Neoadjuvant atezolizumab in combination with dual HER2 blockade plus epirubicin in women with early HER2-positive breast cancer: the randomized phase 2 ABCSG-52/ATHENE trial

In patients with high-risk breast cancer subtypes, pathological complete response (pCR) to neoadjuvant treatment is associated with an improved long-term outcome on an individual patient level¹. Therefore, for most patients with clinical stage II/III human epidermal growth factor receptor 2 (HER2)-positive early breast cancer (EBC), neoadjuvant dual HER2 blockade with trastuzumab and pertuzumab (TP) plus a taxane with or without an anthracycline is considered the standard of care, even when primary breast conservation seems feasible^2,3,4. In phase 2 and 3 trials investigating TP plus polychemotherapy regimens in higher-risk patients (≥T2 and/or ≥N1), pCR rates of 55–62% have been reported^5,6,7. In patients not achieving pCR, postneoadjuvant treatment with the HER2-directed antibody–drug conjugate trastuzumab emtansine has further improved outcomes⁸. Therefore, HER2 positivity has changed from a subtype-defining biomarker conferring poor prognosis to a positive predictive biomarker with currently available HER2-directed treatment options having vastly improved long-term outcomes.

To balance toxicity burden and treatment activity, research in recent years has focused on chemotherapy de-escalation⁹. Owing to cardiotoxicity concerns, especially in combination with anti-HER2 blockade, taxanes were considered the preferred chemotherapy backbone for de-escalation protocols over anthracyclines. In clinical trials investigating neoadjuvant monochemotherapy with a taxane plus TP, pCR rates of 39% and 56–91% were seen in patients with HER2-positive higher-risk and lower-risk EBC, respectively^10,11,12.

Besides their direct cytotoxic effects, conventional chemotherapeutic agents are also known to harbor immunogenic properties¹³. Doxorubicin has been shown to enhance dendritic cell maturation¹⁴, promote the antigen-presenting abilities of mouse dendritic cells¹⁵ and induce the expression of heat shock proteins in vitro¹⁶. Apart from doxorubicin, other anthracyclines (that is, epirubicin and idarubicin) can also trigger immunogenic cell death, a regulated cell death that engages the adaptive immune system¹⁷.

These data provide a sound rationale for combining an anthracycline with immune checkpoint inhibitors (ICBs). Additionally, compared to taxanes, anthracyclines have a more favorable side-effect profile owing to their lack of neurotoxicity and hypersensitivity reactions. Atezolizumab is a monoclonal antibody targeting programmed death ligand 1 (PD-L1)¹⁸. Besides many other indications over a broad spectrum of neoplastic diseases, it is approved by the European Medicines Agency for the treatment of PD-L1 immune cell-positive metastatic triple-negative breast cancer¹⁹.

The role of atezolizumab in addition to standard polychemotherapy plus dual HER2 blockade was investigated in two phase 3 trials^20,21. In IMpassion050, atezolizumab with neoadjuvant dose-dense doxorubicin, cyclophosphamide, paclitaxel and TP did not increase pCR rates versus placebo in the intention-to-treat (ITT; pCR rate of 62.7% in the placebo group and 62.4% in the atezolizumab group, P = 1.00) or PD-L1-positive (pCR rate of 72.5% in the placebo group and 64.2% in the atezolizumab group, P = 0.18) population. In the three-arm APTneo trial, atezolizumab plus an anthracycline-containing regimen (doxorubicin, cyclophosphamide, paclitaxel, carboplatin and TP) increased the pCR rate compared to an anthracycline-free control group (paclitaxel, carboplatin and TP (HPCT); pCR rate of 61.9% versus 52.0%, P = 0.022). In contrast, atezolizumab plus HPCT compared to HPCT showed a similar pCR proportion (53.6% versus 52.0%, P = 0.089).

In the single-arm Keyriched-1 trial, the programmed death 1 inhibitor pembrolizumab was investigated in combination with TP in patients with EBC of a molecular HER2-enriched intrinsic subtype²². With this chemotherapy-free combination, a pCR rate of 46% was seen. Patients with HER2-enriched subtypes have a higher likelihood of achieving pCR following anti-HER2-based neoadjuvant therapy with or without chemotherapy²³. In the single-arm Neo-PATH trial, a combination of neoadjuvant atezolizumab, docetaxel and TP was investigated in HER2-positive patients²⁴. In this monochemotherapy and immunotherapy trial, a pCR rate of 61% was reported.

The results from the phase 3 IMpassion050 trial indicate that ICBs have no additive effect when combined with a standard anti-HER2 therapy plus polychemotherapy regimen²⁰, which has the highest assumed effectiveness in terms of pCR^{5,6,7,10,11,12}. Therefore, we hypothesized that the addition of an ICB to dual HER2 blockade may add activity in regimens with a de-escalated chemotherapy backbone. The optimal cytotoxic drug in such an approach remains to be determined, and the particular role of epirubicin monochemotherapy has not been previously investigated. Therefore, in ABCSG-52/ATHENE, we investigated an anthracycline-based chemotherapy de-escalation immunotherapy regimen in patients with HER2-positive EBC.

Patient and tumor characteristics

Between June 2020 and December 2021, 70 patients were screened and 58 patients (ITT population) were randomized 1:1 to TP plus atezolizumab (TP-A; n = 29) or TP alone (n = 29) at nine sites in Austria (Fig. 1). All patients received at least one dose (safety population). Because one patient in the TP-A group withdrew informed consent before surgery and one patient in the TP group received only one treatment cycle, the final efficacy assessment population consisted of 56 patients (Fig. 2).

Stratification criteria: baseline TILs (<5% versus ≥5%), HR status (positive versus negative) and prognostic stage (≤IIA versus ≥IIB (American Joint Committee on Cancer staging manual version 8.0)). ASCO, American Society of Clinical Oncology; CAP, College of American Pathologists; ECOG, Eastern Cooperative Oncology Group; LVEF, left ventricular ejection fraction; R, randomization; TP-A + E, TP-A plus epirubicin.

Full size image

IC, informed consent.

Full size image

In the ITT population, the median age was 57 years (range 33–82 years). All included patients were women, and 34 patients (59%) were postmenopausal at baseline. Of the enrolled patients, 42 (72.4%) presented with hormone receptor (HR)-positive tumors and 16 (27.6%) had HR-negative disease. According to the clinical prognostic stage, 45 patients (77.6%) were classified as having stage ≤IIA disease and 13 (22.4%) patients had stage ≥IIB disease. The detailed characteristics of the total population and by treatment arm are shown in Table 1.

Primary and secondary endpoints

In the ITT population, pCR (primary endpoint) was observed in 60.3% of patients (n = 35/58; 95% confidence interval (CI) 47.5% to 71.9%), 65.5% (n = 19/29; 95% CI 47.3% to 80.1%) in the TP-A group and 55.2% (n = 16/29; 95% CI 37.5% to 71.6%) in the TP group (difference: 10.3%; 95% CI −14.7% to 35.4%). In patients with an available residual cancer burden (RCB) assessment (secondary endpoint), complete or near-complete remission, defined as RCB category 0 or I, was seen in 80.0% (n = 44/55; 95% CI 67.6% to 88.4%), 85.7% (n = 24/28; 95% CI 68.5% to 94.3%) in the TP-A group and 74.1% (n = 20/27; 95% CI 55.3% to 86.6%) in the TP group (difference: 11.6%; 95% CI −9.4% to 32.6%). The rates of RCB category 0–III per treatment arm are shown in Supplementary Table 2.

In a univariable logistic regression model (Fig. 3), numerically lower pCR rates were observed in peri-/premenopausal patients compared to postmenopausal patients (odds ratio (OR) 0.48; 95% CI 0.16 to 1.40; two-sided P = 0.18), as well as in histological subtypes other than ‘no special type’ (OR 0.37; 95% CI 0.09 to 1.48; two-sided P = 0.16). Higher pCR rates were seen in patients with increased body mass index (BMI; OR per 10-unit increase 1.97; 95% CI 0.62 to 6.22; two-sided P = 0.25). None of the covariates were statistically significantly associated with pCR.

The prognostic stage is given according to the American Joint Committee on Cancer staging manual version 8.0. The column ‘alternative cat.’ refers to the alternative category, and ‘reference cat.’ refers to the reference category of the covariate. An OR of >1 indicates a higher pCR rate of the alternative category (left category in the ‘category’ column) or with higher age and BMI. Events: patients achieving pCR; sample size: 58. P values are from two-sided Wald tests with no adjustment for multiple testing. Covariate effects are presented as ORs including 95% CIs. NST, invasive carcinoma of no special type.

Full size image

Radiological complete response, radiological partial response or radiological stable disease was detected in 21 (37.5%), 29 (51.8%) and 6 (10.7%) patients, respectively. No radiological progressive disease was seen. The overall response rate (radiological complete response + radiological partial response), the other secondary endpoint, was 89.3% (95% CI 78.5% to 95.0%).

Response according to tumor-infiltrating lymphocyte status

In a post hoc exploratory analysis, the mean proportion of stromal tumor-infiltrating lymphocytes (TILs) was 23.9% in the overall population, 22.8% in the TP group and 25.0% in the TP-A group. A lymphocytic-predominant phenotype was seen in 10.3% of patients in the TP group and 17.2% of patients in the TP-A group. No association was detected between TIL proportion at baseline and pCR (OR for a 10-percentage-point increase 1.02; 95% CI 0.78 to 1.34). A moderate positive correlation between the numeric values of TILs and PD-L1 was observed (Spearman r = 0.57, two-sided P < 0.0001).

Response according to PD-L1 status

In a post hoc exploratory analysis, the pCR rate was 69.2% (n = 18/26; 95% CI 50.0% to 83.5%) in PD-L1-negative patients compared to 55.2% (n = 16/29; 95% CI 37.5% to 71.6%) in PD-L1-positive patients. The highest pCR rates were detected in the PD-L1-negative subgroup treated in the TP-A arm (73.3%; 95% CI 48.0% to 89.1%), whereas the lowest pCR rates were observed in PD-L1-positive patients treated in the TP arm (52.9%; 95% CI 31.0% to 73.8%). pCR rates according to treatment arm and PD-L1 status are shown in Table 2.

Adverse events

Treatment-emergent adverse events (AEs) grade ≥3 were reported in 17 patients (29.3%), 9 patients (31.0%) in the TP-A group and 8 patients (27.6%) in the TP group (Table 3 and Supplementary Table 3; AEs per treatment part are listed in Supplementary Tables 4 and 5). The most frequently reported AEs in both treatment groups were nausea (69% in both groups), diarrhea (59% in TP-A, 62% in TP), fatigue (48% in TP-A, 59% in TP) and alopecia (41% in TP-A, 28% in TP; Table 3). No AEs of special interest grade ≥3 were detected (Supplementary Table 6); therefore, none of the predefined boundaries were crossed.

The introduction of ICBs, given either alone or in combination with chemotherapy or small molecules, has considerably changed the outcome in many cancer types and therefore the landscape of standard treatments of neoplastic diseases²⁵. Such positive effects of ICBs on response, disease-free survival or overall survival have been reported when used in the neoadjuvant (for example, lung cancer), adjuvant (for example, melanoma), or advanced and/or metastatic (for example, cancers of the head and neck, lung, esophagus and stomach, liver, and urinary system) setting. In breast cancer, results have been less convincing, except for triple-negative subtypes²⁶, which are considered to carry a higher number of mutations and thus neoepitopes allowing T cells to recognize the neoplastic cells²⁷. Recently, initial promising results of the addition of ICBs to neoadjuvant chemotherapy in patients with high-risk luminal cancer have been presented^28,29. Few data are available concerning the efficacy of ICBs in the HER2-positive subtype, particularly in the neoadjuvant setting. While this subtype can already be treated with high efficacy across all disease stages using HER2-targeted therapies, substantial room for improvement remains in two directions: further increasing pCR rates (and thus the depth of short- and long-term tumor control) and decreasing toxicity and side effects while increasing the quality of life by chemotherapy de-escalation strategies. Following the requirement for both goals, we initiated this proof-of-principle phase 2 randomized trial to test the role of atezolizumab in addition to dual HER2 blockade in an initial induction phase followed by only four cycles of a quadruple regimen combining the three antibody drugs with epirubicin monotherapy in all patients.

pCR has been widely accepted as the primary outcome parameter in the neoadjuvant treatment of breast cancer³⁰ and is the primary endpoint of this trial. After part 1 (two cycles) and part 2 (four cycles) of therapy, we observed a pCR rate of 60.3% (95% CI 47.5% to 71.9%) in the overall population, which exceeded the predefined threshold for effectivity (a pCR proportion of ≥40%).

Comparing these results to previously published data suggests that chemotherapy de-escalation with an intensified immunotherapy and abbreviated chemotherapy regimen, as investigated in ABCSG-52/ATHENE, is effective. In the single-arm Neo-PATH trial²⁴, the combination of six cycles of atezolizumab, pertuzumab, trastuzumab and docetaxel yielded a pCR rate of 61% (90% CI 50% to 71%). This result seems comparable to the pCR rate of 65.5% (95% CI 47.3% to 80.1%) in the TP-A group in our trial, in which only four cycles of epirubicin in combination with six cycles of immunotherapy were applied. However, more patients with a higher risk for recurrence were enrolled in Neo-PATH compared to ABCSG-52/ATHENE (proportion of patients with clinical stage ≥IIB disease: 76% versus 22%). Only a few classes of cytotoxic drugs exert proimmunogenic effects that may synergize with the mechanism of ICBs and vice versa³¹. Both epirubicin and taxanes belong to these groups of drugs, although their mechanisms of interaction with the interplay between cancer cells and immune cells may differ. The differences in results between ABCSG-52 and Neo-PATH might, in part, be due to these discrepancies. The particular role of anthracyclines in combination with ICBs is supported by recently published results from the APTneo trial²¹: the addition of atezolizumab to an anthracycline-containing regimen increased pCR rates compared to an anthracycline-free regimen, whereas no additive effect of atezolizumab was observed when it was added to an anthracycline-free protocol. This is in line with findings from neoadjuvant phase 3 trials investigating combinations of an ICB plus polychemotherapy in triple-negative EBC: pCR rates were increased only with anthracycline-containing regimens^32,33 but not with anthracycline-free, taxane-including protocols³⁴. This hypothesis is further supported by results from the noncomparative phase 2 TONIC trial³⁵. In this study, induction therapy with a short course of anthracycline outcompeted other cytotoxic drugs (cyclophosphamide, cisplatin) in terms of the effect of a subsequent ICB in the metastatic setting of triple-negative breast cancer.

In addition, the combination of ICBs with chemotherapy may be time- and dose-sensitive. We randomized patients to induction with dual HER2 blockade alone versus a combination with atezolizumab for two cycles, followed by four cycles of the quadruple regimen. The underlying concept was that this induction therapy reveals neoepitopes for the priming and activation of antigen-presenting cells and T cells. This could help fully exploit the proimmunogenic effect of epirubicin³⁶. In an exploratory analysis comparing the six versus four applications of atezolizumab (TP-A versus TP), the pCR rate was higher in the six-cycle arm (difference: 10.3%; 95% CI −15% to 35%). This finding may be considered supportive of such a priming effect and suggests that an ICB should be included upfront in the neoadjuvant setting to maximize treatment effects. As no atezolizumab-free treatment arm was included in our phase 2 trial, the quantitative effect of atezolizumab cannot be clarified in this setting. Future translational data may help clarify the biological mechanisms behind our findings.

Regarding pCR rates in subgroups, increased BMI was numerically associated with higher pCR rates on univariable analysis. This is in line with previous findings that BMI was correlated with higher responses to anthracyclines in the neoadjuvant setting³⁷. In our trial, 79% of pre-/perimenopausal women had HR-positive disease compared to 68% in postmenopausal patients. As HR positivity is associated with lower pCR rates in HER2-positive tumors^5,6,7,10, this might explain why numerically lower pCR rates were seen in pre-/perimenopausal women enrolled in our trial.

In ABCSG-52, the numerically highest pCR rates were observed in patients with PD-L1-negative tumors at baseline biopsy and when treated within the TP-A group (pCR rate: 73.3%; 95% CI 48.0% to 89.1%). While this seems counterintuitive, it was shown that trastuzumab-sensitive cancers produce cytokines such as CCL2 (chemokine ligand 2) capable of attracting monocytes, macrophages, dendritic cells and memory T cells in the tumor tissue³⁸. In such a trastuzumab-sensitive microenvironment, PD-L1 was upregulated³⁸. Because of an already sufficient antitumor immune response mediated by the antibody-dependent cellular cytotoxicity and antibody-dependent cellular phagocytosis effects of anti-HER2 antibodies, the addition of anti-PD-L1 antibodies to standard therapy may lack additional effects. Thus, PD-L1 positivity can be understood as a surrogate for a trastuzumab-sensitive microenvironment rather than a predictor of the efficacy of immune checkpoint treatment^39,40. This is in line with our findings and those of the IMpassion050 trial (no additive effect of atezolizumab in PD-L1-positive patients)²⁰. Conversely, trastuzumab-resistant microenvironments have been described as immunosuppressive⁴¹ and having lower PD-L1 expression compared to trastuzumab-sensitive tumors³⁸. In such a scenario, trastuzumab can lead to PD-L1 upregulation in immune and cancer cells, and the addition may resensitize toward HER2 targeting^42,43. This again is in line with the findings of ABCSG-52/ATHENE and IMpassion050 (increased pCR rates in atezolizumab versus placebo in PD-L1-negative patients)²⁰ but in contrast to the findings of the Neo-PATH trial²⁴. Therefore, further clinical and experimental investigations are required.

Anthracyclines not only are effective cytotoxic drugs against breast cancer but also exert cardiotoxic effects in a dose-dependent manner. The exploitation of their proimmunogenic effects is important, provided their cumulative doses can be limited. The fact that only four cycles of epirubicin were needed for a high pCR rate and no cardiotoxic effects were observed despite the combination with dual HER2 blockade and checkpoint inhibitors is promising and reassuring. In fact, the only grade 2 cardiac event occurred in the chemotherapy-free phase of the initial ICB and dual anti-HER2 therapy. Regarding immune-related side effects, continuous toxicity monitoring was included in our trial, and no grade ≥3 toxicities of special interest (immune-related AEs, cardiac disorders grade ≥2 or infusion-related reactions) were detected.

In addition to anthracyclines, taxanes, platinum compounds and alkylants (for example, cyclophosphamide) also have cardiotoxic potential. Whether a regimen with four cycles of epirubicin is more cardiotoxic than an anthracycline-free regimen consisting of six cycles of docetaxel and carboplatin is unanswered. In our view, an anthracycline-containing de-escalation protocol as in ABCSG-52/ATHENE is justified in terms of cardiotoxicities, and our toxicity data support this assumption.

The findings of this study are not without limitations. Because of the phase 2 design, this study was small and did not have an additional arm without ICB therapy over the whole treatment course. Owing to the short observation periods, no details regarding long-term outcomes, such as invasive disease-free survival, can yet be reported. We stratified patients according to the number of stromal TILs as this is an accepted prognostic marker in breast cancer and a stromal TIL proportion of ≥5% was predictive of response to an ICB combined with trastuzumab in pretreated patients in the phase 1/2b PANACEA trial⁴⁴. Future trials should consider both the number of TILs and basal PD-L1 expression.

In summary, our data provide evidence that the addition of anti-PD-L1 inhibitors to abbreviated monotherapy with an anthracycline leads to high pCR rates in HER2-positive breast cancer. Our study also raises interesting questions about the sequencing of chemotherapy and immunotherapy in a combined approach and regarding the biological meaning of PD-L1 expression and its therapeutic inhibition in relation to the sensitivity of tumor cells against HER2 blockade.

Study design

ABCSG-52/ATHENE is a multicenter open-label, two-arm, randomized, single-stage phase 2 study (Fig. 1). Registration of the study in the European Union Clinical Trials Register was performed before the inclusion of the first patient (EudraCT no. 2019-002364-27). The study protocol was reviewed and approved by an independent ethics committee (Ethics Committee of the County of Salzburg, Austria). The first patient was enrolled on 3 July 2020, and the last patient was enrolled on 2 December 2021. The full study protocol is provided in Supplementary Information. The study design and conduct complied with all relevant regulations regarding the use of human study participants. The study was conducted in accordance with the criteria set by the Declaration of Helsinki. The CONSORT (Consolidated Standards of Reporting Trials) guidelines were followed⁴⁵. Randomization, power calculation and statistical tests comply with the ICMJE (International Committee of Medical Journal Editors) guidelines on reporting.

Patients

Patients with previously untreated, histologically confirmed HER2-positive EBC with a clinical prognostic stage of cT1c to cT4a–d, N0–3 and M0, with adequate cardiac (ejection fraction ≥55%), renal, liver and bone marrow function were eligible for this trial. Patients with a history of malignancies other than nonmelanoma skin cancer and in situ carcinomas and those with a history of autoimmune disease, bilateral breast cancer or other concomitant serious medical conditions were excluded from trial participation. All patients signed an informed consent form before study enrollment. Patients were not compensated for clinical trial participation.

Randomization and masking

Patients were randomized 1:1 to two 3-weekly cycles of a chemotherapy-free induction phase (part 1) with TP-A or TP alone. Thereafter, all patients received four cycles of TP-A in combination with epirubicin (part 2). Randomization was done with a centralized web-based system using a minimization algorithm including the three stratification criteria: (1) baseline stromal TILs: <5% versus ≥5%; (2) HR status: HR positive versus HR negative; and (3) prognostic stage: ≤IIA versus ≥IIB (according to the clinical prognostic stage groups defined by the American Joint Committee on Cancer staging manual version 8.0). PD-L1 expression status was neither an inclusion nor a stratification factor.

Procedure

Study treatment

In the ABCSG-52/ATHENE study, treatment consisted of two parts. For part 1, both treatment groups received two 3-weekly cycles of pertuzumab (starting with 840 mg administered intravenously (IV) on cycle 1, followed by 420 mg IV for the subsequent cycles) and trastuzumab (starting with 600 mg administered subcutaneously (SC) or 8 mg kg⁻¹ IV on cycle 1, followed by 600 mg SC or 6 mg kg⁻¹ IV for the subsequent cycles). In the TP-A group, two 3-weekly cycles of atezolizumab (1,200 mg IV per cycle) were added.

For part 2, both groups received four 3-weekly cycles of atezolizumab (1,200 mg IV per cycle), pertuzumab (420 mg IV per cycle), trastuzumab (600 mg SC or 6 mg kg⁻¹ IV per cycle) and epirubicin (90 mg m⁻² per cycle).

Adjuvant treatment was not part of our neoadjuvant trial. If pCR was not achieved, a taxane-based adjuvant chemotherapy was recommended. After the completion of adjuvant standard anti-HER2 therapy and in the case of HR-positive patients, standard endocrine therapy was recommended.

Tumor-infiltrating lymphocytes

TILs in formalin-fixed, paraffin-embedded tissues from diagnostic biopsies performed before treatment initiation were assessed by a breast cancer pathologist. Stromal TILs were counted according to the recommendation of the International TIL Working Group⁴⁶.

PD-L1 expression

The expression of PD-L1 was assessed in formalin-fixed, paraffin-embedded tissues from diagnostic biopsies performed before the start of treatment. For PD-L1 staining, the Ventana PD-L1 SP124 assay was used according to the manufacturer’s protocol (no dilution was required). PD-L1 positivity was defined as at least 1% PD-L1-expressing tumor-infiltrating immune cells⁴⁷.

Outcome parameters

The primary outcome was efficacy with regard to pCR (ypT0/is, ypN0), which was assessed in the overall study population at the time of surgery. This assessment was performed locally at each site.

The secondary outcomes were RCB and overall response rate, which were both assessed in the overall study population at the time of surgery.

Safety control

AEs were assessed and coded according to Common Terminology Criteria for AEs version 5.0. In addition, a strict, continuous safety monitoring program was designed, and the proportion of patients with at least one grade ≥3 AE of special interest, including immune-related AEs, cardiac disorders and/or infusion-related reactions (see Supplementary Table 1 for detailed definitions), was closely monitored. Prespecified Pocock-type boundaries were implemented⁴⁸, and a proportion of 20% was considered safe. The sequential boundaries were calculated such that the trial would have been stopped early with a 5% probability in case the true rate of grade ≥3 AEs of special interest was as high as 20%. Data on AEs were collected until the postsurgery visit (within 7–42 days after surgery but at least 42 days after the last dose of neoadjuvant study treatment).

Echocardiography was required at screening (within 28 days before randomization), before the first administration of epirubicin and before surgery. Additional assessments were performed as clinically indicated. At screening, patients with a left ventricular ejection fraction of <55% were not eligible for study participation. During treatment, patients whose left ventricular ejection fraction decreased to <50% had to permanently discontinue the study treatment.

Statistical analysis

Sample size assumptions were based on the pCR data reported in the NeoSphere trial¹⁰ as well as statistical and medical expert opinions. A pCR proportion of 40% in the overall study population was assumed to indicate relevant clinical activity of this regimen, and the main goal of this trial was to estimate the true pCR proportion with a given level of precision (that is, half-width of 95% CI = 13 percentage points). With these assumptions, a sample size of 55 patients would have been required. However, given that CIs are the widest when the point estimate is 50%, the sample size calculation was based on an assumed pCR rate of 50% and yielded 57 patients. Owing to 1:1 randomization, 58 patients were randomized to achieve a balance between arms. However, the study was considered positive if at least 40% of the ITT population achieved pCR.

The primary endpoint, pCR, was assessed in the full analysis set consisting of all randomized patients. Patients were analyzed according to the ITT principle. The secondary endpoints, RCB⁴⁹ and overall response rate⁵⁰ at surgery, were assessed in all patients with a nonmissing measurement (modified ITT). Furthermore, efficacy endpoints were reanalyzed in the efficacy assessment population, which consisted of all patients who received at least two cycles of the study treatment and were assessable for pCR status.

Safety analyses were conducted based on the safety population, which comprised all patients who received at least one dose of study treatments.

Wilson score 95% CIs were derived for the primary and secondary outcomes, and 95% Wald asymptotic CIs for treatment arm differences were derived. In addition, univariable logistic regression models were used to assess potential associations of clinical covariates with pCR. Covariates used included TIL proportion, HR status, prognostic stage, age, BMI, grade, menopausal status, histological type and PD-L1 status. All indicated P values (Wald tests) are two-sided. No multiplicity adjustments were used.

A nonpredefined exploratory analysis of pCR rate according to PD-L1 status was conducted in the ITT population with available PD-L1 data.

Analyses were performed using SAS software version 9.4 (SAS Institute).

As the trial start coincided with the COVID-19 outbreak in 2020, site initiations were evaluated in close consultation with the trial sites and according to available resources and local circumstances. Because of an increase in the number of infections (following the peak of the COVID-19 outbreak in Austria in the spring of 2020), an evaluation of the benefit–risk balance was performed in cooperation with the ABCSG-52/ATHENE coordinating investigators and in consultation with the investigational medicinal product provider (Roche), and the results were shared with the trial site teams. According to this guidance, study conduct (including enrollment and study treatment) was continued per protocol considering site-specific and/or countrywide COVID-19 government measures and guidance documents (for example, from the Austrian Ministry of Health).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.