Network Meta-Analysis and Indirect Treatment Comparison Services
Pubrica provides network meta-analysis (NMA) and indirect treatment comparison services for HTA and market access teams, pharmaceutical companies, guideline groups and academic researchers. These methods compare treatments that have not been tested head to head, such as a new medicine and the alternatives clinicians prescribe. Pubrica’s biostatisticians and evidence synthesis specialists assess feasibility before the network is built, examine transitivity against the data and report inconsistency in full, and deliver the analysis, the report and reproducible code.
- Feasibility assessed before the network is built, so that disconnected networks and incomparable populations are identified before any analysis
- Transitivity examined, not asserted against the effect modifiers in the study data, since every indirect comparison depends on it
- Inconsistency tested where a loop exists, by node-splitting or an equivalent method, and reported in full
- MAIC and STC where the comparison is unanchored, to NICE DSU TSD 18, with the underlying assumptions stated explicitly
- Rankings reported with their caveats, because SUCRA values and rankograms are often over-interpreted and a rank is not an effect size
Feasibility and network geometryIncluded
Transitivity examined against the dataIncluded
Inconsistency tested where loops existIncluded
Reproducible code handed overIncluded
Analysing a network that does not connectNot performed; reasons reported
Evidence → feasibility → network → tested → comparison
- Feasibility reported before analysis
- Transitivity examined against study data
- Inconsistency tested, not assumed
- Code and data handed over
- Rankings reported with caveats
Service Scope
Our Network Meta-Analysis Services
Pubrica offers four network meta-analysis service areas that can be commissioned individually, because projects reach us at different stages, some as a research question and others with data already extracted.
Who We Serve
Researchers and Organisations We Support with Network Meta-Analysis
We support academic, clinical, HTA and industry teams, most of whom use the results to inform a treatment, reimbursement or guideline decision.
Deliverables
Four Deliverables Included in Every Network Meta-Analysis
Each project assesses whether a valid comparison can be made before any model is run.
Example from a feasibility assessmentA disconnected network cannot be analysed as a single network. A connected network of incomparable populations can be analysed but should not be, and this case carries greater risk because it still produces an estimate.
Example from a transitivity assessmentThe transitivity assumption was considered to be met.Trials in the A-versus-placebo comparison enrolled treatment-naive patients; the B-versus-placebo trials were second-line. Transitivity is questionable for this loop and the indirect estimate is presented with that limitation rather than alongside the others.
Example from a consistency checkWhere direct and indirect evidence disagree, the disagreement is reported as a finding. Pooling them into a single network estimate and reporting only that would conceal the most informative result of the analysis.
Example from a handoverHTA clarification questions often request a scenario analysis at two weeks’ notice, so the client team needs code it can rerun independently.
The samples shown are illustrative and anonymised. Hover over a sample to replay it.
Compliance and Standards
Network Meta-Analysis Methodological Standards
These are the published methods that HTA assessors apply. Any departure from them is justified in the report.
NICE DSU technical support (synthesis methods)
The generalised linear modelling framework for pairwise and network meta-analysis, heterogeneity, inconsistency and population-adjusted indirect comparison. These are the methods UK reviewers apply and a widely accepted default in other settings.
ISPOR–AMCP–NPC good practice (indirect comparison)
The task force reports on interpreting indirect treatment comparisons and network meta-analysis, and the questionnaire assessors use to decide whether a comparison is credible enough to act on.
PRISMA 2020 (and the NMA extension)
The reporting standard for systematic reviews, with the network meta-analysis extension adding the network geometry, the assumptions tested and the ranking caveats a reviewer expects to see stated.
Quality management (cert. H3CR9TCU20251212INDQM10N)
Our certified quality system governs how work is run, recorded and released.
Responsible Use of AI
Responsible Use of AI in Model Selection and Consistency Checking
In network meta-analysis, the main question about AI is whether the model, priors and consistency checks were chosen by a statistician who can justify them. Software will fit a network without indicating that its geometry cannot support the analysis.
Why We Do Not Publish AI Accuracy Figures
Accuracy figures for AI tools are usually reported by the companies that sell them.
Pubrica therefore does not quote a self-generated accuracy figure, because it would carry the same conflict of interest. Instead, we tell you which tool was used on your project, at which stage, and what a specialist changed afterwards.
Methodology
How Our Network Meta-Analysis Process Works
The process has four stages. The feasibility assessment is a deliverable in its own right and can be commissioned on its own.
Evidence review and network mapping
Trials identified from a systematic review, the network drawn, and its geometry described — which comparisons are direct, which are indirect, where the loops are.
Feasibility assessment
Population, comparator and outcome comparability assessed across the network, with effect modifiers identified and differences reported.
Statistical analysis
Fixed and random effects models fitted, heterogeneity and inconsistency examined, and sensitivity to the questionable assumptions tested.
Reporting and handover
Results are reported in the format the submission requires, with assumptions and limitations stated, and the code and data are delivered.
Our Experts
Our Network Meta-Analysis Experts
A named team of subject specialists stays with your project from scoping to delivery. Your project coordinator introduces each team member by name before work begins.
Responsible for every decision that depends on the numbers. Confirms before analysis begins whether the data can answer the question, and hands over the code so the analysis can be re-run independently.
Determines what the work can claim on the basis of the evidence. The subject expert is matched to your topic and tells you clearly when the evidence does not support the intended conclusion.
Reviews the work independently of the first pass and reconciles the two, because most errors in this work are transcription errors rather than errors of reasoning. All disagreements are logged and adjudicated.
Identifies the evidence on which the work is based and documents the search so that it can be reproduced. Writes the search strategy line by line for each platform and delivers it dated and reproducible.
Pricing
Network Meta-Analysis Packages
Pricing depends on the number of trials in the network, the number of outcomes and whether population adjustment is required.
Feasibility assessment
Before committing to the analysis
Quoted per projectOn network size
Delivery Quoted by scope
- ✓Network drawn and geometry described
- ✓Connectivity established
- ✓Population and outcome comparability assessed
- ✓Effect modifiers identified
- ✓A clear recommendation on whether to proceed
Where the network cannot support the analysis, this is the lowest-cost way to find out, and it allows you to adjust your submission plan before analysis begins.
Deliverables Request a quoteNetwork meta-analysis
Connected networks
Quoted per projectOn trials and outcomes
Delivery Quoted by scope
- ✓Fixed and random effects models
- ✓Heterogeneity examined
- ✓Inconsistency tested at every loop
- ✓Rankings with their caveats
- ✓Sensitivity analyses on the weak assumptions
- ✓Code and data handed over
Built in R as standard. WinBUGS or JAGS where an agency or a journal expects it.
Deliverables Request a quotePopulation-adjusted comparison
Where there is no common comparator
Quoted per projectOn IPD availability and covariates
Delivery Quoted by scope
- ✓MAIC or STC as the data allows
- ✓Built to NICE DSU TSD 18
- ✓Effective sample size reported after weighting
- ✓The unanchored assumption stated explicitly
- ✓Sensitivity to covariate selection tested
An unanchored comparison assumes that every prognostic variable and effect modifier has been accounted for. Assessors recognise how strong this assumption is, so the report states it explicitly.
Deliverables Request a quoteSend your trial list and the comparators you have been asked about and we will draw the network at no cost, so you can see whether it connects before commissioning anything.
Benefits
Benefits of a Pubrica Network Meta-Analysis
Network meta-analysis provides four main benefits for HTA, guideline and publication work.
Quality Assurance
Our Service Commitments and Limitations
Our commitments and their limits are set out in full below, so that you can review them before you commission any work.
Our revision commitment
Revision Support Within the Agreed Scope
We continue working within the agreed scope until the work meets the agreed requirements. We offer continued revision support rather than refunds.
01
Unlimited revisions within the agreed scope
There is no cap on revision rounds and no time limit within the agreed scope. If the work does not meet what we agreed to deliver, we revise it. Any widening of the scope is quoted separately, and we confirm which applies before work starts.
02
Corrections at our cost
If the work comes back because of something we ran or wrote, correcting it is free.
03
Delivery on the agreed date
We quote a delivery date we can meet. If a milestone is at risk, we tell you early.
04
Mutual NDA on request
Signed before you send anything, in both directions, at no cost, and we are happy to sign your NDA rather than ours.
05
Final decisions remain with you
We advise, give our view and tell you clearly when we disagree, but decisions about what is published in your name are always yours.
Average client rating
From 3,000 responses to Pubrica’s own post-delivery survey, sent to every client at handover, January 2024 to June 2026. This figure comes from our own records rather than a third-party rating.
Service limitations. We cannot guarantee an outcome that belongs to somebody else — an editor’s decision, a funder’s panel, a regulator’s assessment. What we guarantee, within the agreed scope, is the work: done to the standard named on this page, delivered when we said, and corrected at our cost if we got it wrong.
FAQs
Network Meta-Analysis FAQs
If your question is not answered here, please ask an expert. No quote is required.
A disconnected network cannot be analysed as a network meta-analysis. The options are an unanchored comparison with population adjustment, which carries a much stronger assumption, or accepting that the comparison cannot be made.
We will tell you which applies before quoting the analysis. A forced network produces an estimate that assessors are likely to discount.
You can, with caveats. Rankings are routinely over-interpreted: a treatment can rank first on very sparse evidence, and a rank is not an effect size or a confidence statement.
We report them alongside the effect estimates and their uncertainty, so a reader cannot take the rank without the precision behind it.
R as standard — netmeta for frequentist analysis, gemtc or multinma for Bayesian work. It is easier to reproduce, easier to hand over and easier for your own team to rerun.
WinBUGS or JAGS where an agency, a journal or your own SOP expects it. The choice is yours and we will say when it makes no difference to the answer.
Yes, and that is why the code and network data are handed over. Adding a trial to an existing network is a fraction of the cost of rebuilding it.
Where the analysis supports a live HTA submission, an update on a scheduled basis is often the cheaper arrangement.
Related Services
Related Research Services
Network meta-analysis depends on the systematic review that identifies the trials, and these related services support each stage.
1EVIDENCE SYNTHESIScompleted first
Systematic reviewIdentifies the trials in the network. Meta-analysisPairwise pooling where head-to-head trials exist.2SUBMISSIONSdownstream
EU Joint Clinical AssessmentSupports multi-PICO assessment scopes. Cost-effectiveness modelProvides the effect sizes for the model.3STATISTICAL SUPPORTsame team
BiostatisticsAnalysis plans and reproducible code. Payer evidence reviewFeasibility assessment and comparison.Request a Free Evidence Network Diagram
We will show you whether the network connects and whether the populations are comparable before you commit a budget.