A systematic review and meta-analysis comparing the diagnostic accuracy of initial RT-PCR and CT scan in suspected COVID-19 patients

Abstract

Objective: To perform a systematic review and meta-analysis to compare the diagnostic accuracy of CT and initial reverse transcriptase polymerase chain reaction (RT-PCR) for detecting COVID-19 infection. Methods: We searched three databases, PubMed, EMBASE, and EMCARE, to identify studies reporting diagnostic accuracy of both CT and RT-PCR in detecting COVID-19 infection between December 2019 and May 2020. For accurate comparison, only those studies that had patients undergoing both CT and RT-PCR were included. Pooled diagnostic accuracy of both the tests was calculated by using a bivariate random effects model. Results: Based on inclusion criteria, only 11 studies consisting of 1834 patients were included in the final analysis that reported diagnostic accuracy of both CT and RT-PCR, in the same set of patients. Sensitivity estimates for CT scan ranged from 0.69 to 1.00 and for RT-PCR varied ranging from 0.47 to 1.00. The pooled estimates of sensitivity for CT and RT-PCR were 0.91 [95% CI (0.84–0.97)] and 0.84 [95% CI (0.71–0.94)], respectively. On subgroup analysis, pooled sensitivity of CT and RT-PCR was 0.95 [95% CI (0.88–0.98)] and 0.91 [95% CI (0.80–0.96), p = o.ooo1]. The pooled specificity of CT and RT-PCR was 0.31 [95% CI (0.035–0.84)] and 1.00 [95% CI (0.96–1.00)]. Conclusion: CT is more sensitive than RT-PCR in detecting COVID-19 infection, but has a very low specificity. Advances in knowledge: Since the results of a CT scan are available quickly, it can be used as an adjunctive initial diagnostic test for patients with a history of positive contact or epidemiological history.