Editor’s note: Find the latest COVID-19 news and guidance in Medscape’s Coronavirus Resource Center.
Recently, a patient arrived at the UC San Diego Health medical center with what are now classic symptoms of COVID-19: a history of coughs, pneumonia, and severe respiratory distress that required immediate intubation.
What he didn’t have was a positive SARS-CoV-2 test — neither the first nor the second time clinicians swabbed the back of his throat. SARS-CoV-2 is the virus that causes COVID-19.
“The two negative tests didn’t convince anybody,” said Davey Smith, MD, a virologist and chief of the division of infectious diseases and global public health at UC San Diego School of Medicine. It was only on the third test, when they sampled fluid from a bronchial wash, that they were able to find the virus.
Smith’s patient is not alone. Though almost all experts agree that broad testing for SARS-CoV-2 will be critical to understanding, containing, and eventually treating COVID-19, the effort is hampered by limitations of current tests.
The tests today, experts say, are so new that it’s unclear how reliable they are. Anecdotally, clinicians report false negative rates of anywhere from 2% to 30%, depending on what part of the body is being tested and what means they are using to get a sample, as well as epidemiological and clinical factors.
And the US Food and Drug Administration issued an alert earlier this week warning of false negatives with one of the most commonly used tests, Abbott Labs’ ID NOW rapid test for COVID-19.
Data published earlier this week in the Annals of Internal Medicine show that test accuracy varies widely over the course of the disease in a mixed population of inpatients and outpatients. On the day symptoms appear, the median false negative rate was 38%. That figure dropped to 20% on the third day after symptom onset, but climbed again to 66% about 2 weeks later.
But test results should only be part of the picture. The key is clinical suspicion informed by all the above factors, said Joshua Metlay, MD, PhD, chief of the division of internal medicine at Massachusetts General Hospital, and coauthor of a series of articles on clinical decision-making in the Annals of Internal Medicine.
“How we treat patients is going to depend on understanding this concept,” Metlay told Medscape Medical News. “It isn’t one number. It’s actually much more complicated and very nuanced.” If clinicians don’t understand that, he added, “We’re really going to make mistakes about how to use all these negative tests.”
When Hope Outstrips Reason
A positive SARS-CoV-2 test sets off a cascade of actions, in and out of clinical settings: In patients with symptoms, it triggers a set of protocols, as recommended by the National Institutes of Health and individual hospitals, around use of personal protective equipment (PPE) for staff, whether patients are placed in rooms with others or singly, and specific treatment choices, such as which ventilator protocol to use. By contrast, a negative test, in an ideal situation, should lead a clinician to keep looking for a causative agent or underlying problem. Quality care, in other words, relies on accurate diagnosis.
In patients without symptoms, a positive test means suggesting quarantine and isolation for two weeks, said Colin West, MD, PhD, professor of medicine and biostatistics at the Mayo Clinic in Rochester, Minnesota.
But because of the relatively high rate of false negatives, a negative test in an asymptomatic person can’t confer the kind of relief patients, the public, or policymakers would like it to, West said.
“People can’t relax their physical distancing, their handwashing, their surface hygiene, their mask-wearing” even with a negative test, he said, because they still could be carrying the virus.
“When hope outstrips reason, we sometimes prematurely pin our hopes on tests that aren’t as good as we want them to be,” said West, who wrote a perspective for Mayo Clinic Proceedings warning about the dangers of false negatives. “Smart clinicians all around the country are not believing the test results when their clinical suspicion is high enough.”
Calculating Clinical Suspicion
With false negative rates ranging from 3.2% in a cohort of seriously ill COVID-19 inpatients in New York City, to 66% in the mixed population in the Annals of Internal Medicine study, it’s understandable that clinicians might be skeptical of the results in front of them.
The first thing to understand, Metlay said, is that the sensitivity of the test isn’t the same as the rate of false negatives. Sensitivity only describes the absolute ability of a given test to detect SARS-CoV-2 when it’s present.
A false negative is a combination of the accuracy of the test itself and its handling, along with clinical symptomology, local epidemiology, and the individual behavior of the person in front of you.
“This is why people don’t usually report false negative rates as one number — because that number can only be calculated or interpreted in the context of the group that you’re talking about,” Metlay said. “Two hospitals could report completely different false negative rates and that’s completely reasonable. Even within the same hospital, the false negative rate in the clinic could be different than the false negative rate in the emergency room.”
That is to say, people showing up with symptoms in an emergency department should be met with a higher pre-test probability of having COVID-19 than well people who show up for drive-through testing, he said.
The same holds true for the background rate of COVID-19 in a community.
“If you get into a community where almost no one is infected and you tested everybody, your false-negative rates are going to be very low, because almost everybody there is negative,” he said. In Boston, where Metlay practices, the community rates of COVID-19 are substantial, so even those with mild symptoms might be met with higher suspicion of having SARS-CoV-2. And that’s even before the patient opens their mouth and explains where they’ve been or who they’ve been in contact with.
This helps explain why Rajesh Gandhi, MD, Metlay’s colleague at Mass General, said the inpatient false negative rate there is between 2% and 3%, while Smith from UC San Diego reported that their false negative rates are no more than 10% — or that unpublished data first reported by NPR found that the most commonly used, fastest turnaround tests also churned out 15% false negatives.
No Gold Standard
Another factor in all of this are the tests themselves. Back in March, Chinese researchers at Central South University in Changsha, Hunan province, lamented that “existing PCR methods have very good specificity but low sensitivity, meaning that negative test results cannot exclude the presence of SARS-CoV-2.”
PCR stands for polymerase chain reaction, a means of testing for viral genetic material, currently most commonly used in a nasopharyngeal swab.
But we don’t know how inaccurate the tests actually are, said Stephen Rawlings, MD, PhD, an infectious disease fellow at UC San Diego’s Center for AIDS Research, who has been helping to validate RT-PCR tests for SARS-CoV-2 since repatriated Americans were held in isolation at military bases starting in March.
For one thing, we have nothing to compare current tests with.
“To truly determine false negatives, you need a gold standard test, which is essentially as close to perfect as we can get,” Rawlings said. “But there just isn’t one yet for coronavirus.”
For another, the studies that have been done on the accuracy of the tests themselves are filled with flaws, said Mayo Clinic’s West.
Sensitivity estimates are usually based on testing the tests against people who they already know have COVID-19. But that’s a bias — you know what you’re looking for, West said. Without control groups or blinded testing, it’s impossible to get “good information about where these imperfections lie, or even the magnitude of those imperfections,” said West, who conducted a tweetorial on how to understand the accuracy of current RT-PCR tests.
A recent non-peer reviewed preprint meta-analysis of five COVID-19 studies comprising 957 patients found that the underlying poor quality of data made it impossible to judge how effective the tests in the nation’s labs are at all.
“We’re trying to have informed conversations about how good are these tests, and how helpful are they in ruling in or ruling out diagnosis, when the source literature is so poor,” West said.
This is where the Centers for Disease Control and Prevention should step in, according to UCSD’s Smith. Local labs are doing their best to validate tests on their own, but if they could send their results in a blinded way to the CDC, he’d have a lot more confidence in every test.
“We really need these panels to help with the quality assurance internally,” Smith said. “It’s not about the nasal swabs or not. It’s about the potential sources of error within the lab.”
Human Error and Biological Process
Now add in the human piece of gathering, transporting, and reading an RT-PCR test, said Daniel Griffin, MD, PhD, an infectious disease physician and associate research scientist in the department of biochemistry and molecular biophysics at Columbia University in New York City.
“Basically, I tell my patients that unless you feel like they were trying to biopsy your brain, it wasn’t done correctly,” he said. And the less well done the test, the less likely the results will be reliable.
But even if all that is correct, there’s one more hurdle: What part of the body should be sampled and at what point in the illness? A viewpoint published in JAMA earlier this month synthesized known data on the accuracy of different tests at different points in the disease process.
For instance, it shows that within the first week of exposure before symptoms and in the first week of symptoms, nasopharyngeal swabs are most accurate. But by the end of week 2 of symptoms, bronchoalveolar lavage/sputum is most accurate.
This conforms to what clinicians report anecdotally. The number of copies of the virus in the nose and pharynx is highest in the early days infection, just like the flu, said Columbia University’s Griffin. And that may mean the RT-PCR tests of the nose and pharynx work best in the first few days of infection — when patients still have mild or moderate symptoms.
“If you do the test that first day or so when you’re sick, you’re going to have pretty good sensitivity” with nasopharyngeal swabs, said Griffin, who often provides COVID-19 updates on the podcast “This Week in Virology.”
“The interesting thing is, when people get admitted to the hospital, now they’ve been feeling crummy for a week or more,” he said. “Now it’s day 13 or day 14, and now the virus is actually starting to get to a lower level of activity” in the upper respiratory tract.
And that means, said Griffin, “there’s not as much virus around” the nose and pharynx to test.
That’s when UC San Diego’s Rawlings said they’ve found that passing a catheter through the tracheal tube of someone who’s already intubated can “often find [the virus] there at very high levels.”
This may explain the phenomenon that Smith described, as well as what Neera Ahuja, MD, of Stanford University, said in a “Medicine and the Machine” podcast recently, that “this virus actually moves from proximal upper airway nasal-pharyngeal down to the lower lungs.”
“If you catch it in a stage where it’s already progressed, you may have a false negative,” she said.
Using Clinical Judgment
It’s a lot to take in for a single clinician interpreting a single test result. The good news is that clinicians are literally trained for this, said Carlos del Rio, MD, of Emory University and coauthor with Gandhi of a recent article on mild or moderate COVID-19.
“We as clinicians use our brains,” he said. “We don’t say, ‘Oh the person doesn’t have the disease’ — we use the test in the context of our clinical expertise.”
So clinicians should ask: Are the patients sick themselves? What are their symptoms? Have they traveled to or from COVID-19-endemic areas recently? Have they been in touch with someone they know has COVID-19? Have they been practicing physical distancing, mask-wearing, and other protective behaviors? said Metlay.
They may even want to consider other questions, said Gregorio Millett, MPH, vice president and director of public policy at the American Foundation for AIDS Research (amfAR). Millett and colleagues have unpublished data showing that although disproportionately black counties account for just 22% of US counties overall, they make up 52% of counties with COVID-19 cases and 58% of counties with COVID-19 deaths. In those communities, lack of insurance and living in crowded households were associated with increased risk for acquiring COVID-19 — opening another potential data point to consider in those communities.
The bottom line is that “nobody can integrate all this math in their head every single time” they see a patient, Metlay said. Still, keeping all these factors in mind will help clinicians look at that negative result with clear eyes.
“This,” he said, “is what helps people not get tricked by these tests.”