Letter to the Editor: “Chest CT for detecting COVID-19: a systematic review and meta-analysis of diagnostic accuracy”
by Nicola Panvini, Davide Bellini, Iacopo Carbone (firstname.lastname@example.org)Chest CT for detecting COVID-19: a systematic review and meta-analysis of diagnostic accuracy
We have read with great interest the article entitled “Chest CT for detecting COVID-19: a systematic review and meta-analysis of diagnostic accuracy” by Dr. Buyun Xu and colleagues recently published in European Radiology .
Since the debut of the novel Corona Virus Disease-2019 (COVID-19) pandemic, the role of CT in the disease screening and diagnosing has been widely debated. The first studies coming from China during the early stage of the pandemic reported excellent CT sensitivity for the COVID-19, suggesting that CT could play a pivotal role in the rapid detection of infected patients and limiting the spread of the virus [2–4]. On the other hand, the low specificity and positive predicting value observed, also taking into account the concerns about the radiation exposure and the possible contamination of the scanners, have raised criticisms regarding the use of CT as a screening tool, especially in asymptomatic patients [5–7]. Xu and coworkers did an extraordinary effort to provide, in a short time since the first signs of COVID-19 outbreak, cumulative evidence-based data on diagnostic yield of CT in this setting. However, we do believe that some concerns should be raised.
In their systematic review, Authors were able to include only two studies to summarize CT specificity for COVID-19, reporting a specificity of 25% (95% CI 22-30%)  and 33% (95% CI 23-44%) , respectively. However, Xu and colleagues  considered the whole study population from the Ai et al study (1014 participants) , even though multiple RT-PCR to confirm the initial results were available only for 258 patients. Furthermore, sufficient data to create a 2×2 contingency table were not available from the manuscript even for this subset of patients.
Considering the growing number of evidences suggesting a low sensitivity of the first nasopharyngeal swab for the diagnosis of SARS-CoV-2 infection [3, 9], we strongly believe that accepting the result of a single RT-PCR testing as a solid reference standard to define a patient negative for Sars-CoV-2 infection, is not the correct approach. The strict and proper selection of papers to be included is crucial for a systematic review and meta-analysis. In this case, all the metrics of diagnostic accuracy computed, especially specificity, could be unpaired with the inclusion of a study with such a large cohort.
Furthermore, the study from Ai and colleagues  was deemed at low risk of bias in the index test domain because the Authors reported that the chest CT readers were blinded to the RT-PCR results. Nonetheless, in this study, chest CTs were interpreted as a dichotomous test, without a pre-specified threshold to for defining positive scans. This approach could have represented a considerable source of bias, leading to an underestimation of the real specificity of CT.
Indeed, the use of structured reporting systems for the interpretation of CT scans in patients with suspected COVID-19 has been recently advocated [10, 11]. We believe that these may represent important tools for improving the diagnostic yields of CT for COVID-19, providing preset thresholds for determining positive scans with higher specificity.