A: General information about available data | ||
Data sources | Breast Cancer DataMart (Provided by NCI) Contains IPD from 14 published NCI-sponsored breast cancer studies. 8 studies assessed effect of adding chemotherapy (CT-Y/N) to tamoxifen. Here we include only tamoxifen treated patients. For details see section 4 in Wang et al. | |
Study question | Does effect of CT depend on estrogen receptor (ER) values? | |
Data | CT-N: N = 2982, No ER = 401, in study 2581 CT-Y: N = 3586, No ER = 440, in study 3146 Outcome: disease free survival (DFS) Events: Ct-N = 1342, CT-Y=1601, overall. See Table 2 for more details. We use ER+1 (see “General issues of the three-stage metaTEF approach”) and we truncate all ER values at 1000, which made little difference (see “Estrogen receptor”). | |
Variables | Treatment: CT (Y/N); Outcome: recurrence free survival (RFS); Predictor: ER | |
B: Description of data and main effect of CT | ||
D1 – per study | Patients, Follow-Up, Events, p-values (CT-Y/N) | Table 2 |
D2 – per study | Estimates of DFS rates, combined for CT | Fig. 1 |
D3 – per study | Effect of CT in each study | Fig. A1, p-values in Table 2, check of PH assumption from the Cox model. |
D4: ER – per study | Distribution | |
C: Analyses to investigate interactions of treatment with a predictor | ||
A1 – per study | Interaction of CT with ER? | Table 2, Preliminary analyses |
A2 – ER subgroups studywise and pooled estimate | Effect of CT in 4 ER subgroups | |
A3 – TEF – studywise and pooled estimate | Treatment effect function (TEF) for log ER | Power 0 (log ER) in each study (see Fig. A4), Preliminary analyses |
A4 – studywise FP1 | TEF (FP1, flex3) separately per study | Fig. 2a, metaTEF |
A5 – metaTEF for A4 | Main analysis - Variances, weights per study and metaTEF | Fig. 2b – f |
A6 – studywise and pooled FP2 | As A4 but with FP2 (flex1) | sensitivity analysis FP1 or FP2? |
A7 – metaTEF | Comparison of FP1/FP2 and fixed/random effect | Fig. 3 |
A8 – CI of metaTEF | Pointwise 95% CI for FP1/FP2 and fixed/random effect | Fig. A6 |