Systematic Reviews & Other Review Types

"Network meta-analysis compares multiple interventions simultaneously by analyzing studies making different comparisons in the same analysis."

M. Petticrew et al (2013)

Network meta-analyses are best designed for:

• Conditions with multiple interventions
• Many combinations of direct or indirect interactions 
• Answering more relevant clinical questions
• Making treatment estimates for an entire treatment network instead of scanning each individual pair-wise comparison
• Giving the "full picture" to clinicians
• Gaining precision by considering all available evidence, not just (A vs. B comparisons)
• Potential to more explicitly "rank" treatments using summary outputs

NMA, MTC Meta-Analysis, Multiple Treatment Meta-Analysis, Mixed Treatments Comparison, Multiple Treatments Comparison, Pair-Wise Meta-Analysis, Indirect Treatment Comparison, Multiple Treatment Comparison Meta-Analysis, Live Cumulative Network Analysis (New type)

Timeframe: 12-18+ months.  Same as a traditional systematic review.

Question: Addresses PICO for multiple interventions or treatments (3 or more).  These may be broad or narrow.  Combines indirect (triangular approach) or direct comparisons (pairwise approach). 

Sources and searches: Requires a large search to locate all of the trials using transparent & reproducible methods.

Selection: Same as a systematic review -- based upon clear inclusion/exclusion criteria.  Requires a lot more screening of trials.

Appraisal: Formal quality assessment of all studies.  Still includes Risk of Bias and quality of evidence assessments.

Synthesis: Statistical analysis when possible -- uses different statistical methods than a systematic review. Creates a web-like analysis called a Network Diagram or Network Comparison.  May use bayesian frameworks or meta-regression analysis.

• Requires specialist statistical expertise and software
• Assumes that all interventions included in the "network" are equally applicable to all populations and contexts of the studies included.
• May introduce study selection bias.
• Ease of software may lead researchers to undertake this type of review even though it may not be the most appropriate study design for their research question.
• Complex methodological approaches
• Still an evolving method

M. Petticrew et al (2013) and Li T. et al (2011)

Adapted with permission from Temple University Libraries. https://guides.temple.edu/systematicreviews