Lacosamide adjunctive therapy for partial-onset seizures: a meta-analysis

Protocol and registration

The search strategy, methods of analysis and inclusion criteria were specified in advance and documented in a protocol. The protocol for this systematic review was registered with the Prospective International Register of Systematic Reviews (PROSPERO) and can be found online (Sawh & Newman, 2012).

Information sources/search strategy

Studies were identified by searching the following electronic databases: Medline (OVID 1946 to Week 4, 2012), EMBASE (OVID, 1980 to Week 3 2012), Cochrane Central Register of Controlled Trials (CENTRAL) (Wiley Issue 1 of 12, January 2012).

We contacted the manufacturer of lacosamide and experts in the field for information about unpublished or ongoing studies. The Food and Drug Administration’s (FDA) Approved Drug Products database was searched for clinical trials used to support marketing approval and/or labelling changes in the United States. Conference abstracts and posters were searched from selected meetings of the American Epilepsy Society, World Congress of Neurology, International Epilepsy Congress, and the European Congress on Epileptology. We also searched the metaRegister of Controlled Trials (mRCT) to identify ongoing trials.

Reference lists of all retrieved studies were reviewed for additional relevant studies.

The search was developed and conducted by one of the authors (SS) and reviewed by a Research Librarian (KC). The last search was run February 7, 2012. We used the following search terms to search all trial registers and databases (modified to suit each specific database): randomized controlled trials, epilepsy, seizures, partial epilepsy, lacosamide, and Vimpat. No language restrictions were imposed on the electronic database searches. The online protocol provides the detailed search strategy used in this review.

Study selection

Title and abstract screening was conducted in duplicate to identify potentially eligible papers using a standardized guide for trial inclusion based on title and abstract screening. Two reviewers (JN and SS) underwent a calibration process to identify potential discrepancies in interpretation of the form (with the first 100 citations as a sample). Publications that could possibly have met the selection criteria were retrieved as full-text articles and examined.

Full-text screening was conducted, independently by two reviewers, to confirm eligibility using a standardized screening form (Table S1). We used Fleiss and Cohen’s weighted Kappa (using the program Kappa.exe (Cyr & Francis, 1992)) to assess agreement between the two reviewers on the selection of full-text articles for inclusion (Fleiss & Cohen, 1973). All disagreements were resolved by discussion.

We documented the study selection process in a flow chart as recommended in the PRISMA statement (Liberati et al., 2009) showing the total numbers of retrieved references and the numbers of included and excluded studies, and the reasons for exclusion (Fig. 1).

Data collection process & data items

Data were extracted independently by two reviewers (JN and SS) using an a priori standardized data extraction form with the aid of a data and validity extraction manual. The two sets of extracted data were compared and all discrepancies were resolved by discussion. Data was extracted from each included trial on the following general areas of information:

Risk of bias in individual studies

Two reviewers (JN and SS) independently assessed the risk of bias for each included study using the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins, Altman & Sterne, 2011). Reviewers were not blinded to the study authors, journal or outcome data. We specifically assessed the trial characteristics as specified in the protocol.

• sequence generation;

• allocation concealment;

• blinding of the study (participants, personnel, outcome assessors, data collectors, data analysts) as defined by Akl et al. (2012);

• incomplete outcome data;

• selective outcome reporting;

• other sources of bias.

A summary table and a graph for risk of bias were created using Review Manager software 5.1 (The Nordic Cochrane Centre, 2011).

Synthesis of results

We calculated the pooled relative risks (RRs) and 95% confidence intervals (CI) for dichotomous variables using the Mantel-Haenszel method (Deeks, Higgins & Altman, 2011). For continuous variables measured using the same scales, the mean differences (MD) and its 95% CI were calculated using the inverse variance method. If a continuous outcome variable was measured using different scales across studies, we calculated the standardized mean difference (SMD).

All of our analyses included the total numbers of participants in the treatment groups to which they had been allocated (intention to treat analysis). Participants not completing follow up or with inadequate seizure data were assumed to be non-responders.

We contacted study authors for clarification if more information was needed, and to request missing data.

Randomized trials included multiple dosages of lacosamide in separate randomized arms. For the purpose of the meta-analysis, all lacosamide dosages were combined into one “lacosamide” arm (Cochrane Handbook for Systematic Reviews of Interventions, 2011).

We tested statistically for heterogeneity with a chi-square test and used I² to measure inconsistency (the percentage of total variation across studies due to heterogeneity). We used “small,” ( ≤ 25%), “moderate” (between 25% and 50%) and “large” ( ≥ 50%) to describe the statistical heterogeneity as measured by I² (Higgins et al., 2003). Forest plots were visually inspected for possible sources of heterogeneity.

A summary of findings table was created using GRADEpro software for the three primary outcomes of this review (Brozek, Oxman & Schünemann, 2008). We planned to assess the possibility of publication bias by using funnel plots (Egger & Davey Smith, 1995).

Additional analyses

The following subgroup analyses were pre-specified for primary outcomes: patients younger than 18 years old (if the pediatric outcome data was reported as a discrete subgroup), placebo vs. active comparators, intravenous vs. oral lacosamide, and comparing studies with high vs. low risk of bias. Post-hoc, the potential of a dose-response relationship of lacosamide was explored using subgroup analysis to look at the various dosage levels studied for all three primary outcomes.

The a priori sensitivity analyses for the primary outcomes were: (1) Best case- Participants not completing follow-up or with inadequate seizure data were assumed to be responders in the lacosamide group and non-responders in the control group. For the primary safety outcome, participants not completing follow-up or with inadequate data were assumed to have continued in the trial in the lacosamide arms and discontinued if in the control arm. (2) Worst case- Participants not completing follow-up or with inadequate seizure data were assumed to be non-responders in the lacosamide group and responders in the control group. For the primary safety outcome, participants not completing follow-up or with inadequate data were assumed to have discontinued due to adverse events in their respective lacosamide groups and to have stayed in if in the control group.

Study selection

A total of 11 reports involving 3 studies were identified for inclusion in the review. The search of Medline, EMBASE, and CENTRAL provided a total of 200 citations. The search for unpublished literature (expert survey, manufacturer request, clinical trial registries, and conference abstract proceedings) provided a total of 207 citations. After removing duplicates, 357 citations independently underwent abstract review and 77 citations were considered potentially relevant studies. Of the 77 full-text articles screened, 66 citations were excluded. Three randomized controlled trials (Ben-Menachem et al., 2007; Chung et al., 2010b; Halász et al., 2009) (located as 11 publications (Massie, 2007; Kalviainen et al., 2007; Halász et al., 2006; Chung et al., 2007a; Chung et al., 2009b; Jatuzis et al., 2005; Ben-Menachem et al., 2005; Chung et al., 2007b), that studied 1311 participants, met the inclusion criteria for this review. The weighted kappa statistic for inter-rater agreement on including or excluding potential trials was “excellent” [k = 0.90, 95% CI (0.83, 0.97)] (Fleiss & Cohen, 1973). See flow diagram Fig. 1.

Study characteristics

See Table 1 for the characteristics of the included studies and Table S2 for the table of excluded studies.

Risk of bias within studies

See Fig. 2.

All three studies were randomized-controlled trials, and all studies except Ben-Menachem presented the method of random sequence generation. Allocation concealment and blinding of participants, personnel, and outcome assessors were adequately reported for all trials. Incomplete outcome data reporting was present for all three trials. Selective outcome reporting was noted for all three included trials, as assessed by comparison of the published trials to the studies submitted for FDA approval. None of the three trials did a formal ITT analysis, but used all patients who received at least one dose of study medication as their definition of the study population.

Results of individual studies

Primary outcomes

The mean change in seizure frequency from maintenance phase to baseline was not provided in any of the three included studies. The authors of each study were contacted in an attempt to procure the seizure frequency change data, but no information was provided.

The primary outcome data available from the three trials (“percent reduction in seizure frequency”) is presented in Table 2. There was a larger median percent change (as noted by the trial authors) with the higher dosage arms of lacosamide compared to placebo.

Table 2:

Median percentage reduction in seizure frequency.^*

Trial	Placebo		Lacosamide 200 mg/day		Lacosamide 400 mg/day		Lacosamide 600 mg/day
	N	%	N	%	N	%	N	%
Ben-Menachem et al. (2007)	96	10%	107	26%	107	39%	105	40%
Chung et al. (2010b)	104	20.8%	–	–	201	37.3%	97	37.8%
Halász et al. (2009)	159	20.5%	160	35.3%	158	36.4%	–	–

DOI: 10.7717/peerj.114/table-2

Notes:

* Compares maintenance phase to baseline period.

The 50% responder rate was reported for all three included trials and the results are presented in Fig. 3. In the meta-analysis of this primary outcome (ITT), lacosamide (all dosage arms pooled) was associated with a significantly higher 50% response rate compared to placebo (RR 1.68, 95% CI 1.36, 2.08). There was no evidence of statistical heterogeneity (I² = 0%). The analyses of worst-case scenarios and best –case scenarios both produced similar results to the base analysis (RR = 1.62, 95% CI 1.24, 2.11; I² = 37%) and (RR = 1.73, 95% CI 1.40, 2.13; I² = 0), respectively.

Discontinuation of study drug due to adverse events was reported in all three trials. In the meta-analysis of this outcome (ITT), lacosamide (all dosage arms pooled) was associated with a significantly higher rate of discontinuation with an RR 3.13 (95% CI 1.94, 5.06). There was no evidence of heterogeneity (I² = 0%), see Fig. 4. Best-case and worst-case scenarios were not calculated for this outcome as no patient data was missing.

Secondary outcomes

Adverse effects outcomes

Statistically significant changes (higher rates in the lacosamide pooled dosage arm) were seen in the following adverse event outcomes: ataxia (RR 5.03, 95% CI 2.23, 11.37, see Fig. 5), dizziness (RR 3.49, 95% CI 2.43, 5.01, see Fig. 6), fatigue (RR 2.04, 95% CI 1.08, 3.85, see Fig. 7) and nausea (RR 2.36, 95% CI 1.22, 4.58, see Fig. 8). No heterogeneity was found in any of the adverse events (I² = 0), except for nausea, which showed moderate inconsistency, with an I² = 34%. For the outcome of ataxia, data were included if outcomes were reported as ataxia or “coordination abnormal”.

All other meta-analyses and forest plots including: other adverse event outcomes (headache, somnolence, serious adverse events), seizure-free during the treatment period, the sensitivity analyses for best and worst case scenarios and the dose-response analyses can be found in Figs. S1–S10.

Quality of life outcomes

The quality of life outcomes were incompletely reported across all included studies. Mean change in QOLIE-31 was reported in the Ben-Menachem trial (Ben-Menachem et al., 2007) but no measure of variance (SD) was provided. The measurement of QOLIE-31 was limited by language availability. Since the measurement scale was only available in English, only participants from the United Kingdom or the United States of America were able to contribute to this outcome. The Clinical Global Impression of Change (CGIC) score was not reported as a continuous outcome (mean change), but as a dichotomous outcome from “Very much improved” or “much improved” from baseline to maintenance. No denominators for the groups were provided. There was a numerically larger change in seizure frequency in the lacosamide arms compared to the placebo arms. Although Halász et al. (2009) reported that they would measure quality of life outcomes; these were not reported in the final publication. The quality of outcome scales as reported by Ben-Menachem et al. (2007) are provided in Table 3.

Economic outcomes

No trials reported on hospital admission, length of stay, length of stay in a specialized epilepsy unit or economic outcomes.

Additional analysis

The planned sensitivity analyses for the primary efficacy outcome and 50% responder rate were ultimately not undertaken due to the small number of studies and lack of information relating to the pre-specified subgroups.

Post-hoc exploratory analyses were undertaken to explore a dose-effect relationship with efficacy and safety for two of the primary outcomes. The different dosage arms of lacosamide vs. placebo were placed into separate subgroups to explore the dose response of the outcomes for 50% responder-rate and discontinuation of study drug. Both analyses showed that, as the lacosamide dose increased, so too did the trend in the 50% response rate (P for interaction = 0.26) (Fig. S7) and the discontinuation of study drug due to adverse events (P for interaction = 0.03) (Fig. S8).

As a post-hoc analysis, we also pooled the lacosamide 100 mg BID and 200 mg BID arms (200 mg or less) and compared them to the 300 mg BID arm (greater than 200 mg). These analyses (Figs. S9 and S10) support the previous finding that higher doses of lacosamide are associated with greater risk of drug discontinuation due to adverse events (RR 2.28, 95% CI 1.46, 3.58; I² = 0%).

Risk of bias across studies

No statistical heterogeneity (I² = 0%) was found in the analyses of the primary outcomes and in the majority of the outcomes assessed in the meta-analysis. Funnel plot asymmetry was not tested because only three studies were included in this meta-analysis, rendering this test unreliable (Sterne, Egger & Moher, 2011).

Where heterogeneity did exist in the pre-specified analyses (nausea, 50% responder rate worst-case scenario analysis), it was small to moderate with non-significant p-values. Potential sources of heterogeneity could be: (1) the differing dosage arms of lacosamide used in the trials and, (2) the differing lengths of treatment (dose-titration was 4 weeks in one trial and 6 in the other two trials).

Selective reporting was evident in all three studies included in this review. Consulting the FDA approval documents provided a more complete list of outcomes to be measured in the trials but no details could be obtained on many outcomes, including one of the pre-defined primary outcomes - change in seizure frequency from baseline. This was the stated primary outcome in all three trials, but was not reported in any of the publications.

The quality of evidence was downgraded for risk of bias and imprecision where appropriate (Guyatt et al., 2008). See the Summary of Findings for the statistically significant outcomes from this review (Table 4).

Summary of evidence

Overall, the evidence from three included trials supports that lacosamide improves the 50% responder rate compared to placebo in adults with partial-onset seizures. The reduction in seizures demonstrated by this efficacy outcome must be weighed against the increased rate of discontinuation due to adverse events and the risk of increased ataxia, dizziness, fatigue, and nausea.

Post-hoc analysis demonstrated a possible dose-response relationship with 50% responder rate. While post-hoc analyses should always be interpreted cautiously, the test for interaction for discontinuation of study drug due to adverse events was statistically significant (p = 0.03), indicating that this is an important area for future research.

Incomplete outcome reporting impacted the final results of this systematic review. As study authors did not report the mean seizure frequency in each group, no quantitative analyses could be conducted on the change in seizure frequency, one of the a priori primary outcomes of this review. Quality of life outcomes were only selectively reported, and the data available for this review were not amenable to meta-analysis. All of the authors mentioned that lacosamide had a favourable effect on QOL measures.

None of the included trials reported on economic outcomes. From a strict drug cost perspective, lacosamide is far more expensive compared to other available AEDs (Table S3). Comparative trials with other AEDs would be helpful in determining the most cost effective role in the treatment of seizures.

The findings from this review can be directly applied to ambulatory patients with partial epilepsy, who are refractory to their current AED therapy. Hospitalized patients were not included in any of the trials in this review.

Limitations

This systematic review used a robust search strategy to consider all of the best available published and unpublished evidence of lacosamide in partial-onset seizures in adults; however there were a few limitations to this review. The post-hoc analyses indicate the possible presence of a dose-response for efficacy and adverse effects for lacosamide. The pooling of all lacosamide doses as the comparator arm compared to placebo could underestimate the benefit of the higher doses and/or overestimate the effect of the lower doses with respect to efficacy. This pooling of all dose arms of lacosamide would be expected to similarly affect the adverse events outcomes if a dose-response relationship exists.

All of the included studies (which formed the basis of regulatory approvals) were of a very short duration (three months of maintenance therapy), which may have exaggerated the efficacy of lacosamide as adjunctive AED therapy, given that these therapies are usually administered for many years. These regulatory approval trials tend to have limited external generalizability due to exclusion of patients with co-morbidities which are common in a large percentage of epileptic patients (Chung et al., 2009b; Jatuzis et al., 2005).

The search strategy was not designed to retrieve economic analyses, so relevant economic studies on lacosamide could have been missed.

Incomplete reporting of outcomes precluded meta-analysis of the mean change in seizure frequency, one of the primary outcomes of this review. Overall, included studies were of moderate quality, as selective outcome reporting, using non-intention-to-treat analyses, and imprecision resulted in a downgrade in the quality of evidence of these randomized controlled trials. In this review, the risk for reporting bias, as evidenced by selective outcome reporting for all included trials, is judged to be the biggest threat to validity. Since only three studies were included, publication bias could not be explored.

This review does not apply to the pediatric population, as the two trials including patients less than 18 years old (Chung et al., 2010b; Halász et al., 2009) did not provide separate data on the pediatric participants.