Debunking PCR test "false positives"

joseph's picture

Back in January, the "conspiracy-verse" was abuzz with claims that PCR tests for Covid-19 produced "97% false positive" results.  The pandemic, they said, was being manufactured as a means of social control, and the reported case numbers were vastly inflated by the "false positive" tests.  

Our local conspiracist-in-chief flooded my inbox with such claims, while resisting every effort to correct his misunderstandings and ignoring the plainly visible evidence of a pandemic ravaging long-term care facilities, clogging up ICU units, and the tragic personal experiences of our fellow islanders.  I present here some of the analysis I did then in hopes it helps others who are more interested in understanding the science and facts than pushing a conspiracy theory.

Spoiler alert: PCR tests are a quick, and accurate method of determining if a person has been exposed to Covid-19.  The tests do not indicate viral loads or infectiousness, but are considered the "gold standard" for detecting genetic markers from a viral infection.  The claims about "false positives" are based on an (intentional?) misunderstanding and misconstruing of the science of PCR tests.

The Science

The claims about PCR test "false positives" is based on, in part, this reputable medical research paper:

In particular, the conspiracists quote Figure 1 from this paper which indeed seems to show that at high Ct (cycle threshold), there are a large number of samples that test "positive" (grey bars).  But science is about details - you need to understand what question the authors were asking and what methods they used before jumping to conclusions...


1) Here is the graph that caused all the hubbub:

The conspiracists would have you simply look up Ct value on the X-axis, and then read off the % positive on right-hand Y-axis and interpret that as the "false positive" rate.


2) But that graph does not show "false" positives.   Note the figure caption: "Percentage of positive viral cultures" -- nothing to do with "false" positives.
In other words,  it simply shows the total % of test that returned positive at a given Ct level.


3) The caption clearly states this experiment uses "samples from coronavirus disease 2019 patients" -- in other words, EVERY sample used in this experiment came from a patient who had already been diagnosed with SARS-CoV-2.

Such an experiment can't tell you anything about the false positive rate, since by definition every sample is from a positive patient (i.e., in a population with 100% disease prevalence, the term "false positive" doesn't even have any meaning).


4) In the introduction, the researchers clearly state the question they are interested in addressing:

A major issue related to the outbreak has been to correlate viral RNA load obtained after reverse-transcription polymerase chain reaction (RT-PCR) and expressed as the cycle threshold (Ct) with contagiousness and therefore duration of eviction from contacts and discharge from specialized infectious disease wards.

In other words, they are asking:  when can a patient who has been diagnosed with SARS-CoV-2 be released from hospital without posing an infection risk to others?  
This explains why they only use samples from covid patients.


5) The point of the study isn't about test error rates at all -- it is about the correlation between test results done at different Ct levels and the likely contagiousness of a SARS-CoV-2 patient 1, 2, or 3 weeks following their diagnosis.    
i.e., the researchers want to know if a PCR test can be used to determine infectiousness - they are already confident it is a reliable test for exposure.


6) The researchers finish with a look at the question they need to answer next:

From our cohort, we now need to try to understand and define the duration and frequency of live virus shedding in patients on a case-by-case basis in the rare cases when the PCR is positive beyond 10 days, often at a Ct >30. In any cases, these rare cases should not impact public health decisions.

In other words, to really answer their question, they now have to examine how much virus is being shed by these patients, especially those whose PCR test still shows positive after 10 day, but advise that for the purposes of setting public policy,  10 days seems a practical period.  This perhaps explains where the policy of 10 - 14 days of isolation may come from, or perhaps the authors are confirming that 10 days seems about right.


PCR tests can not tell you much about a person's infectiousness.  
This is the shred of truth that the "97% false positive" narrative is woven around.  

But a PCR test is very reliable and accurate at detecting exposure to a virus.  The very strength of this test is its ability to detect even trace quantities of viral RNA.  People who receive a positive Covid-19 test may rightly want to know if they are actually infectious, but that requires detailed lab analysis.  What cannot be denied is that a positive PCR test is a near certain indication the person has been exposed to the virus and has viral RNA in their system.

In the midst of a pandemic spreading exponentially the most critical public health objective is to stop transmission.  The PCR test provides a quick, relatively inexpensive means of identifying exposures, determining where and how fast the virus is spreading, and preventing further transmission.  As such, it is a critically important tool for assessing public health and guiding public health policy.

Where lies the correct balance between protecting people's liberty and protecting public health?

This is a debate we should be having and dearly wish we could have in public.  Unfortunately, those who choose to misinterpret science and spread misinformation and conspiracy theories make public discourse on such topics nearly impossible.


Post new comment

The content of this field is kept private and will not be shown publicly.
This question tests whether you are a human visitor, to prevent spam submissions.
The answer can easily be found on this site if you don't know it.
Don't stress - if you get it wrong, you'll get another chance, just try again :-)
1 + 1 =
Solve this simple math problem and enter the result. E.g. for 1+3, enter 4.
To prevent automated spam submissions leave this field empty.