When does the use of individual patient data in network meta-analysis make a difference? A simulation study

Table 2 Summary statistics of the mean squared error of treatment-effect for the two meta-regression adjusted NMA models

Scenario	IPD Mean	IPD Median	IPD Std Dev	IPD Range	AgD MR Mean	AgD MR Median	AgD MR Std Dev	AgD MR Range
Mean squared-error
Number of nodes: 3	0.014	0.004	0.035	0.811	2.065	0.015	79.472	3886.193
Number of nodes: 5	0.009	0.002	0.02	0.397	0.044	0.007	0.162	3.587
Number of nodes: 10	0.005	0.002	0.009	0.093	0.014	0.004	0.028	0.349
Proportion edges with IPD: low	0.013	0.004	0.033	0.811	1.833	0.008	79.413	3886.193
Proportion edges with IPD: medium	0.008	0.002	0.02	0.38	0.144	0.007	2.443	115.198
Proportion edges with IPD: high	0.006	0.002	0.016	0.244	0.146	0.007	2.057	91.712
Trial size: equal	0.013	0.003	0.032	0.811	0.155	0.009	2.306	115.198
Trial size: large ipd	0.006	0.002	0.013	0.38	1.26	0.006	64.853	3886.193
Network density: sparse	0.015	0.005	0.033	0.811	1.405	0.019	64.891	3886.193
Network density: well-populated	0.004	0.001	0.007	0.105	0.01	0.004	0.019	0.284
Bias
Number of nodes: 3	0	0	0.119	1.517	0.029	0.005	1.437	73.842
Number of nodes: 5	0.003	0.001	0.093	1.11	0.002	0	0.209	3.773
Number of nodes: 10	0	−0.002	0.072	0.597	−0.004	−0.007	0.118	1.152
Proportion edges with IPD: low	0.003	0	0.115	1.531	0.024	0	1.354	73.842
Proportion edges with IPD: medium	0	−0.003	0.092	1.158	−0.001	−0.004	0.379	14.589
Proportion edges with IPD: high	0	0	0.079	0.977	0.003	0	0.382	13.541
Trial size: equal	0.001	−0.001	0.113	1.531	−0.006	− 0.002	0.393	18.309
Trial size: large ipd	0.001	0	0.076	0.947	0.023	0	1.123	73.842
Network density: sparse	0.002	0.001	0.121	1.531	0.017	0	1.185	73.842
Network density: well-populated	0	−0.001	0.063	0.601	0	−0.002	0.1	0.974

ISSN: 1471-2288