ABSTRACT
OBJECTIVES: We sought to identify the predictors of delayed viral clearance in patients with cancer with asymptomatic COVID-19 when the SARS-CoV-2 Omicron variants prevailed in Hong Kong. METHODS: All patients with cancer who were attending radiation therapy for head and neck malignancies or systemic anticancer therapy saved their deep throat saliva or nasopharyngeal swabs at least twice weekly for SARS-CoV-2 screening between January 1 and April 30, 2022. The multivariate analyses identified predictors of delayed viral clearance (or slow recovery), defined as >21 days for the cycle threshold values rising to ≥30 or undetectable in two consecutive samples saved within 72 hours. Three machine learning algorithms evaluated the prediction performance of the predictors. RESULTS: A total of 200 (15%) of 1309 patients tested positive for SARS-CoV-2. Age >65 years (P = 0.036), male sex (P = 0.003), high Charlson comorbidity index (P = 0.042), lung cancer (P = 0.018), immune checkpoint inhibitor (P = 0.036), and receipt of one or no dose of COVID-19 vaccine (P = 0.003) were significant predictors. The three machine learning algorithms revealed that the mean ± SD area-under-the-curve values predicting delayed viral clearance with the cut-off cycle threshold value ≥30 was 0.72 ± 0.11. CONCLUSION: We identified subgroups with delayed viral clearance that may benefit from targeted interventions.
