Abstract Background: The incidence of central nervous system (CNS) metastases varies with BC subtype, but there is limited data about clinical and molecular predictors of CNS vs non-CNS recurrence risk after neoadjuvant chemotherapy (NACT). Here we present patterns of distant recurrence among patients (pts) in the I-SPY2 trial. Methods: This analysis included 2023 pts with molecularly high-risk stage II and III BC who enrolled in the I-SPY2 trial from 2010-2022 with event-free survival (EFS) data (as of 6/22/25). We analyzed type of recurrence (local vs distant) and site(s) of initial distant recurrence (CNS-only, both CNS- and non-CNS, or non-CNS only). We used a competing risk (Fine-Gray) model to estimate cumulative incidence of site-specific relapse at five years and assessed associations with clinical/molecular subtype, clinical staging (cT, cN), and response to NACT by Residual Cancer Burden (RCB) class. Results: Among 2023 evaluable pts, at a median follow up of 4.7 yrs there were 361 EFS events, most of which were distant recurrences (n=259, 72%). Of the pts with distant recurrences, most had initial non-CNS events (180/259, 69%), 36 pts had initial simultaneous CNS- and non-CNS events (36/259, 14%), and 43 pts had initial CNS-only events (43/259, 17%). Table 1 summarizes the cumulative incidence of initial CNS- and/or non-CNS recurrences at 5 years by clinical/molecular features. Rates of CNS-only recurrence were low and similar across clinical and response predictive subtypes (p=0.47 and =0.19 respectively). Risk of distant recurrence increased in pts with larger tumors and node positive disease at both CNS- and non-CNS sites (see Table 1). Rates of initial non-CNS recurrence increased with the amount of residual disease at surgery, with lowest rates in pts with RCB-0 and highest rates in pts with RCB-III (p0.01). In contrast, the rate of CNS-only recurrence was low but remained similar across the four RCB classes (p=0.67). Because of this, 39% (18/46) of pts with RCB-0/I disease had initial CNS-only recurrence, whereas only 12% (24/204) of pts with RCB-II/III disease had initial CNS-only recurrence. Most CNS recurrences, isolated or not (64/79, 81%), occurred within the first three years after surgery. Conclusions: In pts with molecularly high-risk stage II and III BC who underwent NACT, the incidence of initial CNS-only metastases was low, but the rate was consistent across the four RCB classes, supporting that the CNS is a sanctuary site. Initial CNS-only recurrence accounted for 39% of EFS events for pts with RCB-0/I vs 12% for pts with RCB-II/III. CNS recurrence risk was similar across high-risk BC subtypes, but higher in pts with larger tumors and node positive disease. These results support future research to identify biomarkers that predict CNS recurrence risk and to incorporate CNS-penetrant therapies into early-stage treatment for CNS high-risk pts. Citation Format: L. Huppert, C. Yau, D. Idossa, A. Kahn, F. M. Howard, E. Shagisultanova, A. Zimmer, E. Stringor-Reasor, J. Perlmutter, D. Yee, R. Nanda, N. Hylton, W. F. Symmans, R. Shatsky, C. Isaacs, L. Van't Veer, H. S. Rugo, P. R. Pohlmann, A. DeMichele, L. Pusztai, L. Esserman. Site(s) of Distant Recurrence after Neoadjuvant Therapy for Patients With High-Risk Early-Stage Breast Cancer (BC): Analysis of Data from the I-SPY2 Trial abstract. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PD13-01.
Huppert et al. (Tue,) studied this question.