This study retrospectively evaluates the clinical outcomes of definitive volumetric modulated arc therapy (VMAT) for high-risk or very high-risk locoregional prostate cancer patients from an Asian institution. Consecutive patients who received VMAT (76 Gy in 38 fractions) between January 2017 and June 2022 were included. Whole pelvic radiotherapy (WPRT) (46 Gy in 23 fractions) was employed for clinically node-negative disease (cN0) and a Roach estimated risk of ≥15%, as well as simultaneous integrated boost (SIB) of 55–57.5 Gy to node-positive (cN1) disease. The primary endpoint was biochemical relapse-free survival (BRFS). Secondary endpoints included radiographic relapse-free survival (RRFS), metastasis-free survival (MFS) and prostate cancer-specific survival (PCSS). A total of 209 patients were identified. After a median follow-up of 47.5 months, the 4-year actuarial BRFS, RRFS, MFS and PCSS were 85.2%, 96.8%, 96.8% and 100%, respectively. The incidence of late grade ≥ 2 genitourinary (GU) and gastrointestinal (GI) toxicity were 15.8% and 11.0%, respectively. No significant difference in cancer outcomes or toxicity was observed between WPRT and prostate-only radiotherapy for cN0 patients. SIB to the involved nodes did not result in increased toxicity. International Society of Urological Pathology (ISUP) group 5 and cN1 stage were associated with worse RRFS (p < 0.05). PSMA PET-CT compared to conventional imaging staging was associated with better BRFS in patients with ISUP grade group 5 (p = 0.039). Five-year local experience demonstrates excellent clinical outcomes. PSMA PET-CT staging for high-grade disease and tailored pelvic irradiation based on nodal risk should be considered to maximize clinical benefit.

Background: Urothelial carcinoma is one of the most commonly diagnosed cancers worldwide, with a poor 5-year survival rate. As genomics is the backbone of the precision medicine paradigm, the genetic signature in circulating tumor DNA (ctDNA) is emerging as a pivotal biomarker for detecting early-stage cancer and molecular residual disease (MRD). Objective: We aim to evaluate the feasibility and preliminary effects of a ctDNA-based sequencing approach for detecting MRD and guiding adjuvant chemotherapy in postoperative urothelial carcinomas. Methods: We will perform a stratified 2-arm pilot randomized controlled trial in 2 tertiary hospitals in Hong Kong, involving patients with urothelial carcinomas (pT2-4a N0-2 M0) undergoing radical resection. We plan to recruit 20 patients and determine stratification according to their MRD status before randomization. Patients in each stratum (MRD-positive and MRD-negative groups) will be randomly allocated to either a 4-cycle gemcitabine plus cisplatin chemotherapy arm or a standard management arm in a 1:1 ratio. ctDNA MRD will be tested using a personalized next-generation sequencing panel, which is designed based on the individual’s whole exome sequencing results from the operation specimen. The primary outcome is the feasibility of this trial (ie, recruitment, retention, adherence, and completeness). The secondary outcome is treatment-related adverse events. Exploratory outcomes include radiographic disease-free survival, cancer-specific survival, overall survival, ctDNA clearance in patients with ctDNA MRD-positive status, quality of life, fear of cancer recurrence, and cost-effectiveness. Benchmarks for feasibility evaluation are set as (1) ≥20% recruitment response rate, (2) ≤20% loss to follow-up or withdrawal, (3) ≥80% intervention adherence, and (4) ≤20% missing value rate. Each benchmark will be assigned one score, and a total score of 4, 2 to 3, and 0 to 1 will be deemed high, medium, and low feasibility, respectively. Safety evaluations will be presented as numbers and proportions of the adverse events. ANOVA and the Kruskal-Wallis test will be used for continuous outcome variables, whereas the chi-square test and the Fisher exact test will be used for categorical outcome variables. Hazard ratios will be calculated to compare the preliminary treatment effect of the gemcitabine plus cisplatin arm against the standard management arm on survival within each MRD group. Results: This project was funded in February 2024. Patient recruitment started on May 2, 2024. Recruitment and data collection for the trial are ongoing. Data analysis will be performed in mid-2025 and the results of this study are expected to be published in late 2025. Conclusions: Genetic signature in ctDNA is informative for personalized management of postoperative urothelial carcinomas, including personalized treatment and early detection of disease progression. Trial Registration: ClinicalTrials.gov NCT06257017; https://clinicaltrials.gov/study/NCT06257017 International Registered Report Identifier (IRRID): DERR1-10.2196/72597

BACKGROUND Urothelial carcinoma is one of the most commonly diagnosed cancers worldwide, with a poor 5-year survival rate. As genomics is the backbone of the precision medicine paradigm, the genetic signature in circulating tumor DNA (ctDNA) is emerging as a pivotal biomarker for detecting early-stage cancer and molecular residual disease (MRD). OBJECTIVE We aim to evaluate the feasibility and preliminary effects of a ctDNA-based sequencing approach for detecting MRD and guiding adjuvant chemotherapy in postoperative urothelial carcinomas. METHODS We will perform a stratified 2-arm pilot randomized controlled trial in 2 tertiary hospitals in Hong Kong, involving patients with urothelial carcinomas (pT2-4a N0-2 M0) undergoing radical resection. We plan to recruit 20 patients and determine stratification according to their MRD status before randomization. Patients in each stratum (MRD-positive and MRD-negative groups) will be randomly allocated to either a 4-cycle gemcitabine plus cisplatin chemotherapy arm or a standard management arm in a 1:1 ratio. ctDNA MRD will be tested using a personalized next-generation sequencing panel, which is designed based on the individual’s whole exome sequencing results from the operation specimen. The primary outcome is the feasibility of this trial (ie, recruitment, retention, adherence, and completeness). The secondary outcome is treatment-related adverse events. Exploratory outcomes include radiographic disease-free survival, cancer-specific survival, overall survival, ctDNA clearance in patients with ctDNA MRD-positive status, quality of life, fear of cancer recurrence, and cost-effectiveness. Benchmarks for feasibility evaluation are set as (1) ≥20% recruitment response rate, (2) ≤20% loss to follow-up or withdrawal, (3) ≥80% intervention adherence, and (4) ≤20% missing value rate. Each benchmark will be assigned one score, and a total score of 4, 2 to 3, and 0 to 1 will be deemed high, medium, and low feasibility, respectively. Safety evaluations will be presented as numbers and proportions of the adverse events. ANOVA and the Kruskal-Wallis test will be used for continuous outcome variables, whereas the chi-square test and the Fisher exact test will be used for categorical outcome variables. Hazard ratios will be calculated to compare the preliminary treatment effect of the gemcitabine plus cisplatin arm against the standard management arm on survival within each MRD group. RESULTS This project was funded in February 2024. Patient recruitment started on May 2, 2024. Recruitment and data collection for the trial are ongoing. Data analysis will be performed in mid-2025 and the results of this study are expected to be published in late 2025. CONCLUSIONS Genetic signature in ctDNA is informative for personalized management of postoperative urothelial carcinomas, including personalized treatment and early detection of disease progression. CLINICALTRIAL ClinicalTrials.gov NCT06257017; https://clinicaltrials.gov/study/NCT06257017 INTERNATIONAL REGISTERED REPORT DERR1-10.2196/72597

BACKGROUND As the population of survivors of cancer grows and new information technologies become widespread in Hong Kong, survivors of cancer increasingly seek ongoing care and support information from large language models (LLMs). While these tools provide immediate conversational responses, they carry substantial risks of generating inaccurate, unsafe, or generic medical advice. Evaluating LLM response performance is particularly critical in Hong Kong’s bilingual health care context. OBJECTIVE This study aimed to evaluate the clinical utility, safety, understandability, and readability of 4 popular commercial LLMs answering cancer survivorship questions in English and Traditional Chinese and to explore whether a structured prompting strategy improves response performance. METHODS We conducted a 2-phase evaluation study combining quantitative scoring with qualitative classification of unsafe responses. In phase 1, 25 expert-curated questions spanning 5 survivorship domains were input into 4 LLMs: Google Gemini 3.1 Pro, HK Chat-0.6.2, DeepSeek-V3.2, and Kimi-K2.5. A Delphi expert panel evaluated the baseline responses for clinical utility using a 5-point scale and assessed safety via binary categorization. Understandability was measured using the Patient Education Materials Assessment Tool (PEMAT). Readability was assessed via the Flesch-Kincaid Grade Level for English and lexical richness for Traditional Chinese. In phase 2, experts developed an optimized context, role, audience, format, task, and tone (CRAFT) prompt, and the resulting LLM responses were compared with baseline outputs in an exploratory within-sample analysis using paired statistical tests. RESULTS Baseline evaluations identified Google Gemini 3.1 Pro as the most clinically useful model in English (mean 3.99/5, SD 0.16) and Traditional Chinese (mean 4.21/5, SD 0.20). Unconstrained models demonstrated language-dependent safety vulnerabilities: English errors (up to 24%) centered on service mismatches and definitive interpretations, whereas Traditional Chinese errors (up to 20%) involved prescriptive disease management and unverified adjunctive therapies. Within each language, the models producing the most readable text were not those achieving the highest clinical utility; this relationship was not formally tested. Applying the CRAFT prompt to Gemini 3.1 Pro significantly improved clinical utility across both languages (English: mean 4.68, SD 0.18, P<.001; Traditional Chinese: 4.66, SD 0.23, P<.001). After prompting, the 2 unsafe English baseline responses were reclassified as safe (English 23/25 to 25/25 safe; Traditional Chinese 25/25 at both time points). This transition was not statistically significant on a post hoc exact McNemar test (P=.50) and is reported as descriptive. CONCLUSIONS LLMs exhibited language-divergent clinical safety vulnerabilities when answering cancer survivorship queries, and within each language, the most readable models were not the most clinically useful. In an exploratory within-sample comparison, expert-designed CRAFT prompting improved clinical utility and corrected the unsafe outputs seen at baseline without reducing understandability. Safe deployment of LLMs in multilingual oncology contexts requires strict validation combined with structured, role-restrictive constraints. CLINICALTRIAL

Hong Kong Journal of Radiology

Background: As the population of survivors of cancer grows and new information technologies become widespread in Hong Kong, survivors of cancer increasingly seek ongoing care and support information from large language models (LLMs). While these tools provide immediate conversational responses, they carry substantial risks of generating inaccurate, unsafe, or generic medical advice. Evaluating LLM response performance is particularly critical in Hong Kong’s bilingual health care context. Objective: This study aimed to evaluate the clinical utility, safety, understandability, and readability of 4 popular commercial LLMs answering cancer survivorship questions in English and Traditional Chinese and to explore whether a structured prompting strategy improves response performance. Methods: We conducted a 2-phase evaluation study combining quantitative scoring with qualitative classification of unsafe responses. In phase 1, 25 expert-curated questions spanning 5 survivorship domains were input into 4 LLMs: Google Gemini 3.1 Pro, HK Chat-0.6.2, DeepSeek-V3.2, and Kimi-K2.5. A Delphi expert panel evaluated the baseline responses for clinical utility using a 5-point scale and assessed safety via binary categorization. Understandability was measured using the Patient Education Materials Assessment Tool (PEMAT). Readability was assessed via the Flesch-Kincaid Grade Level for English and lexical richness for Traditional Chinese. In phase 2, experts developed an optimized context, role, audience, format, task, and tone (CRAFT) prompt, and the resulting LLM responses were compared with baseline outputs in an exploratory within-sample analysis using paired statistical tests. Results: Baseline evaluations identified Google Gemini 3.1 Pro as the most clinically useful model in English (mean 3.99/5, SD 0.16) and Traditional Chinese (mean 4.21/5, SD 0.20). Unconstrained models demonstrated language-dependent safety vulnerabilities: English errors (up to 24%) centered on service mismatches and definitive interpretations, whereas Traditional Chinese errors (up to 20%) involved prescriptive disease management and unverified adjunctive therapies. Within each language, the models producing the most readable text were not those achieving the highest clinical utility; this relationship was not formally tested. Applying the CRAFT prompt to Gemini 3.1 Pro significantly improved clinical utility across both languages (English: mean 4.68, SD 0.18, P<.001; Traditional Chinese: 4.66, SD 0.23, P<.001). After prompting, the 2 unsafe English baseline responses were reclassified as safe (English 23/25 to 25/25 safe; Traditional Chinese 25/25 at both time points). This transition was not statistically significant on a post hoc exact McNemar test (P=.50) and is reported as descriptive. Conclusions: LLMs exhibited language-divergent clinical safety vulnerabilities when answering cancer survivorship queries, and within each language, the most readable models were not the most clinically useful. In an exploratory within-sample comparison, expert-designed CRAFT prompting improved clinical utility and corrected the unsafe outputs seen at baseline without reducing understandability. Safe deployment of LLMs in multilingual oncology contexts requires strict validation combined with structured, role-restrictive constraints.