Controversy in coronaViral Imaging and Diagnostics (COVID).

Sir—We are concerned by the “Guidance for pre-operative chest CT imaging for elective cancer surgery during the COVID-19 Pandemic” authored by The Royal Colleges (Surgeons Edinburgh, Surgeons England, Physicians and Surgeons Glasgow, Surgeons Ireland and Radiology). It mistakenly argues that there is a case for screening some asymptomatic patients prior to surgery. We are also aware of pressures in our hospitals to screen patients using computed tomography (CT). There is, however, no good evidence to support such a view, which implies an understanding of the nature of screening for this particular disease, which is not currently justified.

A study of 51 symptomatic patients reported 98% sensitivity for CT compared to 71% for reverse transcription polymerase chain reaction (RT-PCR) in symptomatic patients. A further study of 103 symptomatic subjects showed initial chest CT sensitivity 93%, specificity 53%, positive predictive value 92%, and negative predictive value of 42% compared to positive (up to four) RT-PCR. A further study of 1,049 patients showed a sensitivity of 97%, specificity of 25%, and accuracy of 68% compared to RT-PCR. Table 2 in this paper helpfully gives the numbers of true-positive (580), true-negative (105), false-positive (308), and false-negative (21) tests. The accuracy is affected by the prevalence (in this case 57%).

CT data from asymptomatic individuals from the cruise ship “Diamond Princess”, where 41 of 76 asymptomatic subjects (54%) had lung opacity on CT is not relevant to screening as they were all RT-PCR positive. Even including all the (symptomatic and asymptomatic) subjects, CT only had 61% sensitivity for any lung abnormalities, and a 20% false-negative rate in symptomatic patients. It would seem doubtful that CT would perform better in a low pretest probability cohort of RT-PCR negative patients.

There may be a problem with the RT-PCR test used in our patients; however, it is specific, and also, different to these referenced papers. Replacing it with CT (that in patients with a high pretest probability is sensitive but of poor specificity) is problematic. As prevalence falls, a non-specific test such as CT, which does NOT test for the virus, will become highly misleading. For those who fear an insensitive RT-PCR test, supplementing it with a sensitive, but non-specific, test simply introduces a different error. The clinician is then left not knowing which one to believe. We may as well look at the sediment pattern of tea leaves (tasseography) to determine which one is correct in an asymptomatic subject.

In our department, we have been informed that “clinicians are finding it [CT] helpful”. They are perhaps practising within the realm of belief rather than science as any number of findings on CT could be misinterpreted as classic COVID-19 — heart failure, other infections, air trapping, poor inspiration — meaning a positive CT (using these statistics and based on a 5% prevalence) will have a positive predictive value of 6% meaning a patient is denied appropriate care based on the flimsiest of evidence; the negative predictive value of 99% seems perfect but is in fact only modestly “better” than not doing the test with an overall accuracy of 29%.

The illustration (Fig 1 ) shows a patient referred for CT for “?COVID-19” (but no COVID-19 symptoms). The radiologist reported it as “non-COVID-19” due to the presence of pulmonary oedema (effusions, ground-glass opacity) due to aortic valve stenosis (calcified valve not shown) and left ventricular dysfunction (slow transit of contrast medium on bolus tracking), which was confirmed by echocardiography; however, the C-reactive protein (CRP) was elevated and there was lymphopenia, which led other consultant radiologists to define this as either classic COVID-19 or indeterminate COVID-19. The patient was treated for heart failure and breathlessness improved, with (reportedly) no development of COVID-19 symptoms and no positive RT-PCR. The pulmonary abnormalities were visible on chest radiography (and ignoring the cardiac inferences) “chest” CT did not contribute to the patient's management.

Figure 1

Chest CT image showing bilateral pleural effusions (star) and bilateral ground-glass opacity (arrow).

Centres that wish to research the use of CT in screening subjects should do so, but this should be part of proper research, and using it to influence decision-making is currently wrong. There is no evidence to support screening for COVID-19 in asymptomatic patients using CT. It will produce misleading results, and thus may be dangerous. In our opinion this also includes subjects with acute abdominal pain, though we endorse the opinion of the British Society of Thoracic Radiology on this. It should not be news to doctors that pneumonic processes can present with abdominal symptoms, and “normal clinical” expertise is encouraged.

Conflict of interest

The authors declare no conflict of interest.