Combined Perioperative Lapatinib and Trastuzumab in Early HER2-Positive Breast Cancer Identifies Early Responders: Randomized UK EPHOS-B Trial Long-Term Results

Abstract

Purpose: EPHOS-B aimed to determine whether perioperative anti-HER2 therapy inhibited proliferation and/or increased apoptosis in HER2-positive breast cancer.

Patients and methods: This randomized phase II, two-part, multicenter trial included newly diagnosed women with HER2-positive invasive breast cancer due to undergo surgery. Patients were randomized to: part 1 (1:2:2), no treatment (control), trastuzumab or lapatinib; part 2 (1:1:2) control, trastuzumab, or lapatinib and trastuzumab combination. Treatment was given for 11 days presurgery. Coprimary endpoints were change in Ki67 and apoptosis between baseline and surgery tumor samples (biologic response: ≥30% change). Central pathology review scored residual cancer burden (RCB). Relapse-free survival (RFS) explored long-term effects.

Results: Between November 2010 and September 2015, 257 patients were randomized (part 1: control 22, trastuzumab 57, lapatinib 51; part 2: control 29, trastuzumab 32, combination 66). Ki67 response was evaluable for 223 patients: in part 1 Ki67 response occurred in 29/44 (66%) lapatinib versus 18/49 (37%) trastuzumab (P = 0.007) and 1/22 (5%) control (P < 0.0001); in part 2 in 36/49 (74%) combination versus 14/31 (45%) trastuzumab (P = 0.02) and 2/28 (7%) control (P < 0.0001). No significant increase in apoptosis after 11 days was seen in treatment groups. Six patients achieved complete pathologic response (pCR, RCB0) and 13 RCB1, all but two in the combination group. After 6 years median follow-up, 28 (11%) had recurrence and 19 (7%) died. No recurrences or deaths were observed among patients who achieved a pCR. Ki67% falls ≥50% associated with fewer recurrences (P = 0.002).

Conclusions: Early response after short duration anti-HER2 dual therapy identifies cancers dependent on the HER2 pathway providing a strategy for exploring risk-adapted individualized treatment de-escalation.

Figure 1.

The CONSORT diagram summarizes patients recruited into each part of the trial, patients randomized, patients eligible to start treatment, patients who started treatment, and those who completed perioperative treatment as per protocol. In part 1, 22 patients were allocated to control, 57 to trastuzumab, and 51 to lapatinib; in part 2, 29 were allocated to control, 32 to trastuzumab, and 66 to the combination. Overall, 255 (99%) patients were considered eligible to start treatment and included in the analysis of perioperative endpoints. Of the 204 patients in the treatment groups, 201 patients (99%) received some perioperative treatment, with 190/201 (95%) completing the 11 days of perioperative treatment. The figure also describes how many patients available for analysis of coprimary endpoints Ki67 and apoptosis. Only patients with both paired samples and enough tumor tissue for biomarker analysis were included in the analysis: 223 patients (88%) had paired Ki67 data and 193 (76%) had paired apoptosis data available for analysis. Patients with pCR or 0% breast cellularity were excluded from main analysis of Ki67 and apoptosis.

Figure 2.

Percentage change in Ki67 between pretreatment (baseline) and surgery for part 1 (A) and part 2 (B); Kaplan–Meier estimates by treatment group for relapse free survival (C) and overall survival (D). A, Waterfall plots for part 1 and part 2: for each patient, bar height represents percentage change at surgery from baseline. Percentage change was calculated as [(surgery score + 0.1) − (pretreatment score + 0.1)]/[(pretreatment score + 0.1)]*100. The constant of 0.1 was added to accommodate cases with a value of 0%. Negative values represent decrease from baseline, positive values represent increase from baseline. pCR in breast: patients with pCR (no disease in ether breast or nodes) plus two additional patients with 0% breast cellularity but nodal involvement are represented as bars of height −120% at the left of the figures and noted “pCR in breast;” any existing Ki67 values for these patients have been excluded of the main analysis; in a sensitivity analysis, we imputed a value of −100% change for these patients (Appendix 2). Small triangles indicate patients with RCB1. Disease recurrences are also indicated at the top of each figure with circles and crosses. B, RFS is represented in the time interval of up to 6 years after randomization, as no RFS event occurred later. Overall survival is represented in the fully observed range of values. Log-rank test comparing concurrently randomized treatment groups are reported in the figures. In the figure, trastuzumab and control part 1 and part 2 groups are combined to improve readability. C, control; L, lapatinib; T, trastuzumab; T+L, combination; P1, part 1; P2, part 2; all, P1&P2; P, P value.

Figure 3.

Association of perioperative changes in biological markers with RFS (A) by categories of Ki67 relative change, (B) by categories of Ki67 absolute change, (C) by baseline TILs, (D) by surgery TILs. RFS is represented in the time interval 0 to 6 years, as no RFS events occurred beyond 6 years from randomization. All treatment groups are combined; log-rank tests are stratified by treatment group (P = P value). For A and B, a value of −100% Ki67 change (ΔKi67) has been imputed for patients with a pCR in breast. For B, we have categorized both baseline and surgery Ki67 into high if ≥10% or low if <10%. No patient increased Ki67 from low to high after 11 days of perioperative treatment. Because of small number of patients in the “low–low” group, we have compared patients with “high” value at surgery with patients with “low” value at surgery.