Neoadjuvant T-DM1/pertuzumab and paclitaxel/trastuzumab/pertuzumab for HER2+ breast cancer in the ad

HER2-targeted therapy dramatically improves outcomes in early breast cancer. Here we report the results of two HER2-targeted combinations in the neoadjuvant I-SPY2 phase 2 adaptive platform trial for early breast cancer at high risk of recurrence: ado-trastuzumab emtansine plus pertuzumab (T-DM1/P) and paclitaxel, trastuzumab and pertuzumab (THP). Eligible women have >2.5 cm clinical stage II/III HER2+ breast cancer, adaptively randomized to T-DM1/P, THP, or a common control arm of paclitaxel/trastuzumab (TH), followed by doxorubicin/cyclophosphamide, then surgery. Both T-DM1/P and THP arms ‘graduate’ in all subtypes: predicted pCR rates are 63%, 72% and 33% for T-DM1/P (n = 52), THP (n = 45) and TH (n = 31) respectively. Toxicity burden is similar between arms. Degree of HER2 pathway signaling and phosphorylation in pretreatment biopsy specimens are associated with response to both T-DM1/P and THP and can further identify highly responsive HER2+ tumors to HER2-directed therapy. This may help identify patients who can safely de-escalate cytotoxic chemotherapy without compromising excellent outcome.

HER2-overexpressing breast cancer outcomes have improved dramatically in the last decade with the addition of HER2-directed therapy to chemotherapy for early breast cancer. In the neoadjuvant setting, pathologic complete response (pCR) rates are in the 40–50% range with taxane/trastuzumab combinations, and even higher with the addition of other HER2-targeted agents. pCR has also proven to be a strong surrogate of event-free survival for individual patients1,2, particularly in hormone receptor-negative tumors. This provides multiple opportunities: to determine whether novel investigational agents can further improve pCR rates, assess comparative toxicity, and determine how increased pCR rates affect long-term survival outcomes. As such, the I-SPY2 trial is a neoadjuvant platform trial in which serial tumor samples are collected to identify biomarkers indicative of highly responsive or non-responsive tumors. These biomarkers are essential to optimizing therapy with strategies to de-escalate toxic treatment for highly responsive tumors and escalate therapy for those that are poorly responsive.

T-DM1 (ado-trastuzumab emtansine, Kadcyla) is an intravenous drug–antibody conjugate that links the HER2-targeted monoclonal antibody trastuzumab to emtansine, an active but systemically toxic chemotherapeutic3,4. T-DM1 is currently FDA-approved as a single agent for the treatment of patients with HER2+, metastatic breast cancer who previously received trastuzumab and a taxane5. Pertuzumab (Perjeta) is a humanized monoclonal antibody that targets the extracellular dimerization domain of HER2, distinct from the binding site of trastuzumab. Pertuzumab is FDA-approved for use in combination with trastuzumab and docetaxel in metastatic breast cancer, and in combination with trastuzumab and chemotherapy as neoadjuvant or adjuvant therapy in non-metastatic disease.

This report describes the evaluation of two I-SPY2 trial arms: a neoadjuvant non-chemotherapy regimen, T-DM1 + pertuzumab (T-DM1/P), and a dual-HER2 targeting regimen, paclitaxel + trastuzumab + pertuzumab (THP), compared to paclitaxel + trastuzumab alone (TH) for HER2+ breast cancer in the I-SPY2 trial. Moreover, we examined the potential for pre-specified multi-omic biomarkers reflecting degrees of HER2 pathway activation, estrogen receptor signaling, and proliferation to predict enhanced HER2-directed treatment response in patients already deemed HER2+ by standard CLIA assays. These were examined in three signatures: the overall HER2+ group, as well as the hormone receptor (HR)+ HER2+ and HR–/HER2+ groups.

Fifty-two patients were enrolled in the T-DM1/P arm at 15 clinical sites from June 6, 2013 through August 17, 2015, when the arm graduated in all HER2+ signatures (all patients, HR+/HER2+, and HR–/HER2+). During the same time period, 45 patients were enrolled in the THP arm when the arm graduated in all HER2 signatures and was converted to the HER2+control arm. Thirty-one patients were randomized to the TH control arm across 17 sites from March 10, 2010 through to August 17, 2015 (Fig. 1). Baseline characteristics of the three arms were similar (Table 1), with the exception that the T-DM1/P am had excess Mammaprint ultra-high (MP2) tumors (44%) compared to control (16%; two-sided Fisher exact test p = 0.01). Source data is provided in Supplementary Data 1.

Consort diagram shows the number of patients screening, randomized, and receiving allocated therapy from the start of I-SPY2 to the close of the T-DM1/Pertuzumab and THP arms. I-SPY2 is modified intent-to-treat, where patients receiving their allocated therapy are considered evaluable for analysis.

Full size image

Efficacy

Figure 2 shows the pCR predicted probability curves (associated values are found in Supplementary Table 1). T-DM1/pertuzumab graduated in all three HER2+ signatures. Amongst all HER2+ patients in the TDM-1/P and TH arms, T-DM1/P had a higher estimated pCR rate (55%; 95% PI 41–69%) than TH (25%; 95% PI 11–38%), corresponding to a 99.9% probability that T-DM1/P was superior to TH and a 96% predictive probability of superiority in a 300-patient phase III trial. In the HR+/HER2+ and HR–/HER2+ signatures, the probability that pCR rates with T-DM1/P were superior to control were 99.7 and 98.8%, respectively.

pCR probablility distribution curves of TDM1/P (red) vs TH (blue) and THP (orange) vs TH (blue). Arrows below x-axis indicate 95% probability interval derived from the I-SPY2 Bayesian time and covariate-adjusted logistic model described in the methods.

Full size image

THP also graduated in all three HER2+ signatures, as shown in Fig. 2. Amongst all HER2+patients in the THP and TH arms, THP had a higher estimated pCR rate (56%; 95%PI 42–70%) than TH (25%; 95% PI 11–38%), corresponding to a 99.9% probablility that THP was superior to TH and a 97% predictive probability of superiority in a 300-patients phase III trial. In the HR+/HER2+ and HR–/HER2+ signatures, the probability that pCR rates with THP were superior to control were 99.5 and 99.9%, respectively.

Toxicity

The toxicities of each arm are well described6,7,8 and no new safety signals were seen in the I-SPY2 Trial arms (Table 2; summary of all adverse events in Supplementary Data 2). Overall, adverse events are primarily grade 1/ 2 (see Supplementary Table 2) but one patient on the THP arm died due to respiratory failure during the THP portion of therapy and was unrelated to therapy. The majority of patients in each arm were able to receive AC after initial HER2-directed therapy (Table 2). Toxicities during AC differed between treatment arms: diarrhea and neutropenia were more frequent in the T-DM1/P arm, as was febrile neutropenia, while anemia was more frequent in the THP arm.

Event-Free Survival (EFS)

Follow-up was available for 47 (90%) T-DM1/P patients, 39 (86.7%) THP patients, and 31(100%) control patients, with median follow-up of 4.1 years. Over the follow-up period, a total of 7 events were observed in the T-DM1/P arm, 3 in the THP arm, and 7 in the common TH control arm. Notably, there were five CNS recurrences during the follow up period: three in the T-DM1/P arm, two in the TH arm, and none in the THP arm (Supplementary Fig. 1). Four of the five CNS recurrences observed were HR– (Supplementary Fig. 1).

Three-year EFS was 88% (95% CI: 79–99%) for T-DM1/P, 92% (95% CI: 84–100%) for THP and 87% (95% CI: 75–100%) for TH (Supplementary Fig. 1). In the two arms with brain metastases, EFS estimates were similar for patients who achieved pCR and those who did not: 88% regardless of pCR status for TDM-1/P (95% CI: 77–100% and 74–100% for pCR and non-pCR, respectively), and 88% (67–100%) vs 86% (73–100%) for pCR vs. non-pCR, respectively, in the TH control arm. For THP, EFS was 96% (89–100%) vs 85% (68–100%) for pCR vs non-pCR.

Biomarker assessment

We assessed 10 biomarkers in the HER2, ER/PR, and proliferation pathways as predictors of response to TDM1/P and THP, hypothesizing that highly HER2-activated, non-luminal A or highly proliferative tumors may be more sensitive to anti-HER2 therapies than those that are less HER2-activated, more luminal or quiescent. HER2 and HER2 signaling was evaluated at 5 levels of resolution: IHC,quantitative protein, and phosphoprotein measurements by RPPA (total ERBB2; phosphorylated [p] pERBB2, phosphorylated [p] pEGFR), and mRNA (HER2 amplicon module Module7_ERBB2). These HER2 biomarkers were highly correlated (Supplementary Fig. 2A) and all five HER2 pathway biomarkers were significantly associated with pCR in the TDM1/P and THP arms (Fig. 3A–G, LR p < 0.05 as detailed in Table 4). Nearly all associations retained significance in a model adjusting for HR status, and in HR+/HER2+subsets (Table 3, Supplementary Fig. 2C–H). HER family (HER2 and EGFR) activation/phosphorylation signatures clustered distinctly from total HER2 as measured by mRNA and RPPA (Supplementary Fig. 2A). Plotting pEGFR and pERBB2 expression together revealed a separation into two groups of patients: one with excellent response and one without (Fig. 3G). Since HER2 and EGFR are well-known heterodimerization partners, the fact that both are found to be highly co-activated in the pre-treatment tumor samples of responding patient population provide further evidence of functional HER2-driven pathway activation and signaling coherence.

A Mosaic plot showing the proportion of patients who had pCR (purple) or did not have pCR (yellow) as a function of HER2 IHC level (1+, 2+, 3+; left to right), in the TDM1/P (top), THP (center), and control (bottom) rows. Barplots in B show these data by arm. Pink: IHC 1+; Light Blue: IHC 2+; Blue: IHC 3+. Panels C, D show response-association boxplots of a HER2 mRNA signature overall (C; n = 127) and by arm (D; n = 52 (TDM1/P), 44 (THP), and 31 (Ctr)) Yellow box is non-pCR, light blue box is pCR. Color of data points correspond to IHC level as in panel B. Panels E, F show response-association boxplots of HER2 total protein overall (E; n = 117) and by arm (F; n = 49 (TDM1/P), 43 (THP), and 25 (Ctr)). Yellow box is non-pCR, light blue box is pCR. Color of data points correspond to IHC level as in panel B. G Two-way scatter plot of phosphorylated HER2 (Y1248) (y-axis) and phosphorylated EGFR (Y1173) (x-axis) relative intensity (RI) values generated from the LCM-RPPA data in the pretreatment biopsy samples are shown along with pCR YES (green) and pCR NO (red) designations. For box plots, center line is group median; upper and lower limits of the box correspond to the 1st and 3rd quartile with whiskers extending to 1.5 times the interquartile range from top/bottom of the box.

Full size image

ER/PR signaling, represented by the average expression of ESR1 and PGR, was also evaluated. Higher ER signaling levels were associated with non-response to TDM1/P and THP in the population as a whole (Fig. 4A, B, Table 3) and in the HR+/HER2+ subsets within both arms (Table 3, Supplementary Figure S2J–K). Consistent with the above, tumors classified Luminal A by PAM50 had a lower pCR rate relative to those of other classes (Fig. 4C, Supplementary Fig. 2).

A, B shows HR expression (ESR1 and PGR averaged) response-association boxplots in all patients (A; n = 127) and by arm (B; n = 52 (TDM1/P), 44 (THP), and 31 (Ctr)). For all box plots, yellow box indicates non-pCR, light blue indicates pCR. C Bar plot showing the prevalence of PAM50 subtypes (LumA: light blue; LumB: dark blue; HER2:red; Basal: magenta) in patients achieving pCR compared to non-responders in the population as a whole (left pair of bars, n=126) and by arm (right pairs, n = 51 (TDM1/P), 44 (THP), and 31 (Ctr) n = 51 (TDM1/P), 44 (THP), and 31 (Ctr)). D–F show response-association boxplots of proliferation biomarkers on the mRNA (D; n = 83), protein (E; n = 78) and phospho-protein (F; n = 78) levels in the HR+/HER2+ subset by arm. *Association is significant (LR p < 0.05) only for TDM1/P, broken rectangle. For box plots, center line is group median; upper and lower limits of the box correspond to the 1st and 3rd quartile with whiskers extending to 1.5 times the interquartile range from top/bottom of the box.

Full size image

Finally, we quantitatively assessed proliferation markers at the total protein (RPPA: Ki67), phospho-protein (pAURKA) and mRNA (proliferation signature Module11_Proliferation) levels as predictors of response. In the population as a whole, the mRNA proliferation signature was associated with response in both T-DM1/P and THP, whereas pAURKA was associated with response in only THP. However, all three proliferation biomarkers significantly associate with response to TDM1/P but not THP in the HR+/HER2+ subset (Fig. 4D–F; Table 3).

Of the 10 markers evaluated, the proliferation markers showed the largest predictive performance differences between arms, particularly in the HR+/HER2+ subtype. Other differences include more dramatic HER2-pathway biomarker associations with pCR in the TDM1/P arm relative to THP. Examples include HER2 3+ IHC: OR = 7.1 [1.4–50] vs. 5.2 [0.9–40]; PAM50 HER2: OR = 21 [4–219] vs 1.6 [0.4–6.4]; pERBB2 amplicon: OR/unit increase = 5.5 [2.4–16] vs 2.3 [1.2–6.4]; pERBB2: OR/unit increase = 16.3 [2.6–245] vs. 5.5 [1.5–31]) and is further illustrated by scatter plots of pERBB2 vs. pEGFR by arm (Supplementary Fig. 3). Within the common control arm, of the 10 biomarkers tested only HER2 IHC 3+ significantly associates with response (LR p = 0.046), potentially due in part to small sample size (Table 3).

In HER2+ early breast cancer, neoadjuvant treatment with T-DM1/P or THP demonstrated significant improvement in predicted pCR rates over treatment with TH, when all regimens were followed by AC and surgery. Similar improvement with T-DM1/P and THP over TH was observed in both HR+ and HR– subsets, despite overall higher response rates for HR–disease. Clinically, these results have been corroborated by several other studies and the overall findings with regard to pCR rates and toxicity are similar to what is seen in the I-SPY2 Trial. In fact, THP followed by AC became a care standard for neoadjuvant therapy in HER2+breast cancer during its assessment in I-SPY2, with the FDA approval of pertuzumab based on results from multiple neoadjuvant trials7,9,10. T-DM-1/P alone, however, has not been adopted as a care standard largely because the KRISTINE trial showed better pCR rates with neoadjuvant taxane/platinum with HP (55.7%) over TDM-1/P (44.4%)6. Our pCR estimates in the TDM1/P arm are higher, likely because patients also received AC.

Critically, however, the I-SPY2 trial adds important and novel information to the field in the assessment of biomarkers that differentiate responsiveness to these Her2-directed therapies. Despite all tumors in these arms being classified as HER2+ by standard IHC and/or FISH, pre-treatment HER2 levels (whether measured by IHC, quantitative protein or gene expression), and activation of HER2 signaling (as measured by phosphorylation and co-activation of HER2 and EGFR, or PAM50 HER2 subtype), were all strong predictors of pathologic response to both TDM-1 and THP. Conversely, highly hormone-driven tumor biology as defined by luminal A subtype and low proliferation was inversely correlated with response in both experimental arms. Low levels of expression by protein and mRNA predict a much lower response to HER2 targeted therapy. These findings are critically important to the design of the next generation of HER2+ trials in which these biomarkers can be utilized to assess which patients can potentially de-escalate therapy (by dropping toxic anthracycline or platinum agents) from those who could benefit from treatment escalation with new and more intensive treatment approaches as illustrated in Fig. 5.

Top graphic represents the current treatment paradigm: patients with HER2+ breast cancer as determined by IHC or FISH receive the same neoadjuvant chemotherapy and HER2-directed therapy. Bottom graphic illustrates how implementation of the RPPA assay and gene expression profiling following positive IHC or FISH will enable de-escalation or escalation of chemotherapy and HER2-directed therapy. Original illustration: A. Haymond.

Full size image

I-SPY2 is a biomarker-rich trial where we uniquely apply rigorous pre-specified statistical assessment to pre-defined mechanism-of-action biomarkers that span the DNA–RNA–protein–phosphoprotein biochemical landscape. The design helps us to learn why some patients respond and other do not. The phosphoprotein data generated here provides evidence for a phospho-HER2-EGFR cut point that, once validated, could guide escalation (those predicted to have a pCR) and de-escalation (those predicted to not have a pCR) of therapy (Fig. 3G). Our observation that HR status and estrogen signaling are associated with lower responsiveness to TDM1/P and THP is consistent with previous reports that luminal-type tumors are less responsive to HER2-targeted agents11. As both HER2 and ER biomarkers predict response to both arms, it is unlikely that either pathway biomarker alone would be sufficient to specifically predict response to TDM1/P vs. THP. However, we did observe a more dramatic signal for the HER2 pathway (particularly the pERBB2 and pEGFR) in the TDM1/P arm.

Recently, there has been a resurgence of interest to redefine breast cancer subclasses that respond to HER2-targeted therapies beyond the current repertiore of FDA-approved IHC and FISH testing parameters12. This interest is underpinned by the acknowledged overall lack of clinical sensitivity and specificity of IHC and FISH to accurately idenfity those patients destined for response to HER2-targeted therapy including those both currently defined as HER2+ and HER2–13. Our qualifying gene/protein/phosphoprotein biomarkers were found to significantly associate with response in both T-DM1/P and THP arms, and differentiated HER2 IHC 3+ patients who did not achieve pCR as well as HER2 IHC 1/2+ patients who did not achieve pCR (Fig. 3). In other words, these biomarkers may help with escalation and descalation of HER2-directed therapy. For example, patients whose tumors are HER2 low/indeterminate (0-2+ by IHC) but have high phospho EGFR and HER2 levels) could be enriched for response to FDA-approved/experimental HER2 targeting agents. However, patients whose tumors are HER2 high (3+) but have low phospho EGFR and HER2 levels should potentially receive other experimental therapies combined with HER2-directed agents. Such combinations can be ethically explored in such patients because they are likely non-responsive to current FDA-approved HER2 targeted regimes. While there is a wealth of data that supports using TDM1 as rescue therapy in patients who have residual disease following neoadjuvant therapy14, our data support further investigation of neoadjuvant TDM1/P specifically in HR+/HER2+ selected based on proliferation signatures. This approach could be particularly appealing for those in whom myelosuppression is a particular concern.

The major difference in predictive signal between arms were observed in the proliferation biomarkers in the HR+HER2+ subset, where these biomarkers associated with response in TDM1/P but not THP arm. If confirmed in independent studies, a combination of proliferation and pERBB2/pEGFR may help distinguish likely response to TDM1/P vs. THP in context of HR status. Taken together, these findings suggest that there is a great degree of heterogeneity amongst HER2-overexpressing breast cancers in the I-SPY population of patients with tumors >2.5 cm, which could potentially be exploited for patient selection strategies.

There are several limitations to the current analysis that impact interpretation, most notably the impact of small sample sizes that limit evaluation of both long-term outcomes and biomarkers. The benefit of the adaptive randomization in I-SPY2 is its efficiency in assigning the agents to the subsets of patients that benefit most, and drugs leave the trial when graduation is reached—that is, the minimum number of patients necessary can be exposed to investigational agents in order to determine signal in a statistically robust manner. With this greater degree of efficiency in evaluating pCR rates comes sample size/power limitations with respect to assessing EFS benefits or predictive biomarkers. Importantly, approximately 9% of the patients in KRISTINE who did not have a pCR on the T-DM1 arm received adjuvant chemotherapy, and there were 15 locoregional events prior to surgery in the T-DM1/P arm, suggesting the presence of a subgroup of patients with disease that was primarily refractory to HER2 antibody–drug conjugate-based therapy alone due to the inability to internalize the antibody–toxin, as suggested by Weyergang et al. (submitted). These findings reinforce our I-SPY2 results suggesting that HER2- and HR-related biology of the primary tumor play a large role in responsiveness to T-DM1, and confirming that adjuvant therapy must be considered when examining EFS, with more aggressive adjuvant therapy in non-pCR patients potentially improving outcomes preferentially for this group of patients compared to those with pCR, creating an equalizing effect.

The similar outcomes of patients with and without a pCR that we observed is likely to be related to adjuvant therapy received. Because the primary endpoint of I-SPY2 is pCR, the type or extent of adjuvant therapy is not mandated, and was at the discretion of the treating physican. Adjuvant therapies differed significantly between the three arms and between pCR and non-pCR groups (Supplementary Table 3). For example, seven patients on the non-taxane containing T-DM1/P arm received additional chemotherapy with adjuvant taxane plus trastuzumab, and all of these were in the non-pCR group. No patients in the THP or TH arms received adjuvant chemotherapy.

The CNS is a known sanctuary site for HER2+ breast cancer, even in the setting of pCR, a finding that is further corroborated in our study: the arms where CNS recurrences were found, difference in 3-year EFS between those with pCR vs. non-pCR was not seen. It is difficult to say whether the lack of brain metastases in the THP arm was due to chance, but previous reports indicate that the addition of pertuzumab to trastuzumab may delay onset of CNS disease15. Tucatinib, a recently FDA-approved drug has shown efficacy in metastatic HER2+ breast cancer patients with brain metastases16, will be studied in the post-neoadjuvant setting combined with T-DM1 in the COMPASS-RD (NCT04457596) trial. Combining tucatinib with a less toxic regimen such as T-DM1 and avoiding the toxicity of AC could prove to be a highly effective intervention for women with tumors that highly express HER2 or fall into the HER2 molecular subtype. Development of drugs that penetrate the CNS and prevent CNS recurrence in HER2+ breast cancer is urgently needed to improve outcomes in these high-risk patients.

In summary, both T-DM1/P and THP significantly increased estimated pCR rates over TH in HER2+ patients, within both HR+ and HR– groups. Given similar high pCR rates and toxicity burden in the populations overall, it is important to develop tools that can be applied to individual patients to aid in the selection of the most active therapy in the clinic. Pretherapy gene expression and protein signaling further identifies highly sensitive tumors beyond those that are HER2+ by standard ASCO/CAP guidelines. The goal of I-SPY2 is to identify active agents, and generate useful guidance regarding the biological profiles of patients in whom further studies of an active agent or strategy should be employed. A high percentage of patients can achieve pCR with TDM1/P or THP. Without AC, these could be less toxic combinations that could be effective for HER2 molecular classification and proliferative tumors. The ability to eliminate AC altogether is under investigation in the COMPASS-pCR trial, where patients are treated to receive THP alone neoadjuvantly. The data reported here are useful in planning further definitive studies that will enable optimization of HER2-directed therapy for future phase III trials and shed important light on the heterogeneity of HER2+ breast cancer and its impact on response to targeted therapy.