Seizure recurrence after a 1st unprovoked seizure:

Transcription

1 Seizure recurrence after a st unprovoked seizure: An extended follow-up W. Allen Hauser, MD; Stephen S. Rich, PhD; John F. Annegers, PhD; and V. Elving Anderson, PhD Article abstract-we followed 28 patients identified on the day of their st unprovoked seizure for a ean duration of 4 years. Seizures recurred in 64. Recurrence risks were estiated to be 4%, 29%, and 34% at,3, and 5 years following the st episode. A history of previous neurologic insult (reote syptoatic) was associated with a 2.5-fold increased risk of recurrence. Aong idiopathic cases, a sibling with epilepsy, a generalized spike and wave EEG, or a history of acute syptoatic seizure increased risk for recurrence. Aong reote syptoatic cases, status epilepticus, a prior acute syptoatic seizure, or Todd s paresis increased risk. Depending upon clinical features, recurrence risk at 5 years following a st seizure ranged fro 23% to 8%. TTeatent with anticonvulsant edication was not associated with a decrease in recurrence risks. NEROLOGY 99;463-7 The clinical approach to patients with a single unprovoked seizure reains controversial,,2 at least in part due to conflicting inforation on their risk for further seizures. In an earlier report of prognosis in a cohort followed prospectively for a ean of 22 onths, we estiated the overall risk of further seizures to be 27% at 3 years following the initial event.3 Here we report the results of an extended follow-up of this cohort. We odified case inclusion criteria to address criticis of our previous analy~is.~.~ Follow-up is now ore than 4 years, and a greater nuber of individuals have experienced recurrence, providing a larger effective saple size for ore precise estiates of risk over a longer duration and increased power to evaluate potential factors predictive of recurrence. Methods. Potential cases of st seizure were ascertained through a surveillance syste to identify patients presenting for evaluation of newly identified seizures. This coprised review of hospital adissions, referrals to EEG laboratories, and referrals to the neurology and epilepsy clinics at 4 hospitals affiliated with the niversity of Minnesota. To facilitate patient identification (as well as to assure optial edical care), eergency roo staff agreed to obtain neurologic consultation on all new cases of seizures or epilepsy, and hospital adission fro the eergency roo was expedited for such cases. Once infored consent was obtained, patients were screened to deterine eligibility for a series of studies. Interviews were scheduled to obtain edical (and especially neurologic) history, as well as faily and social history. The interview, requiring up to 3 hours, often could not be copleted at the tie of hospitalization in which case it was scheduled following discharge. Most patients with newly identified unprovoked seizures (74%) had experienced ultiple seizure episodes prior to their st edical contact. The subjects of this report are restricted to those who () had experienced a definite unprovoked seizure as docuented by an eyewitness, (2) on review of past history had not experienced a previous unprovoked seizure, and (3) were identified and had signed consent fors within 24 hours of the index seizure. In contrast to our previous report, we also required that (4) the interview be copleted within 3 days of the index episode. A st unprovoked seizure was defined as st seizure occurring without an identified proxiate precipitant. Seizures occurring in the context of uncertain precipitants such as sleep deprivation or stress were considered unprovoked. All seizures were categorized as partial or generalized onset based upon description of the seizure by an eyewitness according to criteria recoended by the ILAE.5 Generalized onset seizures by description were so categorized independent of findings on EEG or neurologic exaination. The patient s seizures were categorized by etiology3 as idiopathic (seizures in the absence of a historical insult thought to substantially increase the risk for unprovoked seizures) or reote syptoatic (seizures in individuals with a prior history of CNS insult known to substantially increase the risk for subsequent epilepsy, such as head traua? cerebrovascular in~ult,~ CNS infection,!j or static encephalopathy fro birth anifest by ental retardation or cerebral palsygjo). An unexplained localized abnorality identified in the diagnostic evaluation (eg, a reflex asyetry on exaination or focal slowing on EEG) did not odify this categorization but was treated as a potential confounder in the analysis. Individuals reained in the category of initial assignent by etiology and seizure type regardless of subsequent inforation since these are the data a clinician would have at the tie of Fro the G.H. Sergievaky Center (Dr. Hauser), College of Physicians and Surgeons, Colubia niversity, New York, NY; the Departent of Pathology (Dr. Rich) and the Dight Laboratories (Dr. Anderson), niversity of Minnesota, Minneapolis, MN and the School of Public Health (Dr. Annegen), niversity of Texas, Houston, TX. Supported in part by NS 38-. Received June 27,989. Accepted for publication in final for January,99. Address correspondence and reprint requests to Dr. W. Allen Hauser, 63 West 68th Street, New York, NY 32. August 99 NEROLOGY 4 63

2 initial assessent. Thus, an idiopathic case in who a CNS neoplas was identified subsequent to initial evaluation (4 such cases were identified in this cohort) reained in the idiopathic category for analysis of recurrence risks. The definition of single seizure includes status epilepticus (seizures continuing for 3 inutes or ore without interruption) and clusters of seizures (2 or ore) in the sae 24-hour period; excluded are acute syptoatic seizures (seizures occurring in close teporal association-operationally within week of an acute insult to the CNS-or at the tie of an acute systeic etabolic or toxic insult). Because this was an observational study, there was no attept to influence the practice of the treating physicians once patients had been identified. Thus, there was no standardization of treatent (if any) nor systeatic onitoring of adequacy of therapy in those for who anticonvulsant edication (AED) was recoended. We did obtain inforation on the use of AEDs fro the patient as part of the follow-up inforation, but we judged these data to be insufficient to evaluate adequacy of treatent or copliance. Five categories of EEG findings were evaluated () generalized spike and wave (GSW) patterns regardless of syetry, repetition, rate, regularity, or state of alertness during which they occurred; (2) focal epileptifor (sharp waves or focal spikes, ultifocal spikes, bilateral hoologous spikes; patterns of uncertain significance, such as 6-second spike and wave, sall sharp spikes, 4 and 6 positive spikes, were not included in this category); (3) focal slowing (unilateral polyorphic or rhythic slowing); (4) nonspecific slowing (increased theta or delta for age); and (5) noral. For the final analysis, focal slow and focal epileptifor groups were cobined as were the noral and nonspecific groups. Subjects were contacted at 6-onth intervals for 2 years fro date of the st seizure and annually thereafter through 985. If a seizure occurred in the interval between the st seizure and the tie of the initial interview, the case reained in the study, and that date of recurrence was considered an end point. Follow-up was terinated for any of the following: () death, (2) the occurrence of an event usually associated with an increased risk for unprovoked seizures, such as head injury with loss of consciousness or an acute cerebrovascular insult, or (3) request by the patient to terinate participation in the follow-up study. Potential isclassification (ie, st seizures as recurrent seizure or vice versa) was evaluated by coparing hospital records with study interview fors for a rando saple of 2 of the 645 patients classified as newly diagnosed epilepsy (recurrent unprovoked seizures) and all 28 patients classified as st seizures. The hospital records were reassessed to identify st seizure cases inappropriately classified as epilepsy because of a recurrent seizure between the index st seizure and interview. No such cases were found. Rather, our review identified 3 cases diagnosed as a st seizure in hospital records who, based upon our independent study interview, clearly had experienced previous unprovoked seizures (usually coplex partial) prior to the index seizure. Had we relied on hospital records for case ascertainent, these 3 cases of epilepsy would have been included as st seizures. Analytic procedures. The cuulative risks of recurrence were deterined by life-table ethods, with euent defined as an unprovoked seizure recurrence.j2 These ethods take into account the length of follow-up in each individual. The coputed risks, therefore, represent the risk of recurrence conditional on surviving seizure-free through the specified tie (follow-up) interval. nivariate analysis (Kaplan-Meier survival curves) for each variable relates the cuulative percent recurrence to the tie after st seizure.3 For this analysis recurrence in those with the factor in question is copared with those without, irrespective of the presence of copeting risk factors in either group. The proportional hazards odel was used to estiate univariate and ultivariate rate ratios, defined as the ratio of the rate of seizure recurrence in the group of patients with a given factor to the rate of seizure recurrence without that factor (eg, reote syptoatic versus idiopathic).4 Expected risks for specific factors were derived by adding the rates of recurrence estiated fro the subgroup of patients with no statistically significant risk factors (baseline) to the risk estiate of the factor in question. Age was analyzed both as a continuous and as a discrete variable. Findings were considered significant when the bounds of the 95% confidence interval did not include unity. Results. Of,67 newly recognized seizure patients included in the larger series, 28 were identified with a st unprovoked seizure, recruited within 24 hours, and had their interviews copleted within 3 days. These patients were followed prospectively a total of 853 person-years. Sixty-five subjects included in the initial report3 but not interviewed within 3 days of the index event are excluded in this analysis, while another 29 st seizure cases identified and recruited after the date of closure for the earlier analyses are included. Although 3 patients were lost to follow-up, alost all contributed at least 2 years to the analyses. A total of 2 subjects were withdrawn because of death (8), an intervening CNS insult such as head traua thought to increase the risk for seizures (8), or patient request to terinate our follow-up contacts (4). Recurrent unprovoked seizures during the follow-up interval were identified in 64 cases. Overall risk for seizure recurrence-kaplan-meier estiates (figure ). For the total group, recurrence was estiated to be 4%, 29%, and 34% at,3, and 5 years, respectively, following the st convulsive episode. Risk was estiated to be 39% by 7 years following the st episode and, aong the 26 patients followed ore than 7 years (contributing 5.9 person-years to the analysis), there were no further recurrences. The recurrence risk was not constant over the 7 years of follow-up. The risk was highest in the st year following the initial seizure and tended to decrease with increasing tie interval fro the index seizure. There were 3 subjects who had recurrent episodes in association with abrupt discontinuation of alcoholic beverages. These individuals were excluded fro ore detailed analyses, although neither the recurrence rates nor the estiates of recurrence risk for the various subgroups analyzed were altered by their exclusion. Etiology (idiopathic versus reote syptoatic st seizure). A history of an identified insult to the CNS (ie, stroke, infection, head traua) was an iportant predictor of seizure recurrence. The recurrence risk for the 49 patients with an idiopathic seizure was estiated to be lo%, 24%, and 29% at,3, and 5 years, respectively, following the st convulsive episode (figure 2). This contrasts sharply with recurrence risks of 26%, 4%, and 48% at,3, and 5 years aong the 59 patients with a reote syptoatic seizure. Different predictors of recurrence were identified in those with reote syptoatic seizures when copared with those with idiopathic seizures, so results of Kaplan-Meier analyses are presented separately. 64 NEROLOGY 4 August 99

3 Kaplan-Meier Product Liit Analysis A C 7 a I4 PI :.-I 4J rl 5 'Di 6 Kaplan-Meiet Product Liit Analysis - Sibling(s) affected (N=4) - No sibling affected (N=35) 2ov5 I... o $. I. I.,. I. I. I t Figure. Recurrence of seizures aong 28 patients after a st unprovoked seizure. In this and subsequent figures arrows identify standard errors. G & P.d 2.-I P B Kaplan-Meier Product Liit Analysis - Reote Syptoatic (N=59) - Idiopathic (N=49) z Figure 2. Recurrence of seizures aong 49patients with idiopathic seizures and 59 with reote syptoatic seizures. Idiopathic cases. Aong idiopathic cases, significantly increased recurrence risks were observed in subjects with a sibling with epilepsy (figure 3A), in subjects with GSW (figure 3B), and in subjects with histories of prior acute syptoatic seizures (figure 3C). For those with an affected sibling, the recurrence risk was 29% at year and 46% at 5 years copared with 7%, 22%, and 27% at, 3, and 5 years for those without an affected sibling. For those with GSW, recurrence risks were 5% at year and 58% at 5 years copared with 9%, 2%, and 26% at,3, and 5 years for those without. For those with a history of prior acute syptoatic seizures (priarily febrile seizures), risk was % at year and 39% B ) E Q) i, & P 4 rl 7 B C E a ) & PI P.d, rl 5 V I GSW EEG pattern (N=3) - other EEG pattern <N=36) t I... JI ~~ I ' I.,., ~, ~, I I. 2 ll Prior Acute (N=ll) - NO Prior Acute (N=38) o :'. I. I.,.,. I., Figure 3. (A) Recurrence aong patients with idiopathic seizures with and without an affected sibling. (B) Recurrence aong patients with idiopathic seizures with and without GS W EEG pattern. (C) Recurrence aong patients with idiopathic seizures with and without a prior acute event. August 99 NEROLOGY 4 66

4 I A a! fi 2 +I Q V & Q P 4 rl? B V 8 J Kaplan-Meier Product Liit Analysis t Todd's paresis (N=2) - No Todd's paresis (N=47) I. I. I. 8. I n, I. I B, I t 8" 4,I... I 4/ il... - Prior Acute (N=2) - No Prior Acute (N=47).,.,.,.,.,. t Figure 4. (A) Recurrence aong patient with reote syptoatic seizures with and without Todd's paresis. (B) Recurrence aong patients with reote syptoatic seizures with and without a prior acute event. (C) Recurrence aong patients with reote syptoatic seizures with and without ultiple seizures or status epilepticus. at 5 years copared with lo%, 23%, and 27% at,3, and 5 years for those without. Possible risk factors found not to be significant predictors of recurrence for those with a st idiopathic seizure included abnoral neurologic exaination, abnoral EEG other than GSW, status epilepticus or ultiple seizures, age, gender, seizure type, Todd's paresis, history of epilepsy in parents, and history of acute syptoatic seizures in any st-degree relative. Reote syptoatic cases. Aong reote syptoatic cases, significantly increased recurrence risk was observed in those with Todd's paresis (figure 4A), subjects with prior acute syptoatic seizures (figure 4B), and subjects whose index episode was characterized by status epilepticus or ultiple seizures (figure 4C). Recurrence risk after the index episode for those with Todd's paresis was 4% at year and 75% at 5 years copared with 22%, 3%, and 39% at,3, and 5 years for those without. For those with a prior acute seizure, the risk was 6% at year and 8% at 5 years copared with 7%, 33%, and 4% at,3, and 5 years for those without. In those with status epilepticus or ultiple seizures, recurrence risk was 37% at year and 56% at 3 and 5 years copared with 2%, 34%, and 43% at,3, and 5 years for those without. Aong subjects with reote syptoatic seizures, no significant increased risk of recurrence was associated with having a st-degree relative with epilepsy or seizures, abnoral neurologic exaination, seizure type, age, EEG finding, or gender. Rate ratios for recurrence. The univariate rate ratios for the total group, for those with idiopathic seizures, and for reote syptoatic seizures are presented separately (table ). For the total group, factors associated with a significantly increased rate ratio for recurrence included etiology, Todd's paresis, having an affected sibling, a prior acute seizure, ultiple seizures/status epilepticus, and the presence of a GSW EEG pattern. For the idiopathic group, a GSW EEG pattern, a history of an acute syptoatic seizure, or an affected sibling were again significant predictors of increased recurrence risk. The rate ratio was between 2 and 3 for each. The rate ratios were increased (although not significantly) for those with Todd's paresis, ultiple seizures/status, and an abnoral neurologic exaination. Sex, age, and seizure type were not associated with increased rate ratios. For those in the reote syptoatic group, Todd's paresis, ultiple seizures/status, and a history of an acute syptoatic seizure were associated with significantly elevated rate ratios. Rate ratios were elevated (but not significantly) for those with a sibling with epilepsy or an abnoral neurologic exaination. Age, sex, and seizure type were not associated with increased rate ratios. When all factors significant in the univariate analysis were entered into a ultivariate Cox analysis (table l), the sae predictors, with the exception of ultiple seizures/status epilepticus, reained significant for the total group, although the agnitudes of the effects were slightly altered. To evaluate possible interaction with duration of follow-up, rate ratios for recurrence were separately 66 NEROLOGY 4 August 99

5 Table. nivariate rate ratios and ultivariate rate ratios (95% confidence intervals) as estiated by Cox regression analysis Variable in odel Total group (N = 28) nivariate Multivariate Idiopathic (N = 49) nivariate Multivariate Reote (N = 59) nivariate Multivariate Etiology (reote syptoatic) Todd s paresis Sibling affected Multiple seizures/ status epilepticus Abnoral neuro exa Seizure type (partial vs generalized) Prior acute seizures GSW EEG pattern.95* (.27, 2.99).94t (.2,3.3) 2.29* (.3,4.3).797 (.2, 2.86). (.64,.58) 2.55t (.44,4.5).94t (.8,4.5).99t (.9,3.6).47 (.87,2.47).72 (.96,3.).98.4 (.84,.5) (.8,.23) 3.33* 2.69* (2.2, 5.49) (.56,4.62).93t 2.67 (., 3.73) (.7,4.38).27 (.64,2.53) (.25,4.9).26 (.64,2.5).62 (.68,3.85).5 (.85,.29).4 (.69,2.89) 2.5t (.23,5.).43 (.69, 2.96).35 (.56, 3.27).9 (.87,.37) 2.9t.7 (.5,4.53) (.78,3.7) 2.84t 2.69t (.43,5.66) (.28,5.67) 3.3* (.64,6.65).97 (.7,5.43) 2.9t (.5,4.6) 2.9 (.9,4.2).96 (.82,.22) 2.93* (.3, 6.54).95 (.3, 3.2).9 (.49, 2.89) 2. (.98, 4.46). (.77,.29) 4.82* 4.75* (2.29,.4) (.82, 2.42) $ $ * p 5.. t p 5.5. $ Failure to converge in likelihood due to sall nuber of cases in a subset of that category. ( ) 95% confidence intervals. evaluated for recurrence in the st year and for recurrence after that tie. These analyses revealed little change in the ordering of rate ratios for individual factors, although the agnitude of the effect for individual factors was changed. Treatent with continuous AED was recoended for 8% of cases. Medication could not be shown to favorably affect recurrence rate ratios and was in fact associated with a odest but nonsignificant increase in risk. Modeling. We have used the rate ratios derived fro the ultivariate analysis to odel predicted recurrence in idiopathic and reote syptoatic cases (table 2). For an idiopathic case without an affected sibling, with no history of acute syptoatic seizures, no status or ultiple seizures at presentation, no evidence of Todd s paresis, and without a GSW EEG pattern (ie, baseline), the predicted risk for recurrence by 5 years following the initial episode is 2.6%. The predicted 5-year risk increases to 45% for those with an affected sibling, 48% with a GSW EEG pattern, and 34.6% with a history of acute seizure. For the reote syptoatic cases with no other identified factors, observed recurrence risk at 4 years was 36.9% copared with a predicted risk of 53.3% for those with status/ultiple seizures, 83.8% for those with Todd s, and 95.3% with prior acute syptoatic seizures. Discussion. In the present study we find a odest overall risk for recurrence following a st seizure: 34% at 5 years. Studies since 98 have reported seizure recurrence following st seizure fro 3% to 7%5-2 (table 3). Such variance is not surprising, since basic study design, characteristics of populations studied, ean duration of follow-up, and ethods of statistical analysis differ substantially. Soe of these differences are discussed below. Referral versus general population studies. A st seizure in the nited States is seldo a reason for referral to a specialist; rather, initial evaluation is undertaken by the pediatrician, general practitioner, internist, or eergency roo physician. Only cases perceived as being ore coplex will have initial referral to a specialty clinic. Since factors associated with referral ay also be factors that predict seizure recurrence, studies dealing with referral populations start with a unique subgroup of st seizure patients who ay be at exceptionally high risk for recurrence. In the present study, which identified cases at the initial point of edical contact, 25% of idiopathic cases without any risk factor experienced a recurrence at 5 years following a st seizure. Five-year recurrence risk was considerably higher for other subgroups, reaching 8% in those with a reote syptoatic st seizure and a history of acute syptoatic seizures (table 2). Thus, wide variation in total recurrence risk ight be seen depending on the distribution of risk factors in a study population. Siilar referral bias ay apply to studies depending upon referral to EEG laboratories.2 Physicians ay not wish to put patients through the cost or inconvenience of the testing or the patient ay not keep an appointent even when scheduled. Thus, in the nited Kingdo, where socialized edicine progras reove financial ipedient to referral for EEG evaluation August 99 NEROLOGY 4 67

6 Table 2. Observed and expected rates of recurrence in subgroups based upon univariate rate ratios as applied to observed baseline rates of recurrence Risk factors Baseline* (N = 78) Idiopathic with an affected sibling (N = ) Idiopathic with a GSW EEG pattern (N = ) Idiopathic with prior acute seizures (N = 7) Idiopathic with abnoral exaination (N = 3) Idiopathic with abnoral exaination and or ore other features (N = 26) Two or ore features with noral exaination (N = 5) Reote syptoatic with no other significant factors (N = 32) Reote syptoatic with Todd s paresis (N = 4) Reote syptoatic with prior acute syptoatic seizures (N = 3) Reote syptoatic with ultiple seizures or status epilepticus (N = 8) Two or ore factors (N = 2) Months of follow-up % 2.% t (5.3%).% (6.5%).% (.%) 9.3% (.%) 4.3% 3.% 2.% (3.4%) 55.% (32.5%) 4.3% (22.5%) 5.4% (8.6%) 4.3% 6.7% 3.4% (38.3%) 55.% (4.7%) 28.6% (28.7%) 2.3% (23.9%) 22.7% 9.% 3.4% (4.4%) 55.% (43.9%) 28.6% (3.3%) 23.% (27.2%) 22.7% 4.% 4.% 7.% 7.% 5.9% (9.7%).% (48.8%).% (67.6%) 25.% (24.4%) 4.7% 5.9% (37.9%) 25.% (67.5%).% (89.4%) 37.5% (37.4%) 75.% 24.8% (47.%) 5.% (79.%).% (92.8%) 37.5% (47.8%) 75.% 36.9% (5.2%) 5.% (83.8%).% (95.3%) 37.5% (53.3%) 75.% 6 2.6% 45.% (45.5%) 55.% (48.2%) 52.4% (34.6%) 25.8% (29.5%) 22.7% 7.% 36.9% (54.%) * Baseline is defined by idiopathic etiology with noral exaination, no affected siblings, no GSW EEG pattern, no status epilepticus, no Todd s paresis, no prior acute events, and only a single seizure during the initial seizure episode. t Observed recurrence with that expected in parentheses below each value. (but ay introduce other difficulties such as inordinate delay in scheduling the test), only 74% of patients with seizures or epilepsy identified through hospital records and files of general practitioners had an EEG tracing at any tie, uch less at tie of initial diagnosis.22 In a study in the nited States, failure to keep a scheduled EEG appointent was associated with lower recurrence risk.2 Case definition. Risk estiates will vary, depending on the sensitivity and specificity of the inclusion criteria. Our case definition (see Methods) is highly specific (all cases identified are indeed st seizures) but not necessarily sensitive (st seizures not coing to an eergency roo or not witnessed, or not interviewed within 3 days are excluded). Misclassification. Most individuals with newly ascertained unprovoked seizures will have experienced ultiple seizure episodes by the tie of st edical evaluation (74% in the series fro which the present cohort was derived). Studies that rely on histories not specifically addressing the question of prior seizure occurrence (such as record reviews) ay include a substantial proportion of cases identified as st seizures that are, in fact, recurrent. In reviewing a saple of hospital records of recurrent seizure cases fro which the present series was derived, 7% (all with generalized tonic clonic seizures) were labeled as st seizures by house staff and neurology 68 NEROLOGY 4 August 99 attending physicians but had actually experienced prior unprovoked (usually coplex partial) seizures. This history of prior seizures was identified only through detailed questionnaires adinistered by trained interviewers. Since seizure recurrence aong those with 2 or ore unprovoked seizures is 7% or inclusion of such cases would have substantially raised the estiate of risk for recurrence. Tie of ascertainent. Even when groups of patients are followed prospectively, recurrence ay be deterined either fro tie of st visit or tie of st seizure. The invariable lag between the occurrence of a seizure and referral or visit to a referral center eans that soe patients will have experienced recurrence prior to being seen and thus will be ineligible for studies dating recurrence fro st visit. Exclusion of cases with early recurrence ay underestiate the recurrence risk. Hopkins et all8 found a short interval between st seizure and st visit to the referral center to be a predictor of recurrence. Alternatively, studies of referral population which backdate follow-up to tie of st seizure ay overestiate recurrences since seizure recurrence ay be a ajor factor in a patient s decision to keep an appointent. Retrospective ascertainent such as in The Collaborative Perinatal Project (prospective for prenatal and perinatal factors, but retrospective for seizure occurrence and recurrence) is even ore likely to overesti-

7 Table 3. Seizure recurrence after a st unprovoked seizure: Suary of results of studies published since 98 Interi fro Study Source Recurrence Seizure st seizure to Seizure Analytic Study design ofclass Population risk type study entry veriacntion ethods Risk factors Elweal5? Specialty Adulta 7% at3p Generalized Within24hrs Patient Actuarial - clinic interview referrals AnnegerslG Historical Medical Allsgea 48%at3yrs Atypes Within24hra Chart Actuarial EEGetiology, Cohort records- review proportional sz type, eergency hezarda abnoral hospital new exa adission or outpatient Hirtzl'l Historical Annual Children 69Wbyage7 Alltypes ptolyr Parent Proportions Sztype, cohort intemew underage7 (include9 (at tie of interview neonatal sz acute annual syptoatic) history pdate) Hopkin88 Prospective Specialty Adults 5256at3yrs Atypes clinic referral Clelandlg Retrospective Specialty Adults 39% - crude All types clinic referral Shinw Pspwtive Eergency Children 29%at2p Alltypes (P to 8) contads Nonrielr factors Proportion treated None Sex,age,AED 6% Febrile sz, 27% neurologic exa, faily history of afebrile sz, AED pto8wks Patient Actuarial Intervalto Sztype, "Occasional" interview referral, tie age, CT, EEG?, ofday, AED faily history?, febrile sz? p to 6% wks Patient interview Proportions EEG nknown At tie of Patient Actuarial Faily history Age, sex, status 5% 82 (parent) proportional of afebrile sz, epilepticus interview hazards EEG, etiology, acute sz, 82 type Cafield2 Prospective EEGlab Childre 5%-crude Alltypes referrals Preeent Prospective Eergency Allagea 3% at3yrs Alltypes study contacts sz Seizure(s). AED Preacription of antiepileptic drug. nknown Chart review Proportions EEG, abnoral Wake, sleep, 68% new exa, febrile sz, 92 type AED Within 24 Patient or Actuarial, EEG, etiology, Age, ex, AED, 8% hrs of 82 parent proportional faily history 8z type interview hazards ofafebrile sz, acute syptoatic sz, Todd's paresis, statua epilepticus, ultiple BZ ate recurrence, since the ore seizures an individual experiences, the ore likely he or she is to coe to edical attention. Risk factors for recurrence. Despite ajor differences in overall recurrence risk reported fro various studies (perhaps due to differences elucidated above), soe consistency in factors associated with recurrence has been observed (table 3). In ost studies presenting data, recurrence was significantly higher aong cases eeting our criteria for reote syptoatic seizures when copared with idiopathic seizures.6.2 An abnoral EEG has been a consistent predictor of recurrence, although there has not been agreeent on the nature of the predictive EEG abnorality. In the present study, only a GSW pattern was associated with an increased recurrence risk. Other studies report only a focal EEG2 or any abnoral EEG6s2 to be predictive of recurrence. We found that a history of acute syptoatic seizure was associated with an increased risk for recur- rence. Elevated risks for children2 and possibly adults7 with prior febrile seizures have been reported. A sibling with epilepsy was predictive of recurrence in adult@ and children with an abnoral EEG.2 Only other study has analyzed status epilepticus as a risk factor, and, unlike the present study, this factor was not predictive of recurrence.2 No other studies have reported Todd's paresis to be a predictor. This finding is independent of neurologic exaination. nlike the present study and the adult series of Hopkins et a,8 a partial seizure has been a predictor of recurrence in studies dealing exclusively with children7,2,2 and in the study of Annegers et al.6 Although we included subjects of all ages, only 23% of cases were under age 2 and ost of these were teenagers. It is possible that children are different fro adults in this respect. We do not find an abnoral neurologic exaination to be a significant predictor of recurrence, although the point estiates of the risk are siilar to those reported by Annegers et a.6 In agreeent with August 99 NEROLOGY 4 69

8 several studies, neither age at st seizure nor sex were predictors.6j8,2 Although no studies published heretofor have been designed specifically to evaluate AED therapy, soe analyses of the factor have invariably been attepted. No study has shown treatent to reduce recurrence risks, and in soe, such as the present study, prescription of AED was associated with an increased recurrence risk even when controlling for other risk factors. This does not ean that edications are ineffective in the prevention of seizures: the data available are inadequate to deterine whether the results are real or artifacts attributable to irregular edication use, inadequate or inappropriate AED, or AED withdrawal. The proper edication in the proper dose taken on a consistent basis is necessary to confir or deny efficacy. In general, patients in the present study given AED were placed on low dosages, levels were not onitored or adjusted, and any did not take the edication on a regular basis if at all. The question of AED effect can only be addressed in a randoized clinical trial with careful onitoring of AED use in treated patients.24 References. Hauser WA. Should people be treated after a first seizure? Arch Neurol986;43: Hart RG, Easton JD. Seizure recurrence after a first, unprovoked seizure. Arch Neurol 986;43: Hauser WA, Anderson VE, Loewenson RB, McRoberts SM. Seizure recurrence after a first unprovoked seizure. N Engl J Med 982;37: Ewles RDC, Reynolds EH. Should people be treated after a first seizure? Arch Neurol988; Coission on Classification and Terinology of the International League Against Epilepsy. Proposal for revised clinical and electroencephalographic classification of epileptic seizures. Epilepsia 98; Annegers JF, Grabow JD, Groover RV, Laws ER, Elveback LR, Kurland LT. Seizures after head traua: a population study. Neurology 98; Hauser WA, Rairez-Lassepas M, Rosenstein R. Risk for seizures and epilepsy following cerebrovascular insults. Abstract. Epilepsia 984; Annegers JF, Nicolosi A, Beghi E, Hauser WA, Kurland LT. The risk of unprovoked seizures after encephalitis and eningitis. Neurology 988; Benedetti MD, Shinnar S, Cohen H, Inbar D, Hauser WA. Risk factors for epilepsy in children with cerebral palsy and/or ental retardation. Abstract. Epilepsia 986;27:64.. Nelson KB, Ellenberg JH. Antecedents of seizure disorders in early childhood. A J Dis Child 986;4: Cutler SJ, Ederer F. Maxiu utilization of the life table ethod in analyzing survival. J Chronic Dis 958; Elandt-Johnson RC, Johnson NL. Survival odels and data analysis. New York John Wiley & Sons, Kaplan EL, Meier P. Nonparaetric estiations fro incoplete observation. J A Stat Assoc 958;53: Cox DR. Regression odels and life-tables. J R Stat SOC [B] 972; Elwes RDC, Chesteran P, Reynolds EH. Prognosis after a first untreated tonic-clonic seizure. Lancet 985;2: Annegers JF, Shirts SB, Hauser WA, Kurland LT. Risk of recurrence after an initial unprovoked seizure. Epilepsia 986;27: Hirtz DB, Ellenberg JH, Nelson KB. The risk of recurrence of nonfebrile seizures in children. Neurology 984;34: Hopkins A, Garan A, Clarke C. The first seizure in adult life. Lancet 988;: Cleland PG, Mosbue RJ, Steward WP, Fosta JB. Prognosis of isolated seizures in adult life. Br Med J 98;283: Shinnar S, Berg D, Moshe SL, et al. The risk of seizure recurrence following a first unprovoked seizure in childhood a prospective study. Pediatrics 99;85: Cafield PR, Cafield CS, Dooley JM, Tibbles JAR, Fung T, Garner B. Epilepsy after a first unprovoked seizure in childhood. Neurology 985; Goodridge DMG, Shorvon SD. Epileptic seizures in a population of 6. I. Deography, diagnosis and classification, and role of the hospital service. Br Med J 983;287: Hauser WA, Rich SS, Jacobs MP, Anderson VE. Patterns of seizure occurrence and recurrence risks in patients with newly diagnosed epilepsy. Abstract. Epilepsia 983; Mussico M. The effect of drug treatent on risk of recurrence after a first tonic clonic seizure: an Italian ulticenter randoized trial. Abstract. Neurology 989;39(suppl):48. 7 NEROLOGY 4 August 99

9 Seizure recurrence after a st unprovoked seizure: An extended follow up W. Allen Hauser, Stephen S. Rich, John F. Annegers, et al. Neurology 99;4;63 DOI.22/WNL This inforation is current as of August, 99 pdated Inforation & Services Citations Perissions & Licensing Reprints including high resolution figures, can be found at: This article has been cited by 8 HighWire-hosted articles: Inforation about reproducing this article in parts (figures,tables) or in its entirety can be found online at: Inforation about ordering reprints can be found online: Neurology is the official journal of the Aerican Acadey of Neurology. Published continuously since 95, it is now a weekly with 48 issues per year. Copyright 99 by Edgell Counications, Inc.. All rights reserved. Print ISSN: Online ISSN: X.