Circulating tumor DNA to anticipate loco-regional recurrence in early-stage breast cancer: a proof-of-concept study.

Loco-regional recurrence (LRR) poses a clinical challenge for the follow-up of patients treated with curative intent for early-stage breast cancer (EBC). While circulating tumor DNA (ctDNA) has been shown to predict distant metastases, its value for LRR is less characterized. Starting from an index case with documented LRR and available tumor and plasma samples, we report the analysis of the prospective phase III fenretinide prevention trial, which primarily aimed to assess the incidence of second malignancy in women with T1-T2 N0 EBC. Patients were eligible if they had FFPE and/or frozen tissue from primary or recurrent invasive tumor for next generation sequencing, and at least three serial plasma samples for ctDNA analysis by digital PCR. The TP53 R196* mutation was identified in the primary tumor of the index case with a variant allele frequency (VAF) of 29%, and in the LRR with a VAF of 58%. The same mutation was also detected in plasma prior to both the primary and LRR surgeries with VAFs of 0.19% and 0.12%, respectively. Following treatment, the mutation became undetectable in plasma samples during follow-up, consistent with the absence of recurrence. Among 40 eligible patients from the fenretinide prevention trial, 27 (67.5%) had primary tumor somatic variants trackable in plasma. Median age was 55 years (range, 35-78); stage I (16, 59%) and stage II (11, 41%); mostly luminal-like (19, 70%); median follow-up 173 months (range, 98-193); common mutations included PIK3CA (50%), TP53 (30.7%), and PTEN (5.9%). Six patients developed LRR as first event; 4 distant metastases. In all LRR cases, except one, ctDNA was detected prior to surgery and anticipated the clinical diagnosis up to 28 months. Three patients with LRR developed distant metastases 1 to 2 years later. These findings show the potential of ctDNA for the early detection of LRR in EBC, and its promise as a tool for timely interventions and personalized surveillance strategies.