The exact place of PCR and chest CT in screening, staging, and diagnosis of Covid-19: A meta-analysis

Fatemeh Khatami Urology research Center, Tehran University of Medical Sciences, Tehran, Iran Mohammad Saatchi Department of Epidemiology and Biostatistics, School of Public Health, Tehran University of Medical Sciences, Tehran, Iran Seyed Saeed Tamehri Zadeh Urology research Center, Tehran University of Medical Sciences, Tehran, Iran Zahra Sadat Aghamir Faculty of Dentistry, Tehran University of Medical Sciences, Tehran, Iran Alireza Namazi Shabestari Department of Geriatric Medicine, School of Medicine, Tehran University of Medical Sciences, Tehran, Iran Seyed Mohammad Kazem Aghamir (  mkaghamir@tums.ac.ir ) Urology research Center, Tehran University of Medical Sciences, Tehran, Iran


Introduction
In late December of 2019, a cluster of unknown viral pneumonia patients, who later tuned out to be infected with novel coronavirus (nCoV) (the disease named COVID- 19), has been reported in Wuhan City, Hubei Province, China (1). Until 11 March, over 100 countries and 120000 cases have been involved with this beta coronavirus of whom over 4000 died and its incidence is at increasing pace worldwide. O cially, World Health Organization has declared the pandemic of nCoV, therefore, it has attracted an international interest (2).
Currently, due to non-existence of therapeutic vaccine or certain antiviral drugs, both detecting patient at an early stage and immediate patient isolation play mandatory role in ghting against nCoV(3). Although chest computed tomography (CT) scan has shown promising results regarding early diagnosis of COVID-19, reverse transcription polymerase chain reaction (RT-PCR) remain to be the mainstay of this viral disease's diagnosis given the sixth version of diagnosis and treatment of COVID- 19 (3, 4); however, RT-PCR poses several limitations that reduce the sensitivity of this method and cause some patients to be missed (5). Actually CT and PCR are not comparable in the diagnosis of COVID-19 because CT is valuable tool in staging of disease and checking the treatment e cacy but repeating PCR test is the diagnostic tool. Recent studies have focused on chest CT of patients with COVID-19 features and they claim that initial chest CT may enable to detect the disease with higher sensitivity in comparison to RT-PCR (6). Therefore, this systematic review and meta-analysis was performed to determine diagnostic accuracy of initial chest CT scan compare to RT-PCR in patients who are infected with nCoV.

Inclusion criteria
All stages of the study followed the PRISMA guideline. Inclusion criteria were having an observational epidemiological study design, clear report of the number of positive cases by PCR and chest CT, and ability to calculate accuracy indicators.. Articles that did not meet one or more of the study entry criteria were also excluded. All relevant English, Chinese and other language case-series, cross-sectional, and cohort studies were selected. Case reports were excluded and for the remaining literatures scienti c validity was performed.

Search strategy
The study was run by searching all relevant literature from three main databases: MEDLINE (PubMed), Scopus, and EMBASE from January 1st , 2019, to the 27th march 2020 with words the key grouping of "COVID19 virus", "2019 novel coronavirus", "Wuhan coronavirus", "2019-nCoV", "X-Ray Computed Tomography", "Polymerase Chain Reaction", "Reverse Transcriptase PCR", and "PCR Reverse Transcriptase" (Supplementary le 1). In order to obtain more articles and to ensure that the appropriate search is performed in the databases, the list of selected articles was also reviewed. The variables extracted from each targeted article in included the name of the rst author, the year of publication of the article, the country and city of the study, the average age of the subjects, the gender, the study design, total sample size, positive tst result, negative test result, false positive, and false negative.

Data Extraction and Statistical Analysis
To reduce the selection bias, two researchers (SZA and SSTZ) screened articles separately through checking titles and abstracts. Unlikeness and disagreement were removed by the idea of the third ones (FKH). The full text of whole candidate related articles were studied in order to grouping into the included one or excluded one. Actually, included ones had data of con rmed COVID-19 patients by chest CT scan or quantitative real time polymerase chain reaction (RT-PCR). The quality assessment was performed by the Newcastle-Ottawa Scale (NOS) assessment tool. The papers that receive score more than 6 were re ected as the "high quality" one and go to the additional meta-analysis step.
The outcomes of interest, including the diagnostic operation of CT-scan to identify covid-19 by the random effect model, were summary receiver operating characteristic (SROC) curve sensitivity and speci city. Sensitivity, indicating the capacity of index test to identify patients, considered by "Sensitivity = TP/ (TP + FN)". Speci city as the examination to remove disease-free, calculated by "Speci city = TN/ (FP + TN)". Positive Predictive Value (PPV) is probability of disease if the test is positive calculated by "Positive predictive Value = TP/ (FP + TP)". Negative Predictive Value (NPV) is probability of disease-free if the test is negative calculated by "negative predictive value = TN/ (FN + TN)" The Cochran's Q-test of heterogeneity at an importance amount of 5% was used to evaluate statistical heterogeneity of the researches and I2 was practical for quantitative evaluation of heterogeneity between outcomes based on the Higgins classi cation in which an I2 value more than 75% is pinpointing of heterogeneity. Deeks' funnel plot was used to evaluate publication bias by the "Metafunnel". Brie y, to create the funnel plot, the odds ratio was rst calculated using the equations of (TP / FN)/(FP / TN) and after estimating the odds ratio logarithm, the standard error of odds ratio was calculated. Extracted data were collected in Excel 2007 (Microsoft Corporation, Redmond, CA) and analysis was done by using STATA v.14.0SE (College Station, TX, USA) and RevMan 5.

Results
In the present study, at rst 668 (667+1) articles were identi ed through the search of three data bases of Pubmed, Embase, and Scopus. After deleting 188 duplicate articles, the screening was done based on the titles and abstracts which lead to the 89 articles were candidate as the related ones ( gure 1). After that scoring of the articles were done based on NOS screening tool and the number of article were limited to the 47.
Finally, 47 articles with the total sample size of 4283 recruited for the nal analysis. All studies were published in the rst quarter of 2020. A summary of the information in these articles is shown in Table 1.
With the exception of one article from the United States, all articles were from China. According to the articles, the types of studies were 23 case-series, 14 cohorts and 10 cross sectional. The forest plot result is shown in gure 2 and the quantity of False Positives (FP), False Negative (FN) and True Positives (TP), and True Negative (TN) is shown more than sensitivity, speci city, and 95% con dence Intervals (CI) of each study. Sensitivity ranged from 25% to 100% and the speci city was estimated to vary from 19% to 70%. One of the important assumptions in calculating sensitivity and speci city using the Metaprop command is to extract studies that have reported 100% sensitivity or speci city. Therefore, based on this approach and by analyzing 26 articles that reported less than 100% sensitivity, summary sensitivity of CT compared to PCR 86% (95% CI: 83%-90%) and summary speci city of CT based on six articles 43 % (95% CI: 26% -60%). sensitivity is ranging from 25% to 97% and speci city is 25% to 70% in included articles ( gure 3-4). The PPV of CT was 67% (95% CI: 57% -78%) and the negative predictive value was 84% (95% CI: 74% -95%). Variation of estimated numbers is 33% to 97% and 33% to 95% for PPV and NPV respectively ( gures 5,6). The symmetry between the two sides of the funnel plot regression line indicates that the publication is not biased in these articles However, due to the large number of zeros in the FP and TN cells, it was possible to calculate the odds ratio for six studies, and the interpretation of this plot in our study should be done with caution ( gure 7).
The  The mean percentage of primarily con rmed patients divided total con rmed patients in the articles with more than 10 total con rmed patients (the rst 5 articles included), is 62.7%. This means among these 5 articles, RT-PCR test could diagnose 62.7% of the COVID-19 infected patients in the rst test, among all con rmed patients a nd about 37.3% of infected patients couldn't be recognized in the rst place by RT-PCR test. Among these 5 articles, 4 included the exact information of further tests (number 1,2,3 and 5). In these for articles the mean percentage of secondary, thirdly and fourthly con rmed patients divided total con rmed patients is 21.2%, 6.3%, 0.7% retrospectively. Moreover, the percentage of the patients who were not diagnosed after 4 times of repeating test was 8.6% (in the previous 4 articles). The number and sequence of primers and probes could be in uential on PCR sensitivity and speci city which were surveyed in table 3. As we can see in table 3, in the case report by Wendong Hao (57), consuming one pair of primer and probe did not indicate positive result at rst but in the fourth time. On the other hand, Xavier Marchand-Senécal et al. reported a case that was diagnosed initially with one pair primer and probe PCR test. By using 2 pairs of primer and probe, the mean of initially detecting patients divided total con rmed patients, was 65.5%. In addition, Qing Chen (53) ndings implied utilization of 3 pair of primer and probe caused 100% of initially discovering COVID-19 patients (53). Based on limited data that has been found, totally, it seems that the more number of primer and probe have been used, the more initially detected patients have been achieved.
Although more speci c information is needed and demanded from future researchers to determine the exact desired conclusion.

Discussion
Considering the outcomes of RT-PCR as reference, in our meta-analysis, the sensitivity and speci city of initial chest CT scan for detecting patients, who were highly suspicious for nCoV, were 86% and 43% respectively. The positive predictive value and negative predictive value of CT scan were 67% and 84% respectively. In a boarder sense, 67% of individuals with positive chest CT scan had positive RT-PCR and 84% of individuals with negative chest CT scan had negative RT-PCR. Despite numerous studies have been conducted on various subjects of nCoV, the early diagnosis of patients who were infected with nCoV and also, suitable quarantine of them are under the debate. In accordance with previous studies, chest CT scan when it compares to RT-PCR, may have a variety of bene cial features including, but not limited to, less time-consuming and more availability, that makes doctors to take it into consideration as an adjuvant diagnostic tool (61,62).
Tao Ai and his colleagues performed a study on 1014 patients and 888 (88%) of them had positive chest CT scan and 601 patients had positive RT-PCR for nCoV, and the sensitivity, speci city, PPV, and NPV of CT scan were about 97%, 25%, 65%, and 83% respectively that compare to our result sensitivity and speci city were much different. The possible reason behind relatively high sensitivity and low speci city in this study is the odds ratio of positive RT-PCR was low, which was considered as the reference (61).
There have been a number of patients with typical chest CT scan ndings and symptoms for nCoV that their initial RT-PCR results were negative according to previous literatures (4,18,61). For instance, following Y Fang et.al., study that showed that the result of initial RT-PCR in 15 out of 51 patients were negative while their chest CT scan were positive (4,47,61). Hence, negative result of RT-PCR is not able to roll out nCoV and it is very important to pay attention to chest CT scan, epidemiologic features, and clinical symptoms.
Diagnostic approach with favorably sensitivity and speci city is needed to control the rapid distribution of this disease. Therefore RT-PCR is suggested as a method for diagnosis of SARS-CoV-2 by World Health Organization (WHO) (63). The swabs are commonly taken from sputum, nasopharyngeal, pharyngeal and throat which were included in our meta-analysis. Although WHO recommends examining stool specimens, together with nasopharyngeal swabs (63). Furthermore, combination of humoral and cellular immunity, for instance IgG-IgM antibody, alongside RT-PCR could possibly re ne the detection of COVID-19 (25,68).
According to Chen W., and his colleagues research, positive results based on blood samples, represent deterioration of the disease. Additionally, positive results of anal swabs are related to the severe phase of the disease and the RNA load was higher in anal swabs indicates that the virus may replicate in the digestive system (53). Stool swabs were positive in 8 out of 9 con rmed patients but urine and blood samples were negative in all (8). Also, plasma and urine samples caused negative result in the study by Jasper Chan et al, as well (15). A prospective case series including 30 con rmed COVID-19 patients (21 common type, 9 severe type and one of them had conjunctivitis) underwent RT-PCR test for both tear and conjunctival secretions specimens. Surprisingly, the results of the two specimens were positive the only conjunctivitis patient and the other 58 samples were negative. Therefore, tear and conjunctival secretions samples could possibly yield result for RT-PCR test in COVID-19 infected patients with conjunctivitis (23). sensitivity. Moreover, RdRp/Hel analysis did not cross reacted with any human coronaviruses or other respiratory pathogens while RdRp-P2 analysis reacted to SARS-CoV, either (64). Another study expressed that the sensitivity of N gene assay in nding the positive samples, is 10 times more than ORF-1b gene assay (65).
In February, 280 suspected patients with clinical manifestations were enrolled in a hospital in Marseille, RT-PCR assay was conducted. About 51% of the patients were negative for all the common viruses, additionally none of the patients were positive for SARS-COV-2 (66). As reported by  out of 91 suspected patients, were con rmed at least one time by RT-PCR assay in 2 sequential tests with 24 hour interval, and 3 patients with negative outcome reported as con rmed due to clinical manifestation (17). In the study by Tao (36). On the report of Xingzhi Xie (N.2), from total 167 patients, 162 patients were con rmed by the rst test (7 of them had normal CT), while 4 patients turned positive after the second and third test and the other patients became positive after multiple tests. Whereas all these patients had clinical symptoms in addition to GGO pattern on their CT scans (47). In the third study, more than two-thirds of the patients had abnormal CT scans (22). In a retrospective analysis(N.5), in the group of 51 patients, 98% (50/51) had abnormal CT, besides 70.6% (36/51) had positive PCR assay, initially, about 30% of the tests became positive after second, third and fourth scanning (4). from the report of a family cluster (N.7), 4 patients were con rmed by RT-PCR, one patients with negative PCR had GGO patterns in her CT and the other two with the same test result, had mild cough and fever for few days (54). On the report of 5 children (N.9), 4 of them had positive PCR outcome within rst assay but one with CT ndings turned positive after six times of examining (41).
As speci ed by Li Ni (N.10), from 3 suspected patients, 2 patients showed negative PCR result after 3 repeated tests in 11 days and nally a rmed to be infected by typical CT features and clinical manifestations (55). On a report of 2 children, one of them had positive and the other had negative RT-PCR result which became positive in the day after, and at the same time, both had CT symptoms (56). In a case report represented a 56 years old man (N.12), RT-PCR assay was negative for 3 times and the fourth time became positive while CT demonstrated typical features such as ground-glass opaci cation (GGO) (57). Another case report(N.13) presented a patient with negative PCR result but typical features of COVID-19 in CT. nally RT-PCR result became positive in the third examination (58).
What stands out from We acknowledge that our study had some limitations: 1. the speci city of CT scan was not as reliable as the sensitivity of that as a result of the majority of studies' nature, which were case-series and the number of true negative patients in those studies were zero. 2 it has been postulated in A Bernheim et al's study that chance of detecting lung involvement in chest CT scan will be increased if the duration between symptom onset and initial chest CT scan rises and this duration was different among 47 studies. So, the different outcomes would have emerged if the duration between studies were equal.

Declarations
Availability of data and material: Information, data, and photos will be provided if they requested.
Competing interests: All authors claim that there is not any potential competing or con ict of interest Funding: There is no funding.
Authors' contributions: SMK. A. was the principal investigator and supervisor of the project.M. S. was the epidemiologist who runs the statistical analysis of data and provides gures. SS.TZ and SZ. A. individually screened the data and make the data extraction sheet and tables of the article. F. KH had design the search strategy and wrote the manuscript. AN.SH edited the manuscript and co-supervised the project.