Computed tomography of the chest using COVID-19 Reporting and Data System classification system (CO-RADS) displayed a favorable diagnostic performance in symptomatic subjects suffering from coronavirus, assisting its application for triage. 

Chest computed tomography (CT) can aid in the assessment of the temporal disease stage and severity of the coronavirus disease 2019 (COVID-19) pneumonia. In the early stage of coronavirus replication (day 0-4), the ground-glass opacities are the major lesion while the crazy paving patterns define the raised recruitment of inflammatory cells to the lung interstitium in the progressive stage (day 5-8). The peak stage (day 10-13) is characterized by consolidation with fibrosis and diffused alveolar damage.

The radiological lesions are also witnessed in other non-infectious inflammatory lung disorders and viral pneumonia. However, in a pandemic context may harbor the diagnostic potential for coronavirus infection specifically for the patient triage. The reference method for diagnosing coronavirus, SARS-CoV-2 polymerase chain reaction (PCR), is very specific. However, it has variable sensitivity which has beend found to be as low as 70%.

In the medical care settings with long turnaround times and limited PCR capacity, chest CT was suggested as an alternate for coronavirus triage or diagnosis. Studies supporting chest CT as a first-line diagnostic tool for coronavirus displayed numerous methodological concerns. Most of the studies were underpowered, demonstrated major selection biases incorporating only patients suffering from coronavirus symptoms and 40%-50% a priori risk of viral infection, and utilized binary scoring of computed tomography without a standardized definition of coronavirus-compatible computed tomography.

Weighed against the procedural risks and cost of computed tomography, a controversy was sparked.

This resulted in the consensus statements by the American College of Radiology, the Centers for Disease Control and Prevention, the Society of Thoracic Radiology, the American Society of Emergency Radiology, the Fleischner Society, and the Radiological Society of North America (RSNA), opposing computed tomography as the first-line tool for diagnosing coronavirus.

In this study, the diagnostic power of chest computed tomography versus SARS-CoV-2 PCR utilizing CO-RADS was explored. The Dutch Radiological Society developed CO-RADS to categorize the suspicion level for coronavirus pneumonia. It usually aligns with the structured reporting suggested by RNSA, scoring the level of coronavirus suspicion on a scale of 1 to 5, with CO-RADS 1 corresponding to ‘negative’ category, CO-RADS 2 to ‘atypical’, CO-RADS 3 and 4 corresponding to ‘indeterminate’ with ‘lower’ or ‘higher likelihood’, and CO-RADS 5 equaling the RNSA ‘typical’ category. These data may permit a more evidence-based definition of the possible role of chest computed tomography in coronavirus triage.

Rationale behind research

No study prior this one, was conducted to investigate the diagnostic potential of computed tomography CO-RADS for SARS-CoV-2 infection. Thus, this secondary analysis of a single-center prospective trial was performed.

Objective

A study was performed to examine the performance of computed tomography CO-RADS to diagnose SARS-CoV-2 PCR-positivity in patients having coronavirus symptoms and to screen for asymptomatic SARS-CoV-2 infection in control individuals in a setting with an elevated prevalence of viral infections. 

Study outcomes

• Assessment of CO-RADS score was the major outcome.

Outcomes

Baseline: There were no vital differences reported at baseline.

Study outcomes

• CT with CO-RADS scoring has a favorable diagnostic performance for coronavirus pneumonia in both asymptomatic (area under the curve [AUC]=0.70) and symptomatic (AUC=0.89) patients.
• In symptomatic subjects (42% PCR+), a CO-RADS score of ≥ 3 detected positive PCR having acceptable sensitivity (89%) and specificity (73%) leading to a positive predictive value of 70%.
• In asymptomatic subjects (5% PCR+), a CO-RADS score of ≥ 3 detected SARS-CoV-2 infection having low sensitivity (45%) but elevated specificity (89%) and a positive predictive value of 18%.

Fig 2: Diagnostic performance of CT-CORADS scoring in individuals with and without COVID-19 symptoms

This analysis aimed to determine the value of computed tomography of the chest with CO-RADS classification for screening the asymptomatic coronavirus infections and to evaluate its diagnostic potential in patients having coronavirus symptoms during the exponential phase of coronavirus spread. In symptomatic subjects, the pre-test probability of coronavirus infection, as marked by the prevalence of PCR-positivity, was elevated at 41.7%.

A CO-RADS score of ≥ 3 substantially elevated the post-test probability to 69.8% and CO-RADS 5 even to 89.4%. For the infection control policies, the CO-RADS 5 could therefore be utilized as a triage tool to quarantine symptomatic subjects in settings with bottlenecks in PCR testing. Yet, CO-RADS less than 3 was still linked with a post-test probability of 9.7% (corresponding to 90.3% negative predictive value). This indicated that chest computed tomography cannot replace PCR as the diagnostic test.

Prevalence of the SARS-CoV-2 PCR-positivity in the asymptomatic controls was found to be 5.3%, in line with the secondary attack rate at a population level of 6.6% during the exponential phase of coronavirus spread. This control arm was, therefore, suitable to explore if chest computed tomography can screen for asymptomatic coronavirus infection. Also in asymptomatic individuals computed tomography CO-RADS illustrated good diagnostic performance. However, the numerous dichotomization scenarios failed to attain the high sensitivity needed for the screening test.

The CO-RADS ≥ 4 was found to achieve only 31.7% sensitivity. A negative test (CO-RADS <4) shifted pre- to post-test probability only from 5.3% to 3.9%, not sufficient to substantiate the procedural risk of computed tomography. The specificity of CO-RADS ≥ 4 in asymptomatic individuals was elevated (94.4%) and led to a considerable rise in post-test probability to 24.1%.

In the pandemic setting, such incidental findings should be reported as ‘compatible with coronavirus pneumonia’ rather than as ‘viral pneumonia’ as advocated by the RNSA and should trigger SARS-CoV-2 PCR or syndromic panel-based PCR testing for other respiratory pathogens prior to excluding the non-infectious inflammatory lung disorders.

The developers of CO-RADS witnessed a favorable diagnostic performance in a pilot trial on 105 patients suffering from coronavirus symptoms and 50.5% PCR-positivity with AUC under the receiver operating characteristic curve (ROC) curve of 0.91. This was confirmed in this study with comparable AUC on a larger sample. In comparison with prior studies favoring chest computed tomography for screening or diagnosis of coronavirus, this study answered the vital call for well-powered data sets and its prospective design on consecutive, unselected subjects having comparable comorbidities, demographics, and upfront clinical grouping according to presence or absence of coronavirus symptoms, reduces the selection biases.

This study used structured reporting of chest computed tomography data and attribution of the likelihood ratios to each suspicion level. The major studies thus far utilized dichotomization of computed tomography results as negative or positive, frequently without a suitable definition of a positive computed tomography. In China, a trial reported a 97% sensitivity of chest computed tomography for coronavirus diagnosis but it was accompanied by a poor specificity of 25%.  This was possibly clarified by a reduced subjective interpretation threshold to raise the sensitivity.

Similar to specificity and sensitivity, the likelihood ratios are test properties that, in defined patient populations, are independent of the prevalence of the disease. The actual clinical values of a positive test result to confirm (positive predictive value) or negative test result to rule-out disease (negative predictive value) considerably depend on the prevalence of the disorder. Utilizing the likelihood ratios, the post-test probability as an indicator of positive predictive value can be estimated taking the reported prevalence of PCR-positivity as pre-test probability. The negative predictive value is 1 minus the post-test probability. The positive predictive value is mathematically most affected by specificity. 

The meta-analysis displayed a low pooled specificity of dichotomic chest computed tomography of 37% for coronavirus diagnosis with a low associated positive predictive value from 1.5%-8.3% in low prevalence (<10%) settings. The data demonstrated that the CO-RADS categorization enhances specificity and thus discloses an elevated positive predictive value as likelihood ratios rise.

Sensitivity significantly affects the negative predictive value. In the data set of this study, the sensitivity of chest computed tomography was not sufficient for excluding coronavirus infection both in asymptomatic and symptomatic participants. This supports the consensus statement that chest computed tomography should not be utilized as a diagnostic test. While computed tomography screening for asymptomatic coronavirus infections is not suggested, the incidental detection of CO-RADS ≥ 3 in asymptomatic subjects has sufficient positive predictive value to trigger SARS-CoV-2 PCR reflex testing.

In conclusion, computed tomography with structured CO-RADS scoring demonstrated a favorable diagnostic performance for coronavirus pneumonia but cannot replace SARS-CoV-2 PCR as a diagnostic test. It can be utilized as an alternative triage tool in patients having coronavirus symptoms but not for the screening of asymptomatic coronavirus infections as Sensitivity in asymptomatic individuals was not sufficient to justify its usage as a first-line screening technique.