Skip to main content
. 2011 Aug 16;343:d4909. doi: 10.1136/bmj.d4909

Table 3.

 Statistically significant results from direct and adjusted indirect comparisons

Comparison methods Direct comparisons of A and B Total
Significant results (P<0.05) Non-significant results
All trial networks 39 73 112
Indirect comparisons of A and B:
 Significant results 14* 9 23
 Non-significant results 25 64 89
Combination of direct and indirect estimates†:
 Significant results 25 3 28
 Non-significant results 14 70 84

*In one case (CD003262), both direct and indirect estimates were statistically significant but in opposite directions (see web appendix 1).

†DerSimonian-Laird random effects model used to combine direct and indirect estimates.