Circulating tumor DNA as an arbitrator for postmastectomy radiotherapy: bridging molecular residual disease and locoregional control Authors: Gai Liang, Qu Zhang, Mengting Yuan, Guojun Yao, Feng Yuan Published: 2026 ABSTRACT AbstractThe NSABP B-51 trial left unresolved whether postmastectomy radiotherapy benefits ypN0 patients after neoadjuvant chemotherapy. Meanwhile, the PREDICT-DNA study showed that ctDNA clearance after neoadjuvant therapy identifies patients with favorable outcomes irrespective of pathologic response. We propose integrating these datasets: ctDNA status may serve as a tiebreaker for radiotherapy decisions, particularly in patients with residual disease, offering a molecularly guided framework for de-escalation.Subject terms: Biomarkers, Cancer, Oncology The B-51 dilemma The NSABP B-51/RTOG 1304 trial asked a straightforward question: can regional nodal irradiation (RNI) be safely omitted in patients with cT1-3N1 disease who convert to ypN0 after neoadjuvant chemotherapy?1 For the overall population, the answer appears to be yes. The 5‑year invasive breast cancer recurrence‑free interval was 91.8% without RNI and 92.6% with RNI (HR 0.88; 95% CI, 0.60–1.29)—a difference too small to justify routine irradiation. The mastectomy subgroup, however, refuses to cooperate. Among 648 patients who underwent mastectomy, PMRT produced a 3.1% absolute improvement in 5‑year invasive recurrence‑free interval (92.6% vs. 89.5%; HR 0.72; 95% CI, 0.42–1.23). Mutter and Ballman2 note that this 28% relative risk reduction mirrors the ~30% recurrence reduction seen in the EBCTCG meta‑analysis of PMRT in node‑positive disease3. But because B-51 was a superiority trial, the absence of statistical significance tells us precisely nothing about non‑inferiority. We simply do not know whether PMRT can be safely omitted in this group - the trial was never powered to tell us. And so the clinician faces a familiar bind. A mastectomy patient with ypN0 status sits before you. Guidelines, based on the highest clinical or pathologic stage, offer no clear directive. Do you irradiate, accepting the toxicities of chest wall and regional nodal treatment, or observe, accepting the possibility of preventable recurrence? The B-51 data cannot answer this question. Insights from PREDICT-DNA PREDICT‑DNA offers a different way in ref. 4. Using a tumor‑informed, ultrasensitive ctDNA assay capable of detecting tumor DNA at <100 parts per million, the study evaluated 227 patients with stage II/III HER2‑positive or triple‑negative breast cancer receiving neoadjuvant therapy. Two findings matter here. First, ctDNA after neoadjuvant therapy was a substantially stronger prognostic marker than pathologic complete response (pCR). Detectable ctDNA carried a hazard ratio of 8.9 (95% CI, 2.4–33; P = 0.001) for recurrence in triple‑negative disease—independent of pCR status. Second, and more telling, 40% of patients with residual histologic disease had no detectable ctDNA after neoadjuvant therapy. These ctDNA‑negative, non‑pCR patients showed 5‑year invasive disease‑free survival of 92%, indistinguishable from the 90% seen in those with pCR. It is critical to note, however, that in the PREDICT-DNA study, all patients who achieved a pCR were also ctDNA-negative after neoadjuvant therapy (see Supplementary Table S3 in Hunter et al.4). This concordance suggests that in the B-51 population, ctDNA assessment may provide its most significant additional value for patients with residual disease, where it could distinguish between those with and without molecular residual disease. A molecular framework for radiotherapy decisions The implication is straightforward: a negative post‑neoadjuvant ctDNA result identifies patients whose residual disease—whatever its histologic appearance—lacks the biological machinery for dissemination, or has already been functionally eradicated by systemic therapy. In such patients, the 3.1% absolute benefit of PMRT seen in B-51 may evaporate entirely. Conversely, the benefit of PMRT in B-51 is likely concentrated among the minority with persistent molecular residual disease—those whose ctDNA remains detectable despite nodal sterilization. This is more than a statistical re‑framing. Regional nodal irradiation addresses occult microscopic disease in the chest wall and draining nodal basins. The EBCTCG meta‑analysis of regional node radiotherapy showed that RNI significantly reduces distant recurrence and breast cancer mortality (RR 0.87; 95% CI, 0.80–0.94) in modern trials5. If a patient has cleared ctDNA from the systemic circulation at a sensitivity below 100 parts per million, the probability that viable locoregional reservoirs capable of seeding distant metastases remain is arguably low. For those with persistent ctDNA shedding, however, the locoregional sterilizing effect of PMRT may translate into meaningful gains—not just in local control, but in distant disease‑free survival. Limitations and future directions Our analysis is framed by the current evidence, which has inherent limitations. Most notably, the B-51 trial has a median follow-up of only 5 years. Longer-term data are particularly needed for hormone receptor-positive breast cancer, where late recurrence patterns might alter the risk-benefit assessment of PMRT omission, even in patients deemed low-risk by nodal pathology and ctDNA status. Furthermore, the role of non-pathologic complete response in the breast, isolated tumor cells, and differing tumor subtypes may be better delineated with extended follow-up. This hypothesis is testable. Archived plasma samples from the B-51 mastectomy cohort, if available, could be analyzed retrospectively to determine whether the 3.1% absolute benefit of PMRT differs between ctDNA‑positive and ctDNA‑negative subgroups. Such an analysis would require no new trial and could yield practice‑informing evidence relatively quickly. Beyond B-51, future de‑escalation trials should incorporate ctDNA as an integral biomarker. A non‑inferiority trial randomizing ctDNA‑negative mastectomy patients to PMRT versus observation would directly address whether PMRT can be safely omitted in this molecularly favorable subgroup. Conversely, escalation trials could be more efficiently powered by enrolling only ctDNA‑positive patients - those at highest risk of recurrence. The B-51 trial established that nodal pathologic complete response is a necessary but insufficient criterion for omitting PMRT in mastectomy patients. PREDICT‑DNA demonstrates that ctDNA provides an orthogonal layer of risk stratification that transcends histopathology. Taken together, these studies point toward a framework in which locoregional therapy decisions are guided not solely by nodal pathology but by a dynamic assessment of molecular residual disease. The question is no longer whether to de‑escalate PMRT. The question is in whom. ctDNA may finally give us the answer.