Adjuvant nivolumab, capecitabine or the combination in patients with residual triple-negative breast cancer: the OXEL randomized phase II study

Abstract

Chemotherapy and immune checkpoint inhibitors have a role in the post-neoadjuvant setting in patients with triple-negative breast cancer (TNBC). However, the effects of nivolumab, a checkpoint inhibitor, capecitabine, or the combination in changing peripheral immunoscore (PIS) remains unclear. This open-label randomized phase II OXEL study (NCT03487666) aimed to assess the immunologic effects of nivolumab, capecitabine, or the combination in terms of the change in PIS (primary endpoint). Secondary endpoints included the presence of ctDNA, toxicity, clinical outcomes at 2-years and association of ctDNA and PIS with clinical outcomes. Forty-five women with TNBC and residual invasive disease after standard neoadjuvant chemotherapy were randomized to nivolumab, capecitabine, or the combination. Here we show that a combination of nivolumab plus capecitabine leads to a greater increase in PIS from baseline to week 6 (91%) compared with nivolumab (47%) or capecitabine (53%) alone (log-rank p = 0.08), meeting the pre-specified primary endpoint. In addition, the presence of circulating tumor DNA (ctDNA) is associated with disease recurrence, with no new safety signals in the combination arm. Our results provide efficacy and safety data on this combination in TNBC and support further development of PIS and ctDNA analyses to identify patients at high risk of recurrence.

Fig. 1. Changes in peripheral immunoscore #1, and other immune cell subsets after 6 weeks of therapy.

a Heatmap representing the frequency at landmark and 6 weeks of refined classic peripheral blood mononuclear (PBMC) subsets of cell types reflecting known function (Supplementary Table 3) that were used to generate an immunoscore (peripheral immunoscore #1) in patients enrolled in arms A (n = 15), B (n = 14), C (n = 13), and arms A and C combined (n = 28). Each row corresponds to one patient. Peripheral immunoscore #1 is the sum of points assigned to each subset based on tertile distribution as previously described. b The peripheral immunoscore #1 calculated in A before and after 6 weeks of therapy in each treatment arm and arms A and C combined. c Comparison of the percent change after 6 weeks vs. baseline in the peripheral immunoscore in each arm and arms A and C combined. p values are shown; p values were calculated by a two tailed Wilcoxon signed-rank test in B and a two tailed Mann–Whitney test in C, and no adjustments were made for multiple comparisons. Additional immune changes in the peripheral immune profile after 6 weeks of treatment in patients treated with (d) capecitabine (n = 14), e nivolumab (n = 15), and (f) nivolumab plus capecitabine (n = 13). For (d–f), changes in 10 classic PBMC cell types and 148 refined subsets reflective of maturation and function were analyzed with no adjustments made for multiple comparisons. Notable subsets with significant changes at post timepoints vs. landmark are displayed in (d–f) and include those with p < 0.05 (calculated by a two tailed Wilcoxon signed-rank test), difference in medians >0.05, and ≥50% of patients having a >25% change. cDC conventional dendritic cell, DN double negative, EM effector memory, NK natural killer cells, Treg regulatory T cell, MDSC myeloid-derived suppressor cells, PBMC peripheral blood mononuclear cells. Source data are provided as a source data file.

Fig. 2. Median invasive disease-free survival (iDFS) and overall survival (OS).

iDFS (A) and OS (B) stratified by treatment arm. iDFS (C) and OS (D) stratified by the presence or absence of circulating tumor DNA (ctDNA) at landmark in patients enrolled in Arms A, B, and C combined. Blue line: Arm A (nivolumab); Red line: arm B (capecitabine); Green line: Arm C (combination of nivolumab and capecitabine). Source data are provided as a source data file.

Fig. 3. The peripheral immune profile at landmark associates with the development of recurrence after therapy.

The peripheral immune profile at landmark was compared between patients who developed a recurrence (R) and those that did not (no R). Frequency of immune subsets at landmark that associate with recurrence in patients treated with (a) nivolumab, capecitabine, or nivolumab + capecitabine (n = 27 with no R, n = 15 with R), b nivolumab (n = 7 with no R, n = 8 with R), c capecitabine (n = 9 with no R, n = 5 with R), and (d) nivolumab plus capecitabine (n = 11 with no R, n = 2 with R). Differences in 10 classic peripheral blood mononuclear (PBMC) cell types and 148 refined subsets reflective of maturation and function were analyzed. Notable subsets with significant differences are displayed and include those with p < 0.05 (calculated by a two tailed Mann–Whitney test), and difference in medians >0.05 of PBMCs. No adjustments were performed for multiple comparisons. NK/NKT natural killer T cells, PD-1 programmed cell death protein 1, Treg regulatory T cell, HLA-DR Human Leukocyte Antigen – DR isotype, ICOS inducible T cell co-stimulator, PBMC peripheral blood mononuclear cells. Source data are provided as a Source Data File.

Fig. 4. Peripheral immunoscore #2 at landmark associates with disease recurrence in patients receiving nivolumab or nivolumab +/− chemotherapy.

a Heatmap representing the frequency at landmark of refined peripheral blood mononuclear cell (PBMC) subsets of cell types reflecting known function (Supplementary Table 9) that were used to generate an immunoscore (peripheral immunoscore #2) in patients enrolled in arms A (n = 15), B (n = 14), C (n = 13), and arms A and C combined (n = 28). Each row corresponds to one patient. The peripheral immunoscore #2 is the sum of points assigned to each subset based on tertile distribution as previously described. b Association between the peripheral immunoscore #2 calculated in A with disease recurrence following therapy in each arm and arms A and C combined. Peripheral immunoscore #2 was compared in patients with no disease recurrence (no R) vs patients with disease recurrence (R) in Arm A (n = 7 no R, n = 8 R), Arm B (n = 9 no R, n = 5 R), Arm C (n = 11 no R, n = 2 R), and Arms A + C combined (n = 18 no R, n = 10 R). Medians with p values are shown; p values were calculated by a two tailed Mann–Whitney test. c Association between the peripheral immunoscore #2 calculated in A and iDFS in arms A (n = 15), B (n = 14), C (n = 13), and arms A and C combined (n = 28), were analyzed using a Log-rank (Mantel-Cox) test. Hazard ratio and 95% confidence interval, calculated by the Mantel–Haenszel method, are indicated. Solid line: patients with peripheral immunoscore #2 (PIS #2) >the median; dashed line: patients with PIS #2 ≤ the median. PD-1 programmed cell death protein 1, EM effector memory, NK natural killer cells, NKp 30 natural killer cells activating receptor 30, ICOS inducible T cell co-stimulator, Treg regulatory T cell, MDSC myeloid-derived suppressor cells. Source data are provided as a Source Data File.

Fig. 5. Association of the peripheral immune profile at landmark in patients from Arms A, B and C combined with the presence of ctDNA at landmark and recurrence.

The peripheral immune profile was compared at landmark in all arms combined between patients with presence and absence of landmark ctDNA. Frequency of PBMC subsets at landmark that differed between (a) patients with (n = 11) and without (n = 24) ctDNA at landmark, b patients without ctDNA at landmark who recurred (R, n = 4) vs. did not recur (no R, n = 20) after therapy, and (c) patients with ctDNA at landmark who recurred (R, n = 9) and did not recur (no R, n = 2) following therapy. Differences were analyzed in 10 classic PBMC cell types and 148 refined PBMC subsets reflective of maturation and function. Notable subsets with significant differences are displayed and include those with p < 0.05 (calculated by a two tailed Mann–Whitney test), and a difference in medians >0.05 of PBMCs. No adjustments were made for multiple comparisons. ctDNA circulating tumor DNA, PBMC peripheral blood mononuclear cells, PD-L 1 programmed death-ligand 1, Treg regulatory T cell, HLA-DR Human Leukocyte Antigen – DR isotype, ICOS inducible T cell co-stimulator, NK natural killer cells, NKp46 natural cytotoxicity triggering receptor 1. Source data are provided as a Source Data File.