NSABP FB-10: a phase Ib/II trial evaluating ado-trastuzumab emtansine (T-DM1) with neratinib in women with metastatic HER2-positive breast cancer

In 2013, T-DM1 was the first HER2-targeted antibody–drug conjugate (ADC) granted FDA-approval for late-stage metastatic breast cancer after prior trastuzumab. In 2019, T-DM1 was approved as post-neoadjuvant therapy in patients with residual disease based on the KATHERINE trial, demonstrating that post-neoadjuvant T-DM1 was statistically more beneficial than trastuzumab, preventing recurrence of invasive disease or deaths in patients with residual disease in breast or lymph nodes after treatment with trastuzumab ± pertuzumab (hazard ratio for invasive disease or death, 0.05: 95% CI 0.039–0.64; P < 0.001) [1]. KATHERINE required archival HER2-positivity but did not mandate HER2 status at study entry. Because multiple studies have confirmed that HER2 status is plastic with conversion of HER2-positive disease to HER2-low (IHC = 0–1+ or IHC 2+ /FISH-negative) under pressure of therapy [2‐4], this raised the question of ADC efficacy in HER2-low patients—either de novo (HR + /HER2-negative) or acquired from conversion of HER2-amplified to HER2-low. There are now several breast cancer-targeting ADCs in the pipeline in addition to the newly approved trastuzumab deruxtecan (T-DXd). Initial approval of T-DXd was for metastatic HER2-positive breast cancer after prior progression on multiple lines of anti-HER2 therapy (DESTINY-Breast01) [5]. In DESTINY-Breast03, T-DXd improved PFS and OS compared to T-DM1 [6] in patients with metastatic disease with progression on trastuzumab. In heavily pretreated HER2-low breast cancer patients, T-DXd was evaluated in a single arm phase II study. The objective response rate (ORR) to T-DXd by central review was 37%, with median duration of response of 10.4 months [7]. DESTINY-Breast04, a randomized, multicenter trial in patients with unresectable or metastatic HER2-low breast cancer, reported highly significant improvements in PFS and OS in patients receiving T-DXd compared to physician choice of treatment [8]—a particularly striking observation, because neither trastuzumab nor T-DM1 has shown consistent activity in HER2-low populations [9]. HER2 status (expression, mutation, amplification) is thus emerging as a predictor of clinical efficacy for anti-HER2-therapy. In our NSABP phase Ib trial of HER2-targeted therapies using T-DM1 + neratinib in HER2-positive patients, we showed a discordance in HER2 status between archival tissue and a liquid biopsy obtained at study entry. Loss of ctHER2-amp occurred in 63% (17 of 27) patients. Deeper and more durable responses were observed with T-DM1 + neratinib in patients with ctHER2-amplification [10]. We now report on an expanded cohort. Our aims were to: (1) confirm activity of T-DM1 + neratinib in patients progressing on a taxane with trastuzumab + pertuzumab (HP), (2) evaluate discordance in HER2 amplification between archived tissue, contemporaneous tissue, and blood, and (3) compare mutation and gene-expression profiles at different time points. Finally, in a subset of patients, we assessed response by RECIST 1.1 with the Guardant Molecular Response score [11, 12].

Approximately 35% of HER2-positive patients may have a loss of HER2 amplification after undergoing chemotherapy + anti-HER2 therapy [2, 3]. In a retrospective analysis of 525 patients who received neoadjuvant chemotherapy (NAC) + HP, 141 patients with residual disease had HER2 status determined pre-and post-NAC-HP. HER2 was concordant (positive/positive) in 84/141 (60%). HER2 protein expression was lost (IHC 0) in 13/57 (23%) and designated as HER2-low in 44/57 (77%) including IHC 1 + in 31 and IHC 2 + /FISH non-amplified in 13 [4]. HER2 intratumoral heterogeneity is likely one cause of discordant HER2 status between primary and post-treatment residual or metastatic disease [24]. Other possibilities include decreased HER2 expression, which could be a transient change or a result of the selection of HER2-low subclones [4].

We have assessed HER2 status before and after pre- and post-study therapy in not only solid tissue but also blood. We have determined the HER2 status in tissues with CLIA IHC/FISH assays, which is the gold standard for HER2 assessment, plus with Ampli-Seq, because it provided a quantitative analysis of HER2 copy number with a greater dynamic range. Ampli-Seq was able to detect a decrease in HER2 copy number in samples that had not lost HER2 amplification based on IHC/FISH. We have also monitored HER2 status in liquid biopsies, which has several advantages over genomic analysis of tissues. Blood has exposure to all potential metastatic sites allowing for the detection of different variants from different metastatic sites. Thus, blood may be more representative of the metastatic tumor than examination of a single biopsied lesion, and may reflect tumor evolution and intratumoral heterogeneity [25]. Blood samples are more easily collected, making multiple serial collections possible. Collecting multiple serial tissue samples is impractical, costly, and represents a much greater risk to patients than does serial collection of blood. The assessment of ctDNA is a powerful tool, showing very promising results to monitor tumor recurrence and response to therapy, but it does not yet replace the current gold standard, IHC/FISH, for the assessment of HER2 status in solid tumors. However, the monitoring of the HER2 status in ctDNA does provide an indicator of tumor response to therapy.

Loss of HER2 amplification observed after one cycle of study therapy may indicate that responders are either clearing ctDNA completely or that the amplification falls below the limit of detection. In the 5 cases who were ctHER2 DNA amplified, and completely cleared ctDNA, the response rate was 100%.

We applied a Molecular Response VAF ratio calculation to measure the change in ctDNA from baseline to C2D1, with the hypothesis that an early decrease in ctDNA levels would predict response to T-DM1 + neratinib therapy, as measured by PFS and RECIST response. Indeed, Molecular Response was associated with both PFS and best response to therapy. This should be confirmed in larger dataset, however, our findings are in line with other studies demonstrating the ability of ctDNA to predict short- and long-term efficacy. Early data from the PADA-1 trial suggests that changing therapy based on alterations detected in ctDNA, prior to evidence of progression via imaging, may provide clinical benefit. In that trial, patients with ER + /HER2-negative metastatic breast cancer being treated in the first line setting with an aromatase inhibitor (AI) + palbociclib were monitored for hotspot ESR1 alterations via ctDNA using digital droplet PCR (ddPCR). Patients with rising ESR1 VAF on therapy, but no synchronous evidence of disease progression via RECIST 1.1, were randomized to either continue receiving an AI + palbociclib or switched to fulvestrant + palbociclib. PADA-1 met its primary efficacy objective, with patients randomized to receive fulvestrant + palbociclib having a significantly longer PFS compared to those who stayed on an AI + palbociclib (median PFS 11.9 months [95% CI 9.1–13.6 months] versus 5.7 months [95% CI 3.9–7.5 months]; stratified HR 0.61 [95% CI 0.43–0.86], two-sided P = 0.004) [26]. More data on mutational evolution is needed to determine whether similar strategies employing ctDNA to inform change in therapy will be broadly applicable across breast cancer subtypes and therapy classes in order to further prolong OS.

Although a cross comparison of studies can be problematic, phase II and III studies suggest that as patients are more heavily treated with anti-HER2 regimens, the PFS and ORR decrease with each subsequent anti-HER2 therapy [27‐30]. However, in a phase III randomized trial of trastuzumab deruxtecan (T-DXd) versus trastuzumab emtansine (T-DM1) in patients (N = 524) whose disease progressed on anti-HER2 therapy, the reported ORR for patients treated with T-DXd or T-DM1 were 79.7% versus 34.2%, respectively. The landmark analysis of PFS at 12 months was 75.8% with T-DXd as compared to 34.1% with T-DM1 (HR 0.28, 95% CI 0.22–0.37; P < 0.001) [6]. Although both T-DXd and T-DM1 have a trastuzumab backbone, there are substantial differences in the linker-payload chemistry, which favors an increased intracellular payload and a bystander effect with T-DXd [31, 32].

We have shown in our study that the benefit from T-DM1 + neratinib is limited in HER2-non-amplified tumors. This finding is consistent with a study reporting a PFS with T-DM1 of 1.5 months [33] in patients discordant for HER2 in primary and metastatic tissue (HER2-positive/negative). Although loss of HER2-amplification appears to be one mechanism of resistance to T-DM1, half of the patients with HER2-amplified tumors did not respond to T-DM1 + neratinib, indicating that resistance to T-DM1 is not limited to loss of HER2-amplification. Many other mechanisms of resistance to T-DM1 have been proposed, such as altered cellular uptake, intracellular transport, and metabolism of the payload. [32]

Based on the low level of activity of T-DM1 monotherapy in patients failing HP, we speculate that the combination of T-DM1 and neratinib is more effective than T-DM1 monotherapy. We are unaware of any trials in HER2-positive breast cancer in which patients progressing on HP have been randomized to compare the response to T-DM1 as monotherapy with a combination of T-DM1 and an irreversible tyrosine kinase inhibitor (TKI). However, in preclinical lung models with ERBB2 mutation and/or amplification, the combination of T-DM1 + neratinib did show increased efficacy over monotherapy. Anecdotally, enhanced efficacy was demonstrated in a breast cancer patient progressing on monotherapy with T-DM1 who then responded with addition of neratinib. The mechanism of action of the antibody–drug conjugates (ADC) such as T-DM1 and T-DXd involves the recognition and binding of the trastuzumab backbone to the extracellular HER2 surface receptor. The ADC-protein complex is internalized with cleavage of the cytotoxic payload. Irreversible TKIs such as neratinib and afatinib (unlike reversible TKIs such as lapatinib) have been shown to enhance HER2 internalization and lysosomal sorting, which has the potential to increase uptake of bound ADC and release of their cytotoxic payload [34].

The KATHERINE [6] study established T-DM1 as a standard of care in early HER2-positive breast cancer patients with residual disease after neoadjuvant therapy [1]. DESTINY-Breast03 has clearly shown superiority of T-DXd over T-DM1 in HER2-positive metastatic disease. The currently recruiting DESTINY-Breast05 study will compare T-DXd with T-DM1 in high-risk HER2-positive patients with residual disease following NAC-HP (NCT04622319). Another study, DESTINY-Breast06 (NCT04494425), will address the question of HER2-low (IHC 1 + or IHC 2 + /FISH-negative or HER2 IHC > 0, < 1 +) in patients with metastatic hormone-positive disease with progression on at least two lines of endocrine therapy comparing T-DXd with investigator’s choice of chemotherapy [31]. The 30-40% of HER2-positive patients with residual or metastatic disease after neoadjuvant therapy who have lost HER2 amplification, while not directly being addressed with these ADCs studies, warrant further investigation with newer generation ADCs.

Our study has several limitations including the non-randomized design, the logistic difficulties in obtaining samples of blood and tissue on all patients and the small sample size, which limited its power and the ability to perform multivariant analysis. However, the strengths of our study include the multiple temporal sample collections, multiple assessments of HER2 status, and molecular assessment of DNA in both tissue and blood.

Despite the limitations of our study, the findings have generated several hypotheses that should be further investigated. First, retrospective confirmation in a large phase III study that loss of HER2-expression under the pressure of therapy can be detected with liquid biopsy; second, the overall response, depth, and duration of response to anti-HER2 therapy is greater in patients with HER2-amplified than in non-amplified patients; third, the activity of T-DM1 or other ADCs with a trastuzumab-backbone may be enhanced with addition of an irreversible TKI such as neratinib. This hypothesis, testing the interaction of an ADC with a TKI, reversible and irreversible, could be evaluated in patient-derived xenografts or other model systems and should be validated prior to a randomized trial. Finally, a small fraction of HER2-nonamplified patients did benefit from T-DM1 + neratinib. Possible explanations, which require further investigation, include a false negative assay, enhanced internalization of T-DM1 in presence of neratinib, or EGFR becoming the driver in patients with loss of HER2-amplification, and inhibition by neratinib [32‐34]. We also realize that a low ctDNA fraction could have prevented the detection of ctHER2 amplification.

Gene expression analysis revealed several important observations. First, the level of HER2 RNA expression in TP1 tissues was closely correlated with the response rate to study therapy. Second, non-luminal subtypes had a better response rate than luminal tumors, although this difference was not statistically significant. Third, there was a non-significant association of the 8-gene trastuzumab benefit groups with the rate of responses to study therapy. Fourth, changes in intrinsic subtypes were seen between TP0 and TP1 tissue samples, indicating that these changes may be a result of tumors evolving to become resistant to HP. These results highlight the importance of collecting and monitoring molecular changes in tissue samples as patients move through their treatments.

PIK3CA mutations are a known oncogenic driver in breast cancer and drive therapeutic resistance in multiple HER2-targeted therapies [35]. In the EMILIA trial, patients with PIK3CA mutations, treated with capecitabine + lapatinib were associated with a shorter PFS than were patients with wild-type tumors; however, in patients treated with T-DM1 this was not the case [36]. We likewise see that in patients treated with T-DM1 + neratinib, PIK3CA mutations were not associated with outcomes. However, we cannot rule out the possibility that a subset of patients, refractory to T-DM1 + neratinib with PIK3CA mutations, may be responsive to PIK3CA inhibitors.

We demonstrate the usefulness of serial assessment of HER2 status in blood and tissue in patients with an initial diagnosis of HER2-positive disease. Loss of HER2 amplification in ctDNA, or the complete loss of ctDNA on treatment with T-DM1 + neratinib, was associated with clinical benefit. Further, we show that many of the patients with short-lived PR or PD were HER2-low in tissue. These patients may be better treated with the recently approved ADC, trastuzumab deruxtecan. We observed that the ADC, T-DM1, plus neratinib, was well tolerated. The combination with an irreversible tyrosine kinase inhibitor with other ADCs warrants investigation.