Annals of Oncology最新文献

Background: Adding the pan-Akt serine/threonine kinase (AKT) inhibitor capivasertib to first-line paclitaxel in metastatic triple-negative breast cancer (TNBC) led to significantly longer progression-free survival (PFS) and overall survival (OS) versus placebo-paclitaxel in the phase II PAKT trial. CAPItello-290 was designed to further assess capivasertib-paclitaxel, including in patients with PIK3CA/AKT1/PTEN-altered tumours.

Patients and methods: Patients with previously untreated metastatic TNBC were randomised 1 : 1 to paclitaxel 80 mg/m² [day 1, weeks 1-3 (4-week cycle)] plus capivasertib 400 mg or placebo twice daily (days 2-5, weeks 1-3). PIK3CA/AKT1/PTEN alterations were analysed by retrospective central molecular testing. Dual primary endpoints were OS in the overall population and in patients with PIK3CA/AKT1/PTEN-altered tumours; investigator-assessed PFS was a key secondary endpoint.

Results: From July 2019 to February 2022, 812 patients were randomised; 30.7% of patients had PIK3CA/AKT1/PTEN tumour alterations. At final analysis [data cut-off (DCO) 18 March 2024], the median OS for the overall population was 17.7 and 18.0 months with capivasertib-paclitaxel and placebo-paclitaxel, respectively [hazard ratio (HR) 0.92, 95% confidence interval (CI) 0.78-1.08, P = 0.3239] and for patients with PIK3CA/AKT1/PTEN-altered tumours, it was 20.4 months in both arms (HR 1.05, 95% CI 0.77-1.43, P = 0.7602). At PFS DCO (25 May 2022), the median PFS in the overall population numerically favoured capivasertib-paclitaxel (5.6 versus 5.1 months placebo-paclitaxel; HR 0.72, 95% CI 0.61-0.84); this was also the case in patients with PIK3CA/AKT1/PTEN-altered tumours (7.5 versus 5.6 months placebo-paclitaxel; HR 0.70, 95% CI 0.52-0.95). The most frequent adverse event (AE) of grade ≥3 was diarrhoea [12.7% versus 0.7% placebo-paclitaxel (overall population)]. Capivasertib was discontinued due to AEs in 8.5% of patients (4.9% placebo-paclitaxel; overall population); AEs led to death in 4.2% of all patients.

Conclusions: Capivasertib-paclitaxel did not meet the prespecified boundary for improving OS in either population; PFS numerically favoured the combination, especially in PIK3CA/AKT1/PTEN-altered tumours. The safety of capivasertib-paclitaxel was generally manageable and consistent with prior studies.

Background: Germline alterations and smoking status in lung cancer could inform etiology and clinical decisions. We investigated the prevalence of germline alterations in predisposition genes across various lung cancer histologies in two large populations.

Patients and methods: Germline sequencing of 11 740 primary lung cancers was carried out with Tempus xT tumor-normal matched assay (DNA sequencing of 648 genes at an average coverage of 500×, normal specimens at 150× coverage, full transcriptome RNA sequencing). Pathogenic/likely pathogenic (P/LP) potential germline alterations in 46 genes were compared between smokers and never smokers; never smokers somatic EGFR altered (sEGFRalt) and wild type (sEGFRwt); non-small-cell lung cancer (NSCLC) and small-cell lung cancer (SCLC) histologies; and NSCLC sEGFRalt and NSCLC sEGFRwt. P/LP variants were investigated in these 46 genes in 1330 patients with lung cancer from the UK Biobank by smoking status.

Results: Tempus sequencing revealed P/LP alterations in 4.8% of smokers and 5.8% of never smokers, with most alterations in MUTYH (1.3% versus 1.1%), ATM (0.7% versus 1.0%), BRCA2 (0.6% versus 0.9%), and EGFR (<0.1% versus 0.4%). Never smoker sEGFRalt (n = 549) and sEGFRwt (n = 1025) tumors had alterations in MUTYH (1.1% versus 1.1%), ATM (0.7% versus 1.1%), and EGFR (1.1% versus 0%). NSCLC and SCLC tumors had alterations in MUTYH (1.3% versus 0.3%), ATM (0.8% versus 0.3%), and BRCA2 (0.7% versus 0%). sEGFRalt and sEGFRwt NSCLC tumors had germline alterations in MUTYH (1.6% versus 1.3%), ATM (0.5% versus 0.8%), EGFR (1.3% versus 0%), and BRCA2 (0.8% versus 0.6%). UK Biobank patients had similar P/LP alterations: 4.3% of smokers and 5.1% of never smokers, with most germline alterations in ATM (0.8%), BRCA2 (0.79%), and MUTYH (0.62%) in smokers and MUTYH (1.5%) and CHEK2 (1.01%) in never smokers.

Conclusion: Similar distribution of P/LP potential germline alterations in lung cancer subtypes from distinct populations by smoking status suggests that increased next-generation germline sequencing may improve risk assessment.

Background: Manual abstraction of real-world data (RWD) from unstructured health records (HRs) remains resource intensive, error prone, and highly variable across institutions. Large language models (LLMs) offer a scalable alternative, but their performance in multicenter oncology settings is not fully validated.

Patients and methods: We conducted a multicenter study within the French Large & Unified Cancer Cohort (LUCC) consortium to compare the accuracy of artificial intelligence (AI)-based data extraction against manual abstraction by clinical research professionals. A fine-tuned LLM was applied to de-identified unstructured HRs in PDF format to extract 31 variables from lung cancer patients across 10 centers. Ground truth was defined as concordant values across sources, with discrepant cases adjudicated by a blinded expert. The primary endpoint was the extraction error rates. Secondary endpoints included per-variable performance, interinstitutional variability, F1-score for multiple-choice variables, added value of hybrid AI-human workflows, and survival analyses.

Results: Among 10 327 patients with AI-based extraction, 311 were included in the test cohort. Across 8708 datapoints for 28 variables with only one correct answer, the LLM achieved a 7.0% error rate, outperforming manual abstraction (14.2%, P <0.001). The F1-scores of three multiple-choice variables were superior (gene alterations 0.97 versus 0.86, comorbidities 0.86 versus 0.76, metastatic sites 0.71 versus 0.69). Interinstitutional variance was lower with AI (0.12% versus 0.39%). A hybrid approach with targeted human review of 30% of low-confidence AI outputs further decreased error rates to 4.4%. Survival analyses based on AI-extracted data closely matched ground truth, with similar median overall and progression-free survival.

Conclusions: In a multicenter setting, our AI pipeline yielded lower error rates and greater consistency than manual abstraction. These findings support the feasibility of next-generation, AI-enabled multicenter studies to generate high-quality RWD at scale, with potential applicability in prospective clinical trials.

Background: At the primary progression-free survival (PFS) analysis, the phase III EMBER-3 trial in endocrine therapy-pretreated patients with estrogen receptor (ER)-positive, human epidermal growth factor receptor 2 (HER2)-negative advanced breast cancer (ABC) demonstrated significant PFS benefit with imlunestrant versus standard of care (SOC: fulvestrant or exemestane) in patients with ESR1 mutations (ESR1m) and with imlunestrant-abemaciclib versus imlunestrant in all patients, regardless of ESR1m. In this article, we report updated efficacy from a prespecified interim overall survival (OS) analysis.

Patients and methods: Patients with ER-positive, HER2-negative ABC previously treated with aromatase inhibitors ± cyclin-dependent kinase 4 and 6 inhibitors were randomly assigned (1 : 1 : 1) to receive imlunestrant, SOC, and imlunestrant-abemaciclib. Primary endpoints were PFS in imlunestrant versus SOC in patients with ESR1m and all patients, and versus imlunestrant-abemaciclib in all concurrently randomized patients. OS was a key secondary endpoint (tested if the corresponding PFS was statistically significant). Due to only two of three PFS endpoints being met, a limited significance level was passed to the OS comparisons. Exploratory endpoints included time to chemotherapy, chemotherapy-free survival, and PFS2.

Results: A total of 874 patients were randomized (imlunestrant, n = 331; SOC, n = 330; imlunestrant-abemaciclib, n = 213). Median follow-up was 28.5 months; 10.1% of patients remained on treatment (data cut-off: 18 August 2025).In patients with ESR1m, median OS (mOS) was 34.5 months for imlunestrant versus 23.1 months for SOC [hazard ratio (HR) 0.60, 95% confidence interval (CI) 0.43-0.86, P = 0.0043, boundary for significance not reached]. In all patients regardless of ESR1m, mOS was not reached with imlunestrant-abemaciclib versus 34.4 months with imlunestrant (HR 0.82, 95% CI 0.59-1.16, P = 0.2622). Updated PFS demonstrated sustained benefit. Notably, in all patients regardless of ESR1m, the median PFS of imlunestrant-abemaciclib versus imlunestrant was 10.9 versus 5.5 months (HR 0.59, 95% CI 0.47-0.74, nominal P < 0.0001). All prespecified exploratory endpoints favored imlunestrant-based regimens. Safety remains consistent with prior reports.

Conclusions: These findings reinforce the clinical benefit of imlunestrant-based regimens as a potential all-oral, chemotherapy-free treatment option for endocrine-pretreated patients with ER-positive, HER2-negative ABC.

Twenty years after its initial approval, trastuzumab remains a cornerstone in the treatment of patients with human epidermal growth factor receptor 2 (HER2)-positive early breast cancer (eBC). Long-term follow-up from pivotal adjuvant trials has consistently demonstrated significant and durable improvements in survival outcomes across various risk groups and chemotherapy backbones. In parallel, trastuzumab-induced cardiotoxicity (TIC) remains overall infrequent, particularly in patients without prior exposure to anthracycline and appears comparable to the incidence observed in the general population after treatment completion. While real-world data further support the long-term efficacy and safety of trastuzumab, advancements such as more convenient subcutaneous formulations and the widespread availability of more accessible cost-effective biosimilars solidify its ongoing relevance in clinical practice. Conversely, despite two decades of clinical and translational research, no predictive biomarker beyond HER2 overexpression or amplification has yet been validated to guide trastuzumab use. Emerging candidates, including stromal tumor-infiltrating lymphocytes, circulating tumor DNA, and the HER2DX genomic assay, are not yet validated for use in clinical practice, although prospective studies are ongoing. Similarly, while clinical factors and imaging tools may help identify early on patients at higher risk of experiencing TIC, no cardioprotective strategy has yet demonstrated robust and conclusive benefit. Despite the emergence of newer anti-HER2 agents and evolving treatment paradigms, trastuzumab will probably continue to serve as a key therapeutic backbone, especially for patients with lower-risk HER2-positive eBC.

Background: Thymic epithelial tumors (TETs) are rare malignancies that pose significant diagnostic challenges due to their heterogeneous histological patterns and substantial interobserver variability in classification. Despite standardized World Health Organization (WHO) classification criteria, diagnostic concordance remains suboptimal, particularly in non-expert settings, where second-opinion reviews lead to diagnostic reclassification in up to 57% of cases. Deep learning may offer a tool to reduce diagnostic variability and improve the consistency of histological classification.

Methods: We trained a deep learning-based model using hematoxylin and eosin (H&E) whole-slide images from The Cancer Genome Atlas as a training dataset. The model incorporated a novel hierarchical loss function designed to reflect clinically relevant tumor groupings based on treatment strategies and patient outcomes. We validated the model on 112 consecutive cases from the University of Chicago, with diagnoses confirmed by an expert thoracic pathologist. Model performances were evaluated using both a three-group hierarchical scheme and the six-class WHO classification.

Results: In the clinically relevant hierarchical three-group classification (As: A+AB; Bs: B1+B2+B3; Thymic Carcinoma), the model achieved an accuracy of 91.1% with Cohen's κ= 0.859, indicating almost perfect agreement. In the six-class classification (A, AB, B1, B2, B3, Thymic Carcinoma), the accuracy was 77.7% with κ = 0.716. The model demonstrated 100% sensitivity and 94.6% accuracy for thymic carcinoma detection. Notably, 60% of misclassifications occurred within the same clinical management group, thereby limiting their impact on therapeutic decision-making.

Conclusion: This deep learning model demonstrates strong potential as a diagnostic tool for TETs classification, particularly in settings with limited thoracic pathology expertise. The high sensitivity for thymic carcinoma detection and robust performance across different tissue processing conditions suggest its clinical applicability for improving diagnostic consistency and supporting pathological decision-making in both specialized and non-specialized settings.