Network meta-analysis services for comparing many treatments

A pairwise meta-analysis answers one comparison at a time, which is a poor fit for a field where a dozen treatments compete and few have ever been trialled against one another. A network meta-analysis exists for exactly that situation: it uses the trials you do have to estimate the relative effects you do not, so decision makers can rank the whole set rather than reason across a stack of disconnected pairwise meta-analyses.

The power of the method rests entirely on its assumptions. Borrowing strength across the network is only legitimate when the trials are similar enough for indirect comparisons to be fair, so the analytical work is less about running a model and more about earning the right to run it. We treat the connectedness of your evidence and the plausibility of transitivity as the first questions, not an afterthought once the figures are produced.

Each step below carries a specific assumption or diagnostic, and every one is reported so a reviewer can trace how the rankings were reached rather than take them on trust.

1. 1

   Map the network and check connectedness

   We draw the network geometry, confirm every treatment connects through at least one path, and flag comparisons that rest on a single trial or a fragile link before any estimation begins.

2. 2

   Assess transitivity clinically

   We inspect the distribution of effect modifiers, such as dose, population, and follow-up, across the comparisons, judging whether indirect evidence can be pooled fairly before we trust the model output.

3. 3

   Fit the model and test consistency

   We estimate the network with a frequentist or Bayesian model, then test agreement between direct and indirect evidence using a global design-by-treatment test and local node-splitting on each loop.

4. 4

   Rank, rate, and report

   We produce the league table, SUCRA values and rankograms, rate each comparison with CINeMA, and write the synthesis to the PRISMA-NMA reporting standard.

Both frameworks give valid network estimates, and the right choice depends on your question, your reviewers, and how you want uncertainty expressed. We are fluent in both and recommend the one that fits your evidence rather than defaulting to a house preference.

You get a complete, reportable network synthesis rather than a set of raw model outputs: the network diagram, the league table of relative effects, the forest plots of each comparison against a reference, the ranking summaries, the consistency diagnostics, and a methods-and-results write-up aligned with the PRISMA reporting guideline and its network extension. Every figure ships with the analysis script, so the work reruns end to end and your reviewers can see exactly how each estimate was produced.

• A network diagram showing every treatment, comparison, and the trials that inform each link
• A full league table of relative effects with confidence or credible intervals
• SUCRA values, rankograms, and a clearly caveated treatment hierarchy
• Global and local inconsistency checks, including node-splitting on each closed loop
• A CINeMA confidence rating for every comparison that matters to your question
• A methods-and-results section drafted to PRISMA-NMA, plus the annotated netmeta or Bayesian code

The brief shifts by who is asking and what the ranking has to survive, and we shape the modelling and the write-up to match. What every commissioner shares is a decision that turns on comparing more than two options and a need for the method to hold up under close scrutiny.

A network approach is not always the answer. When your evidence is sparse or poorly connected, or your question really is about a single comparison, a well-run pairwise synthesis is the honest choice. The table below shows where the two diverge so you can see which fits your evidence.

We quote a fixed fee per project rather than by the hour, so the cost is known before any work begins. Behind that number sits the real analytical load: the number of treatments and comparisons, how well connected the network is, whether you need a frequentist or a Bayesian model, whether a review protocol and search already exist or need building, and how much of the reporting you want us to draft. If the evidence will not support a credible network, we say so and propose the analysis that will. Send us your included studies and the treatments you want compared, and we will return a plan, a timeline, and a fixed quote with no obligation to proceed. You can also sketch a single comparison first with our free pooled-effect calculator.