Project Summary/Abstract: People living with HIV (PLWH) are living longer because of effective antiretroviral therapy (ART). With longer life expectancy, PLWH are experiencing an increased incidence in non-AIDS defining malignancies.1 Lung cancer is the most common cause of cancer mortality and is one of the most commonly occurring cancers in this population. Lung cancer incidence is nearly three times higher in adults with HIV as compared to those without HIV.2 Screening for lung cancer with low-dose CT (LDCT) reduces mortality by at least 20%.3-6 The reduction in mortality may be even greater in PLWH, as they have been shown to have shortened survival following lung cancer diagnosis largely due to late-stage presentation.7 Currently PLWH are subject to the same screening eligibility criteria as those without HIV, based only on age and smoking history.8 While the prevalence of smoking is higher among PLWH, studies that have controlled for tobacco use still showed a significant increased risk for lung cancer among PLWH compared to adults without HIV.1 This could indicate that HIV modifies the risk for lung cancer independent of smoking intensity and duration. PLWH who are adherent to ART and continue to smoke are significantly more likely to die from lung cancer than from AIDS-related causes.9 The rapid advancements in machine learning with LDCT have significantly enhanced the accuracy of predicting lung cancer incidence and mortality using medical imaging. Our latest deep learning model has achieved AUC > 0.91 when evaluating overall cancer risk in nodules identified during screening. This assessment was conducted using thousands of LDCT scans from both in-house sources and the National Lung Screening Trial (NLST), encompassing data from over 20,000 patients. 10-12 Additionally, we have recently introduced fully automated body composition analysis into lung cancer risk prediction models, which has added value for patients as it can predict outcomes, specifically all-cause mortality.13 However, applying these models to PLWH may not yield the same level of accuracy due to the models being trained primarily on the general population. This discrepancy highlights the well-known issue of domain adaptation, where a model developed for one specific population might not perform as well when applied to a different group. In response to this challenge, our current study aims to assess the effectiveness of our previously validated AI models using real-world data from PLWH. Lung screening may be improved by incorporating blood-based biomarkers with imaging. There is a need for better and more accessible biomarkers to improve lung cancer detection. Biomarkers with multi-protein panels (MPPs) are emerging, but it is unknown if these will improve early detection or improve care. Before MPPs can be used at scale, we need to understand the sensitivity and specificity of these biomarkers, especially in patients who are at increased risk, for example, those living ...