`
`Downloaded from https://academic.oup.com/neurosurgery/article-abstract/62/2/326/2558449 by deeann@ieonline.com on 15 January 2020
`
`Knut Stavem, M.D., M.P.H., Ph.D.
`Medical Division and
`Helse-Øst Health Services Research Centre,
`Akershus University Hospital; and
`Faculty Division,
`Akershus University Hospital,
`University of Oslo,
`Lørenskog, Norway
`
`Helge Bjørnæs, Ph.D.
`The National Centre for Epilepsy,
`Sandvika, Norway
`
`Iver A. Langmoen, M.D., Ph.D.
`Department of Neurosurgery,
`Ullevaal University Hospital,
`Oslo, Norway
`
`Reprint requests:
`Knut Stavem, M.D., M.P.H., Ph.D.,
`Medical Division,
`Akershus University Hospital,
`N-1478 Lørenskog, Norway.
`Email: knut.stavem@klinmed.uio.no
`
`Received, December 7, 2006.
`Accepted, October 3, 2007.
`
`LONG-TERM SEIZURES AND QUALITY OF LIFE
`AFTER EPILEPSY SURGERY COMPARED WITH
`MATCHED CONTROLS
`
`OBJECTIVE: We compared long-term seizure outcome and health-related quality of
`life (HRQoL) of patients who underwent epilepsy surgery and matched medically treated
`nonsurgical controls with intractable epilepsy.
`METHODS: Medically treated controls were identified for patients operated on for
`epilepsy between January 1, 1949 and December 31, 1992. We used a matched cohort
`design, matching for age, sex, and seizure type. The analysis was based on 70 com-
`plete matching pairs. HRQoL was assessed with the Quality of Life in Epilepsy Inventory
`89 questionnaire an average of 15 years after surgery.
`RESULTS: Among surgery patients, 48% were seizure-free during the previous year
`compared with 19% of the controls (P⫽ 0.0004). Fewer surgery patients used antiepilep-
`tic drugs (70%) than controls (93%). The odds of being seizure-free were higher for sur-
`gery patients in total and in subgroups divided according to length of follow-up. The
`mean HRQoL for surgery patients was higher in five of the 17 Quality of Life in Epilepsy
`Inventory 89 dimensions and worse in none. Among patients with more than 7 years of
`follow-up, HRQoL was better in three dimensions and worse in none. Among patients
`with 7 years of follow-up or less, HRQoL was better in two dimensions and worse in the
`language dimension of the Quality of Life in Epilepsy Inventory 89.
`CONCLUSION: After an average of more than 15 years of follow-up, epilepsy surgery
`patients had fewer seizures, used less antiepileptic medication, and had better HRQoL
`in several dimensions of the Quality of Life in Epilepsy Inventory 89 instrument than
`matched medically treated controls with refractory epilepsy, although possibly at a
`slight disadvantage in the language dimension among those with 7 years of follow-
`up or less.
`KEY WORDS: Epilepsy surgery, Health-related quality of life, Health status, Partial epilepsy, Seizures
`
`Neurosurgery 62:326–335, 2008
`
`DOI: 10.1227/01.NEU.0000297049.12530.4E
`
`www.neurosurgery-online.com
`
`Epilepsy surgery is a widely used and important interven-
`
`tion for patients with intractable epilepsy (10, 56). Its use
`has increased, although it is frequently considered an
`underused treatment option (1). Over time, methods of patient
`selection and evaluation have changed with a trend toward
`earlier surgery and the use of more advanced imaging and
`mapping procedures before surgery (1, 10, 45).
`People with epilepsy have high rates of anxiety, depression,
`and low self-esteem (12, 13, 18). They are more likely to be
`under- or unemployed (22, 46), experience lower rates of mar-
`riage (14, 31, 46), and face social stigma and restrictions on
`driving (16). Therefore, epilepsy can have an extensive impact
`on patients’ physical, psychological, and social functioning,
`thereby affecting their quality of life. Recently, there has been
`an increased interest in and an expansion of research into
`
`assessment of health-related quality of life (HRQoL) of people
`with epilepsy (52). Several disease-specific instruments have
`been developed to evaluate the impact of epilepsy on these
`patients’ lives (4, 17, 54).
`In approximately 30 to 40% of individuals with epilepsy,
`the disease is refractory to control with anticonvulsant med-
`ication (9, 43, 50). Nonrandomized studies suggest that sur-
`gery controls seizures in approximately two-thirds of
`selected patients, although there are variations in results
`between series of patients (2, 6, 11, 21, 33), and this has been
`confirmed in a randomized study with 1 year of follow-up
`(58). However, the majority of studies have had no control
`group, and the observation times have been limited to a few
`years after surgery. Hence, the overall effect of surgical treat-
`ment of epilepsy on HRQoL and the relationship of quality
`
`326 | VOLUME 62 | NUMBER 2 | FEBRUARY 2008
`
`www.neurosurgery-online.com
`
`AQUESTIVE EXHIBIT 1104 Page 0001
`
`
`
`LONG-TERM OUTCOME AFTER EPILEPSY SURGERY
`
`Downloaded from https://academic.oup.com/neurosurgery/article-abstract/62/2/326/2558449 by deeann@ieonline.com on 15 January 2020
`
`FIGURE 1. Flow diagram illustrating study recruitment and follow-up.
`
`1988 to 1992
`Between February 1, 1988 and December 31, 1992, 73 adult patients
`were operated on for epilepsy in Norway. Only patients operated on for
`focal epilepsy and with 2 or more years of observation after surgery
`were included in the study. Patients who underwent callosotomy, had
`apparent mental retardation, or were unable to complete self-adminis-
`tered questionnaires were excluded from the study. For each patient
`who underwent surgery, we retrospectively selected a conservatively
`treated control among patients who, at the time of operation ⫾1 year,
`had been hospitalized at the National Center for Epilepsy for intractable
`seizures. The controls were matched for age (⫾1 yr), sex, and seizure
`type (simple partial, complex partial, or partial with secondary general-
`ization). Initial matching was performed by a computerized search
`among all patients admitted between 1987 and 1993 (5239 patient stays),
`selecting possible controls after admission date, age, and sex. The rest of
`the matching process involved review of the discharge summary and/or
`medical record of 994 possible controls. The matched pairs were
`approved in a consensus meeting between two of the authors (IAL, KS).
`For the current study, we found 55 living patients with a matching
`medically treated control, and 53 of the controls were alive at the time
`of the 1995 postal survey (Fig. 1 ). Among the 55 selected controls, 19
`were referred for presurgical evaluation between 1992 and 2000. Ten of
`the controls completed the presurgical assessment, and seven proceeded
`with surgery between 1995 and 1998, after the time of the survey.
`
`Review of Medical Records
`Information regarding the epilepsy surgery patients and controls
`through 1988 was collected by medical record review in three hospitals
`between 1987 and 1989 (24) and from 1988 to 1992 by chart review in
`two institutions in 1994 and 1995 (47). Preoperative seizure frequency
`was converted to the number of seizures per month based on medical
`
`of life outcomes to seizure outcomes have not been fully
`documented (1, 52, 55).
`The objective of this study was to compare the seizure and
`HRQoL outcome of a national cohort of patients who had under-
`gone epilepsy surgery in Norway between 1949 and 1992 with a
`control group of patients with medically treated intractable
`epilepsy selected by a process of individual matching.
`
`PATIENTS AND METHODS
`
`Study Design
`This study used a matched cohort design and was based on patients
`operated on for epilepsy between 1949 and 1992 and individually
`matched medically treated controls. Patients who had: 1) epilepsy as
`the primary indication for surgery, 2) perioperative craniotomy, and 3)
`perioperative corticography were eligible for the study. Only patients
`operated on for focal epilepsy with resective surgery were included.
`The data for this study were collected in a postal survey in 1995, includ-
`ing data on self-reported seizure status and HRQoL. Patients with
`malignant brain tumor at histological examination of the specimen
`were excluded. As brain tumor diagnosis was difficult and could be
`missed before the use of computed tomography and magnetic reso-
`nance imaging (MRI), we also excluded patients with brain tumor as
`the reported cause of death within 5 years of epilepsy surgery.
`Epilepsy surgery has been performed in Norway since 1949.
`Presurgical evaluation has been centralized to the National Center for
`Epilepsy (preliminary selection, neurophysiological and neuropsycho-
`logical workup, and rehabilitation). After 1987, MRI became part of the
`presurgical evaluation protocol. The operations were performed at two
`neurosurgical centers in Oslo: the National Hospital (Rikshospitalet)
`and the Oslo City Hospital (Ullevål sykehus). The indications used to
`select patients for surgery have been described previously (24, 49).
`
`Sample and Matching Process
`Sample selection and matching were performed for two periods
`with slightly different processes. Therefore, they are described sepa-
`rately below. These two samples were pooled before the 1995 postal
`survey.
`
`1949 to 1988
`Between January 1, 1949 and January 31, 1988, 240 patients were oper-
`ated on for focal epilepsy in Norway (49). We excluded 39 patients
`because the indication for surgery was removal of a known brain tumor
`and five who died of a brain tumor within five years of surgery (Fig. 1 ).
`Among the remaining 196 patients who underwent surgery, we searched
`for a conservatively treated control with epilepsy among patients who at
`the time of surgery ⫾3 years had been hospitalized for partial epilepsy
`at the same center performing the presurgical evaluation of surgical can-
`didates, the National Center for Epilepsy. By use of medical record infor-
`mation, the controls were matched for age (⫾3 yr; for most patients, ⫾1
`yr), sex, and seizure type (simple partial, complex partial, or partial with
`secondary generalization). The matching process involved review of
`medical records of 709 possible control patients (24, 49). If there were sev-
`eral possible controls for one patient, we selected the one with the clos-
`est match in age. We found 1:1 matched control patients, in accordance
`with the criteria above, to 139 of the 196 surgical patients. Until 1991, 30
`of the selected controls (22%) had been evaluated as possible surgical
`candidates. At the time of the 1995 postal survey, 105 surgery patients
`and 94 control patients were alive and received questionnaires (Fig. 1 ).
`
`NEUROSURGERY
`
`VOLUME 62 | NUMBER 2 | FEBRUARY 2008 | 327
`
`AQUESTIVE EXHIBIT 1104 Page 0002
`
`
`
`Downloaded from https://academic.oup.com/neurosurgery/article-abstract/62/2/326/2558449 by deeann@ieonline.com on 15 January 2020
`
`removed or added covariates from the model one at a time, examining
`changes in the coefficients and stability of the model.
`In analyses of matched studies with incomplete data (30), it is com-
`mon to supplement with an unmatched analysis of all subjects after
`breaking the matching in the complete pairs. Therefore, we supple-
`mented our analysis with an unmatched analysis. We intended a priori
`to compare groups using analysis of covariance, adjusting for age, sex,
`and predominant seizure type. Because there was no statistically signif-
`icant difference between the groups in these variables, for simplicity,
`we used the t test for independent samples. Furthermore, we per-
`formed an unconditional logistic regression analysis. Because the
`results in the matched and unmatched analyses were similar, we pres-
`ent only the matched analysis in detail in this article.
`Stata Version 8.2 (Stata Corp., College Station, TX) software was used
`for analysis. We chose a significance level of 5%, using two-tailed tests.
`The regional medical ethics review committee approved the study.
`
`RESULTS
`Sample Response and Descriptives
`In total, 109 (68%) of the epilepsy surgery patients and 105
`(71%) of the controls responded and were included in the
`unmatched analysis (Fig. 1 ). A total of 70 surgery patient-
`control pairs (55% of pairs still alive) completed the question-
`naire and were included in the matched analysis (Fig. 1 ).
`In the matched sample, responding surgery patients had fewer
`generalized seizures at study entry, tended to be younger, and,
`hence, had undergone surgery more recently than nonrespon-
`dents; otherwise, there was little difference between respondents
`and nonrespondents (Table 1 ). Similarly, the responding controls
`had more generalized seizures and tended to be nonsignificantly
`younger than nonrespondents. In the matched sample with com-
`plete pairs, surgery patients had more seizures at baseline com-
`pared with the matched controls, and, as expected, there was no
`group difference in the matched variables (Table 1 ).
`Among the surgery patients, approximately 75% had tempo-
`ral lobe resections and 13% underwent frontal resections
`(Table 1 ). Histological examination results of the specimens were
`available for 55 of the 109 responding surgery patients: normal
`tissue in 18 patients (36%), hippocampal sclerosis in nine patients
`(18%), cerebrovascular/trauma in seven patients (14%), malfor-
`mations (angimatous, hamartoma) in seven patients (14%),
`postinflammatory changes in three patients (6%), neoplasms in
`three patients (6%), and focal gliosis in three patients (6%).
`Compared with the control group, a smaller proportion of the
`surgery group had experienced seizures during the previous year,
`and fewer used antiepileptic drugs. There was no difference
`between the two groups in employment or marital status, as
`reported in the 1995 postal survey (Table 2 ).
`
`Seizures
`In a multivariate analysis of the matched samples (complete
`pairs), surgery patients had higher odds of being seizure-free
`than controls (OR, 5.04; 95% CI, 1.64–15.43; P ⫽ 0.005). In addi-
`tion, time since seizure onset influenced the regression coeffi-
`cients and, therefore, was included in the multivariate model
`(OR, 0.88; 95% CI, 0.76–1.02; P ⫽ 0.09) (Table 3 ). Analysis in the
`
`STAVEM ET AL.
`
`record information. Postoperative seizure status the second year after
`epilepsy surgery was abstracted from the medical record at follow-up
`after 2 years (24, 47).
`
`Postal Survey
`In 1995, a comprehensive questionnaire with questions regarding
`sociodemographic data, seizure situation, medication use, and HRQoL
`was mailed to 160 surgery patients and 149 controls (Fig. 1 ). This sur-
`vey was conducted, on average, 17 years (range, 2–42 yr) after surgery
`(n ⫽ 160). The questionnaire was mailed with a cover letter explaining
`the study and a recommendation from a national interest group for
`people with epilepsy. Two reminders were sent; the second was a sim-
`plified three-page questionnaire with items regarding seizures and
`medication use.
`
`Outcome Assessment
`In the survey, the patients reported their seizure status during the
`past year (yes/no) and current use of antiepileptic medication (yes/no).
`HRQoL was assessed with the Norwegian version of the 89-item
`Quality of Life in Epilepsy Inventory (QOLIE-89) (48, 57). This ques-
`tionnaire was developed in the United States as a general HRQoL
`instrument that also contains several epilepsy-specific dimensions. The
`instrument contains 17 different scales, which are aggregated to an
`overall score (57). This measure has been demonstrated to be respon-
`sive to changes in seizure frequency (7).
`
`Statistical Analysis
`Descriptive statistics are presented with the mean and standard devi-
`ation or number (percentage), with the exception of seizure frequency,
`which is presented with the median and interquartile range. The t test or
`χ2 test was used for group comparison of independent samples. For
`paired samples, we used the paired t test, Wilcoxon signed-rank test for
`paired samples, McNemar test, or Stuart-Maxwell test, as appropriate.
`To assess the results over time, we performed separate analyses
`among patients who were included before 1987 (⬎7 yr of follow-up)
`and those included from 1987 to 1992 (ⱕ7 yr of follow-up). This cutoff
`corresponds with the time when MRI became part of the presurgical
`evaluation protocol.
`We analyzed HRQoL outcomes using a matched analysis of the com-
`plete pairs, in which both the epilepsy surgery patient and the control
`responded in the postal survey. In this analysis, we compared HRQoL
`between the groups using the t test for paired samples. We analyzed
`determinants of being seizure-free with bivariate and multivariate con-
`ditional logistic regression analysis, with self-reported seizure status
`during the past year as the dependent variable and stratification for
`pair number. As independent variables, we used possible risk factors
`for seizures or potential confounders available in the medical records
`at the time of surgery or study entry, including epilepsy surgery status,
`age, sex, predominant seizure type, baseline seizure frequency, and
`duration of epilepsy.
`Results of univariate and multivariate analyses are presented as
`odds ratios (ORs) with 95% confidence intervals (95% CIs) and P val-
`ues. The ORs for continuous time variables are presented for increases
`of 1 year. For seizure frequency, we report the OR for an increase of
`20 seizures per month, corresponding to 50% of the interquartile range
`of baseline seizure status in the total sample, with a median of 13
`(interquartile range, 5–45) seizures per month (n ⫽ 295).
`In the multivariate analysis, we included independent variables with
`a P value of 0.25 or less in the bivariate analysis (29). All independent
`variables were initially forced into the model. We then manually
`
`328 | VOLUME 62 | NUMBER 2 | FEBRUARY 2008
`
`www.neurosurgery-online.com
`
`AQUESTIVE EXHIBIT 1104 Page 0003
`
`
`
`Downloaded from https://academic.oup.com/neurosurgery/article-abstract/62/2/326/2558449 by deeann@ieonline.com on 15 January 2020
`
`TABLE 1. Descriptive statistics for responding and nonresponding subjects in the postal survey according to surgery status: matched samplea
`Epilepsy surgery
`Controls
`Surgery
`Controls
`Responding Unpaired or
`Responding Unpaired or
`Respondents:
`matched
`nonrespond-
`matched
`nonrespond-
`complete
`pairs
`ing pairs
`pairs
`ing pairs
`matched pairs
`67–70
`51–58
`67–70
`55–58
`60–70
`60–70
`0.73
`37.0 (12.2)
`40.1 (13.7)
`37.0 (12.1)
`40.0 (13.7)
`37.0 (12.2)
`37.0 (12.1)
`—
`37 (53)
`25 (43)
`—
`—
`—
`—
`0.85
`9.4 (8.2)
`9.7 (9.2)
`9.6 (8.4)
`9.3 (9.0)
`9.4 (8.3)
`9.6 (8.4)
`0.84
`27.7 (12.3)
`29.9 (11.7)
`27.0 (13.0)
`30.5 (13.3)
`27.3 (12.2)
`27.0 (13.0)
`26 (10–60)
`31 (12–79)
`6 (4–16)
`12 (4–35)
`30 (9–60)
`6 (4–18) 0.004
`
`LONG-TERM OUTCOME AFTER EPILEPSY SURGERY
`
`P
`valueb
`
`P
`value
`
`0.19
`
`0.85
`0.15
`0.11
`
`P
`value
`
`0.18
`0.27
`0.84
`0.31
`0.28
`
`No. of patients
`Age at time of survey, yr
`Female, no. (%)
`Age at seizure onset, yr
`Time since seizure onset, yr
`Baseline seizures per month,
`median (IQR)
`Generalized seizures before
`inclusion, no. (%)
`Predominant seizure type
`preoperatively
`Only complex partial seizures
`Complex partial seizures w/
`secondary generalization
`Other
`Known cause, no. (%)
`Age at operation, yr
`Patients 18 yr old or younger at
`surgery, no. (%)
`Time since surgery, yr
`Time of surgery (before/after magnetic
`resonance imaging), no. (%)
`1949–1986
`1987–1992
`Resection type
`Temporal
`Frontal
`Other
`Seizures the second year
`after surgery, no. (%)
`
`22 (31)
`
`26 (49)
`
`c
`
`d
`
`16 (44)
`13 (36)
`
`20 (57)
`12 (34)
`
`7 (19)
`44 (73)
`24.0 (10.6)
`23 (33)
`
`3 (9)
`42 (82)
`24.0 (10.1)
`17 (29)
`
`13.1 (10.7)
`
`16.2 (11.1)
`
`35 (50)
`35 (50)
`
`53 (75)
`9 (13)
`8 (11)
`34 (49)
`
`39 (67)
`19 (33)
`
`39 (68)
`11 (19)
`7 (14)
`28 (55)
`
`0.05
`
`0.36
`
`0.26
`0.99
`0.67
`
`0.11
`0.05
`
`0.75
`
`0.54
`
`20 (29)
`
`27 (47)
`
`0.04
`
`22 (31)
`
`20 (29)
`
`0.67
`
`c
`
`e
`
`0.55
`
`0.81f
`
`17 (43)
`17 (43)
`
`6 (15)
`44 (77)
`
`16 (44)
`13 (36)
`
`7 (19)
`44 (73)
`
`18 (50)
`11 (31)
`
`7 (19)
`44 (73)
`
`1.00
`
`0.11
`
`18 (50)
`11 (31)
`
`7 (19)
`46 (66)
`—
`—
`
`—
`—
`
`—
`—
`—
`—
`
`aIQR, interquartile range. Values are means (standard deviations) unless otherwise
`stated; n ⫽ 256.
`bComparison of complete pairs responding to item on seizure status.
`cn ⫽ 36.
`
`dn ⫽ 35.
`en ⫽ 40.
`fn ⫽ 36, Stuart-Maxwell statistic.
`
`unmatched sample resulted in similar findings as those in the
`matched sample (data not shown).
`Among those with 7 years or less of follow-up, the odds of
`being seizure-free were higher for surgery patients compared
`with controls (OR, 13.00; 95% CI, 1.70–99.37; P ⫽ 0.01). Among
`those with more than 7 years of follow-up, the OR was 5.43
`(95% CI, 0.91–32.47; P ⫽ 0.06) (Table 3 ).
`Health-related Quality of Life
`In an analysis of the complete pairs, the mean HRQoL for
`surgery patients was significantly higher in five of the 17
`QOLIE-89 dimensions with nearly significant differences in
`
`another five dimensions and the overall score. The differences
`in favor of the epilepsy surgery patients were greatest in
`energy/fatigue, health perception, pain, physical role limita-
`tion, and emotional well-being (Table 4 ). The groups were very
`similar in the attention/concentration and memory dimen-
`sions. In the language dimension, there was a reversal in scores,
`with controls scoring somewhat higher than surgery patients,
`although this was not statistically significant (Table 4 ). In the un-
`matched sample, the results were essentially the same as those
`in the complete pairs (data not shown) but with somewhat
`more statistical power than in the matched sample because of
`the larger sample size.
`
`NEUROSURGERY
`
`VOLUME 62 | NUMBER 2 | FEBRUARY 2008 | 329
`
`AQUESTIVE EXHIBIT 1104 Page 0004
`
`
`
`STAVEM ET AL.
`
`Downloaded from https://academic.oup.com/neurosurgery/article-abstract/62/2/326/2558449 by deeann@ieonline.com on 15 January 2020
`
`Included 1948–1986
`Surgery
`Control Pvaluea
`23–36
`23–36
`17 (50)
`25 (74)
`23 (68)
`31 (91)
`11 (31)
`9 (25)
`14 (44)
`11 (34)
`8 (35)
`5 (22)
`5 (22)
`6 (26)
`
`0.06
`0.01
`0.64
`0.37
`0.32
`0.74
`
`Included 1987–1992
`Surgery
`Control Pvaluea
`23–37
`23–37
`19 (54)
`31 (89)
`25 (74)
`32 (94)
`16 (43)
`10 (27)
`20 (57)
`15 (43)
`16 (55)
`12 (41)
`9 (32)
`5 (18)
`
`0.001
`0.008
`0.44
`0.20
`0.10
`0.24
`
`Moreover, seizure frequency was not a predictor of a favorable
`outcome, in agreement with previous findings (36).
`In this study, the proportion of patients in this national
`cohort that was seizure-free and in full-time employment after
`epilepsy surgery was in accordance with other reports (3, 26,
`32, 34, 35, 38, 42), although many studies have reported higher
`seizure-free rates with shorter follow-up (27, 28, 36, 40–42).
`However, previous reports describing epilepsy surgery out-
`comes have very different time spans, for example, including
`patients that were evaluated before and/or after the availabil-
`ity of MRI; use different outcomes or methods for analysis that
`complicate comparisons between studies; and lack control
`groups in all but a few studies, which might lead to an inflation
`of the reported benefit (51).
`The beneficial effect of epilepsy surgery on seizures in the
`present study supports other studies with shorter follow-up:
`compared with a matched control group (25), with patients
`determined unsuitable for surgery after presurgical evaluation
`(5, 26, 34, 39, 41), patients waiting for epilepsy surgery (23), or
`a combined sample of surgery candidates and medically
`treated controls (59).
`Greater improvement in seizure frequency among those with
`observation times of 7 years or less, compared with those with
`longer observation times, could be attributed to differences in
`selection of surgery candidates before and after the introduc-
`tion of MRI for preoperative evaluation, larger resections or
`more effective surgery, or a loss of effectiveness of surgery over
`time with increasing rate of relapse of seizures.
`
`Quality of Life
`Studies have shown an improvement in HRQoL scores after
`epilepsy surgery in patients achieving a 75% reduction in
`seizure frequency compared with presurgery scores (44), pri-
`marily in patients who became entirely seizure-free (39).
`Furthermore, HRQoL scores have been better after surgery
`compared with patients waiting for epilepsy surgery (23), and
`recently HRQoL scores were improved after temporal lobe sur-
`gery in a randomized controlled trial of 80 patients (58), in
`which the controls later underwent surgery. There are, how-
`ever, differences in study design, patient selection, outcome
`
`TABLE 2. Self-reported information from the 1995 postal survey
`Matched analysis (complete pairs)
`Surgery
`Control
`Pvaluea
`45–73
`45–73
`36 (52)
`56 (81)
`48 (71)
`63 (86)
`27 (37)
`28 (38)
`34 (51)
`26 (39)
`24 (46)
`17 (33)
`14 (27)
`11 (22)
`
`No. of patients
`Seizures in past year, no. (%)
`Uses antiepileptic drugs, no. (%)
`Married/cohabiting, no. (%)
`Full-time education/work, no. (%)
`Has one child or more, no. (%)
`Epilepsy in family, no. (%)
`
`0.0004
`0.0003
`0.86
`0.12
`0.07
`0.51
`
`aMcNemar’s test.
`
`Among patients with 7 years or less of follow-up, surgery
`patients reported better scores than controls on the social isola-
`tion and medication effects dimensions and worse scores on
`the language dimensions of the QOLIE-89 (Table 4 ). Among
`those with more than 7 years of follow-up, surgery patients
`reported better results on the pain, energy/fatigue, and emo-
`tional well-being scales than medically treated controls (Table 4 ).
`
`DISCUSSION
`
`Major Findings
`The present study of outcomes in a national cohort of
`epilepsy surgery patients and individually matched medically
`treated controls had a mean follow-up of more than 15 years.
`The surgery group experienced better seizure control and used
`less antiepileptic medication than the matched control group.
`There was no difference in employment status. In the matched
`analysis, surgery patients had better HRQoL than matched con-
`trols in five out of 17 QOLIE-89 dimensions and worse HRQoL
`in none. In patients with 7 years or less of follow-up, surgery
`patients had less frequent seizures, used less antiepileptic med-
`ication, and reported less social isolation, but had more
`reported language problems than the controls.
`In patients with more than 7 years of follow-up, HRQoL was
`better after surgery in three out of 17 QOLIE-89 dimensions
`and worse in none compared with controls, and the odds of
`being seizure-free tended to be higher after surgery, although
`this advantage was statistically nonsignificant.
`
`Seizure Outcome
`According to a recent review, almost all studies of long-term
`(ⱖ5 yr) seizure outcomes after epilepsy surgery describe
`patient cohorts without controls (51). In the present study with
`matched medically treated controls, longer time since seizure
`onset was associated with lower odds of being seizure-free at
`long-term follow-up, in agreement with some previous reports
`(6, 9, 19, 28), but in contrast to another (36). This finding can be
`interpreted in support of the notion that intractable epilepsy
`becomes progressively more resistant to treatment with time.
`
`330 | VOLUME 62 | NUMBER 2 | FEBRUARY 2008
`
`www.neurosurgery-online.com
`
`AQUESTIVE EXHIBIT 1104 Page 0005
`
`
`
`Downloaded from https://academic.oup.com/neurosurgery/article-abstract/62/2/326/2558449 by deeann@ieonline.com on 15 January 2020
`
`LONG-TERM OUTCOME AFTER EPILEPSY SURGERY
`
`TABLE 3. Self-reported determinants for being seizure-free during the last year obtained via surveya
`All
`Included 1948–1986
`
`Included 1987–1992
`
`95% Confi-
`No. of Odds
`pairsb ratio dence interval
`
`P
`value
`
`P
`95% Confi-
`No. of Odds
`pairsb ratio dence interval value
`
`No. of Odds
`pairsb ratio
`
`P
`95% Confi-
`dence interval value
`
`Bivariate analysis
`Surgery vs. control
`Time since seizure onset, per year
`Preoperative seizures,
`increase of 20 per month
`Preoperative seizures, more than
`13 vs. 13 or fewer per month
`Known cause, yes vs. no
`Age at seizure onset, per year
`Multivariate analysis
`Surgery vs. control
`Time since seizure onset, per year
`
`64
`58
`62
`
`64
`
`64
`60
`
`58
`58
`
`4.33
`0.90
`0.98
`
`1.78–10.53
`0.81–0.99
`0.90–1.07
`
`0.0002
`0.01
`0.65
`
`1.29
`
`0.48–3.45
`
`0.62
`
`1.67
`1.05
`
`5.04
`0.88
`
`0.61–4.59
`0.97–1.13
`
`0.32
`0.22
`
`1.64–15.43
`0.76–1.02
`
`0.005
`0.09
`
`36
`32
`36
`
`32
`
`36
`32
`
`32
`32
`
`aConditional logistic regression analysis of the matched sample (complete pairs).
`bOdds ratios are based only on the pairs that were discordant for the outcome.
`
`1.25
`1.14
`
`5.43
`0.82
`
`28
`26
`26
`
`28
`
`28
`28
`
`2.60
`0.88
`0.97
`
`0.93–7.29
`0.76–1.02
`0.89–1.06
`
`0.07
`0.10
`0.54
`
`0.83
`
`0.25–2.73
`
`0.76
`
`13.00
`0.91
`1.89
`
`1.70–99.37
`0.79–1.05
`0.75–4.74
`
`0.01
`0.20
`0.18
`
`4.00
`
`0.45–35.79
`
`0.22
`
`0.34–4.65
`0.98–1.32
`
`0.73
`0.09
`
`2.50
`0.96
`
`0.49–12.89
`0.85–1.09
`
`0.91–32.47
`0.65–1.02
`
`0.06
`0.08
`
`13.00
`28
`— —
`
`1.70–99.37
`—
`
`0.27
`0.54
`
`0.01
`—
`
`assessments, and reporting that make direct comparisons
`among the studies difficult.
`Our finding of better HRQoL in several QOLIE-89 dimen-
`sions than in controls are in agreement with previous findings
`(34, 38, 39, 41, 54) and those of a recent study with several
`control groups (6). Kellett et al. (34) reported a better HRQoL
`in seizure-free patients than in nonsurgery patients, but they
`documented no difference between patients with more than
`10 seizures per month postoperatively and nonsurgery
`patients. The finding of similar results in those with more
`than 7 years of follow-up is reassuring. However, the finding
`of more problems in the language dimension of the QOLIE-89
`in surgery patients among those with 7 years or less of follow-
`up is notable. This contrasts with the previous finding of an
`improvement after anterior temporal lobectomy over 2 years
`compared with a nonsurgical comparison group (39). This
`dimension of the questionnaire consists of five items with six
`response alternatives, asking how often the respondent has
`trouble speaking or how often these problems interfere with
`normal work or living, finding the correct word, understand-
`ing what others are saying, understanding directions, under-
`standing what the respondent read, or writing.
`The findings in the present study also confirm that many of
`the dimensions of the QOLIE-89 questionnaire are able to dis-
`criminate between two groups of subjects who had different
`interventions, thus supporting the validity of this questionnaire.
`Limitations
`Some limitations of the present study and the design with
`a retrospectively selected control group should be noted. Al-
`though there was no apparent baseline difference between
`the groups after the matching process, unrecognized con-
`founding factors may have influenced the results, as in any
`
`observational study. For example, confounding by indication
`or contraindication (37) or possibly bias owing to differential
`detection of outcome may have occurred. The sample size of
`the study was limited, which again reduced the statistical
`power and restricted the analytical techniques. Furthermore,
`the design with matched pairs typically leads to a significant
`fraction of missing values, further reducing the sample avail-
`able for analysis, as in the present study. Therefore, more of
`the results probably would have been statistically significant
`in a larger sample, for example, the OR of 5.43 of seizure-
`freedom for surgery patients among those with follow-up of
`more than 7 years. In such situations, supplementary analysis
`of the unmatched sample can be justified; in the present study,
`it demonstrated similar results as in the matched sample. In
`many studies, however, analysis is only performed in com-
`plete pairs, as in our primary analysis. Thus, the information
`in the incomplete matching pairs is ignored. In this sample,
`seizures were prevalent; therefore, one should be careful not
`to interpret the ORs as risk ratios, which can be justified only
`with rare outcomes.
`A different response rate in the two groups could also cause
`bias. The data used were primarily self-reported, with a possi-
`bility of recall bias or other errors in the data. Some data or cri-
`teria were collected by medical record abstraction, which also
`has inherent limitations that might call into question the relia-
`bility of the data. Moreover, in our sample, not all patients
`underwent standard anterior temporal lobectomy. Hence, the
`mixture of sites of surgery and causes might contribute to
`somewhat lower seizure-free rates than expected.
`In the current study, we gathered self-reported survey data for
`seizure outcome using a rudimentary yes/no classification. On
`the basis of previous experience, we thought that a more detailed
`or sophisticated classification would not work in this setting.
`
`NEUROSURGERY
`
`VOLUME 62 | NUMBER 2 | FEBRUARY 2008 | 331
`
`AQUESTIVE EXHIBIT 1104 Page 0006
`
`
`
`STAVEM ET AL.
`
`Downloaded from https://academic.oup.com/neurosurgery/article-abstract/62/2/326/2558449 by deeann@ieonline.com on 15 January 2020
`
`TABLE 4. Comparison of health-related quality of life between epilepsy surgery patients and controlsa
`All complete matched pairs
`Included 1948–1986
`
`Included 1987–1992
`
`Surgery
`
`Control
`
`Difference: surgery-control
`
`Difference: surgery-control
`
`No. of
`pairs
`
`Mean
`
`Standard
`deviation
`
`Mean
`
`P
`Standard
`deviation valueb
`
`No. of
`pairs
`
`Mean
`
`No. of
`P
`Standard
`deviation va