PURPOSE: The optimal perioperative chemotherapy regimen for patients with nonmetastatic muscle-invasive bladder cancer is not defined. PATIENTS AND METHODS: Between February 2013 and March 2018, 500 patients were randomly assigned in 28 French centers and received either six cycles of dose-dense methotrexate, vinblastine, doxorubicin, and cisplatin (dd-MVAC) once every 2 weeks or four cycles of gemcitabine and cisplatin (GC) once every 3 weeks before surgery (neoadjuvant group) or after surgery (adjuvant group). We report the primary end point of the GETUG-AFU V05 VESPER trial (ClinicalTrials.gov identifier: NCT01812369 ): progression-free survival (PFS) at 3 years. Secondary end points were time to progression and overall survival. RESULTS: Four hundred thirty-seven patients (88%) received neoadjuvant chemotherapy; 60% of patients received the planned six cycles in the dd-MVAC arm, 84% received four cycles in the GC arm, and thereafter, 91% and 90% of patients underwent surgery, respectively. Organ-confined response (< ypT3N0) was observed more frequently in the dd-MVAC arm (77% v 63%, P = .001). In the adjuvant group, 40% of patients received six cycles in the dd-MVAC arm, and 81% of patients received four cycles in the GC arm. For all patients in the clinical trial, 3-year PFS was improved in the dd-MVAC arm, but the study did not meet its primary end point (3-year rate: 64% v 56%, hazard ratio [HR] = 0.77 [95% CI, 0.57 to 1.02], P = .066); nevertheless, the dd-MVAC arm was associated with a significantly longer time to progression (3-year rate: 69% v 58%, HR = 0.68 [95% CI, 0.50 to 0.93], P = .014). In the neoadjuvant group, PFS at 3 years was significantly higher in the dd-MVAC arm (66% v 56%, HR = 0.70 [95% CI, 0.51 to 0.96], P = .025). CONCLUSION: In the VESPER trial, dd-MVAC improved 3-years PFS over GC. In the neoadjuvant group, a better bladder tumor local control and a significant improvement in 3-year PFS were observed in the dd-MVAC arm.

Neoadjuvant cisplatin-based chemotherapy (NAC) followed by radical cystectomy and pelvic lymph node dissection is the optimal treatment for patients with muscle-invasive bladder cancer. In recent years, the VESPER trial showed a statistically significant higher progression-free survival with dd-MVAC (dose dense methotrexate, vinblastine, doxorubicin, and cisplatin) compared to GC (gemcitabine and cisplatin). In the present report, we refine the characterization and outcome of patients whose cystectomy specimens were pathologically free of cancer (pathological complete response, pCR). We confirm that these patients portend a better outcome as compared to patients with invasive disease (≥pT1N0) at cystectomy. Nested variant and lymphovascular invasion were identified as adverse predictive factors of pCR. Progression-free survival probability three years after pCR on cystectomy was about 85%, regardless of the NAC regimen. A lower creatinine clearance and the delivery of less than four cycles were associated with a higher risk of relapse. Predicting the efficacy of NAC remains a major challenge. The planned analysis of molecular subtypes in the VESPER trial could help predict which patients may achieve complete response and better outcome.

Cancer staging plays a critical role in treatment planning and prognosis but is often embedded in unstructured clinical narratives. To automate the extraction and structuring of staging data, large language models (LLMs) have emerged as a promising approach. However, their performance in real-world oncology settings has yet to be systematically evaluated. Herein, we analysed 1000 oncological summaries from patients receiving treatment for breast cancer between 2019 and 2020 at the François Baclesse Comprehensive Cancer Centre, France. Five Mistral artificial intelligence–based LLMs were evaluated (i.e. Small, Medium, Large, Magistral and Mistral:latest) for their ability to derive the cancer stage and identify staging elements. Larger models outperformed their smaller counterparts in staging accuracy and reproducibility (kappa > 0.95 for Mistral Large and Medium). Mistral Large achieved the highest accuracy in deriving the cancer stage (93.0%), surpassing the original clinical documentation in several cases. The LLMs consistently performed better in deriving the cancer stage when working through tumour size, nodal status and metastatic components compared to when they were directly requested stage data. The top-performing models had a test–retest reliability exceeding 97%, while smaller models and locally deployed versions lacked sufficient robustness, particularly in handling unit conversions and complex staging rules. The structured, stepwise use of LLMs that emulates clinician reasoning offers a more efficient, transparent and reproducible approach to cancer staging, and the study findings support LLM integration into digital oncology workflows.