Personalized whole-genome–based ctDNA dynamics during neoadjuvant therapy across breast cancer subtypes: results from MONITOR-Breast Authors: Julia Foldi, Gregory Hogan, Katherine Johansen Taber, Matthew LaBella, Britney Sadler, Ravi Patel, Marija Balic, Derrick Haslem, Marcela Mazo Canola, Judy Peih-Ying Tsai, Therese Cusick, Niyati Nathwani, Yoichi Naito, Chikako Funasaka, Karlton Wong, Alexander Ham, Ashley Acevedo, CJ Battey, Genevieve M Gould, Dale Muzzey, Jeff Jasper Published: 2026 ABSTRACT ABSTRACTPurposePathological complete response (pCR) after neoadjuvant therapy (NAT) strongly associates with reduced breast cancer relapse risk. Circulating tumor DNA (ctDNA) shows promise as a therapy-response biomarker, but prior studies had limited sampling and assay sensitivity. MONITOR-Breast characterized ctDNA dynamics across NAT at high temporal resolution using an ultrasensitive molecular residual disease (MRD) assay.Experimental designIn this prospective observational study, 154 enrolled patients with breast cancer (all subtypes, Stages I-III) were tested at baseline, throughout NAT, and post-surgery using a whole genome sequencing-based MRD assay tracking up to 1,000 variants per patient.ResultsBaseline ctDNA was detected in 93% of patients, with 20% detected in the ultrasensitive range (<100 parts per million). Post-NAT ctDNA positivity strongly associated with residual disease (RD) (odds ratio (OR)>20, p = 1 × 10−7) and post-operative positivity (OR =47, p = 1 × 10−6). All 56 patients with pCR and an evaluable sample had undetectable post-NAT ctDNA. Frequent testing revealed distinct patterns: 78% achieved rapid and/or sustained ctDNA clearance, while 22% exhibited persistent or intermittent positivity, and had significantly higher RD (OR =52, p = 3 × 10−9), and post-operative positivity (OR =25, p = 1 × 10−5).ConclusionsFrequent, ultrasensitive ctDNA assessment provided comprehensive characterization of treatment response, revealing opportunities for de-escalation in early responders and escalation in those with persistent ctDNA.KEYWORDS: ctDNA, breast cancer, molecular residual disease, response monitoring, neoadjuvant therapy Introduction Approximately 25% of patients with early breast cancer in the US undergo neoadjuvant therapy (NAT), most commonly those with triple-negative breast cancer (TNBC), HER2-positive disease (HER2+), or locally advanced disease of any subtype [1–4]. NAT may downstage disease to facilitate surgery and control occult micro-metastatic disease. Pathological response to NAT also provides prognostic information and helps guide adjuvant systemic therapy [3,4]. Patients who undergo NAT and achieve pathological complete response (pCR)—defined as the absence of invasive disease in the breast and lymph nodes—experience a 69% reduction in disease relapse and 78% reduction in mortality compared to those with residual disease (RD) [5,6]. Among patients with RD, residual cancer burden (RCB) provides further prognostic refinement: RCB-I is associated with favorable outcomes, whereas RCB-II and RCB-III are associated with substantially higher recurrence and mortality [7–9]. Both pathological response and RCB are recognized as clinically meaningful endpoints for neoadjuvant trials [5,10]. NAT regimens typically consist of several cycles of multi-agent chemotherapy, often combined with immune checkpoint inhibitors and/or targeted agents depending on molecular subtype [3]. However, regimens vary substantially across institutions, even within each subtype, and new agents are continually being evaluated [11]. This dynamic landscape underscores the need for biomarkers to assess real-time response. Cell-free circulating tumor DNA (ctDNA) is a promising biomarker providing molecular assessment of tumor burden [12–23]. Accumulating evidence demonstrates that ctDNA clearance during and after NAT is associated with pCR and recurrence-free survival (RFS) independent of pathological response [12–17,19,22]. In the neoadjuvant setting, ctDNA clearance at or before the midpoint of NAT predicts pCR, whereas ctDNA persistence identifies patients unlikely to achieve pCR and at elevated risk of metastatic recurrence [13,15–17,19,22,24]. However, prior studies sampled only a few timepoints—typically baseline, midpoint, and end of NAT—and relied on assays with limited sensitivity and low baseline detection rates. A comprehensive characterization of ctDNA dynamics throughout NAT necessitates more frequent sampling with ultrasensitive detection, defined as detection at <100 parts per million (PPM), to determine whether patterns such as sustained clearance or low-level persistent positivity are clinically meaningful. The MONITOR-Breast study is an ongoing multi-center, prospective study employing a tumor-informed, whole-genome sequencing (WGS)-based, ultrasensitive molecular residual disease (MRD) assay to quantify ctDNA levels at frequent sampling intervals across the breast cancer care continuum. Here, we report results from an exploratory analysis of MONITOR-Breast, yielding quantitative ctDNA dynamics at high temporal resolution throughout the NAT period. We evaluated the association of ctDNA clearance with pCR, the clearance patterns captured by frequent sampling, and the performance of the ultrasensitive assay relative to earlier-generation assays. Patients and methods Patients and sample collection Patients were enrolled in the Institutional Review Board (IRB)-approved, prospective, observational MONITOR-Breast study from 10 clinical sites between May 2023 and October 2025 (Central IRB: Advarra, Pro00070267; local IRB: Intermountain Health, 1052007; local IRB: Cleveland Clinic, 24–104). All patients provided written informed consent. Patients were eligible if they had a newly diagnosed cT1a-4 invasive breast cancer, were planning to undergo or previously had undergone surgical treatment for curative intent, had known estrogen receptor (ER), progesterone receptor (PR), and HER2 status determined by immunohistochemistry and/or fluorescence in situ hybridization, and were eligible for chemotherapy. Patients were ineligible if they had metastatic disease, currently active additional cancer diagnosis (except non-melanoma skin cancer or second breast cancer), or previous allogeneic transplant. Patients included in the analyses described herein were enrolled at the time of diagnosis and underwent NAT. Whole blood for ctDNA analysis was collected in Streck Cell-Free DNA BCT® tubes at the following timepoints: baseline (at diagnosis and prior to any systemic treatment), prior to each NAT cycle (approximately every 2–3 weeks depending on regimen and site discretion), following the completion of NAT but prior to surgery, at surgery, and at the post-operative visit (Figure S1A). The post-NAT timepoint was defined as the last plasma sample at the end of treatment or, if unavailable, the last sample collected within 6 weeks prior to surgery. To assess potential variability in ctDNA levels related to treatment-to-surgery intervals, we evaluated the relationship between post-NAT ctDNA status and pathological response with the following time intervals: latest NAT cycle to surgery, latest NAT cycle to post-NAT plasma collection, and post-NAT plasma collection to surgery. No significant associations were seen (all p > 0.05, Wilcoxon test). Three or more plasma timepoints during the NAT period, including one at either baseline or surgery, were required for inclusion in analyses. Tumor specimens were collected at the time of initial diagnostic biopsy prior to treatment initiation. Core needle biopsies were formalin-fixed and paraffin-embedded (FFPE) according to standard clinical protocols. Clinical data included demographics, past medical history, initial diagnosis date, site of diagnosis, American Joint Committee on Cancer (AJCC) clinical stage, pathological stage and grade, complete treatment history, blood and tissue collection dates, dates of clinic visits and surgery, and pathological response. pCR was defined as the absence of invasive disease in the breast and lymph nodes (ypT0/Tis, ypN0). RCB was calculated using standardized methodology as described by Symmans et al. [7]. MRD assay and sample analysis ctDNA results were generated using Precise MRD (Myriad Genetics, Inc, Salt Lake City, UT), an ultrasensitive, tumor-informed, WGS-based MRD assay (Figure S1B) [25]. Tumor and normal WGS was performed to a targeted depth of 30x on libraries with 100ng DNA input (min 10ng) extracted from both the core needle biopsy from the primary tumor and the buffy coat/normal sample collected at the time of enrollment. For patients with multiple primary tumors (multicentric or multifocal disease), a single representative tumor specimen was selected by the pathologist for WGS testing to create a single patient-specific panel. An ensemble somatic variant calling approach identified tumor-specific targets, and a ranking algorithm was used to select an optimal patient-specific panel. Individualized panels were composed of up to 1,000 biotinylated hybridization capture probes targeting patient-specific single-nucleotide variants and insertions/deletions. Cell-free ctDNA was isolated from plasma and subjected to ultra-deep, next-generation targeted sequencing, including error correction, based on the individualized panels. Tumor fractions were estimated by maximum likelihood using a mixture model designed to account for sample- and mutation-type-specific error rates while marginalizing out interference from germline variants and allele frequency outliers. MRD positive status was assessed by calculating a likelihood ratio comparing the maximum likelihood tumor fraction to a tumor fraction of 0, with the MRD positive threshold determined by analyzing a large number of healthy donor samples. Test reports include both a binary “ctDNA detected” or “ctDNA not detected” result and, in the case of a “ctDNA detected” result, a quantitative tumor fraction reported in PPM. Statistical analysis The association of ctDNA levels with NAT response was an exploratory objective of the ongoing MONITOR-Breast study. As such, no formal sample size or power calculation was performed. All eligible patients meeting inclusion criteria during the study period and satisfying requirements for ctDNA analysis were included. Descriptive statistics were used to summarize patient demographics, clinical characteristics, and ctDNA measures. Continuous variables were summarized using medians and ranges, and categorical variables were summarized using counts and percentages. Molecular response, as presented in Section 3.6, was categorized descriptively as follows: complete response, defined as ctDNA clearance; partial response, defined as a >10-fold decline in ctDNA levels during therapy; and little to no response, defined as <10-fold decline in ctDNA levels during therapy. Baseline ctDNA detection status, baseline ctDNA tumor fraction, and sustained or persistent/intermittent ctDNA detection patterns were evaluated for associations with pathological response using Fisher’s exact test. Timing of ctDNA clearance and association with pCR was evaluated using Pearson’s chi-squared test. Differences in ctDNA tumor fraction across ordered clinical categories (clinical stage and T stage) were assessed using the Jonckheere–Terpstra trend test, and comparisons between two groups were performed using the Wilcoxon rank-sum test. To assess the independent predictive value of post-NAT ctDNA tumor fraction while accounting for all clinical-pathological factors simultaneously, we fit a single Firth penalized logistic regression model including log10-transformed ctDNA tumor fraction, clinical stage, T stage, nodal status, and molecular subtype as covariates. Odds ratios (OR) and 95% profile likelihood confidence intervals (CI) were calculated. All statistical tests were two-sided. A nominal p-value < 0.05 was used as a threshold for statistical significance. Analyses were exploratory and were not adjusted for multiple comparisons. Analyses were performed using R (version 4.4.2) with package PMCMRplus (version 1.9.12) for Jonckheere-Terpstra test, logistf (version 1.26.1) for Firth regression analysis, and ggalluvial (version 0.12.5) for the Sankey plot. Results Study cohort Among patients meeting study eligibility criteria, 154 were enrolled at diagnosis and prior to NAT. Time on study ranged from 3 to 37 months (median = 20). Panels were successfully created for 140 patients (91%); 14 had insufficient tissue available. Panels comprised 566–1,000 probes (median = 1,000). Plasma samples from at least three timepoints were received for 125 patients with panels, resulting in 125 patients included in the study cohort (Figure S1C). A total of 949 plasma samples were analyzed across the cohort, ranging from 3 to 12 timepoints per patient (median = 9). Cohort characteristics (Table 1) included all four subtypes and clinical Stages I-III. Histology was mostly invasive ductal carcinoma (n = 117, 94%). All patients received NAT as directed by their treating clinician, including neoadjuvant chemotherapy (100%), HER2-targeted therapy (37%) and immunotherapy (42%). Pathological response data were available for 115 patients (92%); the remaining 10 had not yet undergone surgery or were awaiting pathological results. RCB data were available for 101 patients (81%) (Table 1). Table 1.Study cohort characteristics.CharacteristicAll(n = 125, 100%)TNBC(n = 49, 39%)HR+/ HER2-(n = 28, 22%)HR+/ HER2+(n = 33, 26%)HR-/ HER2+(n = 15, 12%)Median age at enrollment, years [range]55 [31–80]53 [31–74]54 [35–69]51 [36–80]58 [39–75]Female sex, n (%)125 (100)49 (100)28 (100)33 (100)15 (100)AJCC Clinical Stage, n (%) I15 (12)7 (14)2 (7)5 (15)1 (7) II91 (73)35 (71)20 (71)25 (76)11 (73) III19 (15)7 (14)6 (21)3 (9)3 (20)T Stage T122 (18)8 (16)4 (14)6 (18)4 (27) T279 (63)30 (61)18 (64)22 (67)9 (60) T316 (13)6 (12)3 (11)5 (15)2 (13) T48 (6)5 (10)3 (11)0 (0)0 (0)Nodal status, n (%) N067 (54)32 (65)10 (36)21 (64)4 (27) N+56 (45)16 (33)18 (64)12 (36)10 (67) NX2 (2)1 (2)0 (0)0 (0)1 (7)Histology, n (%) Invasive ductal carcinoma117 (94)46 (94)26 (93)31 (94)14 (93) Invasive lobular carcinoma6 (5)2 (4)2 (7)2 (6)0 (0) Mucinous carcinoma1 (1)0 (0)0 (0)0 (0)1(7) Spindle cell carcinoma1 (1)1 (2)0 (0)0 (0)0 (0)Tumor type, n (%) Single primary118 (94)48 (98)26 (93)31 (94)13 (87) Multiple primary7 (6)1 (2)2 (7)2 (6)2 (13)Neoadjuvant treatment type, n (%)c Chemotherapy125 (100)49 (100)28 (100)33 (100)15 (100) Immunotherapy52 (42)48 (98)4 (14)0 (0)0 (0) HER2-targeted therapy46 (37)0 (0)0 (0)31 (94)15 (100)Pathological response (n = 115), n (%)a Pathological complete response (pCR)59 (51)31 (66)6 (24)14 (48)8 (57) Residual disease (RD)56 (49)16 (34)19 (76)15 (52)6 (43)Residual cancer burden (RCB) (n = 101), n (%)b RCB-060 (59)31 (74)7 (30)14 (64)8 (57) RCB-I13 (13)3 (7)2 (9)5 (23)3 (21) RCB-II18 (18)7 (17)6 (26)2 (9)3 (21) RCB-III10 (10)1 (2)8 (35)1 (5)0 (0)aPathological response data were available for 47 TNBC, 25 HR+/ HER2-, 29 HR+/ HER2+, and 14 HR-/ HER2-.bRCB data were available for 42 TNBC, 23 HR+/ HER2-, 22 HR+/ HER2+, and 14 HR-/ HER2+.cNeoadjuvant regimens were individualized based on clinician discretion considering molecular subtype, clinical risk factors, and patient preference.Abbreviations: AJCC, American Joint Committee on Cancer; HR, hormone receptor; TNBC, triple-negative breast cancer. Study cohort characteristics. aPathological response data were available for 47 TNBC, 25 HR+/ HER2-, 29 HR+/ HER2+, and 14 HR-/ HER2-. bRCB data were available for 42 TNBC, 23 HR+/ HER2-, 22 HR+/ HER2+, and 14 HR-/ HER2+. cNeoadjuvant regimens were individualized based on clinician discretion considering molecular subtype, clinical risk factors, and patient preference. Abbreviations: AJCC, American Joint Committee on Cancer; HR, hormone receptor; TNBC, triple-negative breast cancer. Baseline ctDNA detection A baseline plasma timepoint, collected prior to initiation of NAT, was received for 121 patients, and ctDNA was detected in 112 (93%). Baseline detection rates were 100% (28/28) in patients with HR+/HER2- disease, 98% (45/46) with TNBC, 93% (14/15), with HR-/HER2+ disease, and 78% (25/32) with HR+/HER2+ disease (Figure 1, Table S1). Baseline detection rates increased by clinical Stage (I: 80%, II: 93%, III: 100%), T stage (T1: 82%, T2: 93%, T3: 100%, T4: 100%), and nodal status (N0: 89%, N+: 98%) (Figure 1, Table S1). Boxplots showing baseline ctDNA detection in the full cohort and subdivided by Stage, clinical T stage, nodal status, and subtype. Each sample is denoted by a dot located at the tumor fraction detected, with horizontal jitter for visibility. Medians are indicated by a horizontal line. Dots in the light green section at the bottom were not detected at baseline, while those in the gray section were detected in the ultrasensitive range (<100PPM). Baseline tumor fraction ranged from 2.3 to 610,011 PPM (median = 1,291) (Figure 1, Table S1); 88 patients (73%) had tumor fractions of >100 PPM, while 24 (20%) were in the ultrasensitive range (<100 PPM). Among the 24 patients with ultrasensitive detection, 14 (58%) had tumor fractions of 20–100 PPM and 10 (42%) were <20 PPM (Figure 1). Median baseline tumor fraction increased significantly across clinical stage (Jonckheere-Terpstra p = 0.0003) and T stage (Jonckheere- Terpstra p = 0.002), and was significantly higher in node-positive versus node-negative patients (Wilcoxon p = 0.0003) (Table S1). Patients with HR+/HER2+ disease had lower baseline tumor fractions. Pathological response data were available for 115 patients with baseline plasma results (Table 1). No significant association was observed between baseline ctDNA detection and pCR (p = 1.00, Fisher’s exact), nor between baseline tumor fraction category (ultrasensitive versus non-ultrasensitive) and pCR (p = 0.47, Fisher’s exact). Post-NAT and post-operative ctDNA results were strongly associated with pathological response Pathological response was significantly associated with ctDNA negativity at the post-NAT timepoint (OR >20, p = 1 × 10−7, Fisher’s exact) (Figure 2(A-C)). Among the 102 patients with post-NAT ctDNA results and pathological response data, all 56 who achieved pCR were ctDNA- at the post-NAT timepoint (100%, 95% CI 94%-100%). In contrast, among 46 patients with RD, 16 (35%, 95% CI 21%-49%) were ctDNA+ at the post-NAT timepoint. Notably, 11 patients (69%) had ctDNA detected in the ultrasensitive range (<100 PPM). Post-NAT ctDNA detection closely tracked with RCB. All RCB-0 patients (n = 57) were ctDNA-, whereas ctDNA positivity was found exclusively among RCB II (6 of 12) and III (4 of 9) patients. (A). Tumor fraction detected at the indicated timepoints, stratified by those with residual disease (RD) (left) and those who achieved pCR (right). Gray lines connect a patient’s samples if one sample is ctDNA positive after baseline. (B). pCR rate by subtype. (C). Across all subtypes, pathological response was significantly associated with ctDNA status at the post-NAT timepoint. (D). Examples of three distinct patterns of ctDNA results across neoadjuvant timepoints: sustained ctDNA clearance (examples 1 and 2), persistent ctDNA positivity (example 3), and intermittent ctDNA positivity at low tumor fractions (example 4). Proportions of each subtype corresponding to each pattern are shown in the circle chart (blue: sustained clearance, red: persistent positivity, yellow: intermittent positivity). Percent of each subtype in the sustained clearance pattern is indicated inside the circle. (E). ctDNA clearance over time by subtype. Abbreviations: ctDNA, circulating tumor DNA; HR, hormone receptor; MRD, molecular residual disease; NAT, neoadjuvant therapy; pCR, pathological complete response; RD, residual disease; TNBC, triple-negative breast cancer. Post-NAT ctDNA detection in patients with RD varied by clinical characteristics. The proportion of patients with RD who were ctDNA+ at the post-NAT timepoint was 56% in Stage III, 30% in Stage II, and 25% in Stage I. Across subtypes, the proportion was highest in TNBC (7 of 14, 50%), intermediate in HR+/ HER2- (6 of 14, 43%) and lower in HR−/HER2+ (1 of 5, 20%) and HR+/HER2+ disease (2 of 13, 15%) (Figures S2,S3). Post-NAT ctDNA positivity rate was higher in patients with node-positive disease at diagnosis compared to node-negative (OR =4.2, p = 0.015, Fisher’s exact); further, post-NAT ctDNA positivity was associated with 27x higher odds of nodal disease on surgical pathology (p = 2.9 × 10−7, Fisher’s exact). In a multivariable model including post-NAT ctDNA tumor fraction, clinical stage, T stage, nodal status, and subtype, ctDNA tumor fraction remained the strongest associated factor with pCR (OR 0.34, 95% CI 0.10–0.61, p = 6.50 × 10−7). Among subtypes, only TNBC retained independent predictive value. Post-operative plasma samples were available for 100 patients, among which 11 (11%) were ctDNA+; 10 of these 11 (91%) had RD (Figure 2(A)). Among the 94 patients with both post-NAT and post-operative samples, those who were ctDNA+ at the post-NAT timepoint were 47x more likely to be ctDNA+ after surgery (OR =47, p = 1.1 × 10−6, Fisher’s exact). Among the 9 patients with RD who were ctDNA+ post-operatively and had post-NAT samples available, 8 (89%) were also ctDNA+ at the post-NAT timepoint, suggesting ctDNA positivity at the end of NAT is associated with risk of post-operative ctDNA detection, potentially indicative of micro-metastatic disease. Longitudinal ctDNA dynamics and clearance patterns Three distinct ctDNA clearance patterns were observed across the frequent sampling of MRD testing: sustained ctDNA clearance, persistent ctDNA positivity, and intermittent ctDNA positivity (Figure 2 (D)). More than three-quarters (78%, n = 97) of patients experienced sustained clearance, defined as a ctDNA-negative result at or following baseline with ctDNA negativity at every timepoint thereafter (Figure 3(A)). Another 11% (n = 14) of patients experienced persistent positivity. Most of these patients had a decline in ctDNA levels during NAT, often reaching ultrasensitive levels, but never achieved clearance (Figure 3(B)). The remaining 11% (n = 14) had clearance at some point after baseline, followed by intermittent positivity with tumor fraction <20 PPM (Figure 3(B)). (A). Tumor fraction and timing of ctDNA clearance among patients who had sustained clearance during the neoadjuvant time period. Post-operative ctDNA results in this group are shown at the right. (B). Tumor fraction across the neoadjuvant period among those with persistently positive ctDNA results (orange) and those with intermittently positive ctDNA results (yellow). Post-operative ctDNA results in both groups are shown at the right. (C-E). Swimmer plots indicating ctDNA results for each patient at each timepoint, arranged by subtype and clearance pattern ((C): sustained clearance, (D): persistently positive, and (E): intermittently positive). The color of each dot indicates tumor fraction, while colored lines indicate therapy. Boxes indicate the timing of surgery and their color indicates pathological response (blue: pCR, orange: RD). Abbreviations: ctDNA, circulating tumor DNA; HR, hormone receptor; MRD, molecular residual disease; TNBC, triple-negative breast cancer. The proportion of patients with sustained clearance varied by subtype, ranging from 68% in HR+/ HER2- disease to 91% in HR+/ HER2+ disease (Figure 2(D,E)). Timing of clearance also varied by subtype (Figure 2(E)). Among patients with TNBC, 90% of those who ultimately cleared did so by day 45 (6 weeks, 3 days), whereas those with HR+/ HER2- disease showed the slowest clearance, requiring 16 weeks to capture 90% of patients who cleared. Even after clearance rates plateaued, incremental clearance continued up to the time of surgery (Figure 2(E), Figure S1). The difference in clearance kinetics across subtypes suggests that frequent ctDNA assessment is important for identifying patients who ultimately clear ctDNA, particularly for those who clear later in NAT, as well as those who remain at risk for RD. ctDNA dynamics are associated with pathological response In the set of 97 patients with sustained clearance, pathological response data were available for 87 (90%). pCR was observed in 67% (n = 58), significantly higher than the 3.5% (1 of 28) of patients achieving pCR who had persistent or intermittent positivity (OR =52, p = 1.1 × 10−9, Fisher’s exact) (Figure 3(A,C)). Among those with RCB data, 72% with sustained clearance had RCB-0 (Figure S4). Timing of clearance was not significantly associated with pCR (p = 0.57, Pearson chi-squared), suggesting that sustained ctDNA clearance, rather than the time of onset of clearance, is the critical determinant of pCR. Frequent ctDNA assessment may therefore identify additional patients who achieve clearance later in their neoadjuvant course. Patients with sustained clearance continued NAT for an average of 119 days (interquartile range (IQR) 97–155) with undetectable ctDNA. Pathological response data were available for all 14 patients with persistently positive ctDNA results; all 14 (100%) had RD, significantly higher than the proportion of patients with sustained ctDNA clearance who had RD (OR >25, p = 1.5 × 10−6, Fisher’s exact) (Figure 3(B,D)). Pathologic response data were available for all 14 (100%) of the patients with intermittent ctDNA positivity; 13 of 14 (93%) had RD, significantly higher than the proportion of patients with sustained ctDNA clearance who had RD (OR =25, p = 3.0 × 10−5, Fisher’s exact) (Figure 3(B,E)). Overall, 96% of patients with persistent or intermittent ctDNA positivity had RD (Figure 3(D,E)). The rate of RD was not significantly different in those with intermittent ctDNA positivity compared to those with persistent ctDNA positivity (p = 1.0, Fisher’s exact), suggesting that ctDNA positivity, even at levels < 20 PPM, identifies those at risk for RD. Among the patients with RD that were ctDNA- at the post-NAT sample, 7 patients had intermittent ctDNA positivity; thus, testing at each NAT cycle identified 44% more (n = 23) patients at risk of RD compared to a single post-NAT assessment (n = 16 ctDNA+). Recognizing that it may be clinically infeasible to sample at every NAT cycle, we evaluated whether testing at the following clinically meaningful timepoints could provide similar utility: mid-point of the initial regimen, regimen transition (in patients receiving multi-component regimens, such as adriamycin/cyclophosphamide followed by paclitaxel), the mid-point of the subsequent regimen (when applicable), and post-NAT. Testing at those timepoints identified 25% more (n = 20) patients at risk of RD compared with a single post-NAT assessment (n = 16 ctDNA+). Among patients with RCB data, all with persistent positivity and most with intermittent positivity had RCB-II or RCB-III (Figure S4), suggesting that both groups have a higher risk of worse outcomes compared to patients with sustained clearance. Further, patients with persistent or intermittent positivity were at higher risk of being ctDNA+ after surgery; 7 of 11 patients with persistent positivity and 2 of 11 patients with intermittent positivity were also positive at the post-operative visit, representing 25x higher odds compared to patients who achieved sustained clearance (p = 1.1 × 10−5). These results suggest it may be beneficial to measure ctDNA levels across multiple timepoints during neoadjuvant treatment to identify those most likely to experience RD at time of surgery and post-operatively. Response to NAT as measured by MRD has the potential to shape axillary staging and surgical decisions. Persistent and intermittent ctDNA positivity was strongly associated with nodal disease at surgery (OR =13, p = 2.9 × 10−7, Fisher’s exact). 52% (14/27) of patients with nodal disease received axillary lymph node dissection (ALND), and factoring in ctDNA status may better stratify patients who would benefit from ALND. Among patients who did not receive ALND, patients with persistent or intermittent positivity were 9x more likely to have nodal disease (p = 0.0024), and these patients were 29x more likely to have post-operative ctDNA positivity (p = 0.0017, Fisher’s exact). Further, among patients who did receive ALND, patients with sustained ctDNA negativity were 7x less likely to have nodal disease. Tracking treatment response with high-resolution ctDNA monitoring To illustrate the utility of high temporal resolution ctDNA monitoring during treatment regimens, 12 cases representing four response patterns were selected to show ctDNA clearance kinetics around treatment switch points (Figure 4). The A (Figure 4(A)) and B (Figure 4(B)) response patterns demonstrate the effectiveness of multiple-therapy regimens in ctDNA clearance. Patients in response pattern A showed a more than 10-fold decline in ctDNA levels during TX1, but multiple ctDNA+ timepoints suggest the treatment was not on a trajectory toward clearance; initiation of TX2 changed this trajectory and led to ctDNA clearance that was maintained through TX2 and surgery. Patients in response pattern B highlight cases in which surgery was needed to achieve ctDNA negativity; these patients had partial responses to their respective TX1 and TX2, often reaching ctDNA levels <100 PPM by weeks 6–12, but remained ctDNA+ at the post-NAT timepoint. These patients achieved molecular clearance following surgery. Patterns of molecular response exemplified by selected patients in the cohort. Therapy (TX) 1 and TX2 denote sequential components of multi-agent NAT regimens (e.g., carboplatin/taxane followed by anthracycline-based therapy), with treatment switch points defined as transitions between regimen phases. Molecular response to each regimen (TX1, TX2, and surgery) is shown as a colored bar. Green: complete response (ctDNA clearance), yellow: partial response (more than 10-fold decline in ctDNA level during therapy), red: little to no response. Therapy switches and surgery are marked by vertical dashed lines. (A). Pattern of decline in ctDNA levels during TX1 with further declines and clearance after initiation of TX2. (B). Pattern of partial responses to TX1 and TX2 but only clearing after surgery. (C). Pattern of partial response to TX1, but lack of response to TX2 and no clearance after surgery. (D). Pattern of partial response with only one regimen, and no clearance after surgery. Abbreviations: PPM, parts per million. In contrast, patients in response patterns C (Figure 4(C)) and D (Figure 4(D)) did not appear to benefit from a change in treatment and/or surgery. Patients in response pattern C showed partial molecular responses to TX1, with tumor fraction declining to <100 PPM, but subsequent TX2 could not clear residual ctDNA, which was persistently positive in the final weeks prior to surgery and post-operatively. Patients in response pattern D had only one treatment prior to surgery, showing partial response but never clearing, and remaining ctDNA+ post-operatively. Two cases in pattern C (the two left-most panels) and all 3 cases in pattern D are at an elevated risk of recurrence based on persistent post-operative ctDNA positivity. Together, these response patterns suggest that frequent, ultrasensitive ctDNA monitoring can track distinct molecular trajectories throughout NAT, enabling detection of both emerging clearance and early indicators of RD, as well as revealing opportunities to change treatment. These patterns should be interpreted in the context of treatment sequencing, as transitions between regimen components (e.g., TX1 to TX2) may influence ctDNA dynamics; however, this also underscores the ability of ctDNA to reflect response to distinct therapies in real time. Discussion This study is the first to characterize ctDNA dynamics using frequent, serial sampling throughout NAT in breast cancer. By assessing ctDNA at high temporal resolution, we found that sustained ctDNA clearance is a strong indicator of pCR, and persistent or intermittent low‑level positivity is a marker of RD. These findings demonstrate that granular, real‑time ctDNA monitoring can reveal clinically meaningful response patterns, including dynamic clearance and intermittent low-level positivity, that may not be fully captured by single or infrequent measurements alone. The MRD assay used in this study employs deep WGS to identify up to 1,000 high-quality somatic variants to form the basis of individualized tracking panels. Tracking this number of variants enables high sensitivity and confident detection of ctDNA at levels <100 PPM. In contrast, most prior analyses on ctDNA levels in the neoadjuvant period have used earlier-generation tumor-informed MRD assays that track far fewer variants and consequently have limited performance [13,14,22]. With the ultrasensitive assay, baseline detection across the whole cohort was 93%, increasing from 80% at Stage I to 93% and 100% in those with Stage II and III disease, respectively. Earlier generation tests reported lower overall baseline detection rates of 73–77% [13,22]. This difference is more pronounced in those with HR+/HER2- disease; earlier-generation assays detected ctDNA in only 48%-70% of patients, whereas baseline detection reached 100% in this study [13,22]. Notably, 20% of ctDNA+ samples at baseline and 69% at post-NAT had tumor fractions below 100 PPM, underscoring the clinical necessity of an ultrasensitive assay. Post-NAT ctDNA positivity in patients with RD was 35% overall in our study, substantially higher than the 14%-20% reported for earlier-generation assays and demonstrating the ability of the ultrasensitive assay to better identify those who did not have a complete response to NAT [13,22]. Post-NAT negative predictive value (NPV) in our study was 65% (95% CI 55%-75%), twice that of earlier-generation tests (31–32%) [13,22]. NPV is clinically important because a false-negative could lead to the mistaken conclusion that a patient has responded adequately, when additional treatment may still be required. The utility of ctDNA assessments during NAT to monitor therapy response and predict long-term outcomes has been demonstrated in previous studies. ctDNA clearance early after NAT initiation, at the midpoint, and at the end of NAT predicts pCR and is associated with improved distant recurrence-free survival (DRFS) [13,22–24,26]. Conversely, ctDNA persistence, even as early as the midpoint of NAT, is a significant predictor of relapse and metastatic recurrence [13]. Importantly, ctDNA status has been shown to provide prognostic information independent of pathological response: patients who clear ctDNA post-NAT despite RCB-II or III pathologies have outcomes comparable to those achieving pCR, whereas ctDNA+ patients with RD experience significantly worse outcomes (hazard ratio (HR) 5.89 in HR+/HER2- disease, HR 3.79 in TNBC) [14]. Although our study lacks long-term outcomes, our pathological response results are consistent: two-thirds of patients with sustained ctDNA clearance achieved pCR, whereas 96% of patients with persistent or intermittent ctDNA positivity – even at levels <20 PPM – had RD. Post-NAT ctDNA positivity was strongly predictive of RD, yielding a positive predictive value (PPV) of 100% (95% CI 81%-100%), consistent with prior reports [13,22,26]. Also consistent with prior studies, we observed significant differences in ctDNA clearance dynamics and pCR rates by breast cancer subtype, likely reflecting both the inherent biology of these subtypes and their differential sensitivity to standard neoadjuvant regimens [13,22]. The prognostic value of pCR is somewhat limited in HR+ disease compared with TNBC and HER2+ subtypes, underscoring the need for complementary biomarkers that more directly reflect RD burden and recurrence risk. These findings add novel information to prior studies that assessed ctDNA levels at 3–4 timepoints over intervals of several months. By sampling at every NAT cycle, this study enabled a high temporal resolution analysis of ctDNA dynamics throughout the NAT period. We showed that sustained clearance, i.e., persistent negativity prior to surgery, appears to be a critical determinant of pCR across subtypes, underscoring the importance of frequent ctDNA testing to capture clearance dynamics and avoid misclassification based on a single negative result. Frequent sampling identified a subset of patients who appeared to clear ctDNA after 2–3 cycles, but whose ctDNA levels then vacillated between negative and positive at <20 PPM during the latter half of NAT. All but one of these patients had RD, suggesting that even low levels of ctDNA positivity during the second half of NAT may predict poor pathological outcomes. Important considerations regarding optimal sampling frequency remain, as less intensive approaches (e.g., baseline and preoperative timepoints) may miss dynamic patterns, while more frequent sampling enables identification of these patterns and may be further optimized for clinical implementation. These observations, together with the heterogeneity in individual treatment response patterns revealed by high-resolution monitoring, indicate that frequent, ultrasensitive ctDNA profiling can refine risk stratification during NAT, and may offer the opportunity to change treatment. Patients demonstrating rapid and sustained clearance may be candidates for treatment de-escalation, proceeding to early surgery to avoid unnecessary chemotherapy toxicity. In other settings, such as rectal and esophageal cancer, patients with complete clinical response following NAT are increasingly being managed non-operatively or with minimally invasive procedures [27,28]. Ongoing trials in breast cancer are investigating whether ctDNA could be used to identify those who may safely avoid surgical intervention [29]. Conversely, patients with persistent or intermittent low-level positivity may benefit from treatment intensification or altered therapy. Prospective randomized trials are needed to determine whether ctDNA-guided decisions improve outcomes beyond standard approaches. Further, additional effort is needed to evaluate the feasibility and cost of incorporating frequent ctDNA assessment during NAT into clinical workflows, including ensuring that turnaround times are sufficient to support timely treatment decisions. Notably, in our cohort, a set of four clinically meaningful timepoints after baseline identified more patients at risk of residual disease than a single post-NAT assessment, opening actionable opportunities to guide NAT decision-making and inform response-adapted surgical planning. This approach may offer a pragmatic compromise between clinical utility and the cost and logistical demands of frequent sampling. Strengths and limitations Strengths of our study include its prospective design, dense ctDNA sampling timepoints that provide a detailed picture of NAT response, and the inclusion of all breast cancer subtypes, stages, and treatments, reflecting real-world practice. Our study also has limitations. First, long-term outcomes are not yet available. The cohort described here is part of the ongoing prospective MONITOR-Breast study, and analysis of ctDNA dynamics during the neoadjuvant period was an exploratory objective. Consequently, endpoints such as RFS and DRFS will remain blinded until the follow-up period is complete. However, the well-established association between pCR and substantially reduced risk of progression and death, particularly in TNBC and HER2+ disease, as well as its acceptance by the FDA as an endpoint in clinical trials, provides strong rationale for using pCR as an endpoint in this study [5,10,30]. Importantly, our results are consistent with prior studies that have incorporated long-term outcomes. Ongoing follow-up in MONITOR-Breast will enable evaluation of ctDNA dynamics during and after adjuvant therapy and their relationship to long-term outcomes. Second, the pragmatic design did not mandate a uniform NAT regimen. As a result, treatments were heterogeneous, even within subtypes, making it challenging to align sampling timepoints with specific treatment cycles. We therefore reported sampling timepoints by days from NAT initiation, and comparisons with studies using regimen-based timepoints were estimated as precisely as possible. Notably, our ctDNA dynamics results were consistent with findings from other studies, supporting the validity of our time-based approach. Treatment heterogeneity may also influence the interpretation of ctDNA clearance patterns, particularly in patients receiving multi-therapy regimens with distinct switch points, as transitions between therapies could affect ctDNA dynamics and contribute to apparent changes in trajectory independent of underlying tumor biology. Importantly, this heterogeneity reflects real-world clinical practice and highlights the ability of ctDNA to dynamically track response to distinct therapeutic exposures. Third, while our total cohort included 125 patients across diverse subtypes, stages, and nodal involvement, subgroup analyses by these clinical characteristics were limited by sample size, and in some cases, may not have been sufficiently powered to detect differences in ctDNA clearance dynamics across subgroups. Finally, RCB data were available for 81% of patients; the higher proportion of RCB-0 relative to the overall pCR rate suggests that missing RCB data were more common among patients with RD, which may affect the generalizability of RCB-specific analyses. Conclusions Employing an ultrasensitive, tumor-informed MRD assay with frequent serial sampling, we comprehensively and quantitatively characterized ctDNA dynamics across the entire neoadjuvant period. Our results establish the clinical validity of ctDNA in monitoring treatment response, and support continued investigation of ctDNA-guided approaches in prospective clinical trials, with attention to subtype-specific clearance kinetics and the importance of frequent assessment. As ctDNA testing becomes increasingly accessible, integration into neoadjuvant management may enable more personalized treatment approaches, optimizing therapy intensity while minimizing toxicity for patients with breast cancer. Supplementary Material S1_Overview and Consort.jpg Supplemental tables_R1.docx S2_Sankey_RCB.jpg S3_pCR_bySubtype.jpg S4_barplots_RCB_vs_response groups.jpg STROBE_checklist_MONITOR_Breast_R1.pdf Acknowledgments The authors thank the patients who are enrolled in MONITOR-Breast and their clinicians for their participation in the study; members of the Myriad Genetics Clinical Operations, Assay Development, Computational Biology, Biostatistics, and Laboratory Operations teams for assistance in study design, study conduct, data collection and analysis, and sample processing. To ensure accessibility by a wide audience, the plain language summary was developed with assistance from Microsoft M365 Copilot (GPT-5 chat model; Version 19.2605.49101.0); all content was subsequently reviewed and edited by the authors for accuracy and clarity. Smaller subsets of this research was presented at the 2025 San Antonio Breast Cancer Symposium, December 10, 2025, San Antonio, Texas; and the 2026 Annual Meeting of the American Association for Cancer Research, April 20, 2026, San Diego, California. Funding Statement This study was funded by Myriad Genetics Inc., which had roles in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, and approval of the manuscript. Article highlights Frequent, ultrasensitive molecular residual disease (MRD) testing throughout neoadjuvant therapy (NAT) provided a comprehensive view of treatment response and residual disease. Frequent, ultrasensitive molecular residual disease (MRD) testing throughout neoadjuvant therapy (NAT) provided a comprehensive view of treatment response and residual disease. Circulating tumor DNA (ctDNA) status post-NAT was significantly associated with pathological response and post-operative ctDNA positivity. Circulating tumor DNA (ctDNA) status post-NAT was significantly associated with pathological response and post-operative ctDNA positivity. Most patients (78%) showed sustained clearance, often early in NAT, supporting opportunities for ctDNA to guide NAT de-escalation. Most patients (78%) showed sustained clearance, often early in NAT, supporting opportunities for ctDNA to guide NAT de-escalation. 22% of patients were persistently or intermittently ctDNA positive during NAT, and at higher risk of worse outcomes. 22% of patients were persistently or intermittently ctDNA positive during NAT, and at higher risk of worse outcomes. MRD testing during NAT identifies more patients at risk of residual disease (RD) and has potential to improve outcomes as a response measure. MRD testing during NAT identifies more patients at risk of residual disease (RD) and has potential to improve outcomes as a response measure. Disclosure statement Julia Foldi: Consulting fees from BioTheranostics Inc. (a Hologic Company), Daiichi Sankyo, Genentech, Gilead, AstraZeneca, Novartis, and Caris Life sciences. Marija Balic: Research grants from AstraZeneca, Daiichi Sankyo, Novartis, and Stemline; consulting for Gilead; speaker’s bureau member for Novartis, Stemline, AstraZeneca, and Daiichi Sankyo. Derrick Haslem: Consulting fees from Myriad Genetics. Marcela Mazo Canola: Research grants from American Cancer Society and UT San Antonio; consulting for Eli Lilly and AstraZeneca; speaker’s bureau member for Gilead, Stemline, and AstraZeneca. Judy Peih-Ying Tsai, Niyati Nathwani, Chikako Funasaka: No competing interests to disclose. Therese Cusick: Research grants from Natera and W.L. Gore & Associates. Yoichi Naito: Research grants from AbbVie, Ono, Daiichi Sankyo, Taiho, Pfizer, Boehringer Ingelheim, Eli Lilly, Eisai, AstraZeneca, Chugai, Bayer, Takeda, Anthos Therapeutics, and Gilead; consulting fees from AstraZeneca, Eisai, Gilead, Eli Lilly, Novartis, Pfizer, Chugai, Taiho, Bayer, Daiichi Sankyo, and Ono; honoraria for lectures/presentations from AstraZeneca, Eisai, Ono, Guardant, Takeda, Eli Lilly, Novartis, Pfizer, Chugai, PDR Pharma, Nihon Kayaku, Taiho, Bristol, Bayer, Daiichi Sankyo, MSD, and Gilead. Gregory Hogan, Katherine Johansen Taber, Matthew LaBella, Britney Sadler, Ravi Patel, Alexander Ham, Ashley Acevedo, CJ Battey, Genevieve M. Gould, Dale Muzzey, Jeff Jasper: Employees of Myriad Genetics; receive salary and stock compensation. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed. Medical writing support was provided by Emily Gronseth, PhD from Myriad Genetics, Inc. and was funded by Myriad Genetics, Inc. Reviewer disclosures Peer reviewers on this manuscript have received an honorarium from Future Oncology for their review work but have no other relevant financial relationships to disclose. Ethical declaration This study was conducted in accordance with ethical principles outlined in the World Medical Association Declaration of Helsinki, including review and approval by a central IRB (Advarra, Pro00070267) and local IRBs (Intermountain Health, 1052007; Cleveland Clinic, 24–104). All patients provided written informed consent. Data availability statement All relevant, nonproprietary data used to conduct analyses are available within the article. To protect the privacy of patients in this study, clinical data are not made publicly available. De-identified data can be requested from the corresponding author. Supplementary material Supplemental data for this article can be accessed online at https://doi.org/10.1080/14796694.2026.2690165 References Papers of special note have been highlighted as either of interest (•) or of considerable interest (••) to readers. Associated Data Supplementary Materials S1_Overview and Consort.jpg Supplemental tables_R1.docx S2_Sankey_RCB.jpg S3_pCR_bySubtype.jpg S4_barplots_RCB_vs_response groups.jpg STROBE_checklist_MONITOR_Breast_R1.pdf Data Availability Statement All relevant, nonproprietary data used to conduct analyses are available within the article. To protect the privacy of patients in this study, clinical data are not made publicly available. De-identified data can be requested from the corresponding author.