Types of studies

We include randomised controlled trials (RCTs) in which the authors compare rTMS therapy with sham therapy or no therapy. We excluded trials in which the authors report only laboratory parameters.

Types of participants

We include studies with participants of any age or sex after stroke, regardless of the duration of illness or severity of the initial impairment. The clinical definition of stroke was that of the World Health Organization criteria (Stroke 1989), excluding stroke mimics by computerised tomography (CT) or magnetic resonance imaging (MRI) scan.

Types of interventions

We include all trials that evaluated rTMS therapy in patients with stroke, regardless of ipsilateral or bilateral stimulation, frequency, age or duration of illness. The control interventions were sham treatment or other conventional treatment.

We investigated the following comparisons:

1. rTMS only compared with sham treatment;

2. rTMS add‐on baseline treatment compared with sham treatment add‐on baseline treatment;

3. rTMS add‐on baseline treatment compared with baseline treatment alone.

Types of outcome measures

We assessed outcomes at the end of treatment period and scheduled follow‐up.

Electronic searches

We searched the Cochrane Stroke Group Trials Register, which was last searched by the Managing Editor in April 2012 . In addition, and in collaboration with the Cochrane Stroke Group Trials Search Co‐ordinator, we searched the following bibliographic databases:

• the Chinese Stroke Trials Register (April 2012);

• the Cochrane Central Register of ControlledTrials (CENTRAL) (The Cochrane Library 2012, Issue 4);

• MEDLINE (1950 to May 2012) (Appendix 1);

• EMBASE (1980 to May 2012) (Appendix 2);

• ISI Science Citation Index (1981 to April 2012);

• CINAHL (1982 to May 2012) (Appendix 3);

• AMED (the Allied and Complementary Medicine Database (1985 to May 2012) (Appendix 4);

• PEDro (Physiotherapy Evidence Database) (www.pedro.org.au/) (April 2012);

• REHABDATA (www.naric.com/research/rehab/default.cfm) (April 2012);

• CIRRIE Database of International Rehabilitation Research (http://cirrie.buffalo.edu/index.html) (April 2012);

• The China Biological Medicine Database (CBM) (1978 to April 2012);

• The Chinese National Knowledge Infrastructure (CNKI) (1979 to April 2012);

• Chinese Science and Technique Journals Database (VIP) (1989 to April 2012);

• Wanfang Data (www.wanfangdata.com/) (1984 to April 2012).

We also searched the following international trials registers in April 2012:

• ClinicalTrials.gov (www.clinicaltrials.gov/);

• Current Controlled Trials (www.controlled‐trials.com);

• Stroke Trials Registry (www.strokecenter.org/trials/).

Searching other resources

In an effort to identify further published, unpublished and ongoing studies, we searched databases of conference abstracts: Conference Proceedings Citation Index Science (CPCI‐S) and China Medical Academic Conferences (CMAC 1995 to present) in CMCC (Chinese Medical Current Contents) (April 2012), and all reference lists of retrieved articles.

Selection of studies

Two review authors (ZH and DW) independently scanned the titles, abstracts and keywords of records identified from the electronic searches and excluded obviously irrelevant citations. We obtained the full text of the remaining studies and the same two authors selected studies for inclusion based on the criteria outlined previously. We resolved any disagreements through discussion with a third author (ML).

Data extraction and management

Two review authors (ZH and DW) independently extracted details of patient characteristics, methods, interventions and outcomes by using a data extraction form. We resolved disagreements through discussion with a third author (YZ). For dichotomous outcomes we extracted the number of participants experiencing the event and the total number of participants in each arm of the trial. For continuous outcomes we extracted the mean value and standard deviation for the changes in each arm of the trial along with the total number in each group.

Assessment of risk of bias in included studies

We assessed the methodological quality of selected studies as described in the Cochrane Handbook for Systematic Reviews of Interventions (Cochrane Handbook). We created a 'Risk of bias' table and included a description and a judgement (low risk of bias, high risk of bias, or unclear risk of bias) for the following domains for each of the included studies:

1. random sequence generation;

2. allocation concealment;

3. blinding of participants and personnel;

4. blinding of outcome assessment;

5. incomplete outcome data;

6. selective reporting;

7. other sources of bias.

Two review authors independently performed quality assessment; we resolved any disagreements between authors arising at any stage through discussion or with a third author.

Measures of treatment effect

We expressed results for dichotomous outcomes as risk ratios (RR) with 95% confidence intervals (CI), and expressed results for continuous outcomes as mean difference (MD) if the same scale for each trial was available, or standardised mean difference (SMD) if different scales were used. For continuous outcomes, we intended to compare the change scores between groups after treatment and at the end of the follow‐up period.

Unit of analysis issues

For studies with non‐standard designs (e.g. cross‐over trials, cluster‐randomised trials), we planned to manage the data according to the Cochrane Handbook. For example, if we had found any cross‐over trials, we would only have analysed the data from the first period.

Dealing with missing data

If data were missing, we contacted the investigators for additional information. If some data remained unavailable, we considered both best‐case and worst‐case scenarios.

Assessment of heterogeneity

We determined heterogeneity by using the I² statistic. We considered I² greater than 50% to be substantial heterogeneity (Higgins 2003).

Assessment of reporting biases

We used the funnel plot method (Egger 1997).

Data synthesis

We performed statistical analysis using RevMan 5.1 (RevMan 2013) and performed all analyses in accordance with the intention‐to‐treat method. We reported the results as RRs with 95% CIs for dichotomous data and as MDs or SMDs with 95% CIs for continuous data. We used a random‐effects model to combine individual results. If there were no suitable studies, we planned to provide a narrative summary of the study results.

Subgroup analysis and investigation of heterogeneity

We planned a priori subgroup analyses based on:

1. stroke type: ischaemic stroke versus intracranial haemorrhage;

2. ipsilateral or bilateral stimulation;

3. different frequency (low frequency or high frequency);

4. duration of illness;

5. severity of initial impairment;

6. stimulus parameters.

Sensitivity analysis

We planned to perform sensitivity analyses by:

1. excluding studies with inadequate concealment of allocation;

2. excluding studies in which outcome evaluation was not blinded;

3. excluding studies in which loss to follow‐up was not reported or was greater than 10%;

4. re‐analysing the data by removing studies with nonstandard designs if we included these studies;

5. re‐analysing the data by removing studies with assumed values to replace missing data.