Study population and setting
This study used the published and preprint literature (PubMed, bioRxiv, and medRxiv) through April 15, 2020 to identify data for a pooled analysis to estimate the false negative rate of RT-PCR testing for SAR-CoV-2 over time since infection and time since onset of symptoms.
Summary of Main Findings
The authors pooled data from 7 studies which provided data on RT-PCR test results by time since infection. The study population included both inpatients and outpatients, and included 1330 respiratory samples. They used information on sensitivity to model the false positive proportion over time. Since the date of infection was not known, the authors assumed it began 5 days before symptoms began, and time was measured both from the average day of infection and the average day of symptom initiation (on day 5 after infection). The authors found the probability of a false negative RT-PCR test varied over time. In days since assumed infection, 100% of tests were falsely negative (95% CI, 100% to 100%) on day 1 dropping by day 4 to 67% (CI, 27% to 94%). The probability of a false negative decreased after average symptom onset on day 5 to 38% (CI, 18% to 65%) to 20% (CI, 12% to 30%) at 3 days after symptom onset (on day 8), then began to increase to 66% (CI, 54% to 77%) on day 21.
The researchers were able to increase the sample size substantially by conducting a pooled analysis using data from other studies. The study included samples collected at multiple time points after infection, allowing the authors to estimate the false negative proportion over time. Numerous sensitivity analyses were undertaken to ensure the robustness of the analysis.
The authors noted the potential for heterogeneity across studies in how samples were collected that were being pooled, potentially reducing the interpretability of the results of the pooled analysis due to potential differences in the sensitivity of the sample collection method. The estimates had wide confidence intervals, reflecting the limited number of samples when stratified by time since infection. Samples were not collected on the same timeline, which could result in some selection bias. The study also could not separate out rates by sample type (oropharyngeal vs. nasopharyngeal), which may have different false negative proportions.
The study provides the largest sample size to date to the question of false negative results using RT-PCR, which is of critical importance for public health action to ensure those who test negative are truly not infected so they don’t unknowingly infect others.
This review was posted on: 17 June 2020