Factor | Categories | Description and comments |
---|---|---|
Number of nodes (treatments) in the network | 3 | The number of nodes speaks directly to the principal objective of whether too few IPD data will have an affect the estimated treatment-effects in a noticeable manner. More nodes means more data are required for full coverage. |
5 | ||
10 | ||
Proportion of treatment comparisons with IPD | Low | Low implied only a single treatment comparison with IPD. Medium implied multiple edges with IPD, but among the lower number of multiple edges possible for the given network. High allowed for up to 100% of edges having IPD |
Medium | ||
High | ||
Effect-modification | None | The relationship between the covariate X and the relative treatment-effects. None indicates no relationship (treatment-effects are unchanged by varying values of the covariate). Constant indicates that the linear relationship between the covariate treatment-effect has the same slope for all treatment-effects relative to the reference treatment. Exchangeable indicates that the slope between covariate and treatment-effect changes according to the treatments being compared, but that they come from a common distribution of slopes. (see Fig.Ā 1) |
Constant | ||
Exchangeable | ||
Trial sizes | All trials of equal sizes | All trials had 200 patients per arm when set to equal. When IPD were larger, the IPD trials had 500 patients per arm. All trials were 2-arm trials. |
IPD trials are bigger | ||
Network density | Sparse | Sparse networks were star networks with no closed loops. They had 1ā3 trials per treatment comparison. Well-populated trials had closed loops and treatment comparisons with up to 7 trials. |
Well-populated |