Diagnosis Switching and Outcomes in a Cohort of Patients with Potential Epilepsy with Myoclonic-Atonic Seizures

Abstract

INTRODUCTION:

There is overlap in the electroclinical features of many childhood epilepsy syndromes, especially those presenting with multiple seizure types, such as epilepsy with myoclonic-atonic seizures (EMAS) and Lennox-Gastaut syndrome (LGS). This study aimed to determine the frequency of diagnosis switching and the factors influencing epilepsy syndrome diagnosis in a cohort of children with possible EMAS, as well as to explore the relationship between epilepsy syndromes diagnoses, key electroclinical features, and clinically relevant outcomes.

METHODS:

This is a cross-sectional retrospective chart review of children treated at the Children’s Hospital of Colorado with a potential diagnosis of EMAS.

RESULTS:

There were 77 patients that met eligibility criteria, including 39% (n=30) with an initial diagnosis of EMAS and 74% (n=57) with a final diagnosis of EMAS. On average, for the 65% of patients who received more than one epilepsy diagnosis, the first, second, and third diagnoses were received within one year, three years, and ten years after epilepsy onset, respectively. Final diagnosis was significantly related to obtaining at least a six-month period of seizure freedom, p=0.03. Classic LGS traits, including paroxysmal fast activity, slow spike-and-wave, and tonic seizures were present in 50% of the overall cohort, although a minority of these patients had a final diagnosis of LGS. However, the presence of more LGS traits was associated with a higher likelihood of ongoing seizures. Adjusted for age of epilepsy onset, seizure freedom was half as likely for every additional LGS trait observed (0.49[0.31, 0.77], p=0.002).

CONCLUSION:

Current epilepsy syndrome classification has reduced applicability due to overlapping features. This results in diagnosis switching and limited prognostic value for patients with an overlapping clinical phenotype. Future studies should attempt to stratify patients based not only on epilepsy syndrome diagnosis, but also on the presence of various electroclinical traits to more accurately predict outcome.

1. INTRODUCTION

Epilepsy with myoclonic-atonic seizures (EMAS), also known as Doose syndrome, was first described in 1970. In the original report, Herman Doose described an epilepsy syndrome occurring most often in previously healthy and developmentally normal children with a relatively explosive onset of seizures, including myoclonic and / or astatic (now termed atonic) seizures, and often multiple other types of generalized onset seizures (Doose et al., 1970). The presence of myoclonic-atonic seizures is now considered to be critical for the diagnosis of EMAS (Nickels et al., 2018). Clinically the presence of myoclonic-atonic seizures may be preceded by a history of febrile seizures or afebrile generalized tonic-clonic seizures (Caraballo et al., 2013; Doose et al., 1970). Additional features of EMAS include onset between the age of 6 months and 6 years (peak 2–4 years), association with a family history of epilepsy, male predominance (2:1), and eventual generalized polyspike-and-wave discharges on EEG (Berg et al., 2010; Caraballo et al., 2013; Neubauer et al., 2005; Scheffer et al., 2016). Clinical features should not be more consistent with an alternative epilepsy syndrome diagnosis. However, it can be difficult to distinguish many childhood onset epilepsy syndromes, especially those with multiple seizure types, such as EMAS.

Overlapping electroclinical features and evolution of electroclinical features over time pose a challenge to early diagnosis of the underlying epilepsy syndrome. The characteristic myoclonicatonic seizures seen in EMAS are often not the first seizure type experienced by the patient, and the initial interictal EEG for many childhood epilepsy syndromes can be normal (Kelley and Kossoff, 2010). EMAS and Lennox-Gastaut syndrome (LGS) both have onset in early childhood with multiple generalized seizure types. Both can present with what is described by families as sudden spontaneous falls or “drop attacks”. Clinically described “drop attacks” can encompass many different seizure types including myoclonic-atonic, tonic, or atonic seizures, and also epileptic spasms (Blume et al., 2001; Caraballo et al., 2013). Distinguishing these seizure types with similar semiology can be challenging, even with EEG and EMG monitoring. LGS is classically characterized by intellectual disability, frequent tonic seizures, and electrographic slow spike-and-wave often with bursts of paroxysmal fast activity (Berg et al., 2010; Kaminska and Oguni, 2013). However, tonic seizures are also reported in up to 38% of patients with a diagnosis of EMAS, although the incidence varies by study (Caraballo et al., 2013; Doose et al., 1970; Kaminska, 1999; Trivisano et al., 2011). Additionally, the presence of preceding abnormal development, although more suggestive of LGS, was also reported in 19% of patients initially described with EMAS by Herman Doose (Doose, 1979; Doose et al., 1970). Yet, the majority of child neurologists identify normal early development of at least moderate importance in the diagnosis of EMAS (Nickels et al., 2018). Despite the overlapping features between EMAS and LGS, there is a stark difference in the prognosis for each diagnosis. The majority of children with EMAS will obtain seizure freedom over time with variable cognitive outcomes and children with LGS typically have ongoing chronic refractory seizures with moderate to severe intellectual disability (Filippini, 2006; Kaminska, 1999; Kim et al., 2015; Stephani, 2006; Trivisano et al., 2011). The overlapping features of childhood epilepsy syndromes, most notably EMAS and LGS, result in a portion of patients that do not clearly fit into any syndrome and are suggestive of a spectrum between diagnoses (Kaminska, 1999; Kelley and Kossoff, 2010). Unfortunately, despite updates in epilepsy syndrome classification, distinguishing these two epilepsy syndromes remains a clinical challenge.

The primary aim of this study was to determine the frequency of diagnosis switching and identify the factors influencing diagnosis in a cohort of children with possible EMAS. Specifically, we aimed to quantify the prevalence of LGS traits (slow spike-and-wave, paroxysmal fast activity, and tonic seizures) and the correlation between these features and final diagnosis in our cohort. We assessed the association of final epilepsy syndrome diagnosis with clinically important outcomes including seizure freedom and normal development. Finally, we evaluated the effect of these LGS traits and other key electroclinical features as determinants of outcome, regardless of final diagnosis.

2. METHODS

2.1. Study design

A cross-sectional retrospective chart review was performed at Children’s Hospital of Colorado including children with epilepsy onset between May 2004 and April 2017. Approval was obtained from the Colorado multiple institutional review board (COMIRB). Children were identified from a clinical database of patients with a potential diagnosis of EMAS. Eligibility criteria included consideration of a diagnosis of EMAS by a child neurology faculty member at any point in the patient’s course. A definitive diagnosis of EMAS was not required. Children were excluded if on detailed review they never had any clinical or electrographic evidence of drop seizures. Drop seizures were defined as a brief, sudden event causing the patient to fall. Additionally, patients were excluded if they had a clear structural etiology on imaging. Records were reviewed for patient demographics, as well as age of seizure and epilepsy onset, seizure types, treatment, developmental variables, seizure freedom, epilepsy diagnoses, EEG features, and family history. Epilepsy diagnosis was extracted by review of medical records, including child neurology encounter notes and epilepsy diagnosis codes. Family history included the presence of any positive family history of febrile seizures or non-acquired epilepsy. EEG features were determined based on the reports available in the electronic medical record. Slow spike-and-wave was defined as spike-and-wave discharges with a frequency of ≤ 2.5Hz. The presence of focal features was defined by the presence of any of the following; focal seizure semiology, focal electrographic slowing, or focal spikes reported on interictal EEG recordings. Seizure freedom was defined as absence of seizures for at least six consecutive months and remaining seizure free at the last point of contact. Six months was selected as a reasonable, though not perfect, surrogate for long-term seizure freedom. This seizure free duration was selected to maintain overall power of the study given the relatively small sample size with variable duration of follow-up. Developmental variables included the presence of developmental delay prior to seizure onset, developmental regression, and final developmental outcome. Developmental outcomes were split into two categories, either normal or abnormal. This was determined based on clinical documentation by the neurologist, as more specific developmental outcome measures were not universally available or reliable. Patients with less than one year of follow-up were excluded from outcome related analyses of development and seizure freedom. Study data were collected and managed using REDCap (Research Electronic Data Capture) electronic data capture tools hosted at University of Colorado Denver (Harris et al., 2009).

2.2. Statistical analysis

Data analysis was primarily exploratory. Patient characteristics were compared by epilepsy diagnosis, development, and seizure freedom using chi-square and Fisher’s exact tests for categorical variables and t-test and ANOVA for continuous variables. Subject trajectories were plotted with color coding indicating periods of epilepsy specific diagnoses. For patients who experienced more than one diagnosis, the average number of diagnoses per patient within a certain time period was described with a plot of the mean cumulative function (MCF). The MCF is a nonparametric estimate of the cumulative mean number of events (number of diagnoses in this case) and was estimated using the Nelson-Aelen estimator of the cumulative hazard function (REDA package in R). The three LGS traits of paroxysmal fast activity (PFA) positive EEG, slow spike-and-wave (SSW) positive EEG, and tonic seizures were plotted on the subject trajectories. Time until last seizure by number of traits present was examined with Kaplan Meier curves and log rank test (OIsurv package in R). The association was adjusted for age at epilepsy onset using a Cox proportional hazard model. All data analyses were conducted in R with a significance level of 0.05.

3. RESULTS

There were 77 patients that met eligibility criteria, of which 72 patients had at least one year of follow-up and were included in the outcome related analyses. The cohort consisted of 74% males with a mean age of epilepsy onset of 2.8 years (SD 1.3) and a mean follow-up time of 4.6 years (SD 2.8). Approximately two-thirds of the cohort had normal development prior to the onset of epilepsy. There was a family history of either epilepsy or febrile seizures in 16.9% and 45.5% of patients, respectively. Most children presented initially with a generalized tonic-clonic seizure (64.9%) and less frequently myoclonic-atonic seizures (14.3%). Thirty-five patients (46%) had an initial non-specific diagnosis of genetic generalized epilepsy (GGE), although at the last follow-up only seven patients (9.1%) continued to have a non-specific diagnosis of GGE. Of those patients with a final diagnosis of GGE, there was a mean time from epilepsy onset to final diagnosis of 2.6 years (SD 1.7) and a mean follow-up of time of 4.8 years (SD 2.1). There were 30 patients (39%) with an initial diagnosis of EMAS as compared to 57 patients (74%) with a final diagnosis of EMAS, including three patients that changed from a diagnosis of LGS to EMAS. Conversely, eight out of nine patients with a final diagnosis of LGS, had a preceding diagnosis of EMAS. Epilepsy diagnosis for each individual patient over time is depicted in Figure 1a. Twenty-seven patients (35%) received only one epilepsy diagnosis. On average, patients who received more than one epilepsy diagnosis, received the first, second, and third diagnoses within one year, three years, and ten years after epilepsy onset, respectively (Figure 1b).

Figure 1:

Epilepsy diagnosis switching over time since epilepsy onset (a) Epilepsy diagnosis for each individual patient since epilepsy onset, including all epilepsy diagnoses. (b) Mean cumulative function (MCF) with 95% CI (confidence interval) for average number of diagnoses over time for patients who have more than one epilepsy diagnosis (n=50). This demonstrates that for patients who received more than one epilepsy diagnosis, the first diagnosis is received within the first year after onset, the second diagnosis within three years, and the third diagnosis within ten years. EMAS - epilepsy with myoclonic atonic seizures; LGS - Lennox-Gastaut syndrome; CAE - childhood absence epilepsy

Most clinical and demographic patient characteristics were not associated with a patient’s final diagnosis, including initial diagnosis and initial seizure type (Table 1). However, the occurrence of myoclonic-atonic seizures (p=0.02) at any time was associated with a final diagnosis of EMAS. Similarly, the presence of tonic and atypical absence seizures at any time were associated with a final diagnosis of LGS, p=0.008 and p=0.001, respectively. Electrographic features associated with a final diagnosis of LGS included the presence of PFA (p=0.002), SSW (p=0.002), and multifocal spikes (< 0.001) on an EEG at any point in time. Additional analysis revealed a significant relationship between the number of classic LGS traits (PFA, SSW, and tonic seizures) and the final diagnosis such that there were no patients with zero traits who had a final diagnosis of LGS (p = 0.003) (Table 2). However, there were many patients with LGS traits who did not have a final diagnosis of LGS. Notably, 6 of the 14 patients with all three LGS traits had a final diagnosis of EMAS (Figure 3a). The details of epilepsy syndrome diagnosis for each patient with corresponding LGS traits over time is shown in Figure 2. This figure highlights the evolution of LGS electroclinical traits and explores the relationship to epilepsy syndrome diagnosis on an individual patient basis.

Table 1:

Patient characteristics of overall group and by final diagnosis category. Focal features is a composite defined by the presence of any of the following; focal seizure semiology, focal electrographic slowing, or interictal focal spikes. EMAS - epilepsy with myoclonic atonic seizures; LGS - Lennox-Gastaut syndrome

Patient characteristics	Overall (%)	EMAS (%)	LGS (%)	Other diagnoses (%)	p
n	77	57	9	11
Male gender	57 (74)	44 (77)	6 (67)	7 (64)	0.52
Mean age at epilepsy onset (sd)	2.8 (1.3)	2.9 (1.3)	2.5 (1.3)	2.1 (1.1)	0.14
Mean # of anti-seizure medications (sd)	4.5 (2.5)	4.3 (2.1)	6.0 (2.6)	4.8 (3.7)	0.14
Mean age at last visit (sd)	7.4 (3.0)	7.2 (2.9)	8.6 (4.6)	6.9 (1.6)	0.4
Mean years of follow-up (sd)	4.6 (2.8)	4.3 (2.7)	6.1 (3.9)	4.8 (1.7)	0.18
Family history of febrile seizure	13 (17)	11 (19)	0 (0.0)	2 (18)	0.52
Family history of epilepsy	35 (46)	29 (51)	2 (22)	4 (36)	0.23
Abnormal development prior to epilepsy	21 (27)	13 (23)	2 (22)	6 (55)	0.11
Developmental regression	19 (25)	11 (19)	4 (44)	4 (36)	0.15
History of status epilepticus	19 (25)	12 (21)	3 (33)	4 (36)	0.42
History of febrile seizure	13 (17)	8 (14)	2 (22)	3 (27)	0.51
Initial Diagnosis					0.49
EMAS	30 (39)	24 (42)	4 (44)	2 (18)
LGS	3 (4)	3 (5)	0 (0.0)	0 (0.0)
Other	44 (57)	30 (53)	5 (56)	9 (82)
Initial seizure type
Generalized tonic-clonic	50 (65)	37 (65)	7 (78)	6 (55)	0.58
Myoclonic-atonic	11 (14)	10 (18)	1 (11)	0 (0)	0.36
Myoclonic	7 (9)	5 (9)	1 (11)	1 (9)	1
Myoclonic absence	3 (4)	2 (4)	0 (0)	1 (9)	0.6
Absence	1 (1)	1 (2)	0 (0)	0 (0)	1
Atypical absence	0 (0)	0 (0)	0 (0)	0 (0)	n/a
Tonic	3 (4)	1 (2)	0 (0)	2 (18)	0.09
Focal	2 (3)	1 (2)	0 (0)	1 (9)	0.46
Seizure type at any time
Generalized tonic-clonic	63 (82)	46 (81)	8 (89)	9 (82)	1
Myoclonic-atonic	68 (88)	53 (93)	8 (89)	7 (64)	0.02
Myoclonic	68 (88)	52 (91)	8 (89)	8 (73)	0.18
Myoclonic-absence	20 (26)	13 (23)	3 (33)	4 (36)	0.52
Absence	49 (64)	39 (68)	3 (33)	7 (64)	0.15
Atypical absence	21 (27)	10 (18)	7 (78)	4 (36)	0.001
Tonic	26 (34)	15 (26)	7 (78)	4 (36)	0.008
Focal	11 (14)	7 (12)	3 (33)	1 (9)	0.22
EEG characteristics
Slow spike-and-wave	28 (36)	16 (28)	8 (89)	4 (36)	0.002
Paroxysmal fast activity	21 (27)	11 (19)	7 (78)	3 (27)	0.002
Focal features	37 (48)	22 (39)	7 (78)	8 (73)	0.02
Focal spikes	22 (29)	12 (21)	4 (44)	6 (55)	0.05
Multifocal spikes	27 (35)	12 (21)	8 (89)	7 (64)	< 0.001
Generalized spikes	66 (86)	50 (88)	7 (78)	9 (82)	0.65
Generalized polyspikes	59 (77)	43 (75)	8 (89)	8 (73)	0.82
Diffuse background slowing	61 (79)	44 (77)	9 (100)	8 (73)	0.26
Response to photic stimulation	16 (21)	11 (19)	2 (22)	3 (27)	0.81
Response to hyperventilation	11 (14)	7 (12)	1 (11)	3 (27)	0.42

Table 2:

Relationship between the number of classic LGS electroclinical traits present and outcomes of development and final diagnosis. EMAS - epilepsy with myoclonic atonic seizures; LGS - Lennox-Gastaut syndrome

		Development		Final Diagnoses
Number of Traits	n	Normal (%)	p	EMAS (%)	LGS (%)	Other (%)	p
Overall	72	17 (24)	0.2	52 (72)	9 (13)	11 (15)	0.003
0	36	12 (33)		31 (86)	0 (0)	5 (14)
1	14	2 (14)		9 (64)	2 (14)	3 (21)
2	8	2 (25)		6 (75)	1 (13)	1 (13)
3	14	1 (7)		6 (43)	6 (43)	2 (14)

Figure 3:

Relationship between classic LGS electroclinical traits (SSW, PFA, and tonic seizures) with final diagnosis and outcomes. (a) The number of LGS traits, including specific electroclinical feature, by final diagnosis category of either EMAS, LGS, or other. The percentage of patients from each final diagnosis cohort is represented on the y-axis. (b) Kaplein Meier curves demonstrating time to last seizure by number of positive LGS electroclinical traits, regardless of specific trait present. There is a higher likelihood of ongoing seizures with each additional LGS trait present (log rank test p=0.018). EMAS - epilepsy with myoclonic atonic seizures; LGS - Lennox-Gastaut syndrome; PFA - paroxysmal fast activity; SSW - slow spike-and-wave; tonic – tonic seizures.

Figure 2:

Individual epilepsy syndrome diagnosis by category of diagnosis (EMAS, LGS, other) for each patient by the patient’s age over time in years. There is superimposed notation of the presence of classic LGS electroclinical features, including the age at which these occur. This figure demonstrates on an individual level the evolution over time of specific LGS electroclinical features and relationship to epilepsy diagnosis. EMAS - epilepsy with myoclonic atonic seizures; LGS - Lennox-Gastaut syndrome; SSW - slow spike-and-wave; PFA - paroxysmal fast activity.

Within the entire cohort, 38% (n=27) of patients became seizure free for at least six months and 24% (n=17) had normal development at the time of last follow-up (Table 3). The final diagnosis was not significantly associated with developmental outcome (p = 0.16) but was associated with obtaining seizure freedom (p = 0.03). All patients with a final diagnosis of LGS had ongoing seizures. Initial diagnosis was not predictive of outcome and was not associated with either seizure freedom or developmental status (p=0.83 and p=1.0, respectively). Patients with tonic seizures were less likely to have normal development (p=0.04). However, the presence of more LGS traits was not significantly associated with a worse developmental outcome (p=0.2) (Table 2). None of the 21 patients with abnormal development at onset, or the 17 patients with developmental regression following epilepsy onset, had normal development at the time of last contact. On bivariable analysis, SSW (p=0.005) and tonic seizures (p=0.04), but not PFA (p=0.06), were associated with a reduced likelihood of achieving seizure freedom (Table 3). However, there was a higher likelihood of ongoing seizures with each additional LGS trait present (log rank test p=0.018) (Figure 3b). There was only one patient with all three LGS traits who was seizure free for at least six months at the last follow-up. Adjusting for age of epilepsy onset, a Cox proportional hazard ratio determined that seizure freedom was half as likely for every LGS trait observed (0.49[0.31, 0.77], p=0.002).

Table 3:

Patient characteristics and electroclinical features in relation to developmental outcomes and seizure freedom for > 6 months with ongoing seizure freedom at last follow-up time point. Focal features is a composite defined by the presence of any of the following; focal seizure semiology, focal electrographic slowing, or interictal focal spikes. EMAS - epilepsy with myoclonic atonic seizures; LGS - Lennox-Gastaut syndrome

		Development (%)			Seizure free for >= 6 months (%)
Patient Characteristics	Overall	Normal	Abnormal	p	No	Yes	p
n	72	17	55		45	27
Male gender	53 (74)	14 (82)	39 (71)	0.54	33 (73)	20 (74)	1
Mean age at epilepsy onset (sd)	2.7 (1.3)	3.2(1.5)	2.6 (1.2)	0.06	2.6 (1.2)	3.0 (1.5)	0.26
Abnormal development prior to epilepsy	21 (29)	0 (0.0)	21 (38)	0.002	17 (38)	4 (15)	0.06
Developmental regression	17 (24)	0 (0.0)	17 (31)	0.008	13 (29)	4 (15)	0.25
History of febrile seizure	12 (17)	3 (18)	9 (16)	1	9 (20)	3 (11)	0.52
History of status epilepticus	19 (27)	3 (18)	16 (30)	0.51	14 (31)	5 (20)	0.42
Tonic seizures	25 (35)	2 (12)	23 (42)	0.04	20 (44)	5 (19)	0.04
Atypical absence	21 (29)	4 (24)	17 (31)	0.76	17 (38)	4 (15)	0.06
EEG characteristics
Slow spike-and-wave	26 (36)	4 (24)	22 (40)	0.26	22 (49)	4 (15)	0.005
Paroxysmal fast activity	21 (29)	3 (18)	18 (33)	0.36	17 (38)	4 (15)	0.06
Focal features	34 (47)	5 (29)	29 (53)	0.11	24 (53)	10 (37)	0.23
Focal spikes	21 (29)	2 (12)	19 (35)	0.13	14 (31)	7 (26)	0.79
Multifocal spikes	26 (36)	1 (6)	25 (46)	0.003	22 (49)	4 (15)	0.005
Generalized spikes	62 (86)	13 (77)	49 (89)	0.23	39 (87)	23 (85)	1
Generalized polyspikes	55 (76)	12 (71)	43 (78)	0.53	38 (84)	17 (63)	0.05
Diffuse background slowing	57 (79)	11 (65)	46 (84)	0.17	41 (91)	16 (59)	0.002
Response to photic stimulation	16 (22)	2 (12)	14 (26)	0.33	11 (24)	5 (19)	0.77
Response to hyperventilation	10 (14)	3 (18)	7 (13)	0.69	7 (16)	3 (11)	0.73
Initial diagnosis				1			0.83
EMAS	27 (38)	6 (35)	21 (38)		18 (40)	9 (33)
LGS	3 (4)	1 (6)	2 (4)		2 (4)	1 (4)
Other diagnoses	42 (58)	10 (59)	32 (58)		25 (56)	17 (63)
Final diagnosis				0.16			0.03
EMAS	52 (72)	15 (88)	37 (67)		29 (64)	23 (85)
LGS	9 (13)	0 (0.0)	9 (16)		9 (20)	0 (0.0)
Other Diagnoses	11 (15)	2 (12)	9 (16)		7 (16)	4 (15)

4. DISCUSSION

This study represents the first systematic evaluation of the relationship between initial and final epilepsy syndrome diagnosis, key electroclinical traits, and outcomes in a cohort of patients with a clinical presentation concerning for epilepsy with myoclonic-atonic seizures (EMAS). The results support a spectrum of overlapping electroclinical traits that evolve over time. This results in frequent diagnosis switching and suggest a reduced prognostic value of the current diagnostic classifications. While final diagnosis was associated with obtaining a period of seizure freedom, initial diagnosis provided no prognostic value for developmental or seizure free outcomes. This suggests that diagnoses are switching as clinical features evolve, which makes the final diagnosis more reactive than predictive. Using trait-based prognostication to distinguish between syndromes with substantial overlap offers an alternative approach. As a proof of concept, the cumulative presence of each key LGS electroclinical trait allowed for more refined prognostication that was independent from epilepsy syndrome diagnosis.

It is estimated that an epilepsy diagnosis change can occur in 18–33% of patients with epilepsy; however, there is limited detail regarding the number of diagnosis changes for each individual patient or the timing of such changes (Camfield and Camfield, 2007; Shinnar et al., 1999). These studies were also prior to the most recent International League Against Epilepsy (ILAE) revisions regarding seizure classification and epilepsy syndrome diagnosis. In this cohort, 35% of patients had only one epilepsy diagnosis during the follow-up period suggesting that some patients have easy to classify epilepsy syndromes with consistent phenotypic features. However, 65% of patients had some amount of diagnosis switching. The overlap between electroclinical features in EMAS and related disorders suggests that diagnosis switching is more than clinician “misdiagnosis”, but rather a result of ambiguity in diagnostic criteria that does not adequately account for the diversity and evolution of features over time. For example, 50% of the children had a least one electroclinical trait that is typically thought to be characteristic of a diagnosis of LGS, although only 12.5% of patients received a final diagnosis of LGS. These findings are consistent with prior literature reports demonstrating overlapping features of many early childhood epilepsy syndromes and that the electroclinical features more commonly associated with LGS, such as PFA and tonic seizures, are not exclusionary to alternative diagnoses, such as EMAS (Kaminska, 1999; Kaminska and Oguni, 2013; Kelley and Kossoff, 2010; Nickels et al., 2018).

In this study, final epilepsy syndrome diagnosis does have prognostic value, although only in terms of seizure freedom. There were no children with a diagnosis of LGS who obtained seizure freedom for greater than six months during the follow-up period. However, prognosis may be further refined by consideration of additional electroclinical features. On bivariable analysis there were inconsistent associations between individual LGS electroclinical traits (SSW, PFA, or tonic seizures) and six-month seizure freedom, but when combined the three LGS traits provided independent prognostic information such that the presence of each additional trait additively predicted ongoing seizures. Prior studies have also attempted to better understand the importance of various electroclinical features (Doose, 1979; Inoue et al., 2014; Kaminska, 1999), yet many of these features are not part of the underlying epilepsy syndrome classification. For example, the presence of focal electroclinical features (focal seizure semiology, focal electrographic slowing, or focal spikes reported on interictal EEG recordings) was associated with a final diagnosis of LGS, but focal features was not a predictor of abnormal development or seizure freedom as previously reported (Inoue et al., 2014). Larger studies would likely improve on the prognostication this paper offers by incorporating a greater number of electroclinical traits into combined predictive models. This was not possible in the current study due to the number of patients relative to the number of potentially predictive variables.

The retrospective nature of this study limited the assessment of developmental and seizure free outcomes. More complete standardized developmental assessments with longer follow-up duration should be included in future studies. However, even with a dichotomous developmental outcome measure (normal vs abnormal), there were no patients with developmental delay prior to epilepsy onset (29%) or with developmental regression with epilepsy onset (24%) who ultimately had normal development. This finding already has clinical prognostic value but can be refined with more detailed outcome measures.

Additionally, the inclusion criteria in this study, only patients with a potential diagnosis of EMAS and a requirement for drop seizures, limited the spectrum of patients evaluated and resulted in a small sample size of some subgroups (i.e. LGS). Some of the trends observed may reflect the small sample size and future studies would benefit from inclusion of all patients with early childhood epilepsy syndromes. Furthermore, when evaluating the relationship between final diagnosis and outcome, there is an inherent provider bias, such that providers may be more likely to change a patient’s diagnosis to LGS if they have a less favorable outcome. One approach to address this inherent bias and limitation is to also consider electroclinical features independent from the diagnosis received, as done in this study. Alternatively, future prospective studies could consider an independent blinded review and final consensus diagnosis based on a priori criteria. Lastly, the data analysis included numerous comparisons between electroclinical features and outcomes and due to the number of comparisons performed, some of the findings may be spurious. However, this study was intended as an initial exploratory analysis that may prompt further investigation into this topic.

5. CONCLUSION

The current childhood epilepsy syndrome classification has reduced applicability for patients with overlapping electroclinical features, specifically children with features of both EMAS and LGS. This results in difficulty providing a clear epilepsy syndrome diagnosis, frequent diagnosis switching, and overall reduced prognostic value of these epilepsy syndrome distinctions. Future research should attempt to stratify patients based not only on epilepsy syndrome diagnosis, but also on the presence of various electroclinical traits to more accurately predict outcome and guide treatment decisions. Ultimately, this could suggest the need to group these diagnoses and use trait based predictive models to further stratify patients.

Highlights:

• More than half of patients with suspected EMAS undergo epilepsy diagnosis switching.

• A third epilepsy syndrome diagnosis is received, on average, within 10 years after onset.

• Final epilepsy syndrome diagnosis is associated with obtaining seizure freedom.

• Seizure freedom is half as likely for every additional LGS trait observed.

Abbreviations:

EMAS

epilepsy with myoclonic atonic seizures

LGS

Lennox-Gastaut syndrome

PFA

paroxysmal fast activity

SSW

slow spike-and-wave

CAE

childhood absence epilepsy

GGE

genetic generalized epilepsy