Neoadjuvant Trastuzumab and Pyrotinib for Locally Advanced HER2-Positive Breast Cancer (NeoATP): Primary Analysis of a Phase II Study

HER2 has been identified as an oncogene in many cancers, including breast cancer, since the 1980s (1–4). Trastuzumab is the first and indispensable agent targeting HER2. The NOAH trial has corroborated the significant improvement in the pathologic complete response (pCR) rate with the addition of trastuzumab to neoadjuvant chemotherapy (NAC) for HER2-positive breast cancer (5). Furthermore, those who achieve pCR yield superior survival to those who don't (6, 7). Many clinical trials are being or have been conducted to explore how to increase the likelihood of pCR for patients with HER2-positive diseases following neoadjuvant treatment (NAT).

The accumulating evidence has demonstrated that dual HER2 blockade contributes to a higher pCR rate than single trastuzumab when combined with the same NAC regimen for HER2-positive breast cancer (8–10). Nevertheless, nearly half of patients still failed to reach pCR so far. On the other hand, women with HER2-positive tumors seem to derive numerically more benefit in pCR from trastuzumab with lapatinib in the NeoAdjuvant Lapatinib and/or Trastuzumab Treatment Optimization (NeoALTTO) trial than from trastuzumab with pertuzumab in the Neoadjuvant Study of Pertuzumab and Herceptin in an Early Regimen Evaluation (NeoSphere) trial, relative to single anti-HER2 agent (8, 9). Therefore, optimizing the strategy of dual HER2 blockade might make it possible to promote the efficacy in HER2-positive breast cancer.

Pyrotinib is a small molecule, irreversible, pan-ErbB receptor tyrosine kinase inhibitor (TKI). Its potent antitumor activity has been well demonstrated in HER2-positive metastatic breast cancer (11–13). However, the use of pyrotinib for locally advanced breast cancer in the neoadjuvant setting is still far from being fully elucidated. Xuhong and colleagues demonstrated a pCR (ypT0/is ypN0) rate of 73.7% for 19 patients with stage I to III HER2-positive breast cancer administered neoadjuvant pyrotinib orally once per day in combination with four cycles of epirubicin and cyclophosphamide, once every 3 weeks, followed by four cycles of docetaxel and trastuzumab, once every 3 weeks (14). Despite the promising results, dual HER2 blockade was merely used in the second half of the NAT, which may somewhat compromise the efficacy. As a result, we aim to investigate whether dual HER2 blockade with pyrotinib and trastuzumab, when used as early as possible, works well with anthracycline-free NAC for HER2-positive breast cancer. Furthermore, the addition of pyrotinib to chemotherapy and trastuzumab had not been compared with trastuzumab plus the same chemotherapy regimen in the neoadjuvant setting.

On these premises, the NeoAdjuvant Trastuzumab/Pyrotinib (NeoATP plus weekly paclitaxel/cisplatin in patients with HER2-positive breast cancer) trial was conducted in patients with HER2-positive breast cancer to identify the role of neoadjuvant pyrotinib with trastuzumab and anthracycline-free NAC, which might offer implications for future practice.

The NeoATP trial is a single-arm, single-center, open-label, phase II study in women with HER2-positive locally advanced, inflammatory or early breast cancer. This study is registered with ClinicalTrials.gov (NCT04126525).

The protocol and all amendments were approved by the Institutional Review Board of Renji Hospital, School of Medicine, Shanghai Jiaotong University (Shanghai, China). The study was performed in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines. All patients were required to sign the written, informed consents.

All the eligible patients were female aged between 18 and 70 years with histologically confirmed stage IIA to IIIC primary invasive breast cancer according to the anatomic staging system of the 8th Edition American Joint Committee on Cancer Breast Cancer Staging Manual. Tumors had to be HER2 positive, defined as 3+ for IHC or amplified for FISH (15). Ovarian function suppression was also permitted during NAT. Other main inclusion criteria were at least one measurable disease according to RECIST version 1.1 and Eastern Cooperative Oncology Group (ECOG) performance status of 0 or 1. Baseline laboratory tests were required to evaluate whether a patient was eligible with adequate organ function, including leukocyte count, neutrophil count, platelet count, hemoglobin, total bilirubin, alanine aminotransferase (ALT), aspartate aminotransferase (AST), blood urea nitrogen, serum creatinine, creatinine clearance (Cockcroft-Gault Equation), left ventricular ejection fraction (LVEF) as measured by echocardiography, and Fridericia-corrected QT interval (QTcF).

Key exclusion criteria included metastatic disease (stage IV), synchronous bilateral invasive breast cancer, previous treatment with other HER2-targeted agents except trastuzumab, concurrent use of other anticancer therapy except paclitaxel, disease significantly affecting gastrointestinal function, evidence of sensory or motor neuropathy, neurologic or mental disorder, other malignancies (except squamous or basal cell carcinoma of skin, or cervical carcinoma in situ), pregnancy, lactation, and refusal to use contraception.

Histologic grade, tumor type, HER2 status, estrogen receptor (ER) status, progesterone receptor (PR) status, Ki67 index, and PIK3CA status were determined by using core biopsy tissues from the breast at baseline. Hormone receptor (HR)-positive tumors were defined as ER or PR present in no less than 1% of tumor cells by IHC. ER and/or PR present in ≥10% or <10% of tumor cells by IHC were classified into HR high level or HR low level, respectively (16). PIK3CA mutation was assessed using the amplification refractory mutation system (AmoyDx).

Tumor-infiltrating lymphocytes (TIL) were evaluated by a pathologist blinded to the study outcomes, for the stromal compartment on hematoxylin and eosin–stained sections of core biopsy tissues formalin-fixed and paraffin-embedded. The percentage of stromal TILs (sTIL) was calculated in accordance with the recommendations by an International TILs Working Group 2014 (17). Tumors were dichotomized into ≥60% and <60% of sTILs (17, 18).

Residual cancer burden (RCB) index was determined from bidimensional diameters (the dimensions of the largest if multiple tumors) of the primary tumor bed in the resection specimen (d1 and d2), the proportion of primary tumor area containing invasive carcinoma (f_inv), the number of axillary lymph nodes containing metastatic carcinoma (LN), and the diameter of the largest metastasis in an axillary lymph node (d_met). Bidimensional measurements of the primary tumor bed (d_prim) were calculated in millimeters as |$\surd d1d2$|⁠. The parameter f_inv within the cross section of the primary tumor bed was derived from the overall percentage of carcinoma (%CA) and the percentage of in situ carcinoma (%CIS): f_inv = (1 – (%CIS/100)) × (%CA/100). RCB index was estimated as 1.4(f_invd_prim)^0.17 + [4(1 – 0.75^LN)d_met]^0.17 using the RCB Calculator established by MD Anderson Cancer Center (https://www.mdanderson.org/for-physicians/clinical-tools-resources/clinical-calculators/residual-cancer-burden.html). Patients with residual disease (RD) were assigned into three classes: RCB-I (minimal RD), RCB-II (moderate RD), and RCB-III (extensive RD). The cut-off points 0, 1.36, and 3.28 defined subgroups of RCB-0 to RCB-III with increasingly poor prognosis (19).

The histologic response to study treatment was also assessed by the Miller and Payne (MP) grading system with a five-point scale, which mainly takes into consideration the principal manifestation of a reduction in tumor cellularity (20).

The absolute counts of peripheral lymphocyte subsets were determined by the method of flow cytometry using the BD Multitest IMK Kit (a four-color direct immunofluorescence reagent kit) at the Department of Clinical Laboratory, Renji Hospital, School of Medicine, Shanghai Jiaotong University. The mature human lymphocyte subsets in erythrocyte-lysed whole blood were categorized into T lymphocytes (CD3+), B lymphocytes (CD19+), helper/inducer T lymphocytes (CD3+ CD4+), suppressor/cytotoxic T lymphocytes (CD3+ CD8+), and natural killer (NK) lymphocytes (CD3− CD16+ and/or CD56+).

For all patients, paclitaxel at 80 mg/m² was intravenously administered weekly starting on days 1, 8, 15, and 22, and cisplatin at 25 mg/m² on days 1, 8, and 15, every 28 days for four cycles. Trastuzumab was given every week at a loading dose of 4 mg/kg, followed by a maintenance dose of 2 mg/kg thereafter. Pyrotinib was orally given at 400 mg once per day. All patients received study treatment until disease progression, unacceptable toxicity, consent withdrawal, investigator's decision, or study completion. To manage adverse events (AE), dose interruptions or dose reductions were permitted. Dose reescalation was not allowed. After completing NAT, patients underwent surgery and adjuvant therapy as per local guidelines or institutional standards at that time.

We assessed hematologic and biochemical parameters on a weekly basis. ECOG status, vital signs, 12-lead electrocardiograms, and peripheral lymphocyte subsets were examined at baseline, at each cycle and before surgery. Tumor responses for target lesion and regional lymph nodes were evaluated by ultrasound and MRI every two cycles and before surgery as well as by palpation at each cycle and before surgery. Echocardiography was done at baseline and every 3 months thereafter. AEs were monitored until 28 days after the last dose of study treatment and graded according to the NCI Common Terminology Criteria for Adverse Events, version 5.0.

Health-related quality of life (HRQoL) was assessed at baseline, after two cycles of study treatment, and after four cycles of study treatment or before surgery, using the European Organisation for Research and Treatment of Cancer (EORTC) Quality of Life Questionnaire core 30 items (QLQ-C30; ref. 21). The EORTC QLQ-C30 is a cancer-specific HRQoL questionnaire composed of five functional scales (physical, role, cognitive, social, and emotional functioning), three symptom scales (fatigue, pain, and nausea/vomiting), a global health status (GHS)/QoL scale, and six single items (dyspnea, insomnia, loss of appetite, constipation, diarrhea, and perceived financial impact).

Tumor and blood samples for banking were collected at biopsy and at surgery for future identification of potential biomarkers predictive of response or prognosis.

The primary endpoint was the rate of pCR, defined as the absence of residual viable tumor (invasive or noninvasive) cells under the microscopic examination of the breast and the axillary lymph nodes at surgery (ypT0 ypN0). Secondary endpoints included the rate of locoregional pCR, defined as no invasive cancer in the breast and no pathologic involvement of axillary lymph nodes (ypT0/is ypN0); the proportion of RCB-0 and RCB-I (RCB-0/I); the proportion of Grade 4 and 5 (Grade 4/5) in the MP system; disease-free survival (DFS), defined as the time from the date of surgery to the date of local and/or regional recurrence, contralateral breast cancer, distant metastasis, or death as the first event; overall survival (OS), defined as the time from the date of surgery to the date of death from any cause; biomarker analysis; and safety.

With a planned sample size of 47 patients, the study had 80% power to detect an improvement in the pCR rate from 0.50 to 0.70, with an α level of 0.05 by using a two-sided binomial test for superiority. Assuming 10% loss to follow-up, a total accrual of 52 patients was required. The null hypothesis of 0.50 was based on the presumed activity of neoadjuvant paclitaxel/cisplatin chemotherapy combined with trastuzumab in this population according to the previous study (22).

Patient demographics, clinicopathologic characteristics, and safety results were summarized descriptively. The full analysis set (FAS) included all patients who received at least one dose of the study treatment according to the intention-to-treat principle, and the per-protocol set included a subset of the patients in the FAS who were compliant with the requirements of the study protocol. Any patient without a recorded pathology report after NAT was regarded as a nonresponder in both breast and axilla irrespective of endpoint definition. Safety was evaluated in all patients who received any amount of the study treatment.

The analysis of tumor response endpoints is presented as the proportion and two-sided 95% confidence intervals (CI). Subgroups were analyzed by age (≤50 vs. >50), baseline clinical stage (IIA-IIIA vs. IIIB-IIIC), HR status (low level vs. high level), Ki67 index (≤40% vs. >40%), HER2/chromosome enumeration probe 17 (CEP17) ratio (<4 vs. ≥4), average HER2 copy number (<14 vs. ≥14), HER2 IHC status (2+ vs. 3+), and sTILs (<60% vs. ≥60%).

Patients were evaluable for the HRQoL analysis if they completed baseline and at least one subsequent questionnaire. For each EORTC QLQ-C30 scale or item, a linear transformation was used to get the standard score between 0 and 100. For GHS/QoL or functional scales, higher scores suggest a better status of health and functioning, while higher scores indicate worse symptoms for symptom scales. The change from baseline to each timepoint was calculated as the timepoint score minus the baseline score. The changes are summarized as mean ± SD. Clinically meaningful difference in HRQoL was defined as a 10-point decrease or increase from baseline EORTC QLQ-C30 scores, and accordingly patients were categorized into improved, stable, or deteriorated HRQoL.

All the analyses were performed with STATA Statistics SE 14 (Stata Corp LP, College Station, TX) and R statistical software (version 4.1.0). All tests were two-tailed and P < 0.05 was considered statistically significant.

The data generated in this study are available within the article and its Supplementary Data files, or upon reasonable written request from the corresponding author.

A total of 53 patients were eligible from July 2, 2019 to July 27, 2021 and were followed up through September 30, 2021. The median age was 47 years (range, 26 to 66 years) and most patients were premenopausal (n = 35; 66.04%). The majority of patients were clinical stage III (73.58%) and node positive (88.68%) at baseline. More than one third of tumors were cT4 (39.62%) and cN2–3 (41.51%). Patients with high level HR and Ki67 index >40% accounted for 49.06% and 41.51%, respectively (Table 1). The representativeness of study participants is shown in Supplementary Table S1.

Fifty-two of 53 patients completed study treatment without progression and received surgery after study treatment (Fig. 1). Three women combined breast-conserving surgery with axillary dissection, 1 had breast-conserving surgery with sentinel lymph node biopsy, 1 underwent nipple-sparing modified radical mastectomy and immediate reconstruction with a latissimus dorsi flap, and the other 47 patients received modified radical mastectomy. The median number of cycles was 3 (range, 1.25 to 4.5). NAC was delayed in 21 patients mainly due to ALT or AST increase (n = 11), neutropenia (n = 3), and creatinine increase (n = 2). One patient discontinued study treatment because of hospitalization (thromboembolic event).

Overall, 37 of 53 patients (69.81%) achieved pCR. The locoregional pCR rate was 73.58% (n = 39). Only 5 patients (9.43%) had pathologically positive lymph nodes at surgery. The proportion of RCB-0/I was 88.68%. The proportion of Grade 4/5 in the MP system was 92.45% (Supplementary Table S2).

Subgroup analyses revealed that pCR (Supplementary Table S3) was more frequent in cT1/2 (P = 0.008), cN0/1 (P = 0.041), and IIA-IIIA (P = 0.004) tumors. Patients were more easily able to achieve pCR with negative or low level ER (P = 0.008 and P = 0.004), negative PR (P = 0.027), negative HR (P = 0.041; Fig.2), HER2 IHC 3+ (P = 0.016), HER2/CEP17 ratio ≥4 (Supplementary Fig. S1A), and average HER2 copy number ≥14 signals per cell (P < 0.001). No statistically significant difference was observed in terms of pCR between subgroups by PIK3CA status (P = 0.958; Supplementary Fig. S1B), Ki67 index (P = 0.409; Supplementary Fig. S1C), and sTILs (P = 0.729; Supplementary Fig. S1D). In addition, higher locoregional pCR rates (Supplementary Table S4) were also observed in patients with ER negative or low level (P = 0.034 and P = 0.005), HER2 IHC 3+ (P = 0.035), HER2/CEP17 ratio ≥4 (P = 0.004; Supplementary Fig. S2A), average HER2 copy number ≥14 signals per cell (P = 0.001), and IIA-IIIA (P = 0.004) and cT1/2 (P = 0.020) tumors. There was no significant difference in locoregional pCR between subgroups by HR status (P = 0.105; Fig. 2), PIK3CA status (P = 0.753; Supplementary Fig. S2B), Ki67 index (P = 0.452; Supplementary Fig. S2C), and sTILs (P = 0.333; Supplementary Fig. S2D).

As continuous variables, HER2/CEP17 ratio (Spearman P = 0.009; Spearman P = 0.017; Spearman P = 0.014; Supplementary Fig. S3A) and average HER2 copy number (Spearman P = 0.008; Spearman P = 0.005; Spearman P = 0.003; Supplementary Fig. S3B) were positively correlated with pCR, locoregional pCR, and MP grade, accompanied by a positive correlation of sTILs with locoregional pCR (Spearman P = 0.016) and MP grade (Spearman P = 0.013; Supplementary Fig. S3C), while there was no association of Ki67 index with pCR, locoregional pCR, or MP grade (Supplementary Fig. S3D). On the other hand, HER2/CEP17 ratio (Spearman P = 0.011; Supplementary Fig. S4A), average HER2 copy number (Spearman P = 0.003; Supplementary Fig. S4B), and sTILs (Spearman P = 0.011; Supplementary Fig. S4C) were inversely correlated with RCB. However, no association of Ki67 index was observed in terms of RCB (Supplementary Fig. S4D).

When it came to the absolute counts of peripheral lymphocyte subsets, we found a statistically significant difference in the change of average T lymphocytes (Wilcoxon P = 0.013) and helper/inducer T lymphocytes (Wilcoxon P = 0.039) from baseline to 2 cycles of NAT between pCR and non-pCR patients, whereas similar significance failed to be detected in the change of the other subsets from baseline to 2 cycles of NAT as well as in the change of any subset from baseline to the end of NAT or from 2 cycles of NAT to the end of NAT between pCR and non-pCR (Supplementary Fig. S5).

All 53 patients were included in the safety set. Treatment-emergent AEs occurring in 10% or more of the 53 patients are listed in Table 2. In general, the most frequently reported grade 3 to 4 AEs were diarrhea (45.28%), leukopenia (39.62%), and neutropenia (32.08%). One patient had a serious AE (SAE), requiring hospitalization due to a thromboembolic event, considered not related to study treatment. No deaths occurred, and none of the patients experienced febrile neutropenia and grade 4 diarrhea as well as grade 3 or higher cardiac disorders. No LVEF <50% or >10 points drop from baseline to before surgery was reported (Fig. 3; Supplementary Fig. S6).

Fifty-two of 53 patients were included in the HRQoL analysis. Compared with baseline, EORTC QLQ-C30 scores worsened in terms of appetite loss (14.10±41.93) and diarrhea (18.59±44.48) after 2 cycles of NAT. At the end of NAT, the corresponding score for appetite loss still showed increased severity from baseline (11.54±36.09), while the change for diarrhea was not considered clinically meaningful (4.49±37.36). No differences were seen for the other scales or items from baseline to both 2 cycles and the end of NAT (Fig. 4; Supplementary Table S5).

This study substantiates that pyrotinib plus trastuzumab and chemotherapy can be considered as an efficacious and safe option for HER2-positive locally advanced breast cancer in the neoadjuvant setting. It is also, to the best of our knowledge, the first and largest clinical trial to provide evidence that the addition of pyrotinib to trastuzumab yielded stable LVEF and favorable HRQoL on the basis of the anthracycline-free NAC regimen.

Our study first verified that the efficacy of neoadjuvant pyrotinib was independent of PIK3CA status. In the NeoALTTO trial, patients with mutant PIK3CA obtained a significantly decreased pCR rate of 28.6% after neoadjuvant lapatinib plus trastuzumab, whereas the corresponding rate was 53.1% for those with wild-type PIK3CA (P = 0.012; ref. 23). Similar results were also observed in manifold studies (24–27), implying that PIK3CA mutation might serve as a predictor for resistance to trastuzumab, lapatinib, and pertuzumab. In the phase I study of pyrotinib for HER2-positive metastatic breast cancer (MBC), the biomarker analysis suggested that PIK3CA and TP53 mutations in archival primary tumor samples were not associated with objective responses (28), which, to some extent, provides support for our findings. Accordingly, pyrotinib might overcome resistance induced by PIK3CA mutation, requiring further investigation of the underlying mechanisms.

We further found that HER2/CEP17 ratio and average HER2 copy number may help to early predict the likelihood of pCR to neoadjuvant pyrotinib added to trastuzumab and chemotherapy for the first time. Our previous study revealed the predictive value of HER2/CEP17 ratio and HER2 copy number in pCR for patients with HER2-positive breast cancer concurrently receiving NAC and trastuzumab (29), which was supported by numerous studies (30–33). However, little evidence was available for dual HER2 blockade from the prospective clinical trials. Recently, Venet and colleagues confirmed the positive association of HER2/CEP17 ratio and HER2 copy number with pCR in the NeoALTTO cohort (34), partly in keeping with our findings because its specific anti-HER2 regimen was not presented in detail.

Of note, our study also evaluated both local and systemic immune response with the addition of pyrotinib to trastuzumab and NAC for the first time. We found that higher sTILs were associated with superior efficacy in the NeoATP cohort using the correlation analysis. In agreement with our results, large numbers of other studies revealed that more TILs were correlated with higher response and/or better prognosis in the neoadjuvant setting for HER2-positive breast cancer regardless of anti-HER2 regimen (35–40). Besides, in the NeoATP cohort, the change in absolute counts of average T lymphocytes or helper/inducer T lymphocytes during the study treatment conferred response to neoadjuvant pyrotinib combined with trastuzumab plus NAC. These findings suggest that the status or change of the immune cells in either tumor microenvironment or circulating blood might contribute to the sensitivity to various anti-HER2 regimens, which hints at a demand to identify the potential candidates for dual HER2 blockade with promising immune biomarkers.

It has not been elucidated how to optimize the use of neoadjuvant pyrotinib for lack of data. Looking back on limited information of previous literature, we may find some clues. Xuhong and colleagues assigned 19 patients with HER2-positive breast cancer to receive neoadjuvant pyrotinib in combination with four cycles of epirubicin and cyclophosphamide, followed by four cycles of docetaxel and trastuzumab (14). However, the duration of NAT lasted for at least 21 weeks before surgery, which seems too long for patients to complete. In addition, dual HER2 blockade was not given throughout the entire process of NAT, probably in consideration of cardiac toxicity. As a result, we designed the NeoATP trial with an anthracycline-free regimen for a shorter period and dual anti-HER2 agents for the whole course. When compared with the report by Xuhong and colleagues, there were more stage III (73.58% vs. 10%) tumors in the NeoATP trial, with strikingly similar pCR (ypT0/is ypN0) rates (73.58% vs. 73.7%). Therefore, patients with HER2-positive breast cancer might achieve better response by receiving both pyrotinib and trastuzumab from the very beginning of NAC, even with a shorter course of the anthracycline-free regimen.

Our study showed that diarrhea was the most common AE of grade 3 or higher in the NeoATP cohort, which was consistent with the report by Xuhong and colleagues in the neoadjuvant setting (14). Previous reports (8–10, 41) showed that the addition of TKI to NAT might cause more diarrhea of higher grade than the addition of HER2 antibody. Nonetheless, as to the NeoATP cohort, diarrhea was generally reversible with antidiarrhea treatment, dose interruption, or dose reduction and did not lead to treatment discontinuation despite higher rate of all-grade or grade 3 diarrhea, consistent with published clinical trials investigating the role of pyrotinib in MBC (11–13) and the neoadjuvant use of other TKIs (8, 41). The NeoATP trial also revealed that grade 3 diarrhea mainly occurred during the first cycle of study treatment and the incidence trend of grade 3 diarrhea did not increase thereafter (data not shown), similar to the previous data on pyrotinib for MBC (11–13). Besides, none of the patients discontinued study treatment due to diarrhea in the NeoATP cohort. In view of HRQoL analysis, of interest, the score worsened in terms of diarrhea after 2 cycles of NAT compared with baseline, while the change for diarrhea was not considered clinically meaningful at the end of NAT. These findings provide some evidence of adaptation to pyrotinib regarding diarrhea during the trial. As to cardiotoxicity, no significant decrease of LVEF was observed in the NeoATP cohort. Therefore, dual HER2 blockade with pyrotinib and trastuzumab was well tolerated on the whole.

Several limitations exist in our study. Firstly, this analysis was not based on a randomized, controlled clinical trial to evaluate the addition of neoadjuvant pyrotinib. Secondly, the survival analyses of DFS and OS were immature and unreported in this primary analysis of the NeoATP trial. Nonetheless, follow-up is ongoing and the prognostic outcomes are awaited in the near future. Last but not least, all the enrolled patients were Chinese. Additional trials are warranted to validate the racial discrepancies in efficacy and toxicity.

In conclusion, the addition of pyrotinib to trastuzumab plus anthracycline-free NAC regimen is an alternative and promising dual HER2 blockade regimen to the current treatment landscape for patients with HER2-positive breast cancer in the neoadjuvant setting. Exploratory analysis will be performed on biospecimens collected at prespecified timepoints.

No disclosures were reported.

W. Yin: Conceptualization, data curation, formal analysis, supervision, funding acquisition, validation, methodology, writing–original draft, project administration, writing–review and editing. Y. Wang: Conceptualization, data curation, formal analysis, supervision, funding acquisition, validation, visualization, writing–original draft, writing–review and editing. Z. Wu: Data curation, funding acquisition, validation, writing–original draft, writing–review and editing. Y. Ye: Validation, writing–review and editing. L. Zhou: Validation, writing–review and editing. S. Xu: Validation, writing–review and editing. Y. Lin: Validation, writing–review and editing. Y. Du: Validation, writing–review and editing. T. Yan: Validation, writing–review and editing. F. Yang: Validation, writing–review and editing. J. Zhang: Data curation, validation, writing–review and editing. Q. Liu: Data curation, validation, writing–review and editing. J. Lu: Conceptualization, formal analysis, supervision, funding acquisition, validation, project administration, writing–review and editing.

This study was funded by National Natural Science Foundation of China (No. 82103695 and 82173115), Clinical Research Plan of Shanghai Hospital Development Center (No. SHDC2020CR3003A), Science and Technology Commission of Shanghai Municipality (No. 20DZ2201600), Beijing Foundation of Medicine Award (No. YXJL-2020-0941-0737), Shanghai Municipal Key Clinical Specialty, Shanghai ‘Rising Stars of Medical Talent’ Youth Development Program for Outstanding Youth Medical Talents (No. 2018-16), Shanghai Rising-Star Program (No. 22QC1400200), Multidisciplinary Cross Research Foundation of Shanghai Jiao Tong University (No. YG2019QNA28), Clinical Research Innovation Nurturing Fund of Renji Hospital and United Imaging (No. 2021RJLY-002), Nurturing Fund of Renji Hospital (No. PYIII20-09, PY2018-III-15 and PY2018-IIC-01), and Jiangsu Hengrui Pharmaceuticals Co., Ltd.

We thank all the patients and their family members who participated in this study.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.