Mutational Analysis of Circulating Tumor DNA in Patients With Estrogen Receptor–Positive/Human Epidermal Growth Factor Receptor 2–Negative Advanced Breast Cancer Receiving Palbociclib: Results From the TREnd Trial

Abstract

PURPOSE

To identify prognostic circulating biomarkers to cyclin-dependent kinase 4 and 6 inhibitors (CDK4/6i), we performed a mutational analysis on circulating tumor DNA (ctDNA) samples from patients included in the TREnd trial, which randomly assigned patients to receive the CDK4/6i palbociclib alone or with the endocrine treatment (ET) to which they had progressed.

METHODS

Forty-six patients were enrolled in this substudy. Plasma was collected before treatment (T0), after the first cycle of therapy (T1), and at the time of progression (T2). ctDNA hybridization and capture were performed using the Illumina TruSight Tumor 170 Kit. Acquired mutations were confirmed by digital polymerase chain reaction. Progression-free survival analysis was estimated using the Kaplan-Meier method and compared with the log-rank test.

RESULTS

The most frequently mutated genes at T0 were ESR1 (23%), PIK3CA (17%), AR, FGFR2, and TP53 (10%). Mutations in ESR1 at T0 conferred higher risk of progression in the entire population (P = .02) and in patients treated with palbociclib + ET (P = .04). ESR1 mutation effect remained significant after correction for clinical variables (P = .03). PIK3CA mutations at T0 were not prognostic, but higher risk of progression was observed when a broader analysis of PI3K pathway was performed (P = .04). At T2, we observed the emergence of nine new mutations in seven genes.

CONCLUSION

Mutations in ESR1 and in PI3K pathway genes at T0 were associated with worse prognosis in palbociclib-treated patients. We describe the emergence of newly acquired mutations in palbociclib-treated patients, which might potentially affect subsequent treatment.

ctDNA analysis of palbociclib-treated patients revealed an independent prognostic role of ESR1 mutations.

INTRODUCTION

Cyclin-dependent kinase 4 and 6 inhibitors (CDK4/6i) represent a standard option for the first- and second-line treatment of patients with estrogen receptor–positive (ER+) and human epidermal growth factor receptor 2–negative (HER2–) metastatic breast cancer (BC). However, not all patients benefit from CDK4/6i, showing intrinsic or acquired resistance. Studies have investigated the determinants of resistance,^1,2 but to date, to our knowledge, no biomarker has been approved to direct CDK4/6i treatment in ER+/HER2– BC. Also, there is uncertainty about how patients should be treated after CDK4/6i progression.^3-5

CONTEXT

• Key Objective

• Analysis of circulating tumor DNA allows the identification of prognostic biomarkers that might be helpful for patients' stratification and monitoring. Here, through a mutational analysis of a subgroup of patients receiving the cyclin-dependent kinase 4 and 6 inhibitor (CDK4/6i) palbociclib within the TREnd trial, we aimed to identify mutations that might serve as prognostic circulating biomarkers.

• Knowledge Generated

• Of all mutations identified at baseline, those involving ESR1 were significantly associated with outcome at univariate and multivariate analyses. We also identified mutations in additional genes that were likely acquired, not being present before starting treatment.

• Relevance

• In metastatic breast cancer, ESR1 mutations were found to be associated to resistance to endocrine therapy, but their prognostic role in patients treated with CDK4/6i is yet not well understood. We add to the previous knowledge by further defining the prognostic role of ESR1 mutations.

Circulating tumor DNA (ctDNA), the fraction of tumor DNA to the total cell-free DNA (cfDNA), can be easily obtained from plasma samples of patients with advanced BC and assessed for both tumor genetic and epigenetic alterations,^6,7 holding a great potential for biomarker discovery.

Previous studies in patients treated with CDK4/6i demonstrated that both the presence and dynamics of ctDNA alterations might help identifying biomarkers of intrinsic and acquired resistance and monitoring treatments response.^8-14 However, results have not always been univocal.

Here, we performed a mutational analysis on ctDNA obtained from patients with ER+/HER2– advanced BC receiving the CDK4/6i palbociclib. Patients were enrolled in the c-TREnd study, a translational substudy of TREnd, an investigator-initiated phase II randomized trial including patients with ER+/HER2– advanced BC treated with palbociclib alone or palbociclib plus the endocrine agent to which patients had progressed in the previous line of endocrine therapy (ET).¹⁵ The aims were to identify potential circulating DNA-based biomarkers of both intrinsic and acquired resistance to palbociclib.

METHODS

Patients and Sample Collection

TREnd trial is an investigator-initiated, phase II, open-label, multicenter study that randomly assigned 115 patients with ET-resistant ER+/HER2– advanced BC to receive standard-dose palbociclib either as monotherapy or in combination with the ET received in the previous line of treatment.¹⁵ In parallel with TREnd trial, a translational substudy, the c-TREnd, was conducted with the aim to identify potential circulating biomarkers of palbociclib resistance. The study had the approval from the ethics committee of the coordinating center (Comitato Etico Area Vasta Centro) and from the independent local ethics committees of each participating center, according to local regulations. Patients joining this substudy signed a separate informed consent. For the ctDNA analysis, blood samples were collected in EDTA-containing tubes and processed within 1 hour. Plasma samples were then stored at –80°C. cfDNA was extracted using the QIAamp Circulating Nucleic Acid Kit (Qiagen, Hilden, Germany), according to the manufacturer's instructions, and quantified by Qubit fluorometer using Qubit dsDNA HS (High Sensitivity) Assay Kit (Thermo Fisher, Waltham, MA).

Gene Panel Sequencing

Next-generation sequencing experiments, comprising sample quality control, were performed by Genomix4life S.R.L. (Baronissi, Salerno, Italy). cfDNA concentration and size were determined by using Qubit fluorometer (Thermo Fisher Scientific) and the TapeStation 4200 (Agilent Technologies, Santa Clara, CA), respectively. Indexed libraries were prepared from 15 to 20 ng/ea purified ctDNA using the TruSight Tumor 170 Sample Prep Kit (Illumina, San Diego, CA). Libraries were quantified by TapeStation 4200 and Qubit fluorometer and pooled such that each index-tagged sample was present in equimolar amounts, with final concentration of the pooled samples of 2 nM. The pooled samples were subject to cluster generation and sequencing using an Illumina NextSeq550Dx System (Illumina) in a 2 × 75 paired-end format. The raw sequence files generated (.fastq files) underwent quality control analysis using FastQC.¹⁶

Data Analysis

Reads were aligned to the hg19 reference genome with BWA¹⁷ using the GATK (v4.0.11.0)^18-20 workflow (GitHub²¹; downloaded on 2019-4-12). Single-nucleotide variants were detected using LoFreq (v2.1.3.1)²² and annotated using Oncotator (v1.9.9.0)²³ by exploiting both population and cancer databases (ExAC, dbSNP, 1000 Genomes, TCGA, MBC, METABRIC, MSK, and IGR).^24-30

To address the lack of matched control (ie, germline) samples, we performed a custom characterization of all variants to retrieve only the somatic mutations. First, we collected the variant allele counts of all the variants detected by LoFreq at least at one time point per patient using ASEQ.³¹ Then, we divided the variants into three classes according to their variant allele fraction (VAF) using 25% and 75% as cutoffs, and analyzed whether each variant was present in any cancer database and/or in any population database with at least 5% frequency. We excluded all the variants with VAF >75% or with VAF <75% but present only in population databases. Variants were considered somatic if VAF was <25% and found in cancer databases or in no database. Variants were classified as uncertain somatic if VAF was >25% and <75% and found in cancer databases or in no database. Uncertain somatic variants were further reclassified as somatic if VAF was <50% at all time points and the variation between time points (T0-T1, T0-T2, and T1-T0) was significantly different from that of variants with VAF between 25% and 75% and present in population databases (P value < .05 after Benjamini-Hochberg correction for multiple testing). Finally, we removed variants not confirmed by Abemus at any time point as well as those supported by less than two reads.

We considered a mutation to be acquired by a given patient when retrieved by LoFreq only at T2 and with no reads supporting the alternative alleles recovered by ASEQ. Tumor mutational burden (TMB) was defined as the total number of silent and nonsilent mutations per million bases of targeted regions (total size = 521,099).

Droplet Digital Polymerase Chain Reaction

Three patients with acquired mutations and available plasma samples from more than one time point were analyzed by using droplet digital polymerase chain reaction (ddPCR; Bio-Rad, Hercules, CA) to confirm the presence of the mutations identified at T2. Specifically, we analyzed samples ct73 T0 and T2, ct98 T0 and T2, and ct100 T1 and T2. ESR1 (p.Y537N), CCNE1 (p.H306N), AKT1 (p.D48N), BRIP1 (p.E1151Q), ARID1A (p.T1039R), and TP53 (p. E336*) mutations were detected using ddPCR MUT FAM/HEX assays (Bio-Rad). Whenever the assays were not available, custom assays were designed. All the DNA extracted from 1 mL of plasma was used for ddPCR analysis. To simultaneously detect BRIP1 and ESR1 mutations in ct98 samples, ctDNA was equally divided and used for a multiplex ddPCR assay, while a single-plex assay for AKT1 mutation was performed in the same patient. A multiplex assay was performed to allow the simultaneous detection of ARID1A and TP53 mutations in ct100 samples. To set up single-plex and multi-plex ddPCR assays, wild-type (WT) and mutated 251-bp custom-made gBloks Gene Fragments were used (IDT, Leuven, Belgium). These were also used as positive and negative controls in each run. Fractional abundance, the ratio between mutated and total (mutated plus WT) gene copies, was calculated by using Bio-Rad QuantaSoft package.

Statistical Analysis

Progression-free survival (PFS) was computed as the time from treatment initiation to radiologic disease progression or death. Time to treatment failure (TTF) was the time interval between initiation of a new therapy and its discontinuation.

Observation time of patients still on treatment at the time of the data cutoff for the analysis (April 24, 2018) was censored at the last visit.

The distribution of PFS was estimated using the Kaplan-Meier method and compared with the log-rank test. Hazard ratios (HRs) with 95% CIs were calculated with the Cox proportional hazards model. Median follow-up time was estimated according to the Kaplan-Meier reverse method. A multivariate Cox regression model was fitted to evaluate the independent effect of each covariate on PFS. The covariates included in the model were ESR1 mutational status, sites of metastasis at study entry, number of lines of ET before study entry, and duration of last ET before study entry.

RESULTS

Patients and Clinicopathologic Characteristics

Forty-six patients were enrolled in the c-TREnd substudy, 20 from the palbociclib single-agent arm and 26 from the palbociclib + combination ET arm. Before starting trial treatments (T0), we collected 45 samples; 44 samples were collected after the first cycle of treatment (T1) and 37 at the time of progression or before starting a new line of therapy (T2). Of the 126 samples collected, 39 were excluded because of low cfDNA content (<16 ng) or other, mainly technical, reasons (n = 8), leaving 87 samples from 32 patients (14 from the palbociclib single-agent arm and 18 from the combination arm) for the final analysis, as shown in the CONSORT diagram (Data Supplement, Fig S1).

Clinicopathologic characteristics of the final cohort are presented in Table 1. Baseline characteristics were well balanced between the two arms. The majority of patients had performance status 0, visceral disease, and received one line of therapy before entering the trial. Also, around 80% of patients had been on their most recent line of ET before study entry for more than 180 days.

TABLE 1.

Clinicopathologic Characteristics of the Study Population

Characteristic	Palbociclib + ET (N = 18)	Palbociclib (N = 14)
Age, years, No (%)	63.5 (44-82)	64 (52-80)
ECOG, No. (%)
0	14 (77.8)	11 (78.6)
1	3 (16.7)	3 (21.4)
2	1 (5.6)	0 (0)
Disease at diagnosis, No. (%)
De novo metastatic	2 (11.1)	4 (28.6)
Early	15 (83.3)	10 (71.4)
Not classified	1 (5.6)	0 (0)
Sites of metastasis, No. (%)
Visceral	12 (66.7)	11 (78.6)
Bone only	3 (16.7)	1 (7.1)
Other nonvisceral	3 (16.7)	2 (14.3)
Previous lines of ET, No. (%)
One line	12 (66.7)	10 (71.4)
Two or three lines	6 (33.3)	4 (28.6)
Duration of most recent ET, days, No. (%)
≤180	2 (11.1)	3 (21.4)
>180	16 (88.9)	11 (78.6)
Most recent ET, No. (%)
Aromatase inhibitor	9 (50)	9 (64.3)
Fulvestrant	9 (50)	5 (35.7)
Previous chemotherapy for metastatic BC, No. (%)
Yes	5 (27.8)	2 (14.3)
No	13 (72.2)	12 (85.7)

NOTE. Differences between treatment arms were not statistically significant (t-test or chi-square test P > .05).

Abbreviations: BC, breast cancer; ECOG, Eastern Cooperative Oncology Group; ET, endocrine therapy.

A proportion test was used to assess the imbalance in the number of patients according to treatment arm between the TREnd trial and this substudy population, finding no statistically significant imbalance (P = .64).

There was no significant difference in median PFS in the group of patients included in this substudy compared with those who were not (P = .41; Data Supplement, Fig S2).

Mutational Analysis at T0 and PFS Analysis

Of 30 patients with T0, those with at least one reported nonsynonymous mutation were 23 (76.7%; Fig 1). The list of the variants detected in all samples is reported in the Data Supplement (Table S1). The most frequently mutated genes at T0 were ESR1 (7/30, 23.3%), PIK3CA (5/30, 16.7%), AR (3/30, 10%), FGFR2 (3/30, 10%), and TP53 (3/30, 10%). To explore alterations within signaling pathways of interest, genes were grouped on the basis of their functional role (Data Supplement, Table S2). As expected, the most frequently altered pathways were PI3K/AKT signaling, DNA damage, and ER-dependent genes. TMB ranged from 0 to 11.5 with a median value of 3.8 (Data Supplement, Table S3).

FIG 1.

Oncoplot illustrating the mutational landscape, pathway of interest, TMB, and treatment arm of patients included in the study. Mutations were detected by the LoFreq and/or ASEQ (see Methods). ET, endocrine therapy; SNV, single nucleotide variant; TMB, tumor mutational burden.

The median follow-up time for the entire cohort was 28.3 months (95% CI, 23.4 to 31.4).

We aimed to analyze if TMB and the mutations found at T0 were associated with PFS in the entire population or in patients treated with palbociclib plus ET or palbociclib alone. We found that TMB at T0 was not significantly associated with PFS (Fig 2). Among the mutations found at T0, ESR1 mutations were significantly associated with increased risk of progression in the entire population (HR, 3.06 [1.19-7.87]; P = .02). The prognostic effect of ESR1 mutations remained statistically significant in the palbociclib plus ET arm (HR, 4 [1.10-14.59]; P = .04), but not in patients receiving palbociclib alone (HR, 1.88 [0.47-7.56]; P = .38). However, the number of patients in this subanalysis is rather small. Association between ESR1 mutational status and clinical variables showed that ESR1 mutations were more frequently observed in patients who received more than one line of ET before study entry (P = .03; Data Supplement, Table S4). The independent prognostic role of ESR1 mutational status in the entire study population was assessed by multivariate analysis (Table 2). After correcting for multiple variables, ESR1 mutations at T0 remained significantly associated with increased risk of progression (HR, 3.32 [1.14-9.64]; P = .03). PIK3CA and TP53 mutations were not significantly associated with PFS in the entire population or in patients divided according to treatment arms (Fig 2). However, analyzing mutations within PI3K/AKT pathway, a significantly increased risk of progression was observed in the entire population (HR, 2.38 [1.02-5.55]; P = .04). On the other hand, patients with AR mutant tumors tended to have lower, albeit not statistically significant, risk of progression (HR, 0.53 [0.16-1.77]; P = .30). Interestingly, AR and ESR1 mutations appeared to be mutually exclusive and patients with AR mutations had a significantly longer PFS compared with those with mutations in ESR1 (P = .009; Fig 3).

FIG 2.

Forest plots showing the results of PFS analysis according to TMB and selected mutations and pathways (A) in the entire population or in patients divided according to treatment arms: (B) palbociclib or (C) palbociclib plus ET. Forest plots were generated by fitting Cox proportional hazards regression models (coxph R function) to each variate separately. ET, endocrine therapy; HR, hazard ratio; PFS, progression-free survival; TMB, tumor mutational burden.

TABLE 2.

Univariate and Multivariate Analyses

Variable	All	HR (univariable)	HR (multivariable)
ESR1 baseline status, No (%)
WT	23 (76.7)
Mut	7 (23.3)	3.06 (1.19-7.87; P = .020)	3.32 (1.14-9.64; P = .027)
Disease at diagnosis, No (%)
De novo metastatic	6 (20.7)
Early	23 (79.3)	1.35 (0.54-3.37; P = .526)
Sites of metastasis, No (%)
Bone only	4 (13.3)
Other	26 (86.7)	1.78 (0.61-5.17; P = .290)	1.94 (0.64-5.92; P = .242)
Previous lines of ET, No (%)
One line	20 (66.7)
Two or three lines	10 (33.3)	1.46 (0.67-3.19; P = .347)	1.26 (0.51-3.09; P = .616)
Duration of most recent ET, days, No (%)
>180	25 (83.3)
≤180	5 (16.7)	2.18 (0.80-5.89; P = .126)	2.32 (0.82-6.55; P = .112)
Previous chemotherapy for metastatic BC, No (%)
No	24 (80.0)
Yes	6 (20.0)	2.05 (0.82-5.15; P = .125)
Most recent ET, No (%)
Aromatase inhibitor	17 (56.7)
Fulvestrant	13 (43.3)	1.63 (0.76-3.48; P = .210)

Abbreviations: BC, breast cancer; ET, endocrine therapy; HR, hazard ratio; Mut, mutated; WT, wild-type.

FIG 3.

(A) Graph illustrating patients with AR and/or ESR1 mutations in our cohort. (B) Kaplan-Meier curves of PFS according to AR or ESR1 mutational status. PFS, progression-free survival; WT, wild-type.

Mutations at T2 and TTF

At the time of disease progression, we observed the emergence of nine mutations in seven genes (ESR1, AKT1, ARID1A, BRIP1, CCNE1, MTOR, and TP53), not detected at earlier time points (Fig 2).

In patients with available plasma samples, the acquired mutations in ESR1 (p.Y537N), CCNE1 (p.H306N), AKT1 (p.D48N), BRIP1 (p.E1151Q), ARID1A (p.T1039R), and TP53 (p. E336*) were validated by ddPCR, confirming their presence at T2 and their absence at T0 or T1 (Data Supplement, Figs S3 and S4).

We observed that patients with newly acquired mutations at T2 tended to have a shorter TTF after palbociclib progression. Indeed, three of 15 (20%) patients with TTF below the median of the entire population had acquired mutations compared with only one of 15(7%) patients with TTF above the median (Fig 4).

FIG 4.

Swimmer plot showing the TTF and PFS of patients in the study. Highlighted are the most frequent mutations at T0 or the presence of newly acquired mutations at T2. Dashed vertical lines indicate median PFS (left) and median TTF (right). ET, endocrine therapy; PFS, progression-free survival; TTF, time to treatment failure; WT, wild type.

DISCUSSION

The c-TREnd substudy collected blood samples in a subset of patients with ER+/HER2– advanced BC enrolled in the TREnd trial¹⁵ with the aim of identifying circulating biomarkers of response to palbociclib. Albeit small, this subset is representative of the entire TREnd population and does not present a significant imbalance in the treatment arms.

In our study, we found a prevalence of ctDNA mutations in ESR1, PIK3CA, and TP53 comparable with that observed in the PALOMA-3 trial,¹³ a phase III trial of palbociclib plus fulvestrant or placebo plus fulvestrant, including a population similar to TREnd. However, we observed a higher-than-expected mutation rate in the AR gene. AR mutations have been demonstrated in <2% of primary²⁵ and metastatic BC²⁹ but recent findings in ER+/HER2– metastatic BC reported higher mutations rates (8.5%-28.5%).^32-35.

In previous studies, baseline tumor fraction (TF) and variations in tumor content have been variably associated with outcome of patients treated with CDK4/6i.^{10,12,13,36-38} As there is no standardized method to estimate TF with our panel, we did not assess it. A recent report found that high TMB on blood samples from patients receiving palbociclib plus ET was significantly associated with shorter PFS,³⁸ but in our cohort, neither baseline TMB nor variations in TMB at T1 (data not shown) were able to predict outcome on palbociclib. Differences in population and methodologies might explain these discrepancies. Changes in VAF of individual mutations were not explored because of the limited sample size. In the entire population, we found that ESR1 mutations were independently associated with increased risk of progression. ESR1 mutations have been previously assessed in ctDNA of CDK4/6i-treated patients within PALOMA-3 and MONARCH-2, which enrolled patients with ER+/HER2– advanced BC who progressed on ET, similar to TREnd.^39,40 Both trials showed that ESR1 mutations were not predictive of CDK4/6i response. However, in MONARCH-2, a numerical improvement in median PFS with abemaciclib versus placebo was observed in patients with ESR1-mutant tumors. The prognostic role of ESR1 mutations in the CDK4/6i-treated subgroup was not directly investigated. However, in the palbociclib- plus fulvestrant-treated subgroup of PALOMA-3, the median PFS of patients with ESR1-mutant and WT tumors was similar (9.4 v 9.5 months, respectively).^11,40 In MONARCH-2, patients with ESR1-mutant tumors were 59.3% of the analyzed population. In the subgroup of patients treated with abemaciclib and fulvestrant, the median PFS of patients with ESR1-mutant tumors was 20.7 months versus 15.3 for patients with ESR1 WT.³⁹ These unexpected results might be explained by different populations, sample size, and methodology, as discussed by the authors.³⁹ Additional data on ESR1 mutations derive from the PEARL trial, randomly assigning AI-resistant patients with HR+/HER2– BC to receive palbociclib plus exemestane versus capecitabine.³⁷ Baseline ESR1 mutations were not significantly associated with PFS in the overall or in the palbociclib-treated population, but were associated with overall survival (OS) in the overall population.³⁷ In another study analyzing 16 patients with metastatic BC treated with palbociclib and letrozole, PFS was significantly shorter in patients with ESR1-mutant tumors at initial blood draw.¹⁴ Our results confirm a worse prognosis for patients with ESR1-mutant tumors receiving palbociclib. However, an increased risk of progression was observed mainly in patients receiving palbociclib with ET, supporting the hypothesis that ESR1 mutations confer particular resistance to ET.

PIK3CA mutations have been previously analyzed in patients receiving CDK4/6i. PIK3CA status at baseline was not predictive of outcome on palbociclib and fulvestrant in the PALOMA-3 trial, whereas early changes in PIK3CA mutation levels were.^11,41 PIK3CA mutations were not predictive of response in the MONARCH-2 trial but were prognostic in the fulvestrant arm.³⁹In the abemaciclib arm, median PFS of patients with PIK3CA-mutant and WT ctDNA was similar (17.1 months v 16.9 months).³⁹In the PEARL trial, although no significant association between PIK3CA mutations and PFS or OS was observed in the overall population, worse OS was reported for patients in the palbociclib plus fulvestrant arm.³⁷ In our study, PIK3CA mutations were not prognostic, but our results might suggest that a broader evaluation of the mutational landscape within the PI3K pathway might give better prognostic estimations.

We found that AR mutations tended to be associated with reduced risk of progression. To date, there are very limited data regarding AR mutations and prognosis in BC. A recent study demonstrated that the AR variant H384P was significantly associated with increased survival in patients with ER+/HER2– metastatic BC,³⁵ while another study found no association between AR mutations and outcome in patients with ER+/HER2– metastatic BC treated with chemotherapy or CDK4/6i.³² The number of patients with AR mutations in our study is very limited, therefore larger studies are needed to assess their prognostic role in metastatic BC.

At the time of disease progression, we found the emergence of nine mutations in seven genes, confirmed by ddPCR. Of note, in all cases, these were not detectable at earlier time points, suggesting that they might be acquired during treatment. Previous studies showed that ESR1 mutations can be acquired during palbociclib and letrozole treatment.^9,14 In our study, one ESR1 mutation was acquired in a patient treated with palbociclib alone, suggesting that ESR1 mutations might be selected during palbociclib monotherapy. Acquired mutations in TP53 were also found in the PALOMA-3,⁹ and AKT1 mutations were found in a recent study analyzing patients with de novo or acquired resistance to CDK4/6i.⁴² Acquired mutations in CCNE1, ARID1A, BRIP1, and MTOR have not been previously described in patients treated with CDK4/6i, although CCNE1 overexpression/amplification has been associated with CDK4/6i resistance.^43-45 The CCNE1 H306N mutation found in our study has been reported in a metastatic BC sample from a patient with ER+/HER2– tumor²⁹; however, its biological significance is unknown. We found no acquired mutations in PIK3CA, NF1, RB1, ERBB2, KRAS, CDKN1B, or FGFR2, differently from PALOMA-3. Differences in cohorts, treatments, and methodologies coupled with the low allele frequencies of mutations in blood samples might explain discrepancies.

In PALOMA-3, a relationship between longer PFS on palbociclib plus fulvestrant and the acquisition of a new mutations at the end of treatment was found.⁹ However, the possible prognostic effect of the acquired mutations in the next line of therapy was not reported. Our preliminary observation suggests that the emergence of new mutations during treatment with palbociclib might be associated with worse outcome on the postprogression treatment. However, stronger evidence is needed from larger studies.

The major limitations of our study include the limited number of patients, the lack of matched germline samples, and the exploratory nature of the analyses. Strengths encompass the randomized nature of the trial, the availability of samples collected across multiple time points during treatment, the validation of the acquired mutations by ddPCR, and the unique availability for this study of a palbociclib monotherapy arm.

In conclusion, in this study, we assessed potential prognostic biomarkers in patients treated with CDK4/6i. We showed that ESR1 mutations are prognostic in patients treated with palbociclib, particularly those receiving concomitant ET. Given the limited sample size of c-TREnd, further evidence is needed to explore the prognostic role in patients treated with palbociclib alone.

A major unmet clinical question regards the best treatment after progression on CDK4/6i.⁵ Mutations found at the time of progression might be of therapeutic relevance since many drugs are currently being evaluated to target specific alterations.⁴⁶ Here, we identified mutations not previously described in patients treated with CDK4/6i and showed the feasibility of using a targeted gene panel to detect acquired mutations. Further studies are needed to fully understand the associations between these mutations and CDK4/6i treatment.

EQUAL CONTRIBUTION

M.B. and L.M. contributed equally to this work as cosenior authors.

PRIOR PRESENTATION

Presented in part at the ESMO Breast Cancer Meeting 2021, virtual, May 5-8, 2021.

SUPPORT

Supported by Fondazione AIRC per la ricerca sul cancro (IG 22869 and MFAG 18880 to L.M.) and Pfizer via an Investigator-Initiated Research Grant (to L.M.) sponsored by Fondazione Sandro Pitigliani per la lotta contro i tumori ONLUS.

DATA SHARING STATEMENT

The data sets generated and/or analyzed during the current study are not publicly available but might be made available from the corresponding author upon reasonable request.

AUTHOR CONTRIBUTIONS

Conception and design: Ilenia Migliaccio, Giuseppe Curigliano, Matteo Benelli, Luca Malorni

Financial support: Giuseppe Curigliano, Matteo Benelli, Luca Malorni

Administrative support: Giuseppe Curigliano

Provision of study materials or patients: Francesca De Luca, Giuseppe Curigliano, Carmen Criscitiello, Matteo Benelli, Luca Malorni

Collection and assembly of data: Francesca Galardi, Francesca De Luca, Chiara Biagioni, Giuseppe Curigliano, Erica Moretti, Emanuela Risi, Matteo Benelli, Luca Malorni

Data analysis and interpretation: Ilenia Migliaccio, Dario Romagnoli, Chiara Biagioni, Giuseppe Curigliano, Carmen Criscitiello, Alessandro Marco Minisini, Cristina Guarducci, Agostina Nardone, Laura Biganzoli, Matteo Benelli, Luca Malorni

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/po/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).