Lymph node ratio in relation to the survival benefit of immediate breast reconstruction in node-positive breast cancer: A population-based cohort study.

Abstract

e12686 Background: Immediate breast reconstruction (IBR) after mastectomy is increasingly performed, yet its survival impact in node-positive breast cancer remains uncertain and may vary by axillary tumor burden. Lymph node ratio (LNR) may better reflect nodal metastatic extent than conventional N staging. Methods: We identified 14,376 female patients (2010–2017) with non-metastatic, node-positive breast cancer who underwent mastectomy with or without IBR. Propensity score matching yielded 4,050 matched pairs. Multivariable Cox models with interaction terms assessed whether LNR modified the association between IBR and overall survival (OS) or breast cancer–specific survival (BCSS). LNR thresholds for loss of IBR benefit were identified using a stepwise algorithm. Results: A significant interaction between IBR and LNR was observed for both OS (interaction HR = 1.494, 95% CI: 1.111–2.009; p = 0.0078) and BCSS (HR = 1.404, 95% CI: 1.013–1.946; p = 0.0415). IBR was associated with improved OS in patients with LNR ≤ 0.3429 (HR = 0.76, 95% CI: 0.68–0.85; p < 0.001) but not in those with LNR > 0.3429 (HR = 1.02, 95% CI: 0.89–1.17; p = 0.78). For BCSS, the benefit threshold was lower (LNR ≤ 0.1842; HR = 0.82, 95% CI: 0.72–0.93; p = 0.002). Traditional N staging showed inferior discrimination. Conclusions: LNR significantly modifies the survival benefit of IBR in node-positive breast cancer. Incorporating LNR into preoperative decision-making may optimize patient selection for reconstruction. A nomogram integrating LNR and clinicopathologic factors was developed to support individualized risk prediction. Interaction analysis of LNR and surgery type on survival before and after PSM (Cox model). Before or After PSM cohort Before PSMOS Before PSMBCSS After PSMOS After PSMBCSS Model* HR (95% CI), p HR (95% CI) HR (95% CI) HR (95% CI) Model 1 1.628 (1.273–2.082), p < 0.0001 1.422 (1.086–1.863), p = 0.0105 1.416 (1.052–1.906), p = 0.0218 1.327 (0.958–1.839), p = 0.0890 Model 2 1.635 (1.278–2.090), p < 0.0001 1.460 (1.114–1.914), p = 0.0062 1.418 (1.053–1.910), p = 0.0216 1.343 (0.968–1.864), p = 0.0776 Model 3 1.583 (1.239–2.024), p = 0.0002 1.402 (1.070–1.836), p = 0.0141 1.466 (1.089–1.975), p = 0.0118 1.388 (1.000–1.926), p = 0.0500 Model 4 1.634 (1.281–2.085), p < 0.0001 1.433 (1.095–1.874), p = 0.0087 1.497 (1.113–2.013), p = 0.0076 1.406 (1.015–1.949), p = 0.0405 Model 5 1.626 (1.274–2.074), p < 0.0001 1.426 (1.090–1.865), p = 0.0096 1.494 (1.111–2.009), p = 0.0078 1.404 (1.013–1.946), p = 0.0415 *Model 1: Not adjusted Model 2: Model 1 + adjusted for demographic variables (age group, race and origin, marital status, median household income) Model 3: Model 2 + adjusted for clinical and pathology variables (molecular subtype, grade, pathology, T stage) Model 4: Model 3 + adjusted for therapy variables (CT, RT) Model 5: Model 4 + adjusted for surgery related variables (CPM, examined nodes)