Results of the search

The update of searches for this review yielded 22 results; 18 from the Cochrane Register of Studies (CRS Web; 13 September 2010), and four from MEDLINE (1946 to September 13, 2019). After removing five duplicates and one obviously irrelevant item, we identified 16 articles for possible inclusion. On further evaluation of titles and abstracts, none of these articles met the inclusion criteria (Figure 1). Hence, review authors found no additional studies for inclusion in this updated version of this review.

Figure 1

---

---

This diagram refers only to results of the updated searches for the current version of the review

Searches from previous versions of this review identified eight studies that met the inclusion criteria (Brodie 2005; Faught 2001; Guerrini 2013; Lu 2011; Sackellares 2004; Schmidt 1993; Wu 2010; Zhang 2011). A full report for one study was unobtainable, so it was identified as awaiting classification (Anderson 1988). We also found two unpublished studies in the last update of this review (Brigo 2018); however, a full report of them was not available, so both were identified as awaiting classification (NCT00327717; NCT01546688).

Included studies

The eight studies (1636 participants) already included in the previous versions of this review were parallel‐group trials with a stable‐dose phase of 12 or 18 weeks (although Faught 2001 added a cross‐over phase for all participants in the final five weeks, only the parallel group data were included in analysis (Brodie 2005; Faught 2001; Guerrini 2013; Lu 2011; Sackellares 2004; Schmidt 1993; Wu 2010; Zhang 2011). Two studies allowed some dose titration according to seizure response and tolerability (Sackellares 2004; Schmidt 1993). In both studies, the median daily dose for participants completing the study was 400 mg. A third study titrated to 400 mg/day, but participants randomised to zonisamide followed different rates of titration during the first five weeks of the study, which allowed some comparison to placebo, 100 mg/day, and 200 mg/day during this period of the study (Faught 2001). Brodie 2005 randomised participants to placebo, 100 mg, 300 mg, or 500 mg of zonisamide in a 2:1:1:2 ratio, allowing further investigation of dose‐response relationships.  Lu 2011 used two different titration strategies; in the first, a 100 mg/day zonisamide dose was given for the first two weeks after baseline, 200 mg/day for the third week, and 300 mg/day from week four onwards. The initial dose in the second strategy was 100 mg/day of zonisamide, increasing weekly by increments of 100 mg/day until a target of 400 mg/day was reached in week four, and continued for the duration of the trial. Guerrini 2013 titrated zonisamide in weekly increments of 1 mg/kg/day over eight weeks, to a target dose of 8 mg/kg/day (max 500 mg/day), and continued unchanged over the 12‐week maintenance period. Wu 2010 titrated zonisamide as follows: first two weeks 100 mg/day, third week 200 mg/day, and fourth week 300 mg/day, which was maintained to the end, although, according to the participants's situation, doses could be increased to 400mg/day. In the last study, Zhang 2011 started zonisamide at 100 mg/day, and then increased to 300 mg/day (100 mg, three times a day) within three weeks; the dose during maintenance was 300 mg/day.

See Characteristics of included studies for full details of the included studies.

Excluded studies

None of the 16 articles obtained by the updated search strategy appeared to meet the eligibility criteria, and therefore, we considered them not relevant (see Results of the search).

In previous versions of this review, the review authors excluded Shimizu 1988, as no control group was used (see Characteristics of excluded studies; Brigo 2018; Carmichael 2013; Chadwick 2002; Chadwick 2005).

Allocation

In three studies, allocation was concealed by dispensing sequentially numbered packages to each participant (Faught 2001; Sackellares 2004; Schmidt 1993), whilst Brodie 2005 used a telephone randomisation service. In two studies, participants were assigned to groups in blocks of six (Brodie 2005), and four (Schmidt 1993). Randomisation codes were generated centrally (Faught 2001), or by the study sponsor (Sackellares 2004). Lu 2011 used a process of restricted randomisation as zonisamide and placebo were assigned in a ratio of 1:1.

Participant codes were hidden by use of numbered containers (Lu 2011; Sackellares 2004), or sealed envelopes containing an individual participant code (Faught 2001). However, the appearance of the items used to conceal codes was not explicitly described, so one could not be certain how effective these concealment methods were. One study performed randomisation using a stratified random, segmented random, and random distribution list (Zhang 2011). Allocation concealment and random sequence generation were not specified in one study (Wu 2010).

Blinding

We deemed six of the included studies to be at low risk of performance bias (participants), as placebo and zonisamide tablets were identical in appearance (Brodie 2005; Faught 2001; Lu 2011; Sackellares 2004; Schmidt 1993; Guerrini 2013). Two studies did not specify whether placebo and zonisamide tablets were identical in appearance (Wu 2010; Zhang 2011).

Blinding of outcome assessors was not detailed in any of the included studies, and therefore, we classified this as being at an unclear risk of bias. However, participants self‐reported seizure frequency and duration, and therefore, we thought that a lack of detail on the outcome assessors would have a minimal impact overall. One trial report did not provide any detail with regard to blinding of study personnel or outcome assessors, and as such, we classified it as being at unclear risk of bias overall (Schmidt 1993).

Incomplete outcome data

We classified Brodie 2005 and Lu 2011 to be at a low risk of attrition bias overall, as few participants left the study before completion for any reason, and those who did were reasonably evenly spread across the different intervention groups. We assessed Faught 2001 to be at an unclear risk of attrition bias, due to unclear reporting of missing data and study attrition rates. Sackellares 2004 did not clearly report all data in relation to study attrition and missing data. The attrition rate was not clearly reported in Schmidt 1993, and as such, we classified it as being at an unclear risk of attrition bias. Two studies were at high risk of attrition bias(Wu 2010; Zhang 2011).

All of the trials conducted a modified version of an intention‐to‐treat (ITT) analysis, while one trial also conducted a second, unmodified ITT analysis (Faught 2001). Three trials failed to include all randomised participants in their ITT analysis, and instead omitted participants who had not completed the trial for any reason (Brodie 2005; Lu 2011; Sackellares 2004). In these three trials, participants were included in the ITT analysis if they had taken at least one dose of the intervention to which they had been allocated. A fourth trial also conducted a modified ITT analysis, but instead, included participants who had received at least seven days of treatment (Schmidt 1993).

Selective reporting

We contacted all study authors; some provided additional data, but nobody made their trial protocols available to us. All primary and secondary outcomes outlined in the methods section of each trial were analysed and reported in the results section of each study, so we classified them as being at a low risk of reporting bias.

Other potential sources of bias

One potential source of other bias would be the unequal duration of stable‐dose phase, depending on to which dosage group a participant had been allocated. Specifically, in one study, participants allocated to the 100 mg/day zonisamide dose were on a stable dose for 23 weeks in total, as their titration phase was relatively brief in comparison to those participants in the 300 mg/day and 500 mg/day groups who had a stable‐dose phase of 20 and 18 weeks respectively (Brodie 2005).

Five included trials were sponsored by industry (Dainippon or Elan Pharma; Brodie 2005; Faught 2001; Guerrini 2013; Sackellares 2004; Schmidt 1993). No sources of funding were used to assist in the conduct or preparation of Lu 2011, but two different manufacturers of zonisamide provided the drug for the trial (Eisai Co. Ltd and Shenzhen Zifu Co. Ltd). Two studies did not provide details on funding (Wu 2010; Zhang 2011). Only authors of one study explicitly specified their conflicts of interest (Guerrini 2013). Two studies extended the baseline period by four weeks if participants did not experience 15 or more seizures in the first four weeks of baseline, or 30 or more seizures in the first eight weeks of baseline (Sackellares 2004; Schmidt 1993). This manipulation may have artificially inflated the effect of zonisamide on seizure frequency, making any reduction in seizures attributed to zonisamide of a greater magnitude by comparison. Conversely, this extension of the baseline period may have increased the likelihood of regression to the mean. We did not find evidence of any other source of bias in the remaining included trials (low risk of bias; Brodie 2005; Faught 2001; Lu 2011; Wu 2010; Zhang 2011).

Zonisamide versus placebo

Zonisamide versus another antiepileptic drug

We did not find full reports of any studies that conducted studies on zonisamide versus another antiepileptic drug, although three potential studies are awaiting classification.