RESUMO
BACKGROUND: Saliva real-time polymerase chain reaction (PCR) was shown to be sensitive and specific for the detection of congenital cytomegalovirus (cCMV) in universal screening studies. In the current study, we assessed the performance of saliva real-time PCR in newborns undergoing targeted cCMV screening. METHODS: Saliva real-time PCR results were prospectively correlated with reference-standard urine detection in newborns undergoing targeted cCMV screening over a 3-year period, in successive validation (concurrent testing of all saliva and urine specimens) and routine-screening (confirmatory urine testing of positive saliva results) implementation phases. RESULTS: The sensitivity, specificity, and positive and negative predictive values of saliva real-time PCR were 98.3% (95% confidence interval, 90.8%-99.9%), 91.5% (89.3%-93.3%), 45.6% (36.7%-54.7%), and 99.9% (99.2%-99.9%), respectively, in 856 concurrently tested newborns. True-positive saliva real-time PCR detection (defined in relation to urine detection) was associated with earlier saliva sampling (P = .002) and a higher saliva viral load (P < .001). We further identified a saliva viral load cutoff value that reliably distinguished between true-positive and false-positive saliva results. CONCLUSIONS: In newborns undergoing targeted screening for cCMV, saliva real-time PCR is highly sensitive yet has a low positive predictive value, necessitating confirmatory testing. Early sampling and application of a validated viral load cutoff could improve the assay performance and support its large-scale implementation in this growing clinical setting.