Table 3.

HRs Estimating Treatment Effect (chemotherapy v observation) per Subgroup (ACN/PAP v MIP/SOL) for OS, DFS, and SDFS (n = 552)

Subgroup	No. of Deaths		HR	95% CI	P	Interaction
	Chemotherapy	Observation				HR	95% CI	P
OS						0.71	0.43 to 1.17	.18
ACN/PAP	60 of 131	50 of 116	1.00	0.68 to 1.47	.99
MIP/SOL	64 of 141	89 of 164	0.71	0.51 to 0.99	.04
Test of heterogeneity across trials					.27*
					.16†
					.18‡
DFS						0.54	0.34 to 0.86	.009
ACN/PAP	75 of 131	59 of 116	1.11	0.78 to 1.57	.57
MIP/SOL	70 of 141	109 of 164	0.60	0.44 to 0.82	.001
Test of heterogeneity across trials					.10*
					.07†
					.05‡
SDFS
ACN/PAP	68 of 131	53 of 116	1.12	0.77 to 1.61	.56	0.53	0.32 to 0.86	.01
MIP/SOL	62 of 141	101 of 164	0.59	0.42 to 0.81	.001
Test of heterogeneity across trials					.26*
					.04†
					.06‡

NOTE. Similar results were observed when using TNM stage instead of T and N of TNM in multivariable Cox regression model (data not shown).

Abbreviations: ACN, acinar; DFS, disease-free survival; HR, hazard ratio; MIP, micropapillary; OS, overall survival; PAP, papillary; SDFS, specific disease-free survival; SOL, solid.

^*Test evaluating homogeneity of treatment effect across trials in ACN/PAP group.

^†Test evaluating homogeneity of treatment effect across trials in MIP/SOL group.

^‡Test evaluating joint hypotheses of * and †.