Accuracy of saliva tests for SARS-CoV-2 diagnosis during the pandemic in Rio de Janeiro, Brazil

Saliva is a promising non-invasive alternative to nasopharyngeal swabs (NPS) for SARS-CoV-2 diagnosis, yet its longitudinal performance remains underexplored. This longitudinal study evaluated saliva's diagnostic accuracy against NPS in 72 symptomatic individuals across six visits (July 2021-May 2022), analyzing 285 paired RT-qPCR samples in Rio de Janeiro, Brazil. Using NPS as the reference standard, saliva demonstrated high specificity (96.6%; 95% CI 92.9-98.7%), substantial agreement (91.6%; κ = 0.78; 95% CI 70-86% ; P < 0.001), and variable sensitivity (69.2% overall; 95% CI 57.2-79.5%), ranging from 40% during mid-phase infection (visit 3) to 82% during early infection (visit 1). Cycle threshold (Ct) values revealed slightly higher viral loads in NPS (mean Ct = 26.75) than in saliva (mean Ct = 28.75), with a mean difference of 0.79 cycles. Discordant results (8.4%) revealed saliva's utility in detecting late-stage infections missed by NPS. No significant associations were found between diagnostic agreement and participant characteristics or Ct values. In conclusion, our longitudinal data demonstrate that saliva testing achieves 69.2% sensitivity (ranging from 40 to 82% across infection phases), 96.6% specificity, and 91.6% overall agreement with NPS. The 8.4% discordant results included 1.7% of cases where saliva-detected infections were missed by NPS, highlighting its complementary value in late-stage monitoring. These results, coupled with saliva's stability over time, support its implementation as a scalable, cost-effective diagnostic tool-particularly in resource-limited settings-where its high specificity makes it particularly valuable for rule-out testing despite temporal variations in sensitivity.