Aumolertinib with carboplatin-pemetrexed versus aumolertinib for nonsmall cell lung cancer with EGFR and concomitant tumor suppressor genes (ACROSS2): An open-label, multicenter, randomized phase 3 study.

BACKGROUND
Epidermal growth factor receptor (EGFR) mutations are among the most common oncogenic drivers in lung cancer, occurring in 40%–60% of Asian patients with lung adenocarcinoma.
1
Third‐generation EGFR tyrosine kinase inhibitors (EGFR‐TKIs) are established as the standard first‐line therapy for EGFR‐mutated nonsmall‐cell lung cancer (NSCLC), yielding a median progression‐free survival (PFS) of approximately 18–20 months,
2
,
3
,
4
yet clinical outcomes remain heterogeneous across patients. With the increasing use of next‐generation sequencing (NGS), co‐mutations are now frequently identified in EGFR‐mutated NSCLC, with reported rates as high as 92.9%.
5
These alterations include concurrent non‐EGFR driver gene mutations (such as Kirsten rat sarcoma viral oncogene homolog [KRAS], Erb‐B2 receptor tyrosine kinase 2 [ERBB2], and mesenchymal‐epithelial transition factor [MET]) and tumor suppressor gene (TSG) mutations (such as tumor protein p53 [TP53], RB transcriptional corepressor 1 [RB1], and phosphatase and tensin homolog [PTEN]).
6

TP53 mutations represent the most frequent co‐altering event, with a prevalence of 55%–65%. Alterations in other TSG mutations—such as RB1, PTEN, and AT‐rich interaction domain 1A (ARID1A)—are generally detected in fewer than 10% of patients.
7
Our previous BENEFIT study (ClinicalTrials.gov identifier NCT02282267), which evaluated first‐generation gefitinib, reported a median PFS of 9.5 months in the overall population.
8
However, stratification by genomic profile revealed substantial variation: patients with only EGFR‐sensitizing mutations achieved the longest median PFS, whereas those with TSG co‐mutations or non‐EGFR driver gene alterations experienced markedly shorter benefit (13.2 months vs. 9.3 and 4.0 months, respectively).
9
Consistent findings across multiple studies indicate that co‐occurring TSG mutations negatively influence the efficacy of EGFR‐TKIs, a trend observed across several generations of these agents.
7
,
10
,
11

The adverse impact of co‐mutations has raised the question of whether adding chemotherapy to EGFR inhibition could mitigate their detrimental effect. Evidence from the FLAURA2 study (ClinicalTrials.gov identifier NCT04035486) indicated that combining a third‐generation EGFR‐TKI with platinum–pemetrexed significantly prolonged both PFS and overall survival (OS) compared with EGFR‐TKI monotherapy in patients with unselected, EGFR‐mutated NSCLC.
12
,
13
These findings led to regulatory approval of the combination by both the US Food and Drug Administration and the National Medical Products Administration. However, chemotherapy intensification also increased the incidence of adverse events (AEs), primarily hematologic toxicities, and a proportion of patients in the osimertinib monotherapy arm experienced long‐term benefit. These observations suggest that not all patients require combination therapy and highlight the need to define the specific populations most likely to benefit from EGFR‐TKI plus chemotherapy.
Aumolertinib is an innovative, third‐generation EGFR‐TKI approved in China (2020) and the United Kingdom (2025). It irreversibly and selectively inhibits EGFR‐sensitizing mutations (such as exon 19 deletions and L858R) as well as the T790M resistance mutation. The AENEAS2 study (ClinicalTrials.gov identifier NCT04923906) demonstrated that aumolertinib combined with chemotherapy significantly prolonged PFS compared with aumolertinib alone; however, outcomes for patients with co‐mutations were not reported.
14
To address this evidence gap and to clarify the optimal first‐line strategy for EGFR‐mutated NSCLC with co‐occurring alterations, we designed two multicenter, open‐label, randomized phase 3 trials: ACROSS1 and ACROSS2 (ClinicalTrials.gov identifiers NCT04500704 and NCT04500717, respectively). Both studies evaluate the efficacy and safety of aumolertinib plus carboplatin–pemetrexed versus aumolertinib monotherapy in locally advanced or metastatic, EGFR‐mutated NSCLC. ACROSS1, which enrolled patients with concurrent non‐EGFR driver gene mutations, is ongoing and currently in the follow‐up phase for the primary end point. ACROSS2 enrolled patients with EGFR‐sensitizing mutations and concomitant TSG alterations. As the first prospective study specifically powered for this genomically defined population, ACROSS2 addresses the key clinical question of whether chemotherapy intensification can overcome the poor prognostic impact of TSG co‐mutations.

METHODS

Patients
Participants are patients aged 18–75 years with histologically or cytologically confirmed, locally advanced or metastatic NSCLC (stage IIIB/IV; American Joint Committee on Cancer AJCC Cancer Staging Manual, eighth edition) who received no prior systemic therapy for advanced disease. Patients with stage I–III disease who underwent surgery and recurred ≥6 months after completing neoadjuvant or adjuvant chemotherapy, chemoradiotherapy, or targeted therapy also are included in the study. Among the 126 randomized patients, five (4.0%) were documented as having undergone lung surgery; all of them developed subsequent recurrence or metastasis ≥6 months after surgery and had stage IIIB/IV disease at study entry.
All patients had an EGFR‐sensitizing mutation (exon 19 deletion or L858R) together with at least one predefined TSG alteration confirmed by central testing using the OncoCompass 168‐gene panel (Burning Rock Biotech; validated local NGS results were also acceptable). Additional inclusion criteria were an Eastern Cooperative Oncology Group performance status (ECOG PS) of 0–1 (stable for ≥2 weeks), at least one measurable lesion (defined according to Response Evaluation Criteria in Solid Tumors [RECIST] criteria), and not previously treated with local therapy. Written informed consent was required for all participants. The asymptomatic patients who had both treated and untreated central nervous system (CNS) metastases were eligible for this study according to the protocol. Notably, all enrolled, CNS‐positive patients had not received previous brain radiotherapy or any other localized CNS‐directed therapeutic interventions. Key exclusion criteria included prior EGFR‐TKI treatment or any systemic therapy for metastatic NSCLC, clinically significant comorbidities, other malignancies requiring active treatment, etc. Comprehensive details of the trial are provided in the protocol (see Supporting Information S2: Study Protocol, Screening of subject, p 23–25).
The TSGs covered by the central OncoCompass 168‐gene NGS panel were analyzed as an aggregate set; the TSG genes list was defined in protocol (see Supporting Information S2: Study Protocol, Appendix F, p 93). The TSGs included TP53, RB1, APC, PTEN, BRCA2, BRCA1, CHEK2, ATM, CHEK1, PALB2, and RAD51C. Variant calling was performed using the OncoCompass 168‐gene NGS panel. Somatic single nucleotide variants were called at a variant allelic frequency (VAF) ≥1% for hotspot alterations and a VAF ≥2% for non‐hotspot alterations. Similarly, for insertions and deletions (indels), hotspot variants were called at a VAF ≥0.5%, and non‐hotspot variants were called at a VAF ≥2%. TSG alterations were required to be present in the baseline tumor tissue.

Study design
This multicenter, open‐label, phase 3, randomized controlled trial was conducted in Chinese medical centers (see Supporting Information S1, List of sites, p 2). Eligible patients were randomly assigned in a 1:1 ratio to receive either aumolertinib plus carboplatin–pemetrexed or aumolertinib monotherapy to compare the efficacy and safety of the two treatment regimens.
At baseline, patients were required to have at least one measurable lesion (10 mm in greatest dimension; if lymph nodes were involved, the short axis had to be ≥15 mm) assessed according to RECIST, version 1.1. Comprehensive details of the trial are provided in the protocol (see Supporting Information S2, Study design and description, p 20–23).

Randomization and masking
The trial used a central interactive web response system with concealed allocation; randomization codes were released only after all eligibility criteria had been confirmed, thereby minimizing selection bias. Randomization was stratified by EGFR mutation subtype (exon 19 deletion vs. L858R) and the presence or absence of brain metastases at enrolment, as prespecified in the statistical analysis plan. After providing informed consent, eligible patients were randomly assigned in a 1:1 ratio to receive either combination therapy or monotherapy. Screening‐eligible participants were issued a unique randomization number on the day they received the first dose of study treatment (cycle 1, day 1).

Outcomes
Efficacy was assessed by investigators according to RECIST, version 1.1. The primary end point was PFS. Secondary end points included OS, the objective response rate (ORR), the duration of response (DoR), the disease control rate (DCR), the depth of response, and the proportion of patients who developed brain metastases. Exploratory analyses evaluated the association between the EGFR‐mutated allele fraction and treatment efficacy as well as the impact of specific TSG co‐alterations on clinical outcomes.
Safety was monitored continuously from the time of informed consent until 30 days after the last dose. All AEs, laboratory abnormalities, and clinically significant changes in vital signs, physical examination, weight, electrocardiogram, left‐ventricular ejection fraction, ECOG PS, and ophthalmologic findings were recorded. AEs were graded according to the National Cancer Institute's Common Terminology Criteria for Adverse Events, version 4.0, and documented for onset, duration, management, outcome, and relationship to study treatment. Safety end points included the incidence and severity of AEs and serious AEs, the proportion of patients who discontinued treatment because of AEs, and longitudinal changes in hematology, blood chemistry, and urinalysis.

Statistical analysis
A sample size of 120 randomized patients was estimated to provide approximately 80% power to detect a statistically significant difference in PFS between the two treatment groups at a two‐sided alpha of .05. One interim analysis of PFS was planned. Efficacy boundaries for the interim and final analyses were determined using O’Brien–Fleming boundaries with the Lan–DeMets spending function. Based on an assumed hazard ratio (HR) of 0.49 and a median PFS of 17.4 months in the aumolertinib monotherapy group, accounting for potential loss to follow‐up, the final target sample size was approximately 120 patients, and 62 PFS events (disease progression or death) were required. Sample‐size calculations were performed using PASS V15.0.5 (NCCS, LLC).
The intention‐to‐treat (ITT) population comprised all randomly assigned patients. The full analysis set (FAS) included ITT patients who received at least one dose of study treatment, and the safety set consisted of all ITT patients who received any study drug. All efficacy analyses were conducted in the ITT population. PFS was compared between treatment groups using a stratified log‐rank test. Time‐to‐event end points were summarized with the Kaplan–Meier method, and HRs with 95% confidence intervals (CIs) were estimated using a stratified Cox proportional‐hazards model stratified by EGFR mutation subtype (exon 19 deletion vs. L858R) and baseline brain metastasis status (present vs. absent). The ORR and DoR were assessed in the FAS among patients who had measurable disease at baseline and were summarized descriptively. OS and all safety end points were also summarized descriptively. Details of the statistical analysis are provided in the statistical analysis plan (see Supporting Information S3: Statistical Analysis Plan, p 9).
PFS was calculated from randomization to investigator‐assessed progression or death. Event/censoring rules and missing data imputations were prespecified in the SAP software package (version 1.0; June 11, 2025; see Supporting Information S3: Statistical Analysis Plan).

RESULTS

Baseline patient characteristics
Between October 1, 2021, and September 30, 2024, in total, 126 treatment‐naive patients with EGFR‐sensitizing mutations and concomitant TSG alterations were randomly assigned at 26 centers in China. Of 234 screened patients, 126 in the ITT set were randomly assigned: 62 to receive aumolertinib plus carboplatin–pemetrexed and 64 to receive aumolertinib monotherapy. Eight patients in the combination arm never initiated protocol therapy. All 64 monotherapy patients received study drug. The FAS and safety set comprised 118 patients (54 allocated to aumolertinib plus carboplatin–pemetrexed and 64 to aumolertinib monotherapy; Figure 1). At the data cutoff on August 8, 2025, 20 patients (37.0%) in the combination group and 16 (25.0%) in the monotherapy group remained on treatment. Most treatment discontinuations were attributed to disease progression (42.6% vs. 62.5% in the combination and monotherapy groups, respectively), whereas discontinuation because of AEs occurred in ≤4% of patients in each arm. NGS confirmed the presence of EGFR‐sensitizing mutations and predefined TSG alterations in all 126 randomized patients.
Baseline characteristics were balanced between treatment groups (Table 1). Among 126 patients, of 110 patients who had TP53‐mutated NSCLC (87.3%), 23 (18.3%) harbored co‐alterations in at least one additional TSG (co‐TSG group), whereas 87 (69.0%) had a TP53 mutation only (see Table S1).

Survival outcomes
At the data cutoff on August 8, 2025, the median follow‐up for PFS was 25.3 months, during which 63 PFS events (53.4%) were recorded. The investigator‐assessed median PFS was 19.78 months with aumolertinib plus carboplatin–pemetrexed and 16.53 months with aumolertinib monotherapy (unstratified: HR, 0.54; 95% CI, 0.32–0.92; stratified [stratification factors: EGFR mutation subtype and baseline brain metastasis status]: HR, 0.58; 95% CI, 0.34–0.97; Figure 2A). Landmark PFS rates were 78.7% versus 65.3% at 12 months, 67.2% versus 40.8% at 18 months, and 41.0% versus 29.9% at 24 months for combination therapy versus monotherapy, respectively.
Across all prespecified and exploratory subgroups, the combination demonstrated a PFS advantage similar in magnitude to that observed in the overall population (Figure 2B). All p values accompanying exploratory end points are nominal. Point estimates favored combination therapy in all subgroups (women, patients younger than 65 years, never‐smokers, those without baseline CNS metastases, patients with TP53 mutations, and individuals with an ECOG PS of 0), as well as in exploratory analyses of additional covariates. Among the TP53‐mutated subgroup (n = 110), the median PFS was significantly longer with combination therapy than with monotherapy (18.69 vs. 16.30 months; HR, 0.55; 95% CI, 0.32–0.95; see Figure S1e). No benefit was observed in the non–TP53‐mutated subgroup (HR, 0.24; 95% CI, 0.04–2.52; see Figure S1f).
In prespecified EGFR‐mutation subgroups, the magnitude of PFS improvement with combination therapy was consistent with the ITT population, although statistical significance was not reached (see Figure S1a–f). The median PFS was numerically longer with the combination both in patients who had exon 19 deletion (not reached vs. 16.30 months; HR, 0.45; 95% CI, 0.19–1.04) and in those who had L858R‐mutated disease (18.66 vs. 16.53 months; HR, 0.62; 95% CI, 0.31–1.21). A similar trend was observed among patients without baseline CNS metastases (20.96 vs. 16.30 months; HR, 0.51; 95% CI, 0.25–1.05).
Among patients with TP53‐mutated NSCLC (N = 110), 23 (18.3%) harbored co‐alterations in at least one additional TSG (the co‐TSG group), whereas 87 (69.0%) had a TP53 mutation only (see Table S1). In the subgroup with only TP53 mutations, combination therapy significantly prolonged median PFS (19.78 vs. 15.51 months; HR, 0.50; 95% CI, 0.26–0.95). This benefit was not observed in the co‐TSG group (18.66 vs. 16.30 months; HR, 0.86; 95% CI, 0.29–2.53). Notably, patients with TP53 mutations accompanied by TSG alterations other than RB1 (n = 90) derived a significant median PFS benefit (19.78 vs. 15.51 months; HR, 0.50; 95% CI, 0.27–0.94), whereas those with concurrent TP53 and RB1 mutations (n = 20) did not appear to derive a PFS benefit from the addition of chemotherapy (see Figure S2a–d).
Exploratory domain‐based analyses of TP53 mutations showed trends favoring combination therapy for exon 5, 6 and 7 variants. The numerical effect was observed for exon 6 variants (20.96 vs. 12.48 months; HR, 0.13; 95% CI, 0.02–1.14; nominal p = .0306). Similar trends were observed for exon 5 (19.78 vs. 16.56 months; HR, 0.57; 95% CI, 0.16–2.05) and exon 7 (18.66 vs. 12.55 months; HR, 0.74; 95% CI, 0.21–2.58) variants (see Figure S3a–c). When functionally classified, aumolertinib plus carboplatin–pemetrexed improved PFS in both dominant‐negative effect and loss‐of‐function subgroups, as observed in the loss‐of‐function subgroup (18.14 vs. 16.56 months; HR, 0.43; 95% CI, 0.18–1.03) and the dominant‐negative effect subgroup (18.69 vs. 15.51 months; HR, 1.13; 95% CI, 0.47–2.74; see Figure S4a–b). OS data remained immature at the time of analysis, with an event rate of only 4%; longer term follow‐up is ongoing.

Tumor response
In the FAS population, combination therapy was associated with a higher investigator‐assessed ORR compared with monotherapy (72.2% [95% CI, 59.1%–82.3%] vs. 67.2% [95% CI, 55.0%–77.4%]). No complete responses were observed in either treatment arm (Figure 3; see Table S2). The DCR was 92.6% (95% CI, 82.5%–97.1%) with combination therapy and 98.3% (95% CI, 91.7%–99.7%) with monotherapy. Median DoR was 17.28 months in the combination group and 14.92 months in the monotherapy group.

Safety
The median duration of treatment was 14.54 months (range, 8.51–20.53 months) with combination therapy and 11.29 months (range, 7.51–17.51 months) with monotherapy; 79.6% of patients completed four or more cycles of carboplatin, and 81.5% continued pemetrexed maintenance (median, nine cycles; see Table S3).
Treatment‐related AEs occurred in 90.7% of patients receiving aumolertinib plus carboplatin–pemetrexed and in 75.0% of those receiving aumolertinib alone. Grade ≥3 AEs were reported in 14 of 54 patients (25.9%) in the combination group and 11 of 64 patients (17.2%) in the monotherapy group. Serious AEs occurred in 12 of 54 patients (22.2%) and four of 64 patients (6.3%), respectively. No drug‐related deaths were observed. Discontinuation of any study drug because of treatment‐related AEs occurred in six of 54 patients (11.1%) receiving combination therapy and in two of 64 patients (3.1%) receiving monotherapy. Dose interruptions occurred for nine patients in the combination group (16.7%) and for nine patients in the monotherapy group (14.1%), and dose reductions in occurred in five patients (9.3%) versus zero patients, respectively. All dose reductions occurred with chemotherapy and were because of carboplatin in four patients (7.4%) and pemetrexed in five patients (9.3%), whereas no patients required aumolertinib dose reduction.
Chemotherapy‐related hematologic toxicities accounted for most AEs in the combination arm. Among treatment‐related AEs that occurred in ≥20% of patients (Table 2), anemia was the most common AE (57.4% vs. 9.4%), followed by decreases in white blood cell count and increases in aspartate and alanine aminotransferase levels. No new safety concerns were identified.

DISCUSSION
The phase 3 ACROSS2 trial is the first prospective, randomized study to evaluate a third‐generation EGFR‐TKI combined with carboplatin–pemetrexed in a molecularly defined subset of patients with advanced, EGFR‐mutated NSCLC characterized by co‐occurring TSG alterations, which are present in up to 10%–65% of EGFR‐mutated tumors.
7
In this population, aumolertinib plus carboplatin–pemetrexed significantly prolonged PFS compared with aumolertinib monotherapy (19.78 vs. 16.53 months), corresponding to a 42% reduction in the risk of progression. The PFS benefit was generally consistent across all major clinicopathologic subgroups. The DCR was numerically higher with monotherapy (98.3% vs. 92.6%); however, the median DoR was shorter (14.92 vs. 17.28 months). This indicates that short‐term efficacy end points were comparable, whereas the PFS advantage stemmed from a more durable responses with the chemotherapy–TKI combination. Although CNS metastases were a stratification factor, CNS PFS was not a prespecified end point; exploratory analyses revealed lower rates of new brain lesions with the combination (9.7% vs. 12.5%) and a consistent trend in CNS PFS (see Figure S5); formal CNS outcome work is planned. The safety profile of the combination was in line with established experience for this regimen,
12
,
14
,
15
and no new safety signals were identified.
FLAURA2 and AENEAS2 enrolled all patients who had sensitizing EGFR mutations,
12
,
14
whereas ACROSS2 selected patients who had a sensitizing EGFR mutation and co‐occurring TSG alterations. Our data support the concept of selecting the combination therapy for patients who have co‐occurring TSG, such as TP53. Retrospective data from FLAURA 2 have demonstrated no significant difference between the TP53‐mutated and wild‐type subgroups. Median PFS for the TP53‐altered subgroup in the two arms were similar (27.6 vs. 27.6 months; HR, 0.57; 95% CI, 0.29–1.12).
16
However, our current study has prospectively generated the hypothesis that the presence of co‐occurring TSG mutations may be predictive. Direct cross‐trial comparisons should nevertheless be interpreted cautiously because of differences in patient populations, study design, and end point definitions.
TSG alterations like TP53, RB1, and PTEN disrupt genomic stability, apoptosis, and cell‐cycle regulation, promoting clonal diversification and facilitating resistance under EGFR‐TKI pressure.
17
,
18
These disruptions may allow greater subclonal heterogeneity and outgrowth of EGFR‐independent clones.
19
The addition of chemotherapy may counteract this process by exerting cytotoxic effects on both EGFR‐dependent and EGFR‐independent cell populations, thereby delaying progression.
20

In ACROSS2, TP53 was the predominant TSG alteration (110 of 126 patients; 87.3%), consistent with its status as the most frequently mutated TSG in NSCLC. FLAURA2 also included post‐hoc analysis of a TP53 co‐mutation subgroup; however, combination therapy demonstrated no trend of benefit compared with monotherapy in that trial.
16
Comparisons are limited by the post‐hoc analysis of the FLAURA2 trial because it included only a small proportion patients (86 of 557 patients; 15.4%) harboring TP53 co‐mutations and did not encompass patients with co‐TSG mutations, limiting comparability with ACROSS2. In the MARIPOSA trial (ClinicalTrials.gov identifier NCT04487080),
21
293 patients harbored TP53 co‐mutations, a markedly higher population than in the FLAURA2 post‐hoc subset. Among these patients with TP53‐mutated disease, the amivantamab–lazertinib combination reduced the risk of progression by 35% (median PFS, 18.2 vs. 12.9 months; HR, 0.65; 95% CI, 0.48–0.87), closely mirroring the 45% reduction observed in the ACROSS2 study (HR, 0.55; 95% CI, 0.32–0.95).
ACROSS2 enabled detailed characterization of TP53‐mutated subgroups. Exon‐level analyses suggested that DNA‐binding domain mutations (exons 5–7) were primarily responsible for the observed treatment effect, with exon 6 showing the largest point estimate benefit in a hypothesis‐generating, exploratory analysis. This aligns with previous evidence indicating that >80% of TP53 variants are missense mutations clustered within the DNA‐binding domain (exons 4–8), a region critical for sequence‐specific DNA recognition and transcriptional control of cell‐cycle checkpoints and apoptosis.
22
In contrast, TP53 and RB1 co‐mutation appeared to negate the benefit of combination therapy (HR, 0.95; 20 patients overall, nine in the combination arm and 11 in the monotherapy arm). Biallelic loss of TP53 and RB1 promotes neuroendocrine reprogramming and drives intrinsic and acquired resistance to EGFR‐TKIs.
23

RB1 loss also abolishes the G1/S checkpoint and increases reliance on replication‐coupled DNA repair, attenuating platinum‐induced DNA damage and apoptosis.
24
Our findings provide preliminary clues for identifying subtypes of TSGs and suggest that molecular stratification based on functional mutation of TSGs, especially TP53, needs to be further explored to guide clinical practice.
Because chemotherapy is nonselective and is associated with greater toxicity, its use must be judicious. In ACROSS2, aumolertinib plus carboplatin–pemetrexed was associated with a higher incidence of myelotoxicity than monotherapy. These cytopenias were transient, manageable with dose modification and supportive care, and rarely led to treatment discontinuation. No clinically meaningful QT prolongation, interstitial lung disease, or vascular events were observed, consistent with the known safety profiles of aumolertinib and pemetrexed‐based chemotherapy. Overall, the toxicity findings were aligned with those previously reported in AENEAS2.
14

This study has limitations that should be taken into account when interpreting the results. First, all participants were Asian patients of Chinese origin, which may limit generalizability to other populations. Nevertheless, the incidence of co‐mutation was comparable between Asian and non‐Asian EGFR‐mutant populations (corresponding rates were approximately 51% in Chinese and 46% in US lung cancer adenocarcinoma trials).
25
Second, OS was a secondary end point and remains immature (event rate, 4%); longer follow‐up is required to define definitive OS benefit. Third, as an open‐label study, the imaging assessment of ACROSS2 was conducted by investigators rather than by independent central review.
Collectively, ACROSS2 represents the first prospective, randomized, multicenter, phase 3 study to specifically enroll patients who had EGFR‐mutated NSCLC with TSG co‐alterations. It provides clinically actionable evidence supporting a precision‐based intensification strategy for this molecularly defined population. By integrating genomic stratification into first‐line decision‐making, ACROSS2 highlights how targeted and cytotoxic therapies can be synergistically deployed to address genomic complexity in lung cancer. The study also demonstrates that aumolertinib plus carboplatin–pemetrexed delivers statistically significant and clinically meaningful PFS improvement compared with aumolertinib monotherapy and sets the stage for prospective molecular triage algorithms that optimize outcomes while minimizing unnecessary chemotherapy exposure. Therefore, ACROSS2 should be viewed as a proof of concept that genomically enriched chemotherapy–TKI treatment is feasible, and analogous strategies can be tested with osimertinib or other TKIs. For routine adoption, broad, upfront, multigene NGS is essential; in patients for whom reimbursement for chemotherapy–TKI is restricted, clinicians could reserve the combination for the TSG‐positive patients who most likely will derive the PFS gain observed here while awaiting mature OS data.

AUTHOR CONTRIBUTIONS

Jian‐chun Duan: Study design/planning, interpretation of results, and writing–review and editing. Jia Zhong: Study design/planning, interpretation of results, and writing–review and editing. Bo‐yang Sun: Acquisition and analysis of the data, interpretation of results, and writing–review and editing. Wen‐hua Zhao: Acquisition and analysis of the data, interpretation of results, and writing–review and editing. Lin Wu: Acquisition and analysis of the data, interpretation of results, and writing–review and editing. Kai‐lun Fei: Acquisition and analysis of the data, interpretation of results, and writing–review and editing. Qian Chu: Acquisition and analysis of the data, interpretation of results, and writing–review and editing. Qi‐sen Guo: Acquisition and analysis of the data, interpretation of results, and writing–review and editing. Qi‐bin Song: Acquisition and analysis of the data, interpretation of results, writing–review and editing. Yan Yu: Acquisition and analysis of the data, interpretation of results, and writing–review and editing. Da‐xing Zhu: Acquisition and analysis of the data, interpretation of results, and writing–review and editing. Xin‐yan Liu: Acquisition and analysis of the data, interpretation of results, and writing–review and editing. Jun Zhao: Acquisition and analysis of the data, interpretation of results, and writing–review and editing. Zhi‐xiang Zhan: Acquisition and analysis of the data, interpretation of results, and writing–review and editing. Shi Li: Acquisition and analysis of the data, interpretation of results, and writing–review and editing. Lei Nie: Acquisition and analysis of the data, interpretation of results, and writing–review and editing. Jie Lin: Acquisition and analysis of the data, interpretation of results, and writing–review and editing. Xiao‐dong Peng: Acquisition and analysis of the data, interpretation of results, and writing–review and editing. Dian‐sheng Zhong: Acquisition and analysis of the data, interpretation of results, and writing–review and editing. Jin Zhou: Acquisition and analysis of the data, interpretation of results, and writing–review and editing. Li‐hua Li: Acquisition and analysis of the data, interpretation of results, and writing–review and editing. Yun‐fang Chen: Acquisition and analysis of the data, interpretation of results, and writing–review and editing. Chen Hu: Acquisition and analysis of the data, interpretation of results, and writing–review and editing. Tony Mok: Writing–review and editing and supervision. Zhi‐jie Wang: Study design/planning, interpretation of results, and writing–review and editing. Jie Wang: Conceptualization, study design/planning, interpretation of results, and writing–review and editing. All authors had access to all the relevant study data and related analyses, vouched for the completeness and accuracy of the data presented, and had final responsibility to submit this article for publication.

CONFLICT OF INTEREST STATEMENT
Jie Lin reports expert witness fees from the International Association for the Study of Lung Cancer (IASLC) outside the submitted work. Chen Hu reports personal/consulting fees from Belay Diagnostics LLC and Johnson & Johnson outside the submitted work. The remaining authors disclosed no conflicts of interest.

Supporting information
Supporting Information S1

Supporting Information S2

Supporting Information S3