Skip to main content An official website of the United States government Here's how you know Here's how you know Official websites use .gov A .gov website belongs to an official government organization in the United States. Secure .gov websites use HTTPS A lock ( Lock Locked padlock icon ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites. Search Log in Dashboard Publications Account settings Log out Search… Search NCBI Primary site navigation Search Logged in as: Dashboard Publications Account settings Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide PERMALINK Copy As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice ESMO Open . 2026 Mar 31;11(4):106939. doi: 10.1016/j.esmoop.2026.106939 Search in PMC Search in PubMed View in NLM Catalog Add to search Integrating baseline ctDNA-derived tumor metrics enhances risk stratification in HR-positive/HER2-negative advanced breast cancer: a real-world multicenter cohort study from Austria N Dobrić N Dobrić 1 Division of Oncology, Department of Internal Medicine, Medical University of Graz, Graz, Austria Find articles by N Dobrić 1 , SO Hasenleithner SO Hasenleithner 1 Division of Oncology, Department of Internal Medicine, Medical University of Graz, Graz, Austria Find articles by SO Hasenleithner 1 , C Suppan C Suppan 1 Division of Oncology, Department of Internal Medicine, Medical University of Graz, Graz, Austria Find articles by C Suppan 1 , EV Klocker EV Klocker 1 Division of Oncology, Department of Internal Medicine, Medical University of Graz, Graz, Austria Find articles by EV Klocker 1 , D Hlauschek D Hlauschek 1 Division of Oncology, Department of Internal Medicine, Medical University of Graz, Graz, Austria Find articles by D Hlauschek 1 , R Graf R Graf 2 Institute of Human Genetics, Diagnostic and Research Center for Molecular Biomedicine, Medical University of Graz, Graz, Austria Find articles by R Graf 2 , C Beichler C Beichler 2 Institute of Human Genetics, Diagnostic and Research Center for Molecular Biomedicine, Medical University of Graz, Graz, Austria Find articles by C Beichler 2 , C Albertini C Albertini 3 Department of Gynecology, Breast Cancer Center Tirol, Medical University of Innsbruck, Innsbruck, Austria Find articles by C Albertini 3 , D Egle D Egle 3 Department of Gynecology, Breast Cancer Center Tirol, Medical University of Innsbruck, Innsbruck, Austria Find articles by D Egle 3 , D Liu D Liu 4 School of Medicine, Tsinghua Medicine, Tsinghua University, Beijing, China 5 Division of Oncology, Hillmans Cancer Center, University of Pittsburgh, Pittsburgh, USA Find articles by D Liu 4, 5 , AM Starzer AM Starzer 6 Division of Oncology, Department of Medicine I, Medical University of Vienna, Vienna, Austria Find articles by AM Starzer 6 , R Bartsch R Bartsch 6 Division of Oncology, Department of Medicine I, Medical University of Vienna, Vienna, Austria Find articles by R Bartsch 6 , T Moser T Moser 2 Institute of Human Genetics, Diagnostic and Research Center for Molecular Biomedicine, Medical University of Graz, Graz, Austria Find articles by T Moser 2 , G Rinnerthaler G Rinnerthaler 1 Division of Oncology, Department of Internal Medicine, Medical University of Graz, Graz, Austria Find articles by G Rinnerthaler 1 , PJ Jost PJ Jost 1 Division of Oncology, Department of Internal Medicine, Medical University of Graz, Graz, Austria Find articles by PJ Jost 1 , E Heitzer E Heitzer 2 Institute of Human Genetics, Diagnostic and Research Center for Molecular Biomedicine, Medical University of Graz, Graz, Austria 7 Christian Doppler Laboratory for Liquid Biopsies for Early Detection of Cancer, Graz, Austria Find articles by E Heitzer 2, 7 , N Dandachi N Dandachi 1 Division of Oncology, Department of Internal Medicine, Medical University of Graz, Graz, Austria 2 Institute of Human Genetics, Diagnostic and Research Center for Molecular Biomedicine, Medical University of Graz, Graz, Austria 8 Research Unit Epigenetic and Genetic Cancer Biomarkers, Medical University of Graz, Graz, Austria Find articles by N Dandachi 1, 2, 8, ∗, † , M Balic M Balic 1 Division of Oncology, Department of Internal Medicine, Medical University of Graz, Graz, Austria 5 Division of Oncology, Hillmans Cancer Center, University of Pittsburgh, Pittsburgh, USA Find articles by M Balic 1, 5, ∗, † Author information Article notes Copyright and License information 1 Division of Oncology, Department of Internal Medicine, Medical University of Graz, Graz, Austria 2 Institute of Human Genetics, Diagnostic and Research Center for Molecular Biomedicine, Medical University of Graz, Graz, Austria 3 Department of Gynecology, Breast Cancer Center Tirol, Medical University of Innsbruck, Innsbruck, Austria 4 School of Medicine, Tsinghua Medicine, Tsinghua University, Beijing, China 5 Division of Oncology, Hillmans Cancer Center, University of Pittsburgh, Pittsburgh, USA 6 Division of Oncology, Department of Medicine I, Medical University of Vienna, Vienna, Austria 7 Christian Doppler Laboratory for Liquid Biopsies for Early Detection of Cancer, Graz, Austria 8 Research Unit Epigenetic and Genetic Cancer Biomarkers, Medical University of Graz, Graz, Austria ∗ Correspondence to: Assoc. Prof. Nadia Dandachi, Institute of Human Genetics, Diagnostic and Research Center for Molecular Biomedicine, Medical University of Graz, Neue Stiftingtalstraße 6, 8010 Graz, Austria. Tel: +43-316-385-73822 [email protected] ∗ Prof. Marija Balic, Division of Oncology, Hillmans Cancer Center, University of Pittsburgh, Pittsburgh, PA, USA. Tel: +1-412-641-6500 [email protected] † These authors contributed equally to this article and share senior authorship. Collection date 2026 Apr. © 2026 The Author(s) This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). PMC Copyright notice PMCID: PMC13068563 PMID: 41922145 Abstract Background Advances in endocrine therapies for hormone receptor (HR)-positive/HER2-negative advanced breast cancer (ABC) continue to transform care and significantly improve patient outcomes. However, the integration of molecular and clinical risk stratification into guiding individualized treatment selection remains a key challenge. We therefore evaluated whether the integration of baseline circulating tumor DNA (ctDNA)-derived tumor metrics enhances risk stratification among patients receiving early lines of treatment of HR-positive ABC in a real-world multicenter cohort of patients in Austria. Methods Patients with HR-positive/HER2-negative ABC treated at multiple Austrian centers were included. CtDNA was analyzed using a 77-gene panel (AVENIO ctDNA Expanded Kit). Tumor fraction (TFx) was estimated via two complementary approaches: untargeted aneuploidy assessment using mFAST-SeqS, and the highest variant allele frequency (hVAF) from the AVENIO assay. Somatic variants and single, binary, and three-level composite TFx metrics were assessed for their association with progression-free and overall survival (PFS, OS). Results We analyzed 225 ctDNA samples from 184 patients [128 before first-line (1L) and 76 before second-line (2L) treatment], including 40 paired samples. Overall TFx was low (median z-score 2.49; range −0.5-208.3), with higher levels in 2L, although the difference did not reach statistical significance ( P = 0.058). In contrast, the hVAF was significantly higher in the 2L cohort ( P = 0.007). Somatic variant burden was significantly increased in 2L ( P < 0.001), with notably more frequent ESR1 mutations (26.7% versus 7.1% in 1L). Median PFS was 29.2 months in 1L and 6.0 months in 2L, while median OS was 57.3 months and 16.1 months, respectively. TP53 and ESR1 mutations, and all ctDNA-based metrics were significantly associated with PFS and OS, with a three-level composite ctDNA variable showing the highest prognostic discrimination. Conclusions Our findings demonstrate that integrating baseline ctDNA-derived TFx metrics with established clinical variables significantly improves risk stratification in HR-positive/HER2-negative ABC. Key words: HR-positive/HER2-negative breast cancer, metastatic breast cancer, ctDNA, liquid biopsy, next-generation sequencing, precision medicine Highlights • The number of samples with elevated ctDNA increased using a combined metric relative to z-scores and hVAF alone. • ESR1 and TP53 mutations, and all ctDNA-based tumor fraction measures, were significantly associated with survival. • A three-level composite ctDNA tumor fraction variable achieved the highest prognostic discrimination. • Integrating baseline ctDNA-derived metrics significantly improves risk stratification of HR-positive ABC patients. Introduction Breast cancer (BC) is the most frequent malignancy and the leading cause of cancer-related death among women worldwide. 1 , 2 For hormone receptor (HR)-positive/HER2-negative BC, the most common BC subtype, endocrine treatment (ET) represents the first choice of therapy, most commonly in combination with cyclin-dependent kinase 4/6 inhibitors (CDK4/6i). However, due to the substantial molecular heterogeneity of the disease, a considerable proportion of patients derive limited benefit from current standard-of-care treatments, and ultimately, all experience disease progression. 3 , 4 , 5 Over the past two decades, advances in translational research and high-throughput sequencing technologies have greatly improved the genomic characterization of BC. In particular, liquid biopsies have emerged as a minimally invasive, comprehensive approach for tumor molecular landscape profiling, with the most widely used method being the analysis of circulating tumor DNA (ctDNA) from blood. 6 , 7 , 8 Beyond accessibility and repeatability, ctDNA uniquely reflects the current disease state, which is particularly valuable in metastatic cancer, where spatial and temporal heterogeneity can be extensive. This real-time molecular information provides insight into both primary and metastatic tumor biology, the emergence of acquired mutations, and therapeutic response dynamics, offering tangible advantages over tissue biopsies. 7 , 9 , 10 To date, several drugs have received regulatory approval for the management of HR-positive/HER2-negative advanced breast cancer (ABC), often grounded in ctDNA-based molecular profiling. Notable trials such as SOLAR-1, EMERALD, INAVO120, and SERENA-6 are actively redefining strategies for the first-line and early postprogression setting. 11 , 12 , 13 , 14 , 15 , 16 , 17 As the therapeutic landscape continues to evolve, studies in early treatment phases are pivotal, since therapy response remains unpredictable at diagnosis. Recent studies indicate that baseline ctDNA characteristics provide important prognostic information in HR-positive/HER2-negative ABC, reporting that patients with higher tumor fractions (TFx), elevated variant allele frequencies (VAFs), or an increased number of somatic mutations at treatment start experience shorter progression-free and overall survival (PFS, OS). 18 , 19 , 20 , 21 While these studies provide valuable insights, they predominantly represent ancillary analyses from clinical trials and retrospective real-world studies, or employ tumor-informed assays, which may limit their generalizability in the metastatic setting. Furthermore, many focus on single genomic alterations or individual ctDNA parameters, possibly providing an incomplete representation of tumor biology and disease aggressiveness. In this prospective multicenter study, we used ctDNA to characterize the genomic and prognostic landscape of a real-world HR-positive/HER2-negative ABC population in the early lines of therapy, using a tumor-agnostic approach. Our aim was to enable biologically and clinically meaningful patient stratification through the integration of baseline ctDNA-derived tumor metrics, providing a framework for improved risk assessment and the future evaluation of risk-adapted treatment strategies. Patients and methods Study design This prospective, observational study enrolled patients with locally advanced or metastatic BC at three centers in Austria (Medical University of Graz, Medical University of Innsbruck, and Medical University of Vienna) either at ABC diagnosis (baseline) or following progression on early lines of ET. In a prior publication, 22 we reported a methodological comparison of PIK3CA testing in a subset of this cohort. Here, we present expanded data from a larger patient cohort to establish the prognostic value of ctDNA analyses. All participants were clinically diagnosed with HR-positive/HER2-negative ABC, with estrogen receptor (ER) expression, with or without progesterone receptor (PR) expression, and with HER2 negativity characterized either as 0 or 1+ score at evaluation via immunohistochemistry (IHC), or as 2+ IHC score without amplification based on in situ hybridization (ISH). The diagnosis and subtyping were from the most recent tissue samples obtained from either primary or metastatic lesions. Samples with either IHC 1+ or 2+ without amplification were classified as HER2-low. Clinical data were systematically collected and managed via the REDCap electronic data capture platform hosted by the Medical University of Graz. 23 Blood collection and cfDNA extraction All blood samples were collected either before the initiation of treatment of metastatic disease, or immediately before the start of a new systemic therapy. Circulating cell-free DNA (cfDNA) was isolated from blood as previously described, 22 , 24 with minor adaptations. Briefly, 20 ml (range 10-30 ml) of whole blood was collected into PAXgene Blood ccfDNA Tubes (QIAGEN, Hilden, Germany) at the participating centers and shipped to Graz at room temperature. Plasma was separated in two centrifugation steps at 1900 × g for 10 min and subsequently aliquoted into 2 ml tubes, which were stored at −80°C until further processing. cfDNA was extracted from 1-4 ml of plasma using the QIAamp Circulating Nucleic Acid Kit (QIAGEN, Hilden, Germany) or the QIASymphony PAXgene Blood ccfDNA Kit (QIAGEN, Hilden, Germany), following the manufacturer’s instructions. Isolated DNA was quantified using the Qubit 1X dsDNA High Sensitivity Assay Kit (Thermo Fisher Scientific, Vienna, Austria). Library preparation and panel sequencing The AVENIO ctDNA Expanded Kit (Roche, Basel, Switzerland), a hybridization capture-based next-generation sequencing assay targeting 77 cancer-associated genes, was used for molecular profiling of the cfDNA samples. Library preparation was carried out with a cfDNA input of 10-50 ng, following the manufacturer’s protocol. Libraries were quantified using the Qubit 1X dsDNA HS Assay Kit, and 4-16 libraries were pooled equimolarly and quantified using qPCR. Library pools were sequenced in paired-end runs, generating 2× 150 bp reads on an Illumina NextSeq (Mid or High Output kit) or NovaSeq 6000 platform (SP flowcell, XP workflow) (Illumina, San Diego, CA). The sequencing produced an average of 38 million read pairs per sample (range 13-82 million). After generating consensus reads, the average read depth across all samples was 3399× (range 943-8557×), with an average fragment length of 176 bp (range 159-355 bp). Variant calling and filtering Variant calling was carried out using the AVENIO Oncology Analysis Software (version 2.1.0, Roche, Basel, Switzerland), with customized filtration settings: variants with a minor allele frequency ≥1% as defined by ExAC version 1.0 or 1000 Genomes version phase_3_v5b databases, or those listed as common single nucleotide polymorphisms (SNPs) in the dbSNP150 database, were excluded by the software. Putative germline variants (characterized by a VAF ∼50% but a low TFx context) were additionally removed. To enable a high confidence variant call set, variants that passed these filters but had <10 mutated reads or a VAF below the assay limit of detection (LOD), as well as recurrent low-level variants observed in multiple patients (suggesting a sequencing or assay artifact) were flagged and manually excluded. The remaining variants were annotated and classified according to their pathogenicity using Golden Helix VarSeq v2.2.0 (Golden Helix Inc., Bozeman, MT) and the OncoKB database. 25 For the primary analyses, a VAF threshold of 0.5% was applied to prioritize specificity and analytical confidence over maximal sensitivity, given the risk of false-positive variant calls from the hematopoietic background. Additionally, a secondary analysis using a lower detection threshold (LOD = 0.1% VAF) was conducted to assess the impact of including low-VAF variants on overall genomic landscape patterns ( Supplementary Methods , available at https://doi.org/10.1016/j.esmoop.2026.106939 ). Tumor fraction estimation To estimate tumor burden in plasma, TFx was calculated via two complementary methods: untargeted aneuploidy assessment using the modified Fast Aneuploidy Screening Test-Sequencing System (mFAST-SeqS), and the highest observed variant allele frequency (hVAF) among all somatic mutations detected with the AVENIO ctDNA Expanded Kit. When low-pass whole-genome sequencing data was available, TFx was calculated using the ichorCNA algorithm, as an additional exploratory comparator ( Supplementary Methods , available at https://doi.org/10.1016/j.esmoop.2026.106939 ). mFAST-SeqS The mFAST-SeqS assay is an untargeted assay that estimates chromosomal aneuploidy from low-coverage sequencing data by quantifying deviations in the long interspersed nuclear element-1 (LINE1) sequence count from a normal representation in control samples. LINE1 amplicon libraries were prepared as previously described, 26 and sequenced on an Illumina MiSeq or NextSeq platform. The sequencing generated 150 bp single-end or 76 bp paired-end reads, aiming for at least 100 000 reads per sample. Z-score statistics were used to assess over- and under-representation of each chromosome arm, and a genome-wide z-score (z-score) was introduced as a surrogate for TFx. While mFAST-SeqS was originally developed using a z-score threshold of 5, 26 later work has demonstrated that z-scores with an absolute value ≥3 already provide informative TFx estimates, reflecting copy-number gains or losses. 27 We therefore applied a cut-off of 3 to define elevated TFx. Mutation-based tumor fraction Additionally, TFx was assessed based on the VAFs of variants detected with the AVENIO ctDNA Expanded Kit, aiming to complement the genome-wide TFx measure with a targeted approach and to improve risk stratification in patients with ABC. Mutation-based TFx was defined as the TFx estimated from somatic mutations, with hVAF among all detected somatic variants being considered as the representative metric. Considering the detection thresholds of AVENIO and mFAST-SeqS, we defined ctDNA positivity (ctDNA+) as the presence of at least one somatic variant with a VAF above the LOD of 0.5% or a z-score ≥3. Recently, we have demonstrated that combining z-score TFx estimates with hVAF generated from the AVENIO ctDNA Expanded Kit using a 1% VAF cut-off improves the reliability of negative liquid biopsy findings and can help guide therapeutic decision making. 22 Therefore, for the estimation of elevated ctDNA levels and further risk stratification, we set a combined definition that included the hVAF in a sample ≥1% or a z-score ≥3. Based on these definitions, to assess the impact of TFx on prognosis, we used two binary composite metrics: ctDNA+ (detection one or more somatic variant or z-score ≥3) and elevated ctDNA (z-score ≥3 or hVAF ≥1%), as well as a three-level composite variable, which stratified patients into low (z-score <3 and hVAF <1%), intermediate (either z-score ≥3 or hVAF ≥1%), and high (both z-score ≥3 and hVAF ≥1%) TFx. Statistical analysis Statistical analyses were carried out with GraphPad Prism version 10.5.0, STATA 19.5, or R version 4.3.1. Descriptive statistics were used to summarize patient and tumor characteristics, reporting medians and ranges for continuous data and counts/percentages for categorical data. To account for repeated measurements in some patients, we used linear mixed-effects models (STATA routine mixed command) with random intercepts at the patient level for continuous outcomes, and logistic regression models with cluster-robust standard errors for binary outcomes. For count outcomes, we applied a mixed-effects negative binomial regression model to estimate incidence rate ratios. The pairwise co-occurrence and mutual exclusivity among somatic alterations in the mutated genes were assessed using the somaticInteractions function in maftools. 28 Fisher’s exact test was employed to analyze differences in variable frequencies. A McNemar test was applied to paired samples to detect discordant mutation patterns between 1L and 2L samples. PFS was defined as the time from the blood draw to progression or death from any cause. Of note, this represents real-world PFS, determined based on imaging and clinical interpretation of progression, as real-world reporting lacks standardized RECIST criteria. OS was defined as the time from the blood draw to death from any cause. Patients without an endpoint event at the last follow-up visit were censored. Median follow-up time was estimated using the reverse Kaplan–Meier estimator. 29 Differences in survival were tested using the log-rank test and graphically represented by Kaplan–Meier curves. Univariable and multivariable survival analyses, including clinically relevant variables, were carried out using Cox proportional hazards regression. The proportional hazards assumption was tested using Schoenfeld tests. Hazard ratios and 95% confidence intervals (CIs) were estimated. Model discrimination was assessed using Uno’s concordance index (C-index) with 5000 bootstrap resamples. 30 Model performance and goodness of fit were compared using the Akaike information criterion (AIC). 31 All statistical tests were two-sided, and P values <0.05 were considered statistically significant. Ethical considerations This study was approved by the Austrian ethics committees (approval number 32-415 ex 19/20 for the Medical University of Graz, 1495/2023 for the Medical University of Vienna, and 1164/2022 for the Medical University of Innsbruck). Written informed consent was obtained from all patients. All procedures adhered to the ethical standards of the Declaration of Helsinki and the guidelines for good scientific practice required by the Medical University of Graz. Results Patient characteristics and plasma samples In total, 225 blood samples were collected from 184 patients: 128 samples obtained before starting first-line therapy (1L), 76 before starting second-line therapy (2L), and 21 before starting third- or fourth-line therapy (3L, 4L). Forty patients had paired samples, 32 having blood drawn before 1L and 2L. The sample overview is shown in Supplementary Figure S1 A and B, available at https://doi.org/10.1016/j.esmoop.2026.106939 . The summarized clinicopathological characteristics of the patient cohort are displayed in Table 1 . For patients with more than one blood draw, the status at the first one was used for description. Table 1. Patient and tumor characteristics of the study cohort at the time of enrollment N = 184 n (%) Age at blood sample collection, years (range) 66.1 (28.8-93.3) Sex Female 178 (96.7) Male 6 (3.3) Histological type NST 131 (71.2) ILC 34 (18.5) Mixed 10 (5.4) Unknown/other 9 (4.9) HER2 subtype HER2-0 a 63 (34.3) HER2-low b 118 (64.1) Unknown 3 (1.6) Disease setting at diagnosis of advanced disease De novo metastatic 73 (39.7) Metastatic recurrent 104 (56.5) Locally advanced 7 (3.8) Metastases localization Liver c 34 (19.2) Visceral (nonliver) d 58 (32.8) Nonvisceral e 27 (15.2) Bone only 54 (30.5) Other f 4 (2.3) Number of metastatic organs 0 7 (3.8) 1 81 (44.0) 2 55 (29.9) 3 31 (16.9) ≥4 10 (5.4) Time from initial diagnosis to first metastases ≤24 months 7 (6.7) >24 months 97 (93.3) Number of previous palliative lines 0 128 (69.6) 1 44 (23.9) 2 11 (6) 3 1 (0.5) Open in a new tab CNS, central nervous system; IHC, immunohistochemistry; ILC, invasive lobular carcinoma; ISH, in situ hybridization; NST, no special type. a HER2-0: samples with an IHC 0 score. b HER2-low: samples with either an IHC 1+ score, or 2+ without amplification based on ISH. c Liver: alone or with other metastases. d Visceral (nonliver): lung/pleura, peritoneal cavity; if any of the locations were visceral, the patient was counted as visceral. e Nonvisceral: bone (with others), brain/CNS, lymph nodes. f Other: breast, soft tissue, ovary. Genomic landscape Of 225 samples, 221 could be analyzed with the AVENIO assay, while four samples were excluded due to insufficient material. Of the 77 genes covered by the AVENIO panel, at least one mutation was detected in 68 genes across the entire cohort over all time points. Altogether, a total of 529 somatic variants were identified in 160/221 (72.4%) samples, with a median of two variants per sample (range 1-28 variants). Among the samples with detectable somatic variants, the median hVAF was 4.7% (range 0.5%-75.4%). The mutational landscape was analyzed separately for samples collected before 1L ( N = 126) and 2L ( N = 75) treatment. In 1L samples, somatic variants were detected in 82 cases (65.1%), and pathogenic or likely pathogenic (P/LP) variants in 62 cases (49.2%). In 2L samples, somatic variants were detected in 61 samples (81.3%), and P/LP variants in 52 samples (69.3%). A mixed-effects negative binomial model adjusting for elevated TFx (z-scores ≥3) showed that 2L samples had 80% more somatic variants than 1L samples (incidence rate ratio 1.80, 95% CI 1.36-2.38, P < 0.001). Subsequent analyses comparing 1L and 2L samples were restricted to P/LP variants. The genes most frequently mutated for P/LP variants were PIK3CA (32.5% and 40.0%), TP53 (16.7% and 24.0%), and ESR1 (7.1% and 26.7%) ( Supplementary Figure S2 , available at https://doi.org/10.1016/j.esmoop.2026.106939 ), with ESR1 mutations being significantly more frequent in 2L samples (odds ratio 4.7, 95% CI 2.2-10.0, P < 0.001, cluster-robust logistic regression). Consistent with these sample-based findings, in a subset of 31 patients with successfully analyzed paired 1L and 2L samples ( Figure 1 A), ESR1 emerged as the most frequently acquired mutation (25.8%, McNemar test, P = 0.005). RB1, PIK3CA , and TP53 also showed numerical increases in acquired mutations, although these did not reach statistical significance. Mutations affecting the PI3K pathway ( PIK3CA , AKT1 , and/or PTEN genes) were present in 34.9% of 1L and 44.0% of 2L samples. The most common PIK3CA variants across both treatment lines were H1047R (16.7%), E545K (8.4%), and E542K (4.4%). Within ESR1 , D538G was the most frequent variant (6.4%), while Y537S mutations became more prevalent in 2L samples. TP53 variants occurred predominantly within the DNA-binding domain but were heterogeneous in their positional distributions. We also observed evidence of increasing clonal complexity. Among ESR1 -mutated samples, polyclonality was detected in 11% (1/9) of 1L samples and increased to 35% (7/20) in 2L samples. Similarly, multiple concomitant mutations within PIK3CA , TP53 , and RB1 were more frequently detected in 2L samples. Figure 1. Open in a new tab Genomic landscape of patients with HR-positive/HER2-negative ABC. (A) Acquired P/LP somatic alterations detected in a subset of 31 patients with paired samples taken before 1L (gray) and 2L (black) treatment. Bars on the right represent the number of acquired alterations detected in each gene. Colors represent the mutation type: missense (dark green), nonsense (purple), and splice site (light green). (B, C) Co-occurrence plots of the top 10 mutated genes in 1L and 2L samples (pairwise Fisher’s exact test). 1L, first-line; 2L, second-line; ABC, advanced breast cancer; P/LP, pathogenic/likely pathogenic. ∗Statistically significant ( P < 0.05). A sensitivity analysis evaluating variant detection at increased analytical sensitivity (LOD = 0.1%) yielded similar patterns of recurrent alterations, described in the Supplementary Results and shown in Supplementary Figure S3 , available at https://doi.org/10.1016/j.esmoop.2026.106939 . We next explored co-mutation patterns among P/LP mutations in 1L and 2L samples. In both settings, RB1 mutations consistently co-occurred with PIK3CA mutations (1L: 3.1%, unadjusted P = 0.009; 2L: 9.2%, unadjusted P < 0.001, Fisher’s exact test). Among 1L samples, all SMAD4 -mutated samples also harbored a concurrent PIK3CA mutation (3.1%, unadjusted P = 0.009, Fisher’s exact test). Interestingly, a significant co-occurrence was observed between ESR1 and TP53 mutations in 1L samples (3.9%, unadjusted P = 0.039, Fisher’s exact test). In 2L samples, significant co-occurrence was found between ERBB2 and TP53 (5.3%, unadjusted P = 0.039, Fisher’s exact test) ( Figure 1 B and C). Given the growing clinical relevance of the HER2-low phenotype as a therapeutic target for trastuzumab deruxtecan (T-DXd), 32 we compared the mutational profiles of HER2-low and HER2-0 patients in our HR-positive/HER2-negative cohort ( Supplementary Figure S4 , available at https://doi.org/10.1016/j.esmoop.2026.106939 ). The two groups did not show significant differences ( Supplementary Results , available at https://doi.org/10.1016/j.esmoop.2026.106939 ). For a subset of 1L and 2L samples with available low-pass whole-genome sequencing data, we carried out genome-wide copy-number analysis using GISTIC 2.0, which identified recurrent alterations in both groups, with a broader heterogeneity observed in the 1L cohort ( Supplementary Results and Figure S5 , available at https://doi.org/10.1016/j.esmoop.2026.106939 ). Quantification of tumor fractions using ctDNA metrics To quantify ctDNA levels as a proxy for tumor burden, we leveraged two complementary ctDNA-based metrics: the mFASTSeqS-derived z-scores and the AVENIO-derived hVAF. Across 225 samples, TFx estimates derived from mFAST-SeqS indicated overall low tumor content in this cohort (median z-score 2.49, range −0.5-208.3). Using the threshold of z-score ≥3, elevated TFx were observed in 41.3% of samples. The z-scores significantly correlated with hVAF from targeted sequencing (coefficient 1.21, 95% CI 1.05-1.37, P < 0.001; Supplementary Figure S6 A, available at https://doi.org/10.1016/j.esmoop.2026.106939 ). In line with ctDNA-based TFx reflecting tumor burden, both z-scores and hVAF were numerically higher in 2L than in 1L samples. The difference was statistically significant for hVAF (adjusted mean difference 4.74, 95% CI 1.28-8.20, P = 0.007), but not for z-scores (adjusted mean difference 6.33, 95% CI −0.2-12.9, P = 0.058). After adjusting for elevated TFx (z-score ≥3), the association for hVAF remained significant (adjusted mean difference 3.49, 95% CI 0.49-6.49, P = 0.023), indicating a greater somatic variant burden in 2L samples independent of chromosomal instability. While both assay metrics showed overlapping detection profiles, our results also demonstrate their complementary analytical sensitivity. Among 221 evaluable samples, both metrics identified elevated ctDNA in 71 cases (32.1%), whereas 65 samples (29.4%) were positive by hVAF ≥1% only, and 20 samples (9.0%) by z-score ≥3 only. Because mFAST-SeqS and hVAF capture different biological features of ctDNA—genome-wide copy-number imbalance and high-frequency somatic point mutations, respectively—we combined both metrics to better identify patients with elevated ctDNA through either mechanism. Using this combined definition (z-score ≥3 or hVAF ≥1%), the overall detection rate of elevated ctDNA increased to 70.6%, including 62.7% before 1L treatment and 81.3% before 2L treatment, demonstrating improved detection, particularly in samples with low tumor burden. Figure 2 A illustrates assay concordance, and Figure 2 B shows the increase in the number of samples with elevated ctDNA detected using the combined metric relative to each metric alone. Figure 2. Open in a new tab Complementary analytical sensitivity of mFAST-SeqS-derived z-scores and the AVENIO-derived hVAF in 221 samples. (A) Assay concordance shown as the number of samples with elevated ctDNA by one or both metrics. Green bars represent concordant samples and purple bars represent discordant samples. (B) The increase in the number of samples with elevated ctDNA (green) detected using the combined metric (z-score ≥3 or hVAF ≥1%) compared with z-scores and hVAF alone, shown in all samples, as well as in 1L and 2L separately. 1L, first-line; 2L, second-line; ctDNA, circulating tumor DNA; hVAF, highest variant allele frequency. Prognostic value of baseline ctDNA-derived metrics PFS and OS were evaluated across 127 patients starting 1L and 76 patients starting 2L therapy. Median follow-up time was 35.8 months (25th-75th percentile: 17.1-54.4 months) for 1L patients and 25.5 months (25th-75th percentile: 22.1-52.2 months) for 2L patients. During this follow-up, 64 PFS events and 47 OS events occurred in 1L patients, while 65 PFS events and 52 OS events occurred in 2L patients. Median PFS was 29.2 months (95% CI 19.1-37.2 months) in 1L and 6.0 months (95% CI 5.1-7.3 months) in 2L, while median OS was 57.3 months (95% CI 35.3 months-not reached) and 16.1 months (95% CI 12.6-19.8 months), respectively ( Supplementary Figure S7 , available at https://doi.org/10.1016/j.esmoop.2026.106939 ). We comprehensively evaluated multiple ctDNA-derived metrics, including z-scores generated with mFAST-SeqS (z-score ≥3), AVENIO-based hVAF (hVAF ≥1%), and composite variables. Two binary composite metrics were assessed: ctDNA+ (detection one or more somatic variant or z-score ≥3) and a metric defined as z-score ≥3 or hVAF ≥1%. We further refined the ctDNA classification using a three-level composite variable, stratifying patients into low (z-score <3 and hVAF <1%), intermediate (either z-score ≥3 or hVAF ≥1%), and high (both z-score ≥3 and hVAF ≥1%) TFx groups. Additionally, total somatic variant count and the three most frequently mutated genes ( PIK3CA , TP53 , ESR1 ) were assessed. In univariable Cox regression analyses performed in 1L patients, all ctDNA-based measures demonstrated significant associations with PFS and OS, except for PIK3CA mutations, which showed no significant association with PFS ( Figure 3 A and Supplementary Figure S8 A, available at https://doi.org/10.1016/j.esmoop.2026.106939 ). The three-level composite ctDNA variable achieved the highest prognostic discrimination across both endpoints as measured by Uno’s C-index. Kaplan–Meier curves illustrate the stepwise decline in PFS and OS with increasing ctDNA burden ( Figure 3 B-G and Supplementary Figure S8 B-G, available at https://doi.org/10.1016/j.esmoop.2026.106939 ). In 2L patients, ctDNA-derived metrics remained significantly associated with PFS and OS, and the three-level composite ctDNA variable again provided the highest prognostic discrimination across both endpoints ( Supplementary Figure S9 , available at https://doi.org/10.1016/j.esmoop.2026.106939 ). Figure 3. Open in a new tab Association of ctDNA -based measures with PFS of 1L patients. (A) Forest plot representing the performance of individual and composite ctDNA TFx metrics, the total somatic variant count and the most frequently mutated genes, based on univariable Cox regression, in relation to PFS. The prognostic discrimination was measured by Uno’s C-index. (B, C) Kaplan–Meier curves distinguishing patients with high (red) and low (black) TFx estimated by mFAST-SeqS z-scores and AVENIO-based hVAF individually. (D) Kaplan–Meier curves distinguishing ctDNA+ (≥1 detected somatic variant or z-score ≥3, red) and ctDNA− (<1 detected somatic variant and z-score <1%, black) patients. (E) Kaplan–Meier curves illustrating the difference in PFS using a three-level composite variable stratifying patients into high (z-score ≥3 and hVAF ≥1%, red), intermediate (z-score ≥3 or hVAF ≥1%, blue) and low (z-score 3 and hVAF <1%, black) ctDNA burden. This variable achieved the highest prognostic discrimination. (F, G) Kaplan–Meier curves distinguishing patients with (mt, red) and without (wt, black) mutations in TP53 and ESR1 . 1L, first-line; CI, confidence interval; ctDNA, circulating tumor DNA; ctDNA+, ctDNA-positive; ctDNA-, ctDNA-negative; HR, hazard ratio; hVAF, highest variant allele frequency; mt, mutated; NR, not reached; PFS, progression-free survival; TFx, tumor fraction; Uno’s C, Uno’s concordance index; wt, wild-type. To validate these findings in a multivariable setting, we constructed a clinical model for 1L patients that included age, PR status, CDK4/6i treatment, liver metastases, number of metastatic organs, and time to metastasis (categorized as de novo , ≤24 months, or >24 months). In this model, PR positivity (hazard ratio 0.47, 95% CI 0.27-0.82, P = 0.007) and CDK4/6i treatment (hazard ratio 0.22, 95% CI 0.09-0.54, P < 0.001) were independently associated with improved PFS, while liver metastases (hazard ratio 2.03, 95% CI 1.04-3.98, P = 0.039) and time to metastasis ≤24 months (hazard ratio 14.02, 95% CI 4.47-44.00, P < 0.001) predicted shorter PFS ( Figure 4 A). Comparable associations were observed in the multivariable clinical model for OS ( Supplementary Figure S10 A, available at https://doi.org/10.1016/j.esmoop.2026.106939 ). Figure 4. Open in a new tab Multivariable PFS models of 1L patients. (A) Forest plot representing the clinical model including known clinically relevant variables. (B) Forest plot representing the clinical model incorporating the three-level composite ctDNA metric. Increased Uno’s C-index and reduced AIC indicate a significantly improved prognostic performance of the extended model. 1L, first-line; AIC, Akaike information criterion; CDKi, CDK 4/6 inhibitor; CI, confidence interval; ctDNA, circulating tumor DNA; ET, endocrine treatment; HR, hazard ratio; PFS, progression-free survival; PR, progesterone receptor; Uno’s C, Uno’s concordance index. Adding the three-level composite ctDNA metric to this clinical model moderately improved prognostic performance, with Uno’s C-index increasing from 0.701 (95% CI 0.631-0.772) to 0.758 (95% CI 0.698-0.818) for PFS, and a corresponding reduction in AIC ( Figure 4 B). Both intermediate (hazard ratio 2.11, 95% CI 1.06-4.22, P = 0.034) and high (hazard ratio 4.43, 95% CI 2.17-9.03, P < 0.001) ctDNA burden remained independently associated with worse PFS after adjustment for clinical covariates. Similar improvements in prognostic performance were observed for OS ( Supplementary Figure S10 B, available at https://doi.org/10.1016/j.esmoop.2026.106939 ). To confirm that these associations were not driven by treatment heterogeneity, we carried out a sensitivity analysis restricted to the 117 1L patients who received ET plus a CDK4/6i. The results remained consistent in both univariable and multivariable analyses (data not shown). Discussion Our findings demonstrate that integrating baseline ctDNA-derived TFx metrics with established clinical variables significantly improves risk stratification in HR-positive/HER2-negative ABC. The composite ctDNA variable, measured at a single pretreatment time point, provided prognostic information beyond conventional clinicopathologic factors, capturing underlying biological heterogeneity and disease aggressiveness. Our results are consistent with a growing body of evidence demonstrating that baseline ctDNA levels and mutational burden are strong prognostic markers in HR-positive/HER2-negative ABC. Prospective clinical trials and real-world cohorts report that higher ctDNA fractions, elevated VAFs, and increased somatic mutations at treatment initiation are consistently associated with shorter PFS and OS, largely independent of the treatment regimen. 18 , 19 , 20 , 21 , 33 While many earlier analyses focused on single genomic alterations or individual ctDNA parameters considered separately, 34 , 35 more recent studies have highlighted the prognostic relevance of quantitative ctDNA measures and composite metrics that more comprehensively capture tumor burden and biological aggressiveness. 18 , 19 , 20 , 21 However, existing studies vary substantially in design, patient selection, applied assays, and ctDNA metrics evaluated, underscoring the need for complementary evidence from prospective real-world cohorts. Our prospective multicenter cohort addresses this gap, characterizing baseline and early-line disease biology in HR-positive/HER2-negative ABC patients treated in across Austrian centers in a uniform clinical setting. Most detected P/LP alterations were concordant with previously reported mutational profiles in HR-positive/HER2-negative ABC, including recurrent variants in PIK3CA , TP53 , and ESR1 . 20 , 36 , 37 , 38 As anticipated, ESR1 mutations—known to arise predominantly as a mechanism of acquired resistance under selective pressure from aromatase inhibitor (AI) therapy—were identified at a relatively low frequency before 1L treatment (7.1%). 36 However, their presence at baseline was associated with significantly shorter PFS. Since the detection of ESR1 mutations did not inform the choice of ET, as oral selective estrogen receptor degraders (SERDs) were not approved at the time of the study and treatment initiation, and the decision to treat patients with fulvestrant was based on clinical history and not on reported results, this observation likely reflects the underlying biology rather than treatment selection bias. Importantly, it underscores the need for novel endocrine agents with efficacy in ESR1 -mutated tumors and for upfront combination strategies to overcome or delay endocrine resistance. 16 , 39 Notably, among our nine patients harboring ESR1 mutations at 1L, the majority achieved a relapse-free survival exceeding 24 months; two had de novo metastatic disease, and a higher proportion received fulvestrant-based regimens compared with AI. Before the initiation of 2L therapy, ESR1 mutations rose markedly, likely reflecting the selective pressure exerted by AI plus CDK4/6i combinations, which constituted the predominant 1L regimen in our cohort. Overall, these findings are consistent with previously reported genomic evolution patterns in HR-positive ABC. 40 , 41 We observed significant co-occurrence between ESR1 and TP53 mutations in our samples before 1L treatment, which contradicts earlier reports suggesting mutual exclusivity. 42 , 43 In a study by Li et al., the authors analyzed multiple independent patient cohorts and reported that in ER-positive metastatic BC, ESR1 and TP53 mutations were mutually exclusive regardless of the histological subtype or distant metastatic sites, stating that the exclusivity broadly applies to all ER-positive breast tumors. 42 This discrepancy in our findings may be due to the retrospective nature of previous studies, which included patients from different trials who were exposed to different prior lines of therapy. Further supporting this interpretation, Bielo et al. reported mutual exclusivity of ESR1 and TP53 mutations in metastatic ER-positive/HER2-negative BC 44 ; however, their analyzed cohort included only patients previously exposed to ET. In contrast, our study was designed to include patients at defined timepoints, and we report the ESR1 - TP53 mutation co-occurrence specifically in ABC patients naïve for treatment of metastatic disease. Not only ESR1 , but also TP53 mutations detected at baseline were associated with an unfavorable prognosis, indicating that both acquired endocrine resistance mechanisms and intrinsic genomic instability contribute to poorer clinical outcomes in this population. Nevertheless, these findings are considered to be exploratory and should be interpreted with caution. Previously, we have used z-scores generated with mFAST-SeqS as a surrogate for TFx, establishing a cut-off of ≥3 for elevated TFx. 27 However, as these values are often biased toward the degree of aneuploidy in a sample, combined approaches are becoming increasingly used for the assessment of TFx. 45 , 46 In this study, we additionally used hVAF as a TFx measure because it captures low-level ctDNA from copy-number-quiet tumors through focal high-frequency mutations. Nevertheless, copy-number-based approaches detect tumor DNA even when no panel mutations are identified and help contextualize variant clonality. Together, these complementary properties explain differing detection rates and support their combined use. Next, we applied two binary composite metrics—defined as either the detection of one or more somatic variants or a z-score ≥3, or alternatively a z-score ≥3 or hVAF ≥1%—to assess ctDNA positivity and levels. Further refinement using a three-tier composite classification stratified patients into low (z-score <3 and hVAF <1%), intermediate (z-score ≥3 or hVAF ≥1%), and high (z-score ≥3 and hVAF ≥1%) groups. This three-level model demonstrated superior discriminatory capacity and enhanced potential for clinical risk stratification. Our uni- and multivariable analyses demonstrate that integrating ctDNA-derived TFx estimates with established clinical variables moderately improves prognostic stratification in patients with HR-positive/HER2-negative ABC. While established clinical factors, such as PR positivity, early relapse within 24 months, and CDK4/6i exposure, remained significant predictors of outcome, incorporation of the three-level composite ctDNA variable further improved the overall model performance and prognostic discrimination. This improvement suggests that ctDNA-based metrics capture complementary and biologically meaningful information beyond conventional clinicopathologic variables, supporting their integration into standardized reporting and interpretation of ctDNA analysis results. In practical terms, patients characterized as having high ctDNA burden may warrant closer monitoring, earlier escalated therapy, or entry into trials of novel agents, whereas those with low ctDNA burden may be candidates for treatment de-escalation or more conservative management. With the expanding early treatment options for HR-positive/HER2-negative ABC, CDK4/6i-based combinations have become standard. Trials such as PADA-1 and SERENA-6 established ctDNA-guided therapy adaptation based on emerging ESR1 mutations, but focused on a single resistance mechanism. 16 , 47 INAVO120 incorporated both clinical risk and molecular selection, 15 while SONIA demonstrated similar OS regardless of whether CDK4/6i were used in 1L or 2L, highlighting the need to identify patients who may benefit from de-escalated 1L ET alone. 48 Exploratory SONIA analyses and emerging prospective data suggest that ctDNA dynamics may refine treatment sequencing and capture tumor evolution under therapy. Our analysis, limited to baseline ctDNA metrics, provides an initial prognostic framework, but would likely be strengthened by serial assessments. We hypothesize that early on-treatment ctDNA kinetics could enable adaptive escalation or de-escalation strategies, an approach currently under prospective evaluation in our group. Several limitations warrant consideration. Our study was prognostic and not powered to assess predictive treatment effects. The cohort size was modest, and serial ctDNA sampling was unavailable. The AVENIO tumor-agnostic assay prioritized clinical feasibility over maximal sensitivity at very low VAFs, potentially under-representing low-burden disease. Lowering the VAF threshold increased detection but also the uncertainty regarding clonal hematopoiesis, particularly for low-level TP53 variants. 49 As matched white blood cell controls were not available for most patients, this could not be definitively resolved. In addition, limited indel and copy-number detection restricted assessment of key resistance-associated genes (e.g. BRCA1/2 , CCND1 , FGFR1 50 , 51 , 52 ). Finally, co-occurrence analyses were exploratory and require independent validation. Despite these limitations, our findings demonstrate the prognostic value of ctDNA-based metrics in HR-positive/HER2-negative ABC. A composite ctDNA variable integrating quantitative parameters improved risk discrimination beyond clinicopathologic factors, supporting ctDNA as a practical tool for baseline risk stratification. Prospective integration of both baseline and longitudinal ctDNA assessments will be essential to determine whether dynamic changes can further provide predictive and response-adaptive value, enabling more personalized, biology-driven management of HR-positive/HER2-negative ABC. Declaration of Generative Artificial Intelligence and Artificial Intelligence-Assisted Technologies in the Writing Process During the preparation of this work the author(s) used ChatGPT (OpenAI) in order to improve the clarity and phrasing of the text in English. After using this tool/service, the author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the publication. Acknowledgments Funding This work was supported by Novartis, AstraZeneca, Daiichi Sankyo, and Pfizer (no grant numbers), and by the Austrian Federal Ministry for Digital and Economic Affairs (no grant number) (Christian Doppler Research Fund for Liquid Biopsies for Early Detection of Cancer). Disclosure MB received honoraria from AstraZeneca, Daiichi Sankyo, Eli Lilly, Gilead, Menarini, Merck Sharpe Dohme (MSD), Novartis, Pierre Fabre, Pfizer, Seagen, and Stemline, and research funding from AstraZeneca, Daiichi Sankyo, Novartis, and Pfizer. ND received travel/accommodation expense coverage from Daiichi Sankyo. PJJ has had a consulting or advisory role, and received honoraria, research funding, and/or travel/accommodation expenses from AstraZeneca, Bayer, Boehringer Ingelheim, Novartis, Pfizer, Servier, Roche, Bristol Myers Squibb (BMS), Celgene, Pierre Fabre, Janssen/Johnson & Johnson, MSD, Merck, Sanofi/Aventis, Ipsen, Amgen, Cycuria Therapeutics, and Vessel. GR received honoraria from Amgen, AstraZeneca, Daiichi Sankyo, Eli Lilly, Gilead, MSD, Novartis, Roche, Seagen, Stemline, and BMS; has had a consulting or advisory role from AstraZeneca, Daiichi Sankyo, Eli Lilly, Gilead, MSD, Novartis, Pfizer, Roche, and Stemline, and has received travel support from Amgen, Daiichi Sankyo, Gilead, and Roche. RB received honoraria from Amgen, AstraZeneca, BMS, Daiichi Sankyo, Eisai, Eli Lilly, Gilead, Gruenenthal, MSD, MedMedia, Novartis, Pfizer, Pierre Fabre, Roche, Seagen, and Stemline; travel support from AstraZeneca, Daiichi Sankyo, Eli Lilly, MSD, Novartis, and Stemline; and research support from Daiichi Sankyo. ND received travel/accommodation expense coverage from Stemline. SOH has received speaker fees, unrelated funding, and/or travel/accommodation expenses from AstraZeneca and Roche, has an advisory role at CureMatch, and is a co-founder of Vessel FlexCo. AMS received honoraria for lectures from AstraZeneca and travel and congress registration support from PharmaMar, MSD, Eli Lilly, AstraZeneca, and Stemline-Menarini. EVK received honoraria/travel support from AstraZeneca, Daiichi Sankyo, Eli Lilly, Gilead, Novartis, Roche, Stemline, and Pierre Fabre. EH received unrelated funding from Illumina, Roche, Servier, and PreAnalytiX, and received honoraria from Roche, AstraZeneca, and Incyte for advisory boards, not related to our study. CS has received honoraria from AstraZeneca, Daiichi Sankyo, Eli Lilly, Gilead, Novartis, Pfizer, Pierre Fabre, Roche, and Stemline-Menarini and travel support from Daiichi Sankyo, Eli Lilly, MSD, Novartis, and Roche. All remaining authors have declared no conflicts of interest. Contributor Information N. Dandachi, Email: [email protected]. M. Balic, Email: [email protected]. Supplementary data Supplementary Material mmc1.pdf (3.3MB, pdf) References 1. Bray F., Laversanne M., Sung H., et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2024;74(3):229–263. doi: 10.3322/caac.21834. [ DOI ] [ PubMed ] [ Google Scholar ] 2. Giaquinto A.N., Sung H., Miller K.D., et al. Breast cancer statistics, 2022. CA Cancer J Clin. 2022;72(6):524–541. doi: 10.3322/caac.21754. [ DOI ] [ PubMed ] [ Google Scholar ] 3. Başaran G.A., Twelves C., Diéras V., Cortés J., Awada A. Ongoing unmet needs in treating estrogen receptor-positive/HER2-negative metastatic breast cancer. Cancer Treat Rev. 2018;63:144–155. doi: 10.1016/j.ctrv.2017.12.002. [ DOI ] [ PubMed ] [ Google Scholar ] 4. Ribnikar D., Volovat S.R., Cardoso F. Targeting CDK4/6 pathways and beyond in breast cancer. Breast. 2019;43:8–17. doi: 10.1016/j.breast.2018.10.001. [ DOI ] [ PubMed ] [ Google Scholar ] 5. Tang Y., Li J., Liu B., Ran J., Hu Z.Y., Ouyang Q. Circulating tumor DNA profile and its clinical significance in patients with hormone receptor-positive and HER2-negative mBC. Front Endocrinol (Lausanne) 2022;13 doi: 10.3389/fendo.2022.1075830. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 6. Amato O., Giannopoulou N., Ignatiadis M. Circulating tumor DNA validity and potential uses in metastatic breast cancer. NPJ Breast Cancer. 2024;10(1):21. doi: 10.1038/s41523-024-00626-6. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 7. Dickinson K., Sharma A., Agnihotram R.V., et al. Circulating tumor DNA and survival in metastatic breast cancer: a systematic review and meta-analysis. JAMA Netw Open. 2024;7(9) doi: 10.1001/jamanetworkopen.2024.31722. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 8. Heitzer E., Haque I.S., Roberts C.E.S., Speicher M.R. Current and future perspectives of liquid biopsies in genomics-driven oncology. Nat Rev Genet. 2019;20(2):71–88. doi: 10.1038/s41576-018-0071-5. [ DOI ] [ PubMed ] [ Google Scholar ] 9. Moser T., Kühberger S., Lazzeri I., Vlachos G., Heitzer E. Bridging biological cfDNA features and machine learning approaches. Trends Genet. 2023;39(4):285–307. doi: 10.1016/j.tig.2023.01.004. [ DOI ] [ PubMed ] [ Google Scholar ] 10. Klocker E.V., Hasenleithner S., Bartsch R., et al. Clinical applications of next-generation sequencing-based ctDNA analyses in breast cancer: defining treatment targets and dynamic changes during disease progression. Mol Oncol. 2025;19:1897–1917. doi: 10.1002/1878-0261.13671. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 11. André F., Ciruelos E., Rubovszky G., et al. Alpelisib for PIK3CA-mutated, hormone receptor-positive advanced breast cancer. N Engl J Med. 2019;380(20):1929–1940. doi: 10.1056/NEJMoa1813904. [ DOI ] [ PubMed ] [ Google Scholar ] 12. Turner N.C., Kingston B., Kilburn L.S., et al. Circulating tumour DNA analysis to direct therapy in advanced breast cancer (plasmaMATCH): a multicentre, multicohort, phase 2a, platform trial. Lancet Oncol. 2020;21(10):1296–1308. doi: 10.1016/S1470-2045(20)30444-7. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 13. Turner N.C., Oliveira M., Howell S.J., et al. Capivasertib in hormone receptor-positive advanced breast cancer. N Engl J Med. 2023;388(22):2058–2070. doi: 10.1056/NEJMoa2214131. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 14. Bardia A., Cortés J., Bidard F.C., et al. Elacestrant in ER+, HER2− metastatic breast cancer with ESR1-mutated tumors: subgroup analyses from the phase III EMERALD trial by prior duration of endocrine therapy plus CDK4/6 inhibitor and in clinical subgroups. Clin Cancer Res. 2024;30(19):4299–4309. doi: 10.1158/1078-0432.CCR-24-1073. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 15. Turner N.C., Im S.A., Saura C., et al. Inavolisib-based therapy in PIK3CA-mutated advanced breast cancer. N Engl J Med. 2024;391(17):1584–1596. doi: 10.1056/NEJMoa2404625. [ DOI ] [ PubMed ] [ Google Scholar ] 16. Bidard F.C., Mayer E.L., Park Y.H., et al. First-line camizestrant for emerging ESR1-mutated advanced breast cancer. N Engl J Med. 2025;393(6):569–580. doi: 10.1056/NEJMoa2502929. [ DOI ] [ PubMed ] [ Google Scholar ] 17. Cardoso F., Paluch-Shimon S., Schumacher-Wulf E., et al. 6th and 7th international consensus guidelines for the management of advanced breast cancer (ABC guidelines 6 and 7) Breast. 2024;76 doi: 10.1016/j.breast.2024.103756. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 18. Mamann A., Pradat Y., Bidard F.C., et al. Prognostic significance of early on-treatment evolution of circulating tumor DNA in advanced ER-positive/HER2-negative breast cancer. Ann Oncol. 2025;36:1342–1355. doi: 10.1016/j.annonc.2025.06.015. [ DOI ] [ PubMed ] [ Google Scholar ] 19. Jongbloed E.M., Wortelboer N., de Weerd V., et al. Early versus deferred use of CDK4/6 inhibitors in advanced breast cancer: circulating tumor DNA analysis of a randomized phase 3 trial. Nat Med. 2025;31:3662–3667. doi: 10.1038/s41591-025-03935-w. [ DOI ] [ PubMed ] [ Google Scholar ] 20. Pontolillo L., Davis A.A., Gerratana L., et al. Circulating genomic landscape following cyclin-dependent kinase 4/6 inhibitors exposure in HR+/HER2− metastatic breast cancer: a retrospective multi-institutional consortium analysis. NPJ Breast Cancer. 2025;11(1):93. doi: 10.1038/s41523-025-00802-2. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 21. Fuentes-Antrás J., Elliott M.J., Main S.C., et al. Personalized ctDNA monitoring in metastatic HR+/HER2− breast cancer patients during endocrine and CDK4/6 inhibitor therapy. npj Breast Cancer. 2025;11(1):74. doi: 10.1038/s41523-025-00783-2. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 22. Klocker E.V., Dobrić N., Graf R., et al. Clinical impact of single-gene vs. panel sequencing in advanced HR+/HER2− breast cancer: insights and implications. npj Breast Cancer. 2025;11(1):86. doi: 10.1038/s41523-025-00805-z. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 23. Harris P.A., Taylor R., Minor B.L., et al. The REDCap consortium: building an international community of software platform partners. J Biomed Inform. 2019;95 doi: 10.1016/j.jbi.2019.103208. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 24. Suppan C., Graf R., Jahn S., et al. Sensitive and robust liquid biopsy-based detection of PIK3CA mutations in hormone-receptor-positive metastatic breast cancer patients. Br J Cancer. 2022;126(3):456–463. doi: 10.1038/s41416-021-01601-9. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 25. Chakravarty D., Gao J., Phillips S.M., et al. OncoKB: a precision oncology knowledge base. JCO Precis Oncol. 2017;1:1–16. doi: 10.1200/PO.17.00011. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 26. Belic J., Koch M., Ulz P., et al. Rapid identification of plasma DNA samples with increased ctDNA levels by a modified FAST-SeqS approach. Clin Chem. 2015;61(6):838–849. doi: 10.1373/clinchem.2014.234286. [ DOI ] [ PubMed ] [ Google Scholar ] 27. Suppan C., Brcic I., Tiran V., et al. Untargeted assessment of tumor fractions in plasma for monitoring and prognostication from metastatic breast cancer patients undergoing systemic treatment. Cancers (Basel) 2019;11(8):1171. doi: 10.3390/cancers11081171. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 28. Mayakonda A., Lin D.C., Assenov Y., Plass C., Koeffler H.P. Maftools: efficient and comprehensive analysis of somatic variants in cancer. Genome Res. 2018;28(11):1747–1756. doi: 10.1101/gr.239244.118. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 29. Schemper M., Smith T.L. A note on quantifying follow-up in studies of failure time. Control Clin Trials. 1996;17(4):343–346. doi: 10.1016/0197-2456(96)00075-x. [ DOI ] [ PubMed ] [ Google Scholar ] 30. Uno H., Cai T., Pencina M.J., D'Agostino R.B., Wei L.J. On the C-statistics for evaluating overall adequacy of risk prediction procedures with censored survival data. Stat Med. 2011;30(10):1105–1117. doi: 10.1002/sim.4154. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 31. Akaike H. In: Selected Papers of Hirotugu Akaike. Parzen E., Tanabe K., Kitagawa G., editors. Springer; New York: 1998. Information theory and an extension of the maximum likelihood principle; pp. 199–213. [ Google Scholar ] 32. Yang C., Brezden-Masley C., Joy A.A., et al. Targeting HER2-low in metastatic breast cancer: an evolving treatment paradigm. Ther Adv Med Oncol. 2023;15 doi: 10.1177/17588359231175440. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 33. O’Leary B., Cutts R.J., Huang X., et al. Circulating tumor DNA markers for early progression on fulvestrant with or without palbociclib in ER+ advanced breast cancer. J Natl Cancer Inst. 2020;113(3):309–317. doi: 10.1093/jnci/djaa087. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 34. Fuentes-Antrás J., Martínez-Rodríguez A., Guevara-Hoyer K., et al. Real-world use of highly sensitive liquid biopsy monitoring in metastatic breast cancer patients treated with endocrine agents after exposure to aromatase inhibitors. Int J Mol Sci. 2023;24(14) doi: 10.3390/ijms241411419. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 35. Chiu J., Su F., Joshi M., et al. Potential value of ctDNA monitoring in metastatic HR+/HER2 - breast cancer: longitudinal ctDNA analysis in the phase Ib MONALEESASIA trial. BMC Med. 2023;21(1):306. doi: 10.1186/s12916-023-03017-z. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 36. Gerratana L., Davis A.A., Velimirovic M., et al. Cyclin-dependent kinase 4/6 inhibitors beyond progression in metastatic breast cancer: a retrospective real-world biomarker analysis. JCO Precis Oncol. 2023;7 doi: 10.1200/PO.22.00531. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 37. Lefebvre C., Bachelot T., Filleron T., et al. Mutational profile of metastatic breast cancers: a retrospective analysis. PLoS Med. 2016;13(12) doi: 10.1371/journal.pmed.1002201. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 38. Huebner H., Wimberger P., Laakmann E., et al. Cell-free tumor DNA analysis in advanced or metastatic breast cancer patients: mutation frequencies, testing intention, and clinical impact. Precis Clin Med. 2025;8(1) doi: 10.1093/pcmedi/pbae034. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 39. Valenza C., Curigliano G. Positioning oral selective estrogen receptor degraders in patients with metastatic breast cancer. Eur J Cancer. 2025;228 doi: 10.1016/j.ejca.2025.115739. [ DOI ] [ PubMed ] [ Google Scholar ] 40. Davis A.A., Luo J., Zheng T., et al. Genomic complexity predicts resistance to endocrine therapy and CDK4/6 inhibition in hormone receptor-positive (HR+)/HER2-negative metastatic breast cancer. Clin Cancer Res. 2023;29(9):1719–1729. doi: 10.1158/1078-0432.CCR-22-2177. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 41. Bertucci F., Ng C.K.Y., Patsouris A., et al. Genomic characterization of metastatic breast cancers. Nature. 2019;569(7757):560–564. doi: 10.1038/s41586-019-1056-z. [ DOI ] [ PubMed ] [ Google Scholar ] 42. Li Z., Spoelstra N.S., Sikora M.J., et al. Mutual exclusivity of ESR1 and TP53 mutations in endocrine resistant metastatic breast cancer. npj Breast Cancer. 2022;8(1):62. doi: 10.1038/s41523-022-00426-w. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 43. Williams M.M., Spoelstra N.S., Arnesen S., et al. Steroid hormone receptor and infiltrating immune cell status reveals therapeutic vulnerabilities of ESR1-mutant breast cancer. Cancer Res. 2021;81(3):732–746. doi: 10.1158/0008-5472.CAN-20-1200. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 44. Boscolo Bielo L., Guerini Rocco E., Trapani D., et al. Genomic and clinical landscape of metastatic hormone receptors-positive breast cancers carrying ESR1 alterations. ESMO Open. 2024;9(10) doi: 10.1016/j.esmoop.2024.103731. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 45. Guigal-Stephan N., Lockhart B., Moser T., Heitzer E. A perspective review on the systematic implementation of ctDNA in phase I clinical trial drug development. J Exp Clin Cancer Res. 2025;44(1):79. doi: 10.1186/s13046-025-03328-4. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 46. Reichert Z.R., Morgan T.M., Li G., et al. Prognostic value of plasma circulating tumor DNA fraction across four common cancer types: a real-world outcomes study. Ann Oncol. 2023;34(1):111–120. doi: 10.1016/j.annonc.2022.09.163. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 47. Bidard F.C., Hardy-Bessard A.C., Dalenc F., et al. Switch to fulvestrant and palbociclib versus no switch in advanced breast cancer with rising ESR1 mutation during aromatase inhibitor and palbociclib therapy (PADA-1): a randomised, open-label, multicentre, phase 3 trial. Lancet Oncol. 2022;23(11):1367–1377. doi: 10.1016/S1470-2045(22)00555-1. [ DOI ] [ PubMed ] [ Google Scholar ] 48. Wortelboer N., Van Ommen-Nijhof A., Konings I.R., et al. 487MO Overall survival with first versus second-line use of CDK4/6 inhibitors in HR+/HER2− advanced breast cancer. Ann Oncol. 2025;36(suppl 2):S379–S380. [ Google Scholar ] 49. Razavi P., Li B.T., Brown D.N., et al. High-intensity sequencing reveals the sources of plasma circulating cell-free DNA variants. Nat Med. 2019;25(12):1928–1937. doi: 10.1038/s41591-019-0652-7. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 50. Sokol E.S., Feng Y.X., Jin D.X., et al. Loss of function of NF1 is a mechanism of acquired resistance to endocrine therapy in lobular breast cancer. Ann Oncol. 2019;30(1):115–123. doi: 10.1093/annonc/mdy497. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 51. Turner N., Pearson A., Sharpe R., et al. FGFR1 amplification drives endocrine therapy resistance and is a therapeutic target in breast cancer. Cancer Res. 2010;70(5):2085–2094. doi: 10.1158/0008-5472.CAN-09-3746. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 52. Sablin M.P., Gestraud P., Jonas S.F., et al. Copy number alterations in metastatic and early breast tumours: prognostic and acquired biomarkers of resistance to CDK4/6 inhibitors. Br J Cancer. 2024;131(6):1060–1067. doi: 10.1038/s41416-024-02804-6. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Supplementary Materials Supplementary Material mmc1.pdf (3.3MB, pdf) ACTIONS View on publisher site PDF (908.2 KB) Cite Collections Permalink PERMALINK Copy RESOURCES Similar articles Cited by other articles Links to NCBI Databases Cite Copy Download .nbib .nbib Format: AMA APA MLA NLM Add to Collections Create a new collection Add to an existing collection Name your collection * Choose a collection Unable to load your collection due to an error Please try again Add Cancel Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top