Is a completion axillary dissection in patients with early breast cancer with a positive sentinel node necessary? An NCDB-based trial emulation of ACOSOG Z0011.

Abstract

Listen

Translate article

e12592 Background: The ACOSOG Z0011 Study is a practice-changing phase III clinical trial that promotes the de-escalation of axillary surgery in early breast cancer patients, which supports no completion axillary lymph node dissection (ALND) in clinically node negative patients with 1 or 2 positive sentinel lymph nodes (pSLN). Strong evidence for patients with 3 positive SLN was still lacking. We emulated the Z0011 trial using the National Cancer Database (NCDB) to validate the post-trial evidence in the real-world setting followed by a deeper investigation in the subgroups by the number of pSLN. Methods: NCDB PUF 2020 was queried for clinical T1-2 N0 M0 breast cancer patients diagnosed 2010-2017. The inclusion and exclusion criteria were the same as in the Z0011 trial except for the inclusion of 3 positive sentinel lymph nodes. Sentinel lymph node biopsy (SLNB) alone and ALND were defined using Site-Specific Factor 19. The primary endpoint, overall survival (OS), was modeled by the Cox proportional hazard model. The propensity score-based average treatment effect on the treated (ATT) weighting schema was implemented to generate pseudo samples based on this eligible study population and its subgroups as pSLN = 1, 2, and 3, where covariate balance between the two cohorts was achieved for all observed demographic and disease characteristics. Results: 25,774 patients were included in the analysis. 20311 (78.8%) patients were in the SLNB alone group, and 5463 (21.2%) patients were in the ALND group respectively. The utility rate of SLNB increased steadily over the years from 53.7% in 2010 to 91.0% in 2017. The median age was 60, which was older than that in the Z0011 trial (55). The median follow-up time was 6.78 [95% CI: 4.98-8.81] years. By multivariable analysis using the original sample, SLNB alone and the ALND group had similar survival (HR = 0.99 [95% CI: 0.90-1.09]; P = 0.83). By ATT weighting, the SLNB group had significantly improved survival (HR= 0.88 [95% CI: 0.82-0.95]; P = 0.001). The magnitude of HR is close to that in the Z0011 trial (HR= 0.87 [95% CI: 0.62-1.23]), but the statistical significance was mainly driven by a much larger sample size in this study. In the subgroup ATT weighting analysis for pSLN = 1, the SLNB group had significantly better survival compared to the ALND group (HR = 0.85 [95% CI: 0.79-0.93]; P < 0.001). Survival was equivalent for subgroup patients with pSLN = 2 (HR = 1.04 [95% CI: 0.91-1.19]; P = 0.54), while in the subgroup of pSLN = 3 patients, the SLNB group had a significantly worse survival when compared to the ALND group (HR, 1.22 [95% CI, 1.04-1.45]; P = 0.02). Conclusions: This NCDB-based Trial-Emulation study validates the result of the ACOSOG Z0011 clinical trial through a post-trial and real-world setting, where we confirmed that patients with 1 or 2 pSLN are suitable for SLNB alone, but for patients with 3 pSLN, ALND may still maintain a viable option.