2.1 Study design and patients

The detailed design of AENEAS has been published previously [13]. AENEAS was a multicenter, double-blinded, randomized phase III trial conducted at 53 study sites in mainland China. The trial was conducted in accordance with the protocol, applicable local regulations, and the Declaration of Helsinki and the International Conference on Harmonisation Guidelines for Good Clinical Practices principles. The ethics committee of Shanghai Chest Hospital (ID: LS1840) and each participating institution's research ethics board approved the study protocol. All patients provided written informed consent before the initiation of any study-related procedure. The trial was registered at ClinicalTrials.gov with the identifier NCT03849768.

Patients with previously untreated metastatic or locally advanced NSCLC with EGFR sensitizing mutations were enrolled and randomly assigned in a 1:1 ratio to orally receive either aumolertinib (Jiangsu Hansoh Pharmaceutical Group Co. Ltd., Lianyungang, Jiangsu, China) or gefitinib (AstraZeneca Pharmaceutical Co. Ltd., Cambridge, London, UK) in a double-blinded fashion. Patients were stratified by EGFR mutation status (Ex19del versus L858R, confirmed by a central laboratory [Teddy Clinical Research Laboratory Co. Ltd., Shanghai, China] using the Cobas EGFR Mutation Test [version 2; Roche Molecular Systems Inc, Pleasanton, CA, USA], as detected in tumor tissue samples or blood samples) and baseline CNS metastasis status. At baseline, patients were required to have at least one measurable lesion, defined as ≥ 10 mm, and baseline assessment was performed using Response Evaluation Criteria in Solid Tumors version 1.1 (RECIST 1.1). Baseline CNS imaging was mandatory, and for patients with identified CNS metastases, follow-up CNS imaging was performed with the same methodology as baseline and then assessed by BICR. Patients with asymptomatic, stable CNS metastases that did not require steroids for at least 2 weeks before starting the study drug were included. The presence of baseline CNS metastases defined the CNS Full Analysis Set (cFAS), and the CNS Evaluable for Response (cEFR) was the subset of patients with at least one measurable CNS lesion of at least 10 mm in the longest diameter.

2.2 Procedures

An interactive web response system (software developed by Shanghai Shanhu Health Technology Co. Ltd., Shanghai, China) randomly assigned eligible patients 1:1 to receive either 110 mg aumolertinib or 250 mg gefitinib, administered orally once daily. Treatment with the study drug was continued until disease progression, withdrawal of consent, the development of unacceptable adverse effects, or the fulfillment of other discontinuation criteria. Treatment with the study drug beyond disease progression was permitted if the patients continued to derive clinical benefits as assessed by the treating investigator. Upon disease progression, patients in the gefitinib arm who acquired an EGFR T790M mutation were eligible to crossover to aumolertinib treatment. For patients with CNS metastases, follow-up imaging of the same modality as the baseline CNS imaging was performed every 6 weeks (±7 days) from the start of treatment until 15 months, after which imaging was performed at 12-week (±7 days) intervals. CNS response was assessed by neuroradiological-BICR provided a baseline CNS target lesion at least 10 mm in the longest diameter could be identified.

2.3 Outcomes

The data cut-off date of this study was August 06, 2021. The primary objective for this subgroup analysis was to estimate CNS PFS by neuroradiological-BICR (Fantastic Bioimaging Co. Ltd., Shanghai, China), with CNS PFS defined as the time from random assignment until the date of objective CNS progression or death resulting from any cause. CNS endpoints included CNS objective response rate (ORR), duration of response (DoR) and disease control rate (DCR). Responses of CNS lesions require confirmation by neuroradiological-BICR per RECIST 1.1 guidance.

CNS ORR was defined as the proportion of patients with a best overall CNS response of complete response (CR) or partial response (PR) relative to the total number of patients with baseline CNS disease. CNS DCR was defined as the percentage of patients who had a best overall CNS response of CR, PR, or stable disease (SD) ≥ 5 weeks before any progressive disease (PD) event. CNS DoR was defined as the time from documentation of CNS response (CR or PR) to intracranial disease progression or death.

Adverse events (AEs) were monitored continuously from informed consent to 28 days after the last dose of randomized treatment. The severity of AEs was graded using the National Cancer Institute-Common Terminology Criteria for Adverse Events version 4.03.

2.4 Statistical methods

CNS PFS was summarized using the Kaplan-Meier method. The Kaplan-Meier estimates of the 25th, 50th (median), and 75th percentiles, along with their corresponding two-sided 95% CI for PFS, were presented by the randomized treatment arm. A log-rank test that was stratified by EGFR mutation type was used to compare the PFS distributions of the two arms, thus determining whether there was a statistical difference between the arms. The stratified Cox proportional hazards model, as adjusted by EGFR mutation type, was used to estimate the PFS HR, along with the two-sided 95% CI of the two treatment arms. CNS DoR was analyzed using the same methods as the PFS. CNS ORR was analyzed based on the confirmed responses using the Cochran-Mantel-Haenszel test stratified by EGFR mutation type. The exact 95% CI for ORR and DCR was calculated using the Clopper-Pearson method. The incidence of the first event being CNS progression, with no prior non-CNS progression or death, was estimated through competing risk analysis using the Cause-Specific Hazard model. Safety analyses were performed on the safety analysis set, which included all randomized patients with CNS metastases who received the study drug at least once. All the analysis was performed using the Statistical Analysis System (SAS v9.4; SAS Institute, Cary, North Carolina, US).

3.1 Patient disposition and baseline characteristics

Of the 429 patients enrolled and randomized in the AENEAS trial, 106 patients (aumolertinib arm, n = 51; gefitinib arm, n = 55) were found to have CNS metastases at baseline by neuroradiological-BICR. These 106 patients were included in cFAS. Among them, 60 patients (aumolertinib arm, n = 28; gefitinib arm, n = 32) had at least one measurable CNS lesion. These 60 patients were included in cEFR (Figure 1). Demographics were generally balanced between treatment arms in cFAS and cEFR. The aumolertinib arm trended towards a higher Eastern Cooperative Oncology Group (ECOG) performance score. In cFAS, 3 patients (5.9%) in the aumolertinib arm and 4 patients (7.3%) in the gefitinib arm received prior brain radiotherapy (Table 1).

3.2 CNS PFS

Median follow-up time in the overall AENEAS study population was 26.2 months for the aumolertinib arm and 26.3 months for the gefitinib arm, respectively. The median follow-up time for CNS PFS was 20.9 months in the aumolertinib arm and 13.8 months in the gefitinib arm. A total of 53 events (aumolertinib, n = 20; gefitinib, n = 33) were observed among the 106 patients in the cFAS, while 31 events (aumolertinib, n = 10; gefitinib, n = 21) were observed among the 60 patients in the cEFR (Table 2). In the cFAS, median CNS PFS was 29.0 months (95% CI, 12.3-not arrived [NA]) for aumolertinib versus 8.3 months (95% CI, 6.9-9.7) for gefitinib (HR = 0.31; 95% CI, 0.17-0.56; P < 0.001). Similar results were obtained in the cEFR population (HR = 0.26; 95% CI, 0.11-0.57; P < 0.001) (Table 2, Figure 2A-B). In the cFAS, the 12-month CNS PFS rate was 72.5% for aumolertinib, compared to 30.4% for gefitinib. In the cEFR, the 12-month CNS PFS rate was 73.5% versus 21.6% between treatment arms (Table 2).

3.3 Risk of CNS progression

Competing risk analysis showed that the estimated probability of CNS progression without prior non-CNS progression or death was consistently lower with aumolertinib than with gefitinib in patients with and without CNS metastases at baseline (Figure 2C-D). Twelve-month estimated probability of CNS progression was 21.6% (95% CI, 11.4-33.8) with aumolertinib and 39.7% (95% CI, 26.4-52.7) with gefitinib in patients with baseline CNS metastases. Twelve-month estimated probability of CNS progression was 1.2% (95% CI, 0.2-4.1) with aumolertinib and 5.8% (95% CI, 2.9-10.3) with gefitinib in patients without baseline CNS metastases.

3.4 CNS response

Aumolertinib demonstrated a higher CNS CR rate compared with gefitinib in both the cEFR (14.3% vs. 0%) and cFAS (23.5% vs. 5.5%). In the cEFR, CNS ORR was 85.7% (95% CI, 67.3-96.0) with aumolertinib and 75.0% (95% CI, 56.6-88.5) with gefitinib (OR = 1.95; 95% CI, 0.53-7.25; P = 0.312). In the cFAS, CNS ORR was 62.7% (95% CI, 48.1-75.9) with aumolertinib and 49.1% (95% CI, 35.4-62.9) with gefitinib (OR = 1.76; 95% CI, 0.81-3.81; P = 0.149) (Table 2). In the cEFR, the median time to response was similar in patients treated with aumolertinib and gefitinib (6.1 vs. 6.0 weeks) (Supplementary Table S1).

In the cEFR, the median confirmed CNS DoR was 27.7 months in the aumolertinib arm versus 6.9 months in the gefitinib arm (HR = 0.15; 95% CI, 0.05-0.41; P < 0.001). Similar results could be found in the cFAS (Figure 2E-F). In the cEFR, the estimated percentage remaining in CNS response at 12 months was 77.2% for aumolertinib, compared to 21.4% for gefitinib (Table 2).

The median best percentage change from baseline in CNS target lesion size in the aumolertinib arm was −56.5% (range, −100.0% to 32.0%) versus −50.5% (range, −74.1% to −4.5%) in gefitinib arm (Figure 3, Supplementary Table S2). The proportion of patients with ≥ 30% reduction, ≥ 50% reduction, and ≥ 75% reduction in target lesion was 85.7%, 71.4%, and 28.6% in the aumolertinib arm respectively, compared to 87.5%, 56.3%, and 0% in the gefitinib arm, respectively (Supplementary Table S2). More patients in the aumolertinib arm achieved a consistent decreasing trend in CNS target lesions compared with the gefitinib arm (Figure 4).

3.5 Concordance between CNS response and systemic response

By grouping the patients according to the best overall systemic response, we found that most patients with a best overall systemic response of CR and PR showed a decreasing trend in CNS target lesions. A higher proportion of patients with a best overall systemic response of SD and PD in the aumolertinib arm showed an increasing trend in CNS target lesions (Supplementary Figure S1). Changes in CNS lesions were well correlated with systemic responses. In the cEFR, the concordance rate between CNS ORR and systemic ORR in the aumolertinib arm was 89.3%. Among the 28 patients in the aumolertinib arm, 22 (78.6%) had both CNS and systemic response, and 3 (10.7%) had neither CNS nor systemic response. The concordance rate in the gefitinib arm was 78.1%. Among the 32 patients in the gefitinib arm, 20 (62.5%) had both CNS and systemic response, and 5 (15.6%) had neither CNS nor systemic response (Table 3). Regarding systemic efficacy, in the cFAS, median systemic PFS was 15.3 months (95% CI, 10.8-20.9) for aumolertinib and 8.1 months (95% CI, 5.5-8.3) for gefitinib (HR = 0.35; 95% CI, 0.22-0.56; P < 0.001). Median overall survival was NA (95% CI, 23.1-NA) for aumolertinib and 22.8 months (95% CI, 20.1-NA) for gefitinib (HR = 0.73; 95% CI, 0.41-1.31; P = 0.288) in cFAS (Supplementary Figure S2).

3.6 Longitudinal summarization of CNS lesion status

Summarization of CNS lesion status at baseline and data cut-off in the overall study population showed that there were fewer patients with CNS lesions at data cut-off in the aumolertinib arm compared with baseline (Supplementary Figure S3). In patients with baseline CNS metastases, 15.7% in the aumolertinib arm developed new CNS lesions, while 32.7% in the gefitinib arm developed new CNS lesions. The median time to develop new CNS lesions in the aumolertinib and gefitinib arms was 6.2 and 8.2 months, respectively. In patients without baseline CNS metastases, 2.5% in the aumolertinib arm developed new CNS lesions, while 14.4% in the gefitinib arm developed new CNS lesions. The median time to develop new CNS lesions in the aumolertinib and gefitinib arms was 9.0 and 9.7 months, respectively (Supplementary Table S3).

3.7 Safety

The safety profile of aumolertinib in this analysis was consistent with the previous results of the overall trial population as no additional or novel safety signals were observed (Supplementary Tables S4-S5).