NICE has 2 appraisal processes, multiple- and single-technology appraisal (MTA, STA). In
an MTA, a set of competing new products are compared with each other and with standard
comparators, whereas in an STA a single new product is under scrutiny. Although an NMA is
almost always used in MTAs, STAs may be based on pairwise synthesis or NMA. However, the
methods guide requires that the methods of assessment be the same. The evidence synthesis
models and the WinBUGS software for implementing them presented throughout this tutorial
series10 (link),11 (link) have precisely this
property: The software will run either type of analysis without distinction, and the
pairwise model is simply a special case of the NMA model in which only 2 treatments are
being compared. This is not necessarily the case in other models, or software, for NMA.
The Lumley model22 (link)for NMA cannot be run on pairwise meta-analysis or on indirect comparisons, nor in its
published form can it deal correctly with trials comparing more than 2 treatments. These
are serious shortcomings, as manufacturers, clinicians, and decision makers will want to
be assured that there is a single, fair way of estimating treatment effects, regardless of
the structure of the evidence.
an MTA, a set of competing new products are compared with each other and with standard
comparators, whereas in an STA a single new product is under scrutiny. Although an NMA is
almost always used in MTAs, STAs may be based on pairwise synthesis or NMA. However, the
methods guide requires that the methods of assessment be the same. The evidence synthesis
models and the WinBUGS software for implementing them presented throughout this tutorial
series10 (link),11 (link) have precisely this
property: The software will run either type of analysis without distinction, and the
pairwise model is simply a special case of the NMA model in which only 2 treatments are
being compared. This is not necessarily the case in other models, or software, for NMA.
The Lumley model22 (link)for NMA cannot be run on pairwise meta-analysis or on indirect comparisons, nor in its
published form can it deal correctly with trials comparing more than 2 treatments. These
are serious shortcomings, as manufacturers, clinicians, and decision makers will want to
be assured that there is a single, fair way of estimating treatment effects, regardless of
the structure of the evidence.
