This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution

and sharing with colleagues.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/copyright

Author's personal copy

Epilepsy Research (2012) 101, 113—121

jou rn al h om epa ge: www.elsev ier .com/ locate /ep i lepsyres

Predictive model for refractoriness in Temporal LobeEpilepsy based on clinical and diagnostic test data

Pedro J. Serrano-Castro ∗, Manuel Payan-Ortiz, Pablo Quiroga-Subirana,Javier Fernandez-Perez, Tesifon Parron-Carreno

Neurology and Neurophysiology Unit, Hospital Torrecárdenas, Paraje de Torrecárdenas S/N CP: 04009, Almeria, Spain

Received 10 November 2011; received in revised form 5 March 2012; accepted 9 March 2012Available online 1 April 2012

KEYWORDSTemporal LobeEpilepsy;Refractoriness;Prediction model;Prognosis;Epilepsy surgery

SummaryIntroduction: Temporal Lobe Epilepsy (TLE) is frequently resistant to drug treatment, but ahigh percentage of these patients can be free of seizures after epilepsy surgery. Delay in thesurgical decision has been related to quality of life impairment, social and work limitations, andincreased mortality risk. A predictive model for refractoriness based on clinical and diagnosticfactors may allow its earlier detection and a shorter delay before surgery.Material and methods: A case—control study was conducted in TLE patients over 16 years old.The dependent variable was resistance to medical treatment according to ILAE 2010 criteria.Independent variables were clinical, semiological, therapeutic, neurophysiological, radiolog-ical, and neuropsychological variables. A multivariate study was conducted to identify thevariables associated with refractoriness, calculating the positive and negative predictive val-ues and positive likelihood ratios of these variables individually and in combination. These datawere used to construct a refractoriness predictive model.Results: ILAE refractoriness criteria were met by 83 patients (50.9%). In the multivariateanalysis, refractoriness was significantly associated with one semiological variable, one neu-roradiological variable, one neurophysiological variable, and two therapeutic variables but notwith neuropsychological test outcomes. These significant variables were used to construct apredictive model.Conclusion: Assessment of semiological, neurophysiological, and neuroradiological data canserve to stratify the risk of refractory epilepsy in TLE patients.© 2012 Elsevier B.V. All rights reserved.

∗ Corresponding author. Tel.: +34 671562365.E-mail addresses: pedroj.serrano.sspa@juntadeandalucia.es,

pedro.serrano.c@gmail.com (P.J. Serrano-Castro).

Introduction

Temporal Lobe Epilepsy (TLE) is the most frequent typeof human epilepsy (Engel, 2001), and 20—40% of all TLEcases are refractory to medical treatment (Cockerell et al.,1997; Kwan and Sander, 2004). Epilepsy surgery has been

0920-1211/$ — see front matter © 2012 Elsevier B.V. All rights reserved.doi:10.1016/j.eplepsyres.2012.03.009

Author's personal copy

114 P.J. Serrano-Castro et al.

reported to achieve complete seizure remission in 40—90%of patients with refractory TLE, depending on the patholog-ical substrate (Spencer and Hugh, 2008). The effectivenessof anterior temporal lobectomy for refractory TLE was con-firmed by a controlled clinical trial (Wiebe et al., 2001),and a recent meta-analysis concluded that the combinationof surgery and medical treatment was four-fold more likelyto completely eradicate seizures in appropriately selectedpatients with refractory TLE in comparison to medical treat-ment alone (Schmidt and Stavem, 2009).

The surgical decision should be made as early as pos-sible in these patients. The most frequent refractory TLEsyndromes, such as Mesial Temporal Lobe Epilepsy withHippocampal Sclerosis (MTLE-HS), produce progressive neu-ropsychological impairment throughout their natural history,attributable to repeated seizures that can only be preventedby non-pharmacological means, such as surgery (Jokeit andEbner, 1999; Helmstaedter and Elger, 2009). A delay insurgery implies long-term polytherapy with complex medi-cal treatments, which are associated with adverse cognitiveand systemic effects (Carreno et al., 2008) that impair thequality of life of epilepsy patients (Berto, 2002). The socialisolation, academic failure, and inability to achieve workand life goals that characterize patients with epilepsy havebeen related to their repeated seizures and antiepilepticdrug (AED) treatments (Engel, 2004), and the mortality riskis increased by seizure repetition (Sperling et al., 1999;Tomson et al., 2008). Nevertheless, these patients are nottreated with surgery (when available) until 20 years afterdisease onset (Engel et al., 2003) and after undergoing mul-tiple unsuccessful therapeutic trials (Berg, 2008).

An additional factor in this delay is that the concept ofrefractory epilepsy (RE) has only recently become interna-tionally accepted (Kwan et al., 2010); therefore, there arefew published data for results comparison and for estab-lishing consensus guidelines (Perucca, 1998; Berg et al.,2006; Arzimanoglou and Ryvlin, 2008). The ILAE proposeda definition of RE as the failure of adequate trials of twotolerated and appropriate AED regimens in monotherapy orcombination (Kwan et al., 2010). However, a long period cansometimes elapse before this condition is met, during whichthe patient’s quality of life continues to deteriorate. Theability to predict refractoriness before patients fulfill ILAEcriteria would therefore be of high clinical value.

RE has been studied by various authors, and the factormost consistently related to its development is the patho-logical substrate. The condition most frequently associatedwith treatment resistance is MTL-HS, with reports that thesepatients have an 89% likelihood of developing RE, followedby cortical dysplasias (Semah et al., 1998). Refractoriness inTLE has also been related to: early onset, long time course,mental retardation, lack of response to first AED, require-ment for polytherapy, detection of focal alterations in EEG,and radiologic evidence of HS or structural lesion in tempo-ral lobe (Kim et al., 1999; Dlugos et al., 2001; Spooner et al.,2006; Varoglu et al., 2009; Pittau et al., 2009; Aguglia et al.,2011).

In an exhaustive review of the literature, we detectedfour studies on the relationship between clinical TLE vari-ables and the development of treatment resistance (Falipet al., 2003; Varoglu et al., 2009; Pittau et al., 2009;Aguglia et al., 2011), but only partial data were analyzed,

results were always negative, and there was no discussionof semiological issues. We found no report on the possiblerelationship between a specific neuropsychological profileand the development of treatment resistance.

The objectives of this study were: to describe the earlyclinical and diagnostic characteristics of TLE patients whodevelop RE, to stratify RE risk as a function of these charac-teristics, and to propose a predictive model for progressionto refractory TLE based on clinical and diagnostic test cri-teria.

Material and methods

This epidemiological case—control study was performed inpatients with a confirmed diagnosis of TLE under follow-upat the Neurology and Neurophysiology Clinical ManagementUnit of the Torrecardenas Hospital Complex in Almeria(Spain). The diagnosis of TLE was considered when thepatient experienced paroxysmal episodes compatible withepileptic seizures of temporal origin according to clinicalcriteria and there was also a structural lesion evidenced byneuroimaging techniques in one or both temporal lobes ora persistent electrical anomaly in temporal areas in ictal orinterictal EEG records.

A case group was formed by the patients who metthe 2010 ILAE definition of RE, i.e., those who had beenunsuccessfully treated with two tolerated and appropri-ate AED regimens in monotherapy or combination (Kwanet al., 2010). A control group was formed by the remainingpatients.

Sampling was consecutive. Study inclusion criteria were:age ≥16 years; capacity of the patient or direct relativeto give reliable information on study variables (as assessedby the researcher); follow-up period of ≥1 year; avail-ability of adequate data on clinical variables or ability toreliably gather any missing data by prospective interviewwith patient or relative; history of at least 2 interictal EEGrecords separated by >3 months or a prospective video-EEG monitoring study; and a brain MRI study of adequatequality following a specific epilepsy protocol or the oppor-tunity to obtain a prospective study of these characteristics.Exclusion criteria were the presence or well-founded suspi-cion of psychogenic pseudo-seizures and previous surgery forepilepsy.

Definition of variables

The dependent variable was refractoriness to medical treat-ment, defined according to ILAE 2010 criteria. Independent(predictive) variables were: sex, years since epilepsy onset,age at first seizure, history of central nervous system(CNS) infection, history of febrile convulsions (FCs), typeof prevalent seizure according to the ILAE classification(Commission on Classification and Terminology of the ILAE,1981), presence of aura, automatic behavior during seizures,non-automatic ictal motor behavior, secondary general-ization, number of AEDs being taken at study inclusion,etiological classification, neurophysiological findings, neu-ropsychological assessment, and previous CNS surgery of anytype except for epilepsy surgery (see exclusion criteria).

Author's personal copy

Refractoriness in Temporal Lobe Epilepsy 115

The etiology was classified in three groups accordingto the neuroimaging results and following ILAE criteria:unknown origin group (normal neuroimaging), MTLE-HSgroup (neuroimaging compatible with HS), and structuralgroup (neuroimaging compatible with structural diseaseother than HS).

The neuropsychological assessment comprised five tests:visual memory test (complex Rey figure), verbal memorytest (Spain-Complutense verbal learning test [initials inSpanish-TAVEC]), verbal fluency test (Boston Naming Test),visual—perceptual/visual constructive function test (WAISCubes III), and executive function test (controlled oral wordassociation test [COWAT]). Test results were classified asnormal or pathological by comparison with control groupfindings.

Study data were gathered as follows (in this order): ret-rospective analysis of clinical history by means of an adhoc structured questionnaire; personal interview with thepatients and relatives to collect data missing from the clini-cal history; performance of complementary tests for absentor discrepant data, including interictal EEG, ictal video-EEGrecord, high-resolution MRI with specific epilepsy protocol,or standard neuropsychological battery.

Statistical analysis

The Shapiro—Wilks or Kolmogorov—Smirnov tests were usedto assess the normality of continuous variables, which wereexpressed as mean (standard deviation [SD]) or median(range, maximum and minimum values) accordingly. TheStudent’s t or Mann—Whitney U test, as appropriate, wereused to compare continuous variables between the groups,and the chi-square or Fischer’s exact test were used tocompare categorical variables. Variables showing a signifi-cant relationship with the dependent variable in bivariatecorrelation analyses were entered into a logistic regressionmodel, along with variables considered highly clinically ordiagnostically relevant, using the estimation change estima-tion method to assess confounding factors and calculatingodds ratios (ORs) with confidence intervals (CIs). Finally, amultivariate analysis was performed to control for confound-ing factors. Based on the regression coefficients (�) of theindependent variables entered into the model, we obtainedtheir OR with 95% CI. SPSS Version 17 (SPSS Inc, Chicago, IL)was used for the above analyses.

We then calculated the positive (PPV) and negative (NPV)predictive values and positive likelihood ratio (PLR) of thediagnostic variables that were significant in the multivariateanalysis, both individually and in combination, in order toconstruct a refractoriness prediction model; both refractoryand non-refractory groups were included in the analysis inorder to develop a predictive model for all patients withTLE.

The Epidat version 3.1 statistical package was used forthese calculations.

Results

Out of the 163 TLE patients included in the study, ILAE-2010refractoriness criteria were met by 83 (50.9%) and were notmet by 80 (49.1%); 84 (51.5%) were female, 79 (48.5%) were

male. The age at onset ranged from 0 to 74 years, with amean of 20.39 years (SD = 16.61) and median of 16 years,and the mean time from diagnosis to study inclusion rangedfrom 1 to 77 years, with a mean of 24.58 years (SD = 15.71).

The etiology was unknown in 67 patients (41.1%) andstructural in 52 (31.9%), while MTLE-HS radiological crite-ria were met in the remaining 44 patients (27%). Among the44 MTLE-HS patients, an additional alteration was observedin the MRI study of 7 patients, who formed a dual-pathologysubgroup.

Nineteen patients (11.6%) had a history of FC during firstinfancy, which was considerably more frequent in the MTLE-HS group (20.5%) than in the unknown-etiology group (1.9%).Fourteen patients (8.64%) had a history of CNS infectionduring first infancy.

The predominant seizure type was I.B.1 (complex partialseizures [CPS] with aura) in 96 (58.9%) of the patients, typeI.B2 (without aura) in 52 (31.9%), type 1.A (simple partialseizures [SPS] alone) in 14 (8.64%), and type 1.C (partialseizures with secondary generalization) in 1 patient; out ofthe 14 patients with type I.A seizures, 4 had motor SPS alone.

Auras were experienced by 106 patients (65%); the aurawas visceral (epigastric sensation, cephalic oppression, tho-racic oppression, dizziness, sweating, or piloerection) in77 of these patients (72.6%), experiential or cognitive(déjà vu, derealization, fear, forced thought, pleasure, fasttime-passing) in 34 (32%), and sensory (visual, olfactory, gus-tatory, auditory type or paresthesia in tongue or genitals) inthe remaining 19 patients (17.9%).

Out of the 160 patients with available video-EEG recordsor reliable information, 122 (74.8%) showed some typeof automatism, which was oroalimentary in 101 patients(82.7%), manual in 78 (63.9%), and complex in 14 (11.4%).Patients frequently had more than one type of automatism.

Some type of motor behavior, predominantly cephalic,was found in 23 (16.6%) of the 138 patients for whom reliabledata were available (video-EEG monitoring or irrefutableanamnesis).

Out of the 163 patients in the series, 23.3% hadnever experienced secondary generalization, whereas 32.5%reported more than 10 generalized tonic—clonic seizuressince epilepsy onset.

The mean number of AEDs received by the patients was2.15 (SD = 0.99; range 0—4).

The 59 cases with structural etiology (52 with isolatedstructural lesion and 7 with dual disease) comprised: 17brain tumors (12 low grade glial tumors, 3 neuroectodermaltumors, 1 lipoma, and 1 meningioma), 11 temporal vascularmalformations (6 arteriovenous malformations, 3 cavernousangiomas, 1 telangiectasis, and 1 angioma associated withSturge-Weber syndrome), 8 cortical development disorders(6 focal cortical dysplasias, 1 subependymal heterotopia,and 1 hamartoma in tuberous sclerosis), 7 temporal poren-cephalies (5 ischemic vascular, 1 postsurgical, and 1 ofunknown origin), 5 temporal glioses (3 post-traumatic and2 post-encephalitic), and 11 miscellaneous cases, mostlyshowing sequelae of connatal encephalopathies.

Out of the 149 patients for whom neurophysiologic testdata were available, 96 (64.4%) evidenced some type offocal alteration, 67 (44.9%) some type of epileptiformactivity (spikes, spikes-waves, or acute waves), and theremaining 29 (19.4%) slow activity.

Author's personal copy

116 P.J. Serrano-Castro et al.

Table 1 Bivariate analysis of qualitative clinical variables.

Variable Refractory (N = 83) Non-refractory (N = 80) Crude OR (95% CI) P

Sex M/F (%) 40/43 (48.2%/51.8%) 39/41 (48.8%/51.2%) 0.978 (0.529—1.808) NSa

History of CNS infections 8 (9.6%) 6 (7.5%) 1.316 (0.435—3.997) NSa

History of FC 10 (12%) 9 (11.3%) 1.081 (0.415—2.817) NSa

Type of seizure 0.108I.A. alone 2 (2.4%) 12 (15%) 0.14 (0.03—0.647) 0.01a

Presence of SPS (type I.B) 81 (97.6%) 67 (83.8%) 7.858 (1.713—36.055) 0.005a

Any type of automatism 76 (91.6%) 46 (59.7%) 7.317 (2.98—17.964) <0.001a

Manual automatism 50 (60.2%) 28 (36.4%) 2.652 (1.399—5.026) 0.004a

Oroalimentary automatism 62 (74.7%) 39 (50.6%) 2.877 (1.477—5.603) 0.003a

Complex automatism 10 (12%) 4 (5.2%) 2.5 (0.75—8.334) NSa

Non-automatic ictal motor behavior 11 (13.3%) 14 (17.5%) 0.72 (0.306—1.698) NSa

Type of ictal motor behavior NSPresence of any type of aura 54 (65.1%) 49 (61.3%) 1.178 (0.623—2.228) NSa

Type of aura NSVisceral aura 38 (45.8%) 39 (48.8%) 0.888 (0.48—1.643) NSa

Experiential aura 19 (22.9%) 15 (18.8%) 1.286 (0.602—2.75) NSa

Specific sensitive or sensory aura 10 (12%) 9 (11.3%) 1.081 (0.415—2.817) NSa

Secondary generalization 0.178No generalized seizures 16 (19.3%) 22 (27.5%) 0.63 (0.302—1.311) NSa

>10 generalized seizures 27 (32.5%) 26 (32.5%) 1 (0.52—1.929) NSa

No. of AED <0.001Need for ≥3 AEDs 48 (57.8%) 15 (18.8%) 5.943 (2.92—12.095) <0.001a

CNS surgery 15 (18.1%) 2 (2.5%) 8.603 (1.899—38.975) 0.003a

a Correction by Yates continuity.

Table 2 Bivariate analysis of quantitative clinical variables. T test for independent samples. SD = standard deviation.

Variable Refractory (N = 83) Non-refractory (N = 80) P

Years since onset 28.34 (DT = 13.085) 20.69 (DT = 17.268) 0.007Age of first seizure 16.84 (DT = 13.034) 24.08 (DT = 19.048) <0.001

A standard neuropsychological battery was administeredto 43 of the 163 patients in the series. The result of the Reycomplex figure test was pathological in 40 patients (93.02%)and normal in only 3 (6.98%). Frontal executive functiondeficits were detected in 25 patients (58.13%) in the COWATtest, and deficits in verbal fluency and visual—perceptionfunction were found in 22 patients (51.16%) in the TAVECtest.

Tables 1—4 depict the results of bivariate compar-isons between cases and controls in all study variables.With regard to semiological variables, refractoriness wassignificantly associated with predominance of type I.Bseizure (OR = 7.858; 95% CI 1.713—36.055; p = 0.005) andthe presence of any type of automatism (OR = 7.317; 95%CI 2.98—17.964; p < 0.001), while the absence of refractori-ness was associated with the presence of type I.A SPS alone

Table 3 Bivariate analysis of complementary test variables (EEG and MRI).

Variable Refractory (N = 83) Non-refractory (N = 80) OR (95% CI) P

Etiological diagnosis by MRI <0.001Symptomatic etiology (structural lesion

in MRI. Including HS)60 (72.3%) 35 (43.8%) 3.354 (1.746—6.443) <0.001

HS identified by MRI 32 (38.6%) 11 (13.8%) 3.936 (1.814—8.541) <0.001a

Dual disease identified by MRI 7 (8.4%) 0 (0%) 1.092 (1.023—1.166) 0.014b

Some type of anomaly in interictal EEG 60 (72.3%) 39 (48.8%) 2.742 (1.431—5.256) 0.004a

Presence of IED in interictal EEG 40 (48.2%) 27 (33.8%) 1.826 (0.97—3.437) NSa

a Yates’ correction for continuity.b Fisher’s exact test.

Author's personal copy

Refractoriness in Temporal Lobe Epilepsy 117

Table 4 Bivariate analysis. Complementary test variables (neuropsychological study).

Variable Refractory (N = 31) Non-refractory (N = 12) OR (95% CI) P

Visual memory deficit (Rey complex figure test) 29 (93.5%) 11 (91.7%) 1.318 (0.1—16.039) NSb

Verbal memory deficit (TAVEC) 16 (53.3%) 6 (50%) 1.143 (0.299—4.364) NSa

Verbal fluency deficit 18 (58.1%) 4 (33.3%) 2.769 (0.685—11.188) NSa

Visual—perception function deficit 17 (56.7%) 5 (41.7%) 1.831 (0.472—7.104) NSa

Frontal executive function deficit (COWAT) 21 (67.7%) 4 (33.3%) 4.2 (1.01—17.32) 0.08a

a Yates’ correction for continuity.b Fisher’s exact test.

Figure 1 Relative refractoriness (%) by etiology in our patients (p = 0.172, chi-square test).

(OR = 0.14; 95% CI 0.03—0.647; p = 0.01). Refractoriness wassignificantly associated with oroalimentary and manual butnot complex automatisms (Table 1) and with earlier age atonset and longer time since onset (Table 2). It was alsoassociated with the presence of: HS (OR = 3.936; 95% CI1.814—8.541; p < 0.001), any type of structural lesion in tem-poral lobes (OR = 3.354; 95% CI 1.746—6.443; p < 0.001), dualpathology (OR = 1.092; 95% CI 1.023—1.166; p = 0.014), andfocal anomaly in serial interictal EEG (OR = 2.742; 95% CI1.431—5.256; p = 0.004) (Table 3). The refractoriness fre-quency is stratified as a function of the type of structuralanomaly in Fig. 1. No association was found between refrac-toriness and neuropsychological test results (Table 4).

Table 5 exhibits the results of the multivariate analysis,in which five variables were found to be significant: oneclinical variable (any type of automatism), two therapeu-tic variables (previous CNS surgery, receipt of ≥3 AEDs), and

two diagnostic variables (structural lesion in MRI study, focalanomaly in serial interictal EEG). The predictive values cal-culated for these variables and for combinations of thesevariables are given in Table 6.

Discussion

The general characteristics of our patients were very sim-ilar to those described in previous clinical studies of TLEpatients, having the same epileptogenic area but differing inage at onset, etiology, prognosis, and therapeutic response.The onset was earlier and the time since onset longer inthe refractory TLE patients than in the other TLE patients,and these variables were previously reported to be risk fac-tors for refractoriness in epilepsies in general (Kwong et al.,2003; Ramos-Lizana et al., 2009) and in TLE in particular

Table 5 Multivariate analysis model. Forward step-by-step method. Hosmer and Lemeshow �2 test: 5.416; Sig.: 0.712.

Variable Regression coefficients (ˇ) Adjusted OR (95% CI) P

Presence of automatisms 2.348 10.466 (3.341—32.786) <0.001CNS surgery 1.795 6.018 (1.106—32.731) 0.038Need for 3 or more AED 1.586 4.882 (2.161—11.029) <0.001Structural lesion in MRI (including HS) 1.085 2.958 (1.288—6.796) 0.011Some anomaly in interictal EEG 0.939 2.558 (1.129—5.795) 0.024

Author's personal copy

118 P.J. Serrano-Castro et al.

Tabl

e

6

Refr

acto

rine

ss

pred

icti

on

mod

el

base

d

on

the

sign

ifica

nt

inde

pend

ent

vari

able

s.

PPV:

posi

tive

pred

icti

ve

valu

e.

PLR:

posi

tive

likel

ihoo

d

rati

o.

NPV

: ne

gati

ve

pred

icti

veva

lue.

Vari

able

or

com

bina

tion

of

vari

able

s

PPV

%

(95%

CI)

PLR

(95%

CI)

NPV

%

(95%

CI)

Num

ber

of

pati

ents

in

the

grou

p/nu

mbe

r

of

pati

ents

in

glob

al

seri

es

(%)

Foca

l ano

mal

ies

in

inte

rict

al

EEG

s

60.6

1

(50.

48—

70.7

4)

1.48

(1.1

4—1.

93)

64.0

6

(51.

53—

76.6

) 99

/163

(60.

73%)

Pres

ence

of

auto

mat

ism

s

62.3

(53.

29—

71.3

)

1.59

(1.3

—1.

94)

82.9

3

(70.

19—

95.6

6)

122/

163

(74.

84%)

Stru

ctur

al

lesi

on

in

MRI

(inc

ludi

ng

HS)

63.1

6

(52.

93—

73.3

8)

1.65

(1.2

5—2.

19)

66.1

8

(54.

2—78

.16)

95/1

63

(58.

28%)

Pres

ence

of

auto

mat

ism

s +

foca

l ano

mal

ies

inin

teri

ctal

EEG

s72

.00

(61.

17—

82.8

3)

2.48

(1.6

6—3.

70)

67.0

5 (5

6.66

—77

.43)

75/1

63

(46.

01%)

Pres

ence

of

auto

mat

ism

s +

stru

ctur

al

lesi

on

inM

RI

(inc

ludi

ng

HS)

72.6

(61.

69—

83.5

2)

2.55

(1.6

9—3.

86)

66.6

7

(56.

37—

76.9

6)

73/1

63

(44.

78%)

Foca

l ano

mal

ies

in

inte

rict

al

EEG

s +

stru

ctur

alle

sion

in

MRI

(inc

ludi

ng

HS)

73.3

3

(61.

31—

85.3

6)

2.65

(1.6

4—4.

3)

62.1

4

(52.

28—

71.9

9)

60/1

63

(36.

81%)

IED

in

inte

rict

al

EEG

s +

stru

ctur

al

lesi

on

inM

RI

(inc

ludi

ng

HS)

75.0

0

(61.

07—

88.9

3)

2.89

(1.5

7—5.

32)

57.9

8

(48.

69—

67.2

7)

44/1

63

(26.

99%)

Pres

ence

of

auto

mat

ism

s +

foca

l ano

mal

ies

inin

teri

ctal

EEG

s +

HS

in

MRI

81.2

5

(66.

16—

96.3

4)

4.18

(1.8

2—9.

61)

56.4

9

(47.

62—

65.3

6)

32/1

63

(19.

63%)

Pres

ence

of

any

type

of

auto

mat

ism

+

IED

inin

teri

ctal

EEG

+

stru

ctur

al

lesi

on

in

MRI

(inc

ludi

ng

HS)

84.3

8

(70.

23—

98.5

2)

5.20

(2.1

1—12

.85)

57.2

5

(48.

40—

66.1

1)

32/1

63(1

9.63

%)

Pres

ence

of

auto

mat

ism

s +

foca

l ano

mal

ies

inin

teri

ctal

EEG

s +

stru

ctur

al

lesi

on

in

MRI

(inc

ludi

ng

HS)

84.4

4

(72.

74—

96.1

4)

5.23

(2.4

8—11

.03)

61.8

6

(52.

68—

71.0

5)

45/1

63

(27.

6%)

Author's personal copy

Refractoriness in Temporal Lobe Epilepsy 119

(Kim et al., 1999; Falip et al., 2003; Pittau et al., 2009;Varoglu et al., 2009).

We found no significant association between refractori-ness and a history of FC, in agreement with some authors(Falip et al., 2003; Spooner et al., 2006; Varoglu et al., 2009)but in disagreement with others (Kim et al., 1999; Hitiriset al., 2007; Pittau et al., 2009; Aguglia et al., 2011). Weobserved a correlation between radiological evidence of HSand a history of FC, which may explain the association withFC observed in some studies. Our finding supports clinicalpostulates regarding a possible common genetic substratefor FC and HS (Baulac et al., 2001; Claes et al., 2004).

We found no association between a history of CNS infec-tion and refractoriness, confirming previous findings (Kimet al., 1999; Falip et al., 2003; Spooner et al., 2006; Varogluet al., 2009), and it appears adequately demonstrated thatprevious CNS infection does not imply a worse prognosis inTLE patients.

With regard to semiological variables, refractoriness wassignificantly associated with the type of seizure and thepresence of automatisms but not with the presence or typeof aura, the presence or type of non-automatic ictal motorbehavior, or the development of secondary generalization.There have been few studies on the prognostic value of semi-ological data in TLE. Different authors have reported thatthe development of refractoriness is not significantly asso-ciated with the presence of secondary generalization (Pittauet al., 2009; Varoglu et al., 2009) or auras (Varoglu et al.,2009) or with auditory symptoms, autonomic symptoms,consciousness impairment, déjà vu, dysphasia, dystonicpostures, drop attacks, epigastric aura, or secondary gen-eralization (Aguglia et al., 2011). None of these studiesinvestigated the presence of automatic behaviors duringseizures; therefore, its strong correlation with refractori-ness in the present study (OR = 7.317; 95% CI 2.98—17.964;p = <0.001) is a novel finding. Although automatisms are asign of MTLE-HS (Delgado-Escueta et al., 1992; Gil-Nagel andRisinger, 1997) and could therefore represent a confound-ing factor, our multivariate analysis confirmed that theirpresence behaves as an independent predictive variable.We propose that the refractoriness may be associated withmesial lesions that cannot be visualized with available neu-roimaging techniques and that the presence of automatismsmay therefore serve as an early marker of damage of thisarea, consistent with the above-cited studies. The presenceof automatisms has been related to an improved postsurgicalprognosis, although statistical significance was not reached(Hennessy et al., 2001), further supporting the desirabil-ity of an early surgical decision in patients with MTLE andclinical or video-EEG evidence of automatic behavior.

We highlight the protective role against refractorinessof the presence of type I.A seizures alone. Epileptic aurasare highly characteristic of TLEs and are present in 90% ofpatients. However, cases in which type I.A seizures are thesole TLE manifestation are much less frequent, represent-ing only around 8% of our series and forming a subgroupwith differential characteristics. In contrast, CPS was thepredominant seizure in 97.6% of our refractory cases versus83.3% of our non-refractory cases, a significant difference.Hence, it appears exceptional for a TLE to meet refrac-toriness criteria if CPS is not the predominant clinicalmanifestation.

As previously reported, refractoriness was associatedwith the need for antiepileptic polytherapy (≥3 AEDs incombination) and for surgery, which represent the maxi-mum therapeutic effort required by refractory patients. In2001, Dlugos et al. assigned a strong prognostic value to theresponse to the first AED in children newly diagnosed withTLE. Subsequent studies found that replacement with oraddition of a second AED was effective in 27—32% of patients(Mohanraj and Brodie, 2005) but that any further increasein the number AEDs benefited only a very small percentageof patients. The above data are congruent with our results,which were also similar to the findings reported by Kim et al.(1999).

Our finding of an association between refractoriness andradiological evidence of HS, structural lesion, or dual pathol-ogy confirms previous reports, especially with respect to thepresence of HS, which is highlighted by almost all authorswho have analyzed TLE prognostic factors (Kim et al., 1999;Dlugos et al., 2001; Mohanraj and Brodie, 2005; Spooneret al., 2006; Pittau et al., 2009; Varoglu et al., 2009; Agugliaet al., 2011). However, the influence of other types of struc-tural lesion is more controversial; a significant relationshipof non-HS structural lesion with refractoriness was found byPittau et al. (2009) but not by Spooner et al. (2006). Withregard to the underlying disease, we found that dual pathol-ogy, HS, glial tumors, and cortical development disorderswere related to refractoriness but that vascular malforma-tions and porencephalies were not, in line with previousreports (Semah et al., 1998; Liimatainen et al., 2008).

In our series, alterations in interictal EEG records,whether interictal epileptiform discharges (IEDs) or not,conferred a risk of refractoriness development, althoughsignificance was not reached when IEDs alone were consid-ered. Our findings are highly similar to those published byKim et al. (1999), Spooner et al. (2006) (in children), andPittau et al. (2009), while Aguglia et al. (2011) did not obtainsignificant results.

The frontal deficits detected in our neuropsychologi-cal examination are frequently reported in TLE and areattributed to a probable disruption of the normal connec-tivity of temporal lobes with anterior structures (Martinet al., 2000). A novel finding of our study was the morefrequent presence of frontal executive deficits in refrac-tory versus non-refractory patients; although the differencedid not reach statistical significance (p = 0.008; see Table 4),which can likely be attributed to our small sample size.

In the multivariate analysis, the independent variablesassociated with refractoriness were: automatic behaviorduring seizure, structural lesion in neuroimaging (includ-ing findings compatible with HS), pathological interictal EEG(focal slowing or IED), receipt of polytherapy, and previousCNS surgery. Table 6 shows the PPV, NPV and PLR of eachsignificant diagnostic variable and of their different possiblecombinations. A PPV of 84.78% (95%CI; 73.32—96.25%) wasobserved for the combined presence of the three diagnosticvariables, when the likelihood of refractoriness to medicaltreatment would be sufficiently high to order an immedi-ate pre-surgical study of the patient. In this situation, thePPV was reduced to 72.6% (95%CI; 61.69—83.52%) if the EEGwas normal and to 73.33% (95% CI; 62.66—84.01%) if theMRI was normal. Based on these results, we constructed adecision algorithm for TLE (Fig. 2) that could represent a

Author's personal copy

120 P.J. Serrano-Castro et al.

Figure 2 Decision algorithm in temporal lobe epilepsy.

valuable approach to these patients if the present findingsare confirmed in future prospective studies.

The above values are valid assuming a pre-evaluationprevalence of around 50%, although the PLR > 5 obtained forcases with all 3 diagnostic variables can be considered toillustrate the consistency of the predictive model. This isfurther supported by the finding of a deficit in the executivefunction tests in the neuropsychological study.

A search of the literature revealed only one other reporton a similar predictive model, developed in a pediatric TLEpopulation by Dlugos et al. (2001). They obtained a pre-dictive value of 89% (95%CI; 76—96%) for RE within 2 yearsin children with the combination of a history of risk infirst infancy, structural lesion identifiable by MRI, and ini-tial treatment failure. Their study population and definitionof RE differed from our study, which is the first major inves-tigation to use the new definition proposed by the ILAE.

In conclusion, we found that the presence of automaticbehavior during seizure, a structural lesion in neuroimaging(including findings compatible with HS), and a pathologicalinterictal EEG (focal slowing or IED) are independent vari-ables that may predict refractoriness. When these threeconditions were all met, the PPV was as high as 84.44%(72.74—96.14%).

The main limitation of this study is its retrospectivecase—control design, which precludes a direct estimationof the RR. However, we were able to use the OR as a

conceptually and mathematically similar alternativebecause the appropriate conditions were met, i.e., thebase population was stable, and density sampling waspossible because cases and controls were selected simulta-neously from the same population. They all had the sameopportunity to be selected as a case or control because theyreceived the same healthcare, with identical therapeuticresources being delivered by the same professional in acontinuous manner. Nevertheless, a prospective study isrequired to establish a direct causality relationship and tovalidate a predictive model that includes the prognosticfactors identified in this study. A prospective multicenterstudy is currently under way for this purpose.

Funding

This study was supported by the grant ‘‘V Ayuda para lainvestigación neurológica de la Sociedad Andaluza de Neu-rología 2011’’ (Pedro J Serrano-Castro).

References

Aguglia, U., Beghi, E., Labate, A., Condino, F., Cianci, V.,Mumoli, L., et al., 2011. Age at onset predicts good seizureoutcome in sporadic non-lesional and mesial temporal sclerosis

Author's personal copy

Refractoriness in Temporal Lobe Epilepsy 121

based temporal lobe epilepsy. J. Neurol. Neurosurg. Psychiatry82, 555—559.

Arzimanoglou, A., Ryvlin, P., 2008. Towards a clinically mean-ingful definition of drug-resistance. In: Kahane, P., Berg, A.,Lcscher, W., Nordli, D., Perucca, E. (Eds.), Progress in EpilepticDisorder (Vol. 7): Drug Resistant Epilepsies. John Libbey Euro-text, Montrouge, France, pp. 1—6.

Baulac, S., Picard, F., Herman, A., Feingold, J., Genin, E.,Hirsch, E., et al., 2001. Evidence for digenic inheritance in afamily with both febrile convulsions and temporal lobe epilepsyimplicating chromosomes 18qter and 1q25—q31. Ann. Neurol.49, 786—792.

Berg, A.T., Vickrey, B.G., Testa, F.M., Levy, S.R., Shinnar, S.,DiMario, F., et al., 2006. How long does it take epilepsy tobecome intractable? A prospective investigation. Ann. Neurol.60, 73—79.

Berg, A.T., 2008. The natural history of mesial temporal lobeepilepsy. Curr. Opin. Neurobiol. 21, 173—178.

Berto, P., 2002. Quality of life in patients with epilepsy and impactof treatments. Pharmacoeconomics 20, 1039—1059.

Carreno, M., Donaire, A., Sánchez-Carpintero, R., 2008. Cognitivedisorders associated with epilepsy: diagnosis and treatment.Neurologist 14 (6 suppl. 1), S26—S34.

Claes, L., Audenaert, D., Deprez, L., Van Paesschen, W.,Depondt, C., Goossens, D., et al., 2004. Novel locus on chro-mosome 12q22—q23.3 responsible for familial temporal lobeepilepsy associated with febrile seizures. J. Med. Genet. 41,710—714.

Cockerell, O.C., Johnson, A.L., Sander, J.W.A.S., Shorvon, S.D.,1997. Prognosis of epilepsy: a review and further analysis ofthe first 9 years of the British National General Practice Studyof Epilepsy, a prospective population based study. Epilepsia 38,31—46.

Commission on Classification and Terminology of the InternationalLeague Against Epilepsy, 1981. Proposal for revised clinical andelectrographic classification of epileptic seizures. Epilepsia 22,489—501.

Delgado-Escueta, A.V., Swartz, B., Chauvel, P., Bancaud, J.,Walsh, G.O., Halgren, E., et al., 1992. Clinical and CCTV-EEGevaluation in presurgical work-up of temporal and frontal lobeepilepsies. Epilepsy Res. Suppl. 5, 37—54.

Dlugos, D.J., Sammel, M.D., Strom, B.L., Farrar, J.T., 2001.Response to first drug trial predicts outcome in childhood tem-poral lobe epilepsy. Neurology 57, 2259—2264.

Engel, J., 2001. A proposed diagnostic scheme for people withepileptic seizures and with epilepsy: report of the ILAE TaskForce on Classification and Terminology. Epilepsia 42, 796—803.

Engel, J.J., Wiebe, S.M., French, J.M., Sperling, M., Williamson, P.,Spencer, D., et al., 2003. Practice parameter: temporal lobeand localized neocortical resections for epilepsy: report of theQuality Standards Subcommittee of the American Academy ofNeurology, in association with the American Epilepsy Society andthe American Association of Neurological Surgeons. Neurology60, 538—547.

Engel, J., 2004. The goal of epilepsy therapy: no seizures, no sideeffects, as soon as possible. CNS Spectrums 9, 95—97.

Falip, M., Gratacós, M., Santamarina, E., Rovira, R., Padró, L.L.,2003. Factores pronósticos asociados al control médico enpacientes con evidencia radiológica de esclerosis mesial tem-poral. Rev. Neurol. 36, 501—506.

Gil-Nagel, A., Risinger, M.W., 1997. Ictal semiology in hippocam-pal versus extrahippocampal temporal lobe epilepsy. Brain 120,183—192.

Helmstaedter, C., Elger, C.E., 2009. Chronic temporal lobe epilepsy:a neurodevelopmental or progressively dementing disease? Brain132 (Pt 10), 2822—2830.

Hennessy, M.J., Elwes, R.D., Honavar, M., Rabe-Hesketh, S.,Binnie, C.D., Polkey, C.E., 2001. Predictors of outcome

and pathological considerations in the surgical treatment ofintractable epilepsy associated with temporal lobe lesions. J.Neurol. Neurosurg. Psychiatry 70 (4), 450—458.

Hitiris, N., Mohanraj, R., Norrie, J., Sills, G.J., Brodie, M.J.,2007. Predictors of pharmacoresistant epilepsy. Epilepsy Res. 75,192—196.

Jokeit, H., Ebner, A., 1999. Long term effects of refractory temporallobe epilepsy on cognitive abilities: a cross sectional study. J.Neurol. Neurosurg. Psychiatry 67, 44—50.

Kim, W.J., Park, S.C., Lee, S.J., Lee, J.H., Kim, J.Y., Lee, B.I.,et al., 1999. The prognosis for control of seizures with medica-tions in patients with MRI evidence for mesial temporal sclerosis.Epilepsia 40, 290—293.

Kwan, P., Sander, J.W., 2004. The natural history of epilepsy:an epidemiological view. J. Neurol. Neurosurg. Psychiatry 75,1376—1381.

Kwan, P., Arzimanoglou, A., Berg, A.T., Brodie, M.J., AllenHauser, W., Mathern, G., et al., 2010. Definition of drug resis-tant epilepsy: consensus proposal by the ad hoc Task Force ofthe ILAE Commission on Therapeutic Strategies. Epilepsia 51,1069—1077.

Kwong, K.L., Sung, W.Y., Wong, S.N., So, K.T., 2003. Early predictorsof medical intractability in childhood epilepsy. Pediatr. Neurol.29, 46—52.

Liimatainen, S.P., Raitanen, J.A., Ylinen, A.M., Peltola, M.A.,Peltola, J.T., 2008. The benefit of active drug trials is dependenton aetiology in refractory focal epilepsy. J. Neurol. Neurosurg.Psychiatry 79, 808—812.

Martin, R.C., Sawrie, S.M., Gilliam, F.G., Palmer, C.A., Faught, E.,Morawetz, R.B., et al., 2000. Wisconsin Card Sorting perfor-mance in patients with temporal lobe epilepsy: clinical andneuroanatomical correlates. Epilepsia 41, 1626—1632.

Mohanraj, R., Brodie, M.J., 2005. Outcomes in newly diagnosedlocalization-related epilepsies. Seizure 14, 318—323.

Perucca, E., 1998. Pharmacoresistance in epilepsy. How should it bedefined? CNS Drugs 10, 171—179.

Pittau, F., Bisulli, F., Mai, R., Fares, J.E., Vignatelli, L., Labate, A.,et al., 2009. Prognostic factors in patients with mesial temporallobe epilepsy. Epilepsia 50 (suppl. 1), 41—44.

Ramos-Lizana, J., Aguilera-López, P., Aguirre-Rodríguez, J.,Cassinello-García, E., 2009. Early prediction of refractoryepilepsy in childhood. Seizure 18, 412—416.

Schmidt, D., Stavem, K., 2009. Long-term seizure outcome ofsurgery versus no surgery for drug-resistant partial epilepsy: areview of controlled studies. Epilepsia 50, 1301—1309.

Semah, F., Picot, M.C., Adam, C., Broglin, D., Arzimanoglou, A.,Bazin, B., et al., 1998. Is the underlying cause of epilepsya major prognostic factor for recurrence? Neurology 51,1256—1262.

Spencer, S., Hugh, L., 2008. Outcomes of epilepsy surgery in adultsand children. Lancet Neurol. 7, 525—537.

Sperling, M.R., Feldman, H., Kinman, J., Liporace, J.D.,O’Connor, M.D., 1999. Seizure control and mortality in epilepsy.Ann. Neurol. 46, 45—50.

Spooner, C.G., Berkovic, S.F., Mitchell, L.A., Wrennall, J.A.,Harvey, A.S., 2006. New-onset temporal lobe epilepsy in chil-dren: lesion on MRI predicts poor seizure outcome. Neurology67, 2147—2153.

Tomson, T., Nashef, L., Ryvlin, P., 2008. Sudden unexpected deathin epilepsy: current knowledge and future directions. LancetNeurol. 7, 1021—1031.

Varoglu, A.O., Saygi, S., Acemoglu, H., Ciger, A., 2009. Prognosis ofpatients with mesial temporal lobe epilepsy due to hippocampalsclerosis. Epilepsy Res. 85, 206—211.

Wiebe, S., Blume, W.T., Girvin, J.P., Eliasziw, M., 2001. Effective-ness and Efficacy of Surgery for Temporal Lobe Epilepsy StudyGroup. A randomised, controlled trial of surgery for temporallobe epilepsy. N. Engl. J. Med. 345, 311—318.