Serial ctDNA Analysis for Monitoring Molecular Response and Acquired Resistance in Metastatic NSCLC and HR+/HER2‐Negative Breast Cancer: A Real‐World Study

ABSTRACT

Circulating tumor DNA (ctDNA) is a promising biomarker for disease monitoring, yet evidence of its clinical utility in routine care remains limited. We evaluated the utility of serial ctDNA monitoring using digital PCR (ddPCR) for detecting acquired resistance (AR) and monitoring response in metastatic non‐small cell lung cancer (NSCLC) and breast cancer (MBC). This real‐world, longitudinal study included 48 metastatic patients (42 with EGFR‐mutant NSCLC; 6 with HR+/HER2‐negative MBC) treated at a Brazilian oncology reference center. Serial plasma samples were analyzed by ddPCR for activating and resistance mutations at specific time points. In the NSCLC cohort, EGFR T790M was detected in 48.4% of patients who progressed on 1st/2nd‐generation TKIs. For patients on 3rd‐generation TKIs, EGFR C797S was the main AR mechanism (47%), followed by PIK3CA, BRAF mutations, and ERBB2/EGFR amplifications. Longitudinal analysis anticipated radiological progression by approximately 5 months. Early molecular response, characterized by mutation clearance, preceded radiological response. In the MBC cohort, hotspot ESR1 mutations (D538G, L536R, Y537S) were identified in 50% of patients. Variations in ctDNA fractional abundance predicted disease progression up to 9 months before radiological confirmation. This study provides real‐world evidence that ddPCR‐based ctDNA monitoring enables early detection of resistance and predicts disease progression significantly earlier than standard imaging in both NSCLC and MBC. Despite the limited sample size, particularly in the MBC cohort, these real‐world findings support liquid biopsy as a minimally invasive tool with the potential to guide therapeutic decisions and inform personalized management in precision oncology. Larger prospective studies are needed to confirm these observations.

Abbreviations

AI

aromatase inhibitor

AR

acquired resistance

ASCO

American Society of Clinical Oncology

CDK4/6

cyclin‐dependent kinase 4/6

cfDNA

circulating cell‐free DNA

ctDNA

circulating tumor DNA

ddPCR

droplet digital PCR

DP

disease progression

EGFR

epidermal growth factor receptor

EGFR‐act mutation

EGFR‐activating mutation

ER

estrogen receptor

ET

endocrine therapy

FA

fractional abundance

HR+

hormone receptor‐positive

IASLC

International Association for the Study of Lung Cancer

MBC

metastatic breast cancer

NGS

next‐generation sequencing

NSCLC

non‐small cell lung cancer

OS

overall survival

PFS

progression‐free survival

RNAseP

Ribonuclease P

SCLC

small cell lung cancer

SERDs

selective estrogen receptor degraders

TKI

tyrosine kinase inhibitor

1. Introduction

The detection of circulating tumor DNA (ctDNA) in plasma has become a promising biomarker for initial tumor genotyping, monitoring molecular disease response, and predicting and characterizing acquired resistance (AR) in the era of precision oncology [1, 2, 3]. While several studies have confirmed the clinical validity of ctDNA monitoring, real‐world studies are needed to complement clinical trial data by validating clinical utility across diverse patient populations and the complexities of routine care [3, 4, 5, 6, 7, 8, 9, 10]. Although tissue biopsy remains the gold standard for tumor molecular analysis, it is associated with clinical complications, sampling difficulties, and limited capture of tumor heterogeneity. Liquid biopsy, by contrast, is minimally invasive and captures spatial and temporal heterogeneity with high sensitivity [1, 2]. Compared with next‐generation sequencing, digital PCR offers a faster, lower‐cost approach for monitoring well‐characterized, recurrent alterations, such as those driving resistance in lung and breast cancer.

In non‐small cell lung cancer (NSCLC), the epidermal growth factor receptor (EGFR) is one of the most common driver genes; its prevalence varies markedly by ethnicity and geographic region, occurring in approximately 47% of NSCLC adenocarcinomas in Asian populations, compared with approximately 10%–15% in Western/Caucasian populations [11, 12]. Patients with EGFR‐driven NSCLC are eligible for targeted therapies that have revolutionized the treatment of advanced EGFR‐mutated NSCLC [13, 14, 15, 16]. Several clinical trials have shown improved progression‐free survival (PFS) and/or overall survival (OS) with 1st‐generation (gefitinib and erlotinib), 2nd‐generation (afatinib), and 3rd‐generation (osimertinib) EGFR TKIs (Tyrosine kinase inhibitors) compared to chemotherapy [14, 17, 18, 19, 20]. However, tumors inevitably develop resistance to TKI therapies, driven by secondary mutations in the EGFR gene in nearly 50% of cases [21]. Additional resistance mechanisms to 1st‐ and 2nd‐generation TKIs are also present at a lower frequency, including MET and ERBB2 amplification, PIK3CA mutations, BRAF V600E, and histologic transformation from NSCLC to SCLC (small cell lung cancer) [22, 23, 24]. The heterogeneity of resistance mechanisms to osimertinib includes EGFR and MET amplification, EGFR C797S mutation, off‐target mutations in PIK3CA, KRAS, and ERBB2, ROS1 fusions, as well as histologic transformation [25, 26].

Molecular assessment of the tumor is crucial at disease progression to guide personalized treatment strategies in NSCLC patients. The increasing number of NSCLC oncogene alterations identified and targeted therapies developed have enabled ctDNA analysis as a complementary tool to tissue genotyping in treatment decisions.

The International Association for the Study of Lung Cancer (IASLC) updated its statement on liquid biopsy in 2021 [27]. For treatment‐naive patients with advanced or metastatic NSCLC, the IASLC recommends using liquid biopsy alongside tissue biopsy when, even if a tissue sample is available for initial tumor genotyping, the analysis is incomplete, the sample is inadequate, or scarce. If no tissue sample is available for genotyping, the IASLC advises a “plasma first” approach. However, in cases of disease progression (DP) after targeted therapy, the IASLC suggests considering a “plasma first” approach to evaluate AR mechanisms. If the analysis provides uninformative results, a tissue rebiopsy and analysis are then recommended.

For patients with hormone receptor‐positive (HR+), human epidermal growth factor receptor 2 (HER2)‐negative locally advanced or metastatic breast cancer (MBC), cyclin‐dependent kinase 4/6 (CDK4/6) inhibitors combined with endocrine therapy (ET), either aromatase inhibitors (AI) or fulvestrant, are the recommended first‐line standard treatment [28, 29], although resistance is nearly universal and multiple post‐CDK4/6i treatment strategies are under active investigation [30]. Activating mutations in ESR1, which encodes the estrogen receptor alpha protein, are a major mechanism of acquired resistance to ET, affecting 20%–40% of MBC patients [31], and have led to the development of oral selective estrogen receptor degraders (SERDs) such as elacestrant and camizestrant [32, 33]. Based on the EMERALD trial, in which ESR1 mutation status was assessed by ctDNA, the 2023 ASCO guideline recommends ESR1 mutation testing—preferably by ctDNA—to guide therapy at recurrence or progression on ET [34], and serial ctDNA monitoring for emergent ESR1 mutations, as demonstrated in the PADA‐1 trial, may further guide treatment switching [35]. Most recently, the phase 3 SERENA‐6 trial demonstrated that switching to camizestrant upon ctDNA detection of an emerging ESR1 mutation—before clinical or radiological progression—while continuing CDK4/6i, improved progression‐free survival in the first‐line setting [36], directly supporting the rationale for serial ctDNA monitoring explored in the present study.

Here, we conducted a real‐world study to evaluate the clinical utility of serial ctDNA monitoring by droplet digital PCR (ddPCR) for early detection of acquired resistance and molecular treatment response, and to characterize the genetic alterations associated with therapy resistance in patients with metastatic NSCLC and HR+, HER2‐negative MBC treated at Hospital Sírio‐Libanês, a leading private oncology referral center in Brazil.

2. Methods

2.1. Patients

Patients were enrolled from all individuals with metastatic EGFR‐mutant NSCLC starting EGFR‐TKI therapy (NSCLC Cohort), or HR+/HER2‐negative MBC starting ET ± iCDK4/6 (MBC Cohort), at our institution who provided written informed consent to serial research blood collection under an IRB‐approved translational research study (HSL2015‐22). ctDNA monitoring was investigational and did not guide real‐time clinical decision‐making, allowing an unbiased retrospective assessment of its predictive value against the standard‐of‐care radiological follow‐up. Inclusion criteria: Patients aged ≥ 18 years with histologically confirmed metastatic NSCLC harboring a documented EGFR‐activating mutation who were initiating treatment with an EGFR‐TKI, or with histologically confirmed HR+/HER2‐negative metastatic breast cancer initiating ET alone or with an iCDK4/6; written informed consent for serial blood collection and genomic analysis. Exclusion criteria: failure to consent to study participation. All NSCLC patients had a histologically confirmed diagnosis of lung adenocarcinoma. Baseline EGFR‐activating mutations, determined by tissue genotyping at diagnosis, included EGFR exon 19 deletion, EGFR exon 18 mutations, EGFR exon 20 mutations, and EGFR exon 21 mutations, and were the basis for eligibility for first‐line or subsequent EGFR‐TKI therapy. Serial blood samples from both cohorts were drawn throughout the treatment period from December 2016 to March 2021, and for most patients, samples were collected monthly. Radiologic progressive disease was defined according to RECIST v1.1 criteria [37].

2.2. Plasma Processing and Circulating Cell‐Free DNA Extraction

Peripheral blood samples (20 mL) were collected into tubes containing EDTA, and plasma was separated from blood within 2 h as previously described [38]. Circulating cell‐free DNA (cfDNA) was isolated using the QIAamp MinElute Virus Vacuum Kit (Qiagen, Hilden, Germany) and stored at −80°C until needed.

2.3. Droplet Digital PCR (ddPCR)

cfDNA was quantified by ddPCR using the RNAseP (Ribonuclease P) Copy Number Reference Assay (Thermo Fisher Scientific, Carlsbad, CA, USA). A total of 3000 genome equivalents (~10 ng of cfDNA) and 1500 genome equivalents (~5 ng of cfDNA) were analyzed for plasma samples from lung and breast cancer patients, respectively, to achieve a detection sensitivity of 0.2%. Probes and primers for EGFR activating mutation (EGFR deletions (ddPCR Mutation Assay EGFR p.L747_A750: dHsaMDV2516748; ddPCR Mutation Assay EGFR p.L747_S752: dHsaMDS408256419; ddPCR Mutation Assay EGFR E746_A750: #10041170; ddPCR EGFR Exon 19 Deletions Screening: #12002392), EGFR G719A (ddPCR Mutation Assay: dHsaMDS2512974), EGFR S768I (ddPCR Mutation Assay: dHsaMDV2516892) and EGFR L858R (ddPCR Mutation Assay: #10040783)), resistance mutations EGFR T790M (ddPCR Mutation Assay ID: dHsaMDV2010019), PIK3CA E545K (ddPCR Mutation Assay: #10041188), PIK3CA E542K (ddPCR Mutation Assay: #10041187) and BRAF V600E (ddPCR Mutation Assay: #10040779) and gene amplifications (MET (PrimePCR ddPCR Copy Number Assay: dHsaCP2500321), EGFR (PrimePCR ddPCR Copy Number Assay: dHsaCP2500318) and ERBB2 (PrimePCR ddPCR Copy Number Assay: dHsaCP1000116)) were obtained from Bio‐Rad (Hercules, CA, USA). Additionally, primers and probes for EGFR exon 20 insertion (EGFR p.Asp770_Asn771insAlaSerValAsp) and EGFR C797S resistance mutations, which occur at different nucleotides within the codon for the same amino acid 797, were manually designed (EGFR exon 20 insertion: Primer Forward‐5′ AAATCCTCGATGAAGCCTA 3′; Primer Reverse‐5′ GGAGGTGAGGCAGATG 3′; Probe Wild Type‐HEX/BHQ‐CAGCGTGGACAACCC; Probe Insertion‐6FAM/BHQ‐CCAGCGTGGACAACC; EGFR C797S: Primer Forward‐5′ GCCTGCTGGGCATCTG 3′; Primer Reverse‐5′ TCTTTGTGTTCCCGGACATAGTC 3′; Probe Wild Type‐HEX/BHQ‐TTCGGCTGCCTCCTG; Probe Mutation T>A‐6FAM/BHQ‐TTCGGCAGCCTCC; Probe Mutation G>C‐6FAM/BHQ‐CTTCGGCTCCCTCCTG) and obtained from Integrated DNA Technologies (IDT, Coralville, IA, USA). Primers and probes for the detection of ESR1 mutations (L536R, Y537C/N/S, and D538G) were designed by Schiavon et al. [39] and obtained from IDT (Coralville, IA, USA). ddPCR was performed on a QX200 Droplet Digital PCR System (Bio‐Rad, Hercules, CA, USA) according to the manufacturer's instructions, and data were analyzed using the QuantaSoft Software v.1.7.4.0917 (Bio‐Rad, Hercules, CA, USA). ctDNA quantification is presented as fractional abundance (FA—the percentage of the mutant allele in total cfDNA).

2.4. ctDNA Sequencing (ctDNAseq)

Targeted capture libraries were constructed from 25 ng of cfDNA using the SureSelect XT HS Target Enrichment System (Agilent Technologies, Santa Clara, CA, USA) and a panel of 151 cancer genes (ClearSeq Comprehensive Cancer Panel; Agilent Technologies, Santa Clara, CA, USA) according to the manufacturer's instructions, including modifications proposed by Mansukhani et al. [40]. Sequences were generated on a NextSeq 500 sequencing platform (Illumina, San Diego, CA, USA).

2.5. Variant Calling

The 150 bp paired‐end sequences were analyzed using the SureCall software (version 4.0.1.46; Agilent Technologies, Santa Clara, CA, USA) and following modifications proposed by Mansukhani et al. [40]. Variant calls were identified using the SNPPET function from SureCall, and DuplexCaller was used to identify mutations supported by duplex reads [40]. Germline variants were filtered as described elsewhere [41].

2.6. Statistical Analysis

This was a descriptive, real‐world cohort study. Continuous variables (e.g., patient age, time‐to‐progression, lead time between molecular and radiological progression) are summarized as medians with ranges or means, as indicated in the text; categorical variables are summarized as counts and percentages. The denominator (n) for each proportion is stated alongside the result (e.g., the number of evaluable patients for a given mutation or timepoint). The lead time between ctDNA‐based detection of molecular progression/resistance and RECIST‐defined radiological progression was calculated as the mean number of days between the two events for each applicable patient subgroup. All descriptive analyses and figures were generated using Microsoft Excel, GraphPad Prism (version 10.6.1, Build 799), and custom Python scripts (version 3.12.13) utilizing the Pandas (version 2.2.2) and Matplotlib (version 3.10.0) libraries.

3. Results

3.1. Patient Characteristics and Serial ctDNA Analysis

A total of 48 patients were included, comprising 42 with metastatic NSCLC and six with HR+, HER2‐negative MBC, as detailed in Tables 1 and 2, respectively. The median age of metastatic NSCLC patients was 57 years (range: 29–81). Among these, 59.5% (25/42) were female, and 71.4% (30/42) were never smokers. Thirty‐two patients were enrolled during treatment with 1st‐ and 2nd‐generation TKIs (NSCLC Cohort 1) and 10 during therapy with 3rd‐generation TKI (NSCLC Cohort 2). For the MBC cohort, the median age at initiation of endocrine therapy was 58.5 years (range: 48–71). All six patients were postmenopausal, and 83.3% (5/6) were enrolled during combined treatment of endocrine agents and iCDK4/6.

TABLE 1.

Clinical characteristics of metastatic NSCLC patients at baseline.

Characteristics	Cohort 1 (1st/2nd‐gen EGFR TKI)	Cohort 2 (3rd‐gen EGFR TKI)	All patients
N (%)	32 (76.2)	10 (23.8)	42 (100)
Gender
Female, n (%)	21 (65.6)	4 (40)	25 (59.5)
Male, n (%)	11 (34.4)	6 (60)	17 (40.5)
Median age (range)	59 (34–81)	54 (29–73)	57 (29–81)
Smoking status
Never, n (%)	22 (68.8)	7 (70)	29 (69)
Ever or current smoker, n (%)	10 (31.2)	3 (30)	13 (31)
Stage
IV, n (%)	30 (93.8)	8 (80)	38 (90.5)
Other, n (%)	2 (6.2)	2 (20)	4 (9.5)
Line of first TKI treatment
1, n (%)	26 (81.3)	2 (20)	28 (66.6)
2, n (%)	5 (15.6)	5 (50)	10 (23.8)
3, n (%)	0 (0)	2 (20)	2 (4.8)
≥ 4, n (%)	1 (3.1)	1 (10)	2 (4.8)
First TKI treatment
Erlotinib, n (%)	25 (78.1)		25 (59.5)
Afatinib, n (%)	6 (18.7)		6 (14.3)
Gefitinib, n (%)	1 (3.2)		1 (2.4)
Osimertinib, n (%)		10 (100)	10 (23.8)

TABLE 2.

Clinical characteristics of MBC patients at baseline.

Characteristics	All patients
N (%)	6 (100)
Median age at endocrine therapy (ET) start (range)	58.5 (48–71)
Primary tumor histology
Ductal, n (%)	4 (66.7)
Lobular, n (%)	2 (33.3)
Metastatic sites at ET start
Bone only, n (%)	1 (16.6)
Visceral with bone, n (%)	5 (83.4)
Prior lines of treatment in advanced or metastatic setting
Median (range)	3 (1–8)
No. of lines
1, n (%)	2 (33.3)
2, n (%)	0 (0)
≥ 3, n (%)	4 (66.7)
ET at ESR1 mutation analysis
AI, n (%)	4 (66.7)
Tamoxifen, n (%)	1 (16.6)
Fulvestrant, n (%)	1 (16.6)
CDK4/6 inhibitor
Palbociclib, n (%)	4 (66.7)
Ribociclib, n (%)	1 (16.6)

Patient enrollment was successful overall, and monthly blood collection was performed with virtually no difficulties or complications. A variation in cfDNA yield was observed between tumor types, with lung cancer samples showing higher cfDNA concentrations than breast cancer samples. Accordingly, 3000 and 1500 genome equivalents were analyzed by ddPCR for lung and breast samples, respectively; however, this difference did not affect the accuracy or reliability of the results.

In the NSCLC cohorts, previously tumor tissue analysis for EGFR‐activating (EGFR‐act) mutations at the time of disease diagnosis showed that 71.4% (30/42) had EGFR exon 19 deletions, 16.7% (7/42) had EGFR exon 21 mutations, 4.8% (2/42) had EGFR exon 18 mutations, 2.4% (1/42) had EGFR exon 20 mutations, 2.4% (1/42) had two EGFR‐act mutations (EGFR G719S and EGFR S768I), and for 2.4% (1/42) information about the specific EGFR‐act mutation was not available. Among the 42 patients, 36 had DP and a follow‐up of 39.8 months at our institution. A total of 385 plasma samples were collected and activating and resistance mutations to EGFR TKIs were analyzed by ddPCR, averaging approximately nine samples per patient. The patient with the most collections had 48 blood samples over 5 years. Patients who benefited from 1st‐ and 2nd‐generation TKIs had an average survival of 19.9 months following metastatic diagnosis, whereas those who responded to 3rd‐generation TKIs had an average survival of 18.2 months post‐metastatic diagnosis.

For the MBC cohort, 126 plasma samples were collected, and hotspot ESR1 mutations were analyzed using ddPCR, yielding an average of 21 samples per patient, with a maximum of 46 for one patient over 5 years and 6 months. Overall, patients had an average time to DP of 16 months after initiation of endocrine therapy and liquid biopsy monitoring.

3.2. Effectiveness of Liquid Biopsy for Early Detection and Characterization of Acquired Resistance to Treatment

3.2.1. NSCLC Cohorts

Disease progression was observed in 29 of 32 patients (90.6%) from the NSCLC Cohort 1. Acquired resistance EGFR T790M mutation was detected in 14 out of 29 patients (48.4%) at the time of progression. BRAF V600E mutation was found in 1 patient (3.4%), MET amplification in 1 patient (3.4%), and EGFR amplification plus ERBB2 amplification in 1 patient (3.4%). Additionally, transformation from NSCLC to SCLC (small cell lung cancer) was histologically confirmed in 1 patient (3.4%). We were unable to identify the molecular mechanisms of acquired resistance in 11 patients (38%) who progressed to 1st‐ and 2nd‐generation TKIs using liquid biopsy and ddPCR (Figure 1).

FIGURE 1.

Acquired resistance mutations in metastatic NSCLC patients who have progressed to 1st‐ and 2nd‐generation TKIs. Amp, amplification; SCLC, small cell lung cancer.

Retrospective analysis was performed on serial plasma samples from 18 patients who experienced progression during treatment with 1st‐ and 2nd‐generation TKIs and were EGFR T790M and/or EGFR‐act positive at the time of DP to assess the usefulness of liquid biopsies for early detection of AR (Figure 2). The fractional abundance of EGFR‐act and EGFR T790M mutations was determined by ddPCR throughout treatment until disease progression. In 58.3% (7/12) of patients, monitoring of the EGFR T790M mutation revealed increases in fractional abundance that predicted progression, on average, 167 days before radiological disease progression. Similarly, in 56.2% (9/16) of patients, tracking the EGFR‐act mutation indicated disease progression an average of 127.5 days before radiological disease progression.

FIGURE 2.

Early disease progression and resistance detection to 1st‐ and 2nd‐generation TKIs in plasma from metastatic NSCLC patients. Swimmer plot illustrating treatment duration until disease progression, indicated by an “X.” Each horizontal bar represents a patient, with segments indicating different TKI therapies. Circles denote the time points at which each mutation was detected, with mutation types distinguished by circle color as shown in the legend.

Of the 14 patients who progressed on 1st‐ and 2nd‐generation TKIs and were EGFR T790M‐positive, 11 (78.6%) received osimertinib and were further monitored using serial ctDNA analysis. Four additional patients, who were EGFR T790M‐negative and had progressed on these TKIs, also received osimertinib. In addition to these 15 patients, another 10 patients were included in our study during 3rd‐generation TKI treatment (Cohort 2). Of these 25 patients treated with 3rd‐generation TKI, 68% (17/25) experienced progression, and in 13 of them (76.5%), AR mechanisms were identified using liquid biopsies. A total of 8/17 (47%) patients were positive for EGFR C797S. PIK3CA E542K was detected in 1 patient (5.9%), ERBB2 amplification was found in 1 patient (5.9%), and two patients harbored more than one genetic alteration associated with AR: EGFR C797S plus EGFR amplification (5.9%), and BRAF V600E mutation, plus PIK3CA E545K mutation, plus EGFR amplification (5.9%). Additionally, transformation to SCLC was histologically confirmed plus EGFR amplification (5.9%) (Figure 3).

FIGURE 3.

Acquired resistance mutations in metastatic NSCLC patients who progressed to 3rd‐generation TKI. Amp, amplification; SCLC, small cell lung cancer.

Serial plasma samples from 13 patients who progressed during 3rd‐generation TKI therapy and had liquid biopsies positive for AR and EGFR‐act mutations at the time of progression were also retrospectively analyzed to evaluate the utility of liquid biopsies for early detection of AR in patients receiving 3rd‐generation TKI (Figure 4). An increase in the fractional abundance of EGFR‐act predicted clinical DP by an average of 224.4 days in 92.3% (12/13) of patients, while the detection of the EGFR‐C797S mutated allele predicted clinical progression by an average of 62.2 days in 44.4% (4/9) patients. When considering all resistance mutations detected in all 13 patients, radiological progression was anticipated on average 145.9 days using liquid biopsies.

FIGURE 4.

Early disease progression and resistance detection to 3rd‐generation TKI by liquid biopsies in metastatic NSCLC patients. Swimmer plot illustrating treatment duration until disease progression, indicated by an “X.” Each horizontal bar represents a patient, with segments indicating different TKI therapies. Circles denote the time points at which each mutation was detected, with mutation types distinguished by circle color as shown in the legend.

Effectiveness of liquid biopsy in molecular response monitoring was also performed and evaluated only for patients treated with 3rd‐generation TKIs, as those treated with 1st‐ and 2nd‐generation TKIs were enrolled in the study under TKI therapy. Monitoring involved detecting EGFR‐act and EGFR‐T790M mutations in 11 patients who progressed on 1st and 2nd‐generation TKIs and were positive for EGFR‐T790M in ctDNA. These patients received osimertinib and continued to be monitored through longitudinal plasma collections. Nine out of 11 (81.8%) patients who started osimertinib treatment had undetectable levels of the EGFR‐T790M mutant allele in plasma, on average 37.8 days after starting treatment, while EGFR‐act was negative in plasma on average 88 days after initiation of therapy, indicating a good molecular response that was later confirmed by radiological images. Monitoring of molecular response and prediction of resistance to TKI therapy through activating and resistance mutations in the ctDNA of four patients from the NSCLC cohort is illustrated in Figure 5.

FIGURE 5.

Longitudinal monitoring of EGFR‐act and resistance mutations to TKI therapy in ctDNA from patients in the NSCLC cohort. Plasma ctDNA levels are shown as the fractional abundance (FA) of each mutation detected by ddPCR. (A) Patient 6; (B) Patient 7; (C) Patient 11; (D) Patient 14. The red line indicates the limit of detection (0.2%) with FA values above this threshold considered positive for ctDNA detection. DP, disease progression. SBRT, stereotactic body radiation therapy. Chemo, chemotherapy.

3.2.2. MBC Cohort

Disease progression was observed in all six patients (100%) in the MBC Cohort, and hotspot ESR1 mutations associated with acquired resistance to ET were detected in 3 of the 6 patients (50%) (Figure 6). One patient (Pt02) had an ESR1 D538G mutation, and another (Pt03) had an ESR1 L536R mutation detected by ddPCR at the time of progression during treatment with ET combined with iCDK4/6. No ESR1 mutation was detected by ddPCR at disease progression during ET plus iCDK4/6 in the third patient (Pt01). ctDNAseq was performed on the same sample, but no actionable alterations were identified. Subsequently, ctDNAseq was repeated for this patient using a sample collected at progression during fulvestrant therapy, revealing the presence of ESR1 Y537S mutation.

FIGURE 6.

Early disease progression and resistance detection to ET in plasma from MBC patients. Swimmer plot illustrating treatment duration until disease progression, indicated by an “X.” Each horizontal bar represents a patient, with segments indicating different endocrine therapies. Circles denote the time points at which ESR1 mutations were detected, with mutation types distinguished by circle color as shown in the legend.

The presence of these mutations was subsequently monitored by ddPCR throughout the treatment course, as shown in Figure 7. Changes in fractional abundance predicted progression, on average, 9 months before clinical progression.

FIGURE 7.

Longitudinal monitoring of ESR1 mutations in ctDNA from patients in the MBC cohort. Plasma ctDNA levels are shown as the fractional abundance (FA) of ESR1 mutations detected by ddPCR: (A) Patient 01—ESR1 Y537S; (B) Patient 02—ESR1 D538G; (C) Patient 03—ESR1 L536R. The red line indicates the limit of detection (0.2%) with FA values above this threshold considered positive for ctDNA detection. DP, disease progression; Dtx, Docetaxel; Eve, Everolimus; Ex, Exemestane; Letro, Letrozole; Nvb, Vinorelbine; Palb, Palbociclib.

4. Discussion

This real‐world study demonstrates the feasibility and clinical applicability of plasma ctDNA monitoring in patients with metastatic NSCLC and HR+, HER2‐negative MBC, emphasizing the value of liquid biopsies as a minimally invasive, dynamic tool to complement standard imaging and tissue‐based methods. By repeatedly analyzing plasma samples using ddPCR, it was possible to identify molecular mechanisms of AR, predict DP before clinical progression and radiological confirmation, and assess molecular responses to targeted therapies.

In the NSCLC cohorts, our findings support previous reports that ctDNA provides clinically relevant insights into acquired resistance during TKI therapy [42, 43, 44]. EGFR T790M was the most common emerging resistance mutation in patients progressing on 1st/2nd‐generation TKIs (48.4%), and EGFR C797S predominated after 3rd‐generation TKI (47%), consistent with prior series [19, 23, 44, 45]. Longitudinal ctDNA analyses predicted DP, with increased EGFR‐act and EGFR T790M fractional abundance detected about 5 months before clinical and radiographic confirmation, underscoring ctDNA's potential as an early predictor of 1st/2nd‐generation TKI treatment failure. Additionally, after osimertinib initiation, serial ctDNA monitoring proved valuable for tracking tumor dynamics, with the quantification of EGFR‐act mutations providing a more sensitive and informative marker of disease burden than the detection of specific resistance alleles. As truncal mutations, EGFR‐act variants reflect the clonal architecture of the tumor and may therefore offer a more comprehensive view of disease evolution. It is noteworthy that, using the same approach, we previously identified amplification of an EGFR‐act mutation (EGFR exon 19 deletion) as a novel mechanism of resistance to osimertinib [38], further highlighting the versatility of ctDNA analysis in uncovering resistance pathways.

Additionally, in the osimertinib‐treated group (Cohort 2), clearance of EGFR‐act and EGFR T790M mutations shortly after therapy initiation anticipated subsequent radiological responses, confirming ctDNA monitoring as a surrogate marker of treatment effectiveness. This highlights the value of tracking both activating and resistance mutations for improved predictive accuracy.

Our results are consistent with previous studies linking ctDNA dynamics to clinical outcomes. Of note, retrospective longitudinal ctDNA analysis by ddPCR in plasma samples from the FLAURA [46] and AURA [19] phase 3 trials showed that EGFR mutation ctDNA monitoring could detect DP about 3 months before clinical and RECIST‐defined radiologic progression. Additionally, ctDNA clearance after 3–6 weeks of treatment was associated with improved outcomes [44].

In the MBC cohort, ESR1 mutations associated with endocrine resistance emerged in half of patients progressing under ET plus iCDK4/6 therapy. Studies have shown that iCDK4/6 s do not prevent the appearance of ESR1 mutations. However, these mutations do not confer resistance to iCDK4/6s [47]. Our data demonstrate that ESR1 mutations can be tracked to monitor dynamic changes in ctDNA during therapy, offering a more precise method for detecting treatment resistance. ctDNA monitoring revealed that shifts in the fractional abundance of ESR1 mutant alleles predicted clinical progression by nearly 9 months, aligning with previous studies that reported a median lead time of about 3–7 months [31, 48]. This observation supports the hypothesis that early detection of ESR1 mutations could guide therapeutic decisions, as shown in the PADA‐1 trial [35], where switching therapy upon detection of emerging ESR1 mutations improved outcomes. More recently, the phase 3 SERENA‐6 trial provided evidence supporting a ctDNA‐guided approach to detect ET resistance and adjust therapy before clinical and radiological DP in HR+, HER2‐negative advanced breast cancer [36]. Switching from ET to camizestrant, while continuing iCDK4/6 therapy based on liquid biopsy detection of ESR1 mutations before clinical progression, resulted in improved outcomes.

It is noteworthy that the ctDNA analysis of one of the MBC patients revealed the presence of ESR1 Y537S mutation, which is well‐established as a negative predictive biomarker of fulvestrant response [31]. Notably, this mutation was not detected at disease progression on ET plus iCDK4/6 and was only detected afterward during fulvestrant treatment. This case highlights the importance of one of the recommendations in ASCO guidelines: retesting ESR1 wild‐type patients at subsequent progression [34]. Furthermore, a rapid disease progression was observed during treatment with exemestane (AI) plus everolimus (mTOR inhibitor), consistent with findings from the BOLERO‐2 trial. In this trial, the ESR1 Y537S mutation was not associated with any therapeutic benefit from the combination of everolimus and an AI [49].

Taken together, our results highlight the potential of ctDNA monitoring to provide real‐time insights into tumor biology, predict DP, and guide treatment decisions in both NSCLC and HR+, HER2‐negative MBC. The minimally invasive nature of liquid biopsies enables repeated sampling, which is essential for tracking clonal dynamics and emerging resistance mechanisms that are often missed by single‐site tissue biopsies.

Nevertheless, this study has limitations. The single‐center design and relatively small number of patients in the MBC Cohort may limit its applicability. Additionally, ddPCR, while highly accurate and sensitive for predefined mutations, limits the scope of genomic profiling compared to next‐generation sequencing (NGS). In some patients, AR mechanisms could not be identified, reflecting either biological complexity or methodological limitations of using ddPCR. In contrast, for one of the MBC patients, ctDNAseq was performed, revealing an ESR1 mutation, which complemented our analysis.

5. Conclusions

This real‐world study shows that ctDNA monitoring using ddPCR is a valuable tool for managing patients with metastatic NSCLC and HR+/HER2‐negative MBC. Serial plasma analysis enabled the detection of molecular mechanisms of acquired resistance, provided early prediction of disease progression compared to radiological imaging, and correlated with molecular responses to targeted therapies. These findings support integrating liquid biopsies into clinical practice as a complementary, minimally invasive, lower‐cost, and straightforward approach to tissue genotyping and imaging. Although further prospective, interventional trials are necessary to confirm the clinical benefits of ctDNA‐guided treatment decisions, our results emphasize the potential of liquid biopsies to improve personalized cancer care and optimize treatment strategies in precision oncology.

Author Contributions

Maurício Fernando Silva Almeida Ribeiro: resources, formal analysis. Franciele Hinterholz Knebel: conceptualization, resources, methodology, investigation, formal analysis, data curation, funding acquisition, writing – original draft, writing – review and editing. Leandro Jonata Carvalho Oliveira: resources, writing – review and editing. Isabela Flauzino Ferreira: resources, methodology, investigation. Rudinei Diogo Marques Linck: resources. Elisângela Monteiro Coser: resources, methodology. Karina Perez Sacardo: resources. Ernande Xavier dos Santos: resources, methodology. Daniele Coelho Duarte: formal analysis, visualization, writing – review and editing. Felipe Sales Nogueira Amorim Canedo: resources. Andrea Kazumi Shimada: resources, writing – review and editing. Paula Fontes Asprino: methodology, investigation, writing – review and editing. João Victor Machado Alessi: resources. Cibele Masotti: formal analysis, data curation, writing – review and editing. Max Senna Mano: resources. Marco Gerlinger: formal analysis, data curation. Frederico Perego Costa: resources. Fernando Costa Santini: resources. Artur Katz: resources. Dimitrios Kleftogiannis: formal analysis, data curation. Olavo Feher: resources. Louise J. Barber: formal analysis, data curation. Fabiana Bettoni: conceptualization, resources, methodology, investigation, formal analysis, data curation, funding acquisition, writing – original draft, writing – review and editing. Ciro Eduardo de Souza: resources. Gilberto de Castro Junior: resources, writing – review and editing. Anamaria Aranha Camargo: conceptualization, formal analysis, data curation, funding acquisition, writing – original draft, writing – review and editing. Rodrigo Saddi: resources.

Funding

This work was supported by grants from Sociedade Beneficente de Senhoras Hospital Sírio‐Libanês, Ludwig Institute for Cancer Research, and The São Paulo Research Foundation (FAPESP 2015/16854‐4; FAPESP 2016/05375‐0).

Ethics Statement

The study was approved by the Hospital Sírio‐Libanês Ethics Committee (HSL2015‐22), and patients provided written informed consent for participation, clinical data collection, and genomic analysis of blood specimens.

Consent

Patients provided written informed consent for publication of all data generated in the study.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

All data supporting the findings of this study are available from the corresponding author upon request.