Learning from COVID‑19 Viral Loads
How viral loads vary—or don't—across patients can predict the performance of antigen tests in different groups
COVID‑19 test results are usually reported simply as “positive” or “negative.” However, the amount of virus a person produces—the viral load—can vary. As clinical microbiologists responsible for COVID‑19 testing at a major medical center, we estimated viral load for over 40,000 patients who had a positive PCR test at our hospital from 2020-2023 so you can see how viral loads vary—or don't—across age, sex, and so on. Please explore for yourself!
How do viral loads vary across patient groups?
COVID‑19 viral loads can vary a billion fold from person to person. Within each person, it starts low, reaches a peak (often preceding symptoms), and then falls again as the infection comes under control. Based on our observations, we hypothesized that most groups exhibit the same range of viral loads. If true, then antigen tests (see below), which a person can take at home, would be equally effective for most groups. If not, then certain groups might require separate trials to get the most benefit from antigen tests.
A generous grant from the Reagan-Udall Foundation for the FDA allowed us to test this hypothesis. We used fully anonymized data to protect patient privacy. Instead of simply reporting our own observations, we have made the results available here to everyone, so you can explore and compare whatever group or groups that may be of interest to you. This can include comparison of complex subgroups, such as healthy-weight vs. overweight >60-year-old inpatients or sick- vs. well-appearing patients with presumed early vs. delta vs. omicron variants.
To help you assess whether between-group differences are significant, we calculated the p-value for each pair of groups according to a statistical test called the Kolmogorov–Smirnov test (KS test) . The KS test is a commonly used test when data do not follow a bell-shaped curve. The p-value measures how likely it is that two distributions—here, the distributions of viral loads for each pair of groups—are drawn from the same underlying distribution. A large p-value means the two groups in the pair are statistically indistinguishable; a low value mean they differ more than would be expected by chance.
How can specific COVID‑19 antigen tests be expected to perform on the above groups?
Antigen tests are less sensitive than PCR tests but have the advantage that they can be self-administered and used at home. For the antigen tests in the pulldown menu below on the left, the antigen test and a PCR test have been run at the same time on the same people. This real-world data can predict how well these antigen tests will perform on different groups without having to run time-consuming and expensive trials on each group selected above. Select each group using the radio buttons below to compare expected antigen-test performance.
When paired trials that directly compare an antigen test to PCR have not been performed, we can still predict how sensitive an antigen test is for detecting contagiousness by using some measures of the test's analytical performance. The commonly reported measure of a test's performance is the limit of detection (LOD), defined as how high the viral load must be to be detected by the test 95% of the time. However, without another datapoint, our estimate of the performance will be conservative. Therefore if you choose "other test..." you will be able to set a second datapoint, the 50% detection threshold, which is the viral load at which the test will be positive half the time.