Table 7.

HIGH clinical trial: Detection of treatmen-by-subset interaction when K=3

	$\hat{\theta }$global CI95%	$\widehat{{\theta }_A}$CI95%	$\widehat{{\theta }_B}$CI95%	$\widehat{{\theta }_C}$CI95%	Decision	nA	nB	nC	Proportion interaction effect*
Age Partition	Age≤58	58<Age≤68	Age>68
1er interim analysis		1.062[0.484-1.972]	0.804[0.444-1.280]	1.558[0.806-2.969]	Go with entire population	61	74	59	0.21& 0.46
2nd interim analysis		0.913[0.518-1.493]	0.979[0.612-1.450]	1.437[0.920-2.217]	Go with entire population	127	135	126	0.10& 0.12
3th interim analysis		0.755[0.475-1.108]	0.975[0.670-1.381]	1.086[0.782-1.485]	Go with entire population	197	202	186	0.01& 0.00
Final analysis	0.993[0.823-1.18]	0.850[0.570-1.209]	0.986[0.719-1.334]	1.141[0.838-1.499]	Go with entire population	255	268	253	0.00& 0.00

The reported intervals are 95% credibility intervals, defined as [quantile(2.5%), quantile(97.5%)] of the posterior distribution.

^*In case of Gail & Simon, it refers to the posterior probabilities P_quali and P_quanti, respectively, as described in equation (5) & (6).