Taipei Veterans General Hospital, Taiwan, has an ongoing project (VGH-TAYLOR) performing comprehensive genomic profiling on tumor tissues from patients with breast cancer via NGS, which is sponsored by the YongLin Healthcare Foundation. The VGH-TAYLOR study aimed to discover potential biomarkers for recurrence, diagnosis, and prognosis of breast cancer that may enable personalized medicine and improvement in breast cancer treatment; the rationale and design of the study protocol have been previously described (15). Patient and genomic data were collected from the NGS database linked with electronic health records (EHRs) to investigate the mutation prevalence in various subtypes and stages of breast cancer.

In the NGS database, medical records of patients with advanced/metastatic breast cancer who were treated with an mTOR inhibitor or CDK4/6 inhibitor, with treatment initiation (index date) from January 1, 2018 to January 30, 2020, were included in the study once the inclusion/exclusion criteria were met. The EHRs were retrieved 3 years prior to the index date, from January 1, 2015 to January 1, 2018, for baseline characteristics and breast cancer recurrence data collection ( Figure 1 ). Patients with information of first prescription of mTOR or CDK4/6 inhibitors during the 4-year EHR data collection period were also included for analysis for time to treatment failure. The study protocol was approved by the ethics committee of the Taipei Veterans General Hospital.

The pathological features of the tissue samples were determined using immunohistochemistry (IHC) assay. At least 1% of nuclei staining–positive results were defined as ER-positive. As per the American Society of Clinical Oncology (ASCO) and College of American Pathologists (CAP) guidelines, IHC 3+ and IHC 2+ with fluorescence in situ hybridization (FISH) amplification indicated HER2 overexpression. An IHC score of 0 to 1+ is called HER2– and a score of 3+ is called HER2+. A FISH HER2:CEP17 signal ratio >2.2 is called amplification whereas, ratio between 1.8-2.2 is equivocal and <1.8 is non-amplification

Inclusion criteria were defined as follows: (1) female patients aged ≥20 years; (2) patients with confirmed diagnosis of primary invasive breast cancer who are planning to receive treatment for breast cancer; (3) patients who have breast cancer recurrence at screening or Stage IV patients who have received or are currently receiving treatments for breast cancer; (4) Eastern Cooperative Oncology Group (ECOG) performance status ≤3; and (5) life expectancy ≥3 months.

Patients were excluded if they had a primary cancer other than breast cancer within 5 years prior to screening. Archival samples from the biobank (retrospective cohort) will be withdrawn if: (1) tumor content of the FFPE sample is lower than the specified percentage according to the standard of the central laboratory; (2) FFPE samples failed the DNA/RNA quality check. The criteria of the DNA/RNA quality check will follow the standard of the central laboratory. Enrolled patients will be withdrawn if one of the following conditions occurs: (1) patient withdraws consent; (2) patient refuses to provide specimens for evaluation after enrollment; (3) patient for which all samples/specimens fail the DNA/RNA quality check. The criteria of the DNA/RNA quality check will follow the standard of the central laboratory; (4) patient who does not have sufficient FFPE samples, tissues or blood samples for genetic profiling analysis by principal investigator’s discretion; (5) patient who does not return to the clinical site for more than 6 months (based on their medical records) will be considered as lost to follow-up. However, whether this subject should be withdrawn will be based on the PI’s discretion (15).

In this study we retrospectively analyze the database from VGH-TAILOR project (Group 1,2,3). Enrolled patients were categorized into 4 major groups, presented in Figure 2 .

The primary endpoint was the regional prevalence of PIK3CA mutation in breast cancer and by breast cancer subtypes in overall patients with breast cancer, patients with HR+/HER2–, HR−/HER2+, HR+/HER2+ and triple-negative breast cancer. Because the pre- and postmenopausal status present different biological and genetic characteristics in HR+/HER2– breast cancer, the prevalence of PIK3CA mutations was further analyzed by menopausal status in HR+/HER2– breast cancer. Menopause is usually a clinical diagnosis made after ≥12 months of amenorrhea (16).

Secondary endpoints were to assess the frequency of mutation types by breast cancer subtype and by menopausal status and time to treatment failure after mTOR or CDK4/6 inhibitor in metastatic patients stratified by PIK3CA mutation status. Time to treatment failure is defined as the time from the date of first dose of treatment of interest to the date of the patient discontinuing treatment for any reason; the Kaplan-Meier (KM) method was used for the estimation.

The genetic profiles were determined through NGS of FFPE samples. For screening and recruitment of patients for Group 1B and Group 3-2, paired FFPE primary and recurrent tumor samples were collected and sequenced, and for Group 2, paired FFPE diagnostic and post-neoadjuvant specimens were assayed.

The preparation of FFPE was done under standard conditions at the trial site. The DNA/RNA extraction and hematoxylin and eosin (H&E) staining were performed in accordance with the laboratory manual under the guidance of a certified pathologist in the central laboratory. Quality checks for DNA/RNA were performed as per the manual of the Thermo Fisher™ Oncomine™ (TMO) Comprehensive Assay requirement (see below) and additional samples were collected in case of failure. Of the seven unstained FFPE sections retrieved (per subject), one section was prepared for H&E staining and six sections were prepared for TMO comprehensive assay.

Oncomine™ comprehensive assay (TMO comprehensive assay, Thermo Fisher Scientific, Waltham, MA) was used to profile thousands of variants across 161 cancer-relevant genes using FFPE tissues (17). Analyses of TMO comprehensive assay included identification of genes and detection of mutation types such as frameshift, missense, synonymous, single nucleotide variation (SNV), insertion/deletion (Indel), and copy number variation (CNV) observed in individual subject.

Amplicon libraries were constructed with multiplex polymerase chain reaction (PCR) primers for preparation of DNA and RNA (for fusion genes) from FFPE samples. Sequencing was performed with the Ion Gene Studio S5 System and Ion 540 Chips. Raw data process, alignment, and variant calling were performed with Torrent Suite™ Software, with variant calling using the Torrent Variant Caller plug-in. Further management was proceeded by Ion Reporter™ Software with workflow “Oncomine Comprehensive v3 - w3.2 - DNA and Fusions - Single Sample” version 5.10 selected and filter chain “Oncomine Variants” version 5.10 applied. Reference genome was hg19.

Preindex variables, including demographics of all patients, menopausal status, primary diagnosis, previous medical history/comorbidities, and family history of cancer, were extracted from the NGS database, whereas the treatment response of previous breast cancer was taken from the EHRs.

Postindex variables in the follow-up period, including breast cancer genetic mutations, treatment, and treatment response from previous lines if any, were extracted from the NGS database, and breast cancer recurrence data were also extracted from the EHRs.

All statistical tests were conducted with a two-sided alternative hypothesis, and a significance level of 0.05 and a P value of <0.05 were considered statistically significant, unless otherwise specified. Mutations in the PIK3/AKT/mTOR pathway were deciphered. The prevalence of PIK3CA mutations was calculated as the number of PIK3CA mutations divided by the number of patients with breast cancer in the populations stated above. The 95% CIs were presented as appropriate. Analysis of the primary endpoint was descriptive in nature, and no statistical hypothesis or testing was performed. The enrolled set was adopted for both primary and secondary endpoints unless otherwise specified.

A total of 728 patients were included for the PIK3CA mutation analysis, most of whom were categorized into Group 1A (481 patients, 66.07%), followed by Group 2 (93 patients, 12.77%), Group 3-2 (59 patients, 8.10%), Group 3-1 (42 patients, 5.77%), and Group 1B (20 patients, 2.75%), please refer to Figure 2 for what each group stands for. The mean age in all the groups at the time of testing was 51 to 57 years of age. A total of 548 patients (76.2%) with HR+ and 149 patients (20.8%) with HER2+ status were reported. Patient demographic characteristics are detailed in Table 1 .

Overall, 403 of 728 patients (55%) had mutations in the PI3K/AKT/mTOR pathway, of which the 3 most commonly detected mutations were PIK3CA (57%), AKT3 (14%), and PTEN (10%) ( Figure 3 ).

A total of 278 of 728 patients (38%) harbored PIK3CA mutations. With respect to the IHC phenotypes, 29 patients (43.0%) were HR−/HER2+, 96 patients (42.0%) were HR+/HER2– (postmenopausal), and 34 patients (41.0%) were HR+/HER2+. The prevalence of PIK3CA mutation was relatively lower in triple-negative breast cancer (27.0%) and premenopausal patients with HR+/HER2– status (29.0%). The prevalence of PIK3CA mutation by IHC status is presented in Figure 4 .

The majority of the PIK3CA mutations were clustered in exon 9 and exon 20, helical and kinase domains, respectively. Among 278 patients with PIK3CA mutations, the most frequently observed mutations were in exon 20 (H1047R, 41.6%) and exon 9 (E545K, 18.9% and E542K, 10.3%); these are presented in Figure 5A .

The distribution of PIK3CA mutation types is presented in Figure 5B . A total of 27 of 278 patients (9.7%) harbored multiple PIK3CA mutations within one sample. Mutual exclusivity was observed between PIK3CA and AKT1 (P<0.001), PIK3CA and AKT3 (P<0.001), and PIK3CA and PIK3R1 (P=0.007).

PIK3CA hotspot mutations (H1047R, E545K, and E542K) were found in 63.5% of premenopausal patients with breast cancer and in 56.2% of postmenopausal patients with HR+/HER2– breast cancer. Among the premenopausal PIK3CA-mutated patients, H1047R and E545K mutations were the most common (27% each), followed by the E542K mutation (9.5%). A similar distribution was seen in postmenopausal patients, where the H1047R mutation was the most common mutation (32%), followed by E545K (14.5%) and E542K mutations (9.7%).

A total of 19 patients (22.62%) with advanced breast cancer who received CDK4/6 inhibitors (n=84, ribociclib=47, palbociclib=32, abemaciclib=5) and 3 patients (18.75%) who received an mTOR inhibitor (n=16) were reported to have a PIK3CA mutation. For patients treated with CDK4/6 inhibitors, the median time to treatment failure was 16 months (95% CI: 11-23 months) in the PIK3CA wild-type cohort and 12 months (95% CI: 7-21 months) in the PIK3CA mutation cohort, with a hazard ratio (HR) of 1.670 (95% CI: 0.908-3.069) ( Figure 6A ; Table 2 ). The data demonstrated a trend of shorter treatment duration with CDK4/6 inhibitors in patients with PIK3CA mutation. For patients receiving mTOR inhibitor, the median time to treatment failure was 20.5 months (95% CI: 8-33 months) in the PIK3CA mutation cohort and 6 months (95% CI: 2-9 months) in the PIK3CA wild-type cohort, with an HR of 0.244 (95% CI: 0.031-1.922) ( Figure 6B ; Table 2 ).