Early versus late remission in a cohort of patients with newly diagnosed epilepsy

Authors


Address correspondence to Dr. Ettore Beghi, Istituto di Ricerche Farmacologiche “Mario Negri,” Via G. la Masa 19, 20156 Milano, Italy. E-mail: beghi@marionegri.it

Summary

Purpose:  To count patients with newly diagnosed epilepsy entering early and late remission and to identify prognostic predictors of late remission.

Methods:  Children and adults with previously untreated epilepsy from two Italian tertiary centers (Monza, Bari) were the study population. All patients received monotherapy at treatment start; drug choice and schedule were left to the physician’s judgment. A retrospective audit was performed and the following prognostic predictors were identified: age, gender, putative etiology, first electroencephalography (EEG) record, neurologic and psychiatric examination, disease duration at diagnosis, seizure type(s) and number prior to starting treatment, epilepsy syndrome, and first antiepileptic drug. Early remission was defined by 2-year seizure control immediately after treatment start. Late remission was defined by 2-year seizure control achieved at least 24 months after treatment start. Prognostic predictors were assessed by logistic regression analysis, adjusting for age, gender, and center.

Results:  One hundred seventy-four women and 178 men (mean age 31.5 years) were included and followed for 2399.6 person-years. The cumulative time-dependent probability of 2-year remission was 56.3% at 2 years after treatment start, and 62.6, 69.4, and 79.5% at 3, 5, and 10 years. One hundred fifteen patients (23.0%) achieved early remission and 38 patients (10.8%) achieved late remission. The interaction between partial seizures and number of seizures prior to treatment was the only independent predictor of late remission.

Discussion:  The course of epilepsy and the chance of remission are together a complex and dynamic process, possibly explained by the diversity of the mechanisms underlying drug response and the use of an increasing number of drugs.

The chance of seizure remission in epilepsy is an evolving concept. Epidemiologic studies indicate that 60–80% of patients with newly diagnosed epilepsy achieve remission at some time after the onset of the disease (Annegers et al.,1979; Cockerell et al., 1997; Sillanpää et al., 1998). Three different remission patterns have been recently identified: early remission (i.e., remission immediately after or in the first year after diagnosis/start of drug treatment), late remission (i.e., remission starting even several years after diagnosis/start of drug treatment), and periods of remission alternating to relapse (identifying a relapsing-remitting course) (Sillanpää & Schmidt, 2006). Early recognition of these patterns is helpful because of the differing medical and social consequences.

Early remission has been intensively investigated and a number of negative predictors have been identified. These include a documented etiology of seizures (Annegers et al., 1979; Berg et al., 2001; Jallon, 2003), the presence of electroencephalography (EEG) abnormalities (Shafer et al., 1988; Berg et al., 2001), recurrence of seizures during the first months of disease (Brorson & Wranne, 1987; Collaborative Group for the Study of Epilepsy, 1992;Sillanpää, 1993; Arts et al., 1999; MacDonald et al., 2000;Lindsten et al., 2001), and lack of response to the first antiepileptic drug (AED) (Camfield et al., 1997; Kwan & Brodie, 2000; Dlugos et al., 2001). In contrast, the prognostic indicators of late remission are still unknown. Accordingly, the aim of our study was to assess the cumulative incidence of early and late remission in a retrospective cohort of newly diagnosed patients with epilepsy, and to identify predictors of late remission.

Material and Methods

Setting and patients

The study population was included from two tertiary epilepsy centers (Bari, Monza) between January 1, 1989 and December 1, 1999, and followed up until December 31, 2005. In these centers, patients were diagnosed by experienced epileptologists and regularly seen in an outpatient clinic every 6–12 months, or earlier if clinically indicated. History taking and general and neurological examination were performed by the attending neurologist. A standard awake EEG was performed in all cases, whereas a sleep EEG, a video-EEG, or an imaging test (generally magnetic resonance imaging, MRI) was performed only when clinically indicated. Monotherapy was the recommended starting treatment in all cases, but drug choice was left to the judgment of the caring physician. In patients relapsing with the first maintenance dose, further increases were attempted to achieve seizure control or up the highest tolerated dose, whichever came first. When a first monotherapy failed, another drug was given as alternative monotherapy (preferred) or adjunctive therapy based on the physician’s choice. For each patient, baseline and follow-up clinical and demographic data were recorded on case-record forms. Treatment schedules and changes (with indications) were also reported and dated along with seizure recurrences. A patient was included at the time of treatment start, when a definite diagnosis of epilepsy (i.e., two or more unprovoked seizures) (Commission on Epidemiology and Prognosis, International League Against Epilepsy, 1993) was made or, if deemed necessary, at the time of the first unprovoked seizure. For each eligible patient, a number of variables were searched in the case-record forms by three trained abstractors (ADF, GV, and ADP) who completed a semistructured questionnaire. For each eligible case, the following prognostic predictors were identified and recorded: age at diagnosis, gender, disease duration at diagnosis, seizure type(s) and number prior to starting treatment, epilepsy syndrome, putative etiology, first EEG record, neurologic and psychiatric examinations, MRI findings, and first AED.

Definitions

Epileptic seizures and syndromes were defined according to the recommendations of the International League Against Epilepsy (ILAE) (Commission on Classification and Terminology of the International League Against Epilepsy, 1981, 1989). Etiology of seizures was documented by history, clinical examination, and/or neuroimaging findings and coded as present or absent. The EEG record was defined as normal, slow, or epileptiform (with or without slow waves). Neurologic and psychiatric examinations were coded as normal or abnormal according to clinical judgement. Disease duration was assessed as such and stratified into categories (<6 months; 6–12 months; >12 months). Seizure number at admission was counted and also indicated in categories (1; 2–5; 6+). The first AED was recorded as such, regardless of daily dose. For the purposes of this study only the most common active principles were assessed separately.

A period of remission was defined as two or more years of complete seizure control after treatment start. Early remission was defined as a 2-year seizure control immediately after initial antiepileptic treatment. Late remission was defined as a 2-year seizure control achieved at least 24 months after treatment start. Patients who were not controlled immediately but achieved a 2-year seizure control within 2 years were excluded from further analysis. In patients with more than one period of remission, only the first ever episode was considered. All these variables were assessed by an independent abstractor who discussed any difficulty with the treating physician.

Statistics

The 2-year cumulative time-dependent probability of remission was calculated with the Kaplan-Meier method (actuarial analysis). Each prognostic predictor was assessed in patients with early versus late remission and tested with the chi-square for heterogeneity or for trend, as appropriate. Prognostic predictors with significantly different distribution on univariate analysis were also assessed by multiple logistic regression analysis with forward stepwise likelihood ratio, adjusting for age, gender, and center. Any p-value <0.05 was considered statistically significant. Statistical computations were done using SPSS for Windows, release 13.0 (SPSS Inc., Chicago, IL, U.S.A.).

Results

Demographics and baseline features

As shown in Table 1, the sample included 174 women and 178 men aged 3–84 years at diagnosis (mean 31.5 years). One hundred seventy-two patients were from Bari and 180 from Monza. Partial seizures and partial epilepsy syndromes were the predominant categories. Sixteen percent of cases had single seizures at entry. Mean disease duration at diagnosis at the epilepsy centers was 35.5 months (range 0–66). The number of seizures before treatment start was mostly between 2 and 5 (45.7%). Neurologic and psychiatric examinations were abnormal in 16.5% and 9.1% of cases, respectively. EEG at entry was epileptiform in 159 cases (48.9%). The most common first AED was carbamazepine (47.2%), followed by valproate (28.4%) and phenobarbital (13.4%) (Table 2).

Table 1.   General characteristics of the sample at admission (N = 352)
VariableNo. cases%
  1. EEG, electroencephalography; MRI, magnetic resonance imaging; NA, not available.

  2. aBased on history, neurologic examination, and MRI findings.

  3. bNot done in 54.

Gender
 Women17449.4
 Men17850.6
Age (years)
 <153710.5
 15–3520959.4
 35–645515.6
 >655114.5
Family history of epilepsy/seizures
 No28079.5
 Yes6518.5
 NA7 
Documented etiologya
 No23479.6
 Yes6020.4
 NA58 
Disease duration at center diagnosis (months)
 <614742.1
 6–124111.7
 >1216146.1
 NA3 
Number of seizures before treatment
 15515.7
 2–516146.0
 >513438.3
 NA2 
Seizure type
 Partial22664.2
 Generalized12334.9
 Undetermined30.9
Epilepsy syndrome
 Partial idiopathic 339.6
 Partial symptomatic4613.4
 Partial cryptogenic11533.5
 Generalized idiopathic7020.4
 Undetermined216.1
 Isolated seizure5816.9
 NA9 
Neurologic examination
 Normal29483.5
 Abnormal5816.5
Psychiatric examination
 Normal32090.9
 Abnormal329.1
First EEG record
 Normal or aspecific12337.8
 Slow4313.2
 Epileptiform15948.9
 NA27 
MRIb
 Normal19167.7
 Abnormal9132.3
 NA7019.8
Table 2.   First antiepileptic drug: target daily dose (median and range)
DrugNumber of cases%Median dose (mg)Range (mg)
  1. aTarget dose unknown in three cases.

  2. bTarget dose unknown in one case.

Carbamazepinea16647.2800200–2,200
Valproate10028.41,000100–2,800
Phenobarbital4713.410050–250
Phenytoin144.0300100–400
Vigabatrin92.62,0002,000–2,500
Lamotrigine92.6300150–600
Primidone30.820050–300
Oxcarbazepineb20.6600
Clonazepam10.220
Topiramate10.2200

Seizure outcome

Patients were followed for 2399.6 person-years (mean 82.7 months; median 75.0 months). Two hundred twenty patients (62.5%) entered one or more 2-year periods of seizure freedom: 115 patients (32.7%) achieved an early remission and 38 patients (10.8%) achieved late remission. Of these, 101 (87.8%) and 34 (89.5%), respectively, were in remission at last follow-up. Using actuarial analysis, the cumulative time-dependent probability of 2-year remission was 56.3% at 2 years after treatment start, and 62.6%, 69.4%, and 79.5% at 3, 5, and 10 years (Fig. 1). The median period of time needed to achieve 2-year remission was 16 months (range 0–138 months).

Figure 1.

 Cumulative time-dependent probability of 2-year remission in the study cohort.

Prognostic factors

In univariate analysis (Table 3), significant predictors of late remission included partial seizures, six or more seizures prior to treatment, and abnormal (mostly epileptiform) EEG at diagnosis. In multivariate analysis, the only predictor of late remission was the interaction between seizure type and number prior to starting treatment. The odds ratio (OR) of late remission was 2.7 [95% confidence interval (CI) 1.0–6.8] in patients with 2–5 partial seizures and 6.7 (95% CI 2.3–19.3) in patients with more than five partial seizures. Gender, age, documented etiology, first EEG record, neurologic and psychiatric examinations, disease duration at diagnosis, epilepsy syndrome, and first AED were nonsignificant in the logistic regression analysis model. The results were fairly similar when a subanalysis was performed for adults only (i.e., patients aged 18 years or older) (data not shown).

Table 3.   Patients with early remission (N = 115) versus late remission (N = 38) by prognostic factor (univariate analysis)
 Early remission, N (%)Late remission, N (%)p-value
  1. EEG, electroencephalography; MRI, magnetic resonance imaging; NA, not available; NS, nonsignificant

  2. ap-values obtained from chi-squares for trend.

  3. bBased on history, neurological examination and MRI findings.

  4. cIncludes secondary generalized seizures.

  5. dNot done in 27 cases.

Gender
 Women47 (40.9)19 (50.0)NS
 Men68 (59.1)19 (50.0)
Age (years)
 <1510 (8.7)6 (15.8)NSa
 15–3469 (60.0)23 (60.5)
 35–6417 (14.8)6 (15.8)
 >6419 (16.5)3 (7.9)
Family history of epilepsy/seizures
 No90 (79.6)30 (78.9)NS
 Yes23 (20.4)8 (21.1)
 NA2
Documented etiologyb
 No64 (78.0)28 (87.5)NS
 Yes18 (22.0)4 (12.5)
 NA336
Disease duration at diagnosis (months)
 <651 (44.3)15 (39.5)NS
 6–1217 (14.8)3 (7.9)
 >1247 (40.9)20 (52.6)
Number of seizures before treatment
 124 (21.1)6 (15.8)0.046a
 2–565 (57.0)16 (42.1)
 >525 (21.9)16 (42.1)
 NA1
Seizure type
 Partialc63 (54.8)30 (78.9)0.014
 Generalized51 (44.3)7 (18.4)
 Undetermined1 (0.9)1 (2.6)
Epilepsy syndrome
 Partial idiopathic7 (8.4)3 (7.5)0.050
 Partial symptomatic11 (13.3)4 (10.0)
 Partial cryptogenic32 (28.9)21 (52.5)
 Generalized idiopathic27 (22.9)5 (12.5)
 Undetermined10 (6.0)2 (5.0)
 Isolated seizure25 (20.5)3 (12.5)
 NA3
Neurologic examination
 Normal99 (86.1)30 (79.9)NS
 Abnormal16 (13.9)8 (21.1)
Psychiatric examination
 Normal107 (93.0)35 (92.1)NS
 Abnormal8 (7.0)3 (7.9)
First EEG record
 Normal or aspecific48 (43.2)8 (22.2)0.049a
 Slow11 (9.9)6 (16.7)
 Epileptiform52 (46.9)22 (61.1)
 NA42
MRI at entryd
 Normal65 (71.4)23 (74.2)NS
 Abnormal26 (28.6)8 (25.8)
 NA247

Discussion

In our study, late remission was documented in about 10% of patients with newly diagnosed epilepsy followed for a prolonged period. Other reports from population or referral cohorts of patients with apparently drug-resistant epilepsy provide consistent findings on the chance of late remission, even if with different results. Camfield et al., (1997) found that 42% of 72 children failing to respond to the first AED later achieved remission. In a prospective cohort of 613 children followed for a median of 9.7 years, 83 met a stringent definition of intractable epilepsy, defined as failure of two drugs and 1 seizure/month (average) for 18 months (Berg et al., 2006). Of these, 20.5% subsequently entered remission and 13.3% were seizure free at last contact. Late remission (with a mean delay of 9 years) was achieved in 50% of adults with childhood-onset epilepsy (Sillanpää & Schmidt, 2006). In a cohort of patients with uncontrolled chronic epilepsy, 28% of cases had been rendered seizure-free (i.e., complete seizure control for 12 months or longer) during follow-up (Luciano & Shorvon, 2007). The observation of patients entering remission after prolonged periods of active epilepsy indicates that initial failure to enter remission is not always a reliable indicator of long-term failure.

Our results have several implications. First of all, patients experiencing true drug resistance can be identified only after prolonged periods of follow-up. Studies done in population-based cohorts showed that the percentage of patients achieving complete seizure remission was 21% at 4 years (Ecuador), 62% at 5 years (France), 55% at 7 years (Canada), and 68% at 10 years or longer (Switzerland, Finland) (Jallon, 2003). Second, the possibility to enter late remission is against the interpretation of drug resistance as a predetermined, perhaps genetically based, clinical condition. This theory claims that in most cases pharmacoresistance, which is intended as a failure of several drugs to reach the receptor sites to exert their action (Schmidt & Loscher, 2005), has been fully developed before the start of the first AED and even before the first seizure. Third, population-based cohorts followed for several years from the time of the diagnosis of epilepsy present a continuous increase in the number of patients entering remission, albeit at a slower rate, even after five or more years of seizure recurrence (Annegers et al., 1979; Cockerell et al., 1997). Fourth, late remissions may be explained at least in part by the increasing number of new pharmaceutical compounds, each showing remission rates in some patients with chronic active epilepsy recruited in regulatory trials.

The second most important finding in our study was the interaction between seizure type and number in predicting late remission. Compared to generalized seizures, partial seizures have been found to be associated with lower remission rates (Annegers et al., 1979; Cockerell et al., 1997; Arts et al., 1999; MacDonald et al., 2000). However, as shown in this study and confirmed by others (Berg et al., 2006), delayed remission can be observed in patients with focal epilepsies. This finding may be part of a relapsing–remitting pattern (Sillanpää & Schmidt, 2006), which suggests that early remission or intractability may not be enduring outcomes.

The number of seizures experienced by the patient before intake or the onset of treatment or during the first months after treatment start has been consistently found to predict seizure outcome (Brorson & Wranne, 1987; Collaborative Group for the Study of Epilepsy, 1992; Sillanpää, 1993; Arts et al., 1999; MacDonald et al., 2000; Lindsten et al., 2001). A failure to achieve early seizure control has been thought to render epilepsy more resistant to treatment, perhaps because of structural changes in the brain (Reynolds, 1987;Reynolds, 1995). Another explanation is that epilepsy has an inherent severity and response to treatment (Sander, 1993; Chadwick, 1995; Shinnar & Berg, 1996). However, in our study a higher number of partial seizures prior to treatment predicted late remission rather than early remission. This encouraging finding can be explained at least in part by the increasing number of different AEDs for partial seizures now available.

Other factors were not found to influence the outcome in our population, despite past evidence of a prognostic role of age, epilepsy syndromes, or etiology (Jallon, 2003). The study population and design may explain the differing results.

Our study has some limitations. First, ours is a study done in referral patients: eligible patients were referred to tertiary centers either by their general practitioners or by first/second level hospitals. Therefore, selection bias may affect our population, as previously indicated. Second, data were collected retrospectively. Therefore, a predefined standard classification of our putative prognostic factors was impossible. Third, this is mostly an adult sample, and these results may not be applied to a pediatric population. Fourth, a large proportion of our cases had cryptogenic partial epilepsies. Although the syndromic diagnosis was directed by the EEG findings and in several of them with uncontrolled seizures neuroimaging studies were repeated, we cannot exclude that a symptomatic etiology would go undetected in some cases. Fifth, assignment of first AED was not randomized, which prevents us from driving conclusions on the comparative efficacy of drugs assigned at the time of diagnosis. Last, at the time of the opening of the study many new AEDs were not available. This precludes their role on the chance of seizure remission (whether early or late) in newly diagnosed patients with epilepsy.

In conclusion, compared to the rest of our study population, patients with 2–5 partial seizures before starting treatment had a 2.7 chance of late remission. The risk increased to 6.7 in those with 6+ partial seizures. These findings suggest that the outcome of epilepsy and the chance of seizure remission is a fairly complex and dynamic process. This may be explained by the diversity of the mechanisms underlying the response to AEDs. However, the possibility that some drugs used later in the course of the disease play a role cannot be excluded. Because the study was retrospective, we have been unable to investigate more deeply treatment changes and patients’ compliance. This will be the object of future studies.

Acknowledgments

We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

None of the authors has any conflict of interest to disclose.

Ancillary