Background and objective Saliva has been proposed as a potential more convenient, cost-effective, and easier sample for diagnosing SARS-CoV-2 infections, but there is limited knowledge of the impact of saliva volumes and stages of infection on its sensitivity and specificity.

            Methods
            In this study, we assessed the performance of SARS-CoV-2 testing in 171 saliva samples from 52 mostly mildly symptomatic patients (aged 18 to 70 years) with a positive reference standard result at screening. The samples were collected at different volumes (50, 100, 300, and 500 µl of saliva) and at different stages of the disease (at enrollment, day 7, 14, and 28 post SARS-CoV-2 diagnosis). Imperfect nasopharyngeal (NP) swab nucleic acid amplification testing was used as a reference. We used a logistic regression with generalized estimating equations to estimate sensitivity, specificity, PPV, and NPV, accounting for the correlation between repeated observations.
          
            Results
            The sensitivity and specificity values were consistent across saliva volumes. The sensitivity of saliva samples ranged from 70.2% (95% CI, 49.3–85.0%) for 100 μl to 81.0% (95% CI, 51.9–94.4%) for 300 μl of saliva collected. The specificity values ranged between 75.8% (95% CI, 55.0–88.9%) for 50 μl and 78.8% (95% CI, 63.2–88.9%) for 100 μl saliva compared to NP swab samples. The overall percentage of positive results in NP swabs and saliva specimens remained comparable throughout the study visits. We observed no significant difference in cycle number values between saliva and NP swab specimens, irrespective of saliva volume tested.
          
            Conclusions
            The saliva collection offers a promising approach for population-based testing.