Fulvestrant and everolimus efficacy after CDK4/6 inhibitor: a prospective study with circulating tumor DNA analysis

In a prospective study (NCT02866149), we assessed the efficacy of fulvestrant and everolimus in CDK4/6i pre-treated mBC patients and circulating tumor DNA (ctDNA) changes throughout therapy. Patients treated with fulvestrant and everolimus had their ctDNA assessed at baseline, after 3–5 weeks and at disease progression. Somatic mutations were identified in archived tumor tissues by targeted NGS and tracked in cell-free DNA by droplet digital PCR. ctDNA detection was then associated with clinicopathological characteristics and patients’ progression-free survival (PFS), overall survival (OS) and best overall response (BOR). In the 57 included patients, median PFS and OS were 6.8 (95%CI [5.03–11.5]) and 38.2 (95%CI [30.0-not reached]) months, respectively. In 47 response-evaluable patients, BOR was a partial response or stable disease in 15 (31.9%) and 11 (23.4%) patients, respectively. Among patients with trackable somatic mutation and available plasma sample, N = 33/47 (70.2%) and N = 19/36 (52.8%) had ctDNA detected at baseline and at 3 weeks, respectively. ctDNA detection at baseline and PIK3CA mutation had an adverse prognostic impact on PFS and OS in multivariate analysis. This prospective cohort study documents the efficacy of fulvestrant and everolimus in CDK4/6i-pretreated ER + /HER2- mBC and highlights the clinical validity of early ctDNA changes as pharmacodynamic biomarker.

Upon progression on first line CDK4/6i and endocrine therapy (ET), the combination of fulvestrant with everolimus (targeting the PI3K/AKT/mTOR pathway) is recognized as a valid second line treatment option by National Comprehensive Cancer Network (NCCN) and European Society for Medical Oncology (ESMO) [8] guidelines for ER + /HER2-mBC.The underlying evidence mostly consists in two randomized phase 2 trials, PrE0102 [9] and MANTA [10], both conducted in ER + /HER2-mBC patients whose mBC has become resistant to aromatase inhibitors.In these studies, patients treated with fulvestrant and everolimus showed a median PFS of 10.3 and 12.3 months, compared to 5.1 and 5.4 months with fulvestrant alone, respectively.However, these efficacy data were obtained without prior CDK4/6i therapy and no biomarker associated with fulvestrant and everolimus efficacy has been validated so far.
Circulating tumor DNA (ctDNA), a fraction of total cell-free circulating DNA (cfcDNA), has emerged in the last decade as a promising biomarker for diagnosis, prognosis, and monitoring treatment efficacy in different cancer types [11][12][13].In ER + /HER2-mBC treated by ET and CDK4/6i, our group and others have previously shown that early ctDNA decrease was a prognostic factor for PFS [14][15][16] and the efficiency of palbociclib and fulvestrant can be monitored by serial analyses of ctDNA [16].
This study was set up to prospectively document the efficacy of the everolimus plus fulvestrant combination after progression on CDK4/6i and, with the use of personalized droplet digital PCR (ddPCR) assays, to investigate the clinical validity of ctDNA early changes as pharmacodynamic marker in ER + /HER2-mBC patients.

Patient characteristics
From February 2018 to October 2021, 57 patients with ER + / HER2-mBC after progression on a CDK4/6i were enrolled in this study.The main characteristics of the 57 patients are as follows: median age was 57 years (range [36-79]), 30 (52.6%) patients had ≥3 metastatic sites and 33 (57.9%) had visceral metastases.Most patients (N = 48, 84.2%) had received only one prior line of systemic therapy, ET (aromatase inhibitor) and CDK4/6i.Only one patient has already received fulvestrant in the previous lines of metastatic disease.Patient characteristics are shown in Table 1.
Early ctDNA dynamics as pharmacodynamic marker Among 27 patients with positive ctDNA detection at baseline and/ or 3-5 W, 25 patients were available for analysis of correlation with the first tumor assessment (one patient was not followed up and the other patient had no measurable disease according to RECIST criteria).The majority of patients with an increase in ctDNA

Univariate and multivariate analysis
Finally, univariate and multivariate proportional hazards models were performed to evaluate the prognostic value of prospectively collected clinicopathological factors combined with ctDNA detection at different time points.As shown in Supplementary Table 4, PIK3CA mutation detection (in a subgroup of N = 25 (43.8%)patients with PIK3CA mutations), ctDNA detection at baseline and 3-5 W, ctDNA decrease between baseline and 3-5 W, and ctDNA clearance at 3-5 W were significantly associated with PFS and OS in univariate analysis (except for the presence of PIK3CA mutation and ctDNA clearance at 3-5 W, where there was no significant difference in OS), while age and performance status were only associated with OS (HR = 1.00, 95%CI [1.00-1.10],p = 0.03 and HR = 2.90, 95%CI [1.20-6.90],p = 0.02).

DISCUSSION
To our knowledge, this is the first prospective study to evaluate the efficacy of everolimus in combination with fulvestrant after progression on CDK4/6i in ER + /HER2-mBC.We document a median PFS of 6.8 months with everolimus and fulvestrant.This median PFS is apparently shorter than that reported in the PrE0102 and MANTA trials, which included patients not pretreated with CDK4/6i [9,10].A similar shortening of second line PFS in CDK4/6i pretreated patients was previously observed with the fulvestrant-alpelisib combination, with median PFS of 11.0 months when given to CDK4/6i-naïve (SOLAR-1 trial [17]) versus 7.3 (BYLieve trial [18]) and 5.3 (French early access program [19]) months in CDK4/6i-pretreated mBC patients.In the post CDK4/6i setting, few retrospective studies have reported shortlived efficacy of everolimus-based combination (with exemestane or fulvestrant) : median PFS was 4.9 months in N = 12 patients [20], 4.2 months in N = 41 patients [21] and 3.8 months in N = 79 patients [22].These PFS were obtained in patients treated in 2 nd to 5 th lines and whose prior exposure to CDK4/6 inhibitors ranged from 4 to 21 months.Measurable disease according to RECIST was not a prerequisite for inclusion in these trials.Of note, in these retrospective reports, patients received everolimus as second or later line of therapy, which may explain the apparently shorter PFS compared to our prospective study.Finally, the fulvestrant and everolimus combination observed in our study appears somehow similar to the efficacy reported with fulvestrant and capivasertib combination: in the phase III CAPItello-291 trial (Patients were required to have RECIST measurable disease (version 1.1) or at least one lytic or mixed lytic-blastic bone lesion with identifiable soft-tissue components), the AKT inhibitor given in combination with fulvestrant led to a median PFS of 7.2 months and 5.5 months in patients with prior CDK4/6i exposure [23].An appealing characteristic of everolimus is that its label allows using it in all patients, independently of their tumor mutational landscape.PIK3CA mutations have previously been reported to be associated with worse outcome in patients with ER + /HER2-mBC [24], while the efficacy of everolimus in clinical trial data is uncertain.In the neoadjuvant setting, Baselga et al. [25] found that the presence of PIK3CA exon 9 mutations conferred an improved response to the combination of everolimus with letrozole [25].In various metastatic cancers, patients treated with PI3K/AKT/mTOR pathway inhibitors in combination with other treatments (ET, anti-HER2 therapy, or chemotherapy) had a longer time to progression compared to patients without PIK3CA mutations [26].However, in the BOLERO-2 trial, the authors found no association between the presence of a PIK3CA mutation (either by next generation sequencing on tumor tissue or in cell-free DNA by ddPCR) and response to everolimus [27,28].In our study, we found that patients with PIK3CA mutation had a limited benefit, with a median PFS of 3.4 months and a median OS of 22.9 months.This median PFS appears to be shorter than reported with alpelisib in the BYLieve study [18] (7.3 months).The small sample size of this subpopulation in our study makes it difficult to draw definitive conclusions, but the impact of PIK3CA mutation on everolimus and fulvestrant efficacy after CDK4/6i needs to be further explored.
Finally, we investigated the impact of ctDNA dynamics as a predictive biomarker for the efficacy of everolimus plus fulvestrant.We found a prognostic impact of the baseline and 3-5 W ctDNA positivity on PFS and OS, and also a prognostic impact of the variation of ctDNA level between 3-5 W and baseline.However, among the patients with decreased ctDNA, three patients had a PD at the first radiological evaluation and two patients had a PR in spite of an increased ctDNA load.One hypothesis is that the mutations used for ctDNA monitoring in these patients were from a tumor subclone, which eventually couldn't predict and represent the whole tumor evolution under treatment.Our results are in line with a recent study [29] monitoring ctDNA in fulvestrant-exemestane combination, where PFS was particularly short in patients with detectable ctDNA after 14 days of treatment: 2.1 months versus 5.0 months in those without detectable ctDNA (p = 0.012).A statistically significant longer PFS was observed for patients with ctDNA ratio (D14/ baseline) below or equal to the median value (p ≤ 0.061): 4.4 months versus 1.9 months (HR = 2.5, 95%CI [1.0-6.2],Fig. 5 Tumor response and prognostic value of early ctDNA variations.Tumor response according to ctDNA change from baseline (A).*SD ≥ 6 months.Progression-free survival and overall survival according to early ctDNA variation (B, C) and ctDNA clearance (D, E).PD progressive disease, PR partial response, SD stable disease, PFS progression-free survival; OS: overall survival.p = 0.043).In the second line setting where multiple treatments are available and endocrine-resistance therapies are frequent, assessment of ctDNA dynamics during treatment could rapidly predict patients that will experience a tumor progression or response, but further studies are needed to assess its clinical utility.A phase 2 trial ("MONDRIAN", NCT04720729) is underway to evaluate the use of early ctDNA level variation to monitor response to chemotherapy in HER2-mBC.
In conclusion, we report here the first prospective assessment of fulvestrant-everolimus in ER + /HER2-mBC patients who progressed on prior CDK4/6 inhibitor.We document that fulvestranteverolimus is an active regimen in this population, with limited toxicity, whereas ctDNA detection demonstrated a satisfactory clinical validity for monitoring response to therapy.

Patients and blood sampling
All patients signed a written informed consent to this ethically approved study (Comité de Protection des Personnes "Ile de France VI"), registered with clinicaltrials.gov(NCT02866149, cohort #12).Eligibility criteria were: patients aged ≥18 years with ER + /HER2-mBC (immunoreactive for ER in ≥10% of tumor-cell nuclei by immunohistochemistry) treated at Institut Curie Hospitals (Paris and Saint Cloud), who progressed under prior CDK4/ 6i therapy (with no requirement for prior CDK4/6i duration).The presence of measurable disease (per RECIST criteria) was not a prerequisite for patient enrollment.All included patients were then treated with everolimus (10 mg/day, unless decided otherwise by physicians) and fulvestrant, per standard of care.Mucositis prophylaxis was left at the discretion of the physician.Tumor response to therapy was assessed at least every 3 months by CT-scan.The objective response rate (ORR) was calculated only with the patients with RECIST measurable disease, including those with unconfirmed responses.Any patient who did not achieve a complete/ partial response or progressive disease (PD) within 6 months was considered as having a stable disease (SD).Adverse events were coded and graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events (version 5.0).Patient characteristics and survival data were prospectively registered.Patients who discontinued everolimus treatment for any reason other than progression were required to follow the same schedule of assessments until progression.For each patient, longitudinal blood samples were collected: at baseline, after 3-5 weeks (3-5 W) of therapy and at time of disease progression.Variation in ctDNA was calculated using copies/ml.

Plasma sample preparation, storage, DNA extraction and quantification
Blood samples were collected in EDTA tubes and plasma was isolated within 2 h by a two-step centrifugation: 820 × g for 10 min, then 16000 × g for 10 min at 4 °C.DNA was extracted using the QIAmp Circulating Nucleic Acid Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions and quantified by Qubit 2.0 fluorometer (Invitrogen, Carlsbad, CA, USA) using the dsDNA HS Assay (Invitrogen, Carlsbad, CA, USA).Extracted DNA was stored at -20 °C before use.

Identification of trackable somatic mutations
Archived tumor samples, either from the primary tumor or a metastatic deposit, were retrieved from the Pathology department.Tumor DNA was extracted from macro-dissected tumor tissue (Macherey-Nagel nucleospin tissue kit) and subjected to a large amplicon-based targeted NGS panel developed at Institut Curie, covering 483 genes (Qiagen Qiaseq).For patients with no tumor tissue available or failure of tissue DNA sequencing, targeted-NGS was performed on cfcDNA collected at baseline.

Analysis of ctDNA by ddPCR
The identified somatic mutations were used for ctDNA monitoring in plasma by ddPCR.Tumor or synthetic DNA carrying the target mutation was used as positive control.Genomic DNA from healthy donors purchased from Promega (Madison, WI, USA) was used as negative control.Primers and probes for detecting PIK3CA: c.3140A > G mutation are described in Supplementary Table 1.Assays for detecting TP53: c.637 C > T mutation was provided by SAGA Diagnostics (Lund, Sweden).ddPCR assays for detecting other mutations were purchased from Biorad laboratories (Hercules, CA, USA) (Supplementary Table 1).ddPCR assays were performed using the BioRad QX100 system (Biorad laboratories, Hercules, CA, USA) with 900 nM of each primer, 250 nM of each probe with one probe targeting the wide-type sequence and the other one targeting the mutant sequence being labeled with different fluorophores.The amplification was under the following conditions: 95 °C for 10 min (1 cycle); 40 cycles of 94 °C for 30 sec and hybridization at a specific temperature for each mutation for 60 s, and 98 °C for 1 min (1 cycle).The hybridization temperatures used for each assay were in Supplementary Table 1.Cluster thresholding and quantification were performed with QuantaSoft v.1.7.4 software.Droplets were manually assigned as wild-type (WT) or mutant (MUT) based on their fluorescence amplitude: WT: VIC+ or HEX + ; MUT: FAM + .The mutant allelic frequencies (MAFs) were calculated as follows: copy numbers of MUT / (copy numbers of WT + copy numbers of MUT).The total copy number of cfcDNA in each patient was calculated as the sum of the copy numbers of WT and MUT sequences.All experiments met the minimum requirements for digital PCR data [30,31].

ddPCR data analysis
The false-positive rate of each assay was estimated as previously reported [32,33] using ≥ 15 replicates of WT DNA.The limit of blank (LOB) of different assays was defined as the upper 95% confidence limit of the mean false-positive measurements.Samples were considered as positive when MAF was higher than LOB with more than 1-3 positive droplets detected per analysis (Supplementary Table 1).When less than 300 amplifiable genomes were detected by ddPCR, the sample was not considered for further ctDNA analysis.The laboratory performing ddPCR experiments was blinded to patient outcomes.

Statistical analysis
Categorical variables were compared using Fisher's exact test.Survival analysis was performed using Kaplan-Meier plots with significance tested using the log-rank test.Progression-free survival (PFS), defined as the time from initiation of treatment to disease progression or death from any cause, was collected prospectively.Patients with no disease progression were censored at the last follow-up visit.OS was defined as the time from initiation of treatment to the date of death from any cause.Best overall response was defined as the best response recorded from the start of therapy until its discontinuation.
Cox proportional hazards modeling was used for univariate and multivariate analysis to evaluate the prognostic significance of different predictors.Baseline clinical and pathological characteristics with a p value > 0.1 in univariate analysis were excluded from multivariate analysis.Statistical analyses were performed with GraphPad Prism (version 8.0) or R software (version 4.1.1).P-values ≤ 0.05 were considered statistically significant.

Fig. 2
Fig.2Progression-free survival and overall survival in patients treated with everolimus and fulvestrant.In the whole population (A, B) and according to previous palbociclib treatment duration (above or under/equal to median PFS (C, D)).PFS progression-free survival, OS overall survival.

Fig. 3
Fig. 3 ctDNA detection at different time points.A ctDNA levels at different time points quantified by ddPCR.Each dot represents the level of ctDNA of a patient measured.Mann-Whitney test was used to compare ctDNA levels at different time points.B Swimmer plot showing the dynamic changes of ctDNA levels, best overall response and timing of clinical relapse.Each bar represents one patient.Red circles indicate ctDNA detection and blue circles indicate no ctDNA detection.The black triangle indicates disease progression (PD).*SD ≥ 6 months.Bsl: baseline; 3-5 W: 3-5 weeks, PD progressive disease, PR partial response, SD stable disease, NA not available.

Fig. 4
Fig. 4 Prognostic value of ctDNA detection at different time points.Progression-free survival and overall survival according to ctDNA detection at baseline (A, B) and 3 weeks (C, D).PFS progression-free survival, OS overall survival.
*For metastatic disease.

Table 3 .
Multivariate analysis for progression-free survival and overall survival.