保科 勇斗

BACKGROUND: Triplet therapy with darolutamide and docetaxel has reshaped the treatment landscape of metastatic castration-sensitive prostate cancer (mCSPC). However, real-world evidence regarding its oncological outcomes and toxicity remains limited. METHODS: We retrospectively reviewed 112 consecutive patients with mCSPC who were treated with triplet therapy. A deep prostate-specific antigen (PSA) response was defined as a PSA level < 0.2 ng/ml or a ≥ 90% decline from baseline. Castration-resistant prostate cancer (CRPC)-free survival was used as the survival endpoint. Clinicopathological factors associated with CRPC-free survival were analysed using a Cox proportional hazards model. Treatment-related adverse events (AEs) were comprehensively documented by type and grade. RESULTS: During a median follow-up of 17.1 months, 15 patients developed CRPC and four died of prostate cancer; 1-year CRPC-free survival and overall survival rates were 89.8% (95% confidence interval, 83.9%-96.0%) and 98.8% (96.5%-100%), respectively. A deep PSA response was achieved in 73.2% of patients (PSA nadir <0.2 ng/ml) and in 96.4% (PSA ≥ 90% decline). In both the univariable and multivariable analyses, no baseline factor was significantly associated with CRPC-free survival. Grade ≥ 3 hematologic AEs occurred in 75.0% of patients. Among the 31 patients who discontinued docetaxel, 26 discontinued due to non-hematologic AEs. CONCLUSIONS: Triplet therapy demonstrated substantial clinical activity in real-world patients with mCSPC, with high rates of deep PSA response despite frequent AEs. These findings support the feasibility of triplet therapy in routine practice. Given the substantial toxicity burden, careful patient selection and proactive AE management remain important.

OBJECTIVES: To evaluate nerve-sparing status objectively using an artificial intelligence method applied to intraoperative images and to determine whether combining this assessment with preserved urethral length improves prediction of urinary continence recovery after robot-assisted radical prostatectomy. METHODS: We analyzed 195 patients with intraoperative still images from robot-assisted radical prostatectomy performed between May 2020 and 2022. Images were evaluated with a deep learning image classification model that determined whether the neurovascular bundle was preserved. Patients were categorized into two groups: those with both preserved urethral length of at least 26 mm and artificial intelligence-assessed neurovascular bundle preservation, and all other patients. The primary endpoint was time to urinary continence recovery. Kaplan-Meier analysis and Cox proportional hazards models were used. RESULTS: Artificial intelligence reclassified neurovascular bundle preservation in 60 of 195 patients (30.8%), including 48 cases of surgeon overestimation and 12 of underestimation. Patients who met both criteria had significantly better urinary continence recovery than the other patients, with higher continence rates at 30, 60, 90, and 365 days after surgery on log-rank testing. In multivariable Cox models, the combined factor independently predicted continence recovery at 60, 90, and 365 days, with hazard ratios of 1.99, 1.78, and 1.49 and corresponding 95% confidence intervals of 1.07-3.72, 1.07-2.95, and 1.09-2.04 (p values 0.03, 0.03, and 0.01, respectively). CONCLUSIONS: Objective evaluation of nerve-sparing status using artificial intelligence, when combined with preserved urethral length, improves prediction of early urinary continence recovery after robot-assisted radical prostatectomy and may provide an indicator of surgical quality.

PURPOSE: Preserving neurovascular bundles (NVB) during robot-assisted radical prostatectomy (RARP) is vital for reducing postoperative complications such as urinary incontinence and erectile dysfunction. Building on our previous work in ensemble-based NVB classification, we propose the hybrid self-supervised teacher-student model (Hybrid T-S model) that leverages multi-task learning to predict NVB preservation in prostatectomy videos. METHODS: Our approach integrates a self-supervised framework (DINO) as an online self-distillation objective on multi-crop views to learn robust embeddings in a limited data setting, rather than as a stand-alone large-scale pretraining. A teacher encoder, which is an exponential moving average (EMA) of the student encoder, and a reconstruction decoder are trained jointly with a classification head in a single end-to-end framework. This model is evaluated on single frames from patients who underwent RARP surgery. RESULTS: Our experimental evaluation shows that the Hybrid T-S model outperforms previous NVB classification methods. This highlights the benefits of integrating self-supervised learning and multi-task objectives in this surgical context. We achieved an average accuracy of 86.55%, precision of 83.93%, recall of 90.73%, F1-score of 87%, and AUROC of 88.35%, based on fivefold cross-validation. CONCLUSION: Incorporating representation learning through self-distillation, classification, and reconstruction provides complementary signals that enhance the prediction of NVB preservation. Our Hybrid T-S model can assist surgeons in real decision-making and improve patient recovery.

We assessed the efficacy of the lung immune prognostic index (LIPI) in predicting the progression of pathological T3 renal cell carcinoma (RCC). The LIPI scores of patients with pathological T3 RCC were calculated in the pre- and post-operative phases. Patients were divided into zero-point, one-point, and two-point groups according to their LIPI score and into the upstage and non-upstage groups according to the pre- and post-operative increase in LIPI score. Overall survival (OS) was evaluated using Kaplan-Meier curves stratified by group. Univariate and multivariate analyses of OS were performed via Cox proportional hazard regression analysis. LIPI scores were calculated in 80 patients wherein blood sampling data were available. The upstage and non-upstage groups comprised eight and seventy-two patients, respectively. Kaplan-Meier curves showed a significant difference in the pre- to post-operative LIPI score upstage group. LIPI score change was a poor prognostic factor using univariate analysis (OS: hazard ratio (HR) = 4.10, 95% confidence interval (CI) = 1.07-15.61, p = 0.038) and multivariate analysis (OS: HR = 4.38, 95% CI = 1.13-16.89, p = 0.031). An increase in the LIPI score in the pre-operative phase was a poor prognostic factor for pathological T3 RCC.