JLE

Epileptic Disorders

MENU

Measurement of seizure freedom in adjunctive therapy studies in refractory partial epilepsy: the levetiracetam experience Volume 8, numéro 2, June 2006

Auteur(s) : Ilo Leppik1, Katrien De Rue2, Pascal Edrich2, Emilio Perucca3

1University of Minnesota and MINCEP Epilepsy Care, Minneapolis, Minnesota, USA
2UCB S.A., Brussels, Belgium
3Institute of Neurology IRCCS C. Mondino Foundation, Pavia, and Clinical Pharmacology Unit, University of Pavia, Pavia, Italy

Article reçu le 17 Octobre 2005, accepté le 28 Février 2006

Seizure freedom is the ultimate goal of epilepsy treatment: it reduces morbidity and mortality, prevents sudden death from epilepsy, and improves quality of life (Sander and Bell, 2004). Unfortunately, as recently shown in a comprehensive study of 1652 people with epilepsy, a large number of patients (48.3% in this survey) failed to achieve seizure freedom during treatment with an antiepileptic drug (AED) (Moran et al., 2004). This underscores the need for additional research and better treatment paradigms. Regulatory trials have some inherent limitations, such as less diverse populations than in the clinical setting, with fewer comorbid conditions and concomitant medications. They also tend to be only three to six months in duration. However, because of their randomized and placebo-controlled design, they are often the first objective report of a new drug’s clinical efficacy and tolerability.Calculating seizure freedom rates relative to a comparator is an ideal way to document efficacy of a new AED. However, adjunctive therapy trials designed to gain regulatory approval of new AEDs are usually conducted in patients resistant to mono- or polytherapy. Because this population is particularly refractory to drug therapy, only a few of these patients can be expected to become seizure-free. Although placebo-controlled add-on studies of new AEDs are not designed to compare complete seizure control rates, the proportion of patients achieving seizure freedom during the double-blind period and subsequent follow-up represents a clinically meaningful parameter and should be reported.Seizure freedom may appear to be a dichotomous variable: a patient is either seizure-free, or a patient is not free of seizures. However, the duration of the period free of seizures is an important aspect of the measurement; seizure freedom for three years is a more profound effect than seizure freedom for three months. Therefore, with no standard definition, this metric varies widely among trials, and a consensus is not yet available. When comparing an AED in a double-blind way to placebo the duration of seizure freedom assessed should be the whole double-blind period (up-titration period included or not), which is usually around three months. For open-label studies the duration of evaluation can be much longer. In this paper, we use data from double-blind, placebo-controlled studies and long-term follow-up evaluations of levetiracetam (LEV). Pooled data were used because seizure freedom often has low statistical power due to its low event rate. This paper reports the assessment of the impact of different methodologies (( figure 1 )) on seizure freedom results, compares these outcomes with available reports from the literature for other new AEDs, and proposes standard methods for reporting seizure-free data in future studies.

Review of published trials with new AEDs and seizure freedom

We conducted a literature search on MEDLINE in an attempt to obtain seizure freedom data for new AEDs in short-term add-on clinical trials. The search focused on gabapentin (GBP), lamotrigine (LTG), oxcarbazepine (OXC), pregabalin (PGB), tiagabine (TGB), topiramate (TPM), and zonisamide (ZNS), and included a search term for epilepsy. The search focused on double-blind randomized placebo-controlled add-on trials in adults with partial epilepsy. Information was also extracted from a recent comprehensive textbook (Levy et al. 2002). Only papers in English were included. The following studies were excluded as not meeting the search criteria: open-label or observer-blinded studies, monotherapy in newly diagnosed patients, presurgical evaluations, studies of children or healthy volunteers, studies in which primary generalized seizures were the only inclusion criterion, and comparative trials without a placebo control. Meta-analyses were not included, as none were based on seizure freedom (Chaisewikul et al., 2002, Marson et al., 2001). Studies that did not measure seizure frequency or that did not include efficacy as a primary outcome were also excluded. Thus, this analysis was limited to double-blind, placebo-controlled add-on trials which included seizure frequency as an efficacy parameter. We found 47 publications meeting these criteria (table 1( Table 1 )).

Seizure freedom rates were reported in few of these publications, and when they were reported, were accompanied by little or no information describing the methodology for calculation. Overall, six of seven identified GBP publications, nine of 13 LTG publications, one of two OXC publications, two of four PGB publications, four of six TGB publications, and five of 13 TPM publications did not provide any data on seizure freedom. As confirmation of this lack of information, meta-analysis type reviews of several AEDs have reported odds ratios, relative risks, or number needed to treat for 50% responder rates, but none has reported these parameters based on seizure freedom (Chaisewikul et al., 2002, Marson et al., 2001).

Seizure freedom, defined as 100% reduction from baseline seizure frequency, is arguably the most (and arguably the only) meaningful clinical measure in determining an AED’s efficacy. What changes is the treatment period (over the stable dose period, or over the titration and stable dose period or some other defined period) and the population (withdrawals may be counted as seizure-free or not seizure-free, and the intent-to-treat (ITT) population or only completers may be used). When one reviews the literature, one sees how the lack of a clear definition makes it difficult to interpret the reported data.

The AED with the most published seizure freedom data is TPM. Three publications reported pooled seizure freedom rates of 4% to 5% with TPM, as compared to 0% to 1% with placebo (Reife and Pledger, 1997, Reife et al., 2000, Peeters et al., 2003). In the Korean Topiramate Study, seizure freedom in the ITT population was reported to be significantly higher with TPM than placebo (7.9% versus 1.2%, p = 0.04). In all of these publications, no information was provided about how withdrawals were handled, about duration of seizure freedom, and whether or not the up-titration period was included (Korean Topiramate Study Group, 1999). TPM also had a higher reported seizure freedom rate than placebo in a study of add-on therapy to carbamazepine (6% versus 2%) (Guberman et al., 2002). In this study, higher seizure freedom rates were reported, but only during a 4-week maintenance period, and the methodology used for the seizure freedom analysis was not described further. Two other publications reported zero patients out of 60 and two patients out of 47 being seizure-free with TPM, respectively (Tassinari et al., 1996, Sharief et al., 1996), and a third reported results for secondarily generalized seizures only in patients who had such seizures at baseline (Faught et al., 1996). Obviously, the latter analysis provides higher seizure freedom rates than those considering all seizure types.

Seizure-free rates ranging from 3% to 22% were reported for OXC at doses of 600 to 2 400 mg/d as compared to 0.6% with placebo (Barcs et al., 2000). However, it is very difficult to draw conclusions from this apparently high seizure freedom rate, because no information was provided about the methodology used to define seizure freedom. For example, 174 patients were randomly assigned to the highest dose of OXC, but 128 patients (74%) discontinued treatment. It is unclear how long the duration of the evaluation was and how seizure-free patients discontinuing treatment were considered in the seizure-freedom analysis.

Two ZNS publications reported seizure freedom rates of approximately 4% to 6% as compared to 1% to 3% for placebo (Faught et al., 2001, Schmidt et al., 1993). In one publication (Schmidt et al., 1993), the analysis included the ITT population, but no information was given on the handling of withdrawals, whereas information about methodology was not provided in the other (Faught et al., 2001). For TGB, one publication reported that two of 77 patients (2.6%) in the TGB group were seizure-free for the 12-week fixed-dose period, as compared to none of the 77 patients in the placebo group (Kalviainen et al., 1998). In this publication and one other, information was provided on the number of seizure-free days, but this data cannot be used to calculate seizure freedom rates for the whole study (or stable dose) period (Kalviainen et al., 1998, Ben-Menachem, 1995).

Two LTG studies reported seizure freedom during one month of treatment or during five consecutive weeks of double-blind treatment, but neither reported seizure-freedom rates for the entire study period (Matsuo et al., 1993; Stolarek et al., 1994). Seizure-free days were reported in a placebo-controlled crossover study, but none of the patients became seizure-free while receiving LTG (Sander et al., 1990). In a publication on patients with simple partial seizures, more patients in the placebo group than in the LTG group became seizure-free (five of 26, or 19%, versus three of 30, or 10%) (Boas et al., 1996). Incomplete information is also available about seizure freedom with GBP and PGB.

In one GBP publication, anecdotal information was provided for one patient who was reported to be seizure-free for the entire 2-month treatment period while receiving GBP (Crawford et al., 1987).

One PGB publication considered three randomized, placebo-controlled, 12-week studies and reported that across the dose range of 150-600 mg/d, 3% to 17% of patients were seizure-free, but only for the last 28 days of treatment (Brodie, 2004). Additionally, a post hoc analysis was conducted on a restricted ITT population of patients who received > 28 days of treatment and completed > 75% of the daily seizure diary. Using this restricted ITT population, both the 300 and 600 mg/d dose were significantly more effective than placebo, with between 7% and 19% of patients reporting complete freedom from seizures over the 12-week period. No information about the way withdrawals were handled and about how many seizure-free patients withdrew from the studies was provided.

More recently, PGB was evaluated in an international multicenter, 12-week, double-blind placebo-controlled study (Arroyo et al., 2004). The percentage of seizure-free patients was determined using Fisher’s exact test. Again, the seizure-free period was defined as the last 28 days of treatment. This calculation showed 7% and 12% of patients in the 150-mg and 600-mg groups, respectively, to be seizure-free, compared with just 1% in the placebo group (p = 0.065 and p = 0.002 versus placebo, respectively).
Table 1 Summary of seizure freedom data for new antiepileptic drugs.

Drug

Publications

Did not report seizure

(Total = 47)

freedom rates

(Total = 28)

Gabapentin

US Gabapentin Study Group, 1990

Sivenius et al., 1991

Chadwick, 1991

US Gabapentin Study Group No. 5, 1993

Anhut et al., 1994

Fisher et al., 2001

Crawford et al., 1987

Lamotrigine

Binnie et al., 1987

Binnie et al., 1989

Jawad et al., 1989

Loiseau et al., 1990

Smith et al., 1993

Schapel et al., 1993

Messenheimer et al., 1994

Pellock, 1994

Beran et al., 1998

Matsuo et al., 1993

Stolarek et al., 1994

Sander et al., 1990

Boas et al., 1996

Oxcarbazepine

Schmidt and Sachdeo, 2000

Barcs et al., 2000

Pregabalin

Brodie, 2004

French et al., 2003

Miller et al., 2003

Arroyo et al., 2004

Tiagabine

Richens et al., 1995

Loiseau, 1999

Sachdeo et al., 1997

Uthman et al., 1998

Kalviainen et al., 1998

Ben-Menachem, 1995

Topiramate

Ben-Menachem et al., 1996

Privitera et al., 1996

Faught, 1997

Ben-Menachem, 1997

Yen et al., 2000

Reife and Pledger, 1997

Reife et al., 2000

Peeters et al., 2003

Korean Topiramate Study Group, 1999

Guberman et al., 2002

Tassinari et al., 1996

Sharief et al., 1996

Faught et al., 1996

Zonisamide

Faught et al., 2001

Schmidt et al., 1993

Experience from levetiracetam trials: methodological issues in assessing rates of seizure freedom

The calculation of seizure freedom rates in clinical trials is subject to several pitfalls, mainly that the results can be biased by the methodology chosen. For this discussion, an inferential ITT population, including patients and data from three pivotal studies of add-on LEV in partial-onset epilepsy, is used. Data from three pivotal trials, two parallel-group studies (Cereghino et al., 2000, Ben-Menachem and Falter, 2000) and a crossover study (Boon et al., 2002, Shorvon et al., 2000), conducted in Europe and the USA were analyzed and included in a data set of 1088 subjects (placebo, n = 367; LEV 1000 mg/d, n = 277; LEV 2000 mg/d, n = 175; and LEV 3000 mg/d, n = 269). Further details are shown in table 2( Table 2 ). Inferential ITT was chosen consistent with previous methodology (Boon et al., 2002). Specifically, the inferential ITT population includes patients from both parts of and with non-missing data for partial-onset seizure frequency per week in the baseline and stable dose periods (n = 1088, all three studies). Patients who withdrew from a study before entering a stable dose period were therefore excluded (placebo, n = 72; LEV, n = 103), but patients withdrawing during the stable dose period were included. Patients in the crossover study were randomly assigned to different treatments in each part of the study. Each patient is counted once under the relevant treatment in these analyses. Seizures reported during a stable dose period are attributed to the treatment administered during that period. This approach to counting patients from the crossover study results in treating each as if they were two independent subjects.

Baseline demographic and clinical characteristics described in table 2 indicate the severity of the disease in the ITT population. Patients had epilepsy for a mean of > 20 years and had a median baseline seizure frequency per week ranging from 1.9 to 2.6 depending on the treatment arm.

Six different methods for calculating seizure freedom were applied to these data (( figure 1 )). The first method (Method 1) defines seizure freedom as the absence of seizures during the stable dose period, irrespective of whether the patient completed the stable dose period or not. For this calculation, the denominator is the total number of subjects in the inferential ITT population for that treatment group. According to this approach, completion status during the stable dose phase does not influence the definition of seizure freedom. In particular, patients who did not complete the stable dose period but did not have any seizures up to the time of withdrawal are considered to be seizure-free. The percentage of patients meeting the definition of seizure freedom by Method 1 was 0.8% in the placebo group, 4.7% in the LEV 1000-mg group, 6.3% in the LEV 2000-mg group, and 8.6% in the LEV 3000-mg group (( figure 2 )).

A two-sided Fisher exact test was used to make inferential comparisons between each randomized LEV dose group and placebo (table 3( Table 3 )). The seizure freedom odds ratios versus placebo ranged from 6.0 to 11.3 with LEV 1000 to 3000 mg as compared with placebo (p = 0.003 for LEV 1000 mg and p < 0.001 for LEV 2000 and 3000 mg). The inferential analysis underestimates the variability expected from a sample of the same size comprised of real independent subjects, in that the two parts of the crossover trial were counted as if they corresponded to two independent patients.

Table 4( Table 4 ) describes gender differences, age, age at epilepsy onset, and duration of epilepsy in both seizure-free patients and patients not seizure-free (as per Method 1).

An alternative method (Method 2) defines seizure freedom as the absence of seizures during both the titration and stable dose periods for the inferential ITT population. Withdrawals are handled in the same way as with Method 1: seizure-free patients discontinuing during the stable dose period before the end of the study are counted as seizure-free. Therefore, the only difference is that the titration period is also included in the calculation. As shown in table 3, the number of seizure-free patients was lower with Method 2 than with Method 1 in each of the groups. Nevertheless, the data show that the odds ratio versus placebo for seizure freedom was about 11 with LEV 2000 mg and 3000 mg as compared with placebo (p < 0.001).

Method 2 is a more conservative approach: by including the titration period, it offers the advantage of measuring efficacy from the first day of treatment and thus determining whether a drug has a rapid onset of action. However, a disadvantage of Method 2 in analyzing LEV trials is that some patients received placebo during the first 2 weeks of the titration period. The patients who had a seizure while receiving placebo or a low, subtherapeutic LEV dose during titration are not considered seizure-free even when they had no seizures during the stable dose period. In the inferential ITT population, 112 out of 183 patients randomized to receive LEV 1000 mg/d during one of the periods of the crossover study received placebo during the first 2 weeks of the titration period (Boon et al., 2002). In the other studies, there was no placebo during titration but the titration was done gradually, either to 333 mg/d, 666 mg/d, and then 1000 mg/d over 4 weeks (Cereghino et al., 2000) or to 1000 mg/d, 2000 mg/d, and then 3000 mg/d (Cereghino et al., 2000, Ben-Menachem and Falter, 2000). Consequently, the percentage of seizure-free patients in the other groups is also expected to be lower during titration than during the stable dose period, because patients received lower doses of LEV during titration. For placebo-controlled trials, this could be an argument for restricting the assessment of seizure freedom to the stable dose period (Method 1). In clinical trials directly comparing two or more AEDs, however, measuring seizure freedom from the first drug dose, including the titration period (Method 2), will reveal differences between a drug reaching therapeutic doses quickly and a drug needing to be uptitrated slowly. The clinical relevance of stopping seizures as soon as possible is evident.

Seizure freedom rates may be significantly impacted by all methods that allow investigators to define as seizure-free those patients who did not complete the entire treatment period. This could result in artificially inflated seizure-free rates: for example, with Method 1, which considers all patients who discontinue prematurely while being seizure-free as “successes”, there is clearly a bias in favor of poorly tolerated AEDs, because patients would not have necessarily remained seizure-free had they not withdrawn. A highly conservative approach to circumvent this problem is to consider all patients who did not have seizures during the stable dose phase but did not complete such phase as “failures”; i.e., these patients are not considered to be seizure-free (Method 3). In the case of LEV, however, using the inferential ITT population, Methods 1 and 3, both assessing the stable dose period, yielded almost identical results. In the placebo and 1000 and 2000 mg LEV groups, no seizure-free patients withdrew. In the 3000 mg LEV group, only one seizure-free patient withdrew before the end of the stable dose period. Assuming that this patient had remained seizure-free (Method 1), the percentage of seizure-free patients would be 8.6% (( figure 2 )), whereas considering this patient as a failure (Method 3) would yield a seizure freedom rate of 8.2% (table 3). Obviously, the difference between these two rates would be much larger for a drug (or dose) with higher withdrawal rates due to poor tolerability. The finding of similar results obtained with Methods 1 and 3 provides indirect evidence that all drug doses were well tolerated.

A fourth method is the most conservative of all, in that it defines seizure freedom as the absence of seizures during both the titration and the stable dose period (as in Method 2), and considers seizure-free non-completers as failures, as in Method 3. For LEV, using the inferential ITT population, Method 2 and Method 4 yielded identical seizure freedom results (table 3). This finding is consistent with the comparison between Methods 1 and 3.

A fifth approach (Method 5) disregards the inferential ITT population and considers as denominator only the population of patients who complete the stable dose period. In this method, the number of seizure-free patients is the same as determined using Method 3 (i.e., only completers with no seizures during the stable dose phase are considered seizure-free), but the number of patients used in the denominator becomes smaller because only completers are considered. In studies with large numbers of patient discontinuations, the reduction in the denominator can have a significant impact on seizure-free rates. This method can be seriously biased, because it uses the subset of subjects who finish the trial, and excludes those who may have withdrawn because of lack of efficacy or low tolerability. If Method 5 is used, then an analysis based on Method 3 should also be provided: if there is a large difference between the two methods, the analysis suggests that the drug is poorly tolerated. In the LEV trials, the largest difference in seizure-free rates between Method 5 and Method 3 was seen in the 3000-mg group (table 3), but was very minor (8.9% versus 8.2%, respectively). In fact, the number of patients who discontinued was 24 in the placebo group, 20 in the LEV 1000-mg group, 12 in the LEV 2000-mg group, and 22 in the LEV 3000-mg group. An approach similar to Method 5, using completers in the denominator, can also be applied by defining as seizure-free only those patients who complete the trial without seizures during both the titration and the stable dose period (Method 6). In this case, consistency of results should be sought by comparing Method 6 with Method 4.
Table 2 Baseline demographic and clinical characteristics in the inferential intent-to-treat (ITT) population of patients included in placebo-controlled adjunctive therapy trials of levetiracetam (LEV) in refractory partial epilepsy.

Placebo

LEV 1000 mg

LEV 2000 mg

LEV 3000 mg

(n = 367)

(n = 277)

(n = 175)

(n = 269)

Gender, n (%)

Female

179 (48.8)

121 (43.7)

91 (52.0)

121 (45.0)

Male

188 (51.2)

156 (56.3)

84 (48.0)

148 (55.0)

Age, years: mean (± SD)

37.0 (11.4)

37.4 (11.1)

36.7 (11.9)

37.0 (11.2)

Age at epilepsy onset, years: mean (± SD)

15.2 (12.2)

14.1 (11.7)a

13.8 (11.3)

16.2 (13.0)

Duration of epilepsy, years: mean (± SD)

22.3 (11.7)

23.9 (12.3)a

23.4 (13.6)

21.4 (12.1)

Baseline seizure frequency/week: median (Q1-Q3)

2.0 (1.3-3.9)

2.6 (1.6-5.7)

2.6 (1.6-5.0)

1.9 (1.1-3.5)

aOne subject was unevaluable for the age at epilepsy onset and duration of epilepsy. For those two variables, the denominator in thus 276.


Table 3 Seizure freedom estimates based on method of analysis in placebo-controlled adjunctive therapy trials of levetiracetam (LEV) in patients with refractory partial epilepsy.

Method

Placebo

LEV 1000 mg

LEV 2000 mg

LEV 3000 mg

Method 1a

Seizure-free patients, % (n/total)

0.8 (3/367)

4.7 (13/277)

6.3 (11/175)

8.6 (23/269)

Odds ratio (95% CI)

6.0 (1.6–32.9)

8.1 (2.1-45.8)

11.3 (3.4-59.5)

p value versus placebo

0.003

<0.001

<0.001

Method 2b

Seizure-free patients, % (n/total)

0.5 (2/367)

2.2 (6/277)

5.7 (10/175)

5.9 (16/269)

Odds ratio (95% CI)

4.0 (0.7–41.2)

11.1 (2.3–104.5)

11.5 (2.7–104.0)

p value versus placebo

0.081

<0.001

<0.001

Method 3c

Seizure-free patients, % (n/total)

0.8 (3/367)

4.7 (13/277)

6.3 (11/175)

8.2 (22/269)

Odds ratio (95% CI)

6.0 (1.6–32.9)

8.1 (2.1-45.8)

10.8 (3.2-56.8)

p value versus placebo

0.003

<0.001

<0.001

Method 4d

Seizure-free patients, % (n/total)

0.5 (2/367)

2.2 (6/277)

5.7 (10/175)

5.9 (16/269)

Odds ratio (95% CI)

4.0 (0.7–41.2)

11.1 (2.3–104.5)

11.5 (2.7–104.0)

p value versus placebo

0.081

<0.001

<0.001

Method 5e

Seizure-free patients, % (n/total)

0.9 (3/343)

5.1 (13/257)

6.7 (11/163)

8.9 (22/247)

Odds ratio (95% CI)

6.0 (1.6–33.3)

8.2 (2.1–46.2)

11.1 (3.3–58.3)

p value versus placebo

0.003

<0.001

<0.001

Method 6f

Seizure-free patients, % (n/total)

0.6 (2/343)

2.3 (6/657)

6.1 (10/163)

6.5 (16/247)

Odds ratio (95% CI)

4.1 (0.7-41.5)

11.1 (2.3-105.3)

11.8 (2.7-106.5)

p value versus placebo

0.079

<0.001

<0.001

aMethod 1: Seizure freedom defined by absence of seizures during the stable dose period. Patients seizure-free at time of withdrawal are counted as seizure-free.

bMethod 2: Seizure freedom defined by absence of seizures during the titration plus stable dose period for patients reaching the stable dose period. Patients seizure-free at time of withdrawal are counted as seizure-free.

cMethod 3: Seizure freedom defined by absence of seizures during the stable dose period. Withdrawals are counted as not seizure-free.

dMethod 4: Seizure freedom defined by absence of seizures during the titration plus stable dose period for patients reaching the stable dose period. Withdrawals are counted as not seizure-free.

eMethod 5: Seizure freedom defined by absence of seizures during the stable dose period for completers, using completers as denominator.

fMethod 6: Seizure freedom defined by absence of seizures during the titration plus stable dose period for completers, using completers as denominator.


Table 4 Demographic and clinical characteristics in the inferential intent-to-treat (ITT) population included in placebo-controlled adjunctive therapy trials of levetiracetam (LEV) in patients with refractory partial epilepsy, classified according to their seizure freedom status in the trial (calculated according to Method 1a).

Placebo

LEV 1000 mg

LEV 2000 mg

LEV 3000 mg

n = 367

n = 277

n = 175

n = 269

Gender: female/male (%/%)

Seizure-free

33.3/66.7

46.2/53.8

45.5/54.5

34.8/65.2

Not seizure-free

48.9/51.1

43.6/56.4

52.4/47.6

45.9/54.1

Age, years: mean (±SD)

Seizure-free

30.7 (9.5)

35.1 (9.6)

31.2 (13.7)

41.8 (11.8)

Not seizure-free

37.0 (11.4)

37.6 (11.2)

37.1 (11.7)

36.6 (11.1)

Age at epilepsy onset, years: mean (±SD)

Seizure-free

21.2 (14.7)

18.4 (12.8)

12.0 (7.2)

19.9 (15.0)

Not seizure-free

15.2 (12.2)

13.8 (11.6)

13.9 (11.5)

15.8 (12.8)

Duration of epilepsy, years: mean (±SD)

Seizure-free

10.0 (7.2)

17.0 (12.1)

19.6 (16.9)

22.4 (12.6)

Not seizure-free

22.4 (11.6)

24.2 (12.3)

23.6 (13.4)

21.3 (12.0)

Baseline seizure frequency per week: mean (±SD)

Seizure-free

1.4 (0.7)

2.1 (1.3)

2.6 (2.3)

1.4 (1.1)

Not seizure-free

5.3 (13.9)

6.6 (12.5)

5.7 (10.2)

5.2 (15.1)

aMethod 1: Seizure freedom defined by absence of seizures during the stable dose period. Patients seizure-free at time of withdrawal are counted as seizure-free.

Assessing seizure-free rates during long-term follow-up

Long-term follow-up after the double-blind add-on phase can only be performed in an open-label setting for obvious ethical reasons. Some methodological pitfalls in assessing seizure freedom during follow-up are similar to those described for short-term trials, such as the handling of withdrawals and, more importantly, the selective drop-out of non-responders. Other methodological issues are specific to the design of the long-term follow-up. Patients enter these studies at different time points, and usually the studies contain a clearly defined cut-off point, often coinciding with the marketing approval of the drug. Patients will thus receive treatment for variable durations, making the analysis of seizure freedom more complicated and comparisons between different AEDs particularly difficult.

Two general methodologies can be used in calculating seizure freedom in long-term follow-up studies. Seizure freedom can be measured over a specified time period, such as 6 or 12 months, either by determining the patients who were seizure-free “during the last 6 (or 12) months” of treatment or by considering the patients who were seizure-free “for at least 6 (or 12) months” (at any period in the study). By definition, seizure freedom rates will be smaller or at best equal with the “last x month” approach as compared with the “at least x month” strategy. ( figure 3 ) shows data for patients exposed to add-on LEV during long-term pre-marketing follow-up (Radtke and Bennett, 2000). Various populations were evaluated, including the ITT population of 1422 patients, the populations who were exposed to LEV for at least 6 and 12 months, and the population in whom treatment was ongoing at the trial cut-off date. Six-month seizure-free rates (patients seizure-free for at least six months at any time during follow-up) ranged from 12.9% (ITT population) to 26.9% (ongoing population), and similarly, 12-month seizure-free rates ranged from 7.7% to 17.6%.

The ITT population also included patients who were exposed for less than six months or one year, respectively. Patients may have a shorter follow-up because they discontinue treatment, or because the study cut-off date has been reached after less than six months/one year. In this case, the denominator was reduced from 1422 (ITT) to 1053 (exposed for at least six months) or 825 (exposed for at least one year). Thus, the ITT population included patients who by definition could not be counted as seizure-free, because they did not reach the minimum duration of exposure required to be considered as seizure-free. The seizure freedom rate will thus be profoundly affected by whether the study population (denominator) is defined on an ITT basis or as the subgroup exposed to treatment for a specific number of months, making ITT a particularly conservative method for this kind of analysis. However, the ITT method is the most conservative, compared with an analysis of subjects exposed for at least a certain period of time, particularly if subjects were unable to reach the minimum exposure period due to low drug efficacy or intolerable adverse effects.

The method by which withdrawals are handled in long-term studies can also have a considerable impact on seizure freedom rates. The most conservative approach considers all seizure-free patients who discontinue as failures, similar to Method 3 in short-term trials. However, this approach may underestimate the real efficacy of the drug, because patients discontinue for several different reasons, and some would likely have remained seizure-free for a longer period had they remained in the study. In a prolonged follow-up, some patients may remain seizure-free over long time periods, prompting physicians to withdraw one drug or stop all drug therapy. An alternative approach is to consider all withdrawals while seizure-free as successes and include such patients in the calculation of seizure-free rates, similar to Method 1 in the short-term trials. This approach, however, gives an advantage to an effective but poorly tolerated drug with significant numbers of discontinuations due to adverse effects. Such patients would not necessarily have remained seizure-free had they not withdrawn from the study.

As described above, although the ITT approach is associated with the least bias, it is necessary to indicate how the early discontinuations are handled in an ITT population. The following theoretical examples illustrate how these factors influence the seizure freedom results for three different AEDs. Each drug is administered to a theoretical cohort of 100 patients.

Drug A (efficacious and well tolerated AED)

For such a drug, a realistic assumption could be that, in an ITT population of 100 patients, 70 continue treatment with drug one for at least six months; two are seizure-free and discontinue, and 15 patients are seizure-free and still on drug at the end of the follow-up. Using the ITT population, the seizure freedom rate is 17% (17/100) if withdrawals are considered as successes and 15% (15/100) if withdrawals are considered as failures. If the population is defined as the patients exposed for at least 6 months, then the seizure freedom rate is 21% (15/70).

Drug B (efficacious but poorly tolerated AED)

For such a drug, a realistic scenario can be assumed whereby, in an ITT population of 100 patients, 50 continue treatment with the drug for at least six months; 10 are seizure-free but withdraw from treatment, and 15 are seizure-free and still on the drug at the end of the trial. Using the ITT population, the seizure freedom rate is 25% (25/100) if withdrawals are considered as successes and 15% (15/100) if withdrawals are considered as failures. If the population is defined as patients exposed for at least six months, then the seizure freedom rate is 30% (15/50).

Drug C (well tolerated but less efficacious AED)

Here, a reasonable scenario could be that, in an ITT population of 100 patients, 70 continue treatment with drug three for at least six months, one patient is seizure-free and discontinues, and five patients are seizure-free and still on the drug at the end of the trial. Using the ITT population, the seizure freedom rate is 6% (6/100) if withdrawals are considered as successes and 5% (5/100) if withdrawals are considered as failures. If the population is defined as patients exposed for at least 6 months, then the seizure freedom rate is 7% (5/70).

Based on the above examples, it is clear that the seizure freedom rate for a poorly tolerated drug (drug B) is grossly overestimated when withdrawals are considered as successes in the ITT population or when the denominator is defined as the population exposed to drug for at least six months.

Discussion

This review has shown that the literature contains little information on seizure freedom rates with AEDs used as add-on therapy, and even less information about how these rates were determined. Moreover, as reported in this article, lack of uniformity in defining seizure freedom and, more importantly, failure to define seizure freedom in many publications, do not allow any meaningful comparison to be made. Some methods for calculating seizure freedom give highly inflated estimates, particularly with poorly tolerated drugs when last observation carried forward analysis is used, i.e., when seizure-free patients withdrawing from treatment are counted as successes.

Physicians can choose among many different AEDs when selecting treatment for their patients. A variety of criteria enter into the decision-making process, including the efficacy and tolerability of the drug as well as various individual patient characteristics, such as seizure history, pre-existing drug therapy, and comorbidities. At least for adjunctive therapy trials, no direct comparisons between AEDs have been conducted that allow efficacy and tolerability of the drugs to be compared. Moreover, methodological differences among studies make it problematic to differentiate between AEDs.

If we believe that no seizures and no adverse side effects is the ultimate goal of epilepsy treatment (Langfitt and Meador, 2004), then seizure freedom is a very important parameter, and how it is calculated is a vital piece of information in order to evaluate, compare, and apply the data to a particular patient.

Newly diagnosed epilepsy is treated with old or new AEDs, and up to two thirds of these patients may achieve remission with monotherapy (Kwan and Brodie, 2000). Patients who are refractory to mono- or polytherapy may be included in clinical trials of new AEDs, with the goal of achieving regulatory approval for use in add-on therapy. Because this patient population is particularly refractory to AED treatment, high seizure-freedom rates cannot be expected, and seizure freedom often is not reported.

No one can deny, however, that seizure freedom is the single most important determinant of quality of life, and even a modest probability of achieving seizure freedom may well justify trying a new AED in a patient with severely refractory epilepsy. Because of this, information on comparative seizure freedom rates in adjunctive therapy and follow-up studies of new AEDs is clinically meaningful, and it should be reported at all times. For some of the recently introduced AEDs, seizure freedom rates during adjunctive therapy use in severely refractory epilepsy have been modest, but not always negligible. In particular, 6-month seizure freedom rates in long-term follow-up studies of LEV were as high as 26%, depending on the calculation method used. If the ITT calculation method is applied (e.g., ITT analysis in the entire population of 1422 patients included in follow-up studies), the seizure freedom rate was still close to 13%, which is clinically important in a population of patients who had already failed a wide range of drugs.

To be able to comparatively assess the value of the many available AEDs, it would be important to perform a meta-analysis of seizure freedom rates in all studies performed so far with each of the AEDs. Regrettably, a lack of published seizure freedom information makes such a meta-analysis impossible at the present time.

Conclusion

Little emphasis has been placed on seizure freedom in adjunctive therapy studies. Inasmuch as seizure freedom is the ultimate goal of AED therapy, a standardized methodology may need to be developed to ensure seizure freedom rates are reported consistently for all clinical studies, even when such rates are very low or even zero. Direct comparisons between new AEDs remain to be done. In the meantime, indirect comparisons would be a useful, albeit limited alternative; however, they are difficult to conduct because seizure freedom rates are only reported for some AEDs or some trials, leading to potential reporting bias, and because the methods used to calculate them are not always reported and, when they are, differ across studies. Therefore, such comparisons need to be made cautiously. Transparency in the methodology used to assess seizure freedom is mandatory in order to make such indirect comparisons valid. A whole range of different methodologies is available. It is our recommendation that the minimum duration for assessing seizure freedom should be the entire stable dose period in short-term double-blind clinical trials, and the last six months for long-term follow-up studies regardless of the status of the patient’s treatment. At a minimum, the methodology used should be described and at least two different methodologies should be presented, one that considers seizure-free withdrawals as successes, and one that considers them as failures. In the end, it is the number of patients that remain seizure-free throughout the study that may be the most valid measurement. The ultimate goal of antiepileptic treatment is complete seizure control without unacceptable side effects, so reporting seizure freedom is very important, and as such, should not be misrepresented.