16.05.2020

Variation in False-Negative Rate of Reverse Transcriptase Polymerase Chain Reaction–Based SARS-CoV-2 Tests by Time Since Exposure

Diagnostic InfectiologyVirology
Kucirka LM et al
Ann Intern Med

Main result

Evolution of the false negative rate of RT-PCR for SARS-CoV-2 :

  •     100%: on D1 of the contact with SARS-CoV-2
  •     67% (95% CI: 27% - 94%) at D4
  •     at D5, the average day of symptom onset, it is 38% (95% CI: 18% - 65%)
  •     the minimum of 20% (95% CI 12% - 30%) is reached at D8 of the infection, i.e. D3 after the onset of symptoms
  •     then it increases again, it is 66% (54% - 77%) at D21

Negative post-test probability of COVID-19 :

  •     at D3, a negative RT-PCR test decreases the relative probability of COVID-19 by only 3% (95% CI: 0% - 47%), e.g. from a pre-test probability of 11.2% in contacts from the same household, we would move to a post-test probability of 10.9% if a test was performed at D3 of exposure.
  •     On the day of onset of symptoms, i.e. at a median of 5 days after the onset of infection, the probability of COVID-19 would then be reduced by 60% in the case of a negative test (95% CI: 33% - 80%).

Takeaways

Using a negative RT-PCR result to rule out the diagnosis of COVID-19 is tricky. Particularly in the early stages of infection, strong clinical suspicion and epidemiological criteria are to be considered as a priority because RT-PCR has a very high rate of false negatives. The ideal time to minimise the risk of false negative RT-PCR appears to be 3 days after onset of symptoms / 8 days after exposure, but 21% remains.

Strength of evidence Weak

- IF : LOW
- retrospective study (-)
- inclusion of two pre-prints (-)
- inaccurate false negative rate due to design heterogeneity of source studies (-)
- Sensitivity analyses in favor of robust results (+)
- probable underestimation of false-negative rate since COVID-19 patients with no positive test are not considered (-)

Objectives

Estimate the false negative rate of RT-PCR for SARS-CoV-2 per day since the onset of infection

Method

Literature review and analysis of pooled retrospective data.

Review conducted on PubMed, bioRxiv and medRxiv up to 15/04/2020. Selection of studies :
- with data on the performance of RT-PCR for SARS-CoV-2 as a function of time (from exposure or symptom onset).
- using upper airway samples.

Inclusion of 7 studies: 2 pre-prints and 5 peer-reviewed articles allowing 1330 RT-PCR tests to be counted.

Case definition :
- for most studies: at least one positive RT-PCR test
- for two studies: a few cases not confirmed by test but on clinical criteria with often positive anti-SARS-CoV-2 IgG or IgM.

Statistical analyses: construction of a hierarchical Bayesian logistic regression model with the assumption (based on previous studies) that the onset of infection occurred 5 days before the onset of symptoms. Calculation, for each day after infection, of the expected false-negative rate and the post-test probability of infection. The specificity was considered to be 100% from the literature.
Sensitivity analyses:
- exclusion of each study in turn
- repeat analyses with 90% lower specificity
- repeat analysis with variation of incubation time between 3 and 7 days

bibliovid.org and its content are bibliovid property.

Legal Notice