OBJECTIVE: Bacterial resistance is known to diminish the effectiveness of antibiotics for treatment of urinary tract infections. Review of recent healthcare and antibiotic exposures, as well as prior culture results is recommended to aid in selection of empirical treatment. However, the optimal approach for assessing these data is unclear. We utilized data from the Veterans Health Administration to evaluate relationships between culture and treatment history and the subsequent probability of antibiotic-resistant bacteria identified in urine cultures to further guide clinicians in understanding these risk factors. METHODS: Using the XGBoost algorithm, a retrospective cohort of outpatients with urine culture results and antibiotic prescriptions from 2017 to 2022 was used to develop models for predicting antibiotic resistance for three classes of antibiotics: cephalosporins, fluoroquinolones, and trimethoprim/sulfamethoxazole (TMP/SMX) obtained from urine cultures. Model performance was assessed using Area Under the Receiver Operating Characteristic curve (AUC) and Precision-Recall AUC (PRAUC). RESULTS: There were 392,647 prior urine cultures identified in 214,656 patients. A history of bacterial resistance to the specific treatment was the most important predictor of subsequent resistance for positive cultures, followed by a history of specific antibiotic exposure. The models performed better than previously established risk factors alone, especially for fluoroquinolone resistance, with an AUC of .84 and PRAUC of .70. Notably, the models performance improved markedly (AUC = .90, PRAUC = .87) when applied to cultures from patients with a known history of resistance to any of the antibiotic classes. CONCLUSION: These predictive models demonstrate potential in guiding antibiotic prescription and improving infection management.

Immunohistochemistry (IHC) is used to guide treatment decisions in multiple cancer types. For treatment with checkpoint inhibitors, programmed death ligand 1 (PD-L1) IHC is used as a companion diagnostic. However, the scoring of PD-L1 is complicated by its expression in cancer and immune cells. Separation of cancer and noncancer regions is needed to calculate tumor proportion scores (TPS) of PD-L1, which is based on the percentage of PD-L1-positive cancer cells. Evaluation of PD-L1 expression requires highly experienced pathologists and is often challenging and time-consuming. Here, we used a multi-institutional cohort of 77 lung cancer cases stained centrally with the PD-L1 22C3 clone. We developed a 4-step pipeline for measuring TPS that includes the coregistration of hematoxylin and eosin, PD-L1, and negative control (NC) digital slides for exclusion of necrosis, segmentation of cancer regions, and quantification of PD-L1+ cells. As cancer segmentation is a challenging step for TPS generation, we trained DeepLab V3 in the Visiopharm software package to outline cancer regions in PD-L1 and NC images and evaluated the model performance by mean intersection over union (mIoU) against manual outlines. Only 14 cases were required to accomplish a mIoU of 0.82 for cancer segmentation in hematoxylin-stained NC cases. For PD-L1-stained slides, a model trained on PD-L1 tiles augmented by registered NC tiles achieved a mIoU of 0.79. In segmented cancer regions from whole slide images, the digital TPS achieved an accuracy of 75% against the manual TPS scores from the pathology report. Major reasons for algorithmic inaccuracies include the inclusion of immune cells in cancer outlines and poor nuclear segmentation of cancer cells. Our transparent and stepwise approach and performance metrics can be applied to any IHC assay to provide pathologists with important insights on when to apply and how to evaluate commercial automated IHC scoring systems.

OBJECTIVE: Many patients with melanoma treated with immune checkpoint inhibitors (ICIs) do not derive response. Preclinical and retrospective studies identified that inhibition of the cyclooxygenase (COX) pathway may improve response to ICI treatment. METHODS: This prospective single site phase II trial accrued patients with advanced/metastatic melanoma. Participants underwent high-dose aspirin daily combined with pembrolizumab and ipilimumab every 3 weeks for 4 cycles followed by high-dose aspirin and pembrolizumab monotherapy. The primary endpoint was objective response rate (ORR). Longitudinal sampling of blood was performed to assess peripheral immune correlates. RESULTS: Twenty-seven subjects were enrolled with median follow-up of 32 months. An ORR of 62.9% was reached prior to discontinuation due to low likelihood of achieving the pre-specified ORR of 80%. 17 patients (63%) experienced a treatment-related adverse event (TRAEs) grade 3 or higher. A per-protocol analysis showed that patients able to continue aspirin alongside ICI through the induction period experienced significant survival benefit. Ten cytokines and increased regulatory T cells in the periphery correlated with beneficial response. CONCLUSIONS: The addition of high-dose aspirin to combination ICI within this study results in response comparable to ICI alone. Future clinical studies of COX inhibition will need to focus on mitigation of AEs to establish the clinical utility of this combination.