Tran, Phu N.

127 Background: Bone metastases affect up to 90 % of patients with metastatic hormone-resistant prostate cancer (mHRPC) and cause substantial morbidity from skeletal-related events (SREs) such as fracture, spinal-cord compression, and palliative interventions. Bone-modifying agents (BMAs) including bisphosphonates and denosumab reduce SRE risk and are NCCN-recommended, however, real-world use remains inconsistent. Methods: Patients with mHRPC treated from January 2017-September 2024 across Mayo Clinic sites were identified using SQL-based EHR queries confirmed by manual chart review. Bone Metastases were detected through keyword searchers of radiology reports and confirmed through manual chart review. SRE-prevention dosing was defined as zoledronic acid 4mg IV or denosumab 120mg SC; osteoporosis dosing was excluded. SREs were identified using diagnosis codes and text-pattern searches for radiotherapy, surgery, fracture, and spinal-cord compression. Descriptive statistics summarized BMA utilization and SRE incidence. Categorical variables were compared using Chi-square tests, and a risk ratio (RR) with 95% confidence intervals was calculated to evaluate association between early BMA initiation and SRE risk. Results: Among 778 patients with mHRPC, 685 (88 %) had bone metastases (mean age 70.5 years). Of these, 169 (24.7 %) had an SRE, and 261 (38.1 %) received BMAs at SRE-prevention doses (denosumab n = 157 [60 %]; zoledronic acid n = 104 [40 %]). Based on a chi-square test, combined SRE incidence did not differ significantly (No BMA 23 %, denosumab 23 %, zoledronic 33 %; p = 0.069). Bone pain at baseline was more common in patients who later received a BMA (136 [68 %]) than in those who did not (82 [41 %]; p < 0.001). Among 200 randomly sampled non-BMA patients, 83 had no recorded BMA discussion (66 also lacked bone pain), 38 lacked dental clearance, 32 had uncompleted plans and 18 declined therapy. Early BMA initiation (before first bone metastases) was associated with fewer SREs (16 % vs 29 %; RR 0.56, 95 % CI 0.33–0.96). Conclusions: Less than half of bone-metastatic mHRPC patients received BMAs. Non-initiation was often linked to modifiable factors such as delayed dental clearance or lack of treatment discussions, highlighting the need for system-level strategies to improve timely bone-directed care. SREs were common, affecting one quarter of patients. Small sample size precluded meaningful comparison of the impact of BMAs on SRE incidence however early BMA initiation may lower SREs. Incidence of SREs stratified by no BMA vs BMA treatment. SRE No BMA BMA Overall Pathological Fracture 6.4% (27/424) 4.9% (13/261) 5.8% (40/685) Radiotherapy to Bone 13.9% (59/424) 18.8% (49/261) 15.7% (108/685) Spinal Cord Compression 0.7% (3/424) 3.1% (8/261) 1.6% (11/685) Surgery to Bone 2.1% (9/424) 0.3% (1/261) 1.5% (10/685) Combined 23% (98/424) 27.2% (71/261) 24.7% (169/685)

405 Background: Clinical trials are critical to advancing prostate cancer treatment, yet patient enrollment remains a major bottleneck. Despite strong interest in AI for trial matching, real-world use remains limited due to reasons such as complex infrastructure requirements. To address this, we developed a lightweight, scalable AI framework leveraging the Google Healthcare Search API—requiring minimal technical expertise—to automate trial eligibility screening. Methods: We used a pre-trained GPT-5 large language model (LLM) to decompose the eligibility criteria of 13 prostate cancer clinical trials into 269 YES/NO questions, such that for an eligible patient of a given trial, the chart review would yield YES for inclusion and NO for exclusion questions. From 348 patients enrolled in one of the trials, we built a dataset linking each eligibility question to each patient, resulting in 4,947 patient–question pairs (2,034 inclusion pairs with YES labels and 2,913 exclusion pairs with NO labels). The task was framed as binary classification. During evaluation, for each patient–question pair, the LLM extracted search phrases from the question to retrieve relevant EHR information via the Google Healthcare Search API, then used the question and retrieved information to predict YES or NO responses. Performance was measured using accuracy, precision, recall, specificity, and F1 score. Results: Across 4,947 question–patient pairs, our framework achieved 0.81 accuracy, 0.85 precision, 0.65 recall, 0.92 specificity and a F1 score of 0.74. Results for different phases of trials is reported in Table 1, which demonstrates the approach’s robustness across trial phases. Predictions for exclusion criteria were highly precise, while inclusion criteria were more challenging. Most errors were observed in criteria involving complex nested logical conditions, interpretable clinical definitions and data not explicitly measurable or recorded in the EHR (e.g., planned treatments or clinician-assessed scores). Conclusions: This work demonstrates a practical AI framework to support clinical trial matching. The system can identify complex disease states (biochemical recurrence vs radiographic metastasis), castration status (sensitive, resistant) and genomics. Future work should focus on developing human-in-the-loop supervision and improving question decomposition to reduce errors in complex eligibility criteria. Screening performance of the proposed framework across different phases of trials. Trial Phase #Patient-Question Pairs #YES/NO labels Accuracy Precision Recall Specificity F1 score I, II, I/II 4,227 Yes: 1,735 No: 2,492 0.80 0.84 0.66 0.92 0.74 III 720 Yes: 299 No: 421 0.82 0.90 0.63 0.95 0.74 All 4,947 Yes: 2,034 No: 2,913 0.81 0.85 0.65 0.92 0.74 #YES/NO labels: The number of patient-question pairs that are labeled YES/NO among total patient-question pairs of the same row.