Objectives: Several studies suggest the sensitivity of chest computed tomography (CT) is far greater than that of reverse transcription polymerase chain reaction (RT-PCR) in diagnosing COVID-19 patients, and therefore, CT should be included as a primary diagnostic tool. This systematic review aims to stratify studies as high or low risk of bias to determine the true sensitivity of CT for severe acute respiratory syndrome coronavirus-2 infection according to the unbiased (low risk) studies, a topic of particular importance given the insufficient quantity of RT-PCR kits in many countries. We focus on sensitivity as that is the chief advantage perceived of CT.
Materials and methods: This systematic review involved searching the PubMed and Google Scholar databases for articles conducted and published between January 1 and April 15, 2020. The quality assessment tool QUADAS-2 was used to stratify studies according to their risk of bias, and exclusion criteria included not providing the information deemed relevant for such a stratification, such as not indicating if the patients were symptomatic or asymptomatic, or identifying the source of the specimen for the reference standard, RT-PCR (eg, nasal, oropharyngeal, etc). Sensitivity values were then extracted, and random effects meta-analyses were performed.
Results: Of 641 search results, 37 studies (n = 9610 patients) were included in the analysis. The mean sensitivity of RT-PCR for COVID-19 reported by the biased studies was 70% (n = 5409/7 studies; 95% confidence interval [CI], 43-97; I = 99.1%), compared with 78% by unbiased studies (n = 534/4 studies; 95% CI, 69-87, I = 89.9%). For chest CT, the mean sensitivity reported by biased studies was 94% (n = 3371 patients/24 studies; 95% CI, 92-96; I = 93.1%), compared with 75% by unbiased studies (n = 957/10 studies; 95% CI, 67-83; I = 89.5%).
Conclusions: The difference between the sensitivities of CT and RT-PCR for severe acute respiratory syndrome coronavirus-2 infection is lower than previously thought, as after stratifying the studies, the true sensitivity for CT based on the unbiased studies is limited.