• COVID-19
• Clinical Epidemiology

The prototypical study of diagnostic test accuracy compares the results of one or more index tests to an independent ‘reference-standard’ determination of true disease status: ‘disease positive’ (D+) or ‘disease negative’ (D−). If the index test is binary—either ‘test positive’ (T+) or ‘test negative’ (T−)—then sensitivity is the proportion of D+ individuals who have T+ on the index test and specificity is the proportion of D− individuals who have T− on the index test. But what if the comparison is between two or more index tests and there is no reference standard? We cannot calculate sensitivity or specificity, but we can investigate disagreements between the index tests. This is especially valuable when there are three index tests to compare.

In their Thorax paper, Todsen et al addressed the question of how best to obtain a sample to test for SARS-CoV-2 infection using reverse transcriptase PCR (RT-PCR).1 They compared three sample types: (1) nasopharyngeal swab (NPS), (2) oropharyngeal swab (OPS) and (3) saliva. The samples were obtained in randomised sequence from 26 795 adults at two public COVID-19 test centres in Copenhagen between January and May 2021 …