Background: little is known about the incidence and characteristics of multisystem immune-related adverse events (irAEs) associated with dual-agent ipilimumab and nivolumab use. Methods: A retrospective cohort review was completed that included cancer patients seen at the Juravinski Cancer Centre who received at least one dose of ipilimumab and nivolumab from 2018 to 2022. Patient characteristics, cancer types, and irAEs were recorded. Multivariate logistic and cox regressions were completed, comparing those who developed multisystem irAEs, single irAE, and no irAE. Results: A total of 123 patients were included in this study. Out of 123 patients, 72 (59%) had melanoma, 50/123 (41%) had renal cell carcinoma (RCC), and 1/123 (1%) had breast cancer. Multisystem irAEs were seen in 40% of the overall cohort. The most common irAE type was dermatitis (22%), followed by colitis (19%) and hepatitis (17%). Conclusions: Our study demonstrated that multisystem irAEs are prevalent amongst patients receiving ipilimumab and nivolumab. It is important for both physician education and the counseling and consent of patients to monitor the potential for multiple irAEs.

Manual extraction of real-world clinical data for research can be time-consuming and prone to error. We assessed the feasibility of using natural language processing (NLP), an AI technique, to automate data extraction for patients with advanced lung cancer (aLC). We assessed the external validity of our NLP-extracted data by comparing our findings to those reported in the literature. Patients diagnosed with stage IIIB or IV lung cancer between January 2015 to December 2017 at Princess Margaret Cancer Centre who received at least one dose of systemic therapy were included. Their electronic health records were provided to Pentavere's NLP platform, DARWEN^TM, in March 2019. Descriptive statistics summarized baseline patient and cancer characteristics, molecular biomarkers, and first-line systemic therapies. Cox multivariate models were used to evaluate prognostic factors for advanced non-small cell lung cancer (NSCLC) and small-cell lung cancer (SCLC) cohort. NLP extracted clinical information (n = 333 patients) in a total of 8 hours, with only a few missing data for smoking status (n = 2), and Eastern Cooperative Oncology Group (ECOG) status (n = 5). Baseline patient and cancer characteristics summarized from NLP-extracted data were comparable to those in previous studies and population reports. For NSCLC patients, being male (HR 1.44, 95 % CI [1.04, 2.00]), having worse ECOG (1.48 [1.22, 1.81]), and having liver (2.24 [1.45, 3.46]), bone (2.09 [1.48, 2.96]), or lung metastases (2.54 [1.05, 2.26]) were associated with worse survival outcomes. For SCLC patients, having older age (HR 1.70 per 10 years, 95 % CI [1.10, 2.63]) and liver metastases (3.81 [1.61, 9.01]) were associated with worse survival outcomes. Our study demonstrated that automated data extraction using NLP is feasible and time efficient. Additionally, the NLP-extracted data can be used to identify valid and useful clinical endpoints for research. NLP holds significant potential to accelerate the extraction of real-world data for future observational studies.

PURPOSE: Cancer and its treatment cause symptoms. In this study, we aimed to develop machine learning (ML) systems that predict future symptom deterioration among people receiving treatment for cancer and then validate the systems in a simulated deployment across an entire health care system. METHODS: We trained and tested ML systems that predict a deterioration in nine patient-reported symptoms within 30 days after treatments for aerodigestive cancers, using internal electronic health record (EHR) data at Princess Margaret Cancer Centre (3,229 patients; 20,267 treatments). The primary performance metric was the area under the receiver operating characteristic curve (AUROC). The best-performing systems in the held-out internal test set were then externally validated across 82 cancer centers in Ontario (12,079 patients; 77,003 treatments) by adapting techniques from meta-analysis. RESULTS: The best ML systems predicted symptom deterioration with AUROCs ranging from 0.66 (95% CI, 0.63 to 0.69) for dyspnea to 0.73 (95% CI, 0.71 to 0.75) for drowsiness in the internal test cohort. Treatments flagged as high-risk were significantly associated with future symptom deterioration (odds ratios [ORs], 2.53-6.56; all P < .001) and emergency department visits for dyspnea (OR, 1.85; P = .008), depression (OR, 1.84; P = .04), and anxiety (OR, 2.66; P < .001). In the external validation cohort, the AUROCs for different symptoms meta-analyzed across centers ranged from 0.67 (95% CI, 0.66 to 0.68) to 0.73 (95% CI, 0.72 to 0.74). Performance across centers displayed significant heterogeneity for six of nine symptoms (I ² , 46.4%-66.9%; P = .004 for dyspnea, P < .001 for the rest). CONCLUSION: ML can predict future symptoms among people with cancer from routine EHR data, which could guide personalized interventions. Heterogeneous performance across centers must be considered when systems are deployed across a health care system.