1 INTRODUCTION

Epilepsy can be a life-long disorder, but seizures may also discontinue with time. In fact, 70% of patients with epilepsy will enter remission.1 The remission rate is related, among other things, to the duration of follow-up. Long-term remission rates are mostly reported on patient cohorts with 10 to 20 years of follow-up. One existing population study with 50-year follow-up2 was based on baseline data of patients with “present epilepsy” in 1962–1964, and a subsequent questionnaire mailed in 1976 and 2015. Ninety percent of their surviving subjects were in 5-year terminal remission on or off medications. No outcomes other than remission data were given.

We have previously reported very long-term medical outcomes of subjects with childhood-onset epilepsy (COE) in a unique prospective cohort established in 1961–1964.3, 4 This is the only prospective, long-term, population-based interval study ever reported with regular follow-up check-ups from onset of childhood epilepsy to older age. While the seizure outcome was excellent, the subjects showed more neurological soft signs, cognitive and imaging abnormalities than controls.4-6 However, it is unknown if these findings represent a stable outcome (eg, accumulated during years of active epilepsy) or progressive changes related to faster brain aging that could predispose these individuals to adverse outcomes later in life. In order to answer this question, we had two goals for re-examining the cohort: (1) to characterize the cohort's neurological and seizure status, and (2) the longitudinal trajectory of previously reported findings at 55-year follow-up with focus on the last five years. Our hypotheses were that neurological signs would continue to increase and that brain aging would be accelerating in subjects with childhood-onset epilepsy, particularly in subjects with persisting active epilepsy.

2 MATERIALS AND METHODS

The study design is previously described in detail.3, 4, 7 In brief, the initial study consisted of all the children who were resident in the catchment area of Turku University Hospital, Turku, Finland, in 1961–1964, were less than 16 years, and had onset of epilepsy (repeated unprovoked seizures) or had a previous diagnosis of epilepsy with one or more seizures, considered as active epilepsy, in 1961–1964. Children were identified on the basis of several data sources: hospital inpatient and outpatient records; primary healthcare records; private offices patient records; and review of the National Health Service (NHS) records, a register of all patients residing in Finland. Review of the NHS register was of importance from the point of case identification because, according to the rule, all children with epileptic or probable epileptic seizures were to be admitted to hospital care. A total of 245 subjects were identified for a longitudinal, prospective study of outcomes.3 In 1992, age, sex, and domicile matched controls were chosen by the Population Register Centre from the general population of the study area for those 99 still participating subjects, who had uncomplicated epilepsy.7

In 2012, a new nested study, the Turku Adult Childhood-Onset Epilepsy study (TACOE-50) was initiated at 50 years following diagnosis of epilepsy in childhood.4 All subjects with uncomplicated epilepsy and all controls who had returned their completed study questionnaires during the last 10 years and gave written informed consent to participate were recruited. They underwent an extensive two-day investigation program that included clinical neurological examination by a single neurologist, chemical laboratory examinations, EEG recordings, 3T magnetic resonance imaging (MRI), and neuropsychological testing. For chemical laboratory and neuroimaging results, the assessor was single one per each investigation type and blinded to participants whether subjects or controls. The main results have been published previously.4-6

Fifty-one subjects and 52 controls consented in writing and participated to the complete investigation program in 2012 (Figure 1). Given reasons for decline were mostly personal or family health problems, job barriers, reluctance to attend too many medical visits and claustrophobia. Among subjects, the nonparticipants did not significantly differ from the participants in terms of seizure variables including age at onset, remission status, or medication status. The proportions of nonparticipants were similar in subjects and controls by sex, sociodemographic factors, academic or vocational education, working status, lifestyle, or chronic disorders.4

Flow chart of the participant enrollment

In 2017, after extended 5-year follow-up, this process was repeated (TACOE-55 study). Of 51 subjects, 41 (80%) participated as did 46 of the 52 controls (88%). Ten (20%) subjects and six (12%) controls declined. None of the subjects or controls died between the years 2012 and 2017 (Figure 1). The main measures for outcomes were findings in structured clinical neurological examination, chronic comorbidity other than epilepsy, and 3T magnetic resonance imaging (MRI). The MRI equipment using a single scanner (Siemens Skyra Fit 3T, Siemens Healthcare) was unchanged from 2012 to 2017, except for an upgrade from Verio to Skyra fit version before 2017 with comparable image quality, and the assessor was blinded to group (COE subjects vs. controls). White matter changes were assessed separately for periventricular and deep white matter regions. For quality control, every fourth month, the physicists checked the homogeneity of the magnetic field and the coil status and the radiologist continuously followed the possible occurrence of external disturbances like motion artifacts. Voxel-based morphometrics were not undertaken.

The TACOE-50 participants were re-studied using the same parameters including three-dimensional T1, T2, and FLAIR sequences, diffusion tensor imaging (DTI) and T2* sequence. MRIs were visually interpreted by a single neuro-radiologist (R.P.) blinded to the group status. Interpretation included assessment of hippocampal atrophy and age-related white matter changes. Hippocampal atrophy was evaluated using Schelten’s scale from 0 to 4, and white matter changes were evaluated using Fazekas scale from 0 to 3.8, 9 The outcomes were compared between subjects and controls, and between subjects with active epilepsy and remitted epilepsy.

2.1 Definitions

Epileptic seizures and etiology of seizures were defined according to the guidelines for epidemiological research of the International League Against Epilepsy (ILAE).10-12 For the present report epilepsy syndromes, classified initially by the ILAE 1989,11 and re-classified in 199913 in accordance with the contemporary ILAE guidelines,12 were again re-classified to meet the criteria of the current ILAE classification by Scheffer et al. (2017).14 Generalized syndromes included generalized tonic-clonic seizures alone, childhood absence, juvenile absence, and juvenile myoclonic epilepsy. Those who presented with tonic-clonic seizures and had normal EEG were not classifiable. Remission of seizures ever was defined as a seizure-free period of 10 years (10YRE),15 and the subject was considered as drug-responsive or a case of self-limited epilepsy. Terminal remission (10YTR) was defined as 10-year remission at last follow-up. Active epilepsy was defined as failure to enter ten-year terminal remission with or without medication during the last five years. High blood pressure was defined as systolic pressure >130 mmHg or diastolic pressure >80 mmHg according to the Finnish Current Care Guidelines (https://www.kaypahoito.fi/hoi04010). Obesity was defined as a body mass index of >30 kg/m2 according to the WHO guidelines.16 Abnormal neurological status was defined as one or more abnormal findings on examination. Abnormal findings, that is clinical signs, included the assessment of consciousness level, behavior, and orientation to time and place; oral communication; cranial nerve, locomotor, and fine-motor functions; balance; muscular tonus, muscular mass, reflexes, and strength; involuntary movements; and sensory functions. Prescribed daily doses (PDD, WHO Collaborating Centre for Drug statistics 2021)17 of the four most frequently used antiseizure drugs (ASDs), that is, carbamazepine (CBZ), diphenylhydantoin (DPH), phenobarbital (PHB), and valproate (VPA), were calculated and converted to grams for every subject. Life-long ASD load was defined as cumulative sums of all four (CBZ + DPH + PHB + VPA) and separately for diphenylhydantoin and phenobarbital (DPH + PHB).

2.2 Dropout analysis

The previously reported attrition rates at different stages of follow-up2 were low and without any differential dropouts. The attrition rate in the present study from TACOE-50 to TACOE-55 was also low with a non-significant difference between COE subjects and controls (12% vs. 20%, p = .289). Dropouts were more often female than male (23% vs. 3%, p = .004) for unknown reasons; dropouts were more frequent among those with no or less than weekly versus weekly alcohol consumption (22% vs. 4%, p = .015); the effects did not significantly differ between the COE and control subgroups. No other significant differences were found between the participants and nonparticipants. Reasons for dropouts are presented in Figure 1. Of 10 declined subjects, seven reported other diseases (multiple sclerosis, cancer, hearing problems, etc.), one suffered from claustrophobia and two did not present any particular reason. One control person had intensive cancer treatments, and the remaining six did not express any reason for refusal.

2.3 Statistics

The main outcomes were analyzed first as a function of group (COE subjects and controls) and second within the COE group comparing subjects with (a) active versus remitted epilepsy and (b) focal versus idiopathic generalized epilepsy syndrome. Fisher's exact test was used. The data were given as frequencies. The association of lifetime cumulative ASD load with the main outcomes was analyzed retrospectively using Wilcoxon rank sum tests. The data were given as medians (IQR). To detect possible differences in the pace of brain aging between COE subjects and controls during the 5-year follow-up, repeated measures binary or cumulative logistic regression analyses using general estimation equations (GEE) estimation were done. Due to the modest sample size, multiple comparison corrections were not made. As predictors, the models included study group (subjects vs. controls), time (2017 vs. 2012), and their interaction. The data were given as odds ratios (OR) with 95% confidence intervals (95% CI). p-values <.05 were considered as significant. The analyses were done using SAS version 9.4 (SAS Institute).

2.4 Ethics

The Institutional Review Board approved the study design (Diary No. T08/044/17, Study No. T286/2017). Written informed consent was given by both the subjects and controls.

3 RESULTS

At baseline (TACOE-50), the COE subjects had been followed for 50 years (mean 50.1, SD 1.06, median 51, range 49–52). The duration of extended follow-up (TACOE-55) was five years for COE subjects and controls. Total person-years of follow-up for subjects up to TACOE-55 were 2404 years. Age at the end of follow-up (TACOE-55) was 63.2 years (mean, SD 4.14, median 63.2, range 55.8–70.6) for COE subjects and 63.0 years (mean, SD 4.13, median 63.3, range 56.0–69.9) for controls. None of the controls had past or present epileptic seizures during follow-up. One fourth (26%) of the COE group had active epilepsy, and three fourths (74%) were in remission. Generalized epilepsies included 12 with generalized tonic-clonic seizures alone, two childhood absence and two juvenile absence epilepsies and one juvenile myoclonic epilepsy. Focal epilepsies were temporal in 15 and extra-temporal or non-localizable focal in five cases.

Compared with COE patients in remission, in 2017, neurologic signs were significantly more common in COE subjects who were not in remission. Of significant neurologic signs, cerebellar signs without visually observable asymmetry were most frequent followed by pyramidal and peripheral signs (Table 1). Neurologic signs in general and cerebellar signs in particular were, not unexpectedly, significantly more common in those with a history of focal than generalized epilepsies. Hypercholesterolemia was also significantly more common in subjects with focal than generalized epilepsy. High lifetime ASD loads for the four most frequently used ASDs (Figure 2) and for the combination of DPH and PHB (Figure 3) were associated with neurologic signs, especially cerebellar signs and peripheral neuropathy.

TABLE 1. Clinical neurologic and MRI findings in 37 adult subjects with remitted versus non-remitted childhood-onset epilepsy, or with focal versus genetic generalized epilepsy syndrome Epilepsy Active1 Remitted Focal Generalized n (%) n (%) p 2 n (%) n (%) p 2 Neurologic signs Total n 9 28 20 17 Any signs3 9 (100) 15 (54) .015 17 (85) 7 (41) .008 CNS signs Cerebellar 9 6 <.001 12 3 .018 Pyramidal 3 0 .011 2 1 >.99 Cognitive 2 3 .577 4 1 .350 Extrapyramidal 1 4 >.99 3 2 >.99 Cranial nerve 1 4 >.99 3 2 >.99 Peripheral signs 7 8 .017 11 4 .092 Non-neurologic somatic disorders 4 High BP5 9 24 .554 16 17 .109 Hypothyroidism 2 4 .620 4 2 .667 Hypercholesterolemia 2 9 .695 9 2 .037 Obesity6 4 10 .705 10 4 .173 MRI abnormalities Total 8 28 19 17 Any abnormalities3 7 (89) 20 (71) .648 14 (74) 11 (65) .721 Atrophy 4 7 .214 4 4 >.99 Cerebellar7 3 2 .062 0 2 .216 Cerebral 3 3 .109 3 1 .605 Hippocampal8 0 3 >.99 1 1 >.99 Markers of cerebrovascular disease2 6 14 .257 5 4 >.99 White matter changes, Fazekas scale .345 .695 Score 1 2 10 5 3 Score 2 1 2 0 0 Score 3 1 0 0 1 Infarcts 3 3 .109 5 1 .182 1 Seizures during last 10 years and/or ASD medication during last 5 years. 2 Fisher's exact test. 3 Number of participants: An individual may present several signs/disorders. 4 All participants had 1 or more non-neurologic somatic disorder. 5 High blood pressure: systolic >130 mmHg or diastolic >80 mmHg. 6 BMI > 30 kg/m2. 7 Assessed on scale 0–3: All participants with observed abnormalities were on grade 1. 8 Scheltens scale 0–4; all participants with observed abnormalities were on grade 1.

Life-long cumulative ASD load of CBZ, DPH, PHB, and VPA after 55 years of follow-up in 39 adult patients with childhood-onset epilepsy. p-values from Wilcoxon rank sum tests

Life-long cumulative ASD load of PHB and DPH after 55 years of follow-up in 39 adult patients with childhood-onset epilepsy. p-values from Wilcoxon rank sum tests

In COE subjects, there was a significant association between obesity and MRI white matter abnormalities (p = .006), but not between obesity and any MRI abnormality (p = .0827). MRI hippocampal atrophy was significantly more common in COE subjects with high arterial hypertension versus normal blood pressure (100% vs. 21%; p = .017), but did not reach significance in subjects with type 2 diabetes mellitus (p = .083). MRI white matter abnormalities were significantly associated with absence of vocational education (100% vs. 60%; p = .011), but not with basic education (p = .181). None of the MRI abnormalities was significantly more frequent in subjects with continuing seizures than in remitted COE subjects. Subsequent comparisons of epilepsy syndromes were not indicated given the modest sample sizes.

In the combined study cohort of COE subjects and controls, neurologic signs increased both in the COE subjects and in the controls from the TACOE-50 to TACOE-55 study, except for cranial nerve dysfunction that one control reported to have disappeared. While any neurological signs significantly increased in the combined study from 2012 to 2017, no significant difference was found between the groups or in the pace of change within the two groups (time × group interaction p = .8; not included in the final model. (Table 2). In 2012, cerebrovascular abnormalities in MRI were more common in COE subjects than controls. The difference evened during five-year follow-up due to more, mainly mild (Fazekas scale 1) white matter changes in controls than in COE subjects (group*time interaction p < .001) (Table 3).

TABLE 2. Presenting neurologic and other somatic comorbidity in older adults with childhood-onset epilepsy in 2012 and 2017, and in controls 2012 2017 Between subjects and controls Between 2012 and 2017 Subjects Controls Subjects Controls n (%) n (%) n (%) n (%) OR 95% CI OR 95% CI Neurologic signs Total 37 39 37 39 Any signs1 13 (35) 10 (26) 24 (65) 19 (49) 1.67 0.76–3.67 3.22 1.90–5.45 a CNS signs Cerebellar 3 2 15 6 Extrapyramidal 3 0 5 0 Cognitive 0 0 5 0 Pyramidal 1 0 3 1 Cranial nerve 6 4 5 6 Peripheral signs 3 (8) 5 (13) 15 (41) 16 (41) Neurologic disorders Total 37 39 37 39 Any disorders1 7 (19) 3 (8) 8 (22) 5 (13) 2.30 0.77–6.84 1.34 0.73–2.47 Migraine headache 2 3 1 3 Brain tumor 1 0 1 0 Peripheral neuropathy 1 0 2 2 Parkinsonian syndromes 1 0 1 0 Spastic torticollis 1 0 1 0 Mild ataxia 1 0 1 0 Hemisyndrome 1 0 1 0 Non-neurologic somatic disorders Total 37 41 37 41 Any disorders1 37 (100) 36 (88) 37 (100) 38 (93)