Table 6 Summary of steps involved in NMA

Systematic literature review

Prospective registration with PROSPERO

Well-defined research question using the PICOS framework

Searches carried out using a pre-defined search string (specific to each database)

Systematic inclusion/exclusion of studies per the research question

PROSPERO: [36]

Cochrane handbook: [38]

Data extraction and network generation

Quality/risk of bias assessment

Treatment network defined

RoB 2 tool: [40]

Cochrane handbook: [26]

Assessment of NMA assumptions

Similarity: similarity in PICOS criteria of all included studies

Transitivity: no systematic differences in the distribution of effect modifiers between included studies

Consistency: agreement between direct and indirect evidence within the network

Homogeneity: no imbalances in PICOS across direct and indirect comparisons within the network

Cochrane handbook: [26]

Conducting an NMA

Appropriate statistical model used for the available data and/or any specific country requirements

Justified use of FE vs RE methods

Appropriate presentation of results

  For frequentist analysis: estimates of effects and corresponding 95% CIs and associated p-values

  For Bayesian analysis: estimates of effects and corresponding 95% CrIs

Cochrane handbook: [26]

Bucher 1997: [47]

Netmeta: [74]

NICE DSU: [12]

Interpretation of NMA findings

Appropriate and careful interpretation of findings

  For frequentist analysis: ranking of treatments through p-scores. Can be interpreted as statistical significance or absence thereof

  For Bayesian analysis: ranking of treatments through SUCRA. No significance testing

Use of the GRADE framework to assess the confidence in the evidence

Cochrane handbook: [26]

NMA worked example for clinicians: [75]

GRADE resources: [69, 70]

Reporting of NMA findings

Communicated following the PRISMA guidelines for NMA

PRISMA: [37]