ER+, HER2− advanced breast cancer treated with taselisib and fulvestrant: genomic landscape and associated clinical outcomes

Taselisib is a potent β‐sparing phosphatidylinositol 3‐kinase (PI3K) inhibitor that, with endocrine therapy, improves outcomes in phosphatidylinositol‐4,5‐bisphosphate 3‐kinase catalytic subunit alpha (PIK3CA)‐mutated (PIK3CAmut) advanced breast cancer. To understand alterations associated with response to PI3K inhibition, we analysed circulating tumour DNA (ctDNA) from participants enrolled in the SANDPIPER trial. Participants were designated as either PIK3CAmut or PIK3CA no mutation was detected (NMD) per baseline ctDNA. The top mutated genes and tumour fraction estimates identified were analysed for their association with outcomes. In participants with PIK3CAmut ctDNA treated with taselisib + fulvestrant, tumour protein p53 (TP53; encoding p53) and fibroblast growth factor receptor 1 (FGFR1) alterations were associated with shorter progression‐free survival (PFS) compared to participants with NMD in these genes. Conversely, participants with PIK3CAmut ctDNA harbouring a neurofibromin 1 (NF1) alteration or high baseline tumour fraction estimate experienced improved PFS upon treatment with taselisib + fulvestrant compared to placebo + fulvestrant. Broadly, alterations in oestrogen receptor (ER), PI3K and p53 pathway genes were associated with resistance to taselisib + fulvestrant in participants with PIK3CAmut ctDNA. Altogether, we demonstrated the impact of genomic (co‐)alterations on outcomes with one of the largest clinico‐genomic datasets of ER+, HER2−, PIK3CAmut breast cancer patients treated with a PI3K inhibitor.


Introduction
The phosphatidylinositol 3-kinase (PI3K) pathway is one of the most frequently dysregulated signalling pathways in human cancers and is involved in cell growth and proliferation.Activating mutations in PIK3CA, the gene that encodes the p110a catalytic subunit of the Class I PI3K enzyme, leads to dysregulation of PI3K signalling [1].Across solid tumours, PIK3CA mutations (PIK3CAmut) are commonly observed in hormone receptor-positive, HER2-negative (HR+, HER2À) breast cancers, with a prevalence of ~40% [2,3].
The clinical implementation of most PI3K inhibitors to date has been challenged by the toxicities induced by the inhibition of multiple p110 isoforms, particularly pan-isoform inhibitors [4].Taselisib is a potent and selective b-sparing PI3K inhibitor [5] that was investigated in combination with fulvestrant in the phase III clinical trial, SANDPIPER, for patients with oestrogen receptor-positive (ER+), HER2À locally advanced or metastatic breast cancer (NCT02340221) [6].The study met its primary endpoint of improved progression-free survival (PFS) with taselisib plus fulvestrant over placebo plus fulvestrant albeit with modest clinical activity (7.4 vs. 5.4 months; HR = 0.70) [6].
Circulating tumour DNA (ctDNA) is emerging as an important tool that appears representative of the DNA and the overall tumour mutational landscape of a patient's disease state [7,8] and, for patients with ER+, HER2À advanced breast cancer, has prognostic and predictive response value [6,9,10].Moreover, the low-risk, non-invasive procedure of a blood draw reduces the challenges associated with longitudinal sampling that exists for tumour biopsies.Ultimately, next-generation sequencing of ctDNA leverages the relatively high concentrations of ctDNA detected in patients with advanced cancer [11] to enable studies that address questions about disease biology, tumour heterogeneity and mechanisms of resistance to targeted therapies, which were previously impractical because of the infeasibility of repeated tumour biopsies.
To investigate the potential alterations associated with response and/or resistance to endocrine therapy and/or PI3K inhibition in ER+, HER2À breast cancer tumours, we analysed baseline and end-of-treatment ctDNA collected from SANDPIPER participants.

Study design and participants
The study design has been previously described [6].Briefly, SANDPIPER was a randomized, doubleblind, placebo-controlled, phase III trial evaluating the efficacy and safety of taselisib plus fulvestrant (TAS + FUL) versus placebo plus fulvestrant (PBO + FUL) in post-menopausal women with ER+, HER2À locally advanced or metastatic breast cancer who had disease recurrence or progression during or after aromatase inhibitor therapy.Participants were randomized (2 : 1) to either the TAS + FUL or control (PBO + FUL) arm.
The SANDPIPER study protocol, including a description of the exploratory biomarker analyses of all randomized participants with and without PIK3CAmut tumours, was approved by the relevant Institutional Review Board/Ethics Committee at each participating centre prior to study initiation.The trial conformed to the Good Clinical Practice guidelines, the Declaration of Helsinki and applicable local laws.All participants provided signed informed consent, which included consent for the biomarker analyses included in this study.

Plasma ctDNA collection
Plasma samples were collected at different time points throughout the study.Herein, we present data from samples taken at baseline, defined as before the first dose of study treatment (pre-dose at Cycle 1 Day 1) and at the end of study treatment (EOT), defined as the time-point when a participant ceases the study treatment for reasons including but not limited to disease progression, toxicities, administrative reasons or at the investigator's discretion.

Comprehensive genomic profiling
Comprehensive genomic profiling of plasma samples using the FoundationOne Ò Liquid (F1L) assay [12] was performed in a CLIA-certified, CAP-accredited laboratory (Foundation Medicine Inc., Cambridge, MA, USA) using hybrid-capture, adapter ligationbased libraries to identify genomic alterations (base substitutions, small insertions and deletions, copy number alterations and rearrangements/fusion events) for 70 cancer-related genes.Processing of the sequence data and identification of different classes of genomic alterations were performed as previously described [13].Unless otherwise indicated, the analysis is focused on alterations predicted to be pathogenic, defined as of known or likely oncogenic significance.Participants with PIK3CAmut were defined as those with ≥ 1 pathogenic single-nucleotide variant in the PIK3CA gene detected in baseline ctDNA.Participants with PIK3CA no mutation detected (NMD) were defined as those without the detection of a pathogenic single-nucleotide variant in the PIK3CA gene in baseline ctDNA.
Baseline plasma samples from 598 (94.8%) participants underwent genomic profiling, of which 508 (80.5%) samples were successfully sequenced and 90 samples (14.3%) failed processing.Baseline plasma samples were not collected from 33 (5.2%) participants due to withdrawal of consent or local regulatory testing restrictions.EOT plasma samples underwent genomic profiling for a subset of patients (15.8%; n = 100) who exhibited clinical benefit, defined as the best overall response by the investigator of partial response, complete response or stable disease with an extended PFS of > 7 months on study treatment with evidence of tumour shrinkage.
Quantification of the ctDNA fraction derived from tumour cells in blood/plasma samples was performed as previously described (unpublished data) [14] using two complementary methods: the proprietary tumour fraction estimator (TFE) and the maximum somatic allele frequency (MSAF) method.Tumour fraction estimator is based on a measure of tumour aneuploidy, and MSAF uses allele fraction from somatic coding alterations to estimate the ctDNA fraction.A high TFE was defined as ≥ 10% based on prior work that suggested 10% correlated with the proportion of tumour DNA adequate for high-confidence copy number calls [15].

Statistical analyses
Statistical analysis, computation and plotting were performed using R version 3.6.1.Progression-free survival data, from the clinical cut-off date for the primary analysis (15 October 2017) [6], were analysed descriptively for each biomarker subgroup (mutated and NMD), and for each treatment arm (PBO + FUL and TAS + FUL).Kaplan-Meier survival analyses were performed with the log-rank test using a Cox proportional hazards regression model to obtain hazard ratios and 95% confidence intervals.All statistical tests were two-sided.To adjust for multiple comparisons, the Benjamini-Hochberg correction was used, wherein statistical significance was defined as a Benjamini-Hochberg-adjusted P-value (q-value) < 0.05.

Analysis population and participant demographics
As reported previously, 631 participants were enrolled and randomly assigned (2 : 1) to either the taselisib plus fulvestrant (TAS + FUL) or placebo plus fulvestrant (PBO + FUL) arm of SANDPIPER.For the clinical trial, participants were designated to either the PIK3CAmut or PIK3CA no mutation detected (NMD) cohort based on a centralized cobas Ò PIK3CA Mutation Test result from formalin-fixed paraffin-embedded tissue collected prior to randomization, an eligibility criterion for all enrolled participants [6].However, for the analysis herein, PIK3CA mutation status was solely based on results from baseline plasma using the F1L assay.Of note, the majority of participants in SANDPIPER (97.0%) were cyclin-dependent kinase 4/6 (CDK4/6) inhibitor-na€ ıve given that they were randomized between 2015 and 2017 [6], which coincided with the initial approvals of CDK4/6 inhibitors for HR+, HER2À breast cancer treatment.
For the analyses herein (Fig. 1A), PIK3CA mutation status (PIK3CAmut or PIK3CA NMD) was based on the comprehensive genomic profiling (CGP) of baseline plasma samples using the F1L assay.In the 508 (80.5%) participants for whom PIK3CA mutation status is known from both the tissue-based cobas Ò PIK3CA Mutation Test and the baseline ctDNA-based F1L assay, the sensitivity is 78.2% and the specificity is 86.7% [6].These metrics may be reflective of the methodological differences in the assays (PCR vs. NGS) and the temporal differences in when the samples were collected; specifically, plasma samples were freshly collected during the study screening period, whereas tumour tissue samples were largely archival samples, many from the time of primary diagnosis.Of the 339 participants with tumours classified as ctDNA PIK3CAmut, 103 (30.4%) and 236 (69.6%) were treated with PBO + FUL and TAS + FUL respectively.Of the 169 participants with tumours classified as ctDNA PIK3CA NMD, 68 (40.2%) and 101 (59.8%) were treated with PBO + FUL and TAS + FUL, respectively.Demographics and disease characteristics were largely similar between participants with PIK3-CAmut and PIK3CA NMD ctDNA between the treatment arms (Table S1), which was consistent with the total SANDPIPER population [6].Of note, bone metastases (inclusive of participants with and without bone-only disease) were found to occur at a higher frequency in participants with pathogenic alterations detected in baseline ctDNA than in participants with no pathogenic alterations detected in baseline ctDNA [77.8% (n = 369/474) vs. 38.2%(n = 13/34); P < 0.001, Fisher's Exact Test].

Genomic analysis of baseline ctDNA and association with clinical outcomes
To evaluate whether any of the top altered genes in the participants' baseline ctDNA correlated with prognosis or treatment outcomes, we next analysed the association between genomic alteration status and PFS within each of the treatment arms.Consistent with the previously reported hazard ratios (HRs) [6], the PFS HR based on PIK3CA ctDNA status was 0.65 between treatment arms in the entire PIK3CAmut cohort and 0.85 between treatment arms for the entire PIK3CA NMD cohort.

Baseline tumour fraction estimate and clinical outcomes
Recent studies have reported that a larger treatment effect is observed in ER+, HER2À advanced breast cancer patients treated with ET when outcomes are analysed by PIK3CAmut positivity in ctDNA compared to the same analysis by tissue PIK3CAmut positivity [6,10,16].Here, we further dissected the ctDNApositive population using the baseline tumour fraction estimate (TFE), a measure of the quantity of ctDNA shed from the tumour into circulation.

Genomic landscape of EOT ctDNA in participants who experienced clinical benefit from study treatment
End-of-treatment (EOT) plasma samples were submitted for F1L genomic profiling from 100 participants with PIK3CAmut tumours who exhibited clinical benefit.F1L data were evaluable for both the baseline and EOT ctDNA samples for 72 of these 100 participants.Of the 72 participants with paired evaluable ctDNA, 54 of those participants' baseline ctDNA harboured a PIK3CAmut, of which 12 (22.2%)were treated with PBO + FUL and 42 (77.8%)were treated with TAS + FUL; 18 of those participants' baseline ctDNA were PIK3CA NMD, of which 6 (33.3%) were treated with PBO + FUL and 12 (66.7%)were treated with TAS + FUL (Fig. 5A).

Paired baseline and EOT ctDNA analysis of PIK3CAmut participants who experienced clinical benefit from study treatment
To investigate the impact of the study treatment on the genomic landscape of EOT ctDNA in participants who experienced clinical benefit, we compared the genomic alterations that were detected in EOT ctDNA but not detected in baseline ctDNA.Amongst the PIK3CAmut participants, the top newly detected alterations upon PBO + FUL treatment were in ESR1 (33.3%),ERBB2 (16.7%) and TP53 (16.7%); the top newly detected alterations upon TAS + FUL treatment were in ESR1 (28.6%),TP53 (16.7%) and PTEN (9.5%) (Fig. 5B,C and Fig. S7C).Amongst the PIK3CA NMD participants, the newly detected alterations upon PBO + FUL treatment were in PIK3CA (50.0%),CDH1 (16.7%),CHEK2 (16.7%) and ESR1 (16.7%); the top newly detected alterations upon TAS + FUL treatment were in PIK3CA (41.7%),ESR1 (33.3%),ATM (16.7%),CDH1 (16.7%),CHEK2 (16.7%) and FGFR1 (16.7%) (Fig. S8C,D).Within each of the PIK3CAmut and PIK3CA NMD cohorts, no statistically significant difference was observed in the alteration rates of individual genes newly detected at EOT between the treatment arms (q > 0.05, Fisher's Exact Test).Lastly, to study the scope of newly detected mutations specifically in PI3K pathway genes, we assessed the changes in PIK3CA, AKT1 and PTEN detected in EOT ctDNA compared to baseline ctDNA in TAS + FUL-treated participants.We observed the detection of two new AKT1 mutations, eight new PTEN mutations and 10 unique new PIK3CA mutations (Fig. 6A-C).Of note, each of the participants whose EOT ctDNA harboured these newly detected mutations in the PI3K pathway genes was found to also have additional non-PI3K pathway genomic alterations.

Discussion
As the SANDPIPER population had received prior aromatase inhibitor therapy and was predominantly CDK4/6-inhibitor na€ ıve, the genomic landscape of baseline ctDNA was expected to reflect that of an endocrine-resistant population.In both the PIK3CA mut and PIK3CA NMD cohorts, the top altered genes detected in baseline ctDNA included genes associated with the ER (ESR1), PI3K (PIK3CA, PTEN), p53 (TP53, ATM, CHEK2), MAPK (NF1), and/or receptor tyrosine kinase (FGFR1) signalling pathways (Fig. 1B and Fig. S2C).These observations are consistent with the molecular profiles of endocrine-resistant ER+, HER2À breast cancer tumours [3,17,18].Resistance resulting from ESR1 mutations has been wellcharacterized in hormone-refractory disease [19,20] and has been shown to correlate with worse outcomes following ET [21,22].Activation of PI3K signalling through mutations in PIK3CA, AKT1 or PTEN has been shown to confer endocrine resistance in vitro [23,24].Moreover, in vitro data have shown that increased MAPK pathway signalling can promote the loss of ERa expression in breast cancer tumours [25], which may contribute to the poor responses to ET observed in patients whose tumours harbour alterations in genes associated with the MAPK pathway [3].Of note, the prevalence of TP53 alterations detected in baseline ctDNA was higher than that reported in another analysis of advanced HR+, HER2À breast cancer tumour tissue [3] and may be due to the inclusion of TP53 somatic mutations that occur in clonal haematopoiesis, which is detectable in peripheral blood [26].
To expand our understanding of ER+, HER2À breast cancer disease biology, we investigated the association of genomic alterations with PFS to identify patient populations who may be intrinsically resistant to study treatment.Our analysis showed that in PBO + FUL-treated participants with PIK3CAmut baseline ctDNA, alterations in TP53, PTEN and BRAF trended towards poor prognosis.In TAS + FULtreated participants with PIK3CAmut baseline ctDNA, alterations in TP53 and FGFR1 were associated with poor prognosis, and alterations in PTEN, AKT1, GNAS and MYC trended towards poor prognosis.
Numerous studies have demonstrated that TP53 alterations confer clinical resistance to ET [17,[27][28][29], and in vitro studies have suggested that TP53 mutations may confer resistance to PI3K inhibition [30].Furthermore, the loss of the tumour suppressor PTEN, a negative regulator of the PI3K pathway, is associated with clinical resistance to ET [31] and p110a inhibition [32,33].Mutations in PIK3CA and PTEN have been described to be largely mutually exclusive in breast cancers [34]; therefore, we speculated that the detection of cooccurring PIK3CA and PTEN mutations in 6.3% (n = 32/508) of the baseline ctDNA samples from SANDPIPER participants (Fig. 1B) may indicate these alterations are derived from different cancer cells in a single tumour or from entirely different lesions.Recently, FGFR1 alterations have been shown to be associated with worse prognosis compared to FGFR1 NMD in patients with PIK3CAmut tumours treated with ET [35].Of note, in the same analysis, patients with PIK3CAmut tumours that harbour a co-occurring FGFR1 alteration experienced improved PFS upon treatment with alpelisib + FUL compared to PBO + FUL [35].Whereas this latter observation is the opposite association compared to our study, we postulate it may be in part to the small sample sizes of the biomarker-positive cohorts, in addition to differences in the timing of sample collection (i.e.freshly collected blood sample vs. archival or freshly collected tumour tissue sample) and methodologies (e.g.blood-vs.tissuebased NGS assays, challenges associated with calling somatic copy number variation in samples with low fractions of ctDNA) used to determine alteration statuses of the participants.Future studies are warranted to further determine the role of FGFR1 alterations and response to p110a inhibition in PIK3CAmut tumours.
Our search for alterations that may be predictive of response to study treatment identified that PIK3CAmut participants with a co-occurring NF1 alteration in baseline ctDNA experienced significantly improved PFS upon treatment with TAS + FUL compared to PBO + FUL (mPFS: 5.7 vs. 1.9 months; HR = 0.28, 95% CI 0.11-0.67;q = 0.017; Figs 2C and 3D).The trend of participants with NF1 alterations exhibiting shorter PFS compared to participants with NF1 NMD upon PBO + FUL treatment (mPFS: 1.9 vs. 3.7 months; HR = 2.1, 95% CI 1.1-4.1;P = 0.053; q = 0.35; Fig. 3D, Fig. S3A) is consistent with prior studies that have identified neurofibromin (NF1) inactivation as a resistance mechanism to ET [3,27,36].Preclinical data have shown that NF1 functions as a dual repressor of Ras signalling and ER transcriptional activity, and the loss of NF1 results in oestradiol hypersensitivity, contributing to ET resistance [37].Our analysis suggests that PI3K inhibition may overcome the negative impact of NF1 loss in ER+, HER2À advanced breast cancers.While the mechanism remains to be elucidated in breast cancer, prior work suggests that NF1 regulates the proliferation of neural stem cells in a PI3K-dependent manner [38].
To expand beyond genomic alterations, we evaluated the association between ctDNA levels and clinical outcome.We found that PIK3CA mut participants tended to have higher baseline TFE compared to PIK3CA NMD participants.In addition, we demonstrated that participants with high baseline TFE experienced worse PFS compared to participants with low baseline TFE across study treatments, regardless of PIK3CA mutation status in baseline ctDNA.This association between high baseline ctDNA levels and worse clinical outcomes was similarly observed in patients in the PALOMA-3 study (NCT01942135) [17], EVE biomarker study (NCT02109913) [39], and in samples collected from an independent cohort of patients with ER+, HER2À metastatic breast cancer [40].Moreover, this association has been demonstrated in patients with breast cancer in the neoadjuvant [41] and adjuvant settings [42], as well as across multiple cancer types [43].The prognostic value of ctDNA is thought to be partly the result of being detectable upon tumour cell shedding into the bloodstream [11]; as such, the quantity of ctDNA may be indicative of tumour burden and/or aggressiveness [7,[43][44][45].
Lastly, we evaluated the scope of newly detected alterations in EOT samples from participants who exhibited clinical benefit in order to identify biomarkers that may be associated with acquired resistance to study treatment.We focused on the TAS + FUL-treated PIK3CAmut participants who exhibited clinical benefit and clustered genes involved in similar signalling pathways.Newly detected alterations at EOT were predominantly observed in genes associated with the ER, PI3K and p53 signalling pathways (Fig. 6A-C and Fig. S7C), suggesting that the activation or dysregulation of these signalling pathways contributes to clinical resistance to PI3K inhibition plus ET in ER+, HER2À breast tumours.As previously reported, the acquisition of ESR1 or TP53 alterations likely reflects resistance to ET [17,21,22,[27][28][29].The acquisition of PTEN or AKT1 mutations is likely reflective of resistance to PI3K inhibition [32,33,[46][47][48].Collectively, these findings underscore the value of longitudinal ctDNA analysis to elucidate changes in PI3K and associated pathways following treatment with pathway-specific inhibitors.
A key strength of this study is that it leverages one of the largest clinico-genomic datasets to date of participants with PIK3CAmut ER+, HER2À breast cancer treated with a PI3K inhibitor, allowing for detailed subgroup analyses based on genomic alteration status.Nevertheless, our study has several limitations.First, the EOT ctDNA data were derived from the analysis of plasma samples only from PIK3CAmut participants who exhibited clinical benefit [11.7% (n = 12/103) of PBO + FUL-treated and 17.8% (n = 42/236) of TAS + FUL-treated]; thus, the resulting interpretation of the EOT landscape may not be reflective of the total randomized population.Second, taselisib treatment is associated with increased toxicities, as illustrated by the higher proportion of participants treated with TAS + FUL who experienced adverse events leading to discontinuation [16.8% (n = 70/416) with TAS + FUL vs. 2.3% (n = 5/213) with PBO + FUL] [6].As a result, some participants may have discontinued study treatment for reasons other than disease progression; thus, the association of the evaluated oncogenic drivers with clinical outcomes may be underestimated.Moreover, the sample sizes of some biomarker-positive subgroups were small, and therefore, the analysis may be underpowered.Lastly, as CDK4/6 inhibitors in combination with ET are approved for use in first-line HR+, advanced breast cancer, some may question the applicability of our findings for a CDK4/6 inhibitor-treated patient population.That said, patients with PIK3CAmut HR+, HER2À advanced breast cancer have been shown to benefit from treatment with the p110a inhibitor alpelisib plus fulvestrant, regardless of prior treatment with a CDK4/6 inhibitor [49,50].Furthermore, less than 50% of patients with HR+, HER2À advanced breast cancer are prescribed CDK4/6 inhibitors as first-line treatment [51][52][53][54] despite approvals in the first-line setting, illustrating the continued relevance of the SANDPIPER population to a real-world setting.

Conclusion
Our comprehensive baseline ctDNA analysis characterized the genomic heterogeneity of a previously treated ER+, HER2À advanced breast cancer patient population.Although there are no approved treatment options to directly target the alterations we have identified, the detection of a prognostic biomarker from the cumulative molecular landscape of patients may identify those with high-risk diseases and who perhaps warrant additional intervention.Moreover, our analysis of paired baseline and EOT ctDNA samples identified potential mechanisms of resistance to the study treatment that is largely driven by alterations in genes associated with the ER, PI3K, p53, and/or FGFR signalling pathways.This has important clinical implications following the use of PI3K inhibitors in the treatment of patients with ER+, HER2À advanced breast cancer.

Fig. 1 .
Fig.1.Analysis of baseline ctDNA from participants with ER+, HER2À advanced breast cancer.(A) Consort diagram for baseline circulating tumour DNA (ctDNA) analysis of participants enrolled in the SANDPIPER study. 1 PIK3CA mutation status is determined by detection of ≥ 1 pathogenic PIK3CA mutation in baseline ctDNA.2Stratification is based on the detection of ≥ 1 pathogenic alteration in baseline ctDNA.(B) Genomic landscape of baseline ctDNA from participants with PIK3CAmut, ER+, HER2À advanced breast cancer.Only genes that were altered in ≥ 1% of samples (n ≥ 3) are shown in the tile plot.Individual samples may harbour ≥ 1 alterations of the same variant type in a single gene (e.g.≥ 1 TP53 short variants); this information is not denoted in the tile plot.ER, oestrogen receptor; F1L, FoundationOne Liquid; FUL, fulvestrant; HER2, human epidermal growth factor receptor 2; mut, mutated; n, sample size; PBO, placebo; pts, patients; TAS, taselisib.

Fig. 2 .
Fig. 2. Association of progression-free survival (PFS) with (A, B) genomic alteration status and (C) study treatment in participants with PIK3-CAmut baseline ctDNA.Association between PFS and genomic alteration status in (A) PBO + FUL-treated participants and (B) TAS + FULtreated participants.Only genes that were altered in ≥ 1% of samples (n ≥ 3) are shown in the volcano plots; the gene is annotated if the nominal P-value < 0.05.The size of the bubble indicates the frequency of the alterations in the respective gene within the treatment arm.(C)PFS by biomarker status of the top altered genes (altered in ≥ 10% of samples).Log-rank tests using a Cox proportional hazards regression model were used to obtain hazard ratio (HR) and P-values; statistical significance is defined by Benjamini-Hochberg-adjusted P-value (qvalue) < 0.05.CI, confidence interval; FUL, fulvestrant; mPFS, median progression-free survival; mut, mutated; n, sample size; PBO, placebo; TAS, taselisib.

Fig. 5 .
Fig. 5. Paired analysis of baseline and EOT ctDNA from participants enrolled in the SANDPIPER study who exhibited clinical benefit.(A) Consort diagram of the analysis population.Genomic landscape of newly detected alterations at EOT in participants with PIK3CAmut baseline ctDNA treated with (B) PBO + FUL and (C) TAS + FUL.Newly detected alterations are defined as alterations that were not detected in baseline ctDNA but were detected in EOT ctDNA. 1 PIK3CA mutation status is determined by cobas â PIK3CA Mutation testing of baseline tumour tissue. 2 PIK3CA mutation status is determined by detection of ≥ 1 pathogenic PIK3CA mutation in baseline ctDNA.CR, complete response; ctDNA, circulating tumour DNA; EOT, end-of-treatment; FUL, fulvestrant; INV-BOR, investigator-assessed best overall response; INV-PFS, investigator-assessed progression-free survival; mut, mutated; mos, months; n, sample size; NMD, no mutation detected; PBO, placebo; PR, partial response; pts, patients; QC, quality control; TAS, taselisib.

Fig. 6 .
Fig. 6.Circulating tumour DNA (ctDNA) dynamics of PI3K pathway genes in TAS + FUL-treated participants who exhibited clinical benefit.ctDNA dynamics of (A) PIK3CA, (B) AKT1 and (C) PTEN mutations in end-of-treatment (EOT) ctDNA compared to baseline ctDNA.Sample shown only if a change in mutation(s) detected was observed between baseline and EOT ctDNA.Dendrogram includes short variants that were classified as either pathogenic (i.e.functionally relevant) or variants of unknown significance.Samples on the left are from participants with the best overall response by investigator of partial response (PR) or complete response (CR) to TAS + FUL treatment; samples on the right are from participants with stable disease (SD) in response to TAS + FUL treatment who demonstrated extended PFS of > 7 months with tumour shrinkage.AKT1, AKT serine/threonine kinase 1 (also known as protein kinase B); FUL, fulvestrant; mos, months; mut, mutated; NMD, no mutation detected; PBO, placebo; PI3K, phosphatidylinositol 3-kinases; PTEN, Phosphatase and tensin homolog; TAS, taselisib.