I had taken the false positive rate of antigen tests as a fact but the push by gin80 made me go back and look at the study data of the Binax NOW Abbott test and compare with PCR testing for Specificity - the term for true negative with the false positive being the residual. Sensitivity is the true positive rate. Positive predictive rate (PPV) and negative predictive rate (NPV) are based on how common the disease is - key for clinical trial but not germain to this discussion.
Here is a journal article for the eval of the antigen test
Testing among symptomatic participants indicated the following for the BinaxNOW antigen test (with real-time RT-PCR as the standard): sensitivity, 64.2%; specificity, 100%; PPV, 100%; and NPV, 91.2% (
Table 2); among asymptomatic persons, sensitivity was 35.8%; specificity, 99.8%; PPV, 91.7%; and NPV, 96.9%. For participants who were within 7 days of symptom onset, the BinaxNOW antigen test sensitivity was 71.1% (95% CI = 63.0%–78.4%), specificity was 100% (95% CI = 99.3%–100%), PPV was 100% (95% CI = 96.4%–100%), and NPV was 92.7% (95% CI = 90.2%–94.7%). Using real-time RT-PCR as the standard, four false-positive BinaxNOW antigen test results occurred, all among specimens from asymptomatic participants. Among 299 real-time RT-PCR positive results, 142 (47.5%) were false-negative BinaxNOW antigen test results (63 in specimens from symptomatic persons and 79 in specimens from asymptomatic persons).
Here is the PCR False Positive rate
Evidence for a Significant False Positive Rate (FPR) in PCR Tests for COVID-19
Data from External Quality Assessments of PCR tests for other RNA viruses | • FPR usually between 0.8% and 4.0%
• Median = 2.3%; Pooled mean = 3.2% |
Data from External Quality Assessments of PCR tests for COVID-19 | • FPR between <0.4% and 0.7%
• Pooled mean = 0.6% |
Data from actual use of PCR tests for COVID-19 | • FPR usually between 0.2% and 0.9% |
Based on these two studies it the FPR for PCR is 0.6% and the Antigen test (Binax NOW) is 0.2% so less false positive. The real issue is that in all tests for a disease with a low prevalence even small false positive rates stack up. If you test 10,000 for a disease with a 1% prevalence then the numbers are a little crazy -
100 people have the disease, 9,900 do not. Looking only at FPR if it is 0.6% then saddly 59 folks will have a false positive (the PCR data) if 0.2% then 20 folks have a false positive (the antigen test). So asymptomatic screening of lots of folks with a low prevalence disease gets a lot of false positive no matter what test you use.
Based on this I think the antigen test is the way to go - new data but it made me look. Something we all should be doing. I would encourage folks to not just read a newspaper article as there is really poor editing and bias control. Find the real journal articles and ....