key: cord-0958063-yl7xpo3e
authors: Isikbay, Masis; Henry, Travis S.; Frank, James A.; Hope, Michael D.
title: When to Rule Out COVID-19: How Many Negative RT-PCR Tests Are Needed?
date: 2020-08-18
journal: Respir Med Case Rep
DOI: 10.1016/j.rmcr.2020.101192
sha: 1bb406d2480b3c236fd14237e3a45d9e4045f59b
doc_id: 958063
cord_uid: yl7xpo3e

Amidst the COVID-19 pandemic, clinicians have been plagued with dilemmas related to the uncertainty about diagnostic testing for the virus. It has become commonplace for a patient under investigation (PUI) to repeatedly test negative but have imaging findings that are consistent with COVID-19. This raises the question of when the treating team should entertain alternative diagnoses. We present such a case to help provide a framework for how to weigh repeatedly negative test results in clinical decision making when there is ongoing concern for COVID-19.

The spread of Coronavirus disease 2019 (COVID-19) has resulted in a global pandemic and has altered many aspects of daily practice both inside and outside of medicine. While various methods of molecular testing for COVID-19 are now available (1), low reported sensitivities for these tests (2) (3) (4) have lowered clinical confidence in their efficacy in ruling out infection. In this setting, CT evaluation of PUIs has become more popular as a supplement to RT-PCR testing given imaging findings that have been reported as characteristic of infection (5) .

While the radiographic appearance of COVID-19 can be striking, concerns over the true sensitivity and specificity of this modality further complicate the clinical picture (6, 7) . Clinicians now find themselves in a position where patients have imaging findings suggestive of COVID-19 but also have one (or multiple) RT-PCR tests that are negative. We present such a case here and cover the applied clinical statistics to address how many serial RT-PCR tests are needed to effectively rule out COVID-19 infection.

An otherwise healthy 35-year-old presented to the emergency department in April 2020 with three days of subjective fever, cough, chills, myalgia, and diarrhea. He denied sick contacts and had been observing social distancing policies. He had four roommates, none of whom were ill. His labs were notable for a white count of 17.9 K/cmm. A chest radiograph ( Figure 1A) showed bilateral lower lung opacities reported as "typical of viral pneumonia including COVID-19." The respiratory viral panel was negative and a nasopharyngeal COVID-19 RT-PCR swab was obtained. The patient was discharged in stable condition with instructions for supportive care and home self-isolation. Following discharge, the RT-PCR test returned negative.

J o u r n a l P r e -p r o o f Follow-up radiograph at time of representing to the ED two days later (B). The lower lung opacities are more confluent. This was interpreted as "consistent with COVID-19 pneumonia." Axial images from CT pulmonary angiogram performed four days after initial presentation (C and D). There are confluent groundglass opacities particularly in the lower lobes indicative of lung injury. The upper lobes, however, show more discrete airwaycentric nodules, a finding that is typical of EVALI and has not been described in the setting of COVID-19 pneumonia.

Two days later, the patient represented to the emergency department with worsening respiratory distress, productive cough, and 10 -12 episodes per day of watery, non-bloody bowel movements. His temperature was 101.3° F and oxygen saturation was 88% on room air. A repeat chest radiograph ( Figure 1B) showed worsening lower lung opacities and was interpreted as "most consistent with COVID-19 pneumonia." A repeat nasopharyngeal swab RT-PCR was obtained, and the patient was admitted to the respiratory isolation unit for presumptive COVID-19 pneumonia. He was treated with intravenous antibiotics for community-acquired pneumonia and hydroxychloroquine. Stool samples were sent to evaluate for infectious etiologies of diarrhea (Clostridium difficile testing and PCR-based testing for a panel of pathogens). A second nasopharyngeal swab RT-PCR also returned negative.

Over the next three days, his respiratory status continued to decline, eventually requiring transfer to the ICU and high flow nasal cannula at 25 liters per minute (LPM) with the fraction of inspired oxygen at 0.7. A third RT-PCR test returned negative. No infectious cause of diarrhea was identified by laboratory analysis. A chest CT was then performed (Figure 1C and 1D) to further evaluate his worsening respiratory status. While the imaging findings did not exclude COVID-19, the radiologist suggested an alternative diagnosis for the patient's respiratory distress, which prompted a re-evaluation of the presumptive diagnosis of COVID-19.

Rereview of the patient's past medical history was remarkable for vaping of tobacco/cannabis products. Although the history of vaping was elicited at the original ER visit, this detail was not prioritized in the original differential diagnosis, or in his subsequent clinical course. When the history was readdressed after the CT, the patient also reported recently switching THC products just prior to presentation. CT findings were typical for the recently described entity called e-cigarette and vaping associated lung injury (EVALI) (8, 9) . Despite this, J o u r n a l P r e -p r o o f clinical concern for COVID-19 remained and a fourth RT-PCR test was sent which also returned later as negative.

Hydroxychloroquine was stopped after the fourth negative test, and the patient was started on methylprednisolone for presumed EVALI. He received a high dose for 3 days before beginning a taper, despite objection from infectious disease consultants that he might still have COVID-19 pneumonia. Over the first 3 days of steroid therapy, the patient's respiratory status rapidly improved. His diarrhea resolved, and he was transferred out of the ICU, no longer requiring supplemental oxygen to maintain his saturation. The diagnosis of EVALI was made based on the history of vaping, presentation, four negative RT-PCR tests, imaging findings, and response to steroids. In retrospect, the patient's original presentation was typical of EVALI (including his lack of sick contacts and gastrointestinal symptoms), however this was difficult to appreciate in the context of the COVID-19 pandemic. Even when the possibility of EVALI was first entertained, despite the patient's worsening clinical picture, there was concern about administering steroids as this treatment may not be beneficial in COVID-19 and could prolong viral shedding.

While no robust peer reviewed literature currently exists to reliably calculated the sensitivity and specificity of RT-PCR testing, it is widely agreed upon that the specificity is high, approximately 98% (2). The sensitivity is contested with reports suggesting that it may be as low as 60-78% (2) (3) (4) . Disease prevalence also factors greatly into the negative predictive value of RT-PCR. Unfortunately the reported prevalence of COVID-19 is variable and likely unreliable given limited access to testing (10) . As an exercise to highlight the heuristic pitfall of anchoring J o u r n a l P r e -p r o o f bias in this case, we calculated the predicted performance of serial testing with the assumption that each RT-PCR is an independent test. We used this model to estimate the likelihood of COVID-19 infection after serial RT-PCR tests across various values for disease prevalence ( Table 1) . We have conservatively used a sensitivity of 60% for RT-PCR to demonstrate the efficacy of serial PCR testing under "worst case" conditions. For a more detailed description of how these statistics were calculated please refer to our supplementary materials (Section 5), which include standard two by two tables for serial RT-PCR testing (Table S1.1, S1.2). At the time of this patient's presentation, the prevalence of COVID-19 in symptomatic patients in our care network was estimated to be 5%. Based on our conservative calculations, after the second negative RT-PCR test, the probability of COVID-19 was less than one percent

These supplementary materials have been included to explain false negative rates for serial RT-PCR testing have been calculated. One example of our calculations has been provided here.

Using a sensitivity of 60%, specificity of 98%, a disease prevalence of 5%, and a total population of 100,000 patients, a standard two by two table is first created ( To calculate values for the following serial PCR test, a second two by two table is created which uses the entire population that was PCR negative after the first set of testing (Table S1 .2). In this new population the disease prevalence is the same as the false negative rate calculated above (2.1%) as it represents the total number of people who are infected with COVID-19 divided by the entire new study population. The same sensitivity and specificity are, used, and the new false negative rate is calculated

Centers for Disease Control and Prevention

Integrating Emergency Care with Population Health

Sensitivity of Chest CT for COVID-19: Comparison to RT-PCR

Essentials for Radiologists on COVID-19: An Update-Radiology Scientific Expert Panel

Chest CT Findings in Coronavirus Disease-19 (COVID-19): Relationship to Duration of Infection

A role for CT in COVID-19? What data really tell us so far

CT on the Diamond Princess: What Might This Tell Us About Sensitivity for COVID-19? Radiology: Cardiothoracic Imaging

Pathologic, Clinical, and Physiologic Findings of Electronic Cigarette or Vaping Product Use-associated Lung Injury (EVALI): Evolving Knowledge and Remaining Questions

Imaging of Vaping-Associated Lung Disease

MMWR Morb Mortal Wkly Rep