key: cord-0733172-wertj5dz
authors: Iwata, Kentaro; Yoshimura, Kenichi
title: A concern regarding estimated sensitivities and specificities of nasopharyngeal and saliva specimens for SARS-CoV-2 infection
date: 2020-10-26
journal: Clin Infect Dis
DOI: 10.1093/cid/ciaa1655
sha: d636aac0fc0fae91ff60435d7bce551ee3e47b7b
doc_id: 733172
cord_uid: wertj5dz

nan

A c c e p t e d M a n u s c r i p t Dear Editor.

We read with interest a study by Yokota et al estimating sensitivities and specificities of both nasopharyngeal specimens (NPS) and saliva for diagnosing SARS-CoV-2 infection [1] . This report can bring important findings based on large-scale comparative data with qualitycontrolled methods for specimen collection and diagnostics. However, we have concerns regarding the use of Bayesian latent class model for this purpose. With use of 2 sets of diagnostic tests with 2 x 2 tables for tests positivity, the current available diagnostic tests had 6 degrees of freedom. Since unmeasured latent classes estimate both sensitivity and specificity of NSP and saliva specimens, they also have 4 degrees of freedom, with additional 2 degrees of freedom for prior distributions, or prevalence for 2 cohorts. In addition, it is clinically possible and even reasonable to surmise that NPS and saliva specimens are interrelated and not independent, and this adds additional parameter for correlation, making estimation of posterior probability not feasible [2] . We therefore consider that estimated sensitivities and specificities by Yokota et al have impaired credibility. It is essential to explain the method and its limitation in order to interpret it as evidence that can be directly used in daily practice.

Mass screening of asymptomatic persons for SARS-CoV-2 using saliva

Latent Class Models in Diagnostic Studies When There is No Reference Standard-A Systematic Review

A c c e p t e d M a n u s c r i p t