Investigating treatment-effect modification by a continuous covariate in IPD meta-analysis: an approach using fractional polynomials

Table 2 Number of patients and events per study. Percentage of events per treatment group and p-values from the test for interactions (FP1, flex3), overall (pooled data, stratified by trial) p = 0.0215

	Trial	Patients	Events		% events by treatment		p-values	p-values
			No	Yes	CT-N	CT-Y	CT	CT-ER Interact.
1	IBCSG-3	152	25	127	86.7%	80.5%	0.1519	0.0917
2	IBCSG-7	592	177	415	71.1%	69.1%	0.2089	0.8729
3	IBCSG-9	1177	730	447	38.7%	37.2%	0.2945	0.1198
4	NCIC-MA4	731	394	337	46.8%	45.4%	0.5505	0.2195
5	NSABP-B16–1	224	74	150	63.5%	68.7%	0.6368	0.2444
6	NSABP-B16–2	1071	359	712	72.2%	63.6%	0.0009	0.1027
7	NSABP-B20	966	646	320	34.7%	31.5%	0.3501	0.5155
8	SWOG-S8814	814	373	441	58.5%	51.2%	0.0320	0.0081

ISSN: 1471-2288