Sequential use of PI3K/AKT/mTOR pathway inhibitors alpelisib and everolimus in patients with hormone receptor-positive metastatic breast cancer

Abstract

Background

Everolimus, an mTOR inhibitor, is FDA-approved with endocrine therapy (ET) for hormone receptor-positive (HR+)/HER2- MBC regardless of PI3K/AKT/mTOR pathway mutation status. Alpelisib, a PI3K inhibitor, is FDA-approved with fulvestrant for PIK3CA-mutant HR+/HER2- MBC. The efficacy and safety of sequential treatment with these pathway inhibitors is not well described.

Methods

This single-center observational study identified patients with PIK3CA-mutant HR+/HER2- MBC treated with everolimus and alpelisib sequentially from 2012 to 2023. We abstracted patient, tumor, and treatment information from medical records. Patients treated first with everolimus then alpelisib were categorized as Group 1 and those treated first with alpelisib then everolimus as Group 2. Median time to treatment failure on 1st pathway inhibitor (TTF1) and TTF on 2nd pathway inhibitor (TTF2) were computed using the Kaplan-Meier method.

Results

115 patients were included, with 63 (55%) in Group 1 and 52 (45%) in Group 2. Median age was 61 and 53% of patients had visceral disease. Patients received a median 3 prior lines of metastatic treatment (interquartile range [IQR] 2–4.5) and 71% received prior CDK4/6 inhibitor; more patients in Group 2 received prior CDK4/6 inhibitor than Group 1 (87 vs. 59%, p = 0.001). In Group 1, 50 (79%) patients discontinued everolimus for progression and 13 (21%) for toxicity. In Group 2, 33 (63%) patients discontinued alpelisib for progression and 19 (37%) for toxicity. Median TTF1 was 9.2 months (95%CI 5.5–12) in Group 1 and 9.7 months (95%CI 7.1–14) in Group 2 (p = 0.5, Fig. 1, left). In patients who received prior CDK4/6 inhibitor, median TTF1 was 4.3 months (95%CI 3.2–11) in Group 1 and 8.7 months (95%CI 6.4–12) in Group 2 (p = 0.02). In univariable Cox proportional hazards analyses, factors associated with worse TTF1 included prior CDK4/6 inhibitor use and more prior lines of treatment; these remained significant in multivariable analyses. After adjusting for prior CDK4/6 inhibitor use and prior total/intervening lines of therapy, TTF1 was not significantly different between groups (p = 0.2), nor was TTF2 (p = 0.1). Presence of GATA3 mutation was significantly associated with poorer alpelisib-specific TTF (p = 0.016).

Conclusions

Prior CDK4/6 inhibitor use was associated with worse TTF1 and TTF2. There were no significant differences in TTF1 nor in TTF2 between Groups 1 and 2 after adjusting for clinical predictors. Given the recent approvals of capivasertib and inavolisib, prospective studies testing optimal sequencing of pathway agents with integration of biomarker analysis are needed.

Introduction

Approximately 70% of patients with metastatic breast cancer have hormone receptor-positive, human epidermal growth factor receptor 2 (HER2)-negative disease [1]. Standard of care first-line therapy for these patients consists of endocrine therapy combined with a cyclin-dependent kinase 4 and 6 (CDK4/6) inhibitor, which improves progression-free (PFS) and overall survival (OS) when compared with endocrine therapy alone [2, 3]. However, in most cases, disease progression occurs, necessitating a change in therapy.

Genomic testing can inform second-line treatment options, which include agents that target the phosphatidylinositol 3-kinase (PI3K)/AKT/mTOR signaling pathway. Approximately 50% of patients with hormone receptor-positive/HER2-negative breast cancer have activating PIK3CA or AKT1 mutations or PTEN loss of function alterations [4, 5]. Everolimus, an mTOR inhibitor, was the first pathway agent to be given regulatory approval in 2012 following the BOLERO-2 study, which demonstrated improved median investigator-assessed PFS of 6.9 months in combination with exemestane vs. 2.8 months with exemestane alone (hazard ratio [HR] 0.43, 95%CI 0.35–0.54) [6]. Alpelisib, a PI3K inhibitor, was approved in patients with PIK3CA-mutated cancers in 2019 following the SOLAR-1 study, which demonstrated improved median investigator-assessed PFS in combination with fulvestrant of 11.0 months vs. 5.7 months with fulvestrant alone (HR 0.65, 95%CI 0.50–0.85) [7]. Capivasertib, an AKT inhibitor, was approved in 2023 for patients with PI3K/AKT/PTEN pathway alterations following the results of CAPItello-291 which showed median PFS of 7.3 months in combination with fulvestrant vs. 3.1 months in fulvestrant alone (HR 0.50, 95%CI 0.38–0.65) [8]. In the cohort of patients with or without pathway alterations who had prior CDK4/6 inhibitor exposure, median PFS was 5.5 months (95%CI 3.9–6.8 months). Notably, in BOLERO-2 all patients were CDK4/6 inhibitor-naïve and in SOLAR-1, 5.9% had received prior CDK4/6 inhibitor treatment. The BYLieve study of patients treated with alpelisib following prior CDK4/6 inhibitor demonstrated median PFS of 8.0 months (95%CI 5.6–8.6) with fulvestrant plus alpelisib and median PFS of 5.6 months (95%CI 3.7–7.1) with aromatase inhibitor plus alpelisib [9, 10]. Most recently, inavolisib, another PI3K inhibitor, was approved in 2024 as first-line treatment in combination with fulvestrant and palbociclib in patients who had disease recurrence on or within 12 months of completing adjuvant endocrine therapy and whose tumors harbored PIK3CA mutation [11].

Whether there is clinically meaningful efficacy from treating with more than one pathway agent remains unclear; furthermore, the optimal sequencing of pathway agents is unknown. Here, we sought to describe our single-institution experience of patients with PIK3CA mutations treated with both everolimus and alpelisib in either sequence, in combination with any endocrine therapy agent.

Methods

Study cohort

This is an observational, retrospective study of adult patients with PIK3CA-mutant HR + MBC who received both everolimus and alpelisib in either sequence from 2012 to 2023 at Memorial Sloan Kettering Cancer Center (MSKCC). Patients were identified for inclusion by everolimus and alpelisib prescription dates and verified to have ≥ 1 PIK3CA mutation on tissue- or blood-based genomic testing, either on an institutional next generation sequencing (NGS) platform or external testing. Patients were excluded if a prescription was written but upon review of the electronic medical record they never received both everolimus and alpelisib or if they were treated outside of MSKCC. This study was approved by the MSKCC Institutional Review Board. Start and end dates of everolimus and alpelisib were determined from electronic prescribing records and dates were adjusted based on manual chart review. Demographic, clinicopathologic, and treatment data were abstracted from medical records. The date of disease progression following treatment with each pathway agent was recorded and was defined as radiographic or clinical progression leading to treatment change. For patients in whom everolimus and/or alpelisib were discontinued for toxicity, the date of next treatment was recorded. Patients treated first with everolimus and subsequently with alpelisib were categorized as Group 1 while patients treated first with alpelisib and subsequently with everolimus were categorized as Group 2. TTF1 was defined as time from initiation of first pathway agent to treatment discontinuation for progression, or to initiation of next treatment in patients who discontinued first pathway agent for toxicity. Patients who discontinued first pathway agent for toxicity were censored at the time of initiation of next treatment while those discontinuing for progression were counted as events. TTF2 was defined as time from initiation of second pathway agent to treatment discontinuation for progression, or to initiation of next treatment in patients who discontinued second pathway agent for toxicity, or date of last follow up in patients with ongoing treatment. Patients discontinuing second pathway agent for progression were counted as events; patients who discontinued second pathway agent for toxicity were censored at the time of initiation of next treatment or last follow up, and patients who were still receiving treatment with second pathway agent at the end of the study follow up period were censored at their date of last follow-up. For genomic analyses, results of NGS testing from archival tissue using the MSK-IMPACT platform prior to initiation of each pathway agent were recorded.

Statistical analyses

Differences in baseline characteristics between groups were compared using the chi-squared test or Fisher’s exact test for categorical variables and Wilcoxon ranks sum test for continuous variables. Median TTF1 and TTF2 were computed using the Kaplan-Meier method and compared with the log-rank test; TTF1 and TTF2 curves were generated using the Kaplan-Meier method. Associations of patient characteristics and prior treatment with TTF1 and TTF2 between groups were analyzed using univariable and multivariable Cox proportional hazards regression models. Multivariable Cox proportional hazards models were used to estimate adjusted HRs for TTF1 and TTF2. Multivariable analyses (MVAs) were conducted for patients with complete data for all covariates. For descriptive statistics, percentages and means or medians with interquartile range (IQR) were used, unless otherwise noted. Statistical significance was set at a two-sided p-value < 0.05, and binomial 95% confidence intervals (CIs) were calculated. All analyses were performed using R version 4.4.1 with the tidyverse, gtsummary, Hmisc, survival, survminer, and tidycmprsk packages; genomic analyses were performed using packages cbioportalR and gnomeR.

Results

115 patients were included in this study, with 63 (55%) in Group 1 and 52 (45%) in Group 2. Baseline characteristics for the total cohort and by group are shown in Table 1. Patients in Group 2 had fewer prior lines of treatment in the metastatic setting (p = 0.07), more commonly had prior CDK4/6 inhibitor exposure (p = 0.001), and a lower number of intervening lines of treatment between pathway agents (p < 0.001). A greater proportion of patients in Group 1 had intervening chemotherapy use (p < 0.001) between pathway agents. Regarding CDK4/6 inhibitor use, 70 patients (61%) received a CDK4/6 inhibitor prior to the first pathway agent, 12 (10%) received a CDK4/6 inhibitor both prior and as an intervening line of treatment between pathway agents, while 20 (17%) received it only intervening. In Group 1, 29 (46%) received a CDK4/6 inhibitor prior to the first pathway agent, 18 (29%) received it only as an intervening treatment, and 8 (13%) received it both prior and intervening. In Group 2, 41 (79%) received prior CDK4/6 inhibitor, 2 (4%) received it only as an intervening treatment, and 4 (8%) received it both prior and intervening; overall CDK4/6 inhibitor timing differed between the groups (p = 0.001).

Table 1.

Baseline characteristics of the study cohort

Characteristic	Overall	Group 1 (Everolimus first)	Group 2 (Alpelisib first)	p-value
Age, median (IQR)	61 (54–68)	60 (51–68)	63 (56–68)	0.3
Menopausal status, number (%)				0.13
Premenopausal	30 (26%)	20 (32%)	10 (19%)
Postmenopausal	85 (74%)	43 (68%)	42 (81%)
Visceral involvement at diagnosis, number (%)	61 (53%)	36 (57%)	25 (48%)	0.3
Bone-only disease at diagnosis	40 (35%)	24 (38%)	16 (31%)	0.4
Number of prior treatment lines, median (IQR)	3 (2–5)	4 (2–5)	3 (2–4)	0.067
Number of prior endocrine therapies, median (IQR)	2 (1–2)	2 (1–3)	1 (1–2)	0.14
Number of prior chemotherapies, median (IQR)	0 (0–1)	0 (0–1)	0 (0–1)	0.6
Any prior CDK4/6i use, number (%)	82 (71%)	37 (59%)	45 (87%)	0.001
Number of intervening treatment lines, median (IQR)	1 (0–3)	2 (1–3)	0 (0–1)	< 0.001
Intervening chemotherapy use, number (%)	56 (49%)	41 (65%)	15 (29%)	< 0.001
Intervening CDK4/6i use, number (%)	32 (28%)	26 (41%)	6 (12%)	< 0.001

In Group 1, endocrine therapy combination with everolimus was exemestane in 50 (79%), fulvestrant in 6 (9.5%), tamoxifen in 5 (7.9%), and any other aromatase inhibitor in 2 (3%); endocrine therapy combination with subsequent alpelisib was fulvestrant in 48 (76%), aromatase inhibitor in 12 (19%), and tamoxifen in 3 (5%). In Group 2, endocrine therapy combination with alpelisib was fulvestrant in 44 (85%) and aromatase inhibitor in 8 (15%); endocrine therapy combination with subsequent everolimus was exemestane in 26 (50%), fulvestrant in 17 (33%), tamoxifen in 7 (14%), and other aromatase inhibitor in 2 (4%). In Group 1, 50 (79%) of patients discontinued everolimus for progression and 13 (21%) for toxicity; subsequently 50 (79%) discontinued alpelisib for progression and 10 (16%) for toxicity. In Group 2, 33 (63%) patients discontinued alpelisib for progression and 19 (37%) for toxicity; subsequently 44 (85%) discontinued everolimus for progression and 4 (7.7%) for toxicity.

In the overall cohort regardless of pathway sequence, median TTF1 was 9.3 months (95%CI 7.1–12.0 months) and median TTF2 was 4.1 months (95%CI 3.5–5.1 months). By treatment sequence, Group 1 median TTF1 was 9.2 months (95%CI 5.5–12.0) with everolimus first and Group 2 median TTF1 was 9.7 months (95%CI 7.1–14.0) with alpelisib first (Fig. 1A, p = 0.5). In the subgroup of patients who received prior CDK4/6 inhibitor, TTF1 was 4.3 months (95%CI 3.2–11.0) in Group 1 and 8.7 months (95%CI 6.4–12.0) in Group 2 (Fig. 1B, p = 0.02). In univariable analyses, factors associated with worse TTF1 included prior CDK4/6 inhibitor use (HR 2.5, 95%CI 1.5–4.2, p < 0.001) and more prior treatment lines (HR 1.1, 95%CI 1.01–1.2, p = 0.03). There were no significant associations for TTF1 with age, menopausal status, visceral disease, bone-only disease, number of prior endocrine therapies, number of prior chemotherapies, or duration of CDK4/6 inhibitor. In the multivariable model, both prior CDK4/6 inhibitor use (HR 3.6, 95%CI 2.0-6.3, p < 0.001) and total prior treatment lines (HR 1.1, 95%CI 1.0-1.3, p = 0.01) remained significantly associated with TTF1 after adjusting for menopausal status and visceral disease, while treatment group was not significant (p = 0.2). In the multivariable model restricted to patients with prior CDK4/6 inhibitor use, total prior treatment lines (HR 1.29 95%CI 1.05–1.57, p = 0.01) and ET partner (aromatase inhibitor vs. fulvestrant HR 0.47, 95% CI 0.25–0.90, p = 0.02) were significantly associated with TTF1 while the difference in TTF1 between groups was also non-significant in this sub-group (p > 0.9).

Fig. 1.

Time to treatment failure (TTF) on pathway agents

Median TTF2 was 5.0 months (95%CI 4.6–7.9) in Group 1 and 3.5 months (95%CI 3.0-4.1) in Group 2 (Fig. 1C, p = 0.005). In univariable analyses, prior CDK4/6 inhibitor use was associated with worse TTF2 (HR 1.7, 95%CI 1.1–2.7, p = 0.03), whereas total prior lines of treatment before the second pathway agent (HR 0.9, 95%CI 0.8-1.0, p = 0.03) and number of intervening treatments (HR 0.9, 95%CI 0.8-1.0, p = 0.04) were both associated with better TTF2. There were no significant associations with age, menopausal status, visceral disease, bone-only disease, or duration of CDK4/6 inhibitor in univariable analyses for TTF2. In the multivariable model adjusted for number of CDK4/6i lines prior to second pathway agent, total prior treatment lines, visceral disease, and menopausal status, treatment group was not significantly associated with TTF2 (HR 1.50, 95% CI 0.9–2.4, p = 0.10). In the subset of patients with prior CDK4/6 inhibitor use, patients in Group 2 (alpelisib before everolimus) trended towards having worse TTF2 (HR 1.72, 95%CI 1.0–3.0, p = 0.06) after adjusting for ET partner, CDK4/6i treatment duration, prior treatment lines, visceral disease, and menopausal status. Swimmer plots showing TTF1 and TTF2 for each group, denoted with reason for discontinuation (toxicity/ongoing vs. progression of disease) are shown in Fig. 2.

Fig. 2.

Swimmer plots of TTF1 and TTF2 within each group with reason for treatment stop denoted as toxicity/ongoing or progression

In the overall cohort, MSK-IMPACT NGS testing was available in 89 patients prior to alpelisib treatment and in 67 patients prior to everolimus treatment. The most prevalent mutations detected prior to alpelisib treatment were PIK3CA (86 patients, 97%), CDH1 (19 patients, 21%), ESR1 (18 patients, 20%), TP53 (17 patients, 19%), KMT2C (14 patients, 16%), MAP3K1 (11 patients, 12%), and GATA3 (10 patients, 11%). 5 patients (5.6%) had AKT1/PTEN alterations. The most prevalent mutations detected prior to everolimus treatment were PIK3CA (65 patients, 97%), CDH1 (18 patients, 27%), TP53 (13 patients, 19%), KMT2C (12 patients, 18%), ESR1 (11 patients, 16%), ARID1A (8 patients, 12%), and MAP3K1 (7 patients, 10%). The most common PIK3CA mutations were H1047R (34 samples, 37%), E545K (19 samples, 21%), E542K (5 samples, 5.4%), and H1047L (5 patients, 5.4%); 5 patients had ≥1 PIK3CA mutation.

In the subset of patients with genomic testing before first the respective agent, we evaluated TTF with each pathway agent individually regardless of treatment sequence and incorporated genomic data. In univariable analyses in patients with IMPACT testing pre-everolimus, TTF on everolimus was significantly worse with CDK4/6 inhibitor use prior to the first pathway agent and significantly better with increasing age (HR 1.6, 95% CI 1.0-2.7, p = 0.048 and HR 0.9, 95%CI 0.95-1.0, p = 0.03 respectively). In the multivariable model adjusted for total treatment lines prior to everolimus, prior CDK4/6 inhibitor use, visceral involvement at diagnosis, CDH1 mutation, and menopausal status, being in Group 2 (receiving everolimus second) trended toward worse TTF on everolimus (HR: 1.8, 95%CI 1.0-3.3, p = 0.06).

In univariable analyses in patients with IMPACT testing pre-alpelisib, TTF on alpelisib was significantly better in Group 2 (receiving alpelisib first) (HR 0.6, 95%CI 0.32–0.93, p = 0.03). Patients with GATA3 mutations trended towards having worse TTF on alpelisib (univariable HR 2.1, 95%CI 1.0-4.7, p = 0.06), whereas GATA3 mutations were not associated with TTF on everolimus (p = 0.9). In the multivariable model, prior CDK4/6 inhibitor use and GATA3 mutation were significantly associated with poorer alpelisib-specific TTF (HR 2.1, 95% CI1.1-4.2, p = 0.03 and HR 3.0, 95%CI 1.2–7.2, p = 0.02 respectively), after adjusting for treatment group, total treatment lines prior to alpelisib, visceral involvement, and menopausal status. None of the other prevalent gene mutations nor the H1047R and E545K PIK3CA mutations were associated with difference in TTF on alpelisib or everolimus. Among the 5 patients with AKT1/PTEN alterations on pre-alpelisib IMPACT testing, median TTF on alpelisib was 2.7 months (95%CI 2.6-NR).

Discussion

In this observational study, median TTF1 with the first pathway agent was 9.3 months and median TTF2 was 4.1 months. TTF1 was significantly worse among patients who received prior CDK4/6 inhibitor and greater total treatment lines prior to first pathway agent, but did not differ significantly between groups when adjusted for prior CDK4/6 inhibitor use, prior treatment lines, visceral involvement, and menopausal status. Median TTF2 was 5.0 months in Group 1 and 3.5 months in Group 2; prior CDK4/6 inhibitor use was associated with worse TTF2. TTF2 was not significantly different between groups in the multivariable model, although there was a trend toward worse TTF2 in Group 2 in the subset of patients with prior CDK4/6 inhibitor use (p = 0.06). Notably, the two groups differed significantly in baseline treatment characteristics including number of treatment lines in the metastatic setting prior to first pathway agent, proportion of patients with prior CDK4/6 inhibitor use, and number of intervening treatment lines between pathway agents. TTF on each agent individually appeared better when it was given first in the sequence. GATA3 mutation was associated with worse alpelisib-specific TTF, but did not affect everolimus-specific TTF.

The activity of PI3K/AKT/mTOR pathway agents following prior first line CDK4/6 inhibitor use is unclear, as prior CDK4/6 inhibitor use in BOLERO-2 and SOLAR-1 were 0 and 5.9%, respectively [6, 7]. In observational cohort studies, the median PFS of exemestane and everolimus in the subset of patients who received prior CDK4/6 inhibitor ranges from 3.6 to 6.8 months [12–14]. BYLieve was a prospective phase 2 trial testing alpelisib and fulvestrant following aromatase inhibitor plus CDK4/6 inhibitor; median PFS was 8.0 months (95%CI 5.6–8.6 months) [15]. In our study, median TTF1 in patients who received prior CDK4/6 inhibitor was consistent with these data, with 4.3 months with everolimus and 8.7 months with alpelisib. The differences in prior CDK4/6 inhibitor use and total prior lines of treatment between groups is likely explained by the regulatory approvals of everolimus in 2012, the first CDK4/6 inhibitor palbociclib in 2015, and alpelisib in 2019. Since everolimus was approved prior to CDK4/6 inhibitors, fewer patients in Group 1 had received CDK4/6 inhibitor prior to first pathway agent; 14/63 (22%) of patients in Group 1 started everolimus prior to FDA approval of the first CDK4/6 inhibitor, palbociclib. Patients in Group 1 also had greater numbers of prior treatments before alpelisib approval in 2019.

Endocrine therapy agent selection may have impacted TTF2. In Group 1, most patients received exemestane with everolimus followed by switch to fulvestrant with alpelisib. In Group 2, on the other hand, while most patients received fulvestrant with alpelisib, endocrine therapy partner to everolimus was more heterogeneous, with 54% aromatase inhibitor use and only 33% continuing with fulvestrant. Furthermore, 21% of our study population had ESR1 mutations as determined by tissue-based NGS testing, which may underestimate true prevalence as we did not include external blood-based NGS testing in the genomic analyses. Presence of ESR1 mutation is associated with more aggressive disease biology and poorer outcomes, although biomarker analyses of BOLERO-2 and SOLAR-1 demonstrated similar benefit to the addition of everolimus and alpelisib respectively to endocrine therapy in patients with ESR1 mutations compared to endocrine therapy alone [16, 17]. Overall, median TTF2 in our study was modest, particularly in patients who received prior CDK4/6 inhibitor, which is now standard first-line therapy. Most patients with a PIK3CA mutation will now receive fulvestrant with either alpelisib or capivasertib in the second line. In this setting, third-line treatment with an aromatase inhibitor + everolimus appears to offer limited clinical benefit; clearly, biomarkers to identify patients who continue to benefit from endocrine therapy and targeting of the PI3K/AKT/mTOR pathway are needed.

Various mechanisms of resistance to PI3K inhibition have been elucidated, including insulin feedback and reactivation of IGF-1R, PTEN loss, and reactivation of AKT and mTOR [18]. GATA3 loss of function has been implicated in AKT activation, consistent with our study findings of worse alpelisib-specific TTF in patients with somatic GATA3 mutations [19, 20]. In a small real-world study of 13 patients who received alpelisib following everolimus, GATA3 mutation presence vs. absence was not associated with different clinical benefit rate [21]. In the setting of loss-of-function PTEN alterations or AKT1 activating mutations, mTOR inhibition may remain a viable therapeutic strategy; this is supported by our finding that GATA3 mutation did not affect everolimus-specific TTF. Following mTOR inhibition, potential mechanisms that lead to resistance include AKT activation, increased ERK/MAPK signaling, dysregulation of eIF4E, modulation of apoptotic regulators, and increased angiogenesis [22, 23]. Inhibition of mTORC1 may increase mTORC2 activity and may upregulate insulin-like growth factor-1 receptor, both of which lead to AKT activation [24, 25]. In the latter case, subsequent PI3K inhibition may be clinically meaningful, despite PI3K being upstream of mTORC1. Following everolimus use in Group 1 in our study, median TTF2 with alpelisib was 5.0 months and there was a subset of patients who derived longer TTF from alpelisib than everolimus, even after discontinuing everolimus for disease progression rather than toxicity. There were no genomic alterations significantly associated with everolimus-specific TTF.

There are several limitations to our study. First, the retrospective, observational nature of this study precluded controlling for additional confounding factors and our results are hypothesis-generating. Second, given the number of years between regulatory approval of everolimus and alpelisib, Group 1 may have been subject to selection bias for patients with slower-paced disease who were able to receive greater number of therapies and/or ongoing endocrine-based treatment regimens. Third, genomic testing was available for only a subset of patients prior to the start of first pathway agent. Despite these limitations, to our knowledge, this is the largest observational study to-date of patients treated sequentially with PI3K/AKT/mTOR pathway inhibitors and provides evidence for modest but potentially meaningful activity of using a second pathway agent following progression. Furthermore, identification of prior CDK4/6 inhibitor use, greater prior lines of treatment, and lack of GATA3 mutation can inform optimal patient selection for use of more than one pathway agent in sequence.

In conclusion, prior CDK4/6 inhibitor use was associated with worse TTF1 and TTF2. There were no significant differences in TTF1 nor in TTF2 between Groups 1 & 2 after adjusting for clinical predictors. These results may have potential implications for use of recently approved pathway inhibitors capivasertib and inavolisib, and for novel PI3K inhibitors in development such as RLY-2608 (Relay, NCT05216432) and STX-478 (Scorpion, NCT05768139). Prospective studies testing optimal sequencing of pathway agents that incorporate biomarker analysis are urgently needed.

Author contributions

S.S. and K.J. wrote the main manuscript text. M.B. and Y.C. performed statistical analyses and prepared figures. S.S., J.M., A.S., M.A., and P.R. performed or assisted with data collection. All authors reviewed the manuscript.

Funding

This work was supported in part through the National Institutes of Health/National Cancer Institute (NIH/NCI) Cancer Center Support Grant P30 CA008748. SS is supported by the Clinical and Translational Science Center at Weill Cornell Medical Center and Memorial Sloan Kettering Cancer Center CTSA UL1TR00457 grant and the Terri Brodeur Breast Cancer Foundation.

Data availability

Data are available upon written request to the study corresponding author.

Declarations

Ethics approval and consent to participate

This study was approved by the Institutional Review Board of Memorial Sloan Kettering Cancer Center.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.