Saliva sample for testing SARS-CoV-2 infection, 1st update on diagnostic accuracy

Key message

The aim of this review was to investigate the diagnostic accuracy of reverse transcription real time polymerase chain reaction (rRT-PCR) for detection of SARS-CoV-2 in saliva specimens compared to upper airway samples. We performed a systematic literature search focusing on COVID-19 and saliva. The search provided 501 titles and abstracts, from which we inspected 70 records in full text and included 23 studies. From the included studies, we extracted data from 27 sets of comparative analysis of in total 7065 paired samples to calculate sensitivity[1] and specificity[2].

• In five studies with a total of 4299 paired samples from screening a mainly asymptomatic population, we found that sensitivity in the analysis ranged from 61% to 100 % (GRADE certainty of the evidence[3] moderate, ⨁⨁⨁◯). The specificity, ranged from 95% to 100% (GRADE certainty of the evidence high, ⨁⨁⨁⨁).
• In 11 studies with a total of 1612 paired samples from screening mainly symptomatic persons, we found variable, but predominantly good concordance between saliva samples and upper airway samples. Due to risk of bias we have low confidence in estimates based on these studies (GRADE certainty of evidence low ⨁⨁◯◯).
• In 9 studies with a total of 1154 paired samples from re-testing patients with known Covid-19 status, we likewise found variable, but predominantly good concordance. Due to high risk of bias, we have very low confidence in estimates based on these studies(GRADE certainty of evidence very low ⨁◯◯◯).

The results indicate that the diagnostic accuracy of rRT-PCR on saliva samples, compared to upper airway samples, probably is somewhat lower regarding sensitivity, but high regarding specificity. Variations between the analysis may be related to publication bias (more positive results published), risk of bias in the included studies, technical and clinical factors. The studies do not allow conclusions to be made about what factor may have the largest impact on the results.

[1] Sensitivity is the likelihood that a person with a condition (diseased/infected) is correctly diagnosed using the index test compared to the reference test.

[2] Specificity is the likelihood that a person without a condition (healthy/non-infected) is correctly identified as negative by the index test compared to the reference test.

[3] GRADE certainty of evidence reflects our certainty that the estimate is close to a true estimate. The lower the certainty, the higher is our anticipation that new evidence may change the estimates.