Types of studies

We included randomised controlled trials (RCTs) with blinded assessment of participants, personnel, and outcomes.

Types of participants

Participants with a confirmed diagnosis of RRMS or PPMS, according to published criteria (McDonald 2001; Polman 2005; Polman 2011; Thompson 2018), regardless of age, sex, degree of disability, or duration of the disease. And we excluded participants with other clinically significant autoimmune disorder or previous immunosuppressive before.

Types of interventions

Experimental intervention: ocrelizumab alone or associated with other medications at the approved dose of 600 mg every 24 weeks for any course duration.

Comparator: placebo, any other active drug therapy (i.e. corticosteroids, plasmapheresis, beta interferons, glatiramer acetate, natalizumab, alemtuzumab, daclizumab, mitoxantrone, fingolimod, dimethyl fumarate, or teriflunomide).

Concomitant interventions were allowed only if used equally in all arms of the trial.

Types of outcome measures

We assessed the following outcomes at the end of the treatment period.

Electronic searches

We searched the following on 8 October 2021:

1. Cochrane Central Register of Controlled Trials (CENTRAL, the Cochrane Library) (2021 Issue 9);

2. MEDLINE (PubMed) (from 1966);

3. Embase (Embase.com) (from 1974);

4. ClinicalTrials.gov (www.clinicaltrials.gov) for all prospectively registered trials (from 2000);

5. World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP) (apps.who.int/trialsearch) (from 2005).

The full search strategies are listed in Appendix 1, Appendix 2, Appendix 3, Appendix 4, and Appendix 5.

Searching other resources

In addition, we used the following methods.

1. We screened reference lists of relevant review articles and primary studies found.

2. We contacted experts in the field to identify further published or unpublished trials.

3. We contacted the main authors of studies if data reported in the original articles were incomplete.

Selection of studies

Three review authors (ML, JZ, and JL) independently screened titles and abstracts of the citations retrieved by the literature search to obtain titles and abstracts of studies possibly relevant to the review. We obtained full copies of potentially relevant studies for further assessment. We also independently evaluated the eligibility of these studies on the basis of information available in the published data. We excluded irrelevant studies. We resolved disagreements through discussion.

Data extraction and management

Three review authors (ML, JZ and JL) independently extracted data from selected trials using standardised forms, and entered the data into Review Manager 5 (Review Manager 2020).
We extracted the following information from individual studies.

1. Publication details (i.e. year, data, country, journal, authors).

2. Study design and setting: inclusion criteria, exclusion criteria, number of randomised participants and characteristics of participants.

3. Details of intervention (i.e. doses, frequency, scheme, length).

4. Description of outcomes.

5. Risk of bias: sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective outcome reporting, and other bias.

6. Data analyses.

7. Declarations of interest and funding source.

We resolved disagreements by discussion among the review authors.

Assessment of risk of bias in included studies

Three review authors (ML, JZ, and YZ) independently assessed the risks of bias in included studies, using the Cochrane risk of bias criteria (Higgins 2021). We assessed the following domains.

1. Sequence generation.

2. Allocation concealment.

3. Blinding of participants and personnel.

4. Blinding of outcome assessment.

5. Incomplete outcome data.

6. Selective outcome reporting.

7. Other potential sources of bias.

We judged each domain as being at low, high, or unclear risk of bias. And we resolved any disagreements by discussion among all review authors. We judged the overall risk of bias of each included study according to the following criteria.

1. Low risk of bias (plausible bias unlikely to seriously alter the results) if all the above items were met.

2. Unclear risk of bias (plausible bias that raises some doubt about the results) if one or more items were assessed as unclear.

3. High risk of bias (plausible bias that seriously weakens confidence in the results) if one or more items were not met.

Measures of treatment effect

We analysed data using Review Manager 5 (Review Manager 2020). We expressed results for dichotomous outcomes as risk ratios (RRs) with 95% confidence intervals (CIs). We calculated mean differences (MD) with 95% CIs for continuous data. We used hazard ratio (HR) with 95% CIs if calculating time‐to‐event data.

Unit of analysis issues

We included studies with parallel‐group design: participants randomly assigned to intervention or control were analysed at the individual allocation level. We planned to include cross‐over studies by considering only data from the first half of the cross‐over trial, but the search found no cross‐over studies. We performed each separate analysis based on the preset outcomes and different periods of follow‐up (24 and 96 weeks).

Dealing with missing data

We contacted authors of identified studies to obtain additional information. If additional information was not obtained, we analysed the available data.

Assessment of heterogeneity

We evaluated clinical and methodological heterogeneity across included studies by comparing characteristics of participants, interventions, and study designs.

We evaluated statistical heterogeneity among included studies using a Chi² test with an alpha of 0.1, and with the I² test. A P value of less than 0.1 and an I² statistic more than 50% was an indication of substantial statistical heterogeneity (Higgins 2021); we examined potential sources of clinical and methodological heterogeneity.

Assessment of reporting biases

We did not use funnel plots to explore possible publication bias due to an insufficient number of included studies.

Data synthesis

We used Review Manager 5 to conduct formal meta‐analysis (Review Manager 2020). The selection of a fixed‐effect or random‐effects model was mainly based on the results of the Chi² test and I² statistic for heterogeneity (Higgins 2021). If the I² statistic indicated substantial statistical heterogeneity, we explored potential causes of heterogeneity first, to determine whether a subgroup analyses was needed. If the substantial heterogeneity still could not be explained, we adopted a random‐effects model. If the I² statistic indicated no significant statistical heterogeneity, we used a fixed‐effect model.

Subgroup analysis and investigation of heterogeneity

We planned to perform the following subgroups analyses.

1. Different dosages of ocrelizumab.

2. Different duration of treatment.

3. Different degrees of disability.

4. Different co‐interventions.

5. Different types of interferon beta‐1a.

However, we did not carry out subgroup analyses to consider dosages of ocrelizumab, baseline degree of disability, cointerventions, and types of interferon beta‐1a due to lack of available data.

Sensitivity analysis

We planned to perform sensitivity analysis by excluding trials at high risk of bias (i.e. non‐random sequence generation and inadequate allocation concealment, lack of blinded outcome assessor, lack of blinded participants/personnel, or a combination of these). However, because of the limited number of studies, we deemed this analysis inappropriate.

Summary of findings and assessment of the certainty of the evidence

In the summary of findings tables, we included trials with a follow‐up period longer than 12 months. We created two summary of findings tables comparing intravenous ocrelizumab at the approved dose of 600 mg every 24 weeks; one versus subcutaneous interferon beta‐1a 44 μg three times weekly for RRMS at 96 weeks (summary of findings Table 1), and one versus placebo for PPMS at 120 weeks (summary of findings Table 2).

In summary of findings Table 1, we included seven outcomes.

1. Number of participants experiencing at least one relapse.

2. Number of participants experiencing disability progression.

3. Number of participants with any adverse event.

4. Number of participants with any serious adverse events.

5. Number of participants experiencing treatment discontinuation caused by adverse events.

6. Number of participants with gadolinium‐enhancing T1 lesions on MRI.

7. Number of participants with new or enlarging T2‐hyperintense lesions on MRI.

In summary of findings Table 2, we included six outcomes.

1. Number of participants experiencing disability progression.

2. Number of participants with any adverse event.

3. Number of participants with any serious adverse events.

4. Number of participants experiencing treatment discontinuation caused by adverse events.

5. Number of participants with gadolinium‐enhancing T1 lesions on MRI.

6. Number of participants with new or enlarging T2‐hyperintense lesions on MRI.

We used the five GRADE parameters (risk of bias, inconsistency, imprecision, indirectness, and publication bias) to assess the certainty of the body of evidence as it related to the studies that contributed data to the meta‐analyses for prespecified outcomes. We used the methods and recommendations described in Section 8 and Chapter 14 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2021) using the GRADEpro GDT (GRADEpro GDT). We justified all decisions to downgrade or upgrade the certainty of studies in the footnotes and made comments to aid readers' understanding of the review when necessary.