Biomarker Analysis of the Phase III NALA Study of Neratinib + Capecitabine versus Lapatinib + Capecitabine in Patients with Previously Treated Metastatic Breast Cancer

HER2 overexpression occurs in 15% to 20% of breast cancers (1). Before the approval of trastuzumab, HER2-overexpressing breast cancer was the subtype with the worst prognosis (2). Multiple additional HER2-targeted therapies have subsequently been developed, two of which are the small-molecule tyrosine kinase inhibitors (TKI) neratinib and lapatinib, the focus of comparison in this analysis. Neratinib and lapatinib target the adenosine triphosphate–binding domain of HER proteins. Their distinguishing features include their binding mechanisms, preclinical cytotoxicity, and impact on HER family dimerization and downstream signaling and gene expression (3). Neratinib binds covalently/irreversibly to HER2 (4), whereas lapatinib binds noncovalently/reversibly (5). Neratinib binds EGFR, HER2, and HER4 with high affinity; lapatinib targets only EGFR and HER2 with high affinity (3). Both have demonstrated in vitro and in vivo inhibition of growth and proliferation in HER2- or EGFR-amplified preclinical breast cancer models, with lower median inhibitory concentration values for neratinib (6). Finally, neratinib, but not lapatinib, inhibits tumors with a wide spectrum of somatic HER2 mutations (7–9).

Multiple mechanisms of resistance to HER2-targeted therapies have been reported. These include upregulation of alternative receptor tyrosine kinases (10), upregulation of the estrogen receptor (ER) pathway (11), dysregulation of downstream PIK3K/Akt signaling (12), increased expression of p95, a highly active truncated form of HER2 that lacks the extracellular trastuzumab-binding domain (13), and additional upregulation of HER2 via acquired amplifications and/or somatic mutations (14, 15).

The phase III NALA study compared neratinib (240 mg once a day) plus capecitabine (750 mg/m² twice a day on days 1–14 of a 21-day cycle; N+C) with lapatinib (1,250 mg once a day) plus capecitabine (1,000 mg/m² twice a day on days 1–14 of a 21-day cycle; L+C) in patients with HER2-positive (HER2⁺) metastatic breast cancer (MBC) who had received ≥2 prior HER2-directed regimens in the metastatic setting (16). There was a statistically significant benefit in progression-free survival (PFS) favoring N+C (HR = 0.76; 12-month PFS, N+C 29% vs. L+C 15%), translating to a 2.2-month mean PFS improvement without significant benefit in overall survival. Patients treated with N+C had significantly fewer interventions for symptomatic central nervous system disease compared with L+C (cumulative incidence 22.8% vs. 29.2%; P = 0.043). Based on these results, the FDA approved N+C for adult patients with advanced or metastatic HER2⁺ breast cancer who have received ≥2 prior anti-HER2–based regimens in the metastatic setting (17).

The high rate of early events in both arms and the largely indistinguishable PFS curves up until 6 months in NALA suggest that the trial enrolled a group of patients resistant to HER2-directed therapies, capecitabine, or both. In this exploratory analysis, we assessed associations of biomarkers with prognostic or predictive PFS benefit (N+C over L+C) in biomarker-defined subgroups in NALA.

NALA (NCT01808573) was a randomized, international, multicenter, parallel-group, active-controlled phase III trial in patients with HER2⁺ MBC. Patient characteristics, eligibility criteria, and treatment have been described (16). The protocol was approved by national/institutional ethics committees at participating sites and conducted in accordance with the Declaration of Helsinki. All patients provided written informed consent. HER2 positivity was centrally confirmed before enrollment as IHC 3+ or IHC 2+/FISH positive (18, 19).

Tumor samples were collected at screening before randomization and classified by study sites as originating from primary or metastatic tumor, and as archival or fresh biopsy. Sites were instructed to send either one formalin-fixed, paraffin-embedded (FFPE) block or 10 unstained, freshly cut 4-μm FFPE slides, and to utilize tissue from the time of metastasis whenever possible.

Biomarkers and assays were selected based on documented evidence of association with response or resistance to HER2-targeted therapies, and, because of limitations in tissue availability, ability to reliably test minimal amounts of tissue. Hormone receptor status was established via central IHC, with Allred total score ≥3 considered ER/progesterone receptor (PR) positive (ER/PR PharmDx Kit, Dako). All assays were performed blinded to study endpoint and treatment assignment. Additional candidate biomarkers that could not be performed due to lack of sufficient tissue included RNA expression, including but not limited to molecular subtype categorization, and retrospective FISH evaluation of all samples.

The spectrum of somatic mutations was evaluated by whole-exome next-generation sequencing (NGS; ref. 20). DNA extraction was performed from FFPE slides (1–3 × 4 μm) using Maxwell FFPE Tissue LEV DNA Purification Kit (Promega). All slides were pathologist-evaluated to determine ≥20% of tumor area content. If required, microdissection was performed to enrich for tumor content. DNA quality and concentration were measured with double-stranded DNA (dsDNA) Qubit broad-range assay kit (Invitrogen). A minimum of 4 ng of extracted DNA was input for amplicon-based NGS using VHIO-Card, an in-house developed panel of more than 800 primer pairs targeting hotspot mutations in 55 genes (Supplementary Table S1). PCR products were pooled, and libraries prepared with NEBNext Ultra DNA Library Prep Kit for Illumina (New England BioLabs). Indexed libraries were pooled, loaded onto a HiSeq 2500 System (Illumina), and sequencing performed (2 × 100) at average coverage of more than 1000×.

Somatic mutations identified in PIK3CA, ESR1, HER2, AKT1, and KRAS were confirmed by droplet digital PCR (ddPCR; QX200 system Bio-Rad Laboratories). These were of special interest because of involvement in HER2 signaling. Custom Taqman SNP genotyping assays for ddPCR were designed and validated with in-house available mutant samples. A black-hole quencher was used with fluorescent dye reporters, fluorescein amidite (FAM) for mutant probes, and 2′-chloro-7′phenyl-1,4-dichloro-6-carboxy-fluorescein (VIC) for wild-type (WT) probes (Thermo Fisher Scientific).

HER2 IHC and FISH scores were determined according to manufacturer specifications (HercepTest and IQFISH pharmDX, respectively; Dako). To better understand the spectrum of HER2 protein expression, HER2 was retrospectively evaluated by two additional methods: H-score (Targos GMBH) and HERmark (Monogram Biosciences). H-scores were evaluated from original IHC images. The HERmark assay, which quantitatively measures HER2 receptors using a dual-antibody, proximity-based, fluorescence assay (VeraTag technology; refs. 21, 22), was performed on an adjacent section. p95 levels were evaluated using VeraTag technology and categorized as high or low based on previously established cut-off points (23, 24). VeraTag measurements were expressed in units of relative fluorescence per square millimeter of invasive tumor (RF/mm²; ref. 21).

Efficacy endpoints were assessed in individual biomarker populations, defined as any randomized patient with that biomarker result. The primary analysis methods were log-rank test for hypothesis testing and Cox proportional hazards model to estimate HRs and 95% confidence intervals (CI). Multivariable Cox proportional hazard models were used with the following covariates: treatment, time from diagnosis to randomization, time from metastasis to randomization, Eastern Cooperative Oncology Group (ECOG) performance score, previous lines of HER2 therapy, HER2 IHC, histologic grade at diagnosis, region, ER/PR status, and disease location. Kaplan–Meier methodology was used for time-to-event endpoints. Subpopulation treatment effect pattern plot (STEPP) was used to assess patterns of effect size for continuous HER2 protein variables. Analyses were performed using SAS (version 9.1; SAS Institute Inc. RRID:SCR_008567) and R (RRID:SCR_001905; ref. 25).

As previously reported (16), 621 patients with HER2-positive MBC were randomized to N+C or L+C. Most samples originated from primary tumors (n = 393; 63%), with the remainder classified as metastatic (n = 184; 30%) or otherwise uncharacterized (n = 63; 10%). Across the 621 available samples, NGS was successfully performed on 420 patients (68%), H-scoring was possible for 615 (99%) of HER2 IHC-stained slides, HERmark score was reported for 526 patients (85%), and VeraTag p95 was reported for 451 patients (73%; Supplementary Fig. S1). Biomarker-expression characteristics of patients with biomarker analysis results were generally comparable across treatment arms (Table 1). Biomarker-analysis populations were generally comparable with the intent-to-treat population (Supplementary Table S2). HER2 levels were not distributed evenly between hormone receptor–negative and –positive cohorts; higher HER2 was more common in patients who were hormone receptor negative (Supplementary Table S3), consistent with prior reports (26).

After NGS identification, PIK3CA, HER2, HER3, AKT1, and ESR1 activating mutations were orthogonally confirmed by ddPCR for cases with sufficient material (n = 150/176; 86%); 100% concordance was observed. Mutation prevalence was: PIK3CA 34% (143/420); HER2 5% (23/420); HER3 1% (4/420), AKT1 1% (3/420), and ESR1 2% (7/420). PIK3CA and HER2 mutations were balanced between treatment arms (Table 1); mutation frequencies were consistent with prior MBC datasets (27).

Most PIK3CA mutations were kinase-domain hotspots, e.g., H1047R/L (n = 83; 56%), or helical-domain hotspots, e.g., E545K/A (n = 32; 22%) and E542K (n = 24; 16%); patients could have more than one PIK3CA mutation (Supplementary Fig. S2A). Ninety-five of the 143 mutations (66%) were identified from primary tissue, 29 (20%) were from metastases, and 19 (13%) were from otherwise unclassified lymph node tissue. PIK3CA mutations were distributed similarly in hormone receptor–positive and hormone receptor–negative disease: they were detected in samples from 58 of 180 patients (32.2%) with hormone receptor–negative disease and 85 of 240 patients (35.4%) with hormone receptor–positive disease. PIK3CA mutations were also distributed similarly between lower and higher HER2 expression levels (Supplementary Table S4).

The presence of a PIK3CA mutation trended toward shorter PFS, independent of treatment [Fig. 1A; WT vs. mutant PIK3CA: HR = 0.81 (95% CI, 0.64–1.03); median PFS 5.55 vs. 4.34 months, respectively].

HER2 gene mutations (n = 23) included kinase-domain hotspots L755S (n = 5; 22%), D769Y (n = 5; 22%), V777 L (n = 5; 22%), G776V (n = 1; 4%), I767M (n = 1; 4%), and L841V (n = 1; 4%); extracellular-domain hotspot S310F/Y (n = 2; 9%); exon 20 insertion G778_P780dup (n = 1; 4%); transmembrane-domain hotspot R678Q (n = 1; 4%); and non–hotspot mutations L841V (n = 1; 4%) and V669 L (n = 1; 4%; Supplementary Fig. S2B). Fifteen of the 23 HER2 mutations (65.2%) were identified from primary tumor tissue, 5 were from metastases (22%), and 3 were from lymph node tissue (13%). Median PFS was 5.45 months in patients with WT HER2 tumors and 5.72 months in those with mutant HER2 tumors [Fig. 1B; WT vs. mutant; HR = 1.686 (95% CI, 0.967–3.288)].

HER2 mutations were not predictive of PFS benefit for N+C over L+C (P-value_inter = 0.502; Fig. 1C). After adjusting for other prognostic factors in a multivariable Cox model, PIK3CA mutation maintained a trend toward shorter PFS [HR = 0.82 (95% CI, 0.65–1.05, WT vs. mutant)]. PIK3CA mutations were not predictive of PFS benefit for N+C over L+C (P-value_inter = 0.995; Fig. 1C). Too few AKT1, HER3, and ESR1 mutations occurred for meaningful evaluation.

Among 621 patients in NALA, 426 (69%) had IHC 3+, and 195 (31%) had IHC 2+/FISH-positive breast cancer. HER2 protein expression orthogonally measured by H-score and HERmark was well balanced between treatment arms (Table 1). Differences were observed in the extent of quantification dependent on methodology (Fig. 2A): HERmark reported across a larger dynamic range, particularly at highest HER2 protein levels; H-scores ranged from 35 to 300 (median 240), and HERmark scores ranged from 0.9 to 1,179 RF/mm² (median 39 RF/mm²).

Higher HER2 protein expression was prognostic of longer PFS when both treatment arms were combined, using all three methods (Fig. 2B). Higher HER2 protein level also significantly predicted PFS benefit for N+C versus L+C [Fig. 2C; IHC 3+: HR = 0.64 (95% CI, 0.51–0.81); H-score ≥240: HR = 0.54 (95% CI, 0.41–0.72); HERmark positive: HR = 0.65 (95% CI, 0.50–0.84); P-value_inter < 0.001 for IHC and H-score and P = 0.061 for HERmark]. Evaluation of HER2 protein level as a continuous variable using STEPP analysis of H-score and HERmark revealed that increasing HER2 protein levels was associated, although not statistically significantly, with HRs favoring the neratinib combination (Fig. 2D).

High p95 was associated with PFS benefit favoring N+C over L+C [Fig. 2C; high p95 (≥2.8 RF/mm²): HR = 0.66 (95% CI, 0.50–0.86) vs. low p95 (<2.8 RF/mm²): HR = 0.91 (95% CI, 0.61–1.36); P-value_inter = 0.236]. N+C benefit versus L+C was shown in patients with HERmark-positive/p95-high [HR = 0.63 (95% CI, 0.46–0.86)] and HERmark-positive/p95-low [HR = 0.55 (95% CI, 0.29–1.04)] tumors; P-value_inter for HERmark/p95 was 0.063 (Fig. 2C). This supports the hypothesis that patients with high levels of active HER2, whether full-length or truncated, derive greater benefit from N+C compared with L+C.

In the intent-to-treat population, PFS benefit from N+C versus L+C was seen in the hormone receptor–negative subgroup [HR = 0.42 (95% CI, 0.31–0.57)] but not in the hormone receptor–positive subgroup [HR = 1.08 (95% CI, 0.84–1.40); ref. 16]. A multivariable Cox model indicated a significant interaction between treatment and hormone receptor status (P = 0.0038); the effect of HER2 protein levels was therefore investigated by hormone receptor status (Fig. 3). The predictive effect of HER2 level appeared more pronounced in patients with hormone receptor–negative tumors. Patients whose tumors were both hormone receptor–negative and high in HER2 protein expression derived the greatest PFS benefit from N+C compared with L+C [IHC 3+: HR = 0.35 (95% CI, 0.25–0.49); H-score ≥240: HR = 0.32 (95% CI, 0.22–0.46); HERmark-positive: HR = 0.35 (95% CI, 0.24–0.51)].

This preplanned, exploratory analysis examined somatic mutations, HER2 protein expression, and hormone receptor status and their association with outcomes in NALA, providing meaningful information on the potential role of biomarkers in clinical outcomes for this patient population.

PIK3CA mutation has been described as a biomarker of resistance to multiple HER2-targeted therapies in MBC (28–30). The trend observed for shorter PFS with PIK3CA mutation in NALA is therefore as expected. In contrast, no associations between PIK3CA mutations and prognosis have been reported for patients with HER2-positive breast cancer undergoing adjuvant treatment with targeted agents, including neratinib in the extended adjuvant setting (31–33).

HER2 somatic mutations have been adversely associated with prognosis in the absence of HER2-targeted therapy (34). In contrast, HER2 mutations in patients in NALA tended to be associated with better prognosis, although this interpretation is limited by a small sample size (n = 23). This may be due to prior exposure to HER2-targeted therapy in NALA, although neither trastuzumab nor pertuzumab—the most common therapies received before study entry—has proven activity against HER2-mutant MBC (35). Preclinically, tumors with HER2 mutations are more sensitive to neratinib than lapatinib (7–9). In patients with MBC whose tumors harbored somatic HER2 mutations in the absence of HER2 amplification, the SUMMIT trial (NCT01953926) reported responses to neratinib monotherapy or in combination with fulvestrant (15). HER2 mutations coincident with amplification are less well characterized, although HER2 amplification has been described as an acquired resistance mechanism to neratinib in HER2-mutated MBC (15). Notably, HER2 mutations in NALA were identified in tissues from both primary and metastatic samples, suggesting that HER2 mutation and amplification can potentially, albeit very rarely, emerge together.

HER2 protein level was a clear response biomarker in NALA. Not only was higher HER2 expression prognostic of PFS, but patients with tumors expressing higher HER2 levels also benefited more from N+C compared with L+C. This association may help explain the late separation of PFS Kaplan–Meier curves in the primary clinical analysis, as tumors with lower HER2 protein level may simply be less sensitive to HER2-targeted therapies. Notably, because of differences in the maximum tolerated dose of capecitabine with neratinib and lapatinib (36), the capecitabine dose was lower in the N+C arm (750 mg/m² twice a day) than in L+C (1,000 mg/m² twice a day) in NALA; consequently, the capecitabine component in NALA regimens may have been responsible for response to treatment among patients whose tumors were not HER2 driven, including those with low HER2 expression.

Neratinib and lapatinib are both small-molecule TKIs; however, neratinib is considered a more potent inhibitor (3). Our results are consistent with prior phase III clinical trials reporting association of higher HER2 levels with better response to the stronger HER2-targeting regimen in HER2⁺ MBC: CLEOPATRA, EMILIA, and TH3RESA all reported increased PFS favoring the more potent regimen in patients with greater HER2 mRNA expression (28, 29, 37). In the adjuvant setting, APHINITY reported increased benefit with the stronger regimen in patients with higher HER2 protein levels, as measured by IHC (38). Finally, in the neoadjuvant NeoALTTO trial, the predicted probability of pathologic complete response to lapatinib + trastuzumab versus either agent alone was associated with increasing HER2 levels as measured by HERmark (39). Although beyond the scope of this manuscript, it would be interesting to understand whether the same trend holds for the novel HER2-specific, reversible TKI tucatinib, which has been reported to be less potent than neratinib when comparing IC₅₀ values in the context of cell-line models (40). Within the context of HER2 positivity, p95 quantitation did not provide additional predictive or prognostic information, perhaps due to the fact that p95 trended with HERmark score in patients who were HER2 positive, and/or that both neratinib and lapatinib bind to the intracellular domain of HER2.

Negative hormone receptor status in NALA was associated with a trend toward better prognosis and increased response to N+C compared with L+C. The ExteNET extended adjuvant trial, in contrast, reported an association of neratinib with improved invasive disease-free survival in patients with hormone receptor–positive status (41). Cross-talk between the ER and HER2 signaling pathways, which has been well described (42, 43), may explain this apparent discrepancy. Patients with hormone receptor–positive disease received coincident hormonal therapy in ExteNET; however, consistent with standard of care and with other MBC trials in similar populations, hormonal therapy was not permitted in NALA. Preclinical models of ER-positive/HER2⁺ breast cancer evaded HER2 blockade with neratinib via upregulation of ER-pathway signaling, and conversely, inhibition with hormonal therapy directed upregulation of HER2 signaling (44). Dual inhibition with endocrine therapy and neratinib may warrant evaluation in patients with hormone receptor–positive, HER2⁺ MBC.

A primary limitation of this study is the lack of a prerandomization biopsy, which necessitated use of a combination of primary and metastatic samples, resulting in an analysis that may not reflect all the molecular changes acquired due to anticancer therapies administered after biopsy. A second consideration is the limited number of slides that were available for each patient (2–4 slides per patient in general). This restricted the analyses that could be performed, e.g., gene-expression profiling was technically unfeasible for such limited tissue.

In conclusion, meaningful biomarker associations were identified in this analysis of tumor samples from patients included in the NALA study. Patients with MBC with high HER2 protein expression regardless of p95 level may benefit most from N+C. Exploration of alternative therapies in patients with low HER2 may be warranted, and patients with high HER2 and hormone receptor–positive disease, and/or PIK3CA mutations, may derive greater benefit from addition of endocrine therapy or PI3K pathway inhibitors.

The authors declare that the data supporting the findings of this study are available within the article. The authors may be contacted for further data sharing.

C. Saura reports personal fees from AstraZeneca, Daiichi-Sankyo, Eisai, Exact Sciences, Exeter Pharma, F. Hoffmann-La Roche Ltd, MediTech, Merck Sharp & Dohme, Novartis, Pfizer, Philips, Pierre Fabre, Puma, Roche Farma, Sanofi-Aventis, SeaGen, and Zymeworks outside the submitted work. M. Oliveira reports grants from PUMA Biotechnology during the conduct of the study. M. Oliveira also reports personal fees from PUMA Biotechnology and iTEOS; grants, personal fees, and non-financial support from Roche and Novartis; grants and personal fees from AstraZeneca, GSK, and SeaGen; personal fees and non-financial support from Pierre Fabre and Eisai; grants from Genentech, Zenith Epigenetics, Immunomedics, and Boehringer Ingelheim; and personal fees from MSD outside the submitted work. H. Wildiers reports grants from Roche, as well as other support from AbbVie, Biocartes, Novartis, The Planning Shop, Ariez Internat, Lilly, Eisai, AstraZeneca, TRM Oncology, Pfizer, Roche, AstraZeneca Ireland, Daiichi-Sankyo, KCE, Novartis, Orion Corporation, PSI Cro AG, Puma Technology, Sirtex, MSD, Immutep Pty, and Congress Care outside the submitted work; in addition, H. Wildiers is President of SIOG, the International Society of Geriatric Oncology. A.M. Brufksy reports personal fees from Puma during the conduct of the study, as well as personal fees from Daiichi, Roche Genentech, Eisai, Lilly, Novartis, Pfizer, AstraZenenca, and Tyme outside the submitted work. S.H. Waters reports personal fees from Novartis Pharmaceuticals Ltd and Roche Products Ltd outside the submitted work. S.A. Hurvitz reports contracted research support (paid to institution) from Ambrx, Amgen, AstraZeneca, Arvinas, Bayer, Daiichi-Sankyo, Dignitana, Genentech/Roche, Gilead, GSK, Immunomedics, Lilly, Macrogenics, Novartis, Pfizer, OBI Pharma, Phoenix Molecular Designs, Ltd. Pieris, PUMA, Radius, Samumend, Sanofi, Seattle Genetics, and Zymeworks, as well as travel reimbursed by Eli Lilly; in addition, S.A. Hurvitz's spouse holds stocks of ROMTech and Ideal Implant. B. Moy reports grants from PUMA Biotechnology during the conduct of the study. S.-B. Kim reports research funding from Novartis, Sanofi-Aventis, and DongKook Pharm Co.; in addition, S.-B. Kim is a consultant in advisory boards of Novartis, AstraZeneca, Lilly, Dae Hwa Pharmaceutical Co. Ltd, ISU Abxis, and Daiichi-Sankyo, and owns stocks of Genopeaks and NeogeneTC. E. Cronemberger reports grants from Puma Biotechnology during the conduct of the study, as well as other support from Roche, AstraZeneca, Daiichi-Sankyo, and Pfizer outside the submitted work. G.J. Wallweber reports personal fees from Laboratory Corporation of America/Monogram Biosciences during the conduct of the study, as well as personal fees from Genentech/Roche outside the submitted work. K. Keyvanjah reports personal fees and other support from Puma Biotechnology Inc during the conduct of the study; in addition, K. Keyvanjah has been employed by QED Therapeutics Inc outside of the submitted work, is a stockholder of Puma Biotechnology and BridgeBio Pharma, and is currently employed by Amgen. A.S. Lalani reports other support from Puma Biotechnology, Inc. during the conduct of the study, as well as other support from Puma Biotechnology, Inc. outside the submitted work; in addition, A.S. Lalani is an employee and shareholder of Puma Biotechnology, Inc. R. Bryce reports other support from Puma Biotechnology during the conduct of the study. A. Vivancos reports grants and personal fees from BMS, as well as personal fees from Bayer, Novartis, AstraZeneca, and Sysmex outside the submitted work. L.D. Eli reports personal fees from Puma Biotechnology during the conduct of the study. S. Delaloge reports grants from Puma, Daiichi, Roche, Seagen, and Pierre Fabre during the conduct of the study; S. Delaloge also reports grants and non-financial support from Pfizer and AstraZeneca, as well as grants from Sanofi, BMS, Besins, Orion, and MSD outside the submitted work. No disclosures were reported by the other authors.

C. Saura: Conceptualization, supervision, investigation, writing–review and editing. J. Matito: Data curation, validation, investigation, methodology, writing–review and editing. M. Oliveira: Investigation, methodology, writing–review and editing. H. Wildiers: Investigation, writing–review and editing. A.M. Brufksy: Investigation, writing–review and editing. S.H. Waters: Investigation, writing–review and editing. S.A. Hurvitz: Investigation, writing–review and editing. B. Moy: Investigation, writing–review and editing. S.-B. Kim: Investigation, writing–review and editing. W.J. Gradishar: Investigation, writing–review and editing. G.S. Queiroz: Investigation, writing–review and editing. E. Cronemberger: Investigation, writing–review and editing. G.J. Wallweber: Methodology, writing–review and editing. J. Bebchuk: Software, formal analysis, methodology, writing–review and editing. K. Keyvanjah: Conceptualization, writing–review and editing. A.S. Lalani: Conceptualization, resources, supervision, writing–review and editing. R. Bryce: Conceptualization, supervision, writing–review and editing. A. Vivancos: Supervision, investigation, methodology, writing–review and editing. L.D. Eli: Conceptualization, formal analysis, supervision, validation, writing–original draft, project administration, writing–review and editing. S. Delaloge: Conceptualization, investigation, writing–review and editing.

We thank the Independent Data Monitoring Committee, study investigators, research staff, clinical research organizations, and other vendors, as well as the patients who participated in the NALA trial. We would like to acknowledge statistical programming support provided by Yan Yan (Puma Biotechnology). We thank Jeff Sperinde and Weidong Huang (Monogram Biosciences) for additional support for HERmark and p95 assay design and execution. We thank Sandra Bernard (Targos GMBH) for oversight of clinical sample management. We thank Lee Miller and Deirdre Carman (Miller Medical Communications Ltd) for writing/editorial support.

This work was supported by Puma Biotechnology Inc. (no grant number is applicable). The funders of the study designed the trial, were responsible for data collection, data integrity and analyses, and interpretation of the data with oversight from the authors. The manuscript was written with input from authors, and with review and input from the sponsor. The lead and senior authors were responsible for the final decision regarding manuscript contents and submission. Puma Biotechnology Inc. funded the provision of editorial support provided by Miller Medical Communications.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.