Prognostic role of COVID-19 pneumonia signs and other CT-biomarkers for survival in patients with malignant neoplasms: the ARILUS project

BackgroundClinically manifested pneumonia associated with COVID-19 infection in cancer patients has been associated with worse prognosis. The prognostic significance of subclinical manifestations of pulmonary infiltration is poorly understood. AimsTo estimate the survival of cancer patients with signs of asymptomatic pneumonia detected by a multitarget artificial intelligence (AI) algorithm on chest computed tomography (CT) during the COVID-19 pandemic, and to evaluate factors associated with the risk of death in this population. Materials and methodsThis population-based cohort study, conducted within the ARILUS project, included 1,147 examinations of cancer patients performed between August 2020 and May 2021. Signs of pneumonia on CT were detected by the AI algorithm in 556 (48.4%) patients with malignant neoplasms (MN) who had no clinical manifestations of infection at the time of the examination. Overall survival (OS) was assessed using the life table and Kaplan-Meier methods. Multivariate Cox proportional hazards regression analysis was used to identify independent predictors of death and to assess the influence of potential confounding factors. ResultsDuring the follow-up period, 680 deaths were recorded among cancer patients, of which 575 deaths were due to MN progression. Three-year OS differed significantly between the groups: in patients without signs of pulmonary infiltration on CT, it was 60.7% (95% CI 56.6%- 64.7%), whereas in patients with AI-detected signs of pneumonia, it was 45.6% (95% CI 41.3%- 49.8%) (p<0.001). In the multivariable regression analysis, independent factors associated with an unfavorable prognosis (death) were: presence of pulmonary infiltration (adjusted hazard ratio (HR) 1.31; 95% CI 1.09-1.58), male sex (HR 1.25; 95% CI 1.00-1.57), and MN stage (HR 1.78 for stage II, HR 2.93 for stage III, and HR 4.74 for stage IV compared to stage I). Signs of pulmonary emphysema (HR 1.84), aortic aneurysm (HR 1.75), and coronary artery calcification (HR 1.22-1.34), which were significantly associated with the risk of death in the univariable analysis, lost their prognostic significance in the multivariable model after adjustment for other variables. ConclusionsAsymptomatic COVID-19 pneumonia detected by AI on CT is an independent predictor of reduced overall survival in MN patients. Automated AI screening for such changes may be recommended in routine practice.