Single-fraction or multi-fraction stereotactic ablative body radiotherapy followed by atezolizumab in advanced triple-negative breast cancer: a randomized phase II trial Authors: Steven David, Peter Savas, Shankar Siva, Mathias Bressel, Michelle White, Bronte Nikolic, Michael W. Neeson, Shane White, Gavin Marx, Robyn Cheuk, Michelle Grogan, Maria Farrell, Jessica Foudoulis, Paul J. Neeson, Sherene Loi Published: 2026 ABSTRACT Pre-clinical evidence suggests that stereotactic ablative body radiotherapy (SABR) can enhance systemic anti-tumor responses when combined with immune checkpoint inhibition, although optimal fractionation remains uncertain. Here we conducted a phase II multi-center randomized trial (AZTEC trial; ClinicalTrials.gov identifier NCT03464942) to investigate the efficacy and safety of single-fraction (SF: 20 Gy; n = 27) or multi-fraction (MF: 24 Gy in three fractions; n = 27) SABR followed by anti-PD-L1 atezolizumab in 54 women with advanced triple-negative breast cancer (TNBC) as first or second-line treatment. The primary endpoint was progression free survival (PFS). Secondary endpoints included best overall response rate, clinical benefit rate, overall survival and safety. We observed that both treatment arms had similar efficacy: median PFS was 2.5 months (90% CI: 1.7-4.5) for SF and 3.1 months (90% CI: 1.8-3.9) for MF. The null hypothesis (PFS of 2 months) was not rejected in either arm. Clinical benefit (RECIST complete response/partial response or stable disease for ≥24 weeks) was achieved in 5 (20%) SF and 6 (25%) MF patients, all with oligometastatic disease, independent of PD-L1 expression. Adverse events were similar between arms. Prespecified exploratory analysis of peripheral blood found higher CD8 TPEX cells before and during treatment correlated with improved response duration. Combining SABR with atezolizumab is safe, however further testing is required to determine efficacy for advanced TNBC patients with oligometastatic disease. Introduction Therapeutic options for advanced triple-negative breast cancer (TNBC) are currently limited, contributing to its poor prognosis and often associated with significant toxicity1–4. Immune checkpoint inhibitors (ICIs) are being investigated as a potential treatment avenue in advanced TNBC due to the cancer’s high mutational burden and increased expression of programmed death-ligand 1 (PD-L1)5,6. However, ICIs as monotherapy have yielded disappointing tumor response rates7–10 attributed to TNBC’s high levels of immunosuppressive molecules and pathways, including regulatory T cell infiltration, alongside tumor heterogeneity11–14. Consequently, combination approaches targeting the immunosuppressive microenvironment and enhancing the anti-tumor immune response are being explored15,16. Combining chemotherapy with ICIs has yielded positive oncological outcomes in recent randomized phase III trials17,18, leading to their approval for metastatic and neoadjuvant TNBC in combination with chemotherapy. Another strategy involves combining ICIs with radiotherapy. In addition to direct tumoricidal effects, radiotherapy doses in routine use may steer the immune microenvironment towards both pro-inflammatory and immunosuppressive states. A key pro-inflammatory mechanism is the induction of immunogenic cell death (ICD) by radiation, leading to the release of immune-stimulating tumor antigens that potentially transform tumors into in situ vaccines19–22. Both pre-clinical and clinical evidence support the notion that radiotherapy directed at metastatic sites can initiate a systemic immune response, resulting in tumor regression at distant locations—a phenomenon known as the abscopal effect23–26. However, this effect is rare when radiotherapy is used alone, partly due to its inhibitory effects and the naturally suppressive microenvironment of tumors27–29. ICIs have been proposed to partner with radiotherapy to enhance the abscopal effect30. Stereotactic ablative body radiotherapy (SABR) is a highly precise and efficient modality with low toxicity for delivering radiotherapy to multiple targets while exhibiting pro-immune effects, presenting an attractive synergistic partner for ICIs. Thus, combining radiotherapy with ICIs presents a rational strategy in the treatment of advanced TNBC with the potential for synergy, where radiotherapy treatment to metastatic sites acts as an immune stimulant, enhancing the efficacy and durability of systemic tumor response mediated by ICIs. The optimal dose and fractionation for maximizing the systemic immune response remain uncertain; pre-clinical evidence supports both a fractionated regimen of 24 Gy in 3 fractions and a single high-dose approach31–35. Our institutional data confirms the safety and efficacy of 20 Gy in 1 fraction for treating oligometastatic breast cancer36. This regimen may also show synergy when combined with PD-1 checkpoint inhibitors, particularly in multi-target single-fraction treatments for prostate cancer37 and renal cell carcinoma38, without evidence of increased toxicity. However, the recent negative results from the CHEERS trial (24 Gy in 3 fractions)39 suggest that the optimal dose-fractionation schedule for combining radiotherapy with checkpoint inhibitors remains unresolved. Sequencing SABR before ICI is supported by the notion that radiotherapy releases de novo tumor antigens, which subsequent immunotherapy can augment. Both pre-clinical and clinical evidence support this thesis40–42. Here, we present the results of the AZTEC trial, a phase II randomized trial evaluating the efficacy and safety of a combination of two different dose-fractionation schedules of radiotherapy (20 Gy in 1 fraction, or 24 Gy in 3 fractions) to 1–4 metastases followed by the ICI, atezolizumab (a fully humanized, engineered monoclonal antibody of IgG1 isotype against PD-L1), in patients with advanced TNBC. We hypothesized that multi-fraction or single-dose SABR followed by atezolizumab would improve progression-free survival (PFS) in advanced TNBC patients compared to a pre-defined threshold. Results Fifty-four patients were randomized to the AZTEC trial between November 20, 2018 and April 12, 2021, across 8 Australian centers. Four patients were excluded from the final analysis due to inadequate baseline imaging performed in breach of the protocol, resulting in a modified intention-to-treat analysis. The CONSORT flowchart is shown in Fig. 1. The median age was 57 years. Histology was ductal carcinoma, not otherwise specified, in 46 (92%) patients and lobular in 4 (8%) patients. Most patients, 42 (84%), had recurrent metastatic disease. 30 (60%) and 20 (40%) patients had received none or one previous line of chemotherapy in the metastatic setting, respectively. The majority of patients (42, 84%) had oligometastatic disease (2–5 metastases). The most common metastatic sites were bone (25, 50%) and lymph nodes (34, 68%). Most patients received SABR for one (70%) or two (26%) metastases, predominantly to bone. Specifically, 13 (36%) and 14 (45%) bone metastases were irradiated in the single and multi-fraction SABR arms, respectively. Atezolizumab therapy was commenced sequentially in all patients and within five days of completion of SABR in 43 (86%) of patients, as per protocol. Median time from randomization to atezolizumab was 13 days in the 20 Gy arm and 15 days in the 24 Gy arm. Otherwise, the arms were well balanced, and patient characteristics are provided in Supplementary Data 1.Fig. 1CONSORT Diagram.Gy gray, SABR stereotactic ablative body radiotherapy, GCP good clinical practice, C1D1 cycle 1 day 1. CONSORT Diagram. Gy gray, SABR stereotactic ablative body radiotherapy, GCP good clinical practice, C1D1 cycle 1 day 1. Efficacy The median follow-up calculated using the reverse Kaplan-Meier method was 35 months. PFS and OS are shown in Fig. 2A, B. The median PFS for the SABR single-fraction 20 Gy arm was 2.5 (90% CI: 1.7–4.5) months, for the multi-fraction 24 Gy arm was 3.1 (90% CI: 1.8–3.9) months, and for both arms combined was 3.1 (90% CI: 1.8–3.9) months. The median PFS for patients receiving the trial treatment was 3.1 months (95% CI: 1.8–4.5) in the first-line metastatic setting and 1.9 months (95% CI: 1.7–4.6) in the second-line setting.Fig. 2Efficacy analyses.A Progression free survival by treatment arm. B Overall survival by treatment arm. C Overall survival by TIL level. D Swimmer’s plot (n = 50). SABR stereotactic ablative body radiotherapy; HR hazard ratio; TILs tumor-infiltrating lymphocytes; CR complete response; PR partial response; PD progressive disease. Survival analysis performed with Cox regression models and stratified log-rank tests. Tick marks on Kaplan-Meier curves indicate censored observations. Efficacy analyses. A Progression free survival by treatment arm. B Overall survival by treatment arm. C Overall survival by TIL level. D Swimmer’s plot (n = 50). SABR stereotactic ablative body radiotherapy; HR hazard ratio; TILs tumor-infiltrating lymphocytes; CR complete response; PR partial response; PD progressive disease. Survival analysis performed with Cox regression models and stratified log-rank tests. Tick marks on Kaplan-Meier curves indicate censored observations. The median OS was 21 months (95% CI: 12–38) for the SABR single-fraction 20 Gy arm, 15 months (95% CI: 8–37) for the multi-fraction 24 Gy arm, and 17 months (95% CI: 13–33) for both arms combined. Of the 50 patients, 14 (28%) had no target lesions at baseline (9 [35%] in the single-fraction and 5 [21%] in the multi-fraction arms), and four (8%) had all target lesions treated with SABR. Therefore, 32 (64%) patients with unirradiated measurable disease were evaluated for RECIST 1.1 response assessment. Four (4/32, 12%; 95% CI, 4–29%) patients responded in the untreated lesions, two patients in each treatment arm. When considering all patients, clinical benefit (CB) was achieved in 20% (5/25; 95% CI: 7–41%) and 25% (6/24; 95% CI: 10–47%) in single vs multi-fraction arms, and overall was 22% (11/49; 95% CI: 12–37%), with no difference between arms, p = 0.74. One patient withdrew consent at cycle three and was not assessable for CB. Tables 1, 2 show the response assessment for both cohorts combined in all patients (n = 50) and patients with measurable disease (n = 32), by treatment arm.Table 1Response assessments (both cohorts combined)ResponseAll patients (n = 50)Patients with measurable disease (n = 32)CR3 (6%)2 (6%)PR2 (4%)2 (6%)non-CR/non-PD9 (18%)SD13 (26%)13 (41%)PD23 (46%)15 (47%)OR (CR/PR)5 (10%[3,22])4 (12%[4,29])CBa (non-PD > 24 weeks)11 (22%[12,37])7 (23%[10,41])aOne patient withdrew consent at cycle 3 and was not assessable for CB.CR complete response, PR partial response, SD stable disease, PD progressive disease, CB clinical benefit, OR objective response, non-CR/non-PD refers to patients without measurable disease, numbers represent n (% [95% CI]).Table 2Response rate in patients with evaluable disease by RECIST 1.1 by treatment armTreatmentResponseSABR 20 Gy (n = 15)SABR 24 Gy (n = 17)Total (n = 32)Best response by RECIST 1.1CR1 (7%)1 (6%)2 (6%)PR1 (7%)1 (6%)2 (6%)SD5 (33%)8 (47%)13 (41%)PD8 (53%)7 (41%)15 (47%)ORb (CR/PR)2 (13%[2,40])2 (12%[1,36])4 (12%[4,29])CBa (non-PD > 24 weeks)3 (21%[5,51])c4 (24%[7,50])7 (23%[10,41])daOne patient withdrew consent at cycle 3 and was not assessable for CB.bp = 1.00 for the comparison between arms.c One patient in the SABR 20 Gy arm withdrew consent at cycle 3 and was not assessable for CB.dp = 1.00 for the comparison between arms.CR complete response, PR partial response, SD stable disease, PD progressive disease, CB clinical benefit, OR objective response, non-CR/non-PD refers to patients without measurable disease, numbers represent n (% [95% CI]). Table 1Response assessments (both cohorts combined)ResponseAll patients (n = 50)Patients with measurable disease (n = 32)CR3 (6%)2 (6%)PR2 (4%)2 (6%)non-CR/non-PD9 (18%)SD13 (26%)13 (41%)PD23 (46%)15 (47%)OR (CR/PR)5 (10%[3,22])4 (12%[4,29])CBa (non-PD > 24 weeks)11 (22%[12,37])7 (23%[10,41])aOne patient withdrew consent at cycle 3 and was not assessable for CB.CR complete response, PR partial response, SD stable disease, PD progressive disease, CB clinical benefit, OR objective response, non-CR/non-PD refers to patients without measurable disease, numbers represent n (% [95% CI]). Response assessments (both cohorts combined) aOne patient withdrew consent at cycle 3 and was not assessable for CB. CR complete response, PR partial response, SD stable disease, PD progressive disease, CB clinical benefit, OR objective response, non-CR/non-PD refers to patients without measurable disease, numbers represent n (% [95% CI]). Table 2Response rate in patients with evaluable disease by RECIST 1.1 by treatment armTreatmentResponseSABR 20 Gy (n = 15)SABR 24 Gy (n = 17)Total (n = 32)Best response by RECIST 1.1CR1 (7%)1 (6%)2 (6%)PR1 (7%)1 (6%)2 (6%)SD5 (33%)8 (47%)13 (41%)PD8 (53%)7 (41%)15 (47%)ORb (CR/PR)2 (13%[2,40])2 (12%[1,36])4 (12%[4,29])CBa (non-PD > 24 weeks)3 (21%[5,51])c4 (24%[7,50])7 (23%[10,41])daOne patient withdrew consent at cycle 3 and was not assessable for CB.bp = 1.00 for the comparison between arms.c One patient in the SABR 20 Gy arm withdrew consent at cycle 3 and was not assessable for CB.dp = 1.00 for the comparison between arms.CR complete response, PR partial response, SD stable disease, PD progressive disease, CB clinical benefit, OR objective response, non-CR/non-PD refers to patients without measurable disease, numbers represent n (% [95% CI]). Response rate in patients with evaluable disease by RECIST 1.1 by treatment arm aOne patient withdrew consent at cycle 3 and was not assessable for CB. bp = 1.00 for the comparison between arms. c One patient in the SABR 20 Gy arm withdrew consent at cycle 3 and was not assessable for CB. dp = 1.00 for the comparison between arms. CR complete response, PR partial response, SD stable disease, PD progressive disease, CB clinical benefit, OR objective response, non-CR/non-PD refers to patients without measurable disease, numbers represent n (% [95% CI]). The swimmer’s plot is shown in Fig. 2D. The five patients who had objective responses (CR or PR) were all alive at the end of the study period and had durable disease control, subsequently progressing after 44, 72, 72, 92, and 92 weeks. Of the 11 patients with CB, all were oligometastatic ( ≤ 5 lesions in total) at baseline; one patient had five metastases, five had three metastases, and five had two metastases. Safety 46 (92%) patients discontinued atezolizumab prior to the prescribed 34 cycles, primarily due to disease progression, which occurred in 42 (84%) patients, and in only 3 (6%) patients due to treatment-related adverse events (TRAEs). All patients received SABR treatment as planned. Summaries of the most common ( ≥ 10%) treatment-related and immune-related adverse events in both arms combined are shown in Table 3. There were no grade 5 TRAEs and only one grade 4 TRAE. When considering TRAEs by treatment arm, 5 (19% [95% CI: 7–39]) and 2 (8% [95% CI: 1–27]) participants experienced ≥ grade 3 TRAEs in the single-fraction arm and multi-fraction arm, respectively (p = 0.42). Summaries of the worst grade of adverse events, regardless of relationship to treatment, are shown in Supplementary Data 2 (both arms combined) and Supplementary Data 3 (by arm), and adverse events related to treatment in Supplementary Data 4 (by arm). When considering causation, SABR was responsible for only one grade 3 adverse event (fatigue in the 20 Gy arm) and was not responsible for any serious adverse events (Supplementary Data 5).Table 3Common adverse events (worst grade) related to treatment, and immune-related adverse events, both arms combinedGrade (n = 50)Adverse Event1234TotalAutoimmune Disorder00112 (4%)Fatigue741012 (24%)Alkaline Phosphatase Increased10102 (4%)Aspartate Aminotransferase Increased10102 (4%)Hyperglycemia10102 (4%)Atrial Fibrillation00101 (2%)Seizure00101 (2%)Rash53008 (16%)Hypothyroidism32005 (10%)Infusion Related Reaction12003 (6%)Hyperthyroidism31004 (8%)Dermatitis Radiation11002 (4%)Nausea80008 (16%)Diarrhea50005 (10%)Pruritus40004 (8%)Alanine Aminotransferase Increased30003 (6%)Dry Skin20002 (4%)Dyspnea20002 (4%)Arthralgia10001 (2%)Myalgia10001 (2%)Thrombocytopenia10001 (2%)Pneumonitis10001 (2%)Leukopenia10001 (2%)Any Adverse Event†1786132 (64%)†Number of patients whose worst AE was grade 1, 2, 3, 4, and Total. Table 3Common adverse events (worst grade) related to treatment, and immune-related adverse events, both arms combinedGrade (n = 50)Adverse Event1234TotalAutoimmune Disorder00112 (4%)Fatigue741012 (24%)Alkaline Phosphatase Increased10102 (4%)Aspartate Aminotransferase Increased10102 (4%)Hyperglycemia10102 (4%)Atrial Fibrillation00101 (2%)Seizure00101 (2%)Rash53008 (16%)Hypothyroidism32005 (10%)Infusion Related Reaction12003 (6%)Hyperthyroidism31004 (8%)Dermatitis Radiation11002 (4%)Nausea80008 (16%)Diarrhea50005 (10%)Pruritus40004 (8%)Alanine Aminotransferase Increased30003 (6%)Dry Skin20002 (4%)Dyspnea20002 (4%)Arthralgia10001 (2%)Myalgia10001 (2%)Thrombocytopenia10001 (2%)Pneumonitis10001 (2%)Leukopenia10001 (2%)Any Adverse Event†1786132 (64%)†Number of patients whose worst AE was grade 1, 2, 3, 4, and Total. Common adverse events (worst grade) related to treatment, and immune-related adverse events, both arms combined †Number of patients whose worst AE was grade 1, 2, 3, 4, and Total. Association between biomarkers and efficacy PD-L1 status was ascertained for 40 patients, and TIL scores for 42 patients (Supplementary Data 1). Tumor tissue samples used for TIL scoring originated from both primary tumors (n = 11, 26%) and metastatic biopsies (n = 31, 74%). 28% and 52% of patients were classified as PD-L1 positive and negative, respectively (20% of cases had missing data). 52% and 32% of patients were scored as high and low TILs (≥5% and <5%), respectively (16% had missing data). Baseline tumor PD-L1 status (≥1% vs <1%) and TIL status (≥5% vs <5% positive) were not predictive for PFS, OS, ORR, or CBR, regardless of the line of therapy (Tables S1–S7). Comparing cases with TILs ≥ 5% versus <5%, the HR for OS was 0.6 (95% CI: 0.3–1.3), p = 0.2 (Fig. 2C). When analysed as a continuous variable, for each 10% increment in TILs, the odds of clinical benefit increased by 70% (OR 1.7, 95% CI 1.1–2.9, p = 0.02). The discriminatory capacity of TILs for predicting clinical benefit was low, with a receiver operating characteristic curve AUC of 0.62 (95% CI 0.31–0.93). 46 patients had immune cell subpopulations analysed by flow cytometry from peripheral blood samples taken from patients at baseline. Of this population, 45 were assessable for clinical benefit (CB) at baseline, with 9 (20%) demonstrating clinical benefit (CB-Y) and 36 (80%) not (CB-N). PBMC analyses were conducted in this cohort of 45 patients. In the absence of a significant difference in outcome between treatment arms, both arms were pooled to improve the power of PBMC biomarker analyses, and this is presented below. When similar comparisons were conducted within each treatment arm, there was insufficient power to conduct statistical inference, and results are presented descriptively in Figs. S4, S5. CB-Y demonstrated numerically lower proportions of CD8+ T cells and higher proportions of CD4+ T cells at baseline compared with CB-N (Fig. 3B, C). Additionally, a higher baseline CD4+/CD8+ ratio showed a borderline association with CB-Y (p = 0.063, Fig. 3D). No statistical difference was observed in the baseline frequency of CD8+ PD-1+ and CD4+ PD-1+ T cells between the CB-Y and CB-N groups (Fig. 3E, F). Interestingly, we observed that patients with one prior line of chemotherapy in the metastatic setting had significantly lower PD-1 expression on both CD8+ and CD4+ T cells in comparison to no lines of prior chemotherapy (p = 0.025 and p = 0.0097, respectively, Fig. 3G). Following SABR treatment, PD-1+ expression increased in CD8+ but not in CD4+ T cells from baseline (p = 0.04 and p = 0.09, respectively, Fig. 4A, B), before clearly falling again before Cycle 3.Fig. 3Baseline peripheral blood biomarker (PBMC) assessment using flow cytometry.A Proportion of PBMCs that are CD3+ positive at baseline according to clinical benefit. B Proportion of all CD3 T cells that are CD8 + . C Proportion of all CD3 T cells that are CD4 + . D Per-case ratio of proportions of CD4 to CD8 as described in (B, C). E Proportion of PD-1 positive CD8 and F CD4 T cells according to clinical benefit. G Proportion of PD-1 positive CD8 and CD4 T cells at baseline according to prior treatment status. Horizontal bars show the median value. Two-sided Wilcoxon rank sum test applied for group-wise comparisons. Y: clinical benefit; N: no clinical benefit.Fig. 4Peripheral blood biomarker (PBMC) assessment using flow cytometry across time points.A Proportion of PD-1 positive CD8 T cells across time points at baseline before SABR (RT-), post SABR (RT + ) and at the start of Cycle 3 of atezolizumab (C3) in all patients. B Proportion of PD-1 positive CD4 T cells as in (A). C Proportion of PD-1 + TCF1 + CD8 T cells (TPEX) across time points according to clinical benefit. D Proportion of EOMES + CD8 and E CD4 T cells across time points according to clinical benefit. Wilcoxon rank sum test applied for group-wise comparisons. Two-sided Wilcoxon signed rank test applied to paired samples in time course data. Y: clinical benefit; N: no clinical benefit. Baseline peripheral blood biomarker (PBMC) assessment using flow cytometry. A Proportion of PBMCs that are CD3+ positive at baseline according to clinical benefit. B Proportion of all CD3 T cells that are CD8 + . C Proportion of all CD3 T cells that are CD4 + . D Per-case ratio of proportions of CD4 to CD8 as described in (B, C). E Proportion of PD-1 positive CD8 and F CD4 T cells according to clinical benefit. G Proportion of PD-1 positive CD8 and CD4 T cells at baseline according to prior treatment status. Horizontal bars show the median value. Two-sided Wilcoxon rank sum test applied for group-wise comparisons. Y: clinical benefit; N: no clinical benefit. Peripheral blood biomarker (PBMC) assessment using flow cytometry across time points. A Proportion of PD-1 positive CD8 T cells across time points at baseline before SABR (RT-), post SABR (RT + ) and at the start of Cycle 3 of atezolizumab (C3) in all patients. B Proportion of PD-1 positive CD4 T cells as in (A). C Proportion of PD-1 + TCF1 + CD8 T cells (TPEX) across time points according to clinical benefit. D Proportion of EOMES + CD8 and E CD4 T cells across time points according to clinical benefit. Wilcoxon rank sum test applied for group-wise comparisons. Two-sided Wilcoxon signed rank test applied to paired samples in time course data. Y: clinical benefit; N: no clinical benefit. CD8+PD-1+TCF1+ (progenitor exhausted T cells - TPEX or “stem-like”) T cells are reported to have a mechanistic role in mediating response to immune checkpoint blockade43. CB-Y demonstrated higher proportions of circulating TPEX cells before and after SABR, and on atezolizumab therapy, compared with CB-N (p = 0.045 pre-SABR, p = 0.026 post-SABR; p = 0.0093 during atezolizumab; Fig. 4C). The frequency of eomesodermin (EOMES)+ T cells was also evaluated, as it is a transcription factor expressed in effector T cells. CB-Y during atezolizumab treatment also expressed significantly higher levels of EOMES expression on peripheral CD8+ cells compared with CB-N (p = 0.026; Fig. 4D, E). Discussion The TROG 17.05/PMCC/109 AZTEC trial was a randomized clinical trial to investigate the efficacy of two radiotherapy dose-fractionation schedules in combination with the PD-L1 immune checkpoint inhibitor atezolizumab in advanced TNBC patients. This phase II study did not find sufficient evidence to reject the null hypothesis (a 2-month median PFS) with either treatment arm. Consequently, no preferred dose-fractionation SABR schedule was identified for further study based on our results. The median PFS for the SABR single-fraction 20 Gy arm of 2.5 (90% CI: 1.7–4.5) months and for the SABR multi-fraction 24 Gy arm of 3.1 (90% CI: 1.8–3.9) months both compared favorably with data from PD-L1 inhibitor monotherapy trials and encourage further investigation in selected patients. We observed objective responses and disease control seemingly independent of PD-L1 status, although the power to detect differences here was suboptimal. Nevertheless, this suggests that the combination of SABR and atezolizumab may elicit immune-mediated anti-tumor effects through mechanisms beyond PD-L1 expression44,45. Significantly higher levels of circulating TPEX (CD8+ TCF1+ PD-1+) at baseline, post SABR and after the first dose of atezolizumab and CD8+ EOMES+ T cells after the first dose of atezolizumab, were also observed in patients with clinical benefit, suggesting that peripheral biomarkers may help us discriminate early responding patients to atezolizumab, regardless of line of treatment. The results from our study show numerically higher PFS and OS than previously reported monotherapy trials. A phase I trial of atezolizumab9 in 116 patients with heavily pre-treated advanced TNBC reported a median PFS of 1.4 (95% CI, 1.3–1.6) months with 1-, 2-, and 3-year OS rates of 41% (95% CI, 32–50%), 19% (95% CI, 11–26%), and 16% (95% CI, 8–24%) respectively. No PD-L1-negative patients (0 of 21) responded to atezolizumab. Similarly, the PD-1 inhibitor pembrolizumab demonstrated a median PFS of 2.1 months in PD-L1-positive patients in the first-line setting (cohort B of KEYNOTE-086)7. Although these cohorts are not directly comparable to ours due to different levels of prior therapy and disease burden, the survival observed in our study suggests that the addition of SABR to ICI may provide clinical benefit in selected patients. Two smaller clinical trials have previously assessed the safety and efficacy of radiotherapy in combination with immune checkpoint blockade in metastatic TNBC, with similar results to the AZTEC trial. A single-arm phase II trial by Ho et al enrolled 17 patients delivering a radiotherapy dose of 30 Gy in 5 fractions to a single tumor site, followed by the ICI pembrolizumab. This trial reported an ORR of 17.6% and a median PFS of 2.6 months, comparable to the AZTEC trial46. The phase II TONIC trial was a noncomparative study in which 67 patients with metastatic TNBC were randomly assigned to PD-1 blockade with nivolumab with one of five 2-week induction treatments: no induction or 2-week low dose induction, radiotherapy (3 × 8 Gy) to a single index lesion, cyclophosphamide, cisplatin or doxorubicin47. The ORR in the RT arm was 8% (in only 12 patients), leading to early termination due to futility. Although the numbers are small, our trial results compare favorably. When considering the entire study population in TONIC, patients were more heavily pre-treated (27% of patients with 2–3 lines of prior systemic therapy for metastatic disease) and had a higher proportion with visceral metastases (71%), compared with AZTEC. Interestingly, 11 (22%) patients in TONIC demonstrated durable clinical activity beyond 24 weeks, all of whom had oligometastatic disease, with 10 of the 11 patients having two or three metastases in total. This finding suggests that ICI-based treatments may be more effective in patients with reduced tumor burden, as demonstrated in our study. This hypothesis is supported by studies in other tumor types where a lower volume of metastatic disease was associated with an improved response to ICI48–50 and in trials where ICI was used following definitive local therapy to manage microscopic disease51–53. No randomized trials comparing ICI monotherapy to ICI with radiotherapy in any metastatic malignancies have met their primary endpoints54–58. However, the PEMBRO-RT trial in non-small cell lung cancer55 did demonstrate an improved ORR, and a significant OS benefit in the PD-L1 negative subgroup of patients treated with radiotherapy and pembrolizumab. This supports the possibility that SABR may prime PD-L1-negative tumors for response to ICI. Several factors can account for the discrepancy and limited clinical efficacy between pre-clinical studies and our trial. Key among these are the differences in the tumor microenvironment and immune system responses between human and animal models59–61. Additionally, the optimal radiation dose, fractionation, and sequencing with immunotherapy have not yet been established, with some pre-clinical data demonstrating that Trex1 is induced by radiation doses above 12–18 Gy, resulting in reduced immunogenicity33,45. Advanced TNBC may also present inherent immunosuppressive characteristics, tumor heterogeneity, and significant disease burden, which could hinder the effectiveness of such combination therapies62. Currently, the most effective partner for PD-1 inhibitors in TNBC has been cytotoxic chemotherapy, with randomized phase III evidence in KEYNOTE-355 demonstrating improved OS seen in PD-L1 positive patients with advanced TNBC17. The median OS was 23 months in the pembrolizumab–chemotherapy group and 16.1 months in the placebo–chemotherapy group (p = 0.0185), with no benefit seen in PD-L1 negative patients. The median PFS for all patients of PD-L1 status was 7.5 months in the experimental group and 5.6 months in the control group, which were both superior to the PFS in our study, underscoring the importance of chemotherapy as a partner for ICI. First-line atezolizumab in combination with chemotherapy in locally advanced or metastatic TNBC demonstrated a clinically meaningful OS benefit in PD-L1 positive patients, despite no OS benefit in the intention-to-treat population63. The use of SABR in addition to ICI and chemotherapy, particularly in PD-L1 negative patients, may be worth exploring in future studies. While PD-1/PD-L1 expression on tumoral immune cells has been investigated as a biomarker for immune checkpoint inhibitor (ICI) response, results vary64,65. Our study found no association between baseline peripheral immune cell PD-1 expression on the CD8+ or CD4+ T cells and treatment response, aligning with previous findings in TNBC patients65. However, PD-1 expression on circulating CD8+ T cells increased post-SABR, potentially priming cells for a response to atezolizumab, corroborated by other studies66. This suggests that in some patients, SABR to metastases may enhance atezolizumab efficacy irrespective of baseline PD-L1 status, while a history of prior chemotherapy could adversely impact its activity. Also of interest was the observation of increased proportions of CD8+ TPEX in patients with clinical benefit. These data support previous publications of the importance of EOMES and TCF1 in maintaining progenitor function of T cells to ensure the continued availability of functional T cells capable of responding to ICI in cancer67–71. Both could be useful as early peripheral biomarkers of atezolizumab response if further validated. Safety and tolerability in this trial were comparable to previously described ICI monotherapy trials with ≥ grade 3 TRAEs occurring in approximately 10% of patients7,9. The AZTEC trial had a lower rate of ≥ grade 3 TRAEs (14%) compared with 24% in the pembrolizumab and radiotherapy trial of Ho et al46. This study has several limitations. We tested two dose-fractionation schedules without comparison to an atezolizumab monotherapy or chemotherapy arm. Currently, the standard of care for all first-line PD-L1-positive patients with advanced TNBC is a combination of pembrolizumab with chemotherapy. Although it would have been interesting to understand whether SABR could improve disease control in the setting of chemo-immunotherapy, our study was designed before the results of the pivotal KEYNOTE-355 study were known. As our sample size was modest due to the phase II design, the power to reject the null hypothesis may have been influenced by the heterogeneity of metastatic TNBC. This could also have limited our ability to make robust conclusions in the biomarker analysis and translational endpoints. The biomarker findings, however, warrant further validation. In conclusion, this trial did not find sufficient evidence to reject the null hypothesis regarding the efficacy of either single- or multi-fraction SABR when combined with atezolizumab in patients with advanced TNBC. However, we observed acceptable toxicity profiles and durable disease control in patients with oligometastatic disease, independent of PD-L1 expression. Baseline and on-treatment levels of CD8+ PD-1+ TCF1+ TPEX emerged as potential predictive biomarkers that need further validation. Methods The Peter MacCallum Cancer Center (PMCC) ethics committee approved the study (HREC/11/PMCC/109) at all participating sites nationwide. The study was conducted in accordance with the Australian National Statement on Ethical Conduct in Human Research, the Declaration of Helsinki, and applicable Good Clinical Practice guidelines. All participants provided written informed consent, and personally identifiable information was removed from the data to ensure anonymity. The study followed the Consolidated Standards of Reporting Clinical Trials (CONSORT) guideline. The clinical trial protocol and pre-specified statistical analysis plan are available in the Supplementary Information. Study design and patient population This was a multicenter, open-label, phase II randomized clinical trial (NCT03464942) led by PMCC in collaboration with the Trans-Tasman Radiation Oncology Group (TROG 17.05). Key inclusion criteria included histologically confirmed metastatic TNBC, unselected for PD-L1 status, who had de novo metastatic disease or recurred at least six months post completion of (neo)adjuvant chemotherapy, and who had received less than two lines of treatment in the metastatic setting. Participants were between 18 and 75 years of age, with Eastern Cooperative Oncology Group (ECOG) performance status of 0–1, with at least two metastases overall, of which at least one was suitable for SABR and at least one assessable lesion (measurable or non-measurable) as per Response Evaluation Criteria in Solid Tumors version 1.1 (RECIST 1.1), not designated for SABR. Key exclusion criteria were previous high-dose radiotherapy (biologically equivalent dose > 30 Gy) to an area to be treated, previous ICI therapy, evidence of active brain metastases or cord compression, and any contraindications to immunotherapy. Procedures Patients were randomly assigned to receive either 20 Gy in 1 fraction or 24 Gy in 3 fractions for 1 to 4 metastases, with at least one metastasis untreated, based on the physician discretion. Patients were stratified by the number of prior lines of chemotherapy (0 or 1). Clinicians were encouraged to treat the maximum feasible number of lesions with SABR (up to 4). Corticosteroid use was not permitted before or after SABR delivery. Radiotherapy planning, delivery, and quality assurance were overseen by TROG, as per Supplement 3. Since the primary goal of SABR in this trial was to induce a systemic abscopal response, organs at risk were prioritized over target coverage. Any compromises in target coverage required real-time review by the chief investigator. All patients were then administered atezolizumab up to 5 days after the last SABR treatment at a fixed dose of 1200 mg intravenously on day 1 of a 3-week cycle, until progression or intolerable toxicity for a maximum of 24 months. Patients were seen during SABR treatment, then every cycle, 30 days after ceasing treatment, and every 12 weeks after ceasing treatment for two years. Imaging, which included a computed tomography scan of the chest, abdomen, and pelvis, was performed every six weeks ±5 days from day one cycle one of atezolizumab for the first 12 months and then every 12 weeks ±7 days for two years. Other relevant imaging was performed as clinically indicated. Outcomes The primary endpoint was progression-free survival (PFS), defined as the time from randomization to either the first documented disease progression or death, whichever occurred first. PFS was assessed using Response Evaluation Criteria in Solid Tumors (RECIST) 1.1 criteria. Irradiated lesions were excluded from response assessments unless there was unequivocal progression at those sites. Secondary endpoints included best overall response (as per RECIST 1.1), clinical benefit rate (CBR: CR/PR or SD for 24 weeks or more), overall survival (defined as the time from randomization until death from any cause), and safety assessed using NCI-CTCAE version 4.03. Exploratory analysis included efficacy by PD-L1 IHC status (≥1% vs <1%), using the SP142 assay and tumor infiltrating lymphocytes (TIL) quantity (<5% vs ≥5%). We also compared treatment single vs multi-fraction arms, knowing that the power to see a difference was low (~50%, alpha = 0.1). Biomarker analysis Tumor biopsies Fresh metastatic tumor biopsy was taken at baseline, cycle 1 of atezolizumab, and at progression, where feasible. PD-L1 immunohistochemistry (IHC) was performed using the SP142 assay with staining in 1% or more immune cells as positive. Tumor-infiltrating lymphocytes (TILs) were quantified on H&E-stained slides from the most recent biopsy by a single pathologist as per published guidelines72. TIL high was ≥5%73. Both analyses were performed by Nexomics74. Flow cytometry analysis (FACS) Peripheral blood mononuclear cells (PBMCs) were collected from patients at baseline (before SABR), within 5 days after the last SABR treatment but prior to the first dose of atezolizumab, ± 3 days of every subsequent alternating dose of atezolizumab, and at progression. Peripheral blood samples were obtained by 9 mL venipuncture and gently mixed by inverting the tubes 8–10 times. After collection, PBMC samples were taken and securely transported to and stored at The Peter MacCallum Cancer Center. PBMCs were isolated from blood samples using the Ficoll-Paque method, stored using standard protocols, and thawed at later time points to assess PBMC frequencies. Thawed PBMC samples were filtered through a 70 μm filter, counted with a hemocytometer, and aliquoted into a 96-well plate at ~1 × 106/ well. Homogenized cells were labeled with monoclonal antibodies for 30 min at 4 °C in FACS buffer (2% FBS in Dulbecco’s PBS) and then washed twice in FACS buffer. Intracellular staining was performed as per the reagent instructions. The flow cytometry panel included T cell markers CD3, CD4 and CD8, exhaustion marker PD-1, T cell differentiation markers CCR7 and CD45RA, T cell activation marker CD69, intracellular T cell markers EOMES, T-bet, TCF1, IFNγ and TNFα, Natural Killer cell marker CD56 and Regulatory CD4+ T cell (Treg) markers CD25 and FoxP3. Viable cells were revealed using fixable viability dye. Multiparameter FACS data were acquired on the BD Symphony FACS instrument (BD Biosciences) and analysed using FlowJo software (Version 10.4) (Fig. S6). PBMCs from 3 different time points were collected for analysis: prior to any trial therapy (Baseline), within 5 days post SABR prior to the first dose of atezolizumab (Cycle 1), and before the 3rd dose of atezolizumab (Cycle 3). Statistical analysis The sample size was calculated separately for each arm, evaluating whether single or multi-fraction SABR synergizes with atezolizumab. We assumed a median PFS of 2 months was uninteresting (H0) and not worthwhile to study this treatment combination further, as this was expected with atezolizumab monotherapy alone7–9. We also assumed a clinically meaningful improvement of at least 70% in median PFS for SABR in combinations with atezolizumab (i.e., H1 of 3.4 months) to justify further studying this therapeutic approach. With 80% power, 0.05 alpha, 1-sided one-sample log-rank test, 18 months of recruitment, all patients followed until 2 years after the last patient commenced treatment, and an exponential distribution, the required sample size was 24 participants in each arm. Assuming up to 4 participants may be lost to follow-up prior to progression, 52 participants were planned to be recruited to the trial. The body of evidence from each arm was considered when suggesting which arm should be pursued further in a large randomized phase III trial. The maximum toxicity grade per participant for each adverse event was derived and presented in table format according to arm and overall. The analysis of AEs was conducted twice: once for all AEs, regardless of relatedness to treatment, and once considering only AEs related to the treatment (i.e., classified as possibly, probably, or related). The rate of grade 3+ toxicities was compared between arms using Fisher’s exact test. Time-to-event outcomes were described for each arm using Kaplan-Meier methods. The stratified log-rank test was used to compare the curves between arms, and the stratified Cox proportional hazards model was used to estimate the hazard ratio (HR) and 95% CIs as per the pre-specified statistical analysis plan. For the median PFS estimation, a 90% CI was chosen, consistent with the sample size calculation and the signal-seeking nature of the trial design. The number of previous lines of therapy (first-line vs. not first-line) was a stratification variable for the stratified log-rank test and the stratified Cox proportional hazards model. The proportional hazard assumption was verified. ORR (CR/PR) and CBR were compared between arms using Fisher’s exact test. Prespecified exploratory biomarker analyses were conducted where PD-L1 IHC and TILs were classified as positive/negative, and their impact on PFS and OS was assessed using the Cox proportional hazards model, adjusting for treatment effect. Logistic regression was used to assess the relationship of PD-L1 IHC and TILs with CBR. TILs were also analysed as a continuous variable in 10% increments in relation to CBR using logistic regression, with the ability to discriminate patients with clinical benefit determined using receiver operating characteristic curves. Flow cytometry data from PBMCs were analysed and graphed using R (version 4.5.2) with the packages dplyr (1.1.4)75, ggplot2 (3.4.0)76, ggdist (3.3.3)77, patchwork (1.3.2)78, ROCR (1.0-11)79 and pROC (1.19.0.1)80. Wilcoxon rank sum and Wilcoxon signed-rank tests were used to determine statistically significant group-wise differences. All statistical tests were two-sided, with statistical significance defined as P < 0.05, given the exploratory nature of this analysis. For these analyses, participants were grouped according to protocol-defined clinical benefit versus not (CB-Y versus CB-N). Trial oversight The trial was overseen by a trial management committee via TROG and an independent data and safety monitoring committee via the sponsor. The trial protocol and all amendments were approved by the appropriate ethics body at each participating institution. All the patients provided written informed consent before enrollment. All the authors attest that the trial was conducted in accordance with the protocol and its amendments and with the standards of Good Clinical Practice. All authors had access to the data and participated in the manuscript drafting and final approval. The first authors wrote the manuscript without any medical editorial assistance. ClinicalTrials.gov identifier NCT03464942, preregistered March 13, 2018. Patients were randomized between November 2018 and April 2021. Role of funding source This study was designed by the study investigators. The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication. This research was made possible through the imCORE Network, supported by F. Hoffmann-La Roche and sponsored by Peter MacCallum Cancer Center, Melbourne, Australia. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Supplementary Data 4 Supplementary Data 5 Reporting Summary Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Supplementary Data 4 Supplementary Data 5 Reporting Summary Source data Source Data Transparent Peer Review file Source Data Transparent Peer Review file