# NMA / ITC Engine
> Network meta-analysis and MAIC / STC, submission-grade, fully traced.
Source: https://www.visfo.health/product/nma-itc-workbench
Tags: Platform, AI-assisted

_Network meta-analysis and MAIC / STC, submission-grade, fully traced._

Builds the trial network live from the Evidence Synthesis Engine, runs network meta-analysis and indirect treatment comparisons (anchored, unanchored, MAIC, STC) and produces reviewer-ready outputs with full provenance back to each trial. Feeds Model Studio, GVD and the HTA Console.

## What it is

The shared comparator engine for an asset. One trial network, kept current with the Evidence Synthesis Engine, that medical, HEOR and access all reason over rather than re-building per submission.

It is not a black-box statistical service. Every node is a study you can open, every edge is a comparison you can interrogate, every effect carries its model assumptions, its diagnostics and its sensitivity runs. The point is a comparison the HTA reviewer can follow, not a number you have to defend.

## How we run it

1. **Build the network** — Pull the included trial set from the Evidence Synthesis Engine. Construct the comparator network for the locked PICO, with feasibility checks on connectedness and similarity.

2. **Pick the right method** — Anchored NMA where the network connects, MAIC or STC where it does not, with the rationale written down and the population-adjustment covariates declared up front.

3. **Run, diagnose, sensitivity-check** — Fixed and random effects, heterogeneity, inconsistency checks, leave-one-out and pre-specified subgroups. Every diagnostic surfaced, not buried in an appendix.

4. **Publish with provenance** — Submission-grade outputs ship to the synthesis bus. Model Studio, GVD and the HTA Console read the same comparative effect, with the model card, code and trial set attached.

## What you get

- Living trial network with included and excluded studies and the reasons

- Pooled comparative effects with credible intervals, heterogeneity and inconsistency diagnostics

- MAIC or STC results where the network does not connect, with covariate balance and effective sample size

- Reviewer-ready package: model card, code, trial set, sensitivity runs, all versioned

- Live feeds into Model Studio, GVD and the HTA Console so every downstream comparison reads the same number