Retrieval of information

The detailed search strategy made on MEDLINE (Pub Med) was the following:

Advanced

1) Double blind study[MeSH Terms])

2) OR Double blind trial[MeSH Terms])

4) OR placebo[MeSH Terms]

5) AND epilepsy[MeSH Terms]

6) AND antiepileptic drugs[MeSH Terms]

Detailed search strategy made in EMBASE was the following:

1)Double blind study'/exp

2)OR 'double blind trial'/exp

3)OR 'placebo'/exp

4)AND 'epilepsy'/exp

5)AND 'antiepileptic drugs'

6)AND [embase]/lim

7)NOT [medline]/lim

We disregarded registry websites and abstracts of congresses because in general these abstract do not report in detail tolerability data.

We also excluded open-label trials, observational studies, conference proceedings, publications available only in abstract form and studies published in non-English language. We also excluded trials with a pre-randomization run-in response-conditional design or if randomization was preceded by an open-label period.

Data abstraction

Retrieved RCTs were divided in two groups and data were independently extracted by two couple of authors (GZ and AV, and FG and GL) for each group using predefined data fields.

Outcome measures extracted were retrieved from text, tables, and figures. In case of discrepant data, data from tables were chosen.

The following information was retrieved for each trial: (i) trial characteristics—number of treatment arms, description of masking and randomisation procedures, duration of baseline, titration and maintenance periods; design of study (parallel or crossover) (ii) patients’ characteristics— number of patients per arm, number of concomitant AEDs, median seizure frequency during the baseline period; (iii) intervention—type, dose; (iv) outcomes— for each arm, number of responders (all double blind duration), number of withdrawals due to adverse events, number of patients with adverse events, number of patients with each of the following 11 adverse events: dizziness, ataxia/coordination abnormal, diplopia, somnolence, fatigue, headache, memory impairment, tremor, abnormal thinking, anxiety and depression. For each of these AEs, only patients from those studies in which the AE had been reported, were included (we did not consider an AE as absent in those studies in which the AE had not been reported). We did this because, tables reporting adverse effects do not usually report all adverse effects but only those with a frequency higher than a certain threshold which may be more than 3%, more than 5% and, in some cases, even more than 10%. For this reason, we can include adverse effects only from studies reporting that adverse effect. This procedure may introduce a small error consisting in a small overestimation of the frequency of an adverse effect.

When published data reported in tables and illustrations differed, the tabulated data were used.

The Cochrane Collaboration’s tool for assessing risk of bias (Higgins and Altman, 2008) was used to ascertain the validity of eligible RCTs. Six domains were assessed for each study: “sequence generation”, “allocation concealment”, “blinding of personnel and outcome assessors”, “incomplete outcome data”, “selective outcome reporting”, and “other sources of bias”. The domains “incomplete outcome data” and “selective outcome reporting” were assessed for responder rates, withdrawal due to AEs rates, AEs rates, and the 11 above reported AEs.

Criteria for inclusion or exclusion of clinical trials in the systematic review

Inclusion criteria

  • All AEDs
  • Adults (16 years)
  • RCT
  • Double-blinding
  • Placebo-controlled
  • Duration greater or equal to 4 weeks
  • Partial (simple/complex) seizures with or without secondary generalised tonic-clonic seizures
  • Publication in English language
  • Spontaneous adverse events reported in a table or figure (adverse events obtained from a structured recording were not included)

Exclusion criteria

  • Systematic reviews
  • Non-randomized trials
  • Active-controlled trials
  • Double-blind trials performed in generalized epilepsies
  • Double-blind trials performed in children
  • Trials which were of less than 4 weeks duration
  • Trials which used a response conditional design where patients were allocated treatment only if they showed a predetermined response to treatment during a baseline period before randomization
  • Trials which recruited patients into the randomized, double-blind phase following an open-label period
  • Lack of detailed tolerability data or adverse events only reported trough structured recordings