- Email: [email protected]
Biologic factors as predictors of social outcome of epilepsy in intellectually normal children: A population-based study
Biologic factors as predictors of social outcome of epilepsy in intellectually normal children: A population-based study Carol Camfield, MD, Peter Camfield, MD, Bruce Smith, PhD, Kevin Gordon, MD, and Joseph Dooley, MB From the Departments of Pediatrics and Mathematics, Dalhousie Universityand [zaak Walton Killam Children's HospitaL Halifax, Nova Scotia, Canada
We stUdied social outcome for all the normally intelligent children in our province with onset of epilepsy between 1977 and 1985 (excluding absence and "minor motor" seizures). After follow-up averaging 7112 years, the 337 patients were 7 to 28 years of age. Outcome measures were age dependent. Of those old enough to be at risk, the percentage with each unfavorable outcome was as follows: school failure 34%, use of special educational resources 34%, mental health consultation 22%, psychotropic medication 5%, unemployment 20%, social isolation 27%, inadvertent pregnancy 12%, and criminal conviction 2%. In a multivariate model correcting for number of potential unfavorable outcomes (based on age at end of follow-up), many variables related to epilepsy, seizure control, and electroencephalographic findings were not associated with social outcome. Only two variables were associated with at least one unfavorable outcome-learning disorder (p <0.001) and more than 21 seizures before treatment was begun (p <0.03). The only variable with no unfavorable outcome was simple partial seizures (p <0.003). Sensitivity and specificity of this model were 54% and 68%, respectively, indicating that social outcome for these children was often not related to biologiC factors reflected by the medical details and clinical course of their disorder. (J PEDIATR 1993;122:869-73)
Epilepsy is a chronic illness with important effectson social adjustment.' These social problems are often attributed to the biologicfactors of epilepsy,including seizure type,2 seizure severity and frequency,' concomitant cognitive deficits," and side effects of long-term daily medication.! However, social adjustment is also influenced by social factors that are more difficult to measure and that vary from one society to another.f These factors include the stigmatization of chronic illness, exclusion from some social activities, and prejudice toward peoplewith epilepsy." There is controversy about which set of factors is most important. Submitted for publication Oct. 7,1992; accepted Feb. 12, 1993. Reprint requests: Carol Camfield, MD, Izaak Walton Killam Children's Hospital, Box 3070, Halifax, Nova Scotia B3H 3G9, Canada. Copyright @ 1993 by Mosby-Year Book, Inc. 0022-3476/93/$1.00 + .10 9/20/46463
We identified a completecohort of children with normal intelligencein whomepilepsy developed in ourprovince, and we followed them to varying degrees of social maturity, Their seizure outcome will be reported separately. In this study we addressed the question of the degree to which biologic features of epilepsy are related to social adjustment See related article, p. 861.
problems. Only children with normal intelligence are considered, because mental retardation may have an independent effect on social outcome.! METHODS Our previous studieshaveindicatedthat ifa child in Nova Scotia (population 850,000) is seen by a physician because of a history of unprovoked seizures, an electroencephalo869
870
Camfield et al.
The Journal of Pediatrics June 1993
TobIe I. Seizure characteristics
Seizure type Generalizedtonic clonic Simple partial Complex partial Focal, secondarily generalized Rolandic Other No. of seizures beforetreatment <21 >21 Previous febrile seizures Yes No
No.
%
88 15 95 110 28
26.1 4.5 28.2 32.6 8.3
I
280 57
83.1 16.9
46 291
13.6 86.4
gram will be requested." We interpret virtually all pediatric electroencephalograms in the province, so these reports provide a comprehensive case-finding method for children with epilepsy. By reviewing all these reports, and then by conducting a detailed chart and clinical patient review, we were able to identify children (aged I month to 16 years) in whom epilepsy developed between 1977 and 1985 in our province. In 1989 and 1990, children were then followed up in person or by telephone. We personally examined all these patients at some time during their clinical course. To be included, a child must have had at least two afebrile, unprovoked seizures while resident in Nova Scotia during the study period. Only children who were at least 7 years of age at follow-up were included in this study. Patients were included only if they were judged by the attending pediatric neurologist to have normal intelligence at the time of initial diagnosis. At the time of the diagnosis of epilepsy, the presence of a learning disorder was determined by the attending neurologist, who used the clinical data available, including school reports and psychometric testing. Because of other protocol considerations related to seizure outcome, patients with "minor motor" and absence seizures were excluded, as were those with evidence of progressive brain disease. We studied biologic features of epilepsy at the time of diagnosis, during the first year of treatment, and at the end of follow-up. At the time of diagnosis these variables were as follows: age at first seizure, time from first to second seizure, number of seizures before daily medication was initiated, seizure type, cause, previous neonatal or febrile seizures, learning disorder (mental retardation excluded), and neurologic deficits. Initial encephalographic data were scored by background abnormalities, focal slowing, and the presence of focal or generalized spikes. During treatment the variables were as follows: seizure control during the first and second 6-month intervals of treatment, number and
type of medications used, and change in seizure type. At the end of follow-up, patients were considered to be in remission only if they were seizure free and not taking medication. Social outcome was assessed at the end of follow-up. We documented eight adverse social outcomes occurring after the diagnosis of epilepsy. These included (1) school grades repeated or need for special education, including special class or resource teaching, (2) use of psychotropic medication (including methylphenidate), (3) behavior problems sufficient to require a mental health consultation with psychiatric or psychologic services, (4) for those older than 14 years, conviction of a crime, and (5) for girls older than 14 years of age, inadvertent pregnancy. All patients older than 18 years of age were questioned about (6) social isolation, (7) lack of a self-generated source of financial support if no longer a full-time student, and (8) unemployment. Social isolation was noted when, in our judgment, the patient had no consistent relationships outside of his or her immediate family. Thus patients older than 18 years of age at the end of follow-up were eligible for eight adverse social outcomes. For younger subjects, there were fewer potential unfavorable outcomes but never less than three. In the data analysis for each outcome, only the patients eligible for that particular outcome were considered. Patients were divided into those with good social outcome, defined as no adverse factors, or bad social outcome, defined as one or more of the eight adverse factors listed above. Stepwise logistic regression analysis was used to develop a multivariate model based on 337 cases for which there were no missing data for the biologic seizure-related variables listed above. Estimated regression coefficients were used to develop a predictive scoring scheme. A cutoff point was determined to minimize classification errors. The data were also analyzed with a tree model. This procedure constructs the optimal binary classification tree for the data; "optimal" is taken to mean the simplest classification with the smallest number of misclassified cases. The number of branches in the tree structure (terminal nodes) was determined by 10-fold cross-validation.l'' RESULTS During the study period, 504 children were determined to have epilepsy. Only two were lost to follow-up. This report is based on 170 boys and 167 girls eligible for at least three adverse social outcomes at the end of follow-up. Details of their seizure disorders are given in Table 1. Average age at first seizure was 90 months ± 49 SD (range 1 to 192 months). Average age at the end of follow-up was 180 months ± 57 SD (range 84 to 300 months). Total duration of follow-up averaged 90 months ± 30 SD (range 30 to 184 months). Overall, 140 patients had a bad outcome and 197 had a
The Journal of Pediatrics Volume 122, N umber 6
good outcome (Table II ). Social outc ome was not rela ted to remission of epilepsy at the end of follow-up. Virt ually all chi ldren with several seizure-free yea rs had an opportunity to discontinue medication and th erefore fulfill our definition of remission. In fact, fewer than 4% of patients had >4 seizure-free years wit hout an attempt to stop medication. Sixty-seven per cent of chi ldren with good social outcome were in remission, compared with 58% of children with bad social outc ome (p = not significan t). In the multiv ari ate analysis, the signific ant pred ictors of bad outco me wer e learning disorder (p <0 .0001), > 21 seizures before treatment with medication (p >0 .03), and number of potential adverse outcomes (p <0.01) . Good outcome was predicted by seizure type "s imple partial" (p < 0 .008) (Ta ble III). F orty-three per cent of our pati ents with pa rtial complex seizures ha d >20 seizures befor e treatment. Nonetheless , some patients with other seizure types had large numbers of seizures before treatment; only 55% of those with >20 seizure s before tr eatment h ad partial complex seizures. Wi th the use of the regressio n coefficients, a relatively inaccurate predictive sch eme for each pa tient was given by the following equation : (0 .179 X potential num ber of adver se outcomes) + (2.423 if lea rning d isorder ) - (1.708 if simple partial seiz ure type) + (0.707 if >20 seizures before treatment) - 1.0145. If the sum of the equation was >0.0595, bad outcome was predicted. If the su m was <0.0596, good outcome was predicted. This scheme had a total misclassification rate of 106 in 337 (sensitivity 54%; specificity 69%). T he residual deviance of the logistic model wa s 386.7, with 332 degrees of freedom . A goodness-of-fit hypothesis was rejected (p = 0.02 1), indicating substantial va riation in the data that was not explained by the logistic model. The results of the tree model were very similar; the identical variables were identi fied as significant, and the overall misclassification rate was the same. Both the stepwise logistic regression and the tree model identified learning disorder as the most importan t predictive variable, regardless of the status of other predictors. When patients with a learn ing disorder were excluded, a stepwis e logistic model again foun d simple partial seizures (p <0.0001) , number of seizu res >21 before initiation of treatment (p <0. 05), and potential number of unfavorable outcomes (p < 0.024) to be the only significa nt predictors.
DIS C US SIO N T he major finding of our study is that most biologic factors related to the treatment and control of epilepsy in children with normal intelligence do not predict a variety of undesirable social outcomes. Even remiss ion of epilepsy was not pre dictive of social outcome.
Camfield et al.
871
Table II. Psychosocial outcome
Category 1. Grades repeated or
2. 3.
4. 5. 6. 7. 8.
special class Psychiatric or psychologic referral Psychotropic drugs Criminal offense Inadvertent pregnancy Social isolation Financial dependency Unemployment
Eligible sub jects'
Affected sub jects (No.)
No.
%
(95% CI)
101
290
35
(29. 40)
72
316
23
(18,28)
17
5
11
316 187 88
13
(3, 8) (0, 4) (5, 20)
11 21
71 71
16 30
(7, 24) (19, 40)
21
71
30
(19, 40)
4
2
CI, Confidence intervul, "Nu mber sufficiently old where exact information wasknown. Categories I to 3 requiredageat end of follow-up >6.9 years. ca tegory 4 age> 13.9 years, category 5 age > 13.9 years and female, categories 6 to 8 age >17.9 years.
Table 1/1. Variab les predictive of unfavorable outcomes: multivariate model Var iable
Coetrlclent
1. Learning disorder 2. Partial elementary
2.423 -1.708
seizure type 3. Potential number of adverse outcomes 4. Seizures before treatment: > 20
95%CI
p
(1.574, 3.272) <0.0001 (-3 .048, - 0.368) <0 .008
0.179
(0.0436, 0.3144) <0 .01
0,707
(0.031, 1.383)
<0.03
C/, Confidence interval.
We are aware of only two previous population-based stu dies of social success in patients with epilepsy. 8; I I Trostle et al.'! administered the Washington Psychosocial Inventory to nea rly all normally int elligent adults with epilepsy in R ochest er, Minn . They emphasized the importance of studying an entire population and concluded that rates of social disability in those with epilepsy had been overes timated in reports based on referred patients; nevertheless, 19% were noted to have serious social difficulties. Sillanpaa! followed a regional cohort of 178 childrenwith epilepsy for 30 years and noted a striking effect of mental retardation on social outcome. Poor seizur e control and certain seizure types were independen t predictors of some aspects of social outcome. It is difficult to compare his exp erience with ours because there was no overall score for social outcome and because pat ients with norma l intelligence were not considered separately. In our cohort of normally inte1ligent children, the stro ngest predictor of poor social outcome was the presence of a
872
Camfield et al.
learning disorder at the time of diagnosis of epilepsy. This is not surprising because a learning disorder without epilepsy influences social success.P Only two other biologic factors directly related to epilepsy were independently predictive of outcome-more than 20seizures before treatment and the "simple partial" type of seizure. At the end of our follow-up study, young adults were eligible for a greater number of bad outcomes than were younger children. This explains the reason that the number of potential bad outcomes also entered our multivariate model. In other words, the greater the number of possible bad outcomes, the more likely it is that a child will have at least one. The good social outcome for patients with simple partial seizures has not previously been emphasized. Simple partial seizures are relatively uncommon. In one study of 34 adolescents with epilepsy, the two with simple partial seizures had good social success.'! Of all seizure types, the simple partial type involves the smallest amount of cortex, raising the possibility that a broader spread of epileptic discharge through other cortical structures may somehow alter social success. Our data do not allow a direct test of this hypothesis. In contrast to some previous reports, we did not find complex partial seizures to be particularly associated with bad outcome.' It is possible that part of the predictive value of >20 seizures before treatment may be related to partial complex seizures-a seizure type that is often difficult for families to recognize. Nonetheless, more than half of the children with >20 seizures before treatment had other seizure types. Therefore the prognostic value of social outcome for >20 seizures before treatment could be only partly attributed to complex partial seizures. In a previous population-based study of temporal lobe epilepsy, we noted that intellectual ability was the best predictor of social adjustment.!" It is possible that allowing >20 seizures to occur before the diagnosis of epilepsy reflects a chaotic or poorly educated family-features that might alter adjustment to epilepsy but might have an equally negative effect on normal children. By defining bad outcome as one or more adverse social outcomes, we maximized the opportunity for biologic factors to be predictive. We were impressed by the high proportion of patients with at least one of the eight adverse outcomes. In this population-based study (without referral bias), 42% of patients had at least one bad social outcome. We believe that this degree of social disability is beyond what would be expected in a normal, nonepileptic population-a speculation that requires a matched control group for proof. Our study was not designed to establish that children with epilepsy have high rates of social disability; however, we are aware of some information for our province that supports
The Journal of Pediatrics June 1993
this hypothesis. During the study period approximately 11.7% of all children with normal intelligence had resource or special education services at school (personal communication, Nova Scotia Department of Education, 1993), whereas 35% of our patients had such services. Rates of inadvertent pregnancy are not easily obtained. The total rate of pregnancy, both planned and unplanned, for female adolescents 15 to 19 years of age in our province during the study period was about 4/1000 women per year (excluding spontaneous abortion and miscarriagej.P In the same age range, 11 (12.5%) of our 88 patients reported inadvertent pregnancy. We did not record howmany others had planned pregnancies. Twenty-three percent of our patients had been referred to a mental health facility. The rate of such referrals for the overall population of Nova Scotia is not known. However, in Ontario, which has more mental health services for children, it is estimated that 12% of all children 4 to 16 years of age have a mental health referral at some time." Thirty percent of our patients> 18 years of age were unemployed, if no longer in school. In 1988 in Nova Scotia, 17% of male residents and 14% of female residents aged 19 to 24 years were unemployed (Statistics Canada, 1993 [catalog No. 71-001; Labor Force]). Our data do not provide an explanation for the high frequency of poor social outcomes. If it is not related to biologic factors, it presumably is the nonspecific result of a chronic illness, a generic effect of epilepsy, or the result of unmeasured sociologic factors. A recent study addressed the issue of unsatisfactory educational achievement in children with epilepsy and concluded that social issues related to the home environment were important determinants, whereas biologic factors were not.!? Another recent study of 183 children with epilepsy followed in a referral clinic found that social competence, as measured by the Achenbach Child Behavior Checklist, was independently related to the biologic factor of good seizure control and to the social factors of high parental income and a two-parent family.IS However, only 15% of the variance was accounted for by these factors in the statistical model. In general, poor social outcome in adults with epilepsy, whether defined informally or based on standardized rating scales, has not been well explained by biologic features." There is some suggestion that those with either severe seizures or poor control fare more poorly.v!' In several studies, seizure type and laterality, seizure control, medication, age at onset, and duration of illness did not predict social outcome.v 19,20 Concomitant brain problems may be more important determinants than epilepsy.v S Studies that have examined social factors in adults with epilepsy have also not been successful in predicting outcome. In one study, a battery of sociologic and psychologic tests was used to attempt to predict psychologic abnormal-
The Journal of Pediatrics Volume 122, Number 6
ities in adults with poorly controlled epilepsy who were undergoing video monitoring.l? A multiple regression model found no biologic predictors but did identify three social predictors-stressful life events, poor adjustment to epilepsy, and financial stress. These factors accounted for only 23% of the variance in the statistical model, indicating that there are other, more important influences on social outcome than those measured. The generic effect of chronic illness on social function appears to warrant further investigation.I' Only one small study has addressed this issue directly in children with epilepsy.22 Increased rates of social disabilities in children with temporal lobe seizures were similar to those in children with chronic asthma. If the chronic illness factor is proved to be important in the social outcome of epilepsy, interventions used in other chronic diseases might be useful for children with epilepsy. Our study would have been strengthened by direct measures of parental social class and indicators of social disruption. A standardized rating scale might have given different results from our simple clinical measures of social outcome, and a longer follow-up period might have allowed the biologic effects of seizure frequency and severity to have an effect. Nonetheless, our findings and those reported in the literature indicate that the cause of social adjustment problems in people with epilepsy is not well understood. We conclude that unsatisfactory social outcome is common in childhood epilepsy (excluding those with absence and "minor motor" seizures). For most children with epilepsy and normal intelligence, the clinician cannot accurately predict social outcome. Biologic features of the seizures and their control appear to have little relationship with social success. REFERENCES 1. Levin R, Banks 8, Berg B. Psychosocial dimensions of epilepsy: a review of the literature. Epilepsia 1988;29:805-16. 2. Bear DM, Fedio P. Quantitative analysis of interictal behaviour in temporal lobe epilepsy. Arch Neurol 1977;34:454-67. 3. Smith DF, Baker GA, Dewey M, Jacoby A, Chadwick DW. Seizure frequency, patient-perceived seizure severity and the psychosocial consequences of intractable epilepsy. Epilepsy Res 1991;9:231-41. 4. Beran RG, Flanagan PJ. Psychosocial sequelae of epilepsy: the role of associated cerebral pathology. Epilepsia 1987;28:10710.
Camfield et al.
873
5. Trimble MR, Thompson P1. Anticonvulsant drugs, cognitive function and behaviour. Epilepsia 1983;2(suppl 1):855-63. 6. Dodrill CB, Beier R, Kasparick M, Tacke I, Take D, SiangYan T. Psychosocial problems in adults with epilepsy: comparison findings from four countries. Epilepsia 1984;25:17683. 7. Herman BP, Whitman S. Behavioral and personality correlates of epilepsy: a review, methodological critique, and conceptual model. Psychol Bull 1984;95:451-97. 8. Silianpaa M. Children with epilepsy as adults. Acta Paediatr Scand 1990;368(suppl):5-75. 9. Camfield PR, Camfield CS, Dooley JM, Tibbles JA, Fung T, Garner B. Epilepsy after a first unprovoked seizure in childhood. Neurology 1985;35:1657-60. 10. Goldman L, Weinberg M, Weisberg M, et al. A computer-derived protocol to aid in the diagnosis of emergency room patients with acute chest pain. N Eng! J Med 1982;307:588-96. 11. Trostle JA, Hauser WA, Sharbrough FW. Psychologic and social adjustment in Rochester, Minnesota. Neurology 1989; 39:633-7. 12. Rutter M, Graham P, Yule W. A neuropsychiatric study in children. Clinics in Developmental Medicine 1970 (Nos. 35, 36):150-62, 175-85. London: Spastics International Medical Publishers. 13. Clement MJ, Wallace SJ. A survey of adolescents with epilepsy. Dev Med Child Neural 1990;32:849-50. 14. Camfield PR, Gates R, Ronen G, Camfield C, Ferguson A, MacDonald GW. Comparison of cognitive ability, personality profile, and school success in epileptic children with pure right versus left temporal EEG foci. Ann NeuroI1984;15:122-6. 15. LeBlanc JC. Epidemiology of STDs in Nova Scotia. Nova Scotia Medical Bulletin 1992;71:193-5. 16. Cadman D, Boyle M, Offord DR, et al. Chronic illnesses, medical conditions and physical limitations in Ontario children. Can Med Assoc J 1986;135:761-7. 17. Mitchell WG, Chevez JM, Lee H, Gregman BC. Academic underachievement in children with epilepsy. J Child Neurol 1991;6:65-72. 18. Hermann BP, Whitman S, Hughes JR, Melyn MM, Dell J. Multietiological determinants of psychopathology and social competence in children with epilepsy. Epilepsy Res 1988;2:5160. 19. HermanBP, Whitman 8, WylerAR, Anton MT, Vanderzwagg R. Psychosocial predictors of psychopathology in epilepsy. Br J Psychiatry 1990;156:98-105. 20. Collings JA. Psychosocial well-being and epilepsy: an empirical study. Epilepsia 1990;31:418-26. 21. Pless IB. Clinical assessment: physical and psychological functioning. Pediatr Clin North Am 1984;31:33-45. 22. APter A, Aviv A, Kaminer Y, Weizman A, Lerman P, Tyano 8. Behavioral profile and social competence in temporal lobe epilepsy of adolescence. JAm Acad Child Adolesc Psychiatry 1991;30:887-92.
We stUdied social outcome for all the normally intelligent children in our province with onset of epilepsy between 1977 and 1985 (excluding absence and "minor motor" seizures). After follow-up averaging 7112 years, the 337 patients were 7 to 28 years of age. Outcome measures were age dependent. Of those old enough to be at risk, the percentage with each unfavorable outcome was as follows: school failure 34%, use of special educational resources 34%, mental health consultation 22%, psychotropic medication 5%, unemployment 20%, social isolation 27%, inadvertent pregnancy 12%, and criminal conviction 2%. In a multivariate model correcting for number of potential unfavorable outcomes (based on age at end of follow-up), many variables related to epilepsy, seizure control, and electroencephalographic findings were not associated with social outcome. Only two variables were associated with at least one unfavorable outcome-learning disorder (p <0.001) and more than 21 seizures before treatment was begun (p <0.03). The only variable with no unfavorable outcome was simple partial seizures (p <0.003). Sensitivity and specificity of this model were 54% and 68%, respectively, indicating that social outcome for these children was often not related to biologiC factors reflected by the medical details and clinical course of their disorder. (J PEDIATR 1993;122:869-73)
Epilepsy is a chronic illness with important effectson social adjustment.' These social problems are often attributed to the biologicfactors of epilepsy,including seizure type,2 seizure severity and frequency,' concomitant cognitive deficits," and side effects of long-term daily medication.! However, social adjustment is also influenced by social factors that are more difficult to measure and that vary from one society to another.f These factors include the stigmatization of chronic illness, exclusion from some social activities, and prejudice toward peoplewith epilepsy." There is controversy about which set of factors is most important. Submitted for publication Oct. 7,1992; accepted Feb. 12, 1993. Reprint requests: Carol Camfield, MD, Izaak Walton Killam Children's Hospital, Box 3070, Halifax, Nova Scotia B3H 3G9, Canada. Copyright @ 1993 by Mosby-Year Book, Inc. 0022-3476/93/$1.00 + .10 9/20/46463
We identified a completecohort of children with normal intelligencein whomepilepsy developed in ourprovince, and we followed them to varying degrees of social maturity, Their seizure outcome will be reported separately. In this study we addressed the question of the degree to which biologic features of epilepsy are related to social adjustment See related article, p. 861.
problems. Only children with normal intelligence are considered, because mental retardation may have an independent effect on social outcome.! METHODS Our previous studieshaveindicatedthat ifa child in Nova Scotia (population 850,000) is seen by a physician because of a history of unprovoked seizures, an electroencephalo869
870
Camfield et al.
The Journal of Pediatrics June 1993
TobIe I. Seizure characteristics
Seizure type Generalizedtonic clonic Simple partial Complex partial Focal, secondarily generalized Rolandic Other No. of seizures beforetreatment <21 >21 Previous febrile seizures Yes No
No.
%
88 15 95 110 28
26.1 4.5 28.2 32.6 8.3
I
280 57
83.1 16.9
46 291
13.6 86.4
gram will be requested." We interpret virtually all pediatric electroencephalograms in the province, so these reports provide a comprehensive case-finding method for children with epilepsy. By reviewing all these reports, and then by conducting a detailed chart and clinical patient review, we were able to identify children (aged I month to 16 years) in whom epilepsy developed between 1977 and 1985 in our province. In 1989 and 1990, children were then followed up in person or by telephone. We personally examined all these patients at some time during their clinical course. To be included, a child must have had at least two afebrile, unprovoked seizures while resident in Nova Scotia during the study period. Only children who were at least 7 years of age at follow-up were included in this study. Patients were included only if they were judged by the attending pediatric neurologist to have normal intelligence at the time of initial diagnosis. At the time of the diagnosis of epilepsy, the presence of a learning disorder was determined by the attending neurologist, who used the clinical data available, including school reports and psychometric testing. Because of other protocol considerations related to seizure outcome, patients with "minor motor" and absence seizures were excluded, as were those with evidence of progressive brain disease. We studied biologic features of epilepsy at the time of diagnosis, during the first year of treatment, and at the end of follow-up. At the time of diagnosis these variables were as follows: age at first seizure, time from first to second seizure, number of seizures before daily medication was initiated, seizure type, cause, previous neonatal or febrile seizures, learning disorder (mental retardation excluded), and neurologic deficits. Initial encephalographic data were scored by background abnormalities, focal slowing, and the presence of focal or generalized spikes. During treatment the variables were as follows: seizure control during the first and second 6-month intervals of treatment, number and
type of medications used, and change in seizure type. At the end of follow-up, patients were considered to be in remission only if they were seizure free and not taking medication. Social outcome was assessed at the end of follow-up. We documented eight adverse social outcomes occurring after the diagnosis of epilepsy. These included (1) school grades repeated or need for special education, including special class or resource teaching, (2) use of psychotropic medication (including methylphenidate), (3) behavior problems sufficient to require a mental health consultation with psychiatric or psychologic services, (4) for those older than 14 years, conviction of a crime, and (5) for girls older than 14 years of age, inadvertent pregnancy. All patients older than 18 years of age were questioned about (6) social isolation, (7) lack of a self-generated source of financial support if no longer a full-time student, and (8) unemployment. Social isolation was noted when, in our judgment, the patient had no consistent relationships outside of his or her immediate family. Thus patients older than 18 years of age at the end of follow-up were eligible for eight adverse social outcomes. For younger subjects, there were fewer potential unfavorable outcomes but never less than three. In the data analysis for each outcome, only the patients eligible for that particular outcome were considered. Patients were divided into those with good social outcome, defined as no adverse factors, or bad social outcome, defined as one or more of the eight adverse factors listed above. Stepwise logistic regression analysis was used to develop a multivariate model based on 337 cases for which there were no missing data for the biologic seizure-related variables listed above. Estimated regression coefficients were used to develop a predictive scoring scheme. A cutoff point was determined to minimize classification errors. The data were also analyzed with a tree model. This procedure constructs the optimal binary classification tree for the data; "optimal" is taken to mean the simplest classification with the smallest number of misclassified cases. The number of branches in the tree structure (terminal nodes) was determined by 10-fold cross-validation.l'' RESULTS During the study period, 504 children were determined to have epilepsy. Only two were lost to follow-up. This report is based on 170 boys and 167 girls eligible for at least three adverse social outcomes at the end of follow-up. Details of their seizure disorders are given in Table 1. Average age at first seizure was 90 months ± 49 SD (range 1 to 192 months). Average age at the end of follow-up was 180 months ± 57 SD (range 84 to 300 months). Total duration of follow-up averaged 90 months ± 30 SD (range 30 to 184 months). Overall, 140 patients had a bad outcome and 197 had a
The Journal of Pediatrics Volume 122, N umber 6
good outcome (Table II ). Social outc ome was not rela ted to remission of epilepsy at the end of follow-up. Virt ually all chi ldren with several seizure-free yea rs had an opportunity to discontinue medication and th erefore fulfill our definition of remission. In fact, fewer than 4% of patients had >4 seizure-free years wit hout an attempt to stop medication. Sixty-seven per cent of chi ldren with good social outcome were in remission, compared with 58% of children with bad social outc ome (p = not significan t). In the multiv ari ate analysis, the signific ant pred ictors of bad outco me wer e learning disorder (p <0 .0001), > 21 seizures before treatment with medication (p >0 .03), and number of potential adverse outcomes (p <0.01) . Good outcome was predicted by seizure type "s imple partial" (p < 0 .008) (Ta ble III). F orty-three per cent of our pati ents with pa rtial complex seizures ha d >20 seizures befor e treatment. Nonetheless , some patients with other seizure types had large numbers of seizures before treatment; only 55% of those with >20 seizure s before tr eatment h ad partial complex seizures. Wi th the use of the regressio n coefficients, a relatively inaccurate predictive sch eme for each pa tient was given by the following equation : (0 .179 X potential num ber of adver se outcomes) + (2.423 if lea rning d isorder ) - (1.708 if simple partial seiz ure type) + (0.707 if >20 seizures before treatment) - 1.0145. If the sum of the equation was >0.0595, bad outcome was predicted. If the su m was <0.0596, good outcome was predicted. This scheme had a total misclassification rate of 106 in 337 (sensitivity 54%; specificity 69%). T he residual deviance of the logistic model wa s 386.7, with 332 degrees of freedom . A goodness-of-fit hypothesis was rejected (p = 0.02 1), indicating substantial va riation in the data that was not explained by the logistic model. The results of the tree model were very similar; the identical variables were identi fied as significant, and the overall misclassification rate was the same. Both the stepwise logistic regression and the tree model identified learning disorder as the most importan t predictive variable, regardless of the status of other predictors. When patients with a learn ing disorder were excluded, a stepwis e logistic model again foun d simple partial seizures (p <0.0001) , number of seizu res >21 before initiation of treatment (p <0. 05), and potential number of unfavorable outcomes (p < 0.024) to be the only significa nt predictors.
DIS C US SIO N T he major finding of our study is that most biologic factors related to the treatment and control of epilepsy in children with normal intelligence do not predict a variety of undesirable social outcomes. Even remiss ion of epilepsy was not pre dictive of social outcome.
Camfield et al.
871
Table II. Psychosocial outcome
Category 1. Grades repeated or
2. 3.
4. 5. 6. 7. 8.
special class Psychiatric or psychologic referral Psychotropic drugs Criminal offense Inadvertent pregnancy Social isolation Financial dependency Unemployment
Eligible sub jects'
Affected sub jects (No.)
No.
%
(95% CI)
101
290
35
(29. 40)
72
316
23
(18,28)
17
5
11
316 187 88
13
(3, 8) (0, 4) (5, 20)
11 21
71 71
16 30
(7, 24) (19, 40)
21
71
30
(19, 40)
4
2
CI, Confidence intervul, "Nu mber sufficiently old where exact information wasknown. Categories I to 3 requiredageat end of follow-up >6.9 years. ca tegory 4 age> 13.9 years, category 5 age > 13.9 years and female, categories 6 to 8 age >17.9 years.
Table 1/1. Variab les predictive of unfavorable outcomes: multivariate model Var iable
Coetrlclent
1. Learning disorder 2. Partial elementary
2.423 -1.708
seizure type 3. Potential number of adverse outcomes 4. Seizures before treatment: > 20
95%CI
p
(1.574, 3.272) <0.0001 (-3 .048, - 0.368) <0 .008
0.179
(0.0436, 0.3144) <0 .01
0,707
(0.031, 1.383)
<0.03
C/, Confidence interval.
We are aware of only two previous population-based stu dies of social success in patients with epilepsy. 8; I I Trostle et al.'! administered the Washington Psychosocial Inventory to nea rly all normally int elligent adults with epilepsy in R ochest er, Minn . They emphasized the importance of studying an entire population and concluded that rates of social disability in those with epilepsy had been overes timated in reports based on referred patients; nevertheless, 19% were noted to have serious social difficulties. Sillanpaa! followed a regional cohort of 178 childrenwith epilepsy for 30 years and noted a striking effect of mental retardation on social outcome. Poor seizur e control and certain seizure types were independen t predictors of some aspects of social outcome. It is difficult to compare his exp erience with ours because there was no overall score for social outcome and because pat ients with norma l intelligence were not considered separately. In our cohort of normally inte1ligent children, the stro ngest predictor of poor social outcome was the presence of a
872
Camfield et al.
learning disorder at the time of diagnosis of epilepsy. This is not surprising because a learning disorder without epilepsy influences social success.P Only two other biologic factors directly related to epilepsy were independently predictive of outcome-more than 20seizures before treatment and the "simple partial" type of seizure. At the end of our follow-up study, young adults were eligible for a greater number of bad outcomes than were younger children. This explains the reason that the number of potential bad outcomes also entered our multivariate model. In other words, the greater the number of possible bad outcomes, the more likely it is that a child will have at least one. The good social outcome for patients with simple partial seizures has not previously been emphasized. Simple partial seizures are relatively uncommon. In one study of 34 adolescents with epilepsy, the two with simple partial seizures had good social success.'! Of all seizure types, the simple partial type involves the smallest amount of cortex, raising the possibility that a broader spread of epileptic discharge through other cortical structures may somehow alter social success. Our data do not allow a direct test of this hypothesis. In contrast to some previous reports, we did not find complex partial seizures to be particularly associated with bad outcome.' It is possible that part of the predictive value of >20 seizures before treatment may be related to partial complex seizures-a seizure type that is often difficult for families to recognize. Nonetheless, more than half of the children with >20 seizures before treatment had other seizure types. Therefore the prognostic value of social outcome for >20 seizures before treatment could be only partly attributed to complex partial seizures. In a previous population-based study of temporal lobe epilepsy, we noted that intellectual ability was the best predictor of social adjustment.!" It is possible that allowing >20 seizures to occur before the diagnosis of epilepsy reflects a chaotic or poorly educated family-features that might alter adjustment to epilepsy but might have an equally negative effect on normal children. By defining bad outcome as one or more adverse social outcomes, we maximized the opportunity for biologic factors to be predictive. We were impressed by the high proportion of patients with at least one of the eight adverse outcomes. In this population-based study (without referral bias), 42% of patients had at least one bad social outcome. We believe that this degree of social disability is beyond what would be expected in a normal, nonepileptic population-a speculation that requires a matched control group for proof. Our study was not designed to establish that children with epilepsy have high rates of social disability; however, we are aware of some information for our province that supports
The Journal of Pediatrics June 1993
this hypothesis. During the study period approximately 11.7% of all children with normal intelligence had resource or special education services at school (personal communication, Nova Scotia Department of Education, 1993), whereas 35% of our patients had such services. Rates of inadvertent pregnancy are not easily obtained. The total rate of pregnancy, both planned and unplanned, for female adolescents 15 to 19 years of age in our province during the study period was about 4/1000 women per year (excluding spontaneous abortion and miscarriagej.P In the same age range, 11 (12.5%) of our 88 patients reported inadvertent pregnancy. We did not record howmany others had planned pregnancies. Twenty-three percent of our patients had been referred to a mental health facility. The rate of such referrals for the overall population of Nova Scotia is not known. However, in Ontario, which has more mental health services for children, it is estimated that 12% of all children 4 to 16 years of age have a mental health referral at some time." Thirty percent of our patients> 18 years of age were unemployed, if no longer in school. In 1988 in Nova Scotia, 17% of male residents and 14% of female residents aged 19 to 24 years were unemployed (Statistics Canada, 1993 [catalog No. 71-001; Labor Force]). Our data do not provide an explanation for the high frequency of poor social outcomes. If it is not related to biologic factors, it presumably is the nonspecific result of a chronic illness, a generic effect of epilepsy, or the result of unmeasured sociologic factors. A recent study addressed the issue of unsatisfactory educational achievement in children with epilepsy and concluded that social issues related to the home environment were important determinants, whereas biologic factors were not.!? Another recent study of 183 children with epilepsy followed in a referral clinic found that social competence, as measured by the Achenbach Child Behavior Checklist, was independently related to the biologic factor of good seizure control and to the social factors of high parental income and a two-parent family.IS However, only 15% of the variance was accounted for by these factors in the statistical model. In general, poor social outcome in adults with epilepsy, whether defined informally or based on standardized rating scales, has not been well explained by biologic features." There is some suggestion that those with either severe seizures or poor control fare more poorly.v!' In several studies, seizure type and laterality, seizure control, medication, age at onset, and duration of illness did not predict social outcome.v 19,20 Concomitant brain problems may be more important determinants than epilepsy.v S Studies that have examined social factors in adults with epilepsy have also not been successful in predicting outcome. In one study, a battery of sociologic and psychologic tests was used to attempt to predict psychologic abnormal-
The Journal of Pediatrics Volume 122, Number 6
ities in adults with poorly controlled epilepsy who were undergoing video monitoring.l? A multiple regression model found no biologic predictors but did identify three social predictors-stressful life events, poor adjustment to epilepsy, and financial stress. These factors accounted for only 23% of the variance in the statistical model, indicating that there are other, more important influences on social outcome than those measured. The generic effect of chronic illness on social function appears to warrant further investigation.I' Only one small study has addressed this issue directly in children with epilepsy.22 Increased rates of social disabilities in children with temporal lobe seizures were similar to those in children with chronic asthma. If the chronic illness factor is proved to be important in the social outcome of epilepsy, interventions used in other chronic diseases might be useful for children with epilepsy. Our study would have been strengthened by direct measures of parental social class and indicators of social disruption. A standardized rating scale might have given different results from our simple clinical measures of social outcome, and a longer follow-up period might have allowed the biologic effects of seizure frequency and severity to have an effect. Nonetheless, our findings and those reported in the literature indicate that the cause of social adjustment problems in people with epilepsy is not well understood. We conclude that unsatisfactory social outcome is common in childhood epilepsy (excluding those with absence and "minor motor" seizures). For most children with epilepsy and normal intelligence, the clinician cannot accurately predict social outcome. Biologic features of the seizures and their control appear to have little relationship with social success. REFERENCES 1. Levin R, Banks 8, Berg B. Psychosocial dimensions of epilepsy: a review of the literature. Epilepsia 1988;29:805-16. 2. Bear DM, Fedio P. Quantitative analysis of interictal behaviour in temporal lobe epilepsy. Arch Neurol 1977;34:454-67. 3. Smith DF, Baker GA, Dewey M, Jacoby A, Chadwick DW. Seizure frequency, patient-perceived seizure severity and the psychosocial consequences of intractable epilepsy. Epilepsy Res 1991;9:231-41. 4. Beran RG, Flanagan PJ. Psychosocial sequelae of epilepsy: the role of associated cerebral pathology. Epilepsia 1987;28:10710.
Camfield et al.
873
5. Trimble MR, Thompson P1. Anticonvulsant drugs, cognitive function and behaviour. Epilepsia 1983;2(suppl 1):855-63. 6. Dodrill CB, Beier R, Kasparick M, Tacke I, Take D, SiangYan T. Psychosocial problems in adults with epilepsy: comparison findings from four countries. Epilepsia 1984;25:17683. 7. Herman BP, Whitman S. Behavioral and personality correlates of epilepsy: a review, methodological critique, and conceptual model. Psychol Bull 1984;95:451-97. 8. Silianpaa M. Children with epilepsy as adults. Acta Paediatr Scand 1990;368(suppl):5-75. 9. Camfield PR, Camfield CS, Dooley JM, Tibbles JA, Fung T, Garner B. Epilepsy after a first unprovoked seizure in childhood. Neurology 1985;35:1657-60. 10. Goldman L, Weinberg M, Weisberg M, et al. A computer-derived protocol to aid in the diagnosis of emergency room patients with acute chest pain. N Eng! J Med 1982;307:588-96. 11. Trostle JA, Hauser WA, Sharbrough FW. Psychologic and social adjustment in Rochester, Minnesota. Neurology 1989; 39:633-7. 12. Rutter M, Graham P, Yule W. A neuropsychiatric study in children. Clinics in Developmental Medicine 1970 (Nos. 35, 36):150-62, 175-85. London: Spastics International Medical Publishers. 13. Clement MJ, Wallace SJ. A survey of adolescents with epilepsy. Dev Med Child Neural 1990;32:849-50. 14. Camfield PR, Gates R, Ronen G, Camfield C, Ferguson A, MacDonald GW. Comparison of cognitive ability, personality profile, and school success in epileptic children with pure right versus left temporal EEG foci. Ann NeuroI1984;15:122-6. 15. LeBlanc JC. Epidemiology of STDs in Nova Scotia. Nova Scotia Medical Bulletin 1992;71:193-5. 16. Cadman D, Boyle M, Offord DR, et al. Chronic illnesses, medical conditions and physical limitations in Ontario children. Can Med Assoc J 1986;135:761-7. 17. Mitchell WG, Chevez JM, Lee H, Gregman BC. Academic underachievement in children with epilepsy. J Child Neurol 1991;6:65-72. 18. Hermann BP, Whitman S, Hughes JR, Melyn MM, Dell J. Multietiological determinants of psychopathology and social competence in children with epilepsy. Epilepsy Res 1988;2:5160. 19. HermanBP, Whitman 8, WylerAR, Anton MT, Vanderzwagg R. Psychosocial predictors of psychopathology in epilepsy. Br J Psychiatry 1990;156:98-105. 20. Collings JA. Psychosocial well-being and epilepsy: an empirical study. Epilepsia 1990;31:418-26. 21. Pless IB. Clinical assessment: physical and psychological functioning. Pediatr Clin North Am 1984;31:33-45. 22. APter A, Aviv A, Kaminer Y, Weizman A, Lerman P, Tyano 8. Behavioral profile and social competence in temporal lobe epilepsy of adolescence. JAm Acad Child Adolesc Psychiatry 1991;30:887-92.