Statistical interaction in the survival analysis of early breast cancer using registry data: role of breast conserving surgery and radiotherapy

To identify subgroup effects that might influence the survival results of postoperative radiotherapy. Women selected from the Surveillance, Epidemiology, and End Results database, aged 40-69 years, with non-metastasized T1-T2 breast carcinoma, in whom axillary lymph node dissection was performed. Subgroup analyses were performed using proportional hazards models with interactions. Joint significance of subgroups was evaluated with the Wald test. Event was death from any cause. Statistically significant interactions were found between type of surgery (breast-conserving [BCS] or mastectomy [ME]), radiotherapy [RT], T stage, and extent of nodal involvement, but not between treatments and nodal examination. For each treatment combination, ME-no RT, ME+RT, BCS-no RT, BCS+RT, the mortality hazard ratios were respectively: 1, 1.12, 1.11, 0.78 in T1, 0-3 positive nodes; 2.45, 2.77, 2.71, 1.92 in T2, 4+ nodes; 1.31, 1.38, 1.33, 1.19 in T2, 0-3+ nodes; and 3.41, 2.79, 3.44, 2.40 in T2, 4+ nodes. The corresponding joint tests showed: in the absence of radiotherapy, no significant survival disadvantage for breast-conserving surgery vs mastectomy; with radiotherapy, significant survival advantage for breast-conserving surgery irrespective of stage and for mastectomy in T2, 4+ nodes. For mastectomy in less advanced stages receiving radiotherapy, excess breast cancer deaths suggested undocumented adverse selection. The corresponding result was considered inconclusive. The analyses found subgroup effects that should be taken into account to interpret treatment results in breast cancer.