 siblings of our RE patients, and the 188 siblings of the nonepilepsy group. The general methodology for the study has been detailed elsewhere (Clarke et al., 2007). Briefly, children with classic RE and their families were recruited for a genetic study from eight pediatric neurology centers in the northeastern United States (see Acknowledgments for referring physicians). Ascertainment was through a proband, with no other family member required to be affected with RE.
All RE probands were centrally evaluated by a pediatric neurologist and at least one other physician on the study team—a pediatrician or family physician. After evaluation, RE probands were enrolled if they met stringent eligibility criteria for RE, including (1) at least one witnessed seizure with typical features: nocturnal, simple partial seizures affecting one side of the body, or on alternate sides; (2) orofacial–pharyngeal sensorimotor symptoms, with speech arrest and hypersalivation; (3) age of onset for RE between 3 and 12 years; (4) no previous epilepsy type; (5) normal global developmental milestones; (6) normal neurologic examination; (7) at least one interictal EEG with CTS waves and normal background, verified by two independent and blinded readers; (8) and neuroimaging read by two independent and blinded board-certified neuroradiologists that excluded an alternative structural, inflammatory, or metabolic cause for the seizures. Therefore RE probands with unwitnessed episodes, or with only secondary generalized seizures, were excluded, even if EEG was typical.
RE probands had their first seizure at a median age of 8 years (range 3–12 years). Most had <10 lifetime seizures; more than one-third had at least one secondary generalized seizure, but only two had a history of convulsive status epilepticus, and two-thirds had been treated with antiepileptic drugs (AEDs). Table 1 shows their seizure characteristics. Details of EEG and imaging findings have been reported previously (Bali et al., 2007; Boxerman et al., 2007).
Table 1. Clinical descriptors of RE probands
|Lifetime seizure total|
| ≤10||51 (71)|
| >10||21 (29)|
|Relation of seizures to sleep or drowsiness|
| Exclusive||63 (87)|
| Not exclusive||9 (13)|
|Ever treated with antiepileptic drugs||50 (70)|
Nonepilepsy participants were recruited from the pediatric clinics (general, gastroenterology, renal, orthopedic) in the same hospitals as the RE probands, but not from neurologic or psychiatric outpatient clinics. To be eligible, they had to be in the same age range as the RE probands and not have any lifetime history of epileptic seizures. We refer to the index child as the nonepilepsy proband. Comparison of age, sex, and demographic variables in RE probands and nonepilepsy probands is shown in Table 2.
Table 2. Demographic comparison of RE probands and nonepilepsy probands
|Number of probands||72||150|
|Median age of probands, years (range)||9.5 (3–22)||10.0 (2–16)|
|Male probands, n (%)||45 (63)||80 (53)|
|Number of siblings||88||188|
|Mean offspring per family||2.5||2.7|
|Median age of siblings, years (range)||10.0 (1–29)||10.0 (1–31)|
|Male siblings, n (%)||37 (42)||89 (47)|
|Mean parental education level|
| College education||56%||60%|
| Up to high school||44%||40%|
Both parents were interviewed, either together or separately, by a physician who was board certified in pediatric neurology (DKP) or family medicine (TC). The proband and siblings were also interviewed. The investigator administered a 125-item medical history questionnaire including diagnostic criteria from the International Classification of Headache Disorders (ICHD-2), (2004). These criteria have improved diagnostic sensitivity for migraine over previous definitions, especially in pediatric patients (Hershey et al., 2005; Lima et al., 2005). If a subject reported recurrent headaches, then the age-of-onset was recorded; in epilepsy probands the temporal relationship between seizures and headaches was sought: headaches occurring around the time of a seizure were classified as ictal headaches. Migraine equivalents were not assessed.
Data were analyzed using survival analysis because the onset age for migraine is variable (increasing at puberty) and not all subjects had surpassed the susceptible age range (Table 2). Subjects without migraine had their outcomes censored at the date of interview. We used Cox proportional hazards (PH) modeling to assess the degree of association with migraine in RE probands (comorbidity analysis) and siblings, and estimated hazard ratios (HRs) with 95% confidence intervals (CIs). For sibling data, further adjustment for proband comorbidity was performed to take into account familial aggregation of migraine as a confounder for the association of migraine in siblings (Ziegler & Wong, 1967). We also used the log-rank test and compared our results to the PH models because in many analyses we had a limited number of events (see Table 3). We used graphical and testing approaches based on Schoenfeld residuals to assess the validity of the PH assumption, and we used diagnostics based on the Cox-Snell residuals and Nelson-Aalen estimated cumulative hazard function to determine the overall fit of the model.
Table 3. Frequency of migraine in RE probands, nonepilepsy probands, and siblings
|Probands with migraine||11 (15)||11 (7)|
|Probands with ≥1 sibling with migraine||11 (19)||6 (5)|
|Siblings with migraine||12 (14)||8 (4)|
To assess the degree of association of migraine in siblings of RE probands, we undertook two separate analyses, asking (1) does the presence of RE in the family result in an increased risk of having at least one nonepilepsy sibling with migraine (independent sibling analysis), and (2) is a sibling of a RE proband at greater risk of migraine than a sibling of a nonepilepsy proband (correlated sibling analysis, using multiple siblings’ survival data in the analysis).
In the comorbidity analysis, time to migraine onset in the RE probands and nonepilepsy probands was modeled as a function of RE status, adjusting for sex. In the independent siblings analysis, where one sibling per family was incorporated into the model, time to migraine onset in a sibling was modeled as a function of whether the sibling comes from an RE family, adjusted for sex of the sibling and the migraine status of the proband. We then fit a similar model for the correlated sibling analysis, incorporating all siblings in the analysis, not just one per family. We used the robust sandwich variance estimator (Kelly, 2004) in the PH models to take into account the correlation among siblings within a family. Analyses were performed using SAS 9.1.3 for Windows (SAS 2002–2003) and the R Statistical Package for Mac OS X (R Development Core Team 2008). The study was approved by the institutional review boards of the New York State Psychiatric Institute, Columbia University Medical Center, and all collaborating centers. Subjects gave appropriate informed consent.
As expected from study design, RE and nonepilepsy probands and siblings were comparable in terms of demographic factors (Table 2). The one exception was a higher proportion of boys among RE probands, which is to be expected given the epidemiology of RE. Parents were also of similar age and educational background in RE and nonepilepsy families. The median age of onset for migraine was 9 years (range 3–12 years) in RE probands, and 8 years (5–13 years) in nonepilepsy probands. More than half of migrainous probands were girls (59%). Twice as many RE probands met ICHD-2 criteria for migraine compared to nonepilepsy probands (15% vs. 7%) (Table 3), uncorrected for the difference in sex distribution between RE and nonepilepsy probands (Table 2); and three times as many siblings of RE probands met migraine criteria than siblings of nonepilepsy probands (14% vs. 4%), Table 3. There were no cases of migraine with aura, and no isolated periictal, migraine although four cases reported periictal headaches.
The tests and graphical diagnostics for the PH model indicated that the PH assumption held in all three analyses: the comorbidity analysis, the independent sibling analysis, and the correlated sibling analysis. The plot of Cox-Snell residuals versus the Nelson-Aalen estimate of cumulative hazard function was computed in the comorbidity and independent sibling analyses and appeared to be fairly linear, with slope one indicating that there were no large departures from the model assumptions (not shown). The results of the log-rank tests and the PH models in these two analyses were similar.
In the comorbidity analysis, the risk of migraine in RE probands was twice that in nonepilepsy probands, adjusted for sex: HR 2.46, standard error (SE) 1.05. In the independent sibling analysis, the risk of having at least one sibling with migraine in an RE family is almost four times that in a nonepilepsy family, after adjusting for sex of the sibling: HR 3.72, SE 1.91. When one also adjusts for the migraine status of the proband, the risk of migraine is about three times greater among siblings of RE probands compared to siblings of nonepilepsy probands: HR 3.35, SE 1.75.
When all siblings were included in the model, the hazard of migraine in siblings of RE probands was three times that in siblings of nonepilepsy probands, after adjusting for sex of the sibling: HR 3.21, SE 1.56. Even after adjustment for sex of the sibling and migraine status of the proband, the sibling risk in RE families was almost three times that of nonepilepsy families: HR 2.86, SE 1.39, indicating an elevated risk of migraine in RE families independent of the familial risk of migraine. These results are detailed in Table 4.
Table 4. Hazard ratios of association with migraine in RE probands and their siblings
| Probands, adjusted for sex||2.46 (1.06–5.70)||0.036|
|Independent sibling analysis (one sibling per family)|
| Siblings, adjusted for sex||3.72 (1.36–10.17)||0.011|
| Siblings, adjusted for sex and proband migraine status||3.35 (1.20–9.33)||0.021|
|Correlated sibling analysis (all siblings included)|
| Siblings, adjusted for sex||3.21 (1.24–8.30)||0.016|
| Siblings, adjusted for sex and proband migraine status||2.86 (1.10–7.43)||0.031|
Using stringent diagnostic criteria for migraine and RE, these analyses allow us to conclude that a robust and markedly elevated association exists between rolandic epilepsy and migraine in childhood. Furthermore, we also demonstrate an equally strong association of migraine in siblings of rolandic epilepsy probands. The migraine risk for probands and siblings is similar, between two and a half to three times that in comparable nonepilepsy subjects. The combination of positive association in both probands and siblings suggests a mechanism of shared susceptibility to migraine and epilepsy that is not directly mediated by epileptic seizures but may involve a shared predisposition for neuronal hyperexcitability.
There are a number of possible sources of bias and confounding that must be considered when interpreting the results. Bias in selecting comorbid RE probands is the most obvious one, which would serve to inflate estimates of comorbidity. However, most RE probands in the northeastern United States are first diagnosed in pediatric neurology centers (Berg et al., 1999), and so we believe our ascertainment scheme resulted in a relatively unbiased community-based sample. Table 1 shows that RE probands had a distribution of onset age, lifetime seizures, and treatment history representative of “typical” RE patients. Even allowing for the possibility of positive selection for comorbid RE probands, this could not explain the increased risk of migraine in siblings after adjustment for proband migraine status. An alternative possibility is that we had negatively selected for migraine among controls. However, the frequency of migraine in control children is consistent with prevalence estimates in this age group (Lipton, 1997). Therefore, selection bias, operating either on RE or nonepilepsy probands, is unlikely to play a large part in the strong associations demonstrated here.
Recall bias is another important consideration, but the average age of migraine onset was not too distant from the median age of subjects, about 2–3 years. Therefore, the probability of “telescoping,” or of parents of exposed (RE) probands recalling the onset of migraine as being more or less recent than it actually was, is low. Moreover, recurrent headaches are fairly well recalled, and the use of a checklist including the ICHD-2 criteria minimized both recall bias for migraine in favor of other types of headache, as well as possible observer bias in interpretation of symptoms or differential misclassification. Among possible confounding factors, migraine epidemiology is principally influenced by age and sex; ethnic differences have also been noted (Lipton, 1997). Our RE and nonepilepsy probands were of comparable age distribution (Table 2), and we used a survival analysis to allow for varying periods at risk; sex differences in migraine incidence appear mostly in postpubertal children, whereas our sample had an overall median age of 10 years; moreover we adjusted for sex in the Cox PH model. Ethnicities were also equally matched between groups (Table 2). Therefore, the major potential confounders have been accounted for. One other possible confounder is the use of antiepileptic drugs (AEDs), which might theoretically have masked symptoms of migraine in RE probands (e.g., sodium valproate), or conversely exacerbated headache symptoms (e.g., carbamazepine). However, the elevated risk of migraine is still found in the subgroup of RE probands never exposed to AEDs (data not shown) and also would not explain the increased migraine risk in siblings of RE probands.
In previous studies, the incidence of migraine in rolandic epilepsy, and its association, has been unclear, likely as a consequence of a number of methodologic limitations involving ascertainment, study design, migraine definition, and statistical analysis. For example, case series, in which subjects were ascertained through specialist centers or who had survived prolonged follow-up, yielded high prevalence figures for migraine in RE ranging from 62–80% (Bladin, 1987; Giroud et al., 1989; Septien et al., 1991). Such selection pathways are likely to exaggerate the prevalence of comorbid cases. In contrast, it is unclear why migraine might have been underestimated in another center-based study of rolandic epilepsy (Santucci et al., 1985). Possible explanations include: (1) case definitions for RE were not stated, leading to a possible dilution of the association if other forms of epilepsy or nonepilepsy were included; (2) the possibility that RE patients had not surpassed the susceptible age range for migraine at the time of interview (age statistics were not reported); and (3) the less sensitive Prensky criteria were used for defining childhood migraine (Prensky & Sommer, 1979). One fully documented case-control study, and the only one using ICHD-2 criteria, supports the association of RE and migraine, but did not quantify the risk or allow for censoring of observations due to varying durations of follow-up (Wirrell & Hamiwka, 2006).
Migraine has a strong genetic component, both in rare Mendelian forms and as a common complex genetic disorder (Gervil et al., 1999; Wessman et al., 2007). As such, migraine is familial and, therefore, should be expected to aggregate in families of probands with migraine (Ziegler & Wong, 1967), even when probands are comorbid with epilepsy. The results of our study allow us to conclude that, even allowing for the familiality of migraine, the risk of migraine in siblings of rolandic epilepsy probands is elevated approximately 3-fold. This conclusion could be further validated by comparing the risk of migraine in siblings of probands with migraine alone. Although the aggregation of migraine in families of comorbid RE probands has been mentioned before (Bladin, 1987; Septien et al., 1991), it has not been shown in parents (Wirrell & Hamiwka, 2006), possibly due to parental recall bias of distant events, and/or by an association with migraine confined to childhood. This study is the first assessment of migraine–RE association in siblings.
The increased migraine risk in siblings may be explained by shared susceptibility factors for migraine and rolandic epilepsy. Because both RE and migraine are assumed to have complex genetic inheritance (Bali et al., 2005; Wessman et al., 2007; Strug et al., 2009), it could be hypothesized that a part of the genotype shared between siblings and RE probands (probably influencing cortical excitability) increases susceptibility to migraine headaches in childhood. Siblings are at approximately 50% risk of carrying the CTS wave trait (Bali et al., 2007), which has recently been mapped to variants in the ELP4 gene (Strug et al., 2009), and these variants may hypothetically mediate the sibling risk for migraine. In this study we did not have complete EEG data on siblings to assess whether CTS mediated the mechanism of increased sibling migraine risk. However, this is an easily testable hypothesis if ELP4 genotype data and migraine affectedness status are available in family members. Migraine susceptibility in RE families may of course be attributable to genes other than those that influence CTS, or to the same genes that influence migraine without aura in the general population (Wessman et al., 2007). To date, there are no reports of susceptibility genes for common idiopathic epilepsies of complex inheritance that increase migraine risk. An increased migraine risk across all individual childhood epilepsy syndromes has yet to be confirmed. An explanation invoking purely shared environmental factors as a mechanism for increased sibling migraine risk seems unlikely as one would then expect all siblings in a family to be equally at risk.
The familial clustering of migraine in RE families adds to a growing list of co-aggregating traits in RE, including speech sound disorder and reading disability (Clarke et al., 2007); attention impairments are also associated with CTS and may, therefore, be expected in CTS-positive siblings (Kavros et al., 2008). The presence of RE in a family may, therefore, represent a marker for a heritable form of cortical excitability and neurodevelopmental impairment, but the prognostic value of a family history of RE on the course of migraine is currently unknown. Whether migraine attacks remit once the EEG trait disappears, or whether migraine in RE families transforms, persists, or progresses in a pattern similar to that in non-comorbid children with migraine are questions yet to be answered.
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. Our thanks to the families participating in the IRELAND study; and to Frances Rhoads MD; Jerry Boxerman MD, PhD; Jeffrey Rogg MD; Lewis Kull MS; Geoffrey Tremont PhD; Janessa Carvalho BS; Suzanne Foster BS; and to the members of the IRELAND consortium: Cigdem Akman MD; William D Brown MD; Murray Engel MD; John Gaitanis MD; Frank Gilliam MD; Karameh Hawash MD; Huntley Hardison MD; Geoffrey Heyer MD; Steven L Kugler MD; Linda Leary MD; David E Mandelbaum MD, PhD; Patricia McGoldrick RN; Edward Novotny MD, PhD; Steven M Wolf MD; and Maria Younes MD. This study was supported by members of the Partnership for Pediatric Epilepsy Research, which includes the American Epilepsy Society, the Epilepsy Foundation, Anna and Jim Fantaci, Fight Against Childhood Epilepsy and Seizures (f.a.c.e.s), Neurotherapy Ventures Charitable Research Fund, and Parents Against Childhood Epilepsy (P.A.C.E.) (DKP); the Epilepsy Foundation through the generous support of the Charles L Shor Foundation for Epilepsy Research, Inc (DKP); National Institutes of Health grants NS047530 (DKP), HG00-4314 (LJS).
Disclosure: The authors report no conflict of interest.