Table 2.

Reductions in Relative Risks of Recurrence and Death and Absolute Differences in Five-Year Disease-free and Overall Survival, According to ER Status. Relative risks were assessed from multivariate proportional hazards models (adjusted for menopausal status, number of positive axillary lymph nodes, and tumor size). Absolute differences for the individual studies were estimated from Kaplan-Meier survival curves. Absolute differences overall were calculated from the relative risk reduction comparing patients in low-dose CAF of Study 8541 vs those patients modeled as though they receiving biweekly AC followed by P as in Study 9741—see Figure 5 as regards DFS.

	Reduction in Risk		Absolute Difference in 5-Yr Survival
	Recurrence	Death	Disease-free	Overall
Study and ER Status	% (95% Confidence Interval)		%
8541 (high dose vs. low dose)
ER-negative	21 (9 to 31)	17 (4 to 29)	13.9	6.6
ER-positive	9 (−6 to 22)	6 (−11 to 20)	6.6	4.0
9344 (paclitaxel vs. no paclitaxel)
ER-negative	25 (12 to 36)	24 (10 to 37)	8.2	7.4
ER-positive	12 (−3 to 25)	11 (−8 to 26)	2.1	0.0
9741 (every 2 wk vs. every 3 wk)
ER-negative	24 (1 to 42)	28 (1 to 47)	9.1	7.4
ER-positive	8 (−20 to 29)	8 (−28 to 35)	2.8	−0.2
Overall (every 2 wk in 9741 vs low dose in 8541)
ER-negative*	55 (37 to 68)	55 (38 to 69)	22.8	16.7
ER-positive*	26 (−4 to 48)	23 (−17 to 49)	7.0	4.0

^*The interaction between higher chemotherapy and ER status was assessed using proportional hazards model, with added terms for study and main effects for higher chemotherapy and ER status. The interaction was statistically significant (p = 0.02 for recurrence and p = 0.05 for survival) for the overall comparison but not when comparing the studies separately.